When tumor cells start to form nodules within the stroma, they need to communicate with the surrounding microenvironment, which is composed mainly by macrophages, fibroblasts/myofibroblasts, neutrophils, lymphocytes, and dendritic cells. To facilitate angiogenesis tumor cells can either directly release angiogenic factors such as VEGFs to directly stimulate the formation of new blood vessels, or tumor cells cooperate with macrophages, which can release angiogenic growth factors [6–8]. A good example for angiogenesis induced by tumor cells is the vascular variant of squamous cell dysplasia, whereas in atypical adenomatous hyperplasia (AAH) and in well-differentiated adenocarcinomas, angiogenesis seems to relay on cooperating macrophages [9–12] (Figs. 18.1a, b and 18.2a). To understand the function of macrophages, it is necessary to briefly discuss the two different populations of macrophages, the M1 and M2 type. M1 macrophages are acting against tumor cell invasion by secreting interleukin 12 (IL12) which function as tumoricidal by an interaction with cytotoxic lymphocytes and NK cells. M2 macrophages produce IL10, which promote tumor progression. The differentiation of naïve macrophages into either M1 or M2 types is facilitated by Notch, where low Notch via SOCS3 drives macrophages into M2 types [13]. M1 macrophages act pro-inflammatory, inactivate autophagy by production of radical oxygen species, and can also induce apoptosis of tumor cells [14–16]. Notably mutation and inactivation of Notch is found in neuroendocrine carcinomas, whereas activation in other non-small cell carcinomas, which questions the function of this gene as either oncogene or tumor suppressor [17–20]. Most probably different members of the Notch family proteins function differently in squamous cell, small cell, and adenocarcinomas, and in addition act differently during tumor development [21–23].

As the primary tumor grows, usually the formation of new blood vessels cannot keep with this resulting in hypoxia. This is the time when tumor cells are faced with this problem and try to escape apoptosis induced by hypoxia. Some of these mechanisms have been elucidated. HIF-1α is upregulated in areas of tumor hypoxia [24–28] and if translocated into the nucleus and bind to HIF-1β can induce transcription of VEGF, thus increasing the formation of more blood vessels. Apoptosis is also inhibited by growth factors such as IGF and EGF, which are also induced by hypoxia [24, 29]. Carcinoma cells also escape apoptosis and cell death in hypoxic areas by reducing their metabolism and cell division [30]. In mouse models of lung adenocarcinomas driven by the mutated RAS oncogene, invasion was exclusively seen starting in areas of necrosis and hypoxia [31] (Fig. 18.3a, b). This fits well with published data from human tumor research, showing that migration and EMT are increased in hypoxic areas by the release of different proteins [32–36]. If each of these enzymes/proteins act in concert together, or if each of these factors can act independently is presently unknown.

Macrophages also act together with fibroblasts and myofibroblasts to form either a stroma suitable for tumor cell invasion and migration (Fig. 18.2b) or might inhibit migration. This can easily be evaluated by morphology: in case the stroma cells form a classical scar, this means inhibition for the tumor cells to migrate, whereas desmoplastic stroma is a form of stroma remodeling done by myofibroblasts, which enable tumor cells to migrate (Fig. 18.4a, b). Usually fibroblasts in scars do not cooperate with tumor cells, myofibroblasts in contrary cooperate [37, 38]. In some cases even tumor cells undergoing epithelial to mesenchymal transition (EMT) directly form part of the desmoplastic stroma as in pleomorphic carcinomas (Fig. 18.5a, b) and occasionally also in SCLC [39]. Several studies have shown that tumor-associated “fibroblasts” (essentially myofibroblasts in the lung) are different from normal fibroblasts in the lung. They express different genes and proteins related to their function in cancer development. Specifically MLH1 was upregulated, whereas COX1, FGFR4, SMAD3, and p120 were downregulated [40]. Factors have been identified, which drive this differentiation of mesenchymal stem cells into myofibroblasts, namely, TGF-β and IL1β. TGF-β also induces the expression of α−SMA and FAP-α expression [41]. FAP (serine protease fibroblast activation protein) promote tumor growth in an endogenous mouse model of lung cancer driven by the K-rasG12D mutant. On the contrary FAP depletion inhibits tumor cell proliferation indirectly by increasing collagen accumulation, decreasing myofibroblasts in number, and decreasing blood vessel density in tumors [42]. Most importantly myofibroblasts also express metalloproteinases (MMP) such as MMP-2, MMP-9, MMP-8, and MMP-7. So these cells actively take part in remodeling of matrix proteins. In addition MMP-8 actively participates in the process of fibrocyte migration [43].

In this context also changes of matrix proteins, their composition, and their orientation are important for tumor cell migration. Usually the matrix is composed of several proteins such as different types of collagen (I, III, IV, V; predominant collagen I), fibronectin, laminin, elastin, and osteonectin. These proteins provide stability to the stroma by their oriented deposition and network formation by cross-linking. They also serve as orientation molecules providing ligands for migrating leukocytes expressing adhesion molecules. Migrating tumor cells also uses this mechanism. As a further benefit, adhesion of SCLC cells to fibronectin, laminin, and collagen IV through β1 integrins confers resistance to apoptosis induced by standard chemotherapeutic agents. Adhesion to ECM proteins stimulated protein tyrosine kinase (PTK) activity in both untreated and etoposide-treated cells [44]. In NSCLC cells of the tumor, stroma selectively synthesize osteonectin (SPARC, normally only in bronchial cartilages) in case of intratumoral hypoxia and acidity. Osteonectin proven by immunohistochemistry favors cancer cell invasion and migration [45].

Another important factor is the orientation and composition of matrix proteins: high amounts of elastin favor resistance for tumor cell migration, whereas high collagen and organized oriented deposition promote tumor cell migration. Conversely direct cell contacts between tumor cells and fibroblasts can also create migratory-inhibitory matrix composed of unorganized collagens (I, III, IV, and V) and proteoglycans (biglycan, fibromodulin, perlecan, and versican). Thick desmoplastic fiber bundles inhibit the migration and invasion of tumor cells [46]. Matrix protein deposition seems to be in addition regulated by a tumor suppressor gene, frequently lost in lung cancer, RBM5 (RNA binding motif protein 5, chromosome 3p21.3). The encoded protein plays a role in the induction of cell cycle arrest and apoptosis. Loss of RBM5 causes upregulation of Rac1, β-catenin, collagen, and laminin, which in turn increase cell movement. Consequently Rac1 and β-catenin correlate positively with lymph node metastasis in lung cancer patients [47]. Two other matrix proteins are less explored. Expression of periostin is associated with vimentin expression in the stroma or tumor epithelia and correlates with higher stage. The correlation of periostin expression with that of versican and collagen in advanced tumors was less obvious. Opposite to periostin expression of elastin was associated with less advanced disease [39]. However, this observation still needs confirmation as well as more in deep investigation for the function of these proteins.

So invasion, tumor growth, and tumor cell migration of lung cancer cells are regulated by many different factors, as cytokines, adhesion molecules and receptors, and genes acting either directly on tumor cells or cells of the microenvironment.

Usually tumor cells produce many modified proteins, which are recognized as foreign by dendritic cells and lymphocytes. Tumor cells are therefore attacked and destroyed by cytotoxic lymphocytes (CD8+). However, pulmonary carcinoma cells have developed different escape mechanisms to prevent this cytotoxic attack. By modulating the innate immune system, macrophages are preferentially forced to differentiate into M2 types as already explained. Another mechanism to protect tumor cell is to modify the pool of antigen-presenting dendritic cells (DC). Within dendritic cells, several functional quite opposite acting cell populations are discerned: conventional DC will present tumor antigens to T lymphocytes and force the production of cytotoxic T cells, whereas plasmacytoid and monocytoid DCs act as tumor protective cells. For example, bombesin derived from SCLC inhibits IL12 production by DC and their ability to activate T cells [48]. Tumor cells by secretion of TGF-β and prostaglandin E2 induce DCs to differentiate into regulatory DCs with a CD11c(low)CD11b(high) phenotype (also named plasmacytoid DC) and high expression of IL10, VEGF, and arginase I. These regulatory plasmacytoid DCs inhibited CD4+ T-cell proliferation [49] and thus act protumorigenic. Another action to prevent cytotoxic lymphocyte attack is to induce an influx of regulatory T cells (Treg). Treg downregulate the production and influx of cytotoxic T cells and NK cells and promote immune tolerance [50–52]. Finally also bone marrow-derived myeloid precursor cells (MDSC) may downregulate a T-cell-based immune reaction toward growing tumor cells by secreting arginase I [53]. Indoleamine 2,3-dioxygenase (IDO) and IL6 seem to play a regulatory role for these MDSC, as downregulation of IDO resulted in reduced lung tumor burden and improved survival in experimental settings. Loss of IDO resulted in an impairment of protumorigenic MDSC, whereas IL6 recovered both MDSC suppressor function and metastasis susceptibility. In addition vascular density was significantly reduced in Ido1-nullizygous mice [54].

In some carcinomas, preferentially in pulmonary squamous cell carcinomas, eosinophils are found in abundance. The role eosinophilic granulocytes play in NSCLC is not fully understood. It may be that variants of IL17 (IL17E) induce a helper two type of immune response, which in turn by the release of IL4 and IL5 causes tissue eosinophilia. In one study it was shown that IL17E has antitumor activity. Injections of recombinant IL17E resulted in significant antitumor activity. Combining IL17E with chemotherapy increased the antitumor efficacy in a xenograft model [55]. Eosinophils are directly acting cytotoxic against the tumor cells, for example, by releasing cytotoxic basic proteins, which was not explored in this study.

After having established the primary tumor and organized nutrition as well as protection for immune cell attacks, the tumor cells have to acquire changes to migrate to distant sites and establish metastasis. There are two different forms how tumor cells migrate: single cell or small cell cluster movement as it is seen in small cell carcinoma as well as undifferentiated NSCLC and movement by large clusters of organized cells such as in acinar adenocarcinoma or some cases of squamous cell carcinoma (Figs. 18.6a, b, 18.7a, b, and 18.8a, b). For single cell and small clusters, migration seems to be much easier since single cells can more easily adapt, for example, a spindle cell morphology, which enables better movement. Tumor cells during migration reduce or even abolish cytokeratin filaments and increase or de novo express α-actin and vimentin; this is commonly seen in pleomorphic carcinomas (Fig. 18.5b), carcinosarcomas, high-grade squamous cell and adenocarcinomas, and SCLC. Lung adenocarcinomas with high smooth muscle actin gene ACTA2 expression showed significantly enhanced distant metastasis and unfavorable prognosis. ACTA2 downregulation remarkably impaired in vitro migration, invasion, clonogenicity, and transendothelial penetration of adenocarcinoma cells without affecting proliferation. ACTA2 upregulation in lung adenocarcinoma cells was also connected to expression of c-MET and focal adhesion kinase (FAK), whereas ACTA2 targeting by siRNAs and shRNAs resulted in loss of mesenchymal characteristics [56]. Migration within the stroma requires several changes in tumor cells; one of these is formation of invadopodia. Tyrosine kinase substrate 5 (Tks5) is a scaffolding protein necessary for the formation of invadopodia. There are different isoforms, some of them (short isoforms) associated with reduced others (long isoforms) with increased metastasis [57–59]. Expression of Tks5 together with the expression of α-actin is further regulated by cortactin and neural Wiskott-Aldrich syndrome protein (N-WASP), which also regulate the expression of metalloprotease membrane type 1 matrix metalloprotease (MMP-14) [58]. However, migration of tumor cells seems to be regulated by different genes, so probably there is not a single mechanism for each tumor type, but more likely that tumor cells individually have adapted different mechanisms of migration protocols and used it during carcinogenesis. As an example myosin heavy chain 9 (MYH9) and copine III (CPNE3) positively correlate with the migration and invasion properties of lung cancer cell lines. If CPNE3 was knocked down, the metastatic abilities were inhibited in a mouse model. Also CPNE3 protein expression levels were positively correlated with the clinical stage in NSCLC [60]. In another study, nestin protein expression significantly correlated with tumor size and lymph node metastasis in NSCLC and also poor survival in patients with adenocarcinoma. Nestin inhibition by shRNA decreased proliferation, migration, invasion, and sphere formation in adenocarcinoma cells [61]. One of the major studied mechanisms of tumor cell migration is EMT, which again is seen in tumors with single cell or small cluster migration type (Fig. 18.9a–c).

When looking up studies on EMT, a huge amount of published article can be found in databases. The major surprise is that different genes are associated with EMT. And even when focusing on lung cancer studies, there are still different genes found to trigger EMT. One of the most often found EMT-associated genes are Twist, Snail, and TGFβ1.

Suppression of Twist expression in metastatic mammary carcinoma cells inhibits their ability to metastasize from the mammary gland to the lung. Ectopic expression of Twist resulted in loss of E-cadherin-mediated adhesion, activation of mesenchymal markers, and induction of cell motility, suggesting that Twist promotes EMT [62]. In another study, Twist was selectively associated with EGFR-mutated adenocarcinomas. Twist expressed in lung adenocarcinoma cell lines with EGFR mutation showed increased cell mobility. A decrease of EGFR pathway through EGF retrieval or inhibition of Twist expression by small RNA reversed the phenomenon. These findings supported that Twist promotes EMT in EGFR-mutated lung adenocarcinoma [63]. In the study by Pirozzi, the focus was on TGFβ1. They used two epithelial cell lines, which acquired a fibroblast-like appearance when treated by TGFβ1. By inhibiting TGFβ1, vimentin and CD90 were downregulated and cytokeratin, E-cadherin, and CD326 upregulated. TGFβ1 also upregulated Slug, Twist, and β-catenin, thus confirming its EMT property. Interestingly also some stem cell markers as Oct4, Nanog, Sox2, and CD133 were overexpressed too, linking EMT to tumor stem cells [64]. Adhesion plays a major role in EMT, therefore not surprisingly studies have focused on Wnt, Catenin, and GSK3β pathway. Loss of SARI (suppressor of AP-1, also called BATF2) expression initiates EMT, causing repression of E-cadherin and upregulation of vimentin in lung adenocarcinoma cell lines and in human lung adenocarcinomas. By knockdown endogenous SARI in a human lung xenograft mouse model, multiple lymph node metastases developed. SARI has been shown to regulate EMT by modulating the (GSK)-3β-β-catenin signaling pathway [65]. In the study by Blaukovitsch, another pathway for EMT was shown: Snail and Twist were not involved in pulmonary sarcomatoid carcinomas but instead upregulation of c-Jun and consecutive overexpression of vimentin and fascin was seen [66].

When dissecting sites of metastasis, the way EMT is regulated gets more diverse: PREP1 accumulation was found in a large number of human brain metastases of various solid tumors, including NSCLC. PREP1 induces the expression of multiple activator protein 1 components including Fos-related antigen 1 (FRA-1). FRA-1 and PBX1 are required for EMT triggered by PREP1 in lung tumor cells [67]. The study by Shen showed that increased levels of long noncoding RNA MALAT1 promotes lung cancer brain metastasis by EMT, whereas silencing of MALAT1 inhibits lung cancer cell migration and metastasis in the brain [68]. So far nothing similar was investigated for bone metastasis by pulmonary carcinomas.

We have now focused on single cell and small cluster migration. However, in surgical pathology routine, most well-differentiated carcinomas including lung carcinomas move in large cell clusters, for example, acinar adenocarcinomas will show nicely structured acini deep within the stroma and even within blood vessels (Fig. 18.10a, b). The mechanisms how these tumor cells manage their coordinated movement by retaining their epithelial structure is almost unknown. These carcinomas do not undergo EMT. Recently in an investigation using Drosophila border cells as a model, the process of migration of large cell complexes were elucidated. By RNAi silencing 360, conserved signaling transduction genes were knocked down to identify essential pathways for border cell migration. Four genes associated with TGF-β signaling were identified: Rack1 (receptor of activated C kinase), brk (brinker), mad (mother against dpp), and sax (saxophone). Inhibition of Src activity by Rack1 may be important for border cell migration and cluster cohesion maintenance. Although this study focused on signaling pathways involved in collective migration during embryogenesis and organogenesis, these data could be the first step in understanding migration of carcinoma cell complexes in metastasis [69].

Tumor cells orient themselves along adhesion molecules as expressed by matrix proteins, but in addition they also sense for oxygen and most probably also orient themselves for higher oxygen tension [70]. Invasion into blood vessels is very similar to invasion into the stroma: tumor cells have already learned to degrade proteins of the basal lamina at the epithelial border, and similar proteins form the matrix of small blood vessels. Forming holes into the basal lamina of these blood vessels is therefore easily done and tumor cells migrate into the intima. However, a new problem arises for tumor cells within the circulation: shear stress due to tumor cell deformation in small blood vessels and the problem with coagulation.

Shear stress is usually well tolerated by those tumor cells which underwent EMT: tumor cells expressing vimentin and α-actin can adapt to the capillary diameters, but cells still expressing cytokeratins might burst. This is one of the reasons why a majority of tumor cells do not survive within the circulation [71]. With respect to coagulation, tumor cells on one hand have to avoid being trapped within a blood clot, but on the other hand will need a clot to slow down the speed of the bloodstream, attach to the clot, and use it for extravasation [72]. Clot formation might be induced by tissue factors being produced and released by macrophages. Impairment of macrophage function decreased tumor cell survival without altering clot formation, demonstrating that the recruitment of functional macrophages was essential for tumor cell survival [73]. Another way how tumor cells might trigger clot formation has been demonstrated in mucinous adenocarcinomas. Mucins secreted by the tumor cells induced platelet aggregation and furthermore interacted with L-selectin and platelet-derived P-selectin without thrombin generation [74]. This interaction already points to the next step: adherence to vascular walls for extravasation. Coming back to tumor cell trapping by blood clots, it seems that carcinoma cells require the assistance of macrophages and granulocytes for fibrinolysis. In a study of lung carcinomas, fibrinolytic components as tissue plasminogen activators (tPA) and the inhibitors PAI-1 and PAI-2 were all negative in tumor cells, whereas urokinase-specific antibodies stained loosely packed tumor cells and macrophages. Both PAI-1 and PAI-2 were most prominently expressed within interstitial and alveolar macrophages [75]. In another study analyzing pulmonary adenocarcinomas, a positive correlation was found between Ets-1 and urokinase-type plasminogen activator (u-PA) expression [76].

Invasion into lymph vessels is easier than into blood vessels due to the tiny wall of the former. In addition carcinoma cells might already enter the lymphatic stream by the interstitial channels of the lymph draining system. On the contrary tumor cells can easily congest lymph vessels. This can reverse the lymph flow, which might explain unusual sites of lymph node metastasis and so-called skip lesions. In contrast to the situation within blood vessels, carcinoma cells in lymphatics have to deal with the immune system. So survival is dependent on induction of immune cell escape mechanisms (see above).

Whereas carcinoma cells entering the bloodstream might cause early-onset distant metastasis and thus shorten overall survival of the patient [1], propagation of carcinoma cells along the lymphatics will set distant metastasis later. These tumor cells will set primarily metastasis within regional lymph nodes.

Angiogenesis at the metastatic site in one part follows the same principles as in the primary focus; however, there is one major problem: whereas at the primary focus, lung carcinoma cells cross talk with stroma cells by mechanisms and transmitters which have been developed during the process of developing from the precursor lesion to in situ carcinoma to invasive carcinoma, this cross talk is different in the new metastatic site. Brain glia cells or bone marrow stroma cells might respond to other signals than those stroma cells within the lung. So the major developmental step to establish a metastatic focus is communication with the stroma, and further more communication might be different depending on the location. In one investigation a bridge was built between angiogenesis at the primary and metastatic site. CXCL12 was expressed in tumor cells and in tumor vessels; CXCR7 was expressed by tumor and endothelial cells in the primary tumor and in the brain metastasis. CXCR4 showed a nuclear positivity in all samples, but only CXCL12 expression in tumor endothelial cells was significantly correlated with shorter survival [92].

Interaction with stroma: There are no published data, which could highlight general mechanisms by which lung carcinoma cell communicates with their stromal counterparts; however, communication at different organ sites have been studied and therefore will be discussed in this paragraphs.

Research coming from brain metastasis of breast and NSCLC has raised several important findings. First metastatic carcinomas can colonize the brain in different ways. Renal cell carcinomas most often form metastases which are well circumscribed and do not grow out of the microglia pseudocapsule, whereas SCLC tend to form small metastatic foci and tumor cells grow into the microglia pseudocapsule and beyond into brain parenchyma. This has nicely been demonstrated by a coculture system consisting of an organotypic mouse brain slice and epithelial cells embedded in matrigel (3D cell sphere) [96]. In addition the same group of researchers has shown that microglia support invasion and colonization of brain tissue by breast and lung cancer cells (Fig. 18.11a, b). This is under the control of the Wnt pathway, as upregulation of Dickkopf-2 an inhibitor of Wnt inactivates the prometastatic function of microglia. Similar to tumor-dendritic cell interaction, bacterial lipopolysaccharide shifts tumor-educated microglia into a classical M1 phenotype, reduces their proinvasive function, and unmasks inflammatory and Wnt signaling as the most strongly regulated pathways [97]. Several factors have been identified as being specifically involved in regulating brain metastasis, but so far these are still isolated factors, and the main question, how these different factors interact, remains unanswered.

Among the different cells of the brain, astrocytes seem to serve invading carcinoma cells. Astrocytes secrete matrix metalloprotease-2 (MMP-2) and MMP-9 that proactively induced human lung and breast tumor cell invasion and metastasis formation [98]. In addition factors from the coagulation cascade are important. Plasmin acts as a defense against metastatic invasion by converting membrane-bound astrocyte FasL into a paracrine death signal for cancer cells and by inactivating the axon pathfinding molecule L1CAM, which metastatic cells express for spreading along brain capillaries and for metastatic outgrowth. But metastatic carcinoma cells from lung and breast secrete neuroserpin and serpin B2, to prevent plasmin generation and its metastasis-suppressive effects [99]. Within the Wnt pathway, LEF1/TCF4 acts independently of β-catenin in cerebrally metastasized human lung adenocarcinomas [100]. Downregulation of E-cadherin was also observed in a majority of adenocarcinoma and small cell lung cancer samples. LOH of the CDH1 gene was frequently found in SCLC. Altered expression of Dishevelled-1, Dishevelled-3, E-cadherin, and beta-catenin was present in brain metastases of SCLC and adenocarcinoma, again pointing to the importance of the Wnt signaling [101]. In another study, peritumoral brain edema was shown to be associated with increased β-catenin and E-cadherin and decreased CD44v6 and caspase-9 expression in brain metastatic squamous cell carcinoma [102]. These findings were confirmed in another study showing a significant correlation of increased collagen XVII in adenocarcinoma and increased caspase-9 and CD44v6 and decreased cellular apoptosis susceptibility protein (CAS) and Ki-67 in squamous cell carcinoma in brain metastasis [103].

Interestingly when looking up, adenocarcinomas with ALK1 rearrangement FGFR1 gene amplification correlated significantly with brain metastases. Although in these cases there were also higher numbers of visceral metastases, FGFR1 amplifications in brain metastases of adenocarcinomas were fivefold more frequent than in the primary tumors [68]. Also a cross talk of EGFR-MET was reported in adenocarcinomas with brain metastasis. This was not a direct interaction, but a signaling via the activation of mitogen-activated protein kinases (MAPK). EGFR-MET cross talk was independent from the mutation status of EGFR. MET signaling promoted migration and invasion. MET inhibition decreased the incidence of brain metastasis [104]. Also CXCR4 seems to play a role in brain metastasis. CXCR4 protein was highly overexpressed in patients with brain-specific metastasis, but significantly less in NSCLC patients with other organ metastases and without metastases [105].

ADAM9 levels were relatively higher in brain metastases than the levels observed in primary lung tumors. ADAM9 regulates lung cancer metastasis to the brain by facilitating the tPA-mediated cleavage of CDCP1 [106]. In a subsequent study, it was shown that ADAM9 regulated miR-218, which targets CDH2 in aggressive lung cancer cells. The downregulation of ADAM9 upregulated SLIT2 and miR-218, which together downregulated CDH2 expression. This study revealed that ADAM9 activates CDH2 through the release of miR-218 inhibition on CDH2 in lung adenocarcinoma [107].

A lot of interesting studies focused on the comparison of genomic alterations between primary lung carcinomas versus brain and bone metastasis. The hypothesis is that there might be a clonal diversity between these two. It is still not clear if genetic differences between the primary tumor and the metastatic site are a primary event, i.e., clones are existing within the primary tumor or, if these are secondary events, reflecting the interaction of the carcinoma cells with the microenvironment and the cells therein. Mock et al. found gene copy number variations between primary and CNS metastasis by array CGH. Genes with amplified copy numbers in primary and metastatic tumors were related to DNA replication and mismatch repair. Genes only amplified in the metastatic tumor were related to leukocyte migration and organ development. Genes with a lower copy number in the metastatic tumor were related to proteolysis, negative regulation of cell proliferation, and cell adhesion [108].

Many more studies focused on specific genes associated with brain metastasis. Lower LKB1 copy number variations and KRAS mutation were significantly associated with more brain metastasis, even predicted brain metastasis [109]. In the study by Gao, miR-95-3p suppressed tumorigenicity and brain metastasis in vivo and increased overall survival and brain metastasis-free survival. Accordingly the levels of miR-95-3p, pri-miR-95, and ABLIM2 mRNA were decreased, and cyclin D1 was increased in brain metastatic tissues [110]. In another study the expression of ACTN4 (actinin α4) was associated with lung cancer metastasis to the brain. The protein essential for cytoskeleton organization and cell motility was significantly elevated in the metastatic brain tumor but not in the primary lung cancer [111].

As with many marker studies focusing on single genes/proteins, a selection bias or just an over interpretation does occur. We already discussed MALAT1 in the setting of invasion and metastasis. Shen and coworkers found higher levels of MALAT1 in brain metastases compared to other extrapulmonary sites. In the in vitro experiments, it turned out that the major function of this lnRNA is EMT [68]. What can be concluded is that EMT seems to be required for tumor cells invading brain tissues, but MALAT1 is not a brain metastasis gene “per se.” A similar investigation searched for brain metastasis genes and came up with an EMT regulator: pre-B-cell leukemia homeobox (Pbx)-regulating protein-1 (Prep1) overexpression triggered EMT, whereas PREP1 downregulation inhibits the induction of EMT in response to TGF-β. PREP1 modulates the sensitivity to SMAD3 and induces the expression of Fos-related antigen 1 (FRA-1). Both FRA-1 and PBX1 are required for the mesenchymal changes triggered by PREP1 in lung tumor cells. PREP1-induced mesenchymal transformation correlates with increased lung colonization and PREP1 accumulation was found in human brain metastases [67]. Similarly migration seems to play a role in brain metastasis, not surprisingly, since migration is often associated with EMT. Han et al. showed that knockdown of KDM5B and SIRT1 genes specifically inhibits lung cancer cell migration in vitro. SIRT1 was highly expressed in brain metastasis. Using other lung cancer cell lines, the authors showed that the function of SIRT1 correlates with cell migration [112]. There are not many studies looking for the influence of molecules of the adhesin family. Nasser and colleagues investigated E-cadherin expression. Low E-cadherin expression was associated with increased risk of developing brain metastasis. By treating tumor in a mouse model with pioglitazone, a peroxisome proliferator-activated receptor γ-activating drug prevented loss of E-cadherin expression and reduced expression of MMP-9 and fibronectin and furthermore the development of brain metastasis [113].