Altered cell-cycle control, inflammation and adhesion in high-risk persistent bronchial dysplasia

Abstract

Persistent bronchial dysplasia (BD) is associated with increased risk of developing invasive squamous cell carcinoma (SCC) of the lung. In this study, we hypothesized that differences in gene expression profiles between persistent and regressive BD would identify cellular processes that underlie progression to SCC. RNA expression arrays comparing baseline biopsies from 32 bronchial sites that persisted/progressed to 31 regressive sites showed 395 differentially expressed genes (ANOVA, FDR {less than or equal to} 0.05). 31 pathways showed significantly altered activity between the two groups, many of which were associated with cell cycle control and proliferation, inflammation, or epithelial differentiation/cell-cell adhesion. Cultured persistent BD cells exhibited increased expression of polo-like kinase 1(PLK1), which was associated with multiple cell cycle pathways. Treatment with PLK1 inhibitor induced apoptosis and G2/M arrest and decreased proliferation compared to untreated cells; these effects were not seen in normal or regressive BD cultures. Inflammatory pathway activity was decreased in persistent BD, and the presence of an inflammatory infiltrate was more common in regressive BD. Regressive BD were also associated with trends toward overall increases in macrophages and T lymphocytes and altered polarization of these inflammatory cell subsets. Increased desmoglein 3 and plakoglobin expression was associated with higher grade and persistence of BD. These results identify alterations in the persistent subset of BD that are associated with high risk for progression to invasive SCC. These alterations may serve as strong markers of risk and as effective targets for lung cancer prevention.

INTRODUCTION

Mortality associated with non-small cell lung cancer (NSCLC) is higher than the next four tumor types combined and will claim the lives of more than 160,000 people in the United States this year (1,2). Early detection via screening with low dose computed tomography (LDCT) has contributed to improved survival (3,4), but is most effective for detection of peripheral tumors that are predominantly of the adenocarcinoma subtype. LDCT does not detect bronchial dysplasia (BD), the precursor of invasive squamous cell carcinoma (SCC). BD arises in the central airways of the lung, and can be detected by bronchoscopic means especially when advanced techniques such as autofluorescence, narrow-band imaging and other recently developed technologies are employed (5,6). When present, these lesions represent a stage in the development of invasive lung cancer at which application of preventive measures in secondary (patients without a history of lung cancer) or tertiary (patients with a history of lung cancer) settings could be utilized to significantly reduce lung cancer associated mortality.

The relationship between BD and SCC is well established with the demonstration of increased risk when higher grades of BD are present and the description of molecular alterations in BD that parallel those seen in invasive SCC (7–13). However, the large majority of BDs, including high grade lesions (moderate dysplasia or worse), do not progress to invasive SCC (14–16). We have recently shown that patients harboring multiple sites within the airways that persist as or progress to high grade dysplasia represent a subset of patients with aggressive airway disease that have a 7.8-fold increase in risk for the development of SCC (17).

We hypothesize that changes in gene expression that distinguish high risk persistent from low risk regressive BD will elucidate key cellular activities that mediate progression of premalignant lesions to invasive SCC. Cellular processes associated with persistence are further explored to demonstrate a potential role in mediating malignant progression. The description of these gene expression alterations will provide information that can be used to predict risk associated with BD and identify targets for prevention of progression to invasive SCC.

MATERIALS AND METHODS

Subject and biopsy characteristics.

All tissues used in these studies were obtained from subjects who directly or through their guardians signed informed written consents for these types of analyses. These tissues were collected via Colorado SPORE in Lung Cancer bronchoscopy protocols that were approved by the Colorado Multiple Instutitional Review Board (CoMIRB). CoMIRB and University of Colorado investigators follow the policies of the U. S. Common Rule. Biopsies from the treatment arm of the Iloprost chemoprevention trial were excluded as this trial demonstrated a reduction in outcome histology related to Iloprost treatment in former smokers (18). Inclusion criteria consisted of availability of snap frozen biopsy tissue from a site with a known baseline histology as defined from the adjacent formalin fixed, paraffin embedded (FFPE) biopsy, and a known outcome histology as defined in FFPE biopsy at the same bronchial site on a subsequent bronchoscopy (mean time to follow-up biopsy, 12.3 months, see supplemental table S1). Autofluorescence bronchoscopy was used to initially identify dysplastic lesions in the airways. Follow up biopsy sites were determined by reviewing prior images using a common map and algorithm of sequential appearance and orientation of orifices in the airways in protocols developed and employed by bronchoscopists with over 25 years of experience (RLK, YEM and TK). Histology was scored using a system in which numeric assignments were given for each level of atypia as reported previously (17, see figure 1). Inflammation scores generated during microscopic assessment of hematoxylin and eosin (H&E) stained biopsy slides indicate overall cellularity attributed to inflammatory cells as normal (<5%), mild (5–25%), moderate (25–75%) and severe (>75%). Where possible, specimens in test groups were selected to maintain similarity in respect to baseline histology, age, gender, tobacco history and history of NSCLC. Frozen sections were performed on the tissue used for RNA isolation and inclusion required histology that was within one diagnostic score of the adjacent FFPE section (frozen diagnoses were assigned to a site when there was a one score difference). Two-hundred ng of total RNA and RNA integrity number (RIN, Agilent Bioanalyzer) of 2.5 or higher were also required for inclusion. Consistency of overall gene expression levels within groups were evaluated to confirm that differences in RIN did not significantly correlate with differences in gene expression levels. Sixty-three of 107 processed frozen biopsies met criteria for inclusion with 21 excluded for discordant histology and 23 for inadequate RNA quality/quantity. Cultures of primary, un-immortalized bronchial epithelial cells were selected from Colorado SPORE in Lung Cancer tissue bank stores of cultures previously established from fresh biopsy tissue. Because frozen and fresh biopsies were collected only from selected sites at bronchoscopy (usually one or two sites per bronchoscopy) all but three cultures were from sites other than those included in the gene expression analysis. Cultures were classified as persistent or regressive dysplasia based on diagnoses from concurrent immediately adjacent and subsequent FFPE biopsies at the same site.

Figure 1.

Summary of specimens analyzed and differentially expressed genes identified in persistent versus regressive bronchial dysplasia (BD). A) Representative H&E images (400X) from each of the four classifications of biopsy sites according to baseline and follow-up histology scores at a specific site (ND = non-dysplasia). B) Graphical representation of all sites included in gene expression analysis comparing persistent (left) and regressive (right) sites. Lines connect baseline (gray dots) and follow-up (black dots) histologic scores (1 = normal, 2 = reserve cell hyperplasia, 3 = squamous metaplasia without atypia, 4 = mild bronchial dysplasia, 5 = moderate dysplasia, 6 = severe dysplasia and 7 = carcinoma in-situ) for consecutive biopsies from a single site within the airway of an individual. C) Venn diagram showing overlap of genes that distinguish persistent from regressive sites (light gray circle) and those associated with histology score regardless of outcome (dark gray circle).

Gene expression microarray analysis.

Baseline histologic scores of 4 or higher were considered dysplasia and less than or equal to 2 considered non-dysplasia. Processing of biopsies for gene expression analysis was performed with simultaneous evaluation by H&E stained frozen sections. Biopsies frozen in optimal cutting tissue solution (OCT, Fisher Healthcare) at collection and stored in liquid nitrogen were placed in a cryostat at −25^oC and mounted in toto for sectioning. A superficial section was assessed by H&E stain and if demonstrating the appropriate histology, tissue was removed from the cryostat chuck in a manner to prevent thawing and immediately placed in RNEasy (Qiagen, Germantown, MD). RNA was extracted within 60 minutes and resuspended in 12 μl of RNAse free water. One μl was used for bioanalyzer assessment. If the initial H&E section did not show expected histology, deeper sections were taken until the expected histology was identified or until the tissue was exhausted. Labeled RNA was hybridized with the Affymetrix Hu 1.0 array and expression levels measured by address specific signal intensity (GEO accession number: GSE114489). Gene expression data were analyzed by classification of biopsy source in two ways, firstly according outcome histology and secondly by degree of dysplastic change in the baseline biopsy without consideration of outcome.

Culture based PLK1 expression, apoptosis and cell cycle analyses.

Bronchial epithelial cultures were established as previously described (19). Stocks of cells frozen at 3–6 weeks after primary isolation were thawed as passage 3 cultures. All experiments were performed on passage 4–6 cultures and senescence was seen with all cell lines between passages 7–10. Identity of selected cell lines used for extended analyses were confirmed by standard STR (short tandem repeat) testing of cultured cells and associated peripheral blood collected from patients at enrollment. 15–25 μg/lane whole cell lysate from regressive and persistent primary, non-immortalized dysplastic cell lines at 60–90% confluence were analyzed by standard Western Blotting procedures using the Li-Core system. These cultures were also grown to near confluence (80–90% confluent) and either harvested for collection of RNA (RNeasy) or treated with vehicle alone (DMSO, Sigma, St. Louis, MO) or vehicle plus 100 nM PLK1 inhibitor (Volasertib, Selleck, Houston, TX) for 48–96 hours. Parallel cultures were performed in triplicate with one set each used to measure apoptosis via caspase 3/7 activity (ApoTox-Glo, Promega, Madison, WI), cell count by manual cytometer and cell cycle fraction via flow cytometric analysis (University of Colorado Flow Cytometry Core). PLK1 expression levels were measured in TaqMan real-time quantitative PCR analyses using validated QuantiTect Primer and SYBR Green PCR mix based assays for PLK1 and GAPDH (Qiagen, Germantown, MD).

Immunohistochemical and immunofluorescent staining.

Pre-cut slides with 4–8 five-micron sections of bronchial biopsy tissue from the Colorado SPORE in Lung Cancer tissue bank were obtained from persistent or regressive sites. Antigen retrieval was performed in citric acid buffer, pH 6.0 (pH 9.0 for PLK1/Cytokeratin), under pressure for 15 min. to prepare for overnight incubation at 4°C with primary antibodies (IHC/IF) for CD3 (Ventana-2GV6), CD4 (Ventana-SP35/Abcam-EPR6855), CD8 (Ventana-SP57), CD68 (Ventana-KP-1/Abcam-KP-1), HLADRA (Abcam-TAL15B), CD206 (Novus-EPR6828(B)), FoxP3 (Abcam-236A-E7), Junctional plakoglobin (PG) (Fitzgerald-PG 5.1), Desmoglein 3 (AbD Serotec-5G11), PLK1 (Cell Signaling Technology-208G4), and Pan-keratin (Thermo-Fisher-AE1/AE3). Image annotation was used on IHC stained slides to demarcate dysplastic epithelial and underlying stromal compartments for image analysis (Aperio, Leica Biosystems) in which percent of positive pixels were used to indicate relative levels of expression in each stained slide. H&E slides for the cases included in this analysis were reviewed prior to inclusion to confirm the presence of appropriate epithelial histology and adequate associated underlying stroma. The mean area of epithelium and stroma analyzed was larger in the group of persistent BD as compared to the regressive BDs: mean epithelial area 0.139 vs. 0.061 (minimum 0.006 vs. 0.007 and maximum 1.033 vs. 0.327) mm²; mean stromal area 0.173 vs. 0.103 (minimum 0.008 vs. 0.010 and maximum 0.994 vs. 0.447) mm², respectively (p<0.01). Thus, all results were normalized to the measured area scored for each case. The accuracy of the image analysis was confirmed by comparing manual counts of positive cells by two pathologists (DTM, MCO) to the relative levels of positivity as determined by pixel counts to show similar levels of difference in 12 co-scored cases. Immuno-scores (I-score) for DSG3 and PG expression were calculated by manual classification of dysplastic epithelial regions into proportions (percentage) of negative (intensity = 0) to strongly positive (intensity = 3) areas with the I-score calculated by summing the % area x intensity for each level of intensity. Two scorers (DTM and EJD) scored all stained sections and cases with I-scores that were more than 50 units different between the two scorers (six cases) were reviewed to generate a consensus score. Otherwise mean scores were used in analyses.

Statistical methods.

Gene expression analysis was conducted using Affymetrix Human Gene 1.0 ST microarrays, and analyzed using RMA and ANOVA methods. Significant differences in expression were defined using a false discovery rate cutoff of 0.05% based on a corrected T-test. The primary comparison of persistent versus regressive bronchial sites used expression data that was normalized to the mean expression level of a given gene across all groups for the baseline HS that each site demonstrated. The genes meeting criteria for significance were further characterized by employing Ingenuity Pathway Analysis (Qiagen, Germantown, MD) or KEGG Pathway database (GenomeNet) software to identify associations with biological pathways, biological networks and upstream regulators that are associated with the genes comprising the final genelist. A p-value of ≤0.05 was used to represent statistical significance in these analyses. Raw gene expression levels were also assessed for trends over the range of baseline histologic scores. Spearman correlation was applied and r-values of ≥ +/− 0.5 were considered significant. A p-value of ≤0.05 was used to indicate significant evidence of activation or inhibition of upstream regulator activity. An activation z-score was used to rank upstream regulators in respect to degree of activity. Chi-square and Student’s T-tests were employed to assess potential differences between study groups in clinical parameters and in analyses of gene expression, cell cycle, apoptosis and proliferation in cell culture analyses. Validational analyses employ Student’s T-tests of gene expression levels from real-time RT-PCR values normalized to GAPDH expression within each of the triplicate measurements for each cell line. For PLK1 expression analyses, two values in regressive BD derived cell lines were excluded due to outlier status (more than 20-fold difference from the mean of the other two replicate measures for that cell line). T-tests included each measurement for each cell line. A p-value of ≤0.05 is considered significant and p-values of 0.05 – 0.15 are considered to be a trend.

RESULTS

Gene expression analysis of bronchial biopsy tissue

Sixty-three frozen biopsies were selected for gene expression analyses. Twenty-three sites represented persistent dysplasias with mild dysplasia (histologic score = 4) or higher histologic score (HS) at baseline and at least mild dysplasia on follow-up biopsy. The follow up biopsies for this group always showed an HS that was not less than one lower than the baseline score except for one site in which the baseline score was 6 and the follow-up score was 4. This site was included in the persistent group because additional biopsies collected at this site after the initial follow-up showed histologic scores of 6. Fifteen regressive dysplasias were characterized by an HS of 4 or higher on baseline sampling and non-dysplastic HS (normal, HS = 1 or reserve cell hyperplasia, HS = 2) on follow-up biopsy. Nine progressive non-dysplasias (baseline HS = 1–2 and follow-up HS ≥ 4) and sixteen stable non-dysplasias (HS = 1–2 on baseline and follow-up biopsy) were also characterized (figure 1A). Initial comparisons showed differences in gene expression in histologically comparable baseline biopsies from persistent BD versus regressive BD (142 genes, supplemental table S2). While dysplastic biopsies from regressive sites showed fewer than 90 differentially expressed genes compared to histologically distinct stable non-dysplastic sites, persistent sites showed more than 8 times this number versus stable non-dysplasia (780 genes). Additionally, non-dysplastic baseline biopsies that progressed to dysplasia showed 221 differentially expressed genes in comparison to stable non-dysplastic sites. Given the minimal differences in gene expression between regressive bronchial dysplasia and stable non-dysplasia, the final analysis combined groups into those with dysplasia on follow-up (persistent BD and progressive non-dysplasia, referred to as “persistent sites” throughout manuscript) and compared this group to those with non-dysplastic histology on follow-up (regressive BD and stable non-dysplasia, referred to as “regressive sites”) (figure 1B). The baseline HS of dysplastic sites, baseline HS of non-dysplastic sites, time to follow-up biopsy, smoking status, pack year smoking histories, frequency of history of invasive lung cancer, age, gender and ethnic distribution were not statistically different between these groups though a higher percentage of current smokers in the persistent group was near statistical significance (supplemental table S1, specimen specific data table S3). The mean inflammation score was higher in the regressive group (supplemental table S1). ANOVA analysis of normalized gene expression levels revealed 395 genes that were differentially expressed between persistent and regressive groups. The subset (N=169) of genes that were associated with altered cellular pathway activity were selected to represent those most strongly correlated with persistence and supervised hierarchical clustering was performed to further confirm the similarity of the groups combined to form the final persistent and regressive two group comparison. Only 4 cases in the combined persistent (n=32) and 3 in the combined regressive (n=31) groups showed uncharacteristic expression profiles (supplemental figure S1). Compared to regressive sites, 195 of the full genelist were upregulated and 200 downregulated in persistent sites. Additionally, levels of gene expression related to the baseline HS regardless of whether sites showed persistence or regression were assessed. Genes were assessed for the presence of an upward or downward trend in expression with increasing HS. Pearson correlation revealed 797 genes that showed a correlation coefficient of ≥ +/− 0.5 across HS 1–7. Forty percent of the differentially expressed genes associated with persistence were also associated with baseline histology (figure 1C).

The genelists were utilized to identify cellular processes that are associated with persistence and degree of histologic atypia via characterization of activity using the Ingenuity Pathway Analysis program (IPA, Ingenuity, Qiagen, Germantown, MD, USA). Thirty-one of 232 pathways with which the genes could be associated showed statistically significant evidence of altered activity in persistent sites (figure 2A, supplemental table 4). These pathways mostly represented three general categories of cellular activity. Fourteen of these pathways are directly associated with immune responses and predominantly show down-regulation of the genes associated with these activities. Eight of the pathways mediate cell cycle regulation and proliferative activities that are commonly associated with malignant disease. Five pathways are associated with regulation of tissue development encompassing a number of genes that mediate epithelial tissue differentiation, cell-cell interactions and cytoskeleton related signaling. The cell cycle and tissue development groups show more frequent up-regulation of pathway activity. Pathway analysis utilizing the list of genes associated with baseline histology show that 44 of 261 pathways demonstrate statistically significant evidence of altered activity. Eighteen of the top 22 pathways and 26 overall represent proliferation associated pathways that are commonly related to cancer development (supplemental table 5). Cell cycle regulation, frequently via pathways coordinating cell cycle progression with DNA damage responses, is strongly represented. Relatively few pathways (three each) are associated with immune responses and tissue differentiation. Several pathways representing altered metabolic activity and response to cellular injury are correlated with histologic score. All but four of the histology related pathways are associated with predominantly up-regulated genes. Additionally, the persistence and baseline histology gene lists were used for IPA upstream regulator analysis in which groups of genes whose expression are known to be regulated by certain signaling proteins are identified and used to indicate significant increases or decreases in signaling by the related upstream regulatory protein. The upstream regulators with the highest positive activation z-scores to be associated with baseline histology were vascular endothelial growth factor (VEGF) and beta-estradiol (supplemental table S6A). VEGF was also the highest to be associated with persistence (supplemental table S6B).

Figure 2.

Pathway and network analyses showing major cellular processes associated with persistence of BD. A) Pathway analysis showing top 20 pathways with significant alteration of activity in persistent versus regressive BD. Red indicates that the majority of the persistence related genes associated with a given pathway are up-regulated, green indicates predominant down-regulation. Lines connecting pathways indicate that genes are present in both pathways. Intensity of color indicates degree of significance as described in color bar. B) Inflammation associated, differentially expressed genes are predominantly down-regulated and are strongly associated with T-helper lymphocyte and dendritic cell networks. C) The majority of cell cycle progression network associated genes are up-regulated and PLK1 shows the most numerous regulatory interactions. D) The majority of morphology of epidermis network associated genes are up-regulated and include a number of keratins and desmosomal (desmoglein, desmocollins, plakins) genes.

Network analyses provided a broader method by which interactions between genes that are differentially expressed in comparisons of persistent and regressive BD could be assessed. These analyses revealed networks significantly associated with persistence of BD that were reviewed to further focus the cellular processes we chose to study in the validational analyses presented below. Inflammatory networks associated with persistence were generally down-regulated and showed a strong association with helper T-lymphocyte differentiation and macrophage/dendritic cell function networks (figure 2B). The cell cycle progression network included genes that consistently demonstrated up-regulation with persistence and identified polo-like kinase 1 (PLK1) as a central mediator of these activities (figure 2C). The morphology of epidermis network demonstrated that most differentially expressed genes were up-regulated and identified a number of genes associated with squamous differentiation and desmosomal functions (figure 2D).

Polo-Like Kinase 1 (PLK1) expression and activity are increased in persistent as compared to regressive bronchial dysplasia

PLK1 is the central mediator of activity in three of the top 10 pathways most significantly associated with both persistence and baseline histology. To validate this relationship, PLK1 expression was correlated with increased H-score on follow-up biopsy in the Affymetrix gene expression data set (figure 3A), and high expression in persistent BD was also demonstrated in situ by dual immunofluorescence (figure 3B). Expression levels of PLK1 were further assessed in primary cultures of bronchial epithelial cells derived from fresh biopsy tissue. Quantitative real-time PCR based expression levels were normalized to expression levels of PLK1 in a primary cell line derived from a site showing normal histology over multiple timepoints. Triplicate measurements in each of two independent studies of PLK1 mRNA levels demonstrated higher mean levels of PLK1 expression in a group of cultures derived from eight persistent BD sites as compared to measurements from six cultures of regressive BD. Results from a representative analysis are shown in figure 3C where the mean normalized level of PLK1 expression for persistent BD was more than twice that of regressive BD (2.33 vs. 0.88, p=0.002). The majority of the persistent BD derived cultures displayed PLK1 RNA levels that were above the mean for regressive bronchial dysplasia. PLK1 protein expression was also found to be elevated in cultured persistent BD versus cultured regressive BD by western blot analysis (figure 3D).

Figure 3.

PLK1 overexpression is associated with persistence of BD. A) Biopsy sites with PLK1 overexpression (PLK1 OE) show higher histology scores in follow-up biopsies than those with underexpression (PLK1 UE). Baseline histologic scores were not significantly different between OE and UE groups (PLK1 OE=5.14, PLK1 UE=4.75; p=0.11). B) PLK1/Pan-keratin (Pan-Ker) dual immunofluorescence showing representative frequent moderate (i.e. white arrows) and strong (i.e. red arrow) nuclear and cytoplasmic positivity for PLK1 in persistent BD (PBD) that is not seen in regressive BD (RBD). Magnification 600X. C) Cultures of persistent bronchial dysplasia (BD) show higher PLK1 expression than regressive BD. Diamonds represent the mean of triplicate measurements of a single cell line. Horizontal bars indicate mean PLK1 level for the entire group. Vertical bars indicate standard error of mean (based on all included replicates). D) Western blot of PLK1 on protein lysates from normal, RBD and PBD derived cultured cells. By densitometry of duplicates from each case, PBD shows increased PLK1 levels vs. RBD (normalized mean 1.93 vs. 0.99, p=0.02). Student’s T-test, *p<0.01

Squamous cell carcinoma (SCC) of the lung demonstrates one of the highest rates of mutational damage of all malignancies (20). We therefore hypothesized that the role of PLK1 activity in mediating abrogation of cell cycle arrest in the presence of DNA damage at the G2-M checkpoint could represent a key cellular alteration associated with progression of BD to invasive SCC. Additionally, effective inhibitors of PLK1 are commercially available and FDA approved for treatment of some malignancies. In vitro analyses of the role of PLK1 in mediating cellular proliferation and evasion of apoptotic activity were undertaken to confirm increased PLK1 mediated biological activity in persistent BD. Treatment with 100nM of the PLK1 inhibitor volasertib for three days led to significant arrest of persistent BD derived cells in S/G2 phase without an impact on cell cycle progression in normal or regressive BD (figure 4A and B). Similarly, four days of treatment with volasertib induced apoptotic activity as measured by caspase 3/7 activity in persistent BD and to a lesser degree in regressive BD while not having an effect in cultures of normal bronchial epithelium (figure 4C). Volasertib also induced a significant decrease in proliferation in persistent BD, but did not have a significant effect on proliferation in regressive BD or normal cultures (figure 4C).

Figure 4.

PLK1 inhibition inhibits cell cycle progression and induces apoptosis of persistent BD (PBD). A) Flow cytometric analysis of cell cycle fraction in PLK1 inhibitor, volasertib (PLK1i), treated cultured normal bronchial epithelial (Nl BE) and PBD derived cells showing an increase in the fraction of cells in the S (diagonal lines) and G2 (gray peak to the right) phase of the cell cycle in PBD but not normal BE or regressive BD (RBD) derived cells. B) The proportion of PBD cells arrested at the G2/M checkpoint is > 4-fold higher in PLK1 treated versus vehicle (DMSO) alone treated cells and significantly greater than that seen in normal and RBD cells (three replicates each of one normal, two RBD and one PBD cell lines). C) Apoptotic activity and inhibition of proliferation are increased in PBD treated with 100 nM but not in normal BE or RBD (RBD), although a lesser degree of induction of apoptosis is noted in RBD. Caspase activity is normalized to that measured in vehicle only treated parallel cultures (sixteen replicates each of one normal, two RBD and two PBD cell lines). Cell numbers in analyses of proliferation are expressed as ratios compared to initial seeding of 100,000 cells for each cell line/condition (three replicates each). Student’s T-test, *p < 0.05.

Inflammatory responses are down-regulated in persistent bronchial dysplasia

Persistence associated inflammatory pathways indicate significant down-regulation in expression of genes associated with macrophage and T-lymphocyte (T-LC) activity in persistent as compared to regressive BD. Given these observations, inflammation scores that are generated for all biopsies at the time of initial pathological characterization were used to determine if there is a difference in overall levels of inflammation related to persistence/regression. Figure 5A shows that the presence of an inflammatory infiltrate was more frequent at sites that showed regressive, non-dysplastic histology than those with persistent, dysplastic histology in follow-up biopsies. To determine if these gene expression changes correlated with actual levels of inflammatory cell types in tissue, numbers of macrophages and T-LCs were assessed in an independent set of 46 persistent and 39 regressive dysplasias by immunohistochemical (IHC) stains for CD68, CD3, CD4 and CD8. The persistent and regressive BD groups were selected to be matched in respect to inflammation score (mean scores 0.88 vs. 0.83 respectively , p=0.77) and proportion of cases with high grade inflammation (inflammation scores 2–3; 16.3% vs 23.1% respectively, p=0.44), as well as age, gender, tobacco status, pack year smoking history and history of lung cancer (supplemental table S7). Image analysis showed a near significant trend toward decreased overall CD68 positivity per area in persistent versus regressive BD and similar lower magnitude nonsignificant decreases in CD3, CD4 and CD8 (Figure 5B and C). Though no sub-analyses were statistically significant, the same trends were seen when analyses were restricted to only cases with low or high inflammation scores or restricted to the stromal compartment. No changes were observed in the analysis of the epithelial compartment (supplemental figure S2A and B). The gene expression data suggest that alterations in the polarization of inflammatory cell types are key characteristics that distinguish persistent and regressive BD. Direct assessment of the array based gene expression levels for a variety of genes that have been reported in the literature to be associated with specific polarization states of inflammatory cell subsets were evaluated (21–28). The majority of genes with statistically significant or near significant differences when comparing persistent to regressive BD showed decreases in M1 macrophage and T-helper 1 T-lymphocyte markers, while 5 of 9 regulatory T lymphocyte markers were increased in persistent BD (supplemental figure S2C). To investigate these potential differences, dual immunofluorescence (IF) stains to demonstrate subsets of inflammatory cells were used in a small set of representative dysplasias. Figure 5D shows an abundance of dual CD68/HLA-DRA positive M1 macrophages in regressive BD compared to persistent BD. In addition, HLA-DRA expression is noted in epithelial cells of regressive BD, but not that of the persistent BD. Conversely, regulatory T-lymphocytes represent a larger proportion of T-LCs in representative persistent as compared to regressive BD (Figure 5D).

Figure 5.

A) H&E based mild to severe inflammation scores are more frequent in regressive than persistent BD. B) Macrophage and T-lymphocyte markers are more strongly expressed in regressive than persistent BD, though only macrophage marker CD68 shows a trend toward statistical significance (C - representative CD68 immunostains). D) Dual immunofluorescence stains of representative persistent and regressive BD. While the majority of macrophages lack reactivity for M1 marker HLADRA (red arrows, upper middle), in regressive BD nearly all show dual positivity for CD68 and HLADRA (see inset, bottom middle) and additionally, membranous HLADRA staining is seen in epithelial cells (green arrows). Regulatory T-lymphocytes co-expressing CD4 and FoxP3 (Tregs, green arrows, upper and lower right panels) are more abundant in persistent as compared to regressive BD (see inset, upper right). Overall, the numbers of M1 macrophages and Tregs in these biopsies were 63.3% (19 dual positive/30 CD68 positive) and 33.7% (4 dual positive/57 CD4 positive) for persistent BD versus 73.1% (19/26) and 7.0% (28/83) for regressive BD, respectively. Student’s T-test: *p<0.05; **p=0.05–0.15.

Altered Expression of Tissue Differentiation and Cell Adhesion Molecules Distinguish Persistent and Regressive Bronchial Dysplasia

Several pathways associated with persistence are related to cell-cell interactions that contribute to epithelial morphology and nuclear signaling stimulated by physical interactions occurring on the external surface of the cell. These Ingenuity pathways are correlated with gene ontology (GO) related cellular activities regulating organismal development emphasizing the potential role of altered gene expression in mediating altered epithelial morphology as demonstrated histologically by the metaplastic change from glandular to squamous epithelium in BD. Immunofluorescent staining showed that both desmoglein 3 (DSG3) and junctional plakoglobin (PG) demonstrated predominantly membranous staining patterns (figure 6A and B). Immunofluorescence scores (I-scores) based on combined membranous and cytoplasmic staining indicated that high grade BD (HGD) demonstrated significantly higher expression of DSG3 than low grade BD (LGD) and normal bronchial epithelium while PG showed a trend toward higher expression in LGD as compared to normal bronchial epithelium (figure 6C). Persistence of BD was significantly associated with high expression of DSG3 and showed a trend toward association with PG expression using I-score cutoffs of 50 and 100, respectively (figure 6D).

Figure 6.

Expression of desmosomal components is increased in persistent BD. A) Immunofluorescence shows low desmoglein 3 (DSG3) expression in stable reserve cell hyperplasia (left) with increased expression in high grade regressive BD and strongest expression in persistent high grade BD (right). B) A similar pattern of junctional plakoglobin (PG) expression is noted in stable reserve cell hyperplasia (left) with more pronounced increased expression in high grade regressive BD and strongest expression in persistent high grade BD (right). Positive controls = invasive skin SCC (DSG3) and lung SCC (PG) and negative controls = persistent HGDs with secondary Ab only. C) Increased DSG3 expression is seen as histology progresses from normal bronchial epithelium (Nl BE, n=6) to low grade bronchial dysplasia (LGD, n=4) and high grade dysplasia (HGD, n=8). There is a trend toward increased expression of PG in LGD vs normal BE. Bars represent standard error of mean. D) More frequent overexpression of DSG3 in persistent (n=8) versus regressive BD (n=12 [10 for PG]) using an immunofluorescence score (I-score) cut-off of 50 and a trend toward overexpression of PG in persistent BD using a cut-off score of 100. Magnification = 200X. Student’s T-test (C) and Chi-square (D): *p<0.05, **p=0.5–0.15.

DISCUSSION

Premalignant airway disease precedes the development of invasive SCC, and previously described airway-wide, smoking-related gene expression alterations have recently been shown to predict the presence of dysplastic lesions (29,30). However, the rate of progression to SCC of these lesions is low with even the highest grades only progressing in 5–10% of cases (16, 17). Gene expression changes that distinguish high-risk from low-risk BD have not been previously reported. Herein we describe gene expression changes associated with the persistent BDs that are characteristic lesions of aggressive airway disease, a condition associated with a 7.8-fold increase in risk for invasive SCC (17). The persistence associated gene expression profile provides potential markers of risk and indicates alterations in cellular activity that may represent effective targets for prevention of progression to invasive cancer.

Cell cycle control is prominent in both persistence and dysplastic grade associated pathways. While many of the pathways up-regulated in association with increasing histologic grade represent DNA damage repair pathways, these pathways are not represented in the comparisons of persistent versus regressive BD. PLK1 promotes cell cycle progression and also plays a role in regulating the effect of DNA damage on this progression. PLK1 is required for progression through the M-phase of the cell cycle (rev in 31). Additionally, PLK1 plays a key role in regulating progression through the G2-M cell cycle checkpoint. PLK1 activity can override G2-M checkpoint arrest in the presence of DNA damage (32–34). In the presence of significant DNA damage a number of cellular processes will arrest cell cycle mediated proliferation at this checkpoint and induce apoptosis (rev. in 35). Our analyses demonstrate an association of PLK1 overexpression with persistence of BD and show that treatment with PLK1 inhibitor (Volasertib, BI6727) arrests persistent dysplastic cells in the S and G2 phases of the cell cycle while inducing a significant increase in apoptotic activity. These results confirm increased PLK1 activity in persistent BD and suggest that overexpression may promote genomic instability leading to the accumulation of somatic genetic damage potentially associated with the development of malignant behavior. The demonstration that the activity of DNA damage repair pathways increases with increasing histologic grade, but that increased activity of these pathways are not specifically associated with high risk persistent behavior, may indicate that while ongoing DNA damage is common in the airways, it is the control of cell cycle progression that is critical to malignant progression. Indeed, a gene expression study of BD from four cases of resected invasive SCC showed that one of the characteristics that distinguished the premalignant lesions from normal bronchial epithelium was cell cycle M phase activation and modulation of microtubule dynamics (36), both processes that PLK1 influences in its role of promoting cell cycle progression. Thus, our results suggest that PLK1 overexpression could represent a marker of risk for progression and that inhibition of PLK1 may prevent progression of BD to invasive lung cancer.

Altered activity of inflammatory pathways related to persistence of BD suggest that influences of the microenvironment play a role in the progression of premalignant airway disease. The data implicates certain specific aspects of broader cellular processes in progression such as the more frequent identification of alterations in cellular than humoral immune responses and the frequent identification of antigen presentation and T-lymphocyte activity in this subgroup of cellular pathways associated with persistence. Intriguingly, the immune response pathways associated with persistence almost exclusively show down-regulation of the component genes. Our data show that the presence of an inflammatory infiltrate is more common in regressive than persistent BD. In lung cancer, a strong general immune infiltrate has been associated with variable but often poor prognosis, although recent studies have clearly established that the if the composition of immune cells is oriented toward active responses such infiltrates can be associated with improved prognosis (37,38). Potential benefit has been shown when numbers of tumor associated CD4 and CD8 positive T-LCs or M1 macrophages are increased and adverse effects found when M2 type macrophages or regulatory T-LCs are increased in invasive lung cancers (39,40). Our analyses suggest that regressive BD may be associated with decreased regulatory T-LCs and increased antigen presentation capabilities via increased epithelial HLA-DRA expression and the presence of M1 type macrophages which demonstrate robust antigen presentation and are associated with strong anti-tumor responses (41). If results from additional studies confirm alterations such as those identified in our pathway analyses, it may suggest that modulation of immune infiltrates could be a successful approach to prevent progression to lung cancer.

Increased expression of DSG3 and PG, proteins associated with desmosomes, indicate alterations of cell-cell interactions and epithelial differentiation in persistent BD. The structural roles of these factors suggest that increased expression could contribute to squamous metaplastic change that may be more prominent in persistent disease and thus represent potential markers of high risk BD. DSG3 is a key component of desmosomes which sequesters pro-oncogenic desmosomal components associated with epithelial to mesenchymal transition, but has recently also been associated with non-desmosomal cytoplasmic activation of pro-tumorigenic SRC signaling (42). Indeed, overexpression of DSG3 has been described in lung cancer and has been associated with poor prognosis in esophageal SCC and pancreatic adenocarcinoma (43–45). PG, a member of the catenin family of proteins, also demonstrates dual activity involving translocation to the nucleus and interaction with transcription factors, although its transcriptional effects have been associated with anti-oncogenic activities such as inhibition of proliferation, migration and invasion (46–49). While some of the IF stains did show differences in the distribution of DSG3 and PG (i.e. membranous versus cytoplasmic/nuclear), the number of cases studied precluded definitive demonstration of differences related to histologic grade or persistence. Further analysis of the factors associated with both up- and down-regulation of the cell-cell interaction pathways identified in our gene expression analysis will be required to determine if alterations in expression of these factors may directly promote pro-tumorigenic activities that could be targeted for prevention.

The gene expression data also show evidence that increased signaling via VEGF and estrogen distinguishes persistent from regressive bronchial dysplasia. Estrogen receptor expression has been associated with poor prognosis in NSCLC indicating that estrogen signaling contributes to tumor progression (50). Furthermore, administration of estrogen inhibitors in female mice reduced tumor formation providing evidence that estrogen activity may contribute to the development of lung cancer (51). We have previously shown that VEGF and VEGFR2 expression as well as microvessel densities are increased in bronchial dysplasia and invasive lung cancer, especially early stage tumors, as compared to normal or reactive bronchial epithelium (52). Furthermore, VEGF production by dysplastic epithelial cells induced endothelial cell proliferation, migration and tubule elongation (53). A role for VEGF signaling in promoting persistence and progression has also been suggested by observations that BD showing histologic features of increased angiogenesis (angiogenic squamous dysplasia) demonstrate more frequent persistence and that treatment of mice with vandetanib, a strong inhibitor of VEGFR2, reduced the number and size of urethane induced tumors (17,54).

The use of whole biopsy tissue to generate RNA for these studies could represent a drawback as some persistence related gene expression changes might be masked by RNA derived from non-epithelial cells. Additionally, although an extensive set of bronchoscopy derived specimens are available in the Colorado SPORE tissue bank, the availability of frozen tissue from sites with informative follow-up sampling was limited and thus might also have contributed to false negatives regarding changes in the expression of certain genes. One other, recently published study has employed gene expression to assess alterations associated with degree of histologic atypia in precursor lesions of squamous cell carcinoma of the lung (55). Unlike our analysis, this study included gene expression from invasive squamous cell carcinomas as part of the spectrum of progression. Nonetheless, similar pathways associated with cell cycle control and DNA damage repair were also identified lending support to the validity of the findings in both studies.

To our knowledge, this is the first analysis to describe gene expression alterations that distinguish high risk from low risk premalignant squamous cell carcinoma precursor lesions. In summary, the identification of a large number of genes that are differentially expressed between histologically indistinguishable lesions lends strong support to the higher risk associated with persistent BD. Altered activity of pathways associated with these differentially expressed genes show the importance of cell cycle regulation, immune infiltrates and cell-cell interactions in establishing persistent behavior. These findings suggest that markers of these pathways may be useful in predicting which BDs may persist and therefore represent high risk for progression, and also, facilitate identification of certain cellular activities that might be successfully targeted to prevent the development of invasive cancer.

STATEMENT OF SIGNIFICANCE.

Gene expression profiling of high-risk persistent bronchial dysplasia reveals changes in cell cycle control, inflammatory activity, and epithelial differentiation/cell-cell adhesion that may underlie progression to invasive SCC.