5. Research Design and Methods, 5.1 Research Area: This application addresses the broad challenge area “(15) Translational Science” and specific challenge topic, “15-HL-102: Develop new therapeutic strategies for heart, lung, and blood diseases based on microRNA technology”. The project title is “microRNA Regulation of Human Airway Epithelial Cell Phenotype”. 5.2 Challenge and Potential Impact: It is estimated that >22 million people in the USA currently have asthma and that >15 million people have chronic obstructive pulmonary disease (COPD), which is the fourth leading cause of death (data from Centers for Disease Control and Prevention). As well, there are other less prevalent, but individually devastating, obstructive lung diseases such as cystic fibrosis (CF), non-CF bronchiectasis and primary ciliary dyskinesia (PCD). These conditions place massive burdens on health care resources and immeasurably degrade human spirit and potential. All of them are focused on the lung airways, which become characteristically inflamed and remodeled, causing functional impairment. The airway epithelium is strategically located at the lung-environment interface. In addition to its well-known barrier and protective functions, it is now widely appreciated that the epithelium initiates, integrates and orchestrates host immune responses and airway remodeling via secretion of regulatory factors within complex feedback loops {}. Phenotypic changes in the airway such as enhanced epithelial shedding and turnover, deficient repair, mucous secretory (goblet) cell hyperplasia and squamous metaplasia are variably present and are integral to the progression of inflammatory airway diseases. Despite recent advances, many basic mechanisms regulating altered structure and function of hBE cells remain poorly understood, and there are no specific therapies directed at preventing or reversing disease-related phenotypic changes in the airway epithelium. In the last decade, “omics” have revolutionized our understanding of biology and disease, specifically including numerous microarray studies of airway mRNA in asthma and COPD {14248} {14244}. We now appreciate that non-coding, small RNA’s including microRNAs (miRNAs) are critical regulators of genomes and genes. Major effects of miRNA’s are to destabilize mRNA and to inhibit protein translation, but there are also examples of miRNA positive regulation of transcription. In any case, a regulatory network between miRNAs and gene expression is clearly involved in almost every aspect of biology from stem cell maintenance and embryonic development / cell differentiation to cancer (reviewed in {14264}). Recent, but limited, studies provide insight that miRNAs are important in lung development {14253} {14293}, lung inflammation {14238}, the biological response to tobacco smoking {14287}, and lung cancer {14250}. However, we do not know the full miRNA repertoire of the airway epithelium, nor do we have a comprehensive understanding the miRNAmRNA regulatory networks controlling hBE cell phenotype in health and disease. A unique combination of resources and talents at the University of North Carolina is poised to fill this major lung biology knowledge gap. The Randell laboratory is a leading provider of primary human bronchial epithelial (hBE) cells in models recapitulating their normal structure and function and has extensive knowledge of airway epithelial cell biology {}. The Hammond and Hayes laboratories provide cutting edge expertise in miRNA biology and bioinformatics, respectively {14223} {14253} {14252}. Supported by provocative preliminary data that miRNA expression changes dramatically as a function of hBE cell differentiation, these groups have combined forces with a plan to rapidly create new data that will significantly advance our understanding of miRNA regulation of airway epithelial phenotype during normal growth and differentiation and under a series of highly relevant injury/repair conditions. 5.3 Approach, Strategy: Our basic hypothesis is that miRNAs are critical regulators of hBE cell differentiation and the response to injury, that miRNAs become altered in pathologic states, and that modulating miRNAs can change hBE cell phenotype for Figure 1. Time course of hBE cell differentiation in a replicate of ALI therapeutic benefit. This hypothesis is cultures used for preliminary miRNA arrays. Formalin paraffin sections based on initial studies showing stained with H&E, or AB-PAS for glucoconjugates (blue), illustrate the dramatic regulation of miRNAs as a progression from poorly- to well-differentiated. function of hBE cell differentiation in vitro (Figures 1-3). In this experiment, primary hBE cells were harvested from explanted lungs of six individual tissue donors. The initial primary cell culture stage was on plastic, where the cells predictably de-differentiated and proliferated but were unable to polarize and re-differentiate. The cells were then passaged from plastic to a porous support and were grown at an air-liquid interface (ALI) where they again proliferated but were now able to fully differentiate, recapitulating the normal pseudostratified airway epithelial morphology (Figure 1). Total RNA was harvested on days 5, 14 and 35 after seeding under ALI conditions, representing distinct differentiation stages. The 3’ hydroxyl group of miRNA was fluorescently labeled with P-cytidyl-uridyl-Cy3 using T4-RNA ligase and RNA was hybridized to home-spotted microarray slides, consisting of 527 unique oligonucleotides complementary to putative human miRNAs in the Sanger miRBase {14252}. The slides were scanned using the 532-nM channel on a Genepix scanner and raw expression data was extracted and analyzed using GeneCluster, TreeView and Significance Analysis of Microarrays (SAM). Median centered and normalized miRNA expression data of hBE cell cultures from 6 individuals at day 5, 14 and 35, subjected to unsupervised hierarchical clustering (Figure 2), revealed complete separation of the day 5 group from the later time points. The day 14 and 35 groups also tended to cluster but there was some admixture. Significance Analysis of Microarrays (SAM) software called 49 genes as significantly different at a 0% false discovery rate (Figure 3). The predominant expression pattern was a steady increase in expression of several miRNAs from day 5 to day 35, but there were examples of equivalent increases at day 14 and 35. Three miRNAs definitively decreased from day 5 to day 35. Manual observation of the data revealed additional borderline and/or inconsistent changes in several miRNAs. Furthermore, many miRNAs not reliably detected by array technology likely changed but were not visible to us. Manual searching of the literature for functions of highly regulated miRNAs indicated several associated with cellular differentiation in other systems, tumor suppressors, and those with variable expression in different cancers. However, other strongly regulated miRNAs with unknown function were present. These results strongly indicate a key role for miRNAs in regulation of hBE cell proliferation and differentiation and illustrate significant gaps in our knowledge. Figure 2. The above results underlie the strategy for this Unsupervised proposal. Namely, that: 1) miRNAs are important hierarchical clustering of hBE cell miRNA regulators of the phenotype of the airway epithelium; 2) expression data. technology exists enabling comprehensive miRNA Cultures from 6 analysis through both unbiased discovery using deep individuals (1-6) at sequencing, and via state-of-the-art microarray and day (D) 5, 14 and 35 RT-PCR technology for known miRNAs, and 3) in vitro were studied. studies are not plagued by the uncertainty of variable non-epithelial cell contributions to the expression pattern, and are amenable to genetic manipulation for functional testing. Our team has the optimal combination of resources and experience. The hBE cells themselves, which are prohibitively expensive when purchased commercially, will be provided to this project from our existing collection at no expense, which enables more ambitious experimentation. It is important to note, based on our prior array experience that there can be significant variability in miRNA and mRNA expression in hBE cells from human to human, which can be compounded by experimental variability. Our ability to simultaneously study cells from a representative sample of 6 people minimizes one source of experimental variability. Furthermore, we will use triplicate cultures from each donor within each group and will pool RNA to minimize contributions from one untoward, outlier culture well. The cells will be employed in state-of-of-the-art models of key disease processes that are either already well established in our labs or easily achievable. Our team includes world-class, expert co-investigators in Figure 3. Expression map of both miRNA molecular biology and bioinformatics, which are key to achieving miRNAs called as significant by the experimental goals. SAM. Cultures from 6 We note our strategy to perform mRNA expression analysis in the individuals (1-6) at day (D) 5, 14 identical set of cultures as for the miRNA analyses, which entails considerable and 35 were studied. Genes extra effort. Theoretically, we could have explored existing databases for the (rows) are median centered and clustered (tree not shown). mRNA data. However, such data does not exist for all the experimental Arrays (columns) are median conditions, and it is likely that specific technical differences between different centered but not clustered. laboratories would inhibit direct comparisons. Thus, we think that using the identical samples is the optimal strategy to determine the miRNA regulatory network governing phenotypic changes in hBE cells in health and disease. Finally, we already have RNA samples for 2 of the seven sets of experiments in hand, which is compatible with the goal of the ARRA to quickly increase the use of Core facilities, create new hires and generate novel data that accelerates the pace of scientific discovery. Specific Aim 1) Develop a comprehensive portrait of miRNA expression in hBE cells during normal differentiation and in a spectrum of relevant injury/repair conditions. Rationale: miRNAs play a vital regulatory role in many cell processes and their expression is highly cell typeand differentiation stage-specific. Human bronchial epithelial cells are central participants in the pathogenesis of several extremely important lung diseases and, despite provocative early evidence in the literature and preliminary data in this proposal, there is minimal miRNA data in this key cell type. In vitro model systems recapitulating in vivo structure and function are critical milestone tests for therapeutic development and translation. We are uniquely poised to provide a comprehensive picture of the miRNA repertoire in hBE cells and their expression patter and cell localization during normal differentiation and in a series of highly relevant disease models. We will also perform mRNA arrays from the identical cell cultures in order to elucidate miRNA:gene expression networks. Aim 1A. Identify novel hBE cell miRNAs by creating libraries and performing deep sequencing from: 1) the ALI hBE cell differentiation time course; 2) the response of well differentiated cells to wounding; 3) the response to acute and chronic exposure to bacterial products; 4) exposure to the Th2-type cytokine IL-13, which induces mucous secretory cell hyperplasia; 5) exposure to cigarette smoke gas phase; 6) simulated ambient ozone exposures and 7) squamous metaplasia induced by retinoic acid deficiency. Methods: We begin this section with a brief general discussion of the cell culture model and overall experimental design and then a short piece on each of the 7 experimental conditions. Since the downstream analytical methods are common to each of the groups, we follow with a single section with methods for unbiased miRNA discovery by deep sequencing. The ALI hBE cell culture methods have been described in detail previously {}. For all experiments, cells from 6 different tissue donors will be employed. Within each donor, each experimental group (time point, exposure etc.) will consist of 3 replicate wells for RNA harvest, which results in excess RNA, but prevents one potentially bad well from ruining a data point and allows for replicates in downstream quantitative analyses (qRT-PCR). Each well is inspected at each media change, culling any that are unusual. In experiments involving potentially toxic exposures, we assess lactate dehydrogenase release into the apical media and basolateral inflammatory cytokine (IL-8, Gro) production. Replicate wells are also prepared for histology (H&E, other stians and in situ hybridization) using fixation with 4% formaldehyde followed by paraffin embedding and blocks for frozen sections. RNA is extracted using a guanidinum:chloroform/phenol protocol optimized for hBE cells to prevent glycoconjugate (mucous) contamination of RNA. An aliquot is assessed by spectrophotometry (Nano Drop) and using LabChips on an Agilent 2100 Bioanalyzer. The ALI hBE cell differentiation time course. As illustrated above in the Strategy section (Figure 1), the time course of hBE cell growth at an ALI recapitulates a process similar to wound repair in vivo in which poorlydifferentiated cells undergo a phase of rapid proliferation, followed by polarization, and ultimately differentiation into the characteristic airway epithelial cell types, namely basal, secretory and ciliated cells. As noted above, we have chosen 3 time points, days 5, 14 and 35 that represent distinct stages in the process. Cells at day 5 are almost uniformly squamoid and highly proliferative with little expression of differentiation markers, whereas by Day 14 apical membrane polarization and mucous secretory cell development are evident. By Day 35 large numbers of ciliated cells are present as well as fully mature basal and secretory cells. There is a single study of mRNA expression during the ALI hBE cell time course in the literature {} but no miRNA data. Cell culture, RNA harvest and preparation of histology sections of triplicate specimens Figure 4. Wound repair model. A) Complementary pair of probes to from an n = 6 different tissue donors is already complete and ready for wound ALI cultures. B) Probes uniform downstream processing as described below. placed in 12 mm diameter Millicell The response of well differentiated cells to wounding. The normal inserts. C) hBE cells 24 hours after airway epithelium in vivo is mitogenically quiescent but responds wounding, repairing cells (left of dramatically to wounding via cell de-differentiation, migration, proliferation, dash) showing nuclear localization and re-differentiation to ultimately restore an intact epithelium, which is of EGR-1 transcription factor. critical to prevent airway obliteration. We have studied hBE cell wound repair {} and have recently developed a unique device for wounding ALI cultures that enables subsequent harvest of just the non-wounded and wound-repairing cell compartments (Figure 4). To our knowledge there are no miRNA or mRNA gene array studies of hBE cell wound repair. To fill this knowledge gap we will collect RNA for analysis of miRNA and mRNA expression at time 0, and in both cell compartments at 8, 24 and 48 hours after wounding, and subject them to uniform analysis as below. The response to acute and chronic exposure to bacterial products. Repeated and/or chronic infection is characteristic of COPD, CF, non-CF bronchiectasis and PCD. The airway epithelium occupies the interface between the body and luminal infectious agents and is strategically positioned to detect and respond to danger and to orchestrate the host response. There are several studies of the hBE cell transcriptional response to pathogens but many are in cell lines and/or in poorly differentiated cells on plastic. Furthermore, no studies have assessed adaptation to chronic infection, although this is undoubtedly an important feature of the host response. We performed mRNA arrays following acute and chronic exposure of ALI hBE cells to P. aeruginosa bacteria and detected potent induction of inflammatory cytokines and other genes but, interestingly, the cells exhibited partial tolerance (decreased responsiveness) after repeat challenges (Figure 5). We will now perform unbiased detection using deep sequencing and arrays to test the hypothesis that miRNAs are involved in the ALI hBE cell response to acute and chronic bacterial product exposure. We will study cells 4 hours after an initial Figure 5. hBE cell mRNA challenge (acute response) and 24 hours after three challenges (the new expression after P. aeruginosa. A) The acute baseline at 72 hours, representing adaptation to chronic exposure) and 4 hours response of naive cells- 372 after a fourth challenge (76 hours, representing modified responses in adapted probesets changed > 1.5 fold cells). As with the other experimental protocols, the cells will be subjected to with p<0.01. B) The response uniform downstream analysis as below. of adapted cells- only 72 Exposure to the Th2-type cytokine IL-13, which induces mucous secretory probesets met these criteria. cell hyperplasia. Increased mucus production is a feature of multiple lung diseases and is thought to contribute to airway obstruction, especially in acute asthma exacerbations and fatal asthma. Inefficient mucus clearance due to imbalanced, excessive mucous glycoprotein production and deficient ion and water transport creates mucus stasis and infection characteristic of CF and COPD. However, there are no specific therapies directed at preventing or reversing excess mucus production. To understand the role of miRNAs in mucus hyper-production, we will expose ALI hBE cells to 10 ng/ml 1L-13, a Th2-type cytokine known to induce mucous secretory cell hyperplasia. IL-13 treatment will take place every 48 hours beginning at day 25, for up to 10 days, which is a standard protocol in this field, and RNA and histology samples will be harvested at 8, 24, 48 and 240 hours. Exposure to cigarette smoke gas phase. The vast majority of COPD is due to cigarette smoking, and it is vital to understand mechanisms of smoke toxicity. There have been many mRNA array studies of broncho-alveolar lavage cells (mostly macrophages) and airway brushings (epithelium enriched) from smokers and former smokers versus non-smokers, and numerous in vitro studies of smoke exposure of cell lines on plastic to various tobacco preparations, ie. extracts and condensates versus whole smoke. A previous study examined the acute mRNA, but not miRNA, transcriptional response of ALI hBE cells to 1 hour of apical exposure to whole, mainstream smoke and 6 and 24 hours after discontinuation of the single dose {14291}. We will use the identical smoke exposure system, which we posit is the best possible model (see letter from Dr. Tarran) and will extend these studies to include both miRNAs and multiple exposures. These experiments are modeled after the P. aeruginosa experiments (above) and the time points are- 4 hours after exposure to 3 cigarettes over a 4 hour interval Figure 6. Cigarette smoke (acute exposure), 72 hours after 2 additional days of smoke exposure (new model. ALI hBE cells exposed baseline, 18 hours after the end of previous smoke exposure) and again 4 to air or smoke for 3 hours after the fourth day of smoke exposure (altered response in adapted consecutive days modestly cells). As shown in Figure 6, pilot experiments reveal modest LDH and IL-8 released IL-8 and LDH, release under this exposure regimen, suggesting it is appropriate, sustainable, indicating appropriate dosing. and not overtly toxic, which will be verified by histology. N=4 wells/group, mean + SD, *p< 0.05. Simulated ambient ozone exposures. Environmental ground level ozone represents one of the most enduring human air pollution challenges. Especially during the summer, many US municipalities routinely exceed the National Ambient Air Quality Standard of 0.075 ppm over an 8-hour period, which is associated with increased respiratory symptoms and hospitalizations. There have been many mRNA gene array studies after ozone exposure, mostly in plants and in whole animal models, with fewer studies in cells and, to our knowledge, none in ALI hBE cells or examining miRNAs (except one limited study of the arabidopsis response to ozone). Interestingly, laboratory ozone has been associated with technical problems with gene arrays. To examine the interaction of ozone exposure with host defense against S. aureus bacteria, we performed pilot studies of ozone exposure of a novel hBE cell line (UNCN3T cells, {}) capable of differentiation at an ALI (Figure 7). Based on these results, we will perform additional preliminary ozone-exposure studies, now using primary ALI hBE cells, to determine if they have similar toxicity and cytokine induction profiles. We will then expose ALI hBE cells both acutely and chronically to low (~0.2 PPM) and high dose (~0.8 PPM) ozone for 5 hours on 3 consecutive days. The chamber system has been previously well described and will be made available to us by collaborators at the USEPA facility in Chapel Hill (see letter from Dr. Jaspers). Again, samples for RNA and histology will be collected at Figure 7. Ozone model. ALI 4 hours after the completion of the first acute exposure of naive cells, then UNCN3T cells were exposed to again 18 hours after repeated exposures (the new baseline), and finally 4 0.2 and 0.8 PPM O3 for 5 hours hours after the last of 3 exposures (altered responses in adapted cells). on 3 consecutive days and then Squamous metaplasia induced by retinoic acid deficiency. Ongoing challenged with S. aureus injury from cigarette smoking or other stimuli such as chronic infection results filtrates. These doses of O3 did in squamous metaplasia of the bronchial epithelium, a characteristic not cause LDH or cytokine (not phenotypic alteration associated with disease severity, that may become shown) release on their own, but inhibited the normal response to independent of the inciting cause and whose reversibility in vivo remains the bacteria, which may be unknown. Squamous metaplasia is likely a precursor lesion to carcinoma in immunosuppressive. N=4 situ and invasive lung cancer. There are several reports of miRNA wells/group, mean + SD, *p< expression profiles in non-small cell lung cancer, including the squamous 0.05. subtype {14276}, and one study of progressive pre-cancerous lesions {14273}. To our knowledge, there is one relevant mRNA, but not miRNA, array study of an in vitro model of squamous metaplasia induced by prolonged culture on plastic {13600}. We will employ the well-known model of retinoic acid (RA) deficiency to induce squamous metaplasia in ALI hBE cells. RNA and sections for histology will be harvested at days 5, 14 and 35 in the continuous presence or absence of RA and also at 2 time points (24 and 96 hours) after re-addition of RA to squamous cultures, which reverses the phenotype. Creation of small RNA libraries. Since there is not yet a comprehensive hBE cell miRNA expression database, and array technology biases for detection of abundant miRNAs, we will perform novel identification of miRNAs (and potentially other small regulatory RNAs) using small RNA library deep sequencing. Since it is impracticable to create libraries from each of the groups used for Figure 8. Improved miRNA cloning miRNA and mRNA arrays, we will equally pool RNA from all 6 of the strategy. Mature miRNAs are size tissue donors per experimental treatment and time point/condition into fractionated by gel isolation, tailed with poly(A) polymerase and ligated to the RNA one sample. This will result in approximately 60 deep sequencing oligo R1. Reverse transcription is performed runs, anticipating 40 runs in year one and 20 in year 2. using the anchored DNA primer D2 (N refers Cloning was originally described by the Bartel, Tuschl and to any nucleotide, and V refers to any Ambros labs {14299} {14300} {14301}. Their well-established method nucleotide except T; this anchors the primer to the beginning of the tail). PCR is based on ligation of RNA/DNA oligonucleotides to each end of the amplification is performed with primers D1 mature miRNA, followed by PCR amplification and sequencing. We and D3. Sequencing tags are added by a have modified this protocol, eliminating one ligation step (Figure 8). second PCR step, and the population is Instead of ligating an oligonucleotide to the 3' end of the miRNA, we sequenced using a Solexa GAII. use poly(A) polymerase to add a poly(A) tail, providing a primer binding sequence for reverse transcription. This eliminates the cloning background caused by direct 5' and 3' oligonucleotide ligation and reduces the gel purification from three steps to one. Since gel purification is a major source of yield loss, our modification improves the discovery rate of rare miRNA species. Total RNA (up to 10 ug) will be size fractionated on a 15% polyacrylamide/Urea/TBE gel. Isolated RNA will be tailed with poly(A) polymerase, followed by phenol extraction. The tailed small RNAs will be ligated to the 5' oligonucleotide R1. This RNA oligonucleotide lacks a 5' phosphate, therefore cannot concatamerize or circularize, but can only be ligated to an existing 5' phosphorylated small RNA. This product is phenol extracted and reverse transcribed with the anchored oligo d(T) primer D2. The DNA product will be amplified with the primers D1 and D3. Cycle number will be empirically determined, using the minimum to produce a SYBR-gold visible product. The PCR product will be re-amplified to add sequencing tags for Solexa sequencing. Deep sequencing. We will employ the UNC Lineberger Comprehensive Cancer Center (LCCC) High Throughput Genomics Sequencing Facility directed by Dr. Poitr Mieczkowski. The protocols for PCR product sample preparation, submission, prioritization, analysis and data transfer are dictated by facility policy and practice and are routine. Sequencing is via the Genome Analyzer II (GAII)-Illumina (Solexa) platform, enabling >50 million 35 base pair reads per flowcell, 8 lanes per flowcell, 5-7 mil reads per lane and thus generating >1.5 GB of data per single read flowcell, Sequencing data handling. A suite of programs called the “Illumina Genome Analysis Pipeline” is used for data analysis. The data download includes five files: 1) “Raw” sequence file; 2) “ELAND” sequence alignment results file; 3) “Filtered” sequence file; 4) Sequence “Export” file; and 5) “Sorted” sequence file. These files enable compiling and analysis of the sequencing data and bioinformatics analysis including, clustering of sequence similarity groups, construction of clone count tables, comparison of miRNAs between different samples, and hierarchical clustering of samples. Aim 1B. Map new miRNAs to the genome to validate stem-loop potential and vertebrate conservation, and verify expression of bona-fide ~21 nucleotide small RNAs by Northern blotting. Sequencing data analysis. Output from the “Export” file will be mapped to the UCSC genome assembly as a custom track. Reads that correspond to known miRNAs, mRNAs exons, large and small noncoding RNAs (tRNAs, rRNAs, snRNAs, etc) will be removed. The read sequence plus 80 nucleotide flanking genomic sequence (each side separately) will be tested for hairpin folding potential (mFOLD). Candidate miRNAs that pass these filters will be ranked by sequencing clone count and vertebrate conservation. Most miRNAs exhibit a distinctive conservation pattern with high conservation in the mature strand, low conservation in the loop, and moderately high conservation in the second strand of stem (star strand). It is formally possible that we identify novel miRNAs that are not conserved outside of primates, therefore, nonconserved candidates with >5 clone count will be validated. Northern blotting. We will verify that novel sequences are derived from a ~21 nucleotide small RNA, rather than a fragment of a larger RNA species. Total RNA will be analyzed by 12% acrylamide/Urea/TBE northern blot, using an antisense oligonucleotide probe based on the candidate mature miRNA. If necessary, large RNAs will be depleted by precipitation in 12.5% PEG 5000/ 12.5 mM NaCl. Aim 1C. Determine expression of known, annotated miRNAs and mRNAs in the same seven conditions above using Agilent microarrays, facilitating analysis of miRNA regulation of hBE cell gene expression. Methods: Microarrays. The Hayes and Hammond labs have extensive experience in the generation and analysis of microarray data {14223} {14252}, and the UNC Genomics Core is highly qualified to perform the microarray experiments described in this application. This Core is 1 of 3 sites chosen for the pilot phase of The Cancer Genome Atlas (TCGA) for which Dr. Hayes is a co-Investigator {14223}, and also serves as the profiling center for the Cancer and Leukemia Group B project. Figure 9 documents recent productivity, showing classification of malignant brain tumors by expression profiling, for which there is also paired miRNA data (see below). In this example, we Figure 9. Gene expression data. Analysis of 173 glioblastoma samples from the Cancer Genome Atlas project performed at UNC identifies four gene expression subtypes. Samples were ordered based on subtype predictions. document the power of gene expression analysis in our hands to detect molecular subtypes (and the associated genes), which in large part reflect abnormally differentiated cancer cells. In the current proposal, we will use a similar approach to capture miRNA and mRNA expression patterns associated with normal hBE cell differentiation and the response to injury. miRNA profiling. We will use 3rd generation Agilent Human v12 miRNA arrays containing 866 human and 89 viral miRNAs that are regularly updated according to the Sanger miRBase database. Briefly, 400 ng of total RNA are labeled per the manufacturer protocol and samples are processed according to standard protocols in the Genomics Core. Data is stored in the UNC Microarray Database and we favor Quantile normalization for miRNA data. Using this platform and associated protocols, the Hayes lab has analyzed ~1000 arrays In support of the TCGA, and we are the only site in the pilot phase profiling miRNAs. mRNA profiling. Gene expression profiling employs 180,000 feature, custom-designed Agilent long oligonucleotide arrays that contain 25,000 unique human protein-coding gene probes, a Figure 10. miRNA expression data. A) second unique probe for almost all of these protein-coding genes miRNAs define four glioblastoma subtypes. B) miRNA subtypes correlate with subtypes as an independent measure of gene expression, and hundreds derived by gene expression profiling in the of control spots for normalization, mRNA quality and array same sample set. quality control. Probes are spotted in triplicate occupying ~150,000 features. The remaining ~30,000 features will be left open initially, but will be used to include new transcript and/or splicing content as it is discovered by us and others, a primary feature for selection of the Agilent technology over competing array products. The basic protocol for the gene expression arrays has been described in detail elsewhere {14223} and has been performed 10’s of thousands of times in the UNC Genomics Core. Briefly, 1 ug of total RNA is labeled using the Agilent low input RNA linear amplification kit. Microarray hybridizations are carried out using 2 ug of Cy3-labeled common reference sample (Stratagene’s Human Universal Reference RNA) and 2 ug of Cy5labeled experimental sample. Hybridizations are performed using the Agilent hybridization kit and a Figure 11. TMEFF1 gene expression is best predicted Robbins Scientific hybridization oven. The arrays are by hsa-mir-34a versus other genomic events. The 5 incubated overnight, washed and then scanned using an diagonal plots represent histograms of the indicated Agilent Microarray scanner and Agilent Feature data types from ~200 brain tumors (TCGA data). From Extraction and Analysis software. All raw data are then top left to bottom right: TMEFF1 gene expression, loaded into the UNC Microarray Database where a TMEFF1 gene copy number, TMEFF1 locus Lowess normalization procedure is performed. methylation, hsa-mir-34a expression (targets TMEFF1), and hsa-Mir-34a gene copy number. The lower offmiRNA target gene analysis. We document the power diagonal plots graphically illustrate the relationship of pairing microRNA profiling with mRNA gene between the parameters and the corresponding upper expression profiling in a cancer example. First, we off-diagonal plots give the Pearson correlation demonstrate that clustering algorithms of miRNA array coefficient. TMEFF1 gene expression in glioblastoma data define variants (subtypes) of tumors analogous to varies by 4 orders of magnitude (row 1, column 1). The gene expression arrays (compare Figure 10 to Figure influence of TMEFF1 copy number on gene expression is modest (correlation=0.31, row 2, column 1 and row 1, 9), and that these subtypes overlap with tumor subtypes column 2), while there is robust linear anti-correlation defined by gene expression arrays. This is strong between TMEFF1 and hsa-mir-34a expression evidence that shared programs of gene and miRNA (correlation=0.53, row 4, column 1 and row 1, column expression integrate to define fundamental 4). characteristics of the tumors. Since tumor formation frequently includes abnormal cell differentiation, the paired approach will be similarly useful to study normally differentiating hBE cells and the response to injury. In Figure 11, we extend the integrated analysis of gene expression data and miRNA data, focusing on microRNA-gene target pairs. In this cancer example, we examine a gene known to be important in brain differentiation, and document a weak impact of common cancer-specific genomic alterations such as gene loss and gene methylation. In contrast, using paired miRNA and gene expression data, we detect a strong association between miR-has- 34a and its predicted gene target TMEFF1. In the current proposal, we will similarly measure, in a genome wide manner, miRNA-gene target interactions, determining experimentally which of the computationally predicted interactions are measurable in our model systems. Aim 1D. Perform confirmatory qRT-PCR and in situ hybridization localization for select highly regulated miRNA’s identified above. Methods: qRT-PCR. It is conventional to verify accuracy of array data using independent methods. Accordingly, within the scope of this proposal, we will quantitate expression of approximately 5 miRNAs identified on arrays per each of the 7 projects described in Aim 1A. We may assay more than 5 miRNAs in a particular experiment, for example if we discover more than 5 highly regulated, unique or otherwise scientifically provocative miRNAs in a specific project. qRT-PCR may also be used to further characterize novel miRNAs discovered by sequencing. qRT-PCR will be performed exactly as per manufacturers’ instructions (miRNA reverse transcription kit and TaqMan miRNA assay kit, Applied Biosciences). PCR is performed on 96 or 384 well format thermocyclers in the CF Center or in the LCCC, respectively, and analysis is by conventional comparative Ct to 5S rRNA and/or miRNAs reported to not vary significantly across a wide group of specimens {14298}. All assays are performed in duplicate using non-pooled RNA’s from 3 replicate wells from 6 different tissue donors. Results are assessed with descriptive statistics and t-test or ANOVA with post-hoc analysis, or using non-parametric tests when required. In situ hybridization. Since miRNA expression may occur in a cell-type-specific spatial and temporal pattern, we will perform in situ hybridization in the cultures at specific time points or conditions, prioritized based on the sequencing and array results. Since this method is more consumptive than qRT-PCR, we anticipate being able to study ~3 miRNAs per experimental group and will not necessarily examine all subgroups within the project. Briefly, in situ hybridization of formalin paraffin and whole mount culture preparations will be undertaken using specific and control probes (LNA probes, Exiqon) and colorimetric detection as previously described {14253}. Recent studies indicate that frozen sections specifically cross-linked to preserve small RNAs combined with fluorescence detection may be much more sensitive {14237} and frozen blocks for sectioning will be prepared and retained for this purpose. Analysis of staining patterns will be primarily qualitative and complementary to the qRT-PCR data above although antibody immuno-staining for cell-typespecific markers (keratin 5 &14 for basal cells, acetylated tubulin for ciliated cells and MUC5AC or 6 for mucous secretory cells) combined with in situ hybridization will enable correlation of miRNA expression with cell type. Aim 1. Data Analysis/Limitations/Alternatives/Future Directions. A major, novel deliverable from Aim 1 is a comprehensive hBE cell miRNA expression database based on unbiased discovery through deep sequencing as well as arrays and qRT/PCR for known, annotated miRNAs. Studies of this type, by their nature, are hypothesis generating, as opposed to the paradigm of testing specific hypotheses. Nevertheless, the data will be comprehensive and novel, using state-of-the-art cultures, models and analytic methods. It will enable selection of a subset of highly regulated miRNAs for functional analyses as proposed in Aim 2, which is a more mechanistic approach. As described above, we will develop a standardized analytical pipeline for ALI hBE cells using library construction and deep sequencing. We think this is extremely valuable since it is widely appreciated that many miRNAs are highly cell type and differentiation stage specific. For example, the miRNA lsy-6 is expressed in a single neuron in C. elegans. We believe highly restricted cell-specific miRNAs remain to be discovered, and to our knowledge unbiased discovery HAS NOT yet been performed in hBE cells. Drs. Hammond and Hayes have prior experience in unbiased exploration of miRNA and mRNA expression via sequencing. As well as adding to the miRNA repertoire and providing a basis for further quantitation of highly regulated miRNAs, it is conceivable that we may find specific miRNA markers of hBE cells, cell subtypes and functional stages, which will be an important contribution as such markers could be used to direct expression, ablate or purify cells as was recently done for the pri-miR-375 promoter in endocrine pancreas {14254}, but this would be a future direction. We will initially evaluate the time course of normal hBE cell growth and differentiation in vitro and will then exploit the data generating system to efficiently probe highly relevant injury/repair conditions. We anticipate no difficulties regarding the ability to produce the requisite hBE cell cultures, to employ the injury models or to purify adequate RNA and histology samples since the Randell lab has extensive experience and resources for hBE cell ALI culture- the cells already exist in our collection and the models have already been accomplished (wounding, bacteria, smoke, ozone, RA-deficiency) or are quite simple (IL-13). The specific steps for data generation and analyses, including sample size and an outline of technical and statistical considerations for specific assays, have been presented above. These methods have been extensively used and published by members of our group. Many, such as the mechanical generation of libraries, sequencing and, especially, arrays and qRT-PCR are now quite routine. Dr. Hayes has been involved in massive array and sequencing efforts related to cancer for many years. Dr. Hammond has been in the miRNA field for over 10 years and has abundant experience in the analysis of primary sequence and array data and its translation via mapping and, processing, expression and functional assays. Based on the preliminary studies, and the track record/complementary expertise of the team, we are confident that the experiments can be mostly accomplished as planned. Realistically, the plan is highly ambitious in terms of the number of samples to be generated and analyzed. The experiments all follow the same paradigm and should be easier to accomplish as we gain specific experience with the systems employed. However, if experiments must be dropped by necessity we will prioritize experiments in their order, ie. by eliminating the last planned experiment (RA-deficiency) first. In any event, the studies will certainly provide abundant, state-of-the-art profiling of the miRNA repertoire of hBE cells, revealing highly regulated and potentially novel miRNAs almost certainly involved in the dynamic regulation of their growth and differentiation and response to injury. This data will be useful from the perspective of both enhancing our basic understanding of regulation of hBE cell phenotype in health and disease and to prioritize subsequent functional testing as in Aim 2 of this proposal. We expect that this data will be a beginning, a sound foundation, for the design of many experiments and analyses, by others and us, to more fully understand the role of miRNAs in airway biology. Correlative miRNA and mRNA expression analysis will enable exploration of gene expression regulatory networks in these important cells. We expect to reveal many specific relationships between miRNAs and gene expression of their predicted downstream targets, as illustrated in the cancer example above. A commonly employed technique for verifying such relationships is to express a marker gene (luciferase, GFP) fused to the 3’UTR of the putative target gene, sometimes in the cells of interest, but usually in simple conventional culture models. We contemplated using this reporter approach in well-differentiated ALI hBE cells, but like the functional studies in Aim 2 below, it would also require the development and use of lentiviral vectors and infection of the cells while poorly differentiated. Alternatively, adenoviral vectors, which can transduce mature hBE cells, could be used for reporter gene studies. These experiments are possible, but are resource and time (especially adenovirus) consumptive. Based on the anticipated lower payoff of the reporter approach versus manipulation of miRNAs as proof-of-concept for novel therapies (Aim 2), we have prioritized the latter. However, such reporter vectors will be created to screen miRNA mimic and knockdown strategies (Aim 2) and these may be employed in ALI hBE as resources allow. An uncertainty of these studies is that they employ an in vitro model, which may not precisely replicate the in vivo expression pattern. This is certainly true, but the model does allow for more precise cell-typespecific analysis than in vivo studies where cell populations are heterogeneous and variable. Since the model recapitulates many of the functional properties of the in vivo epithelium, it allows for testing of specific hypotheses regarding miRNA participation in normal growth and differentiation, injury and repair. The experiments will clearly complement ongoing in vivo studies by others. A future direction is correlative in vivo studies in humans (measurements in excised human cells and tissues, in situ hybridization of normal and diseased human lung) and also the creation of genetically manipulated mice, but those are beyond the scope and time frame of this already ambitious 2-year proposal and are more realistically considered a future direction. Specific Aim 2) Determine functional consequences of manipulating expression of candidate miRNAs in hBE cells. Rationale: Novel miRNA discovery by sequencing, high throughput screening of miRNA arrays, qRT-pCR and robust northern blots as described above will allow for rigorous identification and quantitation of hBE cell miRNAs, but they must be complemented by over-expression and inhibition in order to determine functional effects. These studies are challenging in ALI hBE cells but are key for establishing whether modulating miRNAs can change phenotype for potentially beneficial therapeutic effects. Aim 2A. 2A. Use an inducible, titratable viral vector system to express miRNA mimics and inhibitors to determine functional consequences of manipulating specific miRNAs in hBE cells. Methods: Manipulating miRNAs in ALI hBE cells. miRNA mimics or inhibitors for gain-of-function and loss-of-function experiments, respectively, enables study of the biological role of specific miRNAs and there are numerous commercial products for this purpose. Lipid transfection or electroporation are typically used to introduce oligonucleotides, modified oligonucleotides or plasmids into tissue culture cells, although some have been applied in vivo via systemic or local administration in animals, with variable success. Well-differentiated ALI hBE cells are notoriously difficult to transfect in vivo or in vitro, precluding typical plasmid and oligonucleotide approaches for genetic manipulation commonly used in routine tissue culture cells. We initially developed lentiviral vectors with constitutive expression to infect hBE cells while poorly differentiated on plastic, followed by Figure 12. Inducible expression in ALI hBE selection and passage to porous supports where they can cells. Primary hBE cells on plastic were differentiate. A potential downside of this approach is that infected with pSLIK lentiviral vectors, selected constitutive expression of inhibitors or genes that alter function, with puromycin, then subcultured at an ALI and treated with doxycycline as indicated. for example that inhibit cell growth would be self-defeating. Thus, The cells were harvested for mRNA (A), and in more recent experiments we have used an inducible system, protein (B), and were also studied in Ussing pSLIK {14302}. Initial studies (Figure 12) examined the epithelial chambers (C) and by confocal microscope sodium channel (ENaC) with shRNA and illustrate knockdown at imaging of airways surface liquid (ASL, D). the RNA, protein and functional (decreased amiloride-sensitive Significant knockdown and corresponding ENaC electrical current and increased airway surface liquid due to changes in function are apparent (see decreased Na absorption). This model enables well-controlled adjacent text). Ctl GFP = pSLIK vector containing GFP only as a control. shENaC functional assessment of the ability of miRNA mimic or inhibitor = pSLIK vector containing GFP and shRNA to expression to modify hBE cell phenotype. ENaC. Dox = doxycycline, which induces miRNA mimic and inhibitor expression in ALI hBE cells. The expression of GFP and shRNA. method for manipulating gene expression in ALI hBE cells involves design and cloning of experimental and control vectors and virus production as well as the downstream characterizations, which are collectively fairly consumptive and will practically limit the number of genes that we can express or knockdown. We anticipate accomplishing at least one example from each of the seven experimental protocols, targeting a highly regulated and/or novel miRNA. This will be selected not only based on expression but also looking for scientifically provocative candidates such as miRNAs highly correlated with cell-type-specific panels of gene expression- for example secretory proteins characteristic of goblet cells or the ciliated cell machinery detected in the mRNA arrays. This approach will be combined with manually scanning the still manageable, but rapidly evolving miRNA biological function database. Ideally, we will target two complementary genes per experimental treatment for both expression and knockdown- one that increases and one that decreases. There has been excellent recent progress, particularly from the Naldini lab, illustrating methods to design and assess the function of introduced natural and chimeric/artificial miRNAs {14823}. They showed that lentiviral vetors enabled exogenous/artificial miRNAs to reach the concentration and activity typical of highly expressed natural miRNAs without perturbing endogenous miRNA maturation or regulation. Similarly to the protocols described in that report, we will initially screen for functional effects of candidate miRNAs in a simple cell line culture system on plastic, using co-transfection of the putative mimic and constructs containing 4 copies of its perfect target cloned into the 3’ UTR of a reporter gene (luciferase and/or GFP). Promising candidates will then be sub-cloned into the pSLIK virus system and viral particles will be produced by triple transfection of 3T3 producer cell lines using routine procedures. hBE cells will be infected and selected as described above, and half of the cells will be exposed to doxycycline at specific time points as dictated by the experimental goals. The loss of function, miRNA knockdown experiments involve a similar approach but now over-expressing miRNA target sequences {14282}, which act as “decoys”. The ideal experiment requires 8 tandem imperfect, bulged matches, which appear to be more effective due to slower RISC complex processing. Furthermore, strong promoters and high vector dosing are needed to result in multiple copy integration and strong expression. The inhibitors will also be tested for knockdown efficiency in simple cell line on plastic model systems to select the optimal candidates for expression in ALI hBE cells as above. In these experiments, we will measure mRNA levels of the endogenous miRNA targets in hBE cells, which are expected to rise. Phenotypic characterization of ALI hBE cells. The Randell laboratory has many years of experience evaluating airway epithelial cell structure and function. A primary method for assessing effects of manipulation of miRNA expression and knockdown will be histology. The simple evaluation of H&E and Alcian blue-PAS stained, routine formalin-paraffin sections is highly informative and cost effective. Semi-quantitative morphological scoring of a series of specific characteristics (epithelial thickness, nuclear density, percent ciliated surface, amount of stored mucosubstances) by multiple viewers blinded to sample identity is concordant with the basic morphometric principle of “do more less well”, but fully quantitative morphometry, will be used in the event of more subtle effects. Additional sections of the same histology blocks will be immunostained with routine airway epithelial cell-type-specific markers (keratin 5 &14 for basal cells, acetylated tubulin for ciliated cells and MUC5AC or MUC5B for mucous secretory cells). Gene (mRNA) and protein expression of the putative target genes and proteins will be assessed using qRT-PCR and western blot/immunostaining, respectively. The function of the putative (and/or verified) target genes as determined in the literature and/or by gene annotation will be used to decide specific, contingent downstream assessments. For example we would logically test the response to inflammatory stimuli for target genes implicated in inflammatory processes. Similarly, targets regulating ion transport would be evaluated as described in Figure 12. We will assess hBE cell proliferation and apoptosis when target genes are involved in cell cycle regulation and/or if hyper- or hypoplasia was seen morphologically. We do not, a priori, plan on performing mRNA (or miRNA) arrays after miRNA modulation, but that is certainly possible, given a sufficiently compelling phenotype. Aim2 Data Analysis/Limitations/Alternatives/Future Directions: The studies in Aim 2 logically extend from Aim 1 and are directed at understanding if ALI hBE cells in vitro can be used as a platform for functionally testing the effects of manipulating highly regulated and/or novel miRNAs. An immediate benefit will be a greater basic understanding of the role of miRNAs in the molecular regulation of airway epithelial cell structure and function, which is still poorly understood. Another ostensible goal is proof-of-concept testing that altering hBE cell phenotype via manipulating miRNA status would be of therapeutic benefit. For example, if we find certain miRNAs to be specifically associated with mucous secretory cell hyperplasia after IL-13 treatment, we will determine whether expression of mimics induced mucous cells and whether inhibition of these miRNAs reduced mucous cells. Results indicating miRNA regulation of mucous secretory cell proliferation and differentiation would suggest a new therapeutic approach to mucus hyper-production. A similar process would be relevant to discovery of anti-inflammatory agents after P. aeruginosa treatment, cyto-protective therapies after smoking or ozone, reversal of squamous metaplasia, and so forth. There are likely to be many examples of highly regulated miRNAs in Aim 1 and we will only be able to test a small subset of the most compelling candidates, but the development of the testing system in the most relevant differentiated cells will clearly facilitate more extensive future studies. There is a certain level of uncertainty in the Aim 2 experiments since regulated and inducible lentiviral manipulation of miRNAs is a new and evolving technology. There are concerns about efficiency, although these appear surmountable based on the literature {14283} {14282}. Similar to siRNA and shRNA, there are concerns regarding off target effects, which may be exacerbated in the case of miRNA due to general effects on related family members or the processing machinery. We will control for these as well as possible using scrambled and irrelevant controls. There are concerns about gene silencing of integrated viral sequences, which we have observed in earlier generations of retroviral vectors (data not shown). We may have to accept reduced sensitivity to see effects if less than 100% of then cells are expressing at the time of analysis. If silencing is profound (ie. >50%), which we do not expect with lenti- versus retroviruses, we could use adenovirus in an up to 7-day expression timeframe. However adenovirus production is more cumbersome and expensive and will limit the number of miRNAs we could study. Alternatively, we could test other lentiviral backbones and promoters or adeno-associated virus to find minimally silencing vector strategies. Finally, since the pSLIK system coordinately expresses a marker gene (GFP) we can employ single cell assays within cultures containing a mixed population of expressing and silenced well-differentiated ALI hBE cells. It will ultimately be important to test the effects of miRNA mimics and antagonists in vivo in genetically manipulated mice, as has been done previously using a simple transgenic lung epithelial over-expression approach {14253}. However, the two-year time frame of the RC1 format does not reasonably leave enough time for the triple transgenic or ES cell knock-in strategies that would allow inducible, airway epithelial-only expression that would prevent the neonatal lethal respiratory phenotypes so common in constitutive lung expression transgenics {14253}. This would be a clear goal for future studies based on positive results from Aim 2. Finally, the ultimate obstacle to miRNA manipulation as a human therapy for airway diseases, perhaps via vector inhalation or instillation, is the long-standing problem of poor transduction of the human airway epithelium in vivo. This challenge is beyond the scope of the current proposal, but novel vectors and/or nanoparticles may be useful in the future. Alternatively, results showing positive phenotypic manipulation of disease-related hBE cell phenotypes will likely spur the development small molecule therapies directed at manipulating specific miRNAs. 5.4 Timeline and Milestones: We are poised to immediately make progress on the studies in this proposal, which is consistent with the goals of the ARRA. We already have RNA in hand for the normal ALI hBE cell time course as well as the RNA from the P. aeruginosa exposure protocol. These will be immediately subjected to size fractionation, library construction, deep sequencing and arrays as described above. Simultaneously, new ALI hBE cells cultures will be sequentially initiated for each of the other exposure/treatment regimens, staggering those more highly demanding (smoke and ozone exposure) to facilitate the workflow. The initial sequencing and array technical procedures and data analysis algorithms will serve as test cases for evaluation of protocols and routines and whether they require modification in subsequent experiments. We anticipate accomplishing 2/3 of the hBE cell cultures, sequencing and arrays in year 1. Simultaneously in year 1, we will begin qRT-PCR and northern blot confirmation of miRNA expression and in situ hybridization for cell localization in specimens from the initial experiments. Also in year 1, we anticipate creating lentiviral vectors for miRNA manipulation and using them to begin to assess functional changes due to modulation of miRNA expression in ALI hBE cells. Additional time in year 1 will be spent on preparation of initil publications. In the beginning of year 2 we will perform the final 1/3 of the cell cultures, sequencing and arrays, and the rest of year 2 will be dedicated to completing the downstream analyses and preparation of publications.