Molecular Profiling of Single Sca-1+/CD34+,− Cells—The Putative Murine Lung Stem Cells

Murine bronchioalveolar stem cells play a key role in pulmonary epithelial maintenance and repair but their molecular profile is poorly described so far. In this study, we used antibodies directed against Sca-1 and CD34, two markers originally ascribed to pulmonary cells harboring regenerative potential, to isolate single putative stem cells from murine lung tissue. The mean detection rate of positive cells was 8 per 106 lung cells. We then isolated and globally amplified the mRNA of positive cells to analyze gene expression in single cells. The resulting amplicons were then used for molecular profiling by transcript specific polymerase chain reaction (PCR) and global gene expression analysis using microarrays. Single marker-positive cells displayed a striking heterogeneity for the expression of epithelial and mesenchymal transcripts on the single cell level. Nevertheless, they could be subdivided into two cell populations: Sca-1+/CD34 − and Sca-1+/CD34+ cells. In these subpopulations, transcripts of the epithelial marker Epcam (CD326) were exclusively detected in Sca-1+/CD34 − cells (p = 0.03), whereas mRNA of the mesenchymal marker Pdgfrα (CD140a) was detected in both subpopulations and more frequently in Sca-1+/CD34+ cells (p = 0.04). FACS analysis confirmed the existence of a Pdgfrα positive subpopulation within Epcam+/Sca-1+/CD34− epithelial cells. Gene expression analysis by microarray hybridization identified transcripts differentially expressed between the two cell types as well as between epithelial reference cells and Sca-1+/CD34+ single cells, and selected transcripts were validated by quantitative PCR. Our results suggest a more mesenchymal commitment of Sca-1+/CD34+ cells and a more epithelial commitment of Sca-1+/CD34 − cells. In summary, the study shows that single cell analysis enables the identification of novel molecular markers in yet poorly characterized populations of rare cells. Our results could further improve our understanding of Sca-1+/CD34+,− cells in the biology of the murine lung.


Introduction
The murine lung contains at least 40 morphologically distinct cell types of mesodermal or endodermal origin [1]. It can be subdivided anatomically in three different regions: large airways, bronchioles, and alveoli. Each region is composed of different cell types whose homeostasis is guaranteed by regionally specific progenitor cells [2,3]. Importantly, Sca-1 + cells have been ascribed decisive functions regarding the maintenance and repair of the airways [2]. In injury models, Sca-1 + cells demonstrated resistance to damage and clonal expansion leading to a restoration of previously experimentally depleted epithelial structures [2]. During the last years, considerable efforts have been undertaken to further characterize a specific subgroup of Sca-1 + cells. These cells were named bronchioalveolar stem cells (BASCs) based on their location at the bronchioalveolar duct junction [4] where they proved to play a key role in distal airway repair [4][5][6][7].
Considering the heterogeneity of Sca-1 + murine lung cells, as it could be demonstrated for the group of BASCs, the isolation and subsequent gene expression analysis of single cells may provide an interesting tool for the identification of novel molecular markers in poorly described rare cells [11,12]. Eventually, this approach could allow better distinction between the different subtypes of Sca-1 + cells and lead to the discovery of novel molecular markers facilitating a better detection and functional evaluation. In this study, we isolated viable Sca-1 + and CD34 + cells on the basis of their protein expression. Following global amplification of single cell mRNA, gene expression profiling was performed to analyze the cell populations at the single cell level. The analysis included markers that have previously been ascribed to BASCs and intended to identify markers that thus far have not been described in Sca-1 + murine lung cells.

Animals and Tissue Preparation
No animal underwent animal experimentation as defined by the German law (1 7 TierschG). All animals were euthanized and killed before organs were taken. According to German law (1 6 Abs 1 TierschG) there is no requirement for an Ethics vote nor a notification of the local Government for dissection and organ use after the death of an animal. Mice were kept according to the guidelines of the Felasa (Federation for Laboratory Animal Science Associations) and the guideline 2010/63/EU and the ETS123 (APPENDIX A to the European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes).
Experiments were conducted on 26 Balb/c mice (10 female, 16 male) between 8 and 12 weeks of age. Mice were euthanized with 100% carbon dioxide, and subsequently the thoracic cavity was opened and the great venous vessels were clamped. The right ventricle was incised and a knop canula introduced to perfuse the lungs with 20-30 ml of body-warm saline. After exsanguination, lungs were dissected and collected in Hanks Balanced Salt Solution (HBSS, pH 7.4, Sigma Aldrich).

Tissue Digestion and Cell Separation
The lungs were finely minced and the tissue was enzymatically digested according to an established protocol [13], with minor modifications. Briefly, each finely minced lung preparation was incubated with 5.4 U/ml collagenase (Roche), 0.03 U/ml dispase (GE Healthcare-Invitrogen) and 2.5 mM CaCl 2 for 45 minutes at 37uC. The suspension was then filtered through nylon membranes (BD; 100 mm, 40 mm). Enzymes were inactivated in HBSS/0.2 M EDTA (pH 7.4), cells were centrifuged (1500 rpm, 4uC, 5 min) and resuspended in PBS (pH 7.4). A Percoll gradient (70%) density centrifugation was performed (2050 rpm, 4uC, 20 min) to separate the mononuclear cell fraction. The cell interphase was recovered and washed in PBS. Following a centrifugation step (1500 rpm, 4uC, 10 min) cells were resuspended in PBS (pH 7.4). Finally, viable cells were identified by Trypan blue assay and counted in a Neubauer chamber.
For FACS analysis, the lung preparations of 5 mice were finely minced. The tissue was enzymatically digested with collagenase (Sigma; 1.5 mg/ml) for 30 minutes at 37uC. The enzymes were inactivated with PBS/BSA (Sigma; 2.5 g/100 ml). The suspension was filtered through a cell strainer (Greiner Bio-One; 40 mm), centrifuged (300 g, 10 min, 4uC) and resuspended in 5 ml red blood cell lysis solution (16) (Miltenyi; order no. 130-094-183) for 15 min at room temperature. Thereafter, cells were washed with PBS, centrifuged und resuspended in PBS to be counted in a Neubauer chamber.
Since quantitative gene expression analysis of the isolated Sca-1 + and CD34 + cells was an important goal of our study, we had to acquire appropriate reference cells for comparison, i.e. pulmonary cells lacking Sca-1 and CD34 expression and not originating from hematopoietic or endothelial cell lines. Therefore, cells of 6 enzymatically digested lungs were double-stained with CD31 and CD45 antibodies in order to collect single cell samples. The same isolation procedure as described above was applied for CD31 2/ CD45 2 reference cells. To exclude cells expressing Sca-1 and CD34, specific PCRs on the corresponding amplified transcripts of those genes were performed.

Whole Transcriptome Amplification (WTA) of Single Cells
The transcriptome of each obtained cell was extracted and amplified according to an established protocol [12] with modifications [11]: One microlitre protease/lysis buffer mix (1:20dilution; Active Motif, Rixensart, Belgium) and 1 ml biotinylated peptide nucleic acids (PNAs; Midi-Kit, Active Motif, dissolved in 400 ml water) were added to each tube for proteolytic digestion. Digestion was performed at 45uC for 10 min, followed by inactivation for 1 min at 70uC and 15 min at 22uC for PNA annealing to the mRNA. mRNA was captured by streptavidinecoated magnetic particles (Active Motif) during 45 min rotation at room temperature. After addition of 10 ml wash buffer 1 (50 mM Tris-HCl (pH 8.3), 75 mM KCl, 3 mM MgCl 2 , 10 mM DTT and 0.25% Igepal), tubes were placed in a magnetic rack. The supernatants containing the genomic DNA were removed and the beads washed with 20 ml wash buffer 2 (50 mM Tris-HCl (pH 8.3), 75 mM KCl, 3 mM MgCl 2 , 10 mM DTT and 0.5% Tween). The supernatants were removed and the beads washed once again with 20 ml wash buffer 1. The mRNA was reverse transcribed using a mix comprising 0.5 mM dNTPs, 30 mM allowed to anneal at room temperature for 10-15 min before the enzyme was added. Reverse transcription was conducted for 45 min at 44uC in an oven with constant steering. Following reverse transcription, the magnetic cDNA-mRNA-hybrids were washed in 20 ml tailing wash buffer (50 mM KH 2 PO 4 (pH 7), 1 mM DTT, 0.25% Igepal), resuspended in 10 ml tailing buffer (10 mM KH 2 PO 4 (pH 7), 4 mM MgCl 2 , 0.1 mM DTT, 200 mM dGTP) and then coated with 40 ml PCR oil. Hybrids underwent denaturation at 94uC for 4 min. The tailing reaction was performed at 37uC for 60 min after addition of 10 U terminal deoxynucleotide transferase (TdT; Amersham, Freiburg). TdT was inactivated at 70uC for 5 min. Next, a hot-start PCR was performed. Therefore, 35 ml of PCR mix 1 containing 4 ml buffer 1 (Expand long template, Roche) and 3% deionized formamide was added to each sample., Next, the samples were heat up to 78uC followed by addition of 5.5 ml of the PCR mix 2 (350 mM dNTPs, 1.2 mM CP2 primer (TCAGAATTCATG[C] n = 15 ) and 5 U Pol Mix (Expand long template)). Forty cycles were run in a MJ research PCR machine: 20 cycles of 15 sec at 94uC, 30 sec at 65uC, 2 min at 68uC and 20 cycles with an elongation of the extension time of 10 sec and a final elongation step of 7 min at 68uC.

Analytical PCR on Specific Transcripts
For quality control of whole transcriptome amplification products, we tested for amplicons of ubiquitously expressed genes Actb (ß-actin) and Gapdh (Glyceraldehyde 3-phosphate dehydrogenase) by PCR. Only cells with at least one positive result were considered for further analysis. For initial molecular characterization of isolated cells, PCR on transcripts of Sca-1, CD34, CD45 and CD31 were performed.
In order to differentiate between a more epithelial or mesenchymal phenotype of isolated cells, we conducted further PCRs specific for epithelial markers Epcam (Epithelial cell adhesion molecule), Itga (Integrin alpha-6) and Sftpc (Surfactant protein C) and mesenchymal markers CD90 (Thy-1) and Pdgfra (platelet derived growth factor receptor alpha, CD140a), as suggested by McQualter et al. [9]. Specificity of all primers was confirmed by restriction digestion, sequences are depicted in Table S1.

Array Hybridization and Data Analysis
Probes of the 29 selected cells were hybridized on Mouse Genome OpArrays (Eurofins MWG Operon; cat # OPMMV4-05). The arrays contain probes for 16,928 genes and have previously been used for hybridization of single cell WTA products [11]. The amplified single cell cDNA was labeled with 0.05 mM digoxygenin-dUTP (Roche) and 0.05 mM aminodigoxygenin-dCTP (PerkinElmer, Rodgau-Jügesheim) in the presence of 3% formamide, 2.4 mM CP2-BGL primer (TCAGAATT-CATGCCGCCCCCCCGGCCC) and dNTPs (0.35 mM dATP and dGTP, 0.3 mM dTTP and dCTP). Reference cDNA was labeled with biotin-dUTP (Roche) and biotin-dCTP (Invitrogen). Primer sequences were then separated from the cDNA sequences in a subsequent digestion step with 30 U of BglI (Fermentas, St. Leon-Rot), and then purified (QIAquick PCR Purification Kit, QIAGEN, Hilden). Test and reference cDNA were co-precipitated with 0.8 ml polyacrylamide carrier, 0.163 M sodium acetate and 2.56 ethanol (100%). Arrays were pre-hybridized with 56 SSC +0.1% SDS +0.1% BSA at 42uC and hybridized in an Arraybooster hybstation (Implen, Munich) at 42uC overnight. The slides were washed twice at 42uC in 26SSC+0.1% SDS for 5 min, twice at room temperature in 0.16 SSC +0.1% SDS for 10 min, and finally twice at room temperature in 0.16SSC for 2 min 30 s. Unspecific binding of labeled proteins was blocked with 1% blocking reagent for nucleic acid hybridization (Roche). Slides were then stained with 16 mg/ml anti-Dig-Cy5 (Jackson Laboratories) and 18 mg/ml Streptavidin-Cy3 (Jackson Laboratories). In order to remove excess antibody/streptavidin, slides were washed with 46 SSC +0.2% Tween-20.
The microarray slides were scanned using the GenePix 4000A Arrayscanner (Molecular Device). The downstream data preprocessing and analysis was done in R using the limma package [14]. Raw probe intensities were background corrected by applying the 'normexpr' method. Analysis was restricted to Cy5 intensities. Loess normalization was used in M versus A plots of individual Cy5 intensities of a given transcript in a given array and the median Cy5 intensity across all arrays of the same transcript. Log2 ratios were calculated from the normalized intensities and quantile-normalization was subsequently applied across all arrays. All further analysis was based on normalized log ratios. Based on the hybridization date the data set is separated into two main batches of microarrays (early 2007/2008 and late 2009), which degrade further into small sub-batches. We used the ComBat algorithm [15] to adjust for the main batches. Transcripts differentially expressed between the groups were identified using regularized linear model as implemented in the limma package [16]. Transcripts were considered as significantly differentially expressed when their corresponding adjusted P-value was #0.05. Adjustment of P-values as correction for multiple testing was done as proposed by Benjamini et al. [17]. The differential gene expression analysis was used to generate candidates of regulated genes. The data have been deposited in NCBIs Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/) and assigned series accession number GSE52215.

Quantitative PCR
The differentially expressed genes for Decorin (Dcn), Gelsolin (Gsn) and Esterase D/formylglutathion hydrolase (Esd) were used to validate microarray results in an additional series of specific analytical and quantitative PCRs (qPCR). qPCR was performed using the LightCyclerH 480 instrument (Roche, Mannheim, Germany). The real-time quantification was conducted using LightCyclerH 480 SYBR Green I Master reagent (Roche, Mannheim, Germany). The cycling procedure consisted of an initial denaturation step (5 min at 95uC) followed by 38 cycles of 20 s at 95uC, 15 s at 58uC and 15 s at 72uC. All reactions were run in three replicas in a final volume of 19 ml containing 5 ml of cDNA template and 7 pmol of each forward and reverse primer. Positive and negative controls were included in all qPCR runs. Additionally, to confirm the specificity of reaction, all experiments included melt curve analysis. For each cell group three representative samples were generated by pooling equal amounts of singlecell cDNA samples assigned to a given group (10, 7 or 12 samples for Sca-1 + /CD34 + , Sca-1 + /CD34 2 and reference cells, respectively). This was done to decrease the measurement noise originating from cell-to-cell variability of gene expression levels. 1006 dilutions of the original cDNA sample representations were used as template for the qPCR. The relative gene expression levels were calculated using the efficiency corrected mathematical model described elsewhere [18]. Quantification was performed in parallel using three different reference genes Actb, Gapdh and Hprt1 (Hypoxanthine phosphoribosyl transferase 1), in each case giving highly comparable results. Group-wise comparison of relative gene expression levels was performed using 2-tailed Student's t-test. A value of p,0.05 was considered to indicate a statistically significant difference.

Identification of Sca-1 + and CD34 + Cells and Selection for Molecular Analysis
In order to isolate Sca-1 + and CD34 + cells 15 Balb/c mice between 8 and 12 weeks of age were sacrificed. Tissue digestion and cell separation started immediately after lung explantation. The number of obtained non-erythrocytic, non-apoptotic cells per experiment ranged from 0.40610 6 to 5.0610 6 (mean 1.6610 6 ; standard error of the mean (SEM) 1.16610 6 ). The number of erythrocytes ranged from 0.20610 6 to 2.6610 6 (mean 0.87610 6 ; SEM 0.73610 6 ), the number of apoptotic cells (i.e. PI + or GFP-A + cells) ranged from 0.92610 6 to 4.0610 6 (mean 1.9610 6 ; SEM 1.3610 6 ). The experimental setup is outlined in Figure 1.
Up to one million cells per animal were screened, equally divided for Sca-1/CD31 and CD34/CD45 immunofluorescence (i.e. up to 0.50610 6 cells for each staining and animal). We identified and isolated 68 Sca-1 + /CD31 2/ PI 2 cells, the number of obtained cells per mouse ranging from 0 to 10 (mean 4.5, SEM 3.0, Table 1 and Figure S1). Additionally, we detected 58 CD34 + / CD45 2/ GFP-A 2 cells, ranging from 0 to 13 in individual mice (mean 3.9; SEM 4.6, Table 1 and Figure S1). All detected cells were isolated and subjected to single cell gene expression analysis following whole transcriptome amplification (WTA). The enzymatically digested lungs of 6 additional mice yielding up to 1.7610 6 non-erythrocytic cells (mean 1.3610 6 cells; SEM 0.32610 6 cells) were stained with antibodies against CD31 and CD45 resulting in the isolation of 35 control cells (CD31 2/ CD45 2 ). All isolated cells were subjected to whole transcriptome amplification (WTA) and WTA quality controlled by expression of Actb and Gapdh. In total, 115/161 (71.4%) samples passed our quality assay and were subjected to further molecular analysis (no difference between cell types, data not shown).
We continued by testing for the presence of transcripts of the proteins targeted by antibodies in immunofluorescence: Sca-1, CD34, CD45 and CD31. Within the two populations of putative BASC, 35 cells completely lacked transcript expression of Sca-1/ CD34, while another 10 cells co-expressed CD31 and/or CD45 (Table 2). We decided to exclude those cells from further analyses which resulted in a cohort of 46 single putative BASCs remaining
In general, gene expression in single cells is subjected to stochastic fluctuations [19]. While in high-dimensional data (as microarrays) normalization generally uses many if not all probes, relative quantification in quantitative PCR (qPCR) experiments relies on stable expression levels of individual genes. This especially holds true for some house-keeping genes, as e.g. Actb or Gapdh, which renders reliable relative quantification using such an approach at the single cell level questionable [20,21]. Although specific genes can be expressed at a stable level between individual single cells of a certain cell type, this approach is not feasible for  Table 2. PCR results of corresponding transcripts in Sca-1+/ CD31-and cells CD34+/CD45-cells.   poorly described cell populations. Therefore, we decided to pool the single cell samples of a pre-defined subgroup for qPCR analyses to validate the results of the microarray analyses. For validation, we chose transcripts for Decorin (Dcn), and Gelsolin (Gsn), which were differentially expressed between Sca-1 + / CD34 + and reference cells, as well as Esterase D/formylglutathion hydrolase (Esd), which was in addition differentially expressed between Sca-1 + /CD34 + and Sca-1 + /CD34 2 cells. Besides the differential expression, the rationale for the selection of the three chosen transcripts was that they represent different functional groups of genes. As expected, cells from the Sca-1 + /CD34 + subpopulation expressed very high levels of all three transcripts in comparison to cells of the other two groups (student's t-test, p,0.05). If comparing Sca-1 + /CD34 2 cells with pulmonary reference samples, Dcn and Gsn were expressed in significantly higher levels in Sca-1 + /CD34 2 cells than Sca-1 2/ CD34 2 cells, while Esd was expressed at significantly lower level in Sca-1 + / CD34 2 cells than in pulmonary reference cells as determined by quantitative PCR (student's t-test, p,0.05, Figure 2B).

Single Cell Analysis of Epithelial and Mesenchymal Transcripts Allows Further Delineation of Sca-1 + /CD34 +,2 Subpopulations
To further clarify the epithelial or mesenchymal commitment of isolated cells, we checked expression of mesenchymal (CD90, Pdgfra) and epithelial transcripts (Epcam, Itga and Sftpc) as recently described [8]. For this purpose, we analyzed all 46 putative BASCs and 21 pulmonary reference cells by analytical PCR (Table S4).
In order to confirm the PCR results we performed FACS analysis of 5 additional digested lung explants using antibodies directed against Sca-1, CD34, CD45, CD31, Epcam and Pdgfra. Within the population of CD45 2/ CD31 2 cells, we could detect 3.61% of Sca-1 + /CD34 2 cells and 14.3% of Sca-1 + /CD34 + cells in mean ( Figure 3B). We proceeded to look at Epcam and Pdgfra expression in these two subpopulations of putative BASCs. As previously described, the majority of Sca-1 + /CD34 2 cells expressed Epcam, while Pdgfra was expressed in the Sca-1 + / CD34 + subpopulation. However, in good correlation to our PCR results, we found a significant subpopulation of Epcam + /Pdgfra + cells within the subpopulation of Sca-1 + /CD34 2 cells ( Figure 3B and C).

Discussion
In this study we investigated the gene expression profile of single Sca-1 + /CD31 2/ PI 2 and CD34 + /CD45 2/ GFP-A 2 cells that were detected by immunofluorescence and subsequently sub-divided into groups based on the protein expression of Sca-1 and CD34, as well as mRNA expression of subgroup-specific markers.
Sca-1 and CD34 have both been ascribed to bronchioalveolar stem cells (BASCs), a rare population of long-living, regional fixed, robust cells residing at the bronchioalveolar duct junction that have been sparsely investigated so far. Only few markers were established in previous studies mainly applying FACS analysis, which enables a rapid screening of very large cell numbers and subsequent comparison between cell populations as defined by applied protein markers. This approach shows high sensitivity and at the same time an elevated risk to isolate false-positive cells, which may impair the detection of extremely rare cells [22]. In this study, we intentionally utilized a microscope-based approach to isolate single cells to analyze cell-to-cell heterogeneity in poorly characterized cells. Considering the low incidence of putative BASCs in lung tissue, we believe that this approach is the method of choice for the molecular analysis of such extremely rare and poorly characterized cells. Beyond that, our workflow additionally allows direct assessment of morphology and viability of the target cells, thereby reducing the risk of contamination with false-positive cells or cell debris to an absolute minimum.
Combined protein and cDNA analysis of Sca-1 and CD34 in our study showed a highly positive correlation between staining and transcript detection (79.2% for Sca-1 and 68.2% for CD34, respectively). In addition, our approach enabled further classification in different subpopulations characterized by Sca-1 and CD34 transcript expression. Analyzed cells consisted of two larger groups of Sca-1 + /CD34 2 (n = 22) and Sca-1 + /CD34 + (n = 17) cells and a smaller group of Sca-1 2/ CD34 + (n = 7) cells.
To further investigate the expression profiles of the different cell types, we performed microarray-based gene expression analysis of single cell WTA products, including 10 Sca-1 + /CD34 + cells, 7 Sca-1 + /CD34 2 cells, and 12 Sca-1 2/ CD34 2 pulmonary reference cells. Comparisons between the different subgroups revealed 107 genes differentially expressed in Sca-1 + /CD34 + cells when compared to the reference cells, whereas no differentially expressed genes could be detected by comparing Sca-1 + /CD34 2 cells with reference cells. The comparison between the two Sca-1 + cell subgroups (Sca-1 + / CD34 + vs. Sca-1 + /CD34 2 ) yielded 8 differentially expressed genes. Validation by quantitative PCRs for the genes Dcn, Gsn and Esd confirmed the microarray results.
During the validation of the microarray results we intentionally abstained from performing qPCR analysis directly on single cell WTA products due to stochastic variation in expression level, a phenomenon that has been repeatedly reported before [19,[23][24][25]. Cell-to-cell heterogeneity in gene expression, which is also detectable in our data sets, may represent stochastic transcriptional bursts that are generated by intrinsic on-off transitions of the corresponding genes occurring at irregular intervals [23,25]. Fluctuations in levels of individual transcripts also affect the housekeeping genes, whose expression serves as reference in qPCR analyses [20,21]. Stable reference is indispensable to reliably quantify relative expression levels of genes of interest. Higher stability in gene expression levels facilitating more reliable downstream analysis can be achieved only by pooling single cell samples representing the same cell population, thereby averaging the stochastic variability of transcript levels in individual cells especially for reference genes. We selected three genes from the list of differentially expressed genes for qPCR validation coding for proteins of different functional groups (Dcn, Gsn, and Esd), representing extracellular matrix components, intracellular/membrane-bound proteins and cytoplasmic metabolic proteins, respectively. Antiproliferative or tissue stabilizing effects as well as the promotion of cellular motility and invasion have been demonstrated for Gelsolin and Decorin [26][27][28][29][30][31]. Both markers have been previously assigned to pulmonary fibroblasts. Consequently, their enhanced expression in Sca-1 + /CD34 + cells would be in line with mesenchymal commitment. On the other hand, an exclusive expression of the novel markers in mesenchymal cells has not been proven so far. Esd has been ascribed an important role in intracellular detoxification [32] making tumor cells more resistant against harmful substances [33].
Our results support an epithelial commitment of Sca-1 + /CD34 2 cells -in contrast to the Sca-1 + /CD34 + cells that displayed an increased incidence of mesenchymal marker expression. However, taking into account the novel subpopulation of Sca-1 + / CD34 2/ EpCAM + /Pdgfra + cells we postulate that single cell isolation and transcription profiling on rare cells represents a sensitive approach to obtain molecular data of single cells of interest. In this study, we showed on single murine lung cells that this approach has the power to deliver novel molecular markers that could contribute to a better understanding of the cellular heterogeneity and hierarchy in the murine lung. The presented approach may, therefore, help to analyze extremely rare and yet poorly characterized cells (e.g. stem cells) in other fields of biology as well.