Introduction

Specific molecular alterations occurring during progression of adenomas to metastatic colorectal cancer (CRC) were elucidated by the pioneering work of Kinzler and Vogelstein [1, 2]. CRC became a role model for molecular events determining the progression of transformed epithelial cells to metastatic carcinoma cells after initiation and subsequent acquisition of multiple tumor-associated mutations due to genomic instability [3]. Depending on the subtype of CRC, specific patterns of chromosomal instability, DNA repair defects or aberrant DNA methylation can be distinguished [3]. These changes culminate in metastasis, which is the cause for the high death toll in patients with CRC [4]. The liver is the major target organ of CRC metastasis as indicated by an investigation covering 4000 tumor patients displaying 11 subtypes of adenocarcinomas [5], however also peritoneal metastasis is not infrequent. While the liver is affected by metastasis in 75 % of these cases, lung metastases were found in only 20 % of these patients [5]. This preference of CRC metastasis can at least partly be explained by the anatomical layout of the vasculature which drains most colon cancer cells through the mesenteric circulation into the liver [6, 7], assuming that the trapping of metastasizing colon cancer cells within the capillary bed of the liver seconds after leaving the colon is a determining mechanism [7, 8]. Indeed, the metastatic dissemination of CRC cells seems to follow, at least in part, a circulation-dependent chronological pattern causing the appearance of liver metastases well before the formation of lung metastases [6]. However, molecular mechanisms may complement or may even overrule the significance of these anatomical aspects according to the well-established ‘seed and soil’ hypothesis [9] that was recently expanded to the concept of pre-metastatic niche formation [10]. The latter addresses the cellular and molecular changes that may occur in the target tissue of metastasis upon the activity of circulating signaling factors as free molecules or packed in exosomes [10] well before a tumor disseminates from the primary site, rendering the target site susceptible to metastatic cells. We here focus on targets that favour dissemination of CRC to the liver (Fig. 1) and discuss approaches for therapeutic intervention. Experimental models are based on syngeneic or xenograft models with a focus on orthotopic liver metastasis models for CRC cells. In a few cases models for experimental metastasis or subcutaneous models are discussed. For target identification and validation CRC cell lines with high tropism of metastasis to the liver in comparison to matching cell lines with low metastatic propensity, have been used. We also describe targets which have been identified by transcriptional profiling of primary CRC versus corresponding liver metastases. We discuss intracellular, plasma-membrane associated, trans-membrane and secreted proteins as targets. They can be derived from tumor cells, but also from stromal cells such as pericytes and fibroblasts.

Fig. 1
figure 1

Schematic display of CRC-related metastasis targets. The circle represents a CRC cell with intracellular, transmembrane and plasma-membrane associated genes. Secreted targets can be derived from tumor cells, stromal cells, or both, depending on the specific target. Abbreviations are explained in the text. ADAM 10 a disintegrin and metalloproteinase 10, APOBEC3G apolipoprotein B mRNA editing enzyme, catalytic polypeptide-like 3G, CD63, CD133 cluster of differentiation 63, 133, CXCR4 CXC chemokine receptor 4, c-MET tyrosine kinase c-MET, HGF hepatocyte growth factor, HIF-1α hypoxia-inducible factor-1α, LIP-C lipase C, L1CAM L1 cell adhesion molecule, MACC1 metastasis associated in colon cancer 1, MMPs matrix metalloproteinases, Notch 1–4 Notch receptors 1–4, PRL3 phosphatase of regenerating liver 3, SDF-1 stromal cell-derived factor-1, S100A4, S100A8, S100P S100 family proteins A4, A8 and 100P, TGFβ transforming growth factor β, TIMP1 tissue inhibitor of metalloproteinase 1, Trop-2 tumor-associated signal transducer 2

Full size image

Dissemination of CRC to the liver

Metastasis is mediated by an orderly sequence of steps comprising local invasion, intravasation into blood or lymph vessels, survival in the circulation, homing and extravasation, survival of dormancy, and colony formation [11]. As it became evident that anatomical conditions cannot always explain efficacy and pattern of metastasis, it was postulated that the compatibility between the seed (disseminated tumor cells) and the soil (distant sites) are essential for the metastatic process [11]. Metastasis initiation mostly involves remodeling of extracellular matrix (ECM) molecules which compose the basement membrane of epithelial tissues including the colon [12]. Furthermore, metastatic cells have to acquire a migratory phenotype as a pre-requisite for the mobility that underlies initial steps of metastasis [12]. These molecular and morphological changes arise as a consequence of the constant and evolving cross-talk with stromal cells [12], which is critical for the survival of metastatic cells at this time point [12]. In CRC it was consequently found that tumor cells at the invasive front accumulate β-catenin in the nucleus triggering their self-renewal capacity [13–15]. Once in the circulation, CRC cells are able to bind to and cover themselves with platelets leading to better protection from shear stress and immune response [16] and also increasing the likelihood of emboli formation in capillary beds of a possible target organ of metastasis at a distant site [16]. In this process, further pathways are activated which promote specific attachment of these tumor cell/platelet aggregates to endothelial cells [16] rather than just the formation of emboli. In the liver sinusoids, the specific ‘vessels’ in the liver microenvironment, the tumor cells are confronted with different cells that may exhibit some anti-tumoral/anti-metastatic activity which would represent another critical step that successful metastatic tumor cells have to overcome [17–19]. These obstacles include Kupffer cells, specialized macrophages lining the walls of the sinusoids [17] and pit cells, large granular lymphocytes with high cytotoxic activity against tumor cells [18]. In addition, hepatic stellate cells in the peri-sinusoidal space can act as antigen-presenting cells which may elicit an anti-tumoral immune response [19]. After extravasation from the sinusoids, colonization of the liver parenchyma is the final step of metastasis to the liver. In this process it is essential that the tumor cells undergo mesenchymal-epithelial transition (MET), the transition from a spindle-shaped mesenchymal phenotype to a polarized, epithelial phenotype with expression of adhesion molecules allowing successful establishment of a metastatic colony in the new environment [19]. It is still under investigation, how the rather newly coined migrating cancer stem cells (MCSC) surpass these different steps of metastasis and whether their mode of action decisively differs from the current concept [19]. Along these steps of the metastatic cascade, one can categorize the metastasis-associated genes in three groups contributing to metastasis initiation, progression, and virulence, respectively [11, 20, 21]. Those genes with metastasis-initiating function ensure self-renewal, resistance to intermittent hypoxia, migration, and invasion, as well as infiltration of pro-tumorigenic bone marrow-derived cells at the primary site and survival in the circulation. Genes ensuring metastasis progression mediate homing, adhesion and extravasation at the distant site as well as survival of dormancy, re-initiation of proliferation and possibly angiogenesis. The virulence-determining genes can be considered as those determining organ-specific colonization or those emerging genes which are expressed and secreted at the primary site by specific tumors or tumor host-interactions, whose gene products are secreted and act over a distance in future target organs of metastasis in order to create a pre-metastatic niche even before the arrival of tumor cells. The set of responsible genes is likely triggered by the specific type of tumor and the distant sites to be colonized, but probably there are also genes that are generically involved in this process.

Metastasis models

The first crucial step in the development of CRC liver metastasis models was the demonstration of metastasis to the liver of CRC cells implanted into the spleen of nude mice [22] in order to recapitulate the observations in the clinic. Subsequently, highly metastatic CRC cells were selected after implantation of surgical specimen into the subcutis, caecum or spleen and serial passage of cancer cells derived from liver metastases [23]. Tumor cell variants with high metastatic potential reveal higher rates of experimental liver metastases (KM12L4 and KM12SM), while higher rates of spontaneous liver metastasis formation were derived from the parental cell line KM12 [23]. Orthotopic implantation of human carcinomas into nude mice provided reliable models for studies of the biology and therapy of metastasis [24]. A crucial finding was the demonstration that the response of tumors to chemotherapy was modulated by the microenvironment by organ-specific upregulation of multi-drug resistance associated P-glycoprotein [25, 26]. Examination of these models revealed several targets for possible therapeutic intervention. It was shown that expression of epidermal growth factor receptor (EGFR) in CRC cells correlates with their potential to form hepatic metastases [27]. Basic fibroblast growth factor (bFGF), interleukin 8 (IL8), and matrix metalloproteinase 9 (MMP9) [28, 29] have emerged as further targets for treatment of liver metastases derived from CRC. The importance of inhibition of platelet-derived growth factor receptor (PDGFR) signaling pathways in tumor-associated endothelial cells and pericytes for inhibition of CRC liver metastasis was demonstrated [30]. CRC tumors consist of heterogeneous cell populations expressing a combination of factors contributing to tumor cell proliferation and stromal cell activation, such as EGFR, PDGFR, and vascular endothelial growth factor receptor (VEGFR). The rationale of simultaneous targeting of these receptors was adopted and indeed, this was the most efficient strategy of therapy in these models [31]. In addition to these receptor tyrosine kinases, activation of intracellular tyrosine kinases, such as SRC, seems to play an important role in human colon cancer cells with high metastatic potential [32] and the presence of activating SRC mutations in a subset of advanced CRC is indicative of this fact [33].

Liver metastasis and inhibition of MMP’s

As invasive processes in general are a pre-requisite for crucial steps of the metastatic cascade, it was hypothesized that proteases, which degrade components of the basement membrane and interstitial ECM would be an ideal target for intervention with metastasis [34]. Among these proteases, the matrix metalloproteinases (MMP’s) were thought to be specifically attractive as they together can in fact degrade all components of the ECM and the basement membrane including collagen type IV [34]. MMP’s are a class of more than 20 Zn-dependent proteolytic enzymes, which exhibit a very conserved catalytic centre that could easily be explored by chemists to develop highly efficient small molecule broad spectrum inhibitors [34]. MMP inhibitors (MMPI’s) were developed with the expectation to inhibit invasion and spread of tumors. This view could not be held up as broad spectrum MMPI’s failed in clinical trials in cancer patients [35]. Retrospectively, it is agreed that the biology of MMP’s is far more complex than anticipated. It turned out that they exert pro- and anti-tumoral functions via their indirect signaling function due to cleavage of a whole cascade of signaling molecules in the context of the protease web [36–39]. This late insight in mind, it is not surprising anymore that broad spectrum MMPI’s can exhibit unforeseen effects including the promotion of liver metastasis formation in different tumor models [40]. One of the underlying mechanisms of this observation may be based on modulation of the physiology of the liver by MMPI’s in the context of a highly connected protease network. The generation of new microenvironments by inhibition of MMP’s can result in changes of susceptibility to circulating tumor cells. In fact, treatment of mice with a broad-spectrum MMPI led to up-regulation of hepatocyte growth factor (HGF, also known as scatter factor) in the liver [40], which, by induction of the invasive growth programme [50] may increase the virulence of metastasis formation in the liver. Unpredictable perturbation of tissue homeostasis affecting metastasis became also evident when MMP9 was ablated in mice. Upon challenge with lacZ-tagged syngeneic murine colorectal carcinoma cells (CT26-lacZ), an 8 fold increase in experimental liver metastasis was observed as compared to the wild-type control [36, 37]. In fact, an MMP9 ablation-adaptive change of transcripts encoding metastasis-associated functions was observed in MMP9-deficient host tissue. Expression of interleukin 6 (IL6) was up-regulated in the bone marrow and in hematopoietic cells localized in the liver leading to increased IL6 levels in the blood [36, 37]. Such elevated IL6 levels are a negative prognostic marker for survival in CRC patients [41].

While the complexity of interference with MMP activity in the host tissue is still under investigation, specific inhibition of MMP’s in tumor cells revealed effects on their metastatic potential. Liver colonization after knock-down of MMP2 or MMP9 in the tumor cells was investigated in the syngeneic L-CI-5 s lymphoma model [42] which preferentially homes to the liver. MMP2 inhibition resulted in inhibition of metastatic foci growth, whereas MMP9 inhibition suppressed the invasive phase of liver colonisation. Along these lines, MMPI’s with greater selectivity/specificity for MMP9 showed greater efficacy against liver metastasis in the model [43]. However, long-term effects on the host, as seen in the knock out model, rule out the option of long-term treatment of CRC in patients with MMPI’s. While synoptic evaluation of the results regarding unspecific versus specific chemical inhibition of MMP’s, MMP9 knock-out mice and liver colonisation as well as MMP2 and MMP9 knock-down experiments would argue against general conclusions regarding the role of MMP’s in liver metastasis, it is clear that they must be ruled out as primary therapeutic targets. It is true that model-specific features such as, mouse strain, tumor cell type, and mode of tumor cell administration might trigger the specific outcome of liver metastasis experiments. Also one should keep in mind that MMP function is not only mediated by its enzymatic function, but also by protein–protein interactions [39].

Even the intuition that the natural MMPI’s, the tissue inhibitors of metalloproteinases (TIMP1, −2, −3, and −4), should exhibit an anti-metastatic effect, had been challenged in the past few years. Recently, it emerged that TIMP1 can create a pre-metastatic niche specifically in the liver for colon carcinoma cells as well as numerous other tumor cells in the respective syngeneic mouse lines as well as in xenograft models where the systemic TIMP1 levels were experimentally increased to similar levels [44] as were found in cancer patients with bad prognosis [45, 46]. In [46] liver metastases were discussed, but post-operative follow-up was not reported. The experimental observations can now explain the so far unresolved paradox that TIMP1 levels correlate with bad prognosis in CRC patients [44]. Moreover, elevated TIMP1 levels were associated with the extent of liver metastasis and frequency of disease relapse in CRC patients [44–46]. Mice with elevated TIMP1 levels were challenged with syngeneic CT26 colon carcinoma cells and a drastic, liver-specific increase of metastasis was found. Further experiments of this study showed that TIMP1 actually creates a liver-specific pre-metastatic niche, an emerging concept providing an explanation of early metastatic events, where homing and successful colonization of an organ is pre-primed by tumor-secreted factors which alter the cellular and molecular signature of the target organ of metastasis before tumor cells reach the site [47, 48]. In consequence, the TIMP1-induced factors that were specifically up-regulated in the liver such as stromal cell-derived factor-1 (SDF-1), fibronectin, transforming growth factor-β (TGF-β), urokinase-type plasminogen activator (uPA) and S100A8 are potentially new targets for anti-metastatic intervention [44]. In fact, pharmacological interference with the interaction of one of these niche-factors, the cytokine SDF-1, with its receptor CXC chemokine receptor 4 (CXCR4) with the widely used small molecule inhibitor AMD3100 led to abrogation of TIMP1-induced pre-metastatic niche formation in the liver [44].

Tyrosine kinase c-MET signaling in tumor cells or liver tissue and generation of liver metastases

c-MET is a pan-epithelial transmembrane tyrosine kinase which can be activated by its ligand HGF [49]. It is overexpressed and activated in many types of carcinomas including colon carcinomas [49–52]. c-MET can be activated by autocrine and paracrine mechanisms [49, 50]. Overexpression of c-MET in tumor cells can lead to ligand-independent activation of the c-MET signaling pathway, resulting in hyperactivation of phosphatidylinositol-3-kinase (PI3 K) and mitogen-activated protein kinase (MAPK) signaling [49]. In tumor cells, this elicits the ‘invasive growth programme’ including induction of proliferation, migration, survival, angiogenesis and morphogenesis [50]. Another level of liver metastasis-promoting c-MET signaling is its modulation of the target tissue of metastasis. This alternative parameter impacting on metastasis formation in the liver is based on the finding that ectopic TIMP1 expression in the liver [53] or elevation of systemic levels of TIMP1 in the host organism [44] formed a potent metastatic niche [53] or even a pre-metastatic niche [44], respectively, leading to increased liver metastasis. In the case of ectopic expression of TIMP1 in liver tissue, cell surface-associated c-MET accumulates by the activity of TIMP1 to inhibit the c-MET sheddase, a disintegrin and metalloproteinase 10 (ADAM-10) [53–55]. The resulting increased c-MET signaling in the liver tissue led to expression of several pro-metastatic factors in this target tissue of metastasis [53] including the protease uPA [56]. Experiments with uPA-ablated mice indicated the crucial role of uPA for the recruitment of neutrophilic cells with elevated TIMP1 levels to the liver and also in increased availability of HGF [56]. HGF promotes a scattered and very invasive pattern of metastases after i.v. inoculation of syngeneic lymphoma cell line LCI.5 s or human fibrosarcoma cells HT 1080 in the liver [56]. In vitro, knock-down of tumor cell-derived TIMP1 prevented HGF-induced phosphorylation of c-MET, ADAM 10 knock-down increased c-MET phosphorylation in metastatic foci and scattering into the surrounding liver parenchyma. TIMP1 expression and c-MET signaling were co-localized within the metastatic colonies and expressed by tumor cells [55].

Similar molecular events take place on the surface of tumor cells themselves rendering them more aggressive. TIMP1 is able to induce hypoxia-inducible factor 1-α (HIF-1α) signaling in tumor cells [57, 58]. Also on tumor cells, TIMP1 induces inhibition of ADAM 10, thereby preventing the shedding of c-MET. This results in cell-associated c-MET accumulation and signaling in host as well as in tumor cells. Making use of i.v. inoculated L-CI.5 s lymphoma cells which home to the liver, it was demonstrated that tumor cell-derived TIMP1 is necessary for maintenance of metastasis-promoting c-MET signaling via inhibition of ADAM 10 [55].

The mode of action of TIMP1 leading to stimulation of liver metastasisis likely more complex, as TIMP1 can exert metalloproteinase inhibition-independent functions. TIMP1 as a cell-binding protein recently was shown to interact in vitro with cluster of differentiation 63 (CD63) by activating the PI3 K signalling pathway dependent on Akt8 virus oncogene cellular homolog (Akt) activation [62] or with CD63 and integrin β1, independent of Akt activation [59, 60]. In fact, we recently found that TIMP1 induces HIF-1α and miR-210 in tumor cells in vitro via CD63, resulting in increased tumorigenicity of the cells [62]. HIF-1α as a mediator of the hypoxic state promotes pro-invasive mechanisms by upregulation of miR-210 which inhibits respiration by inhibition of the subunit D of the succinate dehydrogenase complex (SDHD), regulates the switch to anaerobic glycolysis and induces cell-cycle arrest by inhibition of transcription factor E2F3 in vitro [61–63].

The tumor-promoting function of TIMP1 is still neglected in this field of research [64]. However, the experimental evidence for TIMP1 as a driver of liver metastasis in the past few years explains its strong and independent association with shorter survival time in CRC patients [45, 46]. On the other hand, low TIMP1 was found to be significantly associated with objective response and survival in CRC patients receiving a combination of irinotecan, 5 fluoro-uracil and folinic acid [65].

The relevance of c-MET and its impact on CRC becomes even more evident as it was identified as a transcriptional target of metastasis associated in colon cancer (MACC1) [66]. The src homology 3 (SH3) domain and proline-rich region containing protein MACC1 promotes proliferation, invasion, and HGF-induced scattering of colon cancer cells in vitro and metastasis in mouse models. Mutants of MACC1 lacking the SH3 domain of the proline-rich motif were functionally inactive. MACC1 is a predictor of CRC metastasis. 5 year survival was 80 % for MACC1 low-expressors, but 15 % for individuals who showed high MACC1 expression in their primary tumors [66]. MACC1 mRNA levels were identified as a predictor of recurrence after resection of CRC liver metastasis [67]. Circulating MACC1 transcripts in CRC patients are predictors of metastasis and prognosis [68]. Altogether, c-MET and MACC1 are metastatic pacemakers in CRC patients [69, 70] and therefore important targets for prognosis and therapy of CRC patients. In this context, it was shown that micro RNA-143 (miR-143) targets MACC1, resulting in inhibition of invasion and migration of colon cancer cells in vitro [71].

Phosphatase of regenerating liver (PRL3)

Dysregulation of protein tyrosine phosphatases resulting in aberrant tyrosine phosphorylation has been implicated in development of cancer and its progression [72, 73]. The family of phosphatases of regenerating liver (PRL) is composed of three members. They are farnesylated proteins, which are associated with the inner leaflet of the plasma membrane [72, 73]. Overexpression of PRL3 promotes tumor cell migration, invasion, and metastasis based on Akt activation and subsequent upregulation of MMP’s and EGFR activation as shown in vitro and in vivo [74, 75]. Activation of Akt is based on downregulation of phosphatase and tensin homolog (PTEN) and activation of PI3 K signaling resulting in promotion of epithelial-mesenchymal transition (EMT) in vitro [76]. In a study with 18 patient samples, global gene expression profiling has implicated PRL3 in metastases of CRC to the liver as it is the only gene whose expression was restricted to samples from metastases while no expression in normal colon mucosa and non-metastatic carcinomas [77] was found. PRL3 expression is specific for CRC metastases and was not detected in metastases of other types of cancer [77]. Interestingly, PRL3 can be found in the vasculature of tumors regardless of the tumor type [78], indicating a role in tumor angiogenesis. In fact, the multitasking phosphatase PRL3 is able to promote endothelial recruitment and formation of new blood vessels [79]. It was shown that PRL3 is activated by loss of TGFβ signaling, which often occurs during progression of CRC, by releasing inhibition of its transcription [80].

Taken together, PRL3 seems to be a promising target for treatment of liver metastases derived from CRC, since PRL3 is only expressed during development of the heart, blood vessels and erythrocyte precursors, but not in the corresponding mature tissues [81] so that side effects are unlikely. Proof-of-concept experiments have shown that intracellularly delivered antibodies directed against PRL3 inhibit lung metastasis of human colorectal HT-116 and A2780 ovarian cancer cells in xenograft models and B16 melanoma cells with high PRL3 expression after tail vein injection into syngeneic C57Bl6 mice [81, 82]. In these settings the PRL3 expression status was highly predictive. However, in the context of CRC-derived liver metastasis, PRL3 has not been functionally validated in a respective in vivo system. Functional validation of PRL3 as a target would justify the discovery and development of small molecule PRL3 inhibitors.

L1 cell adhesion molecule (L1CAM)

L1CAM, a member of the L1 family of adhesion molecules, is composed of an extracellular domain consisting of six Ig-like domains and five fibronectin III domains, a transmembrane domain and a cytoplasmic domain [83]. L1CAM expression is involved in axon guidance and neural crest migration and is associated with the invasive front of colon carcinomas [84].

It was shown that MMP2 and MMP9 mediate liver metastasis of the murine lymphoma cell line L-CI.5 s [85]. Increased migration, invasion, and proliferation in vitro as well as increased experimental and spontaneous metastasis to the liver and transcriptional activation of MMP2 and MMP9 could be linked to expression of L1CAM [85, 86], because knock-down of L1CAM abolished these phenomena. Analogous observations were made with the ovarian carcinoma cell line SKOV3ip. The MMP2- and MMP9-specific inhibitor SB-3CT reduced migration and invasion in vitro as well as experimental metastasis to the lungs [87].

Two splice variants have been identified for L1CAM: a full-length form (FL-L1CAM) and an evolutionary conserved splice variant lacking exons 2 and 27 [88]. Exon 2 is involved in homophilic and heterophilic binding to neural ligands contactin and TAX-1 and the cytoplasmic sequence encoded by exon 27 is necessary for clathrin-dependent endocytosis in vitro [89, 90]. Only FL-L1CAM was able to promote experimental metastasis of HT1080 fibrosarcoma, SKOV3ip, and L-CI.5 s to the lungs and to the liver, respectively [91]. Metastasis formation correlated with increased invasive potential, elevated MMP2 and MMP9 expression in vitro, and enhanced proteolytic activity of these MMPs in vivo. These findings might indicate that FL-L1CAM activates distinct signaling pathways upon internalisation due to its endosomal location. It was shown in vitro that L1CAM is able to interact with membrane molecules via its extracellular domain or with other proteins via its cytoplasmic domain [92, 93]. The therapeutic potential of L1CAM antibodies for treatment of cancer has been evaluated [94–96]. They were shown to inhibit growth and dissemination of ovarian carcinoma cells in the peritoneal cavity. L1CAM is a mediator of resistance to chemotherapy and therefore, treatment of pancreatic and ovarian xenografts by combined treatment of L1CAM antibodies and cytotoxic drugs showed increased therapeutic response [97]. From the clinical point of view, a strong correlation between expression of L1CAM in CRC patients and decreased survival has been demonstrated by independent groups [98, 99], confirming the role of L1CAM in liver metastasis. These findings are supported by preclinical data which underline the role of L1CAM in motility and β-catenin/Wnt signaling [100–102].

Tumor-associated calcium signal transducer 2 (Trop-2)

Trop-2 is a transmembrane pan-epithelial antigen related to epithelial cellular adhesion molecule (EpCAM,Trop-1) which is overexpressed in many types of cancer. It has been identified as an oncogene and its upregulation was shown to stimulate cancer growth in vitro and in vivo [103, 104]. Trop-2 can activate the MAPK pathway and support a protumoral signaling network [105, 106]. A possible context of Trop-2 as a metastasis-associated target was indicated by the observation that high levels of Trop-2 correlate with a higher frequency of liver metastasis in 16 CRC patients [107]. Independent from these observations, it was shown that elevated expression of Trop-2 was associated with survival, disease recurrence, and liver metastasis of patients with CRC [108]. Trop-2 was validated as a driver of CRC-related liver metastasis in an experimental in vivo system [109]. After transfection of the CRC cell line KM12SM with Trop-2, the cells gained a dramatically increased metastatic potential as they formed liver metastases after orthotopic injection. Surprisingly, deletion of the whole cytoplasmic domain of Trop-2 increased the metastatic propensity, whereas deletion of the pleckstrin-homology binding domain motif diminished metastasis to the liver. Further validation experiments are required in order to substantiate this issue. Due to the broad expression of Trop-2 in epithelial tissues, therapeutic intervention with antibodies or antibody-related moieties such as antibody–drug conjugates or antibody-fusion proteins with immunotoxins may not be a feasible approach due to possible toxicity issues.

S100 family proteins S100A4 and S100A8

The S100 family consists of 24 members exhibiting a broad range of intracellular and extracellular functions such as regulation of calcium homeostasis and cytoskeletal components, phosphorylation of proteins as well as induction of gene expression when they act as transcription factors [110, 111]. Some S100 proteins are secreted or released and regulate cell functions in an autocrine and paracrine manner via activation of cell surface receptors, such as the receptor of advanced glycation end products (RAGE), G-protein coupled receptors, toll-like receptor 4 (TLR4), and heparin sulphate proteoglycan [100, 111]. S100A4 can interact with epidermal growth factor (EGF) and bFGF, thus enhancing the activity of the appropriate receptors as shown by in vitro experiments [110, 111].

Expression of S100A4 is significantly associated with progression and liver metastasis of CRC patients [112, 113]. An important contribution in this context is the finding that S100A4 is able to bind to tropomyosin and non-muscle myosin II in a Ca-dependent manner, promoting disassembly of myosin filaments in vitro [110]. Systemic RNA knock-down of S100A4 in human CRC cells such as HCT 116, SW 620 and DLD-1 reduces liver metastasis formation in xenograft mouse models [114]. Moreover, S100A4 has been identified in vitro and in vivo as a target of β- catenin/TCF signaling which is often deregulated in CRC [115]. Interestingly, another member of the S100 family, S100A8, is specifically involved in TIMP1-induced pre-metastatic niche formation in liver [44]. It was among those factors that correlated with the increased susceptibility of the liver to tumor cells, including colon carcinoma cells [44]. This suggests that S100 proteins can account for the promotion of liver-specific metastases in patients that have a TIMP1-associated bad prognosis of CRC.

Profiling of highly metastatic versus low-metastatic colon cancer cell lines reveals further genes promoting liver metastasis of CRC cells

These genes were identified by functional selection in an orthotopic mouse model of CRC [116]. The experimental approach was based on non-metastatic CRC cell lines SW480 and SW620 with only lymph node metastasis, their inoculation into the colon of nude mice, isolation of liver lesions (L-1 and L-2) and reinoculation for confirmation of the metastatic phenotype. Transcriptomic microarray analysis of the cell lines [116] and knockdown experiments revealed that four genes are essential mediators of liver metastasis: S100P, CD133, lipase C (LIP-C), and apolipoprotein B mRNA editing enzyme, catalytic polypeptide-like 3G (APOBEC3G). These genes are significantly increased in metastatic colonies as compared to non-metastatic primary tumors. Some impact of these four genes on metastasis is already known:

S100P has been previously identified as a prognostic marker in CRC patients [117]. S100P regulates Ca-signaling transduction to mediate cytoskeletal interactions and is induced by the prostaglandin E2 (PGE2)/EP4) receptor signaling pathway in colon cancer cells [118]. Ectopic expression of S100P in SW 480 CRC promotes invasion and metastasis and decreases sensitivity to 5 FU in vitro [119].

LIP-C is a hydrolase involved in lipoprotein uptake and lipoprotein homeostasis [120]. It has been shown that monoacylglycerol lipase regulates the fatty acid network and promotes cancer pathogenesis [121]. This finding indicates that cancer cells can co-opt a lipolytic enzyme to translate their lipogenic state into pro-tumorigenic signals in vitro and in vivo [121, 122].

CD133 is a stem cell marker in CRC [123]. Expression of CD133 or CD133 and CXCR4 on cancer cells correlates with the presence of lymph node metastasis in CRC patients [124]. However, the role of CD133 in colon cancer metastasis is still controversial, since both CD133+ and CD133− cancer stem cells can initiate tumor formation and the impact on the formation of metastatic colonies is not clear [124].

The fourth gene identified in the SW480 and SW620-based model, APOBEC3G, belongs to the family of cellular cytidine deaminases which deaminate cytidine to uridine in the single-strand DNA substrate [125]. Resulting G to A hypermutations at hot spots in the viral RNA destroy the coding and replicative capacity of the virus [125]. In respect to CRC liver metastasis it was shown that APOBEC3G induces expression of MMP2 by depression of miR29 mediated protein translation inhibition [126].The enzymatic activity of APOBEC3G can be suppressed by recruitment into high-molecular-weight ribonucleoprotein (RNP) complexes [127].

Notch pathway

The importance of Notch signaling in the context of cancer pathogenesis is well-documented [128], but the role of this pathway in liver metastasis of CRC remains to be studied in more detail. There is evidence that inhibition of Notch signaling interferes with CRC liver metastasis [129]. The amino-terminal enhancer of split (Aes) gene product, a member of the family of transcriptional co-repressors, was found to be decreased in liver metastasis compared to primary colon tumors in mice and humans. Aes inhibits Notch signaling in vitro by converting active recombining binding protein suppressor of hairless (rbpj) active transcription complexes into repression complexes on insoluble nuclear matrix. Notch signaling in the context of liver metastasis was shown to stimulate transendothelial migration [129].

TGFβ signaling

The TGFβ signaling pathway has emerged as an other possible target to interfere with CRC liver metastasis [130, 131]. TGFβ signaling is mediated by contraction of sma and mad 2, 3 and 4 (Smad 2, 3, and 4), which are translocated to the nucleus to regulate transcription. Loss of Smad 4 function, as shown with Smad4-null MC 38 and SW 620 colon cancer cells, results in a functional shift of TGFβ from a tumor suppressor to a tumor promoter [132]. TGFβ-receptor tyrosine kinase inhibitor LY2109761 was shown to inhibit liver metastasis of these cell lines by reducing expression of MMP2, uPA, and cyclooxygenase 2 as well as the reduction of E-cadherin expression [130]. Expression of Smad 4 in these cell lines reduced metastasis significantly [132].

Additional experiments support the view that the balance between Smad 4/Smad 7 and the TGFβ pathway may be critical for the metastatic process [133]. In a spleen injection model of liver metastasis, it was shown that abrogation of Smad signaling by ectopic expression of Smad 7 induces liver metastasis. Cells expressing Smad 7 migrated to the liver [133]. Increased junctional staining of claudin 1, claudin 4 and E-cadherin was found in liver metastasis.

The role of cancer-associated fibroblasts (CAF’s) in the context of liver metastasis of CRC has been investigated [134]. Although tumor cells are often non-responsive to TGFβ due to mutational inactivation of the components of the TGFβ pathway [135], CAF’s can respond to TGFβ and secrete tumor-promoting cytokines such as IL11, which can bind to its receptor gp130 on tumor cells and can activate STAT3 signaling in tumor cells, promoting their survival as liver metastases [134].

Concluding remarks

Overall, it should be kept in mind that the mechanisms for invasion and metastasis to the liver can vary between different individual CRC diseases and therefore the involved pathways have to be resolved case-by-case. Also, one should realize that animal models are useful for discovery of new targets and pathways, but are often rather poor models for treatment modalities of human disease.

We have analysed RNA expression of MACC1, MET, PRL3, S100P and TIMP1 in CRC in comparison to matching normal colon mucosa in publicly available RNA Seq data from the Cancer Genome Atlas (TCGA) (Fig. 2). All of them are dramatically overexpressed in tumor tissue. However, we did not find a correlation between progression of primary CRC’s as mesasured by clinical stage annotation and an increase in steady-state RNA levels of these genes (Fig. 3).The elucidation of the status of these genes in liver metastases, preferentially matching with the primary tumor is a pending issue.

Fig. 2
figure 2

Expression of selected genes in colon cancer in comparison to normal colon mucosa. Data were derived from the Cancer Genome Atlas. This cohort represents 421 Colon adenocarcinoma primary tumor samples as well as 40 matched normal samples. Expression was measured by whole transcriptome sequencing and values provided represent normal read counts (log2) as derived from the Broad FIREHOSE portal. ‘Solid tissue normal’ represents matched adjacent normal tissue of cancer patients. Data are shown for MACC1, MET, PRL3, S100P, and TIMP1. For all those genes expression levels between tumor and matched normal tissue are significantly different (t test, p value < 0.0001)

Full size image
Fig. 3
figure 3

RNA expression of selected genes in different stages of colorectal cancer progression. Data were derived from whole transcriptome RNA sequencing and values shown represent normalized read counts (log2) as provided by the Broad FIREHOSE portal. This cohort represents 421 colon adenocarcinoma primary tumor samples (72 stage i, 34 stage ii, 124 stage iia, 9 stage iib, 22 stage iii, 13 stage iiia, 52 stage iiib, 36 stage iiic, 43 stage iv, 16 stage iva). Expression is shown in correlation to pathological stage annotation. Selected genes: MACC1, MET, PRL3, S100P and TIMP1

Full size image

Prevention of liver metastasis of CRC after resection of the primary tumor based on new molecular targets might be a future option. It is of high clinical relevance that it is not yet possible to identify prospectively those patients (20–30 %) belonging to the high risk group with locally restricted stage II colon cancer but will ultimately experience disease recurrence [136]. In the subset of metastatic cases who undergo liver-metastasis resections, half were cured and the other half relapsed with metastases elsewhere in the liver and/or in the lungs. Thus the latter half of the patients could have been cured if early metastasis could be blocked post-operatively by metastasis-prevention, e.g. by employing adjuvant therapies to surgery [137]. Also, identification of a validated gene expression signature that is predictive for the metastatic potential would be helpful for the design of clinical trials [136]. Another clinical objective is the conversion of non-resectable to resectable disease based on new targets expressed in liver metastasis. A significantly higher resection rate (25 vs. 7.4 %) was observed in a randomized controlled clinical trial of cetuximab plus chemotherapy in patients with KRAS wild-type non-resectable colorectal liver-limited metastasis [138]. In addition to direct targeting of liver metastasis, identification of new pathways conferring dependence of liver metastasis on stromal interactions for their maintainance and propagation is an exciting issue for further studies. It is presently unclear, whether the targets described in this review can be translated into clinically relevant targets.