Cell surface marker profiling of human tracheal basal cells reveals distinct subpopulations, identifies MST1/MSP as a mitogenic signal, and identifies new biomarkers for lung squamous cell carcinomas

Normal human tracheal tissue was obtained as surgical waste from lung transplant operations with written informed consent from lung donors, and with the approval of the University Health Network Research Ethics Board (study 08-0318-T). The tracheal tissue was healthy and was obtained from lungs deemed to be of suitable quality for lung transplants, but was not necessary for the transplant procedure. For high throughput antibody screening, the three strains of basal cells were derived from patients with the following characteristics: Strain 23 (age 47, male, non-smoker), Strain 38 (age 59, male, non-smoker), and Strain 37 (age 38, male, non-smoker). The 8227 leukemic cell line was derived from a primary human AML with written informed consent from the donor, and with approval of the University Health Network Research Ethics Board (study 02-0763-C).

Basal cells were isolated following the procedure of Karp et al. [28]. Briefly, tracheal tissue was digested with 1.4 mg/ml Protease type XIV (Sigma) in minimal essential media at 4°C for 24–48 hours. After digestion, tracheal epithelial cells were pelleted, rinsed with minimal essential medium, resuspended in LHC-9 media [20] with 10% fetal bovine serum and 10% dimethylsulfoxide, and stored in cryovials in a liquid nitrogen freezer.

Primary tracheal basal cells were expanded in LHC-9 media containing 50 μg/ml gentamycin, 1 unit/ml penicillin, 1 μg/ml streptomycin, and 1.25 μg/ml amphotericin B, and additional supplements as described [20], on plastic Petri dishes coated with 48 μg/ml PureCol (Advanced BioMatrix) diluted in 0.01 N hydrochloric acid. Cells were grown at 37°C in a 5% CO₂ incubator, and media was changed every other day. To subculture basal cells, cells were treated with 0.025% trypsin/EDTA, which was subsequently neutralized with an equal volume of Trypsin Neutralizing Solution (Lonza).

Rat tracheal xenografts were performed as described [29],[30] with the approval of the UHN Animal Care Committee. Briefly, the endogenous epithelium from adult rat tracheas was removed by three rounds of freeze/thaw. 1 × 10⁶ primary human tracheal basal cells were seeded per trachea, and one tracheal cassette was implanted subcutaneously onto the back of a single adult male NOD/SCID mouse. After implantation, tracheas were irrigated once per week with Ham’s F-12 media, and after three weeks, were irrigated twice per week. Tracheas were harvested between five to eight weeks after transplantation.

For immunofluorescence staining of primary tracheal basal cell cultures, basal cells were cytospun onto charged glass slides and fixed in 4% paraformaldehyde for 10 minutes at room temperature. Following fixation, excess paraformaldehyde was quenched by three five minute washes in 100 mM glycine/PBS. Cells were permeabilized with 0.1% Triton X-100/PBS, blocked with 3% bovine serum albumin (BSA)/0.1% Triton X-100/PBS for 30 minutes at room temperature, and then incubated with primary antibody diluted in 3% BSA/0.1% Triton X-100/PBS for one hour at room temperature. Primary antibodies were α-KRT5 (ab24647, Abcam, 1:100), α-TP63 (sc-56188, Santa Cruz, 1:100), α-MUC5AC (MS-145-P1, Neomarkers, 1:100), α-MUC16 (sc-33344, Santa Cruz, 1:100), and α-BTUB4 (T7941, Sigma, 1:100). After primary antibody incubation, cells were washed three times with 0.1% Triton X-100/PBS and incubated with secondary antibody in 3% BSA/0.1% Triton X-100/PBS for two hours at room temperature or overnight at 4°C. Secondary antibodies were Alexafluor 488 goat anti-rabbit (Invitrogen, 1:500) and Alexafluor 568 goat anti-mouse (Invitrogen, 1:500). After secondary antibody staining, cells were washed three times with 0.1% Triton X-100/PBS and mounted in Vectashield mounting media containing DAPI (Vector Laboratories).

Human tracheal and rat tracheal xenograft tissue were fixed in 10% buffered formalin overnight at room temperature, then soaked in 70% ethanol overnight at room temperature, and then paraffin embedded. Paraffin blocks were cut into 8 μm sections, placed on positively charged glass slides, and then dried at 55°C overnight. For immunofluorescence staining, sections were deparaffinized through successive incubations in xylene and decreasing concentrations of ethanol and antigens were retrieved in 10 mM citrate buffer, pH 6.0, using the 2100 Retriever (Aptum Biologics, Ltd.). Hematoxylin/eosin staining was performed using standard methods. Anti-MST1R/RON (Santa Cruz, sc-322, 1:400) immunohistochemistry was performed using the Vantana Benchmark XT autostainer with the iVIEW DAB detection kit (Vantana Medical Systems). Images were captured using an Axioimager Z1 microscope and either an AxioCam MRm camera with Axiovision software (Zeiss) or a Canon EOS color camera.

Fluorescence In Situ Hybridization (FISH) to detect human DNA in xenografted rat tracheal epithelia was performed on paraffin-embedded tissue using the human specific CEP3 Alpha Satellite DNA Spectrum Orange probe (Abbott Molecular) and the manufacturer’s instructions.

Primary human tracheal basal cells were expanded on plastic until ~70% confluent (7–10 days), removed with 0.025% trypsin, incubated with trypsin neutralizing solution, and then pelleted and resuspended in FC buffer (Hanks buffered salt solution with 1% fetal bovine serum). In control experiments with antibodies that detected heterogeneously expressed markers, we found that of 64 tested antibodies, only three (4.7%) detected epitopes that were trypsin-sensitive (Additional file 1: Table S3). Furthermore, this sensitivity was seen with a two-fold higher concentration of trypsin (0.05%), and was only partial, with marker expression reduced 2-3-fold. Thus, our trypsinization conditions did not have a major effect on marker detection. High-throughput staining and analysis were generally performed as described [31]. 50,000 tracheal basal cells in 100 μl were aliquoted per well into round-bottom 96-well plates, with each well containing a 2X concentration of each of 334 different fluorochrome-conjugated cell-surface targeted antibodies (PE, FITC, or APC; see Additional file 2: Table S1 for vendors). Wells containing buffer only were included as negative controls. The final antibody dilution was 1:50 for all antibodies. Staining was performed on ice for 20 minutes, after which plates were centrifuged, buffer aspirated, and pellets washed twice with 200 μl of FC buffer. Finally, pellets were resuspended in 100 μl of FC buffer containing 1 μg/ml of DAPI to allow exclusion of dead cells. Cells were then analyzed on a BD LSRII equipped with a high throughput sampler. A minimum of 10,000 events per well were collected using FACSDiva software, and resulting FCS 3.0 files exported to FlowJo version 9.3 for analysis. Debris, dead cells (DAPI-positive), and live cell doublets were excluded from analysis by using the gating strategy outlined in Additional file 3: Figure S1. Validation experiments were either performed in a high-throughput manner (with Strain 37) or a single tube format. For single tube staining, approximately 2 × 10⁵ cells were stained in 100 μl of FC buffer with a 1:50 dilution of the test antibody. Alternatively, α-CD45-APC (555485, BD Pharmingen) and α-CD326/EpCAM-FITC (sc-66020, Santa Cruz) were used at 1:10 and 1:25, respectively. CD44 was detected with α-CD44 (14-0441-81, eBioscience, 1:50) and goat α-mouse-APC (BD Pharmingen, 1:500).

Images were downloaded from the online resource at http://​www.​proteinatlas.​org. Antibodies used in the downloaded images included EGFR (Zymed, 28–0005), MST1R/RON/CD136 (SDIX, 1659.00.02), PDPN (MBL, D189-1), CD6 (Santa Cruz, sc-65249), CD9 (Novocastra, NCL-CD9), CD49C (Sigma, HPA008572), CD51 (Novocastra, NCL-CD51), CD54 (Sigma, HPA004877), CD55 (Sigma, HPA024386), CD66C (Sigma, HPA011041), CD71 (Zymed, 13–6800), CD91 (Sigma, HPA022903), CD99 (Dako Cytomation, M3601), CD116 (Santa Cruz, sc-21764), CD117 (Dako/Cytomation, A4502), CD138 (Sigma, HPA006185), CD142 (Santa Cruz, sc-20160), CD146 (Novocastra, NCL-CD146), CD226 (Sigma, HPA015715), CD227 (Dako Cytomation, M0613), CD268 (MBL, D201-3), CD276 (Sigma, HPA009285).

To examine surface marker gene expression in cultured, normal human bronchial basal cells, microarray gene expression data were downloaded from GEO (http://​www.​ncbi.​nlm.​nih.​gov/​geo/​) (accession number: GSE24337) [18]. All of the surface markers that were expressed in three or more isolates of basal cells (Additional file 6: Table S2), as well as all of the markers representing receptors for cleaved or secreted factors (Table 1), were selected for mRNA expression analysis. Surface marker genes were mapped to Affymetrix probes using ArrayTrans (http://​www.​ocigc.​ca/​arraytrans/​). 111/119 of these markers could be mapped to microarray probes. mRNA expression values were log normalized, and for each gene, the probe data with the largest expression value was used. A Pearson’s product moment correlation coefficient was then used to test for association between mRNA expression for a given marker and its average frequency in the population of normal basal cells (as determined by FACS).

To examine surface marker gene expression in primary lung cancers, RNAseq data (Tier 3) for adenocarcinomas and squamous cell carcinomas from The Cancer Genome Atlas were downloaded from cBioportal (http://​www.​cbioportal.​org) [32]. For this analysis, 110 out of the above 119 markers could be mapped to genes in the RNAseq data. Log quantile normalized RSEM values were used for testing differential expression between the two cell types, and a Bonferroni correction was used to adjust p-values for multiple testing.

AlamarBlue was purchased from Life Technologies and used according to the manufacturer’s instructions. Cell cultures were incubated with alamarBlue for four hours at 37°C, after which the red fluorescence of the cellularly reduced resorufin was measured in the media with a plate reader at 555/570 nm. To determine the correlation between alamarBlue readings and cell number, passage one (P1) human tracheal basal cells were seeded in duplicate in 12-well dishes at 2500 cells/cm². At days 1, 3, 5, and 7, alamarBlue readings were taken, after which the cell number was determined by staining cells with SYTO 60. Multiwell dishes were photographed on an inverted Olympus IX81 microscope with a 10× objective lens using a Rolera MGi Plus camera (QImaging) and MetaMorph software (Molecular Devices). The average number of cells in the field was determined from photographs of duplicate wells and was used as an approximation of cell growth. MST1/MSP was purchased from R&D Systems and was dissolved in PBS. For MST1/MSP growth assays, passage one (P1) human tracheal basal cells were seeded at 1000 cells/cm² in triplicate in 6-well dishes. Positive control cultures were grown in LHC-9 medium and negative control cultures were grown in LHC-9 medium lacking 5 ng/ml EGF and 100 μg/ml BPE. Cells were fed with fresh medium and growth factor every other day beginning at day 1, at which point alamarBlue readings were also taken.

To examine expression of hundreds of cell surface receptors at a single cell level in human tracheobronchial basal cells, we chose a FACS-based approach using well characterized FACS antibodies, which would be more efficient than optimizing tissue staining for hundreds of antibodies. The basal cells were extracted from healthy adult human tracheal tissue (Figure 1A) using previously described methods [27],[28],[30],[33]. Although our goal was to profile basal cell surface marker expression in as close to an in vivo setting as possible, we found that effective isolation of these cells from tissue required pronase digestion, which appeared to cleave some surface markers including the canonical basal cell marker CD44 [2],[4]. Expression of CD44 and other surface markers by FACS was determined in individual live tracheal cells with the general gating strategy outlined in Additional file 3: Figure S1. Anti-CD44 FACS analysis of tracheal cell suspensions derived from fresh pronase-digested tissue identified fewer than the expected number of basal cells (predicted to be ~30% by previous studies [16],[34] and empirically determined by us using anti-TP63 staining) (Additional file 7: Figure S2). Furthermore, the CD44-positive cells in these suspensions had a lower overall fluorescence intensity than control basal cells that had recovered in vitro and were removed with trypsinization (Additional file 7: Figure S2). To circumvent the pronase problem and increase the yield of cells for high-throughput surface marker profiling, we transiently expanded pronase-digested tracheal suspensions over 7–10 days without passaging. By potentially mimicking an injury state, this approach also had the advantage of possibly expanding subpopulations of basal cells that might be rare in a quiescent epithelium in vivo. During this time, tracheal cells were grown in a serum-free medium that is known to support tracheobronchial stem cell/progenitor activity [20],[26]. After expansion, cells were removed with a mild trypsinization protocol, which we found cleaved few surface markers (see below and later).

The purity of the expanded primary tracheal cell cultures was first assessed by antibody staining for intracellular epitopes on canonical basal cell and differentiated columnar cell markers. 99% and 97% of the cells stained positive for the basal cell markers KRT5 and TP63, respectively (Figure 1B,C,K), while none of the cells stained positive for BTUB4, a ciliated cell marker [35], or MUC5AC and MUC16, tracheal mucinous cell markers [36],[37] (Figure 1E-K). By FACS, 100% of the cells were positive for CD44 (Additional file 8: Figure S3A). These CD44-positive cells were not hematopoietic since none of the tracheal cells stained positive for the pan leukocyte CD45 marker [38] (Additional file 8: Figure S3B) using an anti-CD45 antibody clone that recognizes a trypsin-insensitive epitope [39]. In contrast, ~100% of cells from a control hematopoietic cell line were positively stained with the same anti-CD45 antibody (Additional file 8: Figure S3B). These data indicate that our primary tracheal cultures comprise only basal cells. KRT14, which is heterogeneously expressed in human and mouse large airway basal cells in vivo [11],[13], was also expressed heterogeneously in the ex vivo cultures (Figure 1D,K) as reported in other ex vivo work [13], providing evidence for retention of some in vivo heterogeneity ex vivo.

We next verified that our primary human tracheal basal cell cultures maintained functional stem cell/progenitor activity. It has been reported that with the isolation and expansion conditions we used, human tracheal basal cells retain multipotentiality, as revealed in rat tracheal xenograft assays [8]. In this assay, human tracheal basal cells are seeded into denuded rat tracheas and implanted as xenografts onto the backs of immunocompromised mice [8],[30]. This assay is one of the most stringent assays for human tracheal stem cell activity. It allows for the generation of the most in vivo-like pseudostratified mucociliary epithelium, and it uniquely allows assessment of submucosal gland lineage potential [8]. To verify that our primary tracheal basal cell culture conditions maintain stem cell activity, we assessed the ability of basal cells derived from two different donor tracheas to repopulate denuded rat tracheas. After five weeks, engrafted primary tracheal basal cells gave rise to morphologically well-differentiated mucociliary epithelia and occasional submucosal glands (Figure 2). These epithelia and glands did not form in the absence of transplanted cells (Figure 2A), closely resembled their counterparts in normal human tracheal tissue (Figures 1A and 2B, C-E), and were of human origin, as assessed by FISH for a human-specific centromeric probe (Figure 2F). Thus, our isolation and culture conditions generally support the ex vivo expansion of tracheal basal cells capable of giving rise to multiple differentiated lineages.

To investigate expression of cell surface receptors at a single cell level and to identify new signaling pathways that potentially modulate tracheobronchial basal cell behavior, we used flow cytometry to profile expression of 332 surface markers in primary, non-passaged human tracheal basal cell cultures. Human tracheal basal cells derived from two different donor tracheas were stained and analyzed by FACS in a high-throughput manner. 314 out of the 332 tested antibodies yielded good quality FACS data in both basal cell strains (Additional file 2: Table S1). 108 of these antibodies stained less than 0.1% of the total population in both strains of basal cells, which we set as a cut-off for detectable expression (Additional file 2: Table S1). They provided a large number of isotype controls that indicated basal cells are not generally “sticky” to all antibodies. Some of these antibodies included anti-CD31, which is commonly used to detect endothelial cells [40], as well as three different anti-CD45 antibodies (Additional file 2: Table S1), and support our earlier findings that our tracheal cell cultures are not contaminated with non-basal cell types. 157 markers were detected in both basal cell strains and were expressed on varying numbers of cells (Table 2). These shared markers represented 80% and 95% of all expressed markers in each strain, indicating an overall strong similarity in the immuno-surface phenotype between the strains (Table 2). To further confirm the epithelial origin of rarer subpopulations of tracheal cells, we co-stained primary tracheal cell cultures with antibodies to a few of these markers, as well as CD45 and CD326/EpCAM. CD326 is only expressed in epithelial cells, including bronchial epithelial cells [41],[42]. By gating on cells that were CD326-positive and CD45-negative, we confirmed that these rare subpopulations are indeed, epithelial (Figure 3).

We then chose to validate 110 of the shared expressed markers in one or two additional basal cell strains derived from different donor tracheas (Additional file 2: Table S1). These validation strains did not have detectable contamination with CD31 or CD45-positive cells (Additional file 2: Table S1). 105 of the 110 basal cell markers were detected in these additional strains, identifying a large set of markers expressed in three to four independently isolated basal cell strains. The high degree of validation (95%) suggests that most of the markers identified as being shared between the initial screening strains can be extended to all isolates of human tracheal basal cells. Within the group of 105 markers, 74 were expressed with a mean frequency of less than 90% (Additional file 6: Table S2), suggesting that they are heterogeneously expressed within the basal cell bulk population. Heterogeneous expression of these markers (especially low expression) was not due to our mild trypsinization protocol. Prior control studies indicated that the anti-CD31 and anti-CD45 antibodies, as well as most of the antibodies to the heterogeneously expressed markers recognize trypsin-insensitive epitopes [31] (summarized in Additional file 1: Table S3). In these control experiments, a cocktail of collagenase, hyaluronidase, dispase, and a two-fold higher concentration of trypsin only affected FACS detection of 17 of 68 tested markers of heterogeneity. Furthermore, of these 17 markers, detection of only three (CD62L, CD195, CD213a2, 4.4% of number assayed) was affected by trypsin in isolation. Since expression of these three markers was only affected two to three-fold while their mean frequencies in the bulk basal cell population ranged from 0.6-3.7%, even partial sensitivity to trypsin cannot account for their heterogeneous expression. For 47 markers, mean frequencies were measured between 0.1 to <90% of the total basal cell population with 95% confidence (Additional file 6: Table S2), identifying good candidates for heterogeneously expressed markers. Examples of representative FACS plots for these markers are shown in Figure 4. Some markers displayed a lot of variation in the number of positive cells in different basal cell isolates, while others, especially those expressed on fewer than 6% of all basal cells, were detected at consistently similar frequencies (Additional file 6: Table S2). By looking at marker frequencies with non-overlapping 95% confidence intervals, it appeared that using several different cut-offs, subpopulations of basal cells could be grouped into more (e.g. ~ > 13%) or less (e.g. ~ < 6%) abundant subpopulations (see Figure 5A for examples).

While a complete analysis of the total number of subpopulations represented by marker heterogeneity was beyond the scope of this work, we, nevertheless, investigated potential overlap between some subpopulations. CD90-expressing cells (~3% of the total population) did not overlap with CD36L1 (0.75%) or CD227 (1.31%)-positive cells, indicating that CD90-positive cells are distinct from these other subpopulations (Figure 5B). Furthermore, CD36L1 and CD227 did not mark identical subpopulations since >90% of CD36L1, but only 12% of CD227-positive cells expressed CD66c (Figure 5C).

To corroborate the surface marker heterogeneity observed by FACS by another method, we investigated if marker FACS data correlated with marker mRNA expression in the bulk population. We focused on the 105 surface markers that were validated in three or more basal cell isolates (Additional file 6: Table S2), as well as 14 additional signaling receptors that were not analyzed by FACS beyond the initial high-throughput screening (part of Table 1). mRNA levels were determined from five microarray datasets of gene expression in primary human bronchial basal cells that were grown under the same culture conditions used by us [18]. Probes for 111/119 markers were on the microarrays yielding 92% coverage. Overall, we found that the percentage of marker-positive cells detected by FACS was correlated with mRNA expression for that marker (Figure 6A). These data corroborate the accuracy of the antibodies in detecting their antigens, and support the FACS data indicating basal cells exist as distinct, molecularly identifiable subpopulations. We also found that in the microarray analysis [18], 21/111 of our markers were independently annotated as basal cell-enriched genes relative to the other large airway epithelial cells (Figure 6B). These data additionally suggest that a number of our markers are not only expressed on basal cells, but also define basal cell lineage-specific markers.

To determine if some of the basal cell markers we identified in vitro might mark basal cells in normal airway tissue in vivo, we used the Human Protein Atlas (http://​www.​proteinatlas.​org), which is an expansive online resource for normal and cancer tissue immunostaining for the human proteome [43]-[47]. Normal tissues included the nasopharynx and bronchus, which share the same type of epithelium as the trachea. Good tissue sections were available for 83 of the 105 markers we identified in three or more basal cell isolates. 63 of these markers showed some expression in airway epithelia. Expression ranged from all epithelial cells to exclusively all basal cells, to only subsets of basolaterally located cells (but not necessarily, basal epithelial cells). By using the traditional morphological criteria that were used to establish basal cell-specific expression of the canonical basal cell markers (cuboidal shape and occupying exclusively and uniformly basolateral positions) KRT5, KRT14, CD44, and GSA-I-B4 reactivity [3],[4],[48]-[50], it appeared that CD55, CD49C, CD276, CD146, CD9, CD142, and the EGFR, which were detected on most basal cells in vitro, were also expressed on most basal cells in vivo (Figure 7). While PDPN, CD91, CD138, and possibly CD51, also appeared to be enriched on most basal cells in vivo (Figure 9), in vitro, these markers were expressed only on subsets of basal cells. Collectively, these eleven markers identify pan-basal cell lineage markers. CD54, CD227, CD99, and possibly CD6, were expressed heterogeneously in vitro and at basolateral positions in vivo (Figure 7). However, in these cases, it is not clear if they mark subsets of basal cells or rare, infiltrating other cell types.

To further support the in vivo relevance of the new basal cell markers and to begin to explore their potential relevance to cancer biology, we also examined marker expression in squamous cell lung cancers (SQCCs), which likely arise from basal cells. SQCCs develop in the large airways and >90% of SQCCs express pan-basal cell lineage markers such as KRT5 and TP63 [51],[52]. To strengthen the interpretation that some SQCC marker expression reflects the basal cell origin of SQCCs, we also examined marker expression in control lung adenocarcinomas (ADCs), most of which are not likely to arise from basal cells. ADCs tend to develop distally, with KRT5 and TP63 expression being infrequent [51],[52]. For this analysis, we used the same panel of 119 markers that was employed in the microarray analysis. 110 of these markers could be mapped to genes in The Cancer Genome Atlas lung cancer RNAseq datasets [32]. 22% (24/110) of these markers had significantly higher expression (p-value < 0.05) in SQCCs compared to ADCs (Additional file 4: Table S4, Figure 8). This degree of enrichment is much larger than expected by chance, as using the same statistical analysis and threshold for significance (p-value < 0.05), only 1.4% (290/20,725) of total genes was estimated to be significantly upregulated in SQCCs relative to ADCs. The degree of basal cell marker enrichment in SQCCs was also greater than observed for 101 control markers not expressed on basal cells (expressed on <0.1% of all basal cells) (7%, Additional file 5: Table S5, Figure 1). In this latter case, the 7% enrichment is likely an overestimate of true enrichment due to the misclassification of some markers as being not expressed on basal cells. Some FACS antibodies may have yielded false negative results due to their inability to recognize all protein isoforms and their modified variants for a given gene. Notably, among the control markers, the two most SQCC-enriched proteins, JAG1 and FGFR2, were not detected on basal cells by FACS, but by mRNA and other protein analyses, are known to be expressed on human airway basal cells growing in vitro and in vivo [18],[53]-[55].

While the fold-differences in marker mRNA expression between the SQCCs and ADCs were not dramatically different, many of these differences are likely to be meaningful at the protein level. Protein expression of 14/24 of the SQCC-enriched markers has been detected in SQCCs [56]-[67], and at least in seven cases (CD98/SLC3A2, CD44, CD221/IGF1R, CD276, CD141/THBD, EGFR, PDPN), marker protein expression was reported to be greater or more common in SQCCs as compared to ADCs [58],[62]-[66],[68]-[70]. Furthermore, expression of three markers has been reported to be prognostic with regards to overall survival in SQCCs (CD98/SLC3A2, CD44, PDPN) [60],[63],[71]-[73]. To obtain protein data to corroborate the mRNA data for the remaining SQCC-enriched markers not previously interrogated in SQCCs, we examined marker protein expression in lung cancer samples through the Human Protein Atlas (http://​www.​proteinatlas.​org). While the small number of cancer samples did not allow assessment of statistical significance, several of the new markers identified through mRNA analysis also showed trends of increased protein expression in SQCCs relative to ADCs (CD71/TFRC, CD9, PDPN, CD138/SDC1, CD99) (Figure 9, Additional file 9: Table S6). Using the Human Protein Atlas, we also found that some of the heterogeneously expressed basal cell markers that were not identified as SQCC-enriched markers, were, nevertheless, still expressed in SQCCs (Figure 9, Additional file 9: Table S6, CD227/MUC1, CD54/ICAM1).

Since some of the SQCC-enriched markers have already been shown to have prognostic significance to SQCCs [60],[63],[71]-[73], we examined the remainder of these markers for the relationship of their expression to overall survival by Kaplan Meier survival analysis. We used KM-plotter (http://​kmplot.​com/​analysis/​) [74], an online resource that compiles gene expression and overall survival data from the Gene Expression Omnibus (http://​www.​ncbi.​nlm.​nih.​gov/​geo/​) and The Cancer Genome Atlas (http://​cancergenome.​nih.​gov/​). Using the entire SQCC and ADC datasets, we found that expression of four markers had significant associations with overall SQCC patient survival (Figure 10). Two markers were associated with worse overall survival (CD49E/ITGA5, CD141/THBD) and two markers were associated with better overall survival (CD49F/ITGA6, CD222/IGF2R). For CD141/THBD and CD49F/ITGA6, the direction of the association with overall survival was unique to SQCCs. This finding could relate to these markers having different functions in SQCCs and ADCs and their distinct cells of origin. CD49E/ITGA5 and CD222/IGF2R were prognostic in the same direction in both SQCCs and ADCs, suggesting that despite elevated expression in SQCCs, these proteins may perform similar functions in the two tumor types and their cells of origin. Collectively, these normal and tumor tissue data support the relevance of our newly identified markers to in vivo populations of normal and transformed basal cells.

Among the surface markers detected on human tracheal basal cells were 33 receptors that are involved in signaling by cleaved or secreted factors (Table 1). Receptors for EGF, IGF, insulin, NGF, MST1/MSP, IFNγ, FASL, and some TNF family growth factors were expressed on most basal cells while receptors for leptin, LIF, VEGF, SCF, chemokines, interleukins, and other TNF family members were expressed on fewer cells. These observations suggest that some factors may act on all basal cells while others may regulate distinct subpopulations. Some of these receptors are known to be functional as EGF and insulin are essential components of the serum-free medium used to culture human tracheal basal cells [19],[20],[22] and NGF promotes survival of basal cells after RSV infection [75].

Since the biological function of most of these receptors is unknown in large airway basal cells, we chose to investigate one receptor in more detail. CD136/MST1R/RON is the receptor for MST1/MSP and by FACS, was expressed on all basal cells (Table 1, Additional file 2: Table S1). Although a biological function for MST1/MSP has not been reported in basal cells, it has previously been shown to directly bind cultured human tracheal basal cells and to induce phosphorylation of MST1R/RON [76]. Furthermore, MST1R/RON missense mutations and expression of an alternatively spliced oncogenic MST1R isoform have been reported in lung SQCCs [32],[77], suggesting this receptor may promote growth of basal cells, as it does in TP63-positive keratinocytes [78]. To address this possibility, we cultured tracheal basal cells in the absence of EGF and bovine pituitary extract, which allowed the cells to remain viable, but reduced cell growth. We then added two different doses of MST1/MSP and measured cell growth by alamarBlue, which we found is directly proportional to cell number (Additional file 10: Figure S4). Both doses of MST1/MSP promoted growth, with the higher dose yielding half the maximal cell number obtained with optimized amounts of EGF and bovine pituitary extract (Figure 11A). We also verified that MST1R/RON was expressed in tracheal basal cells in vivo by immunostaining human tracheal tissue with anti-MST1R antibodies. MST1R/RON expression was detected in all cells of the tracheal epithelium, including basal cells (Figure 11B). This expression pattern was corroborated by the Human Protein Atlas with a different anti-MST1R/RON antibody in the nasopharyngeal epithelium (Figure 11C), where MST1/MSP has been reported to increase the beat frequency of ciliated cells [76]. Thus, MST1/MSP has distinct effects on basal and non-basal cell types in the large airways.