This study aimed for a better understanding of the genetic events associated with dysplasia in a genetic model of lung cancer induced by overexpression of the c-Raf-1 kinase.
We aimed to determine the regulatory gene networks associated with dysplasia. This enabled identification of candidate genes specifically regulated at the edge of malignant transformation. Based on hierarchical gene cluster analysis and principal-component analysis we were able to distinguish the dysplastic cells from transgenic and non-transgenic lung tissue.
To further validate the microarray data, we employed qRT-PCR of selected genes, and confirmed by immunohistochemistry regulated proteins in dysplastic foci.
Notably, RT-PCR data is suggestive for the microarray to underestimate changes in the gene expression even though both methods supported the general directions of changes. An important finding of our study was that about 10% of the differentially expressed genes regulated by ≥2 fold are already known to be associated with lung cancer.
Specifically, in dysplasia genes coding for nucleic acid metabolism or cell cycle regulation were unchanged but two EGF-ligands, namely amphi- and epiregulin were highly significantly up-regulated. These ligands enable autocrine loops to foster undue EGF-signaling. Notably, Areg (amphiregulin) is a 252-amino acid transmembrane glycoprotein and consist of two major soluble forms of 78 and 84 amino acids, respectively. Areg was originally isolated from conditioned medium of the human breast carcinoma cell line MCF-7 and was found to be a heparin-binding growth factor. Areg promotes neoplastic growth in mammary epithelial cells, fibroblasts, and keratinocytes 
and was shown to function in an autocrine manner to drive the proliferation of malignantly transformed cells of colon, breast, cervix, prostate, and pancreas 
. This ligand of the EGF tyrosine kinase is commonly over expressed in cancers of human colon, stomach, breast, and pancreas, in which the level of Areg correlates with tumor progression and poor patient survival 
. Notably, it was shown in patients with advanced non-small cell lung cancer that the survival time of amphiregulin-negative patients treated with gefitinib was significantly longer than that of amphiregulin-positive patients. Therefore, an increase in serum amphiregulin can be viewed as a predictor of resistance to gefitinib 
. Moreover, Areg was also reported as an inhibitor of apoptosis in non-small cell lung cancer cell lines 
. Our data indicate that Areg secreted from dysplastic cells are in an autocrine loop to foster malignant transformation.
A further growth factor strongly induced in dysplasia is Ereg (epiregulin). The gene codes for a transmembrane precursor before being proteolytically cleaved to release a 46-amino-acid activated protein 
. Ereg is promiscuous, binds and activates the Egfr family member Erbb4 via heterodimeric interactions with Erbb2 
. Although Ereg expression is highly correlated with survival from bladder cancer 
, others suggest epiregulin to be important for pancreatic and prostate cancer development 
. The strongest evidence for a role of epiregulin during tumorigenesis was obtained in Ki-ras-mediated signaling of colon cancer cells 
The fact that transformed cells grow faster than unaltered cells is also sustained by our data with regard to the number and strength of up-regulated genes affecting cellular growth. Specifically, the cell surface membrane proteins play an important role in the behaviour of cells to allow for communication with other cells, cell movement and migration, adherence to other cells or structures and recognition by the immune system. Alterations of the plasma membrane in malignant cells may thus be inferred from a variety of properties that characterize their growth and behaviour.
Our study is suggestive for changes in the glycosylation patterns and in cell surface glycolipids. Changes in glycosylation can include the presence of new carbohydrate structures that are not detected in the normal epithelial cells and/or short carbohydrate chains usually masked by larger epitopes. For example, we found St8sia6 to be up-regulated in dysplasia. St8sia6 (alpha-2,8-sialyltransferase VI) belongs to a family of sialyltransferases that synthesize sialylglycoconjugates. The most frequently described aberrant glycosylation in cancer cells include the synthesis of highly branched and heavily sialyted glycans 
. Because of the negative electric charge of sialic acid at the terminal non-reducing end of glycoprotein oligosaccharides, it plays a key role in mediating biological recognition events, including those responsible for metastasis. Next, we found an overexpression of B4galt6 (UDP-GAL:Beta-GlcNAc Beta-1,4-Galactosyltransferase, polypeptide 6). This enzyme is involved in the production of GlcNAc (galactose beta-1,4-N-acetylglucosamine). GlcNAc is the first sugar residue to be linked to serine or threonine and serves as acceptor for the elongation with further GlcNAc molecules. The resulting branched structure may be elongated by the sequential addition of galactose, fucose and sialic acid 
Changes of the glycosylation pattern in dysplasia enables further variability in molecular assembly of membranes and offers a wide range of flexibility in response to cellular environment. Their detection on the cancer cell surface may provide useful diagnostic or prognostic information but until now is incompletely understood and therefore requires further elucidation.
Changes in the expression pattern of cell-adhesion molecules and components of intercellular junctions are related to loss of epithelial organization, proper cell layer and tissue polarity. With regard to adhesion molecules, we found up-regulation of Chl1 (cell adhesion molecule with homology to L1CAM), a member of the L1 gene family of neural cell adhesion molecules, and of several claudins as exemplified by the up-regulation of Cldn2 (claudin2), Cldn4 (claudin4), and Cldn8 (claudin8), which are important components of the tight junctions. Adhesion properties greatly impacts cell-to-cell interaction and growth of cancer cells.
In this regard, the cytoskeleton, which represents a complex of interconnected fibrillar elements, has been determined as an important factor in mediating adhesion-independent and dependent signalling. During morphogenesis, they determine cell shape and polarity, and promote stable cell-cell and cell-matrix adhesions through their interactions with cadherins and integrins, respectively. Surprisingly, in dysplastic cells we did not identify regulation of genes involved in the cytoskeleton.
Instead, we identified changes in cell to cell communication. In our study Gjb3 (gap junction membrane channel protein beta 3) and Gjb4 (gap junction membrane channel protein beta 4) were significantly up-regulated. It was shown that the gap junctional intercellular communication is a form of cell–to-cell signalling thereby mediating the exchange of small molecules between neighbouring cells 
. The regulation of connexins and gap junctions is a hallmark of carcinogenesis, while their induction in cancer cells leads to reversal of the cancer phenotype, induction of differentiation, and regulation of cell growth 
We also focused on genes coding for surfactant lipids in respiratory epithelium. Surfactant is a complex mixture of lipids and proteins that reduces surface tension at the air–liquid interface and prevents alveolar collapse during respiration. Four surfactant proteins (SP) with unique properties have been identified. SP-A and SP-D are relatively hydrophilic proteins and contribute to innate defence of the lung and surfactant homeostasis. SP-B and SP-C are hydrophobic proteins that enhance surface-active properties of surfactant phospholipid films 
. In dysplastic epithelium we could not detect regulated genes coding for surfactant proteins. However, nearly all of the genes grouped under lipid metabolism that have been previously linked to transport and secretion of lipid were regulated. For example, Apoa1 (apolipoprotein A-I) is known to be the major protein moiety of high-density lipoprotein (HDL) and is mainly produced in the liver and the intestine but is also produced in vitro by some differentiated cell lines established from human colorectal tumors 
We found Apoa1 (apolipoprotein A-1), Adcyap1 (adenylate cyclase activating polypeptide 1), Ltb4dh (leukotriene B4 12-hydroxydehydrogenase), Fst (follistatin), Inhbb (inhibin beta-B), Clu (clusterin), Hnf4α (hepatic nuclear factor 4, alpha), Prokr1 (prokineticin receptor 1), Pla2g1b (phospholipase A2, group IB), Sult2b1 (sulfotransferase family, cytosolic, 2B, member 1), St8sia6 (ST8 alpha-N-acetyl-neuraminide alpha-2,8-sialyltransferase 6), Cyp1b1 (cytochrome P450, family 1, subfamily b, polypeptide 1) and the Pthlh (parathyroid hormone-like peptide) to be regulated with inferred roles in lipid metabolism.
For example, Adcyap1 (adenylate cyclase activating polypeptide 1) is a member of the secretin/glucagons/vasoactive intestinal peptide (VIP) family of peptides. It has been localized by immunohistochemistry to the central nervous system, digestive tract and was shown to exhibit a variety of biological activities. It is involved in synthesis of sulfatides, glycolipid, sphingolipid and plays a role in the accumulation of estrogen and progesterone. It was also reported that Adcyap1 can regulate the proliferation and differentiation 
and was shown to be overexpressed in neuroblastoma 
and breast cancer 
. Likewise, Ltb4dh (leukotriene B4 12-hydroxydehydrogenase) catalyzes the conversation of leukotriene B(4) into 12-oxo-leukotriene B(4) and is involved in reduction of 15-keto prostaglandin E1 and prostaglandin. It could be shown that Ltb4dh was overexpressed in Ta urothelial cell carcinoma 
. We also observed Pthlh (parathyroid hormone-like peptide) to be induced in dysplasia. This protein is responsible for most cases of humoral hypercalcemia of malignancy. Pthlh expression in tumor samples has also been correlated with poor prognosis in breast cancer 
, renal carcinoma 
and colorectal tumors 
Because cellular growth and differentiation is depended on highly co-ordinated transcriptional networks we also searched for master regulatory proteins of respiratory epithelium. Notably, morphogenesis and branching of respiratory epithelium depends on the timely expression and activity of transcription factors. Dynamic changes in transcriptional activation of lung-specific genes are required to get an appropriate positioning of respiratory epithelial cells with the mesenchyme-derived endothelial cells. If this critical process is disturbed, this can lead to malignancy 
. Transcription factors involved in differentiation of pulmonary epithelium includes HNF3B (hepatocyte nuclear factor 3 beta also termed Foxa2) 
, Ttf1 (transcription termination factor, RNA polymerase I) 
, Foxf1 (forkhead box F1a) 
, Gata6 (GATA binding protein 6) 
and Gata5 (GATA binding protein 5) 
In our study some of these transcription factors were highly significantly regulated in dysplasia i.e. Hnf4α, Foxa3 and Foxp2. Notably, Hnf4α (hepatocyte nuclear factor 4 alpha), a member of the steroid/thyroid hormone receptor superfamily is a transcriptional activator 
whose ligand has been identified as tightly bound endogenous fatty acids 
. We found Hnf4α to be highly significantly induced (12.8-fold). Hnf4α regulates constitutive expression of a large number of target genes encoding enzymes, transporters and other nuclear receptors.
Among the genes regulated by HNF4α we found at least 11 target genes up-regulated in dysplasia i.e. Arg2 (arginase type 2), Apoa1 (apoliporotein A1), Cyp1b1 (cytochrome P450, family 1, subfamily B, polypeptide 1), Cldn2 (claudin 2), Fetub (fetuin beta), Gpx2 (glutathione peroxidase 2), Gsta4 (glutathione S-transferase A4), Gpc6 (glypican 6), Hpn (hepsin), Orm1 (orosomucoid 1) and Lad1 (ladinin1) 
, therefore providing an important link between induction of this transcription factor and up-regulation of target genes.
Likewise, we found Foxa3 (forkhead box A3) to be induced 7.5-fold. This protein belongs to a group of endoderm-related developmental factors that are members of the forkhead box (Fox) superfamily of transcription factors. They were first discovered by their ability to bind to promoters of liver specific genes encoding α1-antitrypsin and transthyretin 
. Foxa genes are regulated in early mouse embryo development 
and are responsible, at least in part, for metabolic regulation. Mice that lack Foxa3 display an increase in the mRNA levels of various serum proteins and glycolytic enzymes and show a low serum glucagon level 
We also found Foxp2 (forkhead box P2) to be up-regulated 4.5-fold. This protein is a member of the new subfamily of winged-helix/forkhead DNA binding domain transcription factor. It could be shown that Foxp2 are expressed in the lung restricted to the distal epithelium and may regulate lung epithelial-specific gene transcription during embryonic development 
. Foxp2 has been characterized as a transcriptional repressor. All the known Fox genes that were implicated in regulating lung expression were characterized as transcriptional activators 
. This finding suggests that Foxp2 plays a role in the balance of transcriptional activation and repression that is involved in regulating epithelial cell identity and development of the lung. These processes are also crucial for the transdifferentiation of alveolar type I to type II epithelial cells, which are responsible for gas exchange and surfactant protein expression essential for lung function.
Additionally, we have confirmed the up-regulation of Areg, Ereg, Hnf4α, Foxa3 and Foxp2 by immunohistochemical staining. The immunostaining showed a distinct pattern of immunoreactivity in dysplastic cells with no staining in unaltered transgenic cells. For example, immunohistochemical staining of Areg showed a cytoplasmic immunoreactivity in dysplastic cells but not each cell expressed this molecule equally strong.
In dysplasia, however, we could not evidence altered transcriptional networks of the MAPK pathway and we did not observe regulated genes coding for cell cycle or nucleotide metabolism. Both processes are highly active in tumor cells but do not play a big role in dysplastic cells prior to malignant transformation. In this context we found two EGF ligands, Areg and Ereg to be specifically up-regulated in dysplasia. While amphiregulin binds exclusively to EGFR epiregulin binds to HER4 and mediates aberrant activation of this receptor. Indeed, exaggerated EGF tyrosine kinase activity is a probable cause for lung cancer. We now evidence de novo expression of two ligands of EGFR in dysplasia that are frequently over expressed in cancers. We did, however, not observe increased transcript expression of the cancer stem cell markers CD44, CD133 and the epithelial cell adhesion molecule EpCAM. This may support the notion that dysplasia is a facultative cancer where additional events need to occur that enable malignant transformation. Taken collectively, the whole genome expression data provided important information in the multistage process of lung cancer. Our study revealed interesting novel candidate genes and pathways that dissected the programme of respiratory epithelium from transgenic into dysplasia and eventually lung cancer. In the second part of our study we report the additional genetic events that take place from dysplasia to malignantly transformed cells and thus provide a molecular rational for the multistage process in lung cancer.