MicroRNAs (miRNAs) are short, non-coding RNAs, which post-transcriptionally regulate gene expression and are proposed to play a key role in the regulation of innate and adaptive immunity. Here, we report a next generation sequencing (NGS) approach profiling the expression of miRNAs in primary bovine mammary epithelial cells (BMEs) at 1, 2, 4 and 6 hours post-infection with Streptococcus uberis, a causative agent of bovine mastitis. Analysing over 450 million sequencing reads, we found that 20% of the approximately 1,300 currently known bovine miRNAs are expressed in unchallenged BMEs. We also identified the expression of more than 20 potentially novel bovine miRNAs. There is, however, a significant dynamic range in the expression of known miRNAs. The top 10 highly expressed miRNAs account for >80% of all aligned reads, with the remaining miRNAs showing much lower expression. Twenty-one miRNAs were identified as significantly differentially expressed post-infection with S. uberis. Several of these miRNAs have characterised roles in the immune systems of other species. This miRNA response to the Gram-positive S. uberis is markedly different, however, to lipopolysaccharide (LPS) induced miRNA expression. Of 145 miRNAs identified in the literature as being LPS responsive, only 9 were also differentially expressed in response to S. uberis. Computational analysis has also revealed that the predicted target genes of miRNAs, which are down-regulated in BMEs following S. uberis infection, are statistically enriched for roles in innate immunity. This suggests that miRNAs, which potentially act as central regulators of gene expression responses to a Gram-positive bacterial infection, may significantly regulate the sentinel capacity of mammary epithelial cells to mobilise the innate immune system.
Citation: Lawless N, Foroushani ABK, McCabe MS, O’Farrelly C, Lynn DJ (2013) Next Generation Sequencing Reveals the Expression of a Unique miRNA Profile in Response to a Gram-Positive Bacterial Infection. PLoS ONE 8(3): e57543. https://doi.org/10.1371/journal.pone.0057543
Editor: Huaijun Zhou, University of California, Davis, United States of America
Received: October 22, 2012; Accepted: January 25, 2013; Published: March 5, 2013
Copyright: © 2013 Lawless et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This project was supported by Teagasc (RMIS 6018). NL and AF are supported by the Teagasc Walsh Fellowship Scheme. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
MicroRNAs (miRNAs) are an abundant class of highly conserved, small (19–24 nt long), non-coding, double-stranded RNA molecules. They act as post-transcriptional regulators of gene expression, altering mRNA stability and translation efficiency by hybridizing to the 3′ untranslated regions (UTRs) of certain subsets of mRNAs (collectively as many as 60% of all mRNA transcripts) . Since their initial discovery in Caenorhabditis elegans in 1993 , researchers have gained much insight into the prevalence of miRNAs in other species. The latest miRBase database (release 19) contains 21,264 precursor miRNAs, expressing 25,141 mature miRNA products, in 193 species .
miRNAs have been shown to play key roles in the regulation of innate and adaptive immunity in humans and mice . miR-146a, for example, regulates the innate immune response to bacterial infection, targeting TNF receptor-associated factor 6 (TRAF6) and Interleukin-1 receptor-associated kinase 1 (IRAK1) , while miR-150 regulates the production of mature B cells . Studies elucidating the regulatory roles of miRNAs in bovine infection and immunity, however, are more limited. Bovine miRNAs have been shown to be expressed in a wide range of tissues, including immune-related ones , , but only a handful of studies have investigated how the expression of bovine miRNAs are altered in response to infection. A recent RT-qPCR study, for example, highlighted the differential expression of five inflammation related miRNAs (miR-9, miR-125b, miR-155, miR-146a and miR-223) in response to E. coli lipopolysaccharide (LPS) and S. aureus enterotoxin B stimulation of bovine monocytes . Two other recent studies have used a similar approach to identify several miRNAs that were differentially expressed in the mammary gland tissue of cattle with mastitis , . These and other studies suggest roles for individual miRNAs in regulating bovine immunity, however, according to Ensembl v66 ,  there are over 1,300 annotated miRNAs in the bovine genome. Therefore, studies which adopt genome-wide approaches are required to gain greater insight into the repertoire of bovine miRNAs involved in immunity and infection.
Although microarray technologies to profile miRNA expression have been around for some time , next generation sequencing (NGS) based technologies are revolutionising the field and provide the opportunity to profile the expression of known miRNAs with discriminating resolution and accuracy, and also to identify novel miRNAs . Furthermore, these technologies allow one to differentiate between the expression of alternative mature miRNAs from the same precursor and to identify the differential expression of miRNA isomiRs . To date, a limited number of studies have applied these approaches to profile miRNAs in different bovine tissues , ,  and only one study has used an NGS approach to investigate the expression of bovine miRNAs in response to infection .
In this study, we implemented a NGS approach to profile the expression of bovine miRNAs at multiple time-points in primary mammary epithelial cells infected in vitro with Streptococcus uberis, a causative agent of bovine mastitis. This inflammatory disease of the mammary gland has significant economic impact on the global dairy industry –. To the best of our knowledge, this study represents the most comprehensive NGS study to date that profiles the host miRNA response to infection, in any species. In comparison to previous studies, we have sequenced un-pooled miRNA libraries to a previously unprecedented sequencing depth from multiple replicates and controls across multiple time-points, allowing us to explore the statistically significant temporal changes in miRNA expression in response to infection.
Materials and Methods
Bovine Mammary Epithelial Cell Culture
Primary bovine mammary epithelial cells, which had been isolated from mammary parenchyma, were purchased from AvantiCell (AvantiCell Science Ltd., Ayr, UK) , . The source animal was in her third trimester of first pregnancy, was aged between 26–30 months, and was negative for bovine viral diarrhoea and Bovine spongiform encephalopathy. Cells were plated (seed density of 1×106) directly onto collagen coated plastic flasks (Greiner-Bio-One GmbH, Frickenhausen, Germany) and immersed in AvantiCell medium – (199/Ham’s F12 (50∶50) pH 7.4 containing 5% (v/v) horse serum, 5% (v/v) fetal bovine serum, 5 µg/ml bovine insulin, 1 µg/ml hydrocortisone, 3 µg/ml cortisol, 10 ng/ml epidermal growth factor (EGF), 2 mM sodium acetate, 10 mM Hepes, U/ml penicillin/streptomycin and single strength Fungizone™).
Media was initially replaced after 48 h. Cells were split twice (75 cm2 and 175 cm2) and were then seeded at a concentration of 1.8×105 cells/well into collagen coated 6-well plates. Media was then changed after 24 h, and cells were inspected under microscopy for confluence. Cells were harvested by washing with Hanks Balanced Salt Solution (HBSS) pH 7.4 and treated with 4 ml of 0.25% trypsin for approximately 5 min at 37°C. An equal volume of medium to trypsin (1∶1) neutralised trypsin.
Infection of Cells with Streptococcus uberis 0140J
Streptococcus uberis 0140J was purchased from the American Type Culture Collection (ATCC), Virginia, USA (Cat# BAA-854). S. uberis 0140J was first isolated in milk obtained from a clinical case of bovine mastitis in the United Kingdom in 1972. S. uberis was cultured as per ATCC instructions. BMEs were challenged with S. uberis 0140J at a multiplicity of infection (MOI) of 50, over a time course of 1, 2, 4, & 6 h. Three replicates were infected at each time point and three replicate uninfected controls were also maintained for each time point.
Total RNA and small RNA was extracted from each of the 24 samples using the mirVana™ miRNA Isolation Kit (Life Technologies, Carlsbad, CA, USA). Procedures were performed according to the manufacturer’s protocol. Briefly, cells were lysed using 500 µl lysis/binding solution directly on a culture plate. 50 µl of miRNA homogenate was added; solution was mixed by vortexing and left on ice for 10 min. 500 µl of acid-phenol chloroform was added and solution was mixed by vortexing for ∼60 sec. The solution was then centrifuged for 5 min at 10,000×g at room temperature to separate phases. Aqueous phase was removed and transferred to a separate tube. 1/3 volume of 100% ethanol was added to the aqueous phase and mixed by vortexing. Samples were passed through a filter cartridge (glass-fiber filter) by centrifuge for ∼30 sec at 10,000×g. The filtrate was collected (residue on filter contained RNA <200 nt, was retained for later use) and 2/3 volume 100% ethanol was added and mixed by vortexing. Filtrate was passed through a second filter cartridge by centrifuge for ∼30 sec at 10,000×g. The flow through was discarded, and the filter was washed with 700 µl wash solution 1 and 500 µl wash solution 2 (twice). After discarding all flow through after each step, the filter was centrifuged for a further 1 min. 50 µl of pre-heated (95°C) nuclease free water was applied to the filter for 1 min, and the filter was centrifuged for 30 sec. Eluate was collected and stored at −80°C. Total RNA integrity was measured by the Agilent RNA 6000 Nano Kit using the 2100 Bioanalyzer (Agilent Technologies, Colorado Springs, CO, USA). The Agilent Small RNA Kit (Agilent Technologies) was used to quantify miRNA.
Small RNAseq Library Preparation and Sequencing
Twenty-four indexed miRNA libraries were prepared using the ScriptMiner™ Small RNAseq Library Preparation Kit (Epicentre, Madison, WI, USA). Procedures were performed according to the manufacturer’s protocol. Briefly, a 3′-tagging sequence was added to the 3′- end of the RNA followed by treatment with a degradase enzyme to reduce excess 3′ adaptor oligo. A tagging sequence was then added to the 5′- end of the RNA. The RNA, now tagged at both ends (di-tagged), was purified using the Zymogen RNA Clean and Concentrator (Zymogen, Irvine, CA, USA). The di-tagged RNA was then reverse transcribed into cDNA, and the remaining RNA was removed using RNase. The PCR step used by the ScriptMiner™ kit, is a two stage process, firstly an analytical PCR step was carried out to optimise the number of cycles necessary for amplification. Once this was determined, the libraries were amplified and adaptors were added. Size fractionation of the miRNA libraries to separate them from adapter dimers was achieved by electrophoresis on an 8% TBE polyacrylamide gel (Life Technologies, Carlsbad, CA, USA) (1.00 mm×10 well). The libraries were then purified from the gel and the Agilent High Sensitivity DNA Kit (Agilent Technologies, Colorado Springs, CO, USA) was used to quantify the molarity and size of finished miRNA-seq libraries. miRNA libraries were randomised across three lanes of a flowcell, with eight indexed samples on each lane. Libraries were sequenced on an Illumina HiSeq 2000 by the Norwegian Sequencing Centre with TruSeq v3 reagents. Fastq files were produced using the CASAVA pipeline v1.8.2. Barcodes (indexes) and adaptor sequences for multiplexed samples are provided (Table S1).
Small RNAseq Analysis
Preliminary quality control analysis of the 24 fastq files was carried out with FASTQC software v0.10.0 (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Cutadapt v1.1 (http://www.cutadapt/) was then used to trim 3′ adaptor sequences. Reads which were shorter than 18 nucleotides after trimming were discarded. Trimmed reads were then further filtered using the fastq quality filter (http://hannonlab.cshl.edu/fastx_toolkit/) v0.0.13. Reads where at least 50% of the bases had a Phred score <20 were removed . Finally, reads passing all the above filters were also trimmed at their ends to remove low quality bases (Phred score <20). Reads which successfully passed filtering were aligned to the bovine genome (UMD3.1) using novoalign version 2.07.11 (http://www.novocraft.com) using the “-m” miRNA mode. Reads that did not uniquely align to the genome were discarded. HTSeq version 0.5.3p3 (http://www-huber.embl.de/users/anders/HTSeq/doc/overview.html) using the union model was used to assign uniquely aligned reads to Ensembl (v66) bovine gene and miRNA annotation (separately).
miRNAseq fastq files have been submitted to the NCBI Gene Expression Omnibus (GEO) database  with experiment series accession number GSE41278.
Differential Expression Analysis
Prior to assessing differential expression, count data were first normalised across libraries using either the trimmed mean of M-values (TMM) normalisation method  or upper-quantile normalisation . Differential expression analysis of miRNAseq data has been shown to be sensitive to the normalisation approach implemented . To address this issue, we identified differentially expressed miRNAs in three alternatively normalised datasets; TMM-normalised, upper-quantile normalised and no normalisation. Only miRNAs which were identified as differentially expressed across all three datasets were considered further i.e. the differential expression of these miRNAs was robust to the normalisation procedure. As an aside, we found that the two different normalisation approaches resulted in very similar miRNAs being detected as differentially expressed.
The R (version 2.14.1) Bioconductor package EdgeR (v2.4.6) , which uses a negative binomial distribution model to account for both biological and technical variability was applied to identify statistically significant differentially expressed miRNAs. Only miRNAs that had at least 1 count per million in at least 3 samples were analysed for evidence of differential gene expression. The analysis was undertaken using moderated tagwise dispersions. Differentially expressed miRNAs were defined as having a Benjamini and Hochberg  corrected P value of <0.05.
Novel miRNA Discovery
In addition to profiling the expression of known miRNAs, miRNAseq data can also be used to identify the expression of potentially novel miRNAs. To do this, miRNAseq data from this study was analysed using the software package miRDeep2 v0.0.5 . The miRDeep2 algorithm mines high-throughput sequencing data for the presence of multiple sequenced RNAs corresponding to predicted miRNA hairpin structures in the genome. It then uses Bayesian statistics to score the fit of sequenced RNAs to the biological model of miRNA biogenesis. MiRDeep2 predicted a large number of potentially novel miRNAs from our miRNAseq data. We further parsed this data using a number of different parameters to identify those novel miRNAs that have the highest likelihood of being true positives. Specifically, we identified those predictions where both the mature and star stands were expressed with a minimum of 5 reads each; where miRDeep2 predicted that the miRNA had >90% probability of being a true positive; where the hairpin structure had a significant Randfold p-value and where the novel miRNA was independently predicted in two or more different miRNAseq samples.
Customised Perl scripts were also written in house to examine the miRDeep2 output for the presence of miRNA isomiRs. These scripts were used to identify isomiRs that were expressed at a level of at least 100 reads and to identify cases where the expression of the isomiR was higher than the expression of the miRBase consensus mature sequence. Furthermore, we identified whether isomiRs were modified at the 5′ or 3′ ends (first and last 5 nucleotides). IsomiRs with >1 mismatch to the reference sequence were excluded from the analysis.
miRNA Target Predictions
Target genes that are potentially regulated by differentially expressed miRNAs were predicted using the consensus of two computational approaches, miRanda v3.3a  and TargetScan v6.2 , , . Given the high false positive rates for miRNA target prediction, we identified only those potential target genes that were predicted by both methods. More specifically, we first established a broad pool of potential targets by applying miRanda to bovine mature miRNA (miRBase v18) and cDNA sequences (UMD3.1, Ensembl v66) under default threshold settings. This resulted in the prediction of thousands of possible target genes per differentially expressed miRNA. To narrow down this pool of potential targets, we used TargetScan to independently identify conserved targets with a PCT-score above 0.9 and/or non-conserved targets with a context+ score above 90%. Target genes that were not corroborated by one of the two methods were discarded. Pathway analysis of predicted gene targets was undertaken using the SIGORA R package (http://cran.r-project.org/web/packages/sigora/index.html) with KEGG pathway annotations . Genes that have an annotated role in innate immunity were identified using www.innatedb.com , a curated database of innate immunity genes, pathways and molecular interactions.
Isolation of Small RNA from Bovine Mammary Epithelial Cells
Small RNA was isolated for NGS sequencing from S. uberis infected primary bovine mammary epithelial cells at 1, 2, 4 and 6 hours post-infection (hpi) (n = 3 infected and n = 3 controls at each time-point). Total RNA and small RNA were examined for quantity and integrity in each of the 24 samples (Table S2). Total RNA was assessed to be of high quality based on both Bioanalyzer and 28S/18S analysis. RNA integrity numbers (RIN) for total RNA were >8 for each sample. The concentration of miRNA in each sample was also assessed (Table S2). Sufficient quantities were present to proceed with RNAseq library preparation.
High-throughput Sequencing of Small RNA Libraries Prepared from Bovine Mammary Epithelial Cells
Small RNA libraries were prepared from size selected RNA (<200 nucleotides). Libraries were prepared using the ScriptMiner™ protocol with indexing before cluster generation, sequencing and imaging on an Illumina Hiseq 2000. Samples were randomly multiplexed over 3 flowcell lanes for sequencing. Sequencing of small RNA libraries yielded more than 450 million raw sequence reads from mammary epithelial cells. Following a pipeline of adaptor removal, quality filtering and the removal of sequences that were too short, more than 213 million reads were retained for further analysis (78,604,161 and 134,850,887 for control and infected replicates, respectively). These filtered reads were then aligned to the reference Bos taurus UMD 3.1 genome. Over 116 million reads aligned uniquely to the genome (Table S1). Reads that aligned to more than one position in the genome were discarded. Uniquely aligning reads were then assigned to known miRNAs using HTseq (http://www-huber.embl.de/users/anders/HTSeq/doc/overview.html) based on Ensembl v66 ,  annotation of the bovine genome.
Repertoire of RNA Species in Small RNA Libraries
The proportion of reads (averaged across 24 samples) uniquely aligning to different RNA biotypes demonstrates that miRNAs are the dominant ncRNA species sequenced in our small RNA libraries (Figure 1). The vast majority (>90%) of reads that align uniquely to known ncRNAs align to known miRNAs. There was no significant difference in the proportion of reads aligning to different RNA biotypes in the infected and control samples. The majority of the remaining reads primarily mapped to snoRNAs (Figure 1) . Although the vast majority of reads align to known bovine ncRNAs, a low density of reads can be observed along each chromosome (Figure S1). These possibly represent mRNA degradation products.
The Expression of miRNAs in Primary Bovine Mammary Epithelial Cells
To characterise the bovine mammary epithelial cell microRNome, miRNAs that were expressed at an appreciable level (based on mapped read counts in tags per million sequenced (tpm)) were identified. 276 miRNAs had a count of greater than 1 tpm (Table S3). Of these, 114 miRNAs were expressed at a level >100 tpm. To determine whether these miRNAs were expressed from related genomic regions, we examined all miRNAs with >100 tpm for genomic clustering (Figure 2). There was no evidence of a substantial genomic bias from which these miRNAs were encoded.
The position of other annotated bovine miRNAs is shown (blue).
The top 10 highly expressed miRNAs, which accounted for >80% of all aligned reads (Figure 3), were evolutionarily conserved across multiple species. These miRNAs represent seven different miRNA families; miR-let-7 (bta-let-7i & bta-miR-3596), miR-21 (bta-miR-21), miR-27 (bta-miR-27a & bta-miR-27b), miR-28 (bta-miR-151), miR-184 (bta-miR-184), miR-200 (bta-miR-200a & bta-miR-200b), and miR-205 (bta-miR-205). Many of these miRNAs have been shown to have pleiotropic roles in other species (Table 1). miR-21 and miR-205, have been shown to have role in cancer, regulating tumour suppressor genes such as VEGF-A and TGFI-R2 , , , .
Of particular interest to our study is the fact that several of the most highly expressed miRNAs in BMEs have been shown to have a role in immunity. miR-27b, for example, has been shown to negatively regulate the mRNA stability of peroxisome proliferator-activated receptor gamma (PPARgamma), a transcriptional regulator of the inflammatory response . Interestingly, miR-27b has also been found to be degraded by a viral transcript in lytic murine cytomegalovirus (MCMV) infection, further highlighting its role in immunity . miR-21 has also recently been shown to be the most highly expressed miRNA in Mycobacterium leprae infected monocytes and to negatively regulate the Vitamin D-dependent antimicrobial pathway . Evidence from previous studies also suggests that highly expressed miRNAs in BMEs may also regulate each other. miR-184, for example, has been demonstrated to antagonise miR-205 to maintain SHIP2 levels in epithelia .
Multiple miRNAs are Differentially Expressed in Response to S. uberis Infection
Once we had characterised which miRNAs were expressed in unchallenged bovine mammary epithelial cells, we then utilised the EdgeR statistical package  to determine which miRNAs were significantly differentially expressed in response to S. uberis infection at 1, 2, 4 and 6 hpi. It has been suggested that differential expression analysis of miRNAseq is sensitive to the normalisation approach implemented . To address this issue, we identified differentially expressed miRNAs in three alternatively normalised datasets; TMM-normalised , upper-quantile normalised and no normalisation. Only miRNAs which were identified as differentially expressed across all three datasets were considered as significantly differentially expressed i.e. the differential expression of these miRNAs was robust to the normalisation procedure. We found that the two different normalisation approaches actually resulted in very similar miRNAs being detected as differentially expressed.
Fifteen different miRNAs were identified as being significantly up-regulated in response to the S. uberis challenge. No miRNAs were identified as differentially expressed at 1 hpi. At 2 hpi, 2 miRNAs, bta-mir-29e and bta-mir-708, were found to be up-regulated (Figure 4). bta-mir-29e was subsequently observed to be down-regulated at 6hpi. At 4 hpi, bta-let-7b and bta-miR-98 were up-regulated (Figure 5). Additionally, bta-miR-let-7c and bta-miR-708 were observed to be up-regulated at 4 hpi when the miRNAseq count data were normalised (both methods), but this was not observed in the un-normalised data. At 6 hpi, 12 miRNAs were found to be up-regulated (Figure 6), including bta-let-7b, which was also up-regulated at 4 hpi.
The more red the color the more highly expressed that miRNA is. The heatmap was generated using the R v2.14.1 heatmap package.
The more red the color the more highly expressed that miRNA is. The heatmap was generated using the R v2.14.1 heatmap package.
Seven different miRNAs were identified as down-regulated in response to the S. uberis challenge. No miRNAs were down-regulated at 1 or 2 hpi. At 4 hpi, bta-miR-29b-2, bta-miR-193a, and bta-miR-130a were down-regulated. At 6 hpi, bta-miR-29b-2, bta-miR-29c, bta-miR-29e, bta-miR-100, bta-miR-130a and Ensembl predicted miRNA ENSBTAG00000047296, were down-regulated. Two miRNAs, bta-miR-29b-2 and bta-miR-130a were down-regulated at both 4 and 6 hpi. Additionally, bta-miR-15a, bta-miR-17, bta-miR-26a-2, bta-miR-29a, bta-miR-29b-1, and bta-miR-193a were identified as down-regulated in the normalised data (both methods), but not in the un-normalised data. Fold changes in expression for miRNAs that are differentially expressed at 4 and 6 hpi are shown in Figure S2 and S3.
These results indicate that there are rapid temporal changes in the expression of miRNAs in response to a Gram-positive infection, with different miRNA repertoires being identified as differentially expressed at time-points that are 2 hours apart.
The miRNA Response to the Gram-positive S. uberis is Markedly Different to the LPS miRNA Response
To date, many immune-relevant miRNAs have been identified as part of the host response to lipopolysaccharide (LPS) stimulation , , which is frequently used to mimic a Gram-negative bacterial infection. We have completed a literature survey and identified over 145 miRNAs that have been shown to be differentially expressed in response to LPS across multiple different species and tissues (Table S4). Eighty-four of the 145 LPS inducible miRNAs were found not be expressed above 1 tpm in BMEs. Of the 21 miRNAs that we identified as being differentially expressed in response to the Gram-positive S. uberis, only 9 of these (bta-let-7d, bta-let-7b, bta-mir-98, bta-miR-100, bta-mir-130a, bta-miR-193a, bta-miR-210, bta-miR-494, bta-miR-652) have also been reported to be differentially expressed in response to LPS in other species. Furthermore, 5 of these 9 (bta-miR-98, bta-miR-100, bta-miR-193a, bta-miR-210, bta-miR-494) show an inverse response to S. uberis infection in comparison to LPS. Most notably, bta-miR-100 and bta-miR-494, which were previously identified as up- and down-regulated, respectively, in mouse lung 6h post-stimulation with LPS, showed the inverse response at the same time-point in response to S. uberis infection , . This would suggest that the miRNA response to Gram-positive bacteria may be markedly different to Gram-negative.
Predicted Targets of Down-regulated miRNAs are Enriched for Genes with a Role in Innate Immunity
Target genes that are potentially regulated by differentially expressed miRNAs in response to S. uberis infection at 2, 4 and 6 hpi were predicted using two computational approaches, miRanda ,  and TargetScan , , . Given the high false positive rates for miRNA target prediction, we identified only those potential target genes that were predicted by both methods (Table 2). Target genes that were not corroborated by one of the two methods were discarded. In total 1,417 unique genes were predicted to be targeted by differentially expressed miRNAs (Table S5). This resulted in 2,491 miRNA-target interactions; 477 of these were targeted by down-regulated miRNAs; 1,921 were targeted by up-regulated miRNAs; and 93 were targeted by both up and down-regulated miRNAs. Because of the difficulties in accurately predicting miRNA targets, it is more appropriate to examine whether broad functional categories of genes are statistically over-represented among predicted target genes, rather than focusing on individual gene predictions. Statistical analysis (Hypergeometric test) revealed that the predicted target genes of down-regulated miRNAs at 4 and 6 hpi were significantly enriched (P = 0.01) in genes annotated by www.innatedb.com  as having a role in innate immunity (Figure 7). The predicted target genes of up-regulated miRNAs were not enriched for a role in innate immunity suggesting that up and down-regulated miRNAs target different processes in response to S. uberis infection.
Red arrow shapes represent up-regulated miRNAs at 4 hpi; green down-regulated. Red circles represent target genes that are annotated by www.innatedb.com as having a role in innate immunity. The network was constructed and visualised in Cytoscape v2.8.2. .
Pathway analysis of the predicted gene targets of up-regulated miRNAs (at 4 and 6 hpi), revealed that pathways which have been previously implicated in mastitis are statistically enriched among the predicted gene targets of up-regulated miRNAs (Table 3). These pathways include MAPK signalling; Cytokine-Cytokine Receptor Signalling and the JAK-STAT Signalling Pathway. The MAPK signalling pathway, for instance, has been identified as one of the top canonical pathways highlighted in a microarray study examining the bovine mammary tissue response to mastitis, 20 hpi with S. uberis . Many of the other pathways that we identified as statistically enriched among the predicted gene targets of up-regulated miRNAs were also highlighted as significant in this previous microarray study.
Taken together, these analyses strongly suggest that miRNAs that are differentially expressed during infection of BMEs with S. uberis are key regulators of the host response to this pathogen.
MicroRNA isomiRs are heterogeneous variants of canonical miRNA species, which are, increasingly, being suggested to be of functional importance . It has been suggested that these miRNA variants can be cell type specific, have functional differences, and vary in their response to biological stimuli . Evidence suggests that although isomiRs show similar expression patterns to their equivalent canonical miRNA, their targets can vary . Deletions at both the 5′ and 3′ end of isomiRs may change the specificity of the seed binding region effecting miRNA function.
We have found that the expression of isomiRs was common for the majority of BME expressed miRNAs (Table 4). 100 known miRNAs were found to have at least one isomiR expressed at a level of >100 reads and more than 1,000 different isomiRs were identified. Notably, in 40% of cases at least one isomiR was more highly expressed than the miRbase consensus sequence, suggesting that the isomiR should in fact be annotated as the consensus. On further examination, we found that isomiR ‘nibbling’ was 1.4 more times likely than post-transcriptional additions and isomiR editing was 2.3 times more likely to be 3′ modified than 5′ modified, agreeing with current literature . That said, almost 40% of isomiRs were 5′ modified, potentially impacting on which targets they regulate, though the majority of 5′ modified isomiRs were expressed at low levels. Further work is required to determine whether isomiRs have a functional role in response to infection.
Novel miRNA Discovery
In addition to profiling the expression of known miRNAs, miRNAseq data can also be used to identify the expression of potentially novel miRNAs. To do this, miRNAseq data from this study was analysed using the software package, miRDeep2 . We identified 21 high-confidence, putatively novel, bovine miRNAs that were independently predicted in multiple BME miRNAseq datasets (Table 5). Homology searching of the miRBase database (v 19)  using BLAST  identified that 2 of the novel miRNAs had 100% identity to known miRNAs in other species, ssc-miR-664-3p (pig) and hsa-miR-219-1 (human). Additionally 5 of the novel bovine miRNAs had significant homology with the bta-mir-2285 family. The bta-mir-2285 family has over 40 members spanning the entire bovine genome, . Two additional novel miRNAs showed homology to the bta-mir-2284 family. The remaining miRNAs did not show significant homology to other known miRNAs in other species. However, given the very high read counts observed for several of these predicted miRNAs, and the fact that were independently predicted in multiple different samples, it would suggest that many of these predictions represent true novel bovine miRNAs.
In this study, we have used a next generation sequencing approach to profile the expression of bovine miRNAs at multiple time-points in primary bovine mammary epithelial cells (BMEs) infected in vitro with S. uberis, a causative agent of bovine mastitis. In comparison to previous NGS studies investigating the host miRNA response to infection, we have sequenced un-pooled miRNA libraries to a previously unprecedented sequencing depth from multiple replicates and controls across multiple time-points, allowing us to explore statistically significant temporal changes in miRNA expression in response to infection. Analysing over 450 million sequencing reads, we found that approximately 20% of known bovine miRNAs are expressed in BMEs. A similar diversity of miRNA expression has also been recently reported in other tissues, including bovine retinal microvascular endothelial cells (RMECs)  and in testicular and ovarian tissues . As has also been reported in other studies, there is a significant dynamic range in the expression of known miRNAs in BMEs. A few miRNAs are expressed at very high levels, with the majority being expressed at low levels. The top 10 most highly expressed miRNAs account for >80% of all aligned reads and are highly conserved across species. Whether or not the other more lowly expressed miRNAs play a significant biological role remains an open question.
We have also found that the expression of isomiRs was common for many of the BME expressed miRNAs. Most significantly, in 40% of cases, at least one isomiR was more highly expressed than the miRbase consensus sequence, indicating that studies such as ours can also be used to improve miRNA annotation. In particular, changes to the 5′ of the consensus sequence could lead to dramatically different target genes being computationally predicted. We identified a large number of 5′ isomiRs although they were mostly expressed at low levels.
In addition to profiling known miRNAs, we have also analysed the sequencing data to identify potentially novel bovine miRNAs. Twenty-one high-confidence, putatively novel, bovine miRNAs were identified independently across multiple different samples. The mature sequences of two of the novel miRNAs were 100% identical to known miRNAs in other species. Seven of the other predicted miRNAs exhibited significant homology to two bovine miRNA families, bta-mir-2284 and bta-mir-2285.
Few studies have previously investigated temporal changes in global miRNA expression using an NGS approach , although several have used microarray technology , , , . Thus far, many immune-relevant miRNAs have been identified as part of the host response to LPS stimulation , . We completed a literature survey and identified more than 145 miRNAs across multiple different species and tissues that have been shown to be LPS responsive. In our study, the miRNA response to the Gram-positive S. uberis was markedly different to the reported LPS miRNA response. Over the 6 hour time-course, we identified 21 known bovine miRNAs as significantly differentially expressed in response to the S. uberis challenge. Only 9 of these miRNAs have also been reported to be differentially expressed in response to LPS. For those in common, an inverse pattern of expression was observed in 5 of the 9 cases suggesting that the miRNA response to Gram-positive bacteria may be markedly different to Gram-negative. Further global studies of the miRNA response to Gram-positive and Gram-negative bacteria in the same tissue at the same time-points will be required to confirm this.
It is notable that we found most miRNAs were differentially expressed at different time-points post-infection, suggesting that miRNAs exhibit rapid dynamics in their temporal expression. It is also notable that the majority of miRNAs that we report as being differentially expressed exhibit relatively subtle changes in gene expression in response to infection. This subtle change in expression is in line with existing literature and strengthens the hypothesis that miRNAs are fine-tuners of gene expression , . For example, miR-let-7d, miR-652 and miR-494 demonstrated similar levels of differential expression 6 hours post LTA stimulation in mouse tissues .
Computational analysis revealed that the predicted target genes of S. uberis down-regulated miRNAs were statistically enriched for roles in innate immunity. This would suggest that these miRNAs may significantly regulate the sentinel capacity of mammary epithelial cells to mobilise the innate immune system . Pathway analysis of the predicted targets of up-regulated miRNAs has also identified the statistical over-representation of several pathways previously implicated in the host response to mastitis, such as the MAPK, JAK-STAT and other cytokine signaling pathways. Furthermore, several of the differentially expressed miRNAs have been shown to have roles in the immune systems of other species. For example, bta-let-7 miRNAs were up-regulated at both 4 and 6 hours post-infection with S. uberis. The let-7 family has been extensively described in the literature for having a role in immunity. The down-regulation of let-7 family members, for example, was shown to promote expression of IL-10 and IL-6 in HeLa cells infected with Salmonella enterica serovar Typhimurium . The observed up-regulation of let-7 miRNAs in our study may lead to the repression of anti-inflammatory cytokines to promote innate immunity.
We also report the down-regulation of two other miRNAs, bta-miR-29b-2 and bta-miR-130a, both of which have known roles in immunity and infection in other species. miR-29a/miR-29b down-regulation has been demonstrated to facilitate IFN-γ up-regulation in NK cells and TH1 cells , . IFN- γ is well known as an innate inflammatory mediator and its secretion promotes host resistance against viral and intracellular bacteria. Furthermore, IFN-γ mRNA expression has been demonstrated in human mammary epithelial cells , suggesting that this may be a relevant target in our model. LPS induced TNF-α expression in neonatal and adult monocytes has been shown to be greatly suppressed by the induction of miR-130a . Taken together, the evidence strongly suggests that the differentially expressed miRNAs identified in this study are likely central regulators of the innate immune response to S. uberis and thus represent potential therapeutic targets or novel biomarkers of infection and inflammation.
Read coverage along chromosome 26 (25 nt windows). The higher the read density the darker the red colour. Green regions represent positions where the read density is < 5 reads, Grey = no reads.
Fold changes of differentially expressed miRNAs at 4 hours post-infection (hpi).
Fold changes of differentially expressed miRNAs at 6 hours post-infection (hpi).
Summary read statistics, barcodes and adaptor sequences for each mammary epithelial cell miRNAseq library.
RNA integrity, quantity, 28S/18S ratio, and miRNA quantity for each of the 24 samples.
miRNAs that are expressed at ≥1 tag per million in bovine mammary epithelial cells.
Summary of lipopolysaccharide (LPS) responsive microRNAs. 145 miRNAs that have been identified as being differentially expressed in response to LPS across multiple different species and tissues.
The authors would like to thank Dr. Kieran Meade and Dr. Orla Keane in Teagasc for contributions to the experimental design. The sequencing service was provided by the Norwegian Sequencing Centre (www.sequencing.uio.no), a national technology platform hosted by the University of Oslo and supported by the “Functional Genomics” and “Infrastructure” programs of the Research Council of Norway and the Southeastern Regional Health Authorities.
Conceived and designed the experiments: NL DJL. Performed the experiments: NL MM. Analyzed the data: NL AF COF DJL. Contributed reagents/materials/analysis tools: NL DJL. Wrote the paper: NL COF DJL.
- 1. Bi Y, Liu G, Yang R (2009) MicroRNAs: novel regulators during the immune response. J Cell Physiol 218: 467–472.
- 2. Lee RC, Feinbaum RL, Ambros V (1993) The C. elegans heterochronic gene lin-4 encodes small RNAs with antisense complementarity to lin-14. Cell 75: 843–854.
- 3. Kozomara A, Griffiths-Jones S (2011) miRBase: integrating microRNA annotation and deep-sequencing data. Nucleic Acids Res 39: D152–157.
- 4. O’Connell RM, Rao DS, Chaudhuri AA, Baltimore D (2010) Physiological and pathological roles for microRNAs in the immune system. Nat Rev Immunol 10: 111–122.
- 5. Williams AE, Perry MM, Moschos SA, Larner-Svensson HM, Lindsay MA (2008) Role of miRNA-146a in the regulation of the innate immune response and cancer. Biochem Soc Trans 36: 1211–1215.
- 6. Xiao C, Calado DP, Galler G, Thai TH, Patterson HC, et al. (2007) MiR-150 controls B cell differentiation by targeting the transcription factor c-Myb. Cell 131: 146–159.
- 7. Coutinho LL, Matukumalli LK, Sonstegard TS, Van Tassell CP, Gasbarre LC, et al. (2007) Discovery and profiling of bovine microRNAs from immune-related and embryonic tissues. Physiol Genomics 29: 35–43.
- 8. Jin W, Grant JR, Stothard P, Moore SS, Guan LL (2009) Characterization of bovine miRNAs by sequencing and bioinformatics analysis. BMC Mol Biol 10: 90.
- 9. Xu G, Zhang Y, Jia H, Li J, Liu X, et al. (2009) Cloning and identification of microRNAs in bovine alveolar macrophages. Mol Cell Biochem 332: 9–16.
- 10. Dilda F, Gioia G, Pisani L, Restelli L, Lecchi C, et al. (2011) Escherichia coli lipopolysaccharides and Staphylococcus aureus enterotoxin B differentially modulate inflammatory microRNAs in bovine monocytes. Vet J 192: 514–516.
- 11. Naeem A, Zhong K, Moisa SJ, Drackley JK, Moyes KM, et al. (2012) Bioinformatics analysis of microRNA and putative target genes in bovine mammary tissue infected with Streptococcus uberis. J Dairy Sci 95: 1–12.
- 12. Hou Q, Huang J, Ju Z, Li Q, Li L, et al. (2011) Identification of Splice Variants, Targeted MicroRNAs and Functional Single Nucleotide Polymorphisms of the BOLA-DQA2 Gene in Dairy Cattle. DNA Cell Biol 31: 739–744.
- 13. Flicek P, Amode MR, Barrell D, Beal K, Brent S, et al. (2012) Ensembl 2012. Nucleic Acids Res 40: D84–90.
- 14. Hubbard TJ, Aken BL, Ayling S, Ballester B, Beal K, et al. (2009) Ensembl 2009. Nucleic Acids Res 37: D690–697.
- 15. Kim VN, Han J, Siomi MC (2009) Biogenesis of small RNAs in animals. Nat Rev Mol Cell Biol 10: 126–139.
- 16. Buermans HP, Ariyurek Y, van Ommen G, den Dunnen JT, ‘t Hoen PA (2010) New methods for next generation sequencing based microRNA expression profiling. BMC Genomics 11: 716.
- 17. Lee LW, Zhang S, Etheridge A, Ma L, Martin D, et al. (2010) Complexity of the microRNA repertoire revealed by next-generation sequencing. RNA 16: 2170–2180.
- 18. Huang J, Ju Z, Li Q, Hou Q, Wang C, et al. (2011) Solexa sequencing of novel and differentially expressed microRNAs in testicular and ovarian tissues in Holstein cattle. Int J Biol Sci 7: 1016–1026.
- 19. Chen X, Gao C, Li H, Huang L, Sun Q, et al. (2010) Identification and characterization of microRNAs in raw milk during different periods of lactation, commercial fluid, and powdered milk products. Cell Res 20: 1128–1137.
- 20. Guduric-Fuchs J, O’Connor A, Cullen A, Harwood L, Medina RJ, et al. (2012) Deep sequencing reveals predominant expression of miR-21 amongst the small non-coding RNAs in retinal microvascular endothelial cells. J Cell Biochem 113: 2098–2111.
- 21. Glazov EA, Kongsuwan K, Assavalapsakul W, Horwood PF, Mitter N, et al. (2009) Repertoire of bovine miRNA and miRNA-like small regulatory RNAs expressed upon viral infection. PLoS One 4: e6349.
- 22. Hogeveen H, Pyorala S, Waller KP, Hogan JS, Lam TJGM, et al.. (2011) Current status and future challenges in mastitis research. National Mastitis Council Proceedings. Available: http://www.milkproduction.com/Library/Scientific-articles/Animal-health/Current-status-and-future-challenges-in-mastitis-research/. Accessed 25 October 2012.
- 23. Berry DP, Amer PR (2005) Derivation of a health sub-index for the Economic Breeding Index in Ireland. Technical report to the Irish Cattle Breeding Federation. Available: http://www.icbf.com/publications/files/Derivation_of_health_sub-index_with_loco.doc. Accessed 25 October 2012.
- 24. Blatchford DR, Quarrie LH, Tonner E, McCarthy C, Flint DJ, et al. (1999) Influence of microenvironment on mammary epithelial cell survival in primary culture. J Cell Physiol 181: 304–311.
- 25. Blatchford DR, Wilde CJ, Matsuda T, N A (1999) Milk secretion in cultured mammary epithelial cells. Animal Cell Technology: Basic and Applied Aspects: Proceedings of the 10th Annual Meeting of the Japanese Association for Animal Cell Technology, Nago 10: 141–145.
- 26. Cock PJ, Fields CJ, Goto N, Heuer ML, Rice PM (2010) The Sanger FASTQ file format for sequences with quality scores, and the Solexa/Illumina FASTQ variants. Nucleic Acids Res 38: 1767–1771.
- 27. Barrett T, Troup DB, Wilhite SE, Ledoux P, Evangelista C, et al. (2011) NCBI GEO: archive for functional genomics data sets–10 years on. Nucleic Acids Res 39: D1005–1010.
- 28. Robinson MD, McCarthy DJ, Smyth GK (2010) edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics 26: 139–140.
- 29. Bullard JH, Purdom E, Hansen KD, Dudoit S (2010) Evaluation of statistical methods for normalization and differential expression in mRNA-Seq experiments. BMC Bioinformatics 11: 94.
- 30. Garmire LX, Subramaniam S (2012) Evaluation of normalization methods in mammalian microRNA-Seq data. RNA 18: 1279–1288.
- 31. Benjamini Y, Hochberg Y (1995) Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. J R Stat Soc Ser B Stat Methodol 57: 289–300.
- 32. Mackowiak SD (2011) Identification of novel and known miRNAs in deep-sequencing data with miRDeep2. Curr Protoc Bioinformatics Chapter 12: Unit 12 10.
- 33. Betel D, Wilson M, Gabow A, Marks DS, Sander C (2008) The microRNA.org resource: targets and expression. Nucleic Acids Res 36: D149–153.
- 34. Lewis BP, Burge CB, Bartel DP (2005) Conserved seed pairing, often flanked by adenosines, indicates that thousands of human genes are microRNA targets. Cell 120: 15–20.
- 35. Grimson A, Farh KK, Johnston WK, Garrett-Engele P, Lim LP, et al. (2007) MicroRNA targeting specificity in mammals: determinants beyond seed pairing. Mol Cell 27: 91–105.
- 36. Friedman RC, Farh KK, Burge CB, Bartel DP (2009) Most mammalian mRNAs are conserved targets of microRNAs. Genome Res 19: 92–105.
- 37. Kanehisa M, Goto S, Sato Y, Furumichi M, Tanabe M (2012) KEGG for integration and interpretation of large-scale molecular data sets. Nucleic Acids Res 40: D109–114.
- 38. Lynn DJ, Winsor GL, Chan C, Richard N, Laird MR, et al. (2008) InnateDB: facilitating systems-level analyses of the mammalian innate immune response. Mol Syst Biol 4: 218.
- 39. Wu H, Zhu S, Mo YY (2009) Suppression of cell growth and invasion by miR-205 in breast cancer. Cell Res 19: 439–448.
- 40. Yue X, Wang P, Xu J, Zhu Y, Sun G, et al. (2012) MicroRNA-205 functions as a tumor suppressor in human glioblastoma cells by targeting VEGF-A. Oncol Rep 27: 1200–1206.
- 41. Yang R, Schlehe B, Hemminki K, Sutter C, Bugert P, et al. (2010) A genetic variant in the pre-miR-27a oncogene is associated with a reduced familial breast cancer risk. Breast Cancer Res Treat 121: 693–702.
- 42. Yu Y, Kanwar SS, Patel BB, Oh PS, Nautiyal J, et al. (2012) MicroRNA-21 induces stemness by downregulating transforming growth factor beta receptor 2 (TGFbetaR2) in colon cancer cells. Carcinogenesis 33: 68–76.
- 43. Jennewein C, von Knethen A, Schmid T, Brune B (2010) MicroRNA-27b contributes to lipopolysaccharide-mediated peroxisome proliferator-activated receptor gamma (PPARgamma) mRNA destabilization. J Biol Chem 285: 11846–11853.
- 44. Marcinowski L, Tanguy M, Krmpotic A, Radle B, Lisnic VJ, et al. (2012) Degradation of cellular mir-27 by a novel, highly abundant viral transcript is important for efficient virus replication in vivo. PLoS Pathog 8: e1002510.
- 45. Liu PT, Wheelwright M, Teles R, Komisopoulou E, Edfeldt K, et al. (2012) MicroRNA-21 targets the vitamin D-dependent antimicrobial pathway in leprosy. Nat Med 18: 267–273.
- 46. Yu J, Ryan DG, Getsios S, Oliveira-Fernandes M, Fatima A, et al. (2008) MicroRNA-184 antagonizes microRNA-205 to maintain SHIP2 levels in epithelia. Proc Natl Acad Sci U S A 105: 19300–19305.
- 47. Qi J, Qiao Y, Wang P, Li S, Zhao W, et al. (2012) microRNA-210 negatively regulates LPS-induced production of proinflammatory cytokines by targeting NF-kappaB1 in murine macrophages. FEBS Lett 586: 1201–1207.
- 48. O’Neill LA, Sheedy FJ, McCoy CE (2011) MicroRNAs: the fine-tuners of Toll-like receptor signalling. Nat Rev Immunol 11: 163–175.
- 49. Moschos SA, Williams AE, Perry MM, Birrell MA, Belvisi MG, et al. (2007) Expression profiling in vivo demonstrates rapid changes in lung microRNA levels following lipopolysaccharide-induced inflammation but not in the anti-inflammatory action of glucocorticoids. BMC Genomics 8: 240.
- 50. Hsieh CH, Rau CS, Jeng JC, Chen YC, Lu TH, et al. (2012) Whole blood-derived microRNA signatures in mice exposed to lipopolysaccharides. J Biomed Sci 19: 69.
- 51. Enright AJ, John B, Gaul U, Tuschl T, Sander C, et al. (2003) MicroRNA targets in Drosophila. Genome Biol 5: R1.
- 52. Moyes KM, Drackley JK, Morin DE, Bionaz M, Rodriguez-Zas SL, et al. (2009) Gene network and pathway analysis of bovine mammary tissue challenged with Streptococcus uberis reveals induction of cell proliferation and inhibition of PPARgamma signaling as potential mechanism for the negative relationships between immune response and lipid metabolism. BMC Genomics 10: 542.
- 53. Cloonan N, Wani S, Xu Q, Gu J, Lea K, et al. (2011) MicroRNAs and their isomiRs function cooperatively to target common biological pathways. Genome Biol 12: R126.
- 54. Neilsen CT, Goodall GJ, Bracken CP (2012) IsomiRs - the overlooked repertoire in the dynamic microRNAome. Trends Genet 28: 544–549.
- 55. Peng Z, Cheng Y, Tan BC, Kang L, Tian Z, et al. (2012) Comprehensive analysis of RNA-Seq data reveals extensive RNA editing in a human transcriptome. Nat Biotechnol 30: 253–260.
- 56. Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ (1990) Basic local alignment search tool. J Mol Biol 215: 403–410.
- 57. Cui L, Guo X, Qi Y, Qi X, Ge Y, et al. (2010) Identification of microRNAs involved in the host response to enterovirus 71 infection by a deep sequencing approach. J Biomed Biotechnol 2010: 425939.
- 58. Mukhopadhyay P, Brock G, Pihur V, Webb C, Pisano MM, et al. (2010) Developmental microRNA expression profiling of murine embryonic orofacial tissue. Birth Defects Res A Clin Mol Teratol 88: 511–534.
- 59. Reinsbach S, Nazarov PV, Philippidou D, Schmitt M, Wienecke-Baldacchino A, et al. (2012) Dynamic regulation of microRNA expression following Interferon-gamma-induced gene transcription. RNA Biol 9: 978–989.
- 60. Li Z, Branham WS, Dial SL, Wang Y, Guo L, et al. (2010) Genomic analysis of microRNA time-course expression in liver of mice treated with genotoxic carcinogen N-ethyl-N-nitrosourea. BMC Genomics 11: 609.
- 61. Wang FZ, Weber F, Croce C, Liu CG, Liao X, et al. (2008) Human cytomegalovirus infection alters the expression of cellular microRNA species that affect its replication. J Virol 82: 9065–9074.
- 62. Bartel DP (2009) MicroRNAs: target recognition and regulatory functions. Cell 136: 215–233.
- 63. Swamy M, Jamora C, Havran W, Hayday A (2010) Epithelial decision makers: in search of the ‘epimmunome’. Nat Immunol 11: 656–665.
- 64. Schulte LN, Eulalio A, Mollenkopf HJ, Reinhardt R, Vogel J (2011) Analysis of the host microRNA response to Salmonella uncovers the control of major cytokines by the let-7 family. Embo J 30: 1977–1989.
- 65. Ma F, Xu S, Liu X, Zhang Q, Xu X, et al. (2011) The microRNA miR-29 controls innate and adaptive immune responses to intracellular bacterial infection by targeting interferon-gamma. Nat Immunol 12: 861–869.
- 66. Smith KM, Guerau-de-Arellano M, Costinean S, Williams JL, Bottoni A, et al. (2012) miR-29ab1 Deficiency Identifies a Negative Feedback Loop Controlling Th1 Bias That Is Dysregulated in Multiple Sclerosis. J Immunol 189: 1567–1576.
- 67. Khalkhali-Ellis Z, Abbott DE, Bailey CM, Goossens W, Margaryan NV, et al. (2008) IFN-gamma regulation of vacuolar pH, cathepsin D processing and autophagy in mammary epithelial cells. J Cell Biochem 105: 208–218.
- 68. Huang HC, Yu HR, Huang LT, Chen RF, Lin IC, et al. (2012) miRNA-125b regulates TNF-alpha production in CD14+ neonatal monocytes via post-transcriptional regulation. J Leukoc Biol 92: 171–182.
- 69. Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, et al. (2003) Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res 13: 2498–2504.