DAPK1 loss triggers tumor invasion in colorectal tumor cells

Colorectal cancer (CRC) is one of the leading cancer-related causes of death worldwide. Despite the improvement of surgical and chemotherapeutic treatments, as of yet, the disease has not been overcome due to metastasis to distant organs. Hence, it is of great relevance to understand the mechanisms responsible for metastasis initiation and progression and to identify novel metastatic markers for a higher chance of preventing the metastatic disease. The Death-associated protein kinase 1 (DAPK1), recently, has been shown to be a potential candidate for regulating metastasis in CRC. Hence, the aim of the study was to investigate the impact of DAPK1 protein on CRC aggressiveness. Using CRISPR/Cas9 technology, we generated DAPK1-deficient HCT116 monoclonal cell lines and characterized their knockout phenotype in vitro and in vivo. We show that loss of DAPK1 implemented changes in growth pattern and enhanced tumor budding in vivo in the chorioallantoic membrane (CAM) model. Further, we observed more tumor cell dissemination into chicken embryo organs and increased invasion capacity using rat brain 3D in vitro model. The novel identified DAPK1-loss gene expression signature showed a stroma typical pattern and was associated with a gained ability for remodeling the extracellular matrix. Finally, we suggest the DAPK1-ERK1 signaling axis being involved in metastatic progression of CRC. Our results highlight DAPK1 as an anti-metastatic player in CRC and suggest DAPK1 as a potential predictive biomarker for this cancer type.


Introduction
Colorectal cancer (CRC) is the fourth most frequently diagnosed malignancy worldwide. Approximately 8.5% of patients with CRC die from their cancer 1 . Development of metastasis predominantly in liver and lung is the fatal leading cause of death and a serious obstacle of curing CRC. Recently, it has been shown that Death-associated protein kinase 1 (DAPK1) might be a potential candidate regulating metastasis in CRC [2][3][4] . DAPK1 as an actin filament-associated calcium/calmodulin-regulated, stressresponsive serine/threonine kinase was reported to be an extremely pleiotropic molecule due to its unique multidomain structure [5][6][7][8][9] . Chen et al. 3 have postulated that three main action modes for metastasis suppression by DAPK1 might exist: the increase in susceptibility of tumor cells to apoptotic signals, inhibitory role on integrinmediated cell adhesion and migration and modulation of tumor microenvironment. Several reports have shown that primary tumors of patients with metastases (pM1) show decreased DAPK1 protein expression in comparison to primary tumors without metastases (pM0) 3,4 . Recently we have reported that the loss of anti-migratory function of DAPK1 could be one of the reasons for tumor cell dissemination 4 . In this regard it was not surprising that DAPK1 was nearly lost in tumor buds at the invasion front of CRC. Tumor buds are defined as single cells or small clusters of up to four cells 4 and their evaluation at the stroma border of colorectal tumors is supposed to be indicative for tumor aggressiveness 4 . Moreover, tumor budding seems to be an independent and a reliable predictor of local lymph node metastasis, tumor recurrence and survival 10 . Since DAPK1 is lost in this aggressive cell subpopulation at the tumor invasion front we suggest a tumor suppressor role for DAPK1. Interestingly, DAPK1 was found to be involved in the cross talk of tumor cells with macrophages of the stroma environment 11 . At the onset of cancer, loss of DAPK1 may provide selective advantage for hyperproliferative tumor cells of evading the p53-dependent apoptotic checkpoint during transformation 12 . Indeed, DAPK1 loss has been shown in a high number of T1 colorectal tumors 13 . At later stages of tumorigenesis and during metastasis, tumor cells seem to benefit from loss of DAPK1 resulting in reduced sensitivity to detachment from extracellular matrix (ECM) 14 . So far, the molecular mechanisms behind metastasis in CRC and especially the role of DAPK1 in this process remain little understood.
To mirror-image DAPK1 loss in tumor buds and to shed more light into the role of DAPK1 in CRC aggressiveness we designed unique CRISPR/Cas9 DAPK1 knockout (ko) cell lines and characterized the knockout phenotype in vitro and in vivo. Although we observed a definite variation in functional readouts for the different clones, we were able to extract a uniform DAPK1-stroma specific gene expression pattern. In the chorioallantoic membrane (CAM) model, we showed an increased tumor budding and metastatic potential when DAPK1 was lost in tumor cells supporting its role as a tumor suppressor. We propose the DAPK1-ERK axis to be involved in metastasis suppression. Loss of adhesion molecule ICAM1, gain in TACSTD2 expression, and increased tumor cell-ECM interaction under DAPK1 loss seem to be central events in this process.

Nuclear/cytoplasmic fractionation of proteins
Sub-cellular fractions of the HCT116, HCT 7/6, and HCT 21/9 cells were prepared using REAP cell fractionation method 18 . Briefly, cell pellets were resuspended in 500 μl of ice-cold 0.1% NP40 (Calbiochem, CA, USA) in PBS, triturated five times using a p1000 micropipette and centrifuged for 10 s in 1.5 ml micro-centrifuge tubes. The supernatants were transferred to the new tubes and kept on ice (this is the cytoplasmic fraction). The pellets were washed with 1 ml of ice-cold 0.1% NP40-PBS lysis buffer, centrifuged for 10 s, and the supernatants were discarded. The remaining pellet was dissolved in 100 µl 0.1% NP40-PBS lysis buffer (this is the nuclear fraction). All lysates were analyzed by Western Bloting.

WST-8-based cell proliferation assay
Proliferation rate was determined using the colorimetric Cell Counting Kit-8 (CCK-8, Dojindo, Munich, Germany) according to the manufacturer's instructions. Briefly, cells (10 × 10 4 cells/well) were seeded in a 96 well flat-bottom microplate and cultured in 200 µl of culture medium at 37°C and 5% CO 2 overnight to allow adherence and further cultured for 0, 3, 6, 24 and 48 h. After given time points, old medium was replaced by 100 µl fresh medium supplemented with 1% WST-8 reagent and incubated for further 2 h at 37°C. Thereafter, 100 µl supernatant was transferred into a new 96 well and the absorbance was measured at 450 nm using the multilabel reader VIC-TOR TM X3 (Perkin Elmer, Rodgau, Germany). Results are presented as mean ± SEM. A value of P < 0.05 was considered to be statistically significant.

Chicken CAM assay
Fertilized, specific pathogen-free (SPF) chicken eggs (VALO Biomedia, Osterholz-Scharmbeck, Germany) were maintained at 37°C and 80% constant humidity. On day 8, a window of 1.5-2.0 cm diameter was cut in the shell at the more rounded pole of the egg and sealed with tape (Durapore silk tape, 3 M). The next day, 1.0 × 10 6 human tumor cells per pellet were embedded in growth factorreduced matrigel (Corning, Wiesbaden, Germany) serving as matrix and were transplanted onto the CAM. The window was sealed again and eggs were incubated for additional 5 days. Tumor growth was monitored over time using a light microscope (×10, SU 1071 Traveler). Tumors were sampled with the surrounding CAM on day 5, fixed in 4% formaldehyde, paraffin-embedded and cut into 3-5 µm sections for immunohistochemical evaluation. For gene expression and protein expression analysis, fresh tumors were cryopreserved in liquid nitrogen and stored at −80°C for long-term storage (44). For optical imaging, tumor cells were stained with CytoPainter Cell Proliferation Staining Reagent-Deep Red Fluorescence (ab176736, Abcam, Cambridge, UK) as described previously 19 .

Immunohistochemistry
Serial sections (3-5 µm) of 4% formalin-fixed and paraffin-embedded CAM tumors were deparaffinized with xylene and rehydrated in graded alcohol. Validated protocols established for the clinical routine were applied for hematoxylin-eosin (HE), phospho-Histone 3 (pHH3, Cell Signaling, Frankfurt am Main, Germany), pan-Cytokeratin (pan-CK, Thermo Fisher Scientific, Waltham, MA, USA) staining. Anti-ICAM1 and anti-phospho-ERK1/2 Thr202/Tyr204 (both from Cell Signaling, Frankfurt am Main, Germany) and anti-TACSTD2 (Abcam, Cambridge, UK) stainings were performed as follows: for antigen retrieval, slides were cooked in 1 mM Tris-EDTA buffer (pH 8.5) at 120°C for 5 min. After peroxidase blocking (Dako/Agilent, Munich, Germany) for 5 min at RT, slides were incubated with primary antibodies anti-ICAM1 (1:50) and anti-phospho-ERK1/ 2 Thr202/Tyr204 (1:50) for 30 min and with anti-TACSTD2 (1:2000) overnight at RT. ICAM1 and anti-phospho-ERK1/2 Thr202/Tyr204 antibody binding was visualized by incubating the sections with the EnVision Detection System (Peroxidase/DAB, Dako) for 30 min at RT and subsequently with DAB substrate (Dako/Agilent, Munich, Germany) for 10 min at RT. For detection of TACSTD2 antibody binding, the ABC-kit from Vector Laboratieries (Burlingame, CA, USA) was used according to the manufacturer's recommendations. Sections were counterstained with hematoxylin (Merck) for 1 min. Immunohistological stainings were brightfield imaged at a magnification of ×20 with the Olympus BX51 microscope and Olympus XC50 camera (Olympus Corporation, Hamburg, Germany) or were scanned using a Panoramic MIDI system (Camera type: CIS VCC-FC60FR19CL; objective: Plan-Apochromat; magnification: ×40; Camera adapter magnification: ×1, 3DHISTECH, Ludwigshafen, Germany) for digital analysis. Tumor budding 10 of CAM tumors was determined using pan-CK stained sections and the high-power-field (HPF, ×40) method 20 . It was calculated as the average number of buds in 4-10 HPFs per sample. A respective budding score was defined as low-grade with an average of ≤1 and high-grade with an average of >1 buds per 4-10 HPFs. Vessel areas of CAM xenografts were determined in scans of HE stained sections. The tumor as well as the intratumoral vessels were annotated manually using the CaseViewer software (3DHISTECH, Ludwigshafen, Germany). The percentage of vessel area was calculated by relating the total area of intratumoral vessels to tumor area and expressed in percent (%). ICAM1, TACSTD2 and cytoplasmic and nuclear pERK immunoreactive scores (IRS) were determined by multiplying staining intensities (0-3) and the respective percentage of positive cells (0-100%, no positive cell −0; all cells positive -100) according to Remmele and Stegner 1 . The product was divided by 10 to achieve an IRS between 0 and 27. The mitotic rate of cells in the CAM micro-tumors was determined using pHH3 stained sections. Quantification was conducted using the QuPath software (https://qupath.github.io/) semi-automatically analyzing 10 HPFs (×20 magnification) per section. The mitotic index per section is expressed as the mean number of positive mitotic figures in % of 10 HPFs.

Ex ovo optical imaging
After performing the CAM assay using HCT116 and clone 21/9 cells, chicken embryos were sacrificed by decapitation on day 6 of tumor growth and were placed in an in vivo optical whole body imaging system (IVIS Spectrum CT, Perkin Elmer Rodgau, Germany). On the camera of the imaging system, images of the whole embryo were taken and overlayed with optical signals of deep red fluorescence labeled cells (CytoPainter Cell Proliferation Staining Reagent, Abcam, Cambridge, UK) which were acquired with the following parameters: Epiillumination using an excitation filter for 605 nm and an emission filter for 660 nm, exposure time of 2 s and optical fields of view (FOV B) of 6.6 cm. After background correction (embryos with unstained tumor cells), the average radiant efficiency within the embryos was determined by selecting a rectangular ROI that covered the entire embryo. Embryos without tumor cell grafts served as baseline control.

Organotypic brain slice culture
The brain slice culture system using genetically identical slices of 6-day-old rat (Wistar strain, Charles River) brains were utilized in order to study colorectal carcinoma cell invasion. HCT116, clone 7/6, clone 10/8 and clone 21/9 cells were labeled with a green fluorescent vital dye (Abcam, Cambridge, UK) for tracking invading cells and subsequently transplanted onto the brain slices (3.0 × 10 5 cells per brain slice). For each cell line, brain precision-cut tissue slices (PCTS, 3 slices per 6 well) of 5 different rats were used. Slices of a sixth rat without tumor cell transplants served as negative control. Propidium iodide (PI) staining of all brain slices served as control for equal loss of vitality of the brain tissue (data not shown). The integrated fluorescence intensities of live invading tumor cells and PI signals were imaged (×2.0 magnification) and quantified after 1 and 3 days of tumor growth under a fluorescence microscope (Olympus IX71, Olympus, Hamburg, Germany). Data are reported as mean fluorescent intensities per brain slice.

3D-tumor spheroid-based migration assay
For spheroid formation, HCT116 cells and clone 7/6, clone 10/8 and clone 21/9 cells (2.0 × 10 3 cells per 96 well) were seeded into a 96 well round-bottom plate. After 3 days, preformed multicellular spheroids were transferred into an uncoated 96-well flat-bottomed plate (one spheroid per well). Tumor cell migration/dissemination was monitored for 48 and 96 h and imaged using a light microscopy (Leica Dmi1 light microscope, ×10 HI Plan I objective, Leica, Munich, Germany). The area of migration was annotated using GIMP (GNU Image Manipulation Program, Version 2.8) and determined using a selfprogrammed macro using ImageJ 1.46r (National Institutes of Health) software as published before 19 . The assay was performed in technical replicates (n = 8-11). A comparable experiment was conducted using 1.0 × 10 3 cells per spheroid (n = 10-12) showing analogous results.

3D-tumor spheroid-based invasion assay
Upon spheroid formation after 72 h, the invasive potential of tumor cells was analyzed by embedding the generated 3D-tumor spheroids in an artificial ECM. For this, compact spheroids were transferred into a 96-Well ULA round-bottom plate by pipetting 100 µL (using a 1250 µL pipette tip without filter on top of a 100 µL pipette tip) of medium containing one single spheroid into each well. In this way, each well contained only one spheroid. When the spheroids were already generated in a 96-well ULA round-bottom plate it was sufficient to gently remove 100 µL medium from each well. After thawing growth factor reduced Matrigel ® on ice, the 96well plate was placed on ice for 5 min to cool the wells. In this way an early polymerization of the Matrigel® was prevented. Then, 100 µL Matrigel ® were dispensed carefully into each well and mixed with the remaining medium by slowly pipetting up and down several times without introducing air bubbles into the mixture. To ensure that the spheroids were in a central position, the plates were centrifuged at 300g for 3 min at 4°C. The plates were then placed in an incubator at 37°C for 1 h until the Matrigel ® had polymerized. Afterwards, 100 µL of complete growth medium was added to each well. The plates were incubated at 37°C and 5% CO 2 for up to 96 h. Tumor invasion was documented by taking pictures with an inverted light microscope in the brightfield channel at ×4 and ×10 magnification every 48 h.

NanoString sample preparation and nCounter assay
Gene expression analysis was performed using the human nCounter ® PanCancer Progression Panel (Nano-String Technologies, Hamburg, Germany). Total RNA was isolated from frozen cell pellets (48 h culture) by QIAzol-chloroform extraction followed by RNeasy Mini Kit (Qiagen, Hilden, Germany) preparation. Isolated RNA (100 ng) was processed through the NanoString nCounter Prep Station. Briefly, the hybridization reaction (3 µl Reporter CodeSet, 5 µl hybridization buffer, 2 µl Capture ProbeSet and 5 µl total RNA) containing 100 ng RNA (HCT116: n = 3; clone 7/6: n = 2; clone 10/8: n = 3; clone 21/9: n = 1) was conducted for 16 h at 65°C. Subsequently, samples were processed according to the manufacturer's instructions and signals of reporter probes were counted and tabulated using the nCounter Digital Analyzer (NanoString Technologies, Hamburg, Germany). Finally, housekeeping gene (geometric mean of 30 genes) normalization for quantitating gene expression levels, positive control normalization for background noise correction and data analysis was performed using the nSolver™ Analysis Software 3.0 (NanoString Technologies, Hamburg, Germany) and standard settings. The foldchange of counts was determined by averaging results per DAPK1 ko cell line and comparing them to average counts of HCT116 cells. Only significant (P < 0.05) fold changes >−1.5 or > + 1.5 fold were considered as differentially expressed in DAPK1 ko clones compared to the wildtype. Transcripts with a RNA count of <5 of all samples were excluded as they were considered as nonexpressed. For statistical analysis of fold changes, p values of pairwise t-tests calculated by nSolver TM were consulted. Expression profiling data are available online (GEO accession number: GSE130488).

Gene list enrichment analysis
Gene list enrichment analysis of significantly up-and down-regulated transcripts in DAPK1 ko clones vs. HCT116 wt were performed using the Enrichr tool 21 and the integrated KEGG 2016 and REACTOME 2016 geneset databases. Lists of either significantly upregulated (n = 22) or down-regulated transcripts (n = 12) were used as input and only gene sets with an adjusted P value < 0.05 were considered significantly enriched.

Calculation of the stromal scores for RNAs deregulated by DAPK1 ko in HCT116 cells
A "stroma score" for RNAs with either "unaltered" (absolute fold change < 1.05), "down-regulated" (fold change < 1.25) or "upregulated" (fold change > 1.25) expression between DAPK1 ko clones and HCT116 wildtype cells were calculated to illustrate that up and down-regulated RNAs in DAPK1 ko clones have more frequent stromal cell origin (in CRC tumors) as compared to unaltered transcripts, which are primarily of epithelial origin, i.e. deregulated transcripts are likely to affect tumor microenvironment ("cell extrinsic") rather than "cell intrinsic" processes. The "stroma scores" were calculated using a similar approach as in Bramsen et al. 22 and is the fraction of mouse (i.e. stroma cell derived) vs. human (i.e. epithelial cancer cell derived) transcripts in human PDX tumors from mice evaluated by RNA sequencing by Isella et al. 23 (used scores are listed in Supplementary Table 1). Hereby, a high and low stroma score indicate that transcripts are primarily of stroma cell (mouse) or cancer epithelial (human) origin in CRC tumors, respectively.
Pre-ranked gene-set enrichment analysis (GSEA) of the DAPK1 ko expression signature Pre-ranked GSEA was performed using the GSEA V3.0 software 24,25 using standard settings and all gene sets included in the Molecular Signatures Database v6.2. As input we used expression fold change values (DAPK1 ko clones vs. HCT116) for the 770 transcripts profiled by the nCounter ® PanCancer Progression Panel (NanoString Technologies Hamburg, Germany). Gene sets were manually categorized into functional gene-set groups representing biological properties relevant to this paper using the following keywords for each category: ECM matrix (keywords: matrisome, ECM, extracellular collagen); EMT/invasion (keywords: mesenchymal, invasive, epithelial cell migration); integrin pathways (keyword: integrin); cell-substrate adhesion (keywords: matrix adhesion, substrate adhesion, focal adhesion); and healing/ inflammation (keywords: wound, "inflam"); Cell-cell adhesion (keywords: cell cycle); chromosome organization (keywords: chromosome, chromatin). All gene sets containing the terms "positive regulation of", "Negative regulation of", "UP" and "DN" were omitted from the analysis.
Nearest template prediction (NTP) classification of TCGA COREA samples and comparison to their CMS classification status CRC samples from the TCGA project was classified into three categories "DAPK1 ko up sign.", DAPK1 ko down-sign.", and DAPK1 ko unaltered sign." using the NTP module 26 of GenePattern 2.0 27 . As input we used RNA sequencing profiles for 434 COREAD samples acquired from the UCSC XENA Public Data Hubs (https://xena.ucsc.edu/public-hubs/) as log2 (FPKM + 1) normalized RNA expression values for 20.530 genes. As classification templates we used the transcripts with either "unaltered" (absolute fold change < 1.05), "downregulated" (fold change < −1.25) or "upregulated" (fold change > 1.25) expression between DAPK1 ko clones versus HCT116 wildtype cell line. NTP classifications were considered robust for samples with a Benjamini-Hochberg <0.05. Consensus molecular subtype calls for the COREAD samples were acquired provided by the Colorectal Cancer Subtyping Consortium 28 (CRCSC; "CMS final network plus RFclassifier in non-consensus samples". The analysis was restricted to CRC samples for which a CMS annotation was provided by the CRCSC.

ECM cell adhesion assay
The adhesion capacity of DAPK1 ko cells to different protein substrates was analyzed using the ECM Cell Adhesion Array Kit, colorimetric (Merck, Millipore, Darmstadt, Germany) according to the manufacturer's recommendations. Briefly, after rehydration of the plate strips, 0.5 × 10 6 cells per 96 well were plated in triplicates per cell line and incubated for 1 h at 37°C and 5% CO 2 . Thereafter, the supernatant was carefully removed, cells were washed with PBS and 100 µl per well of Cell Stain Solution was added. After an incubation of 5 min at RT, cells were washed with ddH 2 O and were solubilized for 15 min in 100 µl Extraction Buffer. The absorbance was then quantified at 570 nm in the multilabel reader VIC-TOR TM X3 (Perkin Elmer, Rodgau, Germany). BSAcoated wells served as negative control. The data were expressed as BSA-corrected absorbances as percent increase of HCT116 cells.

Second-harmonic generation (SHG) microscopy
We applied SHG microscopy in freshly harvested CAM tissues 3-5 days after tumor cell implantation using an upright multiphoton microscopy system (TriM-Scope II; LaVision BioTec GmbH, Bielefeld, Germany). A femtosecond Ti:sapphire laser was used for imaging at the excitation wavelength 800 nm. The CAM was brought into focus using an HC Fluotar L25x/0.95 W Visir water immersion objective (Leica, Munich, Germany).

STRING bioinformatic analysis
For interpretation of the newly discovered DAPK1 ko gene expression pattern, protein-protein interactions and the interplay with the DAPK1-ERK2 axis were investigated using STRING interaction database (https://stringdb.org/).

Transient siERK2 transfection experiment
To obtain an ERK2 knockdown, HCT116 and DAPK1 ko clone 7/6 and 21/9 cells were grown to 70% confluence in a 6 well culture plate and transfected with Dharma-FECT reagent and 100 nM of siRNA (SMARTpool: ON-TARGETplus Human MAPK3 (ERK2) siRNA (both from Dharmacon, Lafayette CO, USA) according to the manufacturer's instructions and incubated for 48 and 72 h. Transfection with non-targeting SMARTpool siRNA served as negative control. The knockdown efficiency was determined by Western Blotting. ERK2 knockdown experiments were repeated in two independent experiments and representative Western blots are shown.

Statistical analysis
All statistical tests (using two-tailed Mann-Whitney test, one-way ANOVA multiple comparison, unpaired ttest, multiple t-test and Pearson's correlation) were performed using Prism 7 (San Diego, California, USA). Differences were considered statistically significant according to values of two-tailed *P < 0.05, **P < 0.01, ***P < 0.001. Types of the tests are indicated in the figure legends.

Results
Evaluation of HCT116-derived CRISPR/Cas9-mediated DAPK1 ko clones Basal level of DAPK1 protein in the HCT116 cell line was determined using Western blot (Fig. 1a). HCT116 express low to moderate DAPK1 compared to DLD1 and HT29, which express high levels of DAPK1, and in comparison to SW480, SW620 and SW837, where DAPK1 levels were not detected (Fig. 1a). To mirrorimage the loss of DAPK1 protein observed at tumor invasion front of colorectal tumors, we first established CRISPR/Cas9-driven DAPK1 ko HCT116 cell lines using two different single guide RNAs (sgRNA1 and 2) (Supplementary Fig. 1a, b). The presence of Cas9 protein, which serves as evidence for an active CRISPR/ Cas9 system was shown in the lysate of HCT116 cells 24 h after transient transfection with sgRNA1 or sgRNA2 vectors by Western Blot whereas it was no longer detectable in the established monoclonal ko cell clones ( Supplementary Fig. 1c). Sanger sequencing analysis revealed homozygeous and sgRNA-specific insertion mutations resulting in a reading frame shift (Supplementary Fig. 1d). Three monoclonal DAPK1 ko cell lines named clone 7/6, 10/8 and 21/9 were randomly chosen from a panel of several successfully generated subclones and the lack of endogenous DAPK1 protein was verified by immunofluorescence staining (Fig. 1b; Supplementary  Fig. 2a) and Western Blotting (Fig. 1c). To exclude CRISPR/Cas9-caused off-target effects on other highly related DAPK family members, we determined the protein expression of DAPK2 (DRP-1), DAPK3 (ZIPK), DRAK1 and DRAK2 in Western Blotting (Fig. 1c). Steady state protein levels in HCT116 cells and in 10/8 and 21/9 revealed no ko effects on these molecules. The protein levels of pMLC, a well known DAPK1 target, were decreased in the DAPK1 ko clones (Fig. 1c). Regarding "stemness", we found differences between the single clones with clone 7/6 had the highest expression of CD133 and CD44 markers and clone 21/9 did not express CD44 at all (Fig. 1d). No epithelial-to-mesenchymal transition (EMT) was observed in DAPK1 clones visualized by steady E-cadherin protein levels and lack of vimentin expression (Fig. 1d).
In confocal immunofluorescence images we showed that pERK1/2 in HCT116 cells is predominantly localized in the cytoplasm (Fig. 1e), whereas in the three DAPK1 ko clones pERK1/2 expression was remarkably increased in the nuclei of tumor cells in vitro (Fig. 1e). These findings were confirmed analyzing nuclear and cytoplasmatic protein fractions in Western Blot (Fig. 1f) indicating that the loss of DAPK1 is associated with nuclear pERK1/ 2 shuttling under endogeneous conditions. The functional DAPK1 ko was verified in a 3D migration spheroid system. We evaluated the whole areas since the border of the cores were not well defined. At day 4 we confirmed an anti-migratory role of DAPK1 in all three clones ( Supplementary Fig. 3a, b). Moreover, in agreement with our previous findings 29 with DAPK1 as a proapoptotic player, the TNF-α induced phosphorylation of Cofilin was remarkably diminished in DAPK1 ko clones ( Supplementary Fig. 3c). In summary, these findings strongly support the functional ko in our newly generated DAPK1 ko cell lines.
In vivo growth pattern of DAPK1 ko clones Next, we used the CAM assay to examine the role of DAPK1 loss for tumor growth. DAPK1 ko clones and parental HCT116 cells were transplanted onto the chicken CAM and were cultured in ovo for 5 days as schematically illustrated in Fig. 2a. Examples of CAM xenografts are shown in Fig. 2b. Conventional HE staining of HCT116-derived CAM tumors reflected the typical growth pattern of a microsatellite instable colorectal carcinoma with well-differentiated epithelial-like glandular structures and clear margins pushing back the chicken connective tissue (Fig. 2c). In contrast, DAPK1deficient tumors showed a shift to loosely packed tumor masses and a highly infiltrative growth pattern intruding the CAM (Fig. 2c). Pan-CK staining of CAM xenografts was used to identify human CRC cells in the chicken mesodermal layer and to highlight tumor budding (Fig. 2d). The avian epithelial monolayers were also Pan-CK positive. Quantification of peritumoral budding, single cells or small clusters of up to four cells ahead of the invasive front 10 , revealed a trend of an increase in budding in DAPK1 ko tumors at day 5 (not significant, Fig. 2e). When evaluating a budding score with low-grade budding defined as an average of ≤1 and highgrade budding as >1 buds per 10 high-power-fields (HPFs; ×40), the DAPK1-dependent difference was remarkable. Immunofluorescence staining of DAPK1 (green) in parental HCT116 cells and DAPK1 ko clones. Cells were counterstained with phalloidin for F-actin (red) and nuclear Hoechst (blue). Fluorescence microscopy was performed using a x 100 oil immersion objective. Representative images of two independent experiments are shown. Scale bar = 20 µm. c Protein expression of DAPK1 family members DAPK1, DAPK2, DAPK3, DRAK1, DRAK2 and DAPK1 phosphorylation target pMLC in HCT116 wildtype cells and DAPK1 ko clones were detected by Western Blotting using specific primary antibodies. Representative images of two independent experiments are shown. * images were cropped here; all samples were analyzed on the same SDS-PAGE gel; ** GAPDH blot has been used twice see Fig. 5b, proteins have been loaded on the same membrane. d Western Blot analysis of stem cell markers (CD133, CD44) and EMT markers (epithelial marker: E-cad = E-cadherin, mesenchymal marker: Vimentin). Representative images of two independent experiments are shown.*images were cropped here; all samples were analyzed on the same SDS-PAGE gel. e Representative images of endogenous phospho-ERK1/2 (red) levels of immunostained HCT116 cells and DAPK1 ko clones examined by confocal immunofluorescence microscopy (63x; enlarged: cropped and zoomed in). Cells were nuclear counterstained with Hoechst (blue). Immunofluorescence was repeated in two independent experiments and representative images are shown. White arrows: empty nucleus; dashed arrow: nuclear expression of pERK1/2. Scale bar = 50 µm. f pERK1/2 expression analyzed by Western Blot in cytoplasmic (C) and nuclear (N) protein fractions. Representative images of two independent experiments are shown. GAPDH served as loading control in total and cytoplasmatic protein fractions. Lamin A/C was used for nuclear loading control.
DAPK1 loss was associated with high-grade tumor budding (50% of evaluated tumors) demonstrating that every second HPF of DAPK1 negative tumors was given a high-grade score whereas 89% of all HCT116 wildtype tumors were low-grade budders (Fig. 2f). Since CAM assay is a classical in vivo model for angiogenesis, we analyzed neovascularization in HCT116 and DAPK1 ko tumors. DAPK1 ko tumors showed higher vessel area compared to HCT116 ( Fig. 2g; Supplementary Fig. 4) indicating DAPK1 being an anti-angiogenic factor in CRC. Proliferation did not differ in a DAPK1-dependent manner in vitro (Fig. 3a) and in vivo when evaluating the pHH3 positive tumor cell population (Fig. 3b, c) in the CAM model which is in close agreement with the rather similar tumor volumes of CAM xenografts (Fig. 3d).
For in vitro analysis of invasive potential of HCT116 and DAPK1 ko clones 7/6, 10/8 and 21/9 tumor cells, 3Dtumor spheroids were embedded in matrigel and cell penetration from the spheroid core into the environment was monitored over time (Fig. 4a). Even though none of the cell lines developed any invadopodia-like spikes, we observed outgrowth of compact protrusions in HCT116, clone 10/8 and 21/9 after 48 h and even more pronounced bulges after 96 h. Clone 7/6, however, failed to develop protrusions. Applying semi-automated quantification of the invasion area, we investigated that DAPK1 ko clones 10/8 and 21/9 exhibited significantly enhanced (***P < 0.001) invasive capacity compared to HCT116 after 96 h, while clone 7/6 revealed significantly smaller invasive areas after 48 h (*P < 0.05) and 96 h (**P < 0.01; Fig. 4b).
Next we confirmed the highly invasive behavior of DAPK1 ko tumor cells in another 3D ex vivo model, where tumor cells were applied and cultured on precision-cut tissue slices of rat brain (Fig. 4c). These brain slices at least partly reflect the complexity of an organotypic environment. Tracking green fluorochromelabeled tumor cells, we found larger areas of invasion for DAPK1 ko clones one day after transplantation reflecting their increased invasive capability in a physiological 3D matrix in contrast to HCT116 wildtype cells (*P < 0.05; **P < 0.01; Fig. 4c, d). Although a high variability, significance levels were reached for all three ko clones at day one. However, at day three after transplantation, tumor areas reached nearly equal fluorescence intensities (Fig.  4e) suggesting that DAPK1 ko cells have a better early adaptation capability to the 3D environment. Notably, just like the CAM tumors (Fig. 2c), HCT116 cells generated dense tumor masses while DAPK1 ko clones formed more loosely packed and more spacious tumors as reflected by a more diffuse green fluorescence (Fig. 4c). PI staining of empty brain slices and slices with transplanted HCT116 and DAPK1 ko cells for quality control showed equal mean intensities of PI signal at day 1 and day 3 of the experiment (Supplementary Fig. 5a). The slight decrease in tumor area at day 3 might be caused by a loss of fluorescent dye due to proliferation of cells.
Furthermore, we examined if the increase in tumor budding was associated with an enhanced potential to form metastases in the chicken embryonic organs. Since DAPK1 loss is associated with anoikis resistance 30 , we hypothesized that DAPK1 ko tumor cells might have a survival benefit in the vascular system reflected by a higher number of disseminating tumor cells in the embryonic chicken organs. To test this hypothesis, prelabeled HCT116 and DAPK1 ko clone 21/9 tumor cells were transplanted onto the CAM and whole bodies of freshly sacrificed chicken embryos were screened for optical signals using an intravital imaging system (IVIS Spectrum, Perkin Elmer; Fig. 4f, Supplementary Fig. 5b). Clone 21/9 seems to be representative since it showed more migration, more invasion and an elevated number of budding. Results exposed significantly higher average radiant efficiency (**P < 0.01) within animals loaded with clone 21/9-derived DAPK1 ko tumors (Fig. 4f) suggesting that DAPK1 ko cells showed more disseminating tumor cells, preferentially accumulating in the liver, heart, and brain of the chicken embryo.
We inspected deregulated genes to identify biological processes possibly affected by DAPK1 loss in the ko clones. Foremost, gene list enrichment analysis of the 34 significantly deregulated transcripts suggested that they were involved in regulating tumor-stroma interactions, in particular altered ECM organization and matrix adhesion was characteristic for upregulated transcripts (Fig. 6a). In agreement with a role in tumor-stroma modulation, we found that both sets of up-and downregulated transcripts had a significantly higher "stroma score" (i.e. transcripts primarily expressed by nonepithelial cells within the tumor stroma) than unaffected genes when analyzing all transcripts (Fig. 6b). To closer investigate which tumor-stroma processes were up-and down-regulated upon DAPK1 ko we performed preranked gene-set enrichment analysis (pre-ranked GSEA) using all gene sets in the BROAD Database (MsigDB) gene-set collection v6.2. Handpicked gene sets representing the biological properties that correlate with the DAPK1 ko phenotype in Fig. 6c. We found that gene sets for organization and binding to ECM and mesenchymal transition had higher Normalized Enrichment Scores (NESs) in DAPK1 ko clones, whereas gene sets for immune pathways, cell cycle and chromosome/chromatin organization had negative NESs (Fig. 6c). Collectively, these analyses suggested that loss of DAPK1 in HCT116 cells affected transcripts involved in enhancing tumor-stroma binding and mesenchymal transition. Quantification of the invasive potential of n 3D spheroids from two independent experiments (1. experiment: displayed in black; 2. experiment: displayed in blue). All values represent means ± S.D. and statistical analysis was performed using two-way ANOVA followed by Dunnett's multiple comparisons test (HCT116 (•): n = 5/5; clone 7/6 (■): n = 5/5; clone 10/8 (▲): n = 5/5); clone 21/9 (◆): n = 5/5). (*P ≤ 0.05 and ***P ≤ 0.001). c HCT116 and DAPK1 ko cells were implanted on living organotypic brain slices. Areas of tumor cell invasion one and three days after implantation were analyzed. Representative images (×2.0) of fluorescence intensities of two independent experiments are shown. Relative tumor invasion was detected by mean fluorescence intensities (green). d one day (*P < 0.0368, **P < 0.0028 compared to HCT116) and (e) three days after implantation and presented as fold-change to HCT116. Black/gray dots present the first (n = 15), blue dots the second (HCT116: n = 14; clone 7/6: n = 14; clone 10/8: n = 14; clone 21/9: n = 15) experiment. f Optical overlay of representative surface images and fluorescence signal of deep-red fluorescence labeled HCT116 and clone 21/9 cells in chicken embryos 6 days after tumor cell implantation. Embryos with unstained tumor cell implants served as background control. The color bar indicates the range of average radiant efficiency ×10 7 from minimal (black) to maximal (yellow); Average radiant efficiency as detected by fluorescence imaging. B: brain; H: heart; L: lung." HCT116: n = 8; clone 21/9: n = 5. (**P < 0.0062; Mann-Whitney test).
Notably, classification of TCGA CRC samples (COREAD) using NTP suggested that the aggressive CMS4 subtype tumors were indeed associated with the gene signature upregulated in the DAPK1 ko clones as compared to down-or unaffected gene signatures (Fig. 6d).
To support the suggestion that ECM interaction was altered in DAPK1 ko clones we performed an ECM cell adhesion array, which revealed that DAPK1 ko cells significantly benefit from loss of DAPK1 regarding binding to collagen I, II and IV (*P < 0.05, **P < 0.01; **P < 0.01; Fig.   6e). Indeed, SHG images of the surrounding CAM showed differences in collagen alignment for HCT116 and DAPK1 ko tumors. Whereas there were areas of very similar collagen alignment as found in the control CAM without xenografts (Fig. 6f-1 versus 2,6), DAPK1 ko engrafted CAMs showed a higher variation in the extend and arrangement of collagen fibers. DAPK1 ko was majorly associated with a more regular and polarized collagen fiber structure (Fig. 6f-7,8), abundant thicker bundles (Fig. 6f -5,7,8) with network building (Fig. 6f-5,7,8). In CAMs Gene expression analysis was performed using the human nCounter® PanCancer Progression Panel (NanoString Technologies, Hamburg, Germany). Only significant (P < 0.05) fold changes >−1.5 or > +1.5 fold were considered as differentially expressed in DAPK1 ko clones compared to the wildtype. Transcripts with an RNA count of <5 of all samples were excluded as they were considered as non-expressed. The average of clones (clone 7/6: n = 2; clone 10/8: n = 3; clone 21/9: n = 1) vs. HCT116 (n = 3) ± is presented. P value was calculated by nSolver software (Nanostring Technologies) using t-test. Expression profiling data are available online (GEO accession number: GSE130488) engrafted with HCT tumors we could hardly detect any aligned fibers in the matrix and the fibers were oriented in a multitude of directions ( Fig. 6f -2-4).

DAPK1/ERK2 signaling axis targets TACSTD2 expression
It has been reported that ERK when phosphorylating DAPK1 at Ser 735 induces a negative feedback loop since DAPK1´s death domain is then binding to ERK, blocking its nuclear translocation and finally triggering apoptosis induction 31 . To analyze oncogenic players of the DAPK1-ERK2 mediated regulation axis, we performed a STRING analysis including the 34 deregulated DAPK1-dependent genes ( Table 1) and added DAPK1 and ERK2 (MAPK3) into this network (Fig. 7). Indeed, ERK2 was situated in close neighborhood of DAPK1 (Fig. 7). The network itself consisted of two major Three independent experiments per cell line are shown with one dot representing the mean of three technical replicates. Error bars represent mean ± SD. Unpaired t-tests were performed to compare qRT-PCR data. b Western Blot analysis of endogenous TACSTD2 and ICAM1 protein levels in HCT116 cells and DAPK1 ko clones. GAPDH served as loading control. Representative blots of two independent experiments are shown. c qRT-PCR of TACSTD2 (**P = 0.0082) and ICAM1 (***P < 0.001) gene expression in 5-day-old CAM micro-tumors. B2M was used as reference gene. Three independent experiments per cell line are shown with one dot representing the mean of three technical replicates. Error bars represent mean ± SD. Unpaired t-tests were performed to compare qRT-PCR data. d Western Blot analysis of TACSTD2 and ICAM1 expression in CAM tumor tissue (HCT116: 3 eggs, clone 7/6: 3 eggs, clone 10/8: 3 eggs). GAPDH served as loading control. e Representative images of ICAM1 and TACSTD2 immunohistochemistry of CAM micro-tumors (formalin-fixed and embedded in paraffin). Representative images are shown. Scale bar = 50 µm. ICAM1 immunoscores were obtained by multiplying staining intensity with percentage positivity divided by 10 (HCT116: n = 8; clone 7/6: 7, clone 10/ 8: 6, clone 21/9: 4). TACSTD2 immunoscores were obtained by multiplying staining intensity with percentage positivity (HCT116: n = 14; clone 7/6: n = 9; clone 10/8: n = 7; clone 21/9: n = 9). A Mann-Whitney test was performed to calculate significance between two groups (**P = 0.0037, ***P < 0.0001).
nodules: NOTCH1 and ECM modulators. Interestingly, DAPK1 seems to regulate the network from outside linking the two clusters. Notably, so far, CLEC2B, Heg1, GALNT7, CALCRL, GPR124, CGN, and TACSTD2 have not been studied for their association with the DAPK1 network. The DAPK dependent gene signature involved six different dysregulated ECM components (Fig. 7).
inhibition Western Blot analysis showed decreased TACSTD2 levels in HCT116 cells and in 21/9 cells whereas ICAM1 remained undetectable (Fig. 8e) suggesting that DAPK1-ERK2 axis might be partly involved in regulation of TACSTD2, but not in regulation of ICAM1.

Discussion
In this study, we firstly report a novel DAPK1-mediated network associated with metastatic potential of colorectal tumors. We show that under DAPK1 loss, colon tumor cells gain ability to modulate the ECM consequently increasing their metastatic potential in vitro and in vivo. We hypothesize that this hallmark of cancer could be partly driven by the DAPK1-ERK axis.
So far, there have been only a few hints obtained from CRC patient data suggesting that DAPK1 functionally acts as a metastasis suppressor in the colonic epithelium. It has been shown in two independent studies that DAPK1 expression was decreased in tumors that had already metastasized at time of diagnosis 3,4 . In papillary thyroid cancer, pituitary tumors, hepatocellular, and esophageal carcinoma DAPK1 promoter hypermethylation was correlated with advanced tumor stages and worse prognosis [32][33][34][35] . The molecular basis of these observations has only been little understood.
Here, we give novel experimental evidence for DAPK1´s role in migration, invasion, and ECM modulation and antimetastatic potential in CRC cells. DAPK1 seems to exert sensor functions under conditions where an interaction with ECM and basement membrane components is possible. Our study required a long-term and stable gene knockout which was implemented by the CRISPR/Cas9 technology. Slight differences in molecular and functional readouts among the randomly picked DAPK1 ko clones are caused by the monoclonal strategy which was applied to the heterogeneous HCT116 parental cell line. Thus, as a result of mixed or inefficient gene editing (a per-cell rate of 30-60% CRISPR ko is reported), an efficient conversion to an uniform cell population which was required for proper gene ko assessment, can be reached only by singlecell cloning 36 . Although a complete loss of DAPK1 expression was verified, we showed that the genetic background of single clones was varying especially regarding stem cell marker expression. Such genetic background pattern might modify the DAPK1 signature, however, can be mostly neglected in our approach since we only focused on genes that were commonly deregulated in all three DAPK1 clones.
We identified a DAPK1-driven network of 22 genes upregulated under loss of DAPK1 in HCT116 cells. Literature research (Pubmed, September 2019) revealed that 73% (IL11, TACSTD2, CLEC2B, LAMA4, CCL5, SER-PINE1, TGFB2, COL7A1, GALNT7, PLAUR, CGN,  ITGA1, NOTCH1, TBXA2R, LAMA5, TNS1) of upregulated genes found in our study have already been described in previous studies in CRC. In particular, IL11 promotes growth of neoplastic epithelium 37 , and is associated with poor differentiation, a large tumor size, lymph node metastasis and overall low survival of CRC patients 38 . CLEC2B belongs to C-type lectins which facilitate the tumor metastasis in many cancers 39 . Interestingly, two laminins LAMA4 and LAMA5 were upregulated in DAPK1 ko signature 40,41 . LAMA5 is required for growth of hepatic metastases where it promotes branching angiogenesis and regulates Notch signaling 42 . In accordance with the observed increased collagen binding in vitro and in vivo in DAPK1 ko clones, ITGA1, which encodes the alpha 1 subunit of integrin receptors (collagen surface receptor), has been found to be upregulated in the DAPK1 ko clones. Moreover, the negative regulator of the MMP-associated proteolytic network, the serine protease inhibitor SERPINE1 (plasminogen activator inhibitor-1) seemed to be associated with aggressive tumor behavior in CRC 43,44 . It was reported to maintain an angiogenic "scaffold" and stabilizes nascent capillary structure which is in agreement with the observed proangiogenic phenotype under DAPK1 loss.
For 23% (P3H2, CHRDL1, COL5A1, HEG1, AGRN) DAPK1 dysregulated genes their role in CRC is  Table 1. completely unknown. Taken together, the identified overall gene expression signature in the DAPK1 ko cell lines confirms our hypothesis, that DAPK1 might inhibit CRC progression and metastasis.
Then we focused further on TACSTD2 (TROP2), a cell surface receptor that transduces Ca 2+ signals. Although TACSTD2 was found to be highly expressed in many cancer types, only a few very recent reports are available about its functional role in cancer 45 . TACSTD2 has been described as an oncogene promoting cell proliferation, epithelial-to-mesenchymal transition, and metastasis in bladder and colon cancer 46,47 . EpCAM positive circulating breast cancer cells were enriched for TACSTD2 expression linking it to a higher metastatic capability of tumor cells 48 . In our study, we showed in vitro and in vivo that DAPK1 loss led to higher TACSTD2 expression in an ERKdependent manner. Thus, one of the tumor suppressor functions of DAPK1 might be partly mediated by suppressing the metastasis associated TACSDT2. Loss of DAPK1 in tumor buds at tumor invasion front of CRC as we previously described 4 would consequently add more aggressive properties to disseminating tumor cells. This hypothesis is supported by findings that TACSTD2 promotes cell motility in prostate cancer cells by modulating the ß1 integrin signaling and increases wound healing by promoting stem cell survival 49,50 . DAPK1 is known to inactivate integrin β1 dependent matrix survival signals and TACSTD2 would potentiate this signaling, in a DAPK1 loss situation also further reducing the integrin-dependent ICAM1 signaling. Indeed we found an up-regulation of integrin β1 expression (1.22-fold, P = 0.07) when DAPK1 is lost. During metastatic cascade, tumor cells interact not only with several immune cells from the tumor environment, but also with fibronectin, laminin, and type I collagen of the basement membrane. In our study, we showed for the first time that the loss of DAPK1 was associated with a higher collagen binding. The reorganized collagen fiber meshwork of the CAM around DAPK1 ko xenografts seems to trigger cell invasion and distant metastasis formation. This might also explain the better and faster adaptation of DAPK1 loss tumor cells in the physiological environment of PCTS experiment. DAPK1 loss might promote dissemination seen as increase in number of tumor buds and remodeling of the CAM collagen matrix triggering the colonization of tumor cells at a secondary site since we found significantly more metastasis in chicken embryonic organs from DAPK1 ko xenografts.
Tumor progression with metastasis is associated with severe alterations in cell-cell and cell-matrix interactions. One of the genes remarkably deregulated under DAPK Nuclear and (c) cytoplasmic immunoscore of phospho-ERK1/2 expression in CAM micro-tumors obtained by multiplying staining intensity with percentage positivity divided by 10 (HCT116: n = 6; clone 7/6: n = 7; clone 10/8: n = 4; clone 21/9: n = 4). A Mann-Whitney test was performed to calculate significance between two groups (*P = 0.021; ns P = 0.1539). d Western Blot analysis of pERK Thr202/Tyr204 , ERK1/2, PARP and cleaved PARP, ICAM1 and TACSTD2 protein expression in HCT116 cells treated with 0, 10, and 20 µM of ERK1/2 inhibitor (FR180204) for 48 h. Representative images are shown. GAPDH served as loading control. e Effects of siERK2 in HCT116, clone 21/9 and clone 7/6 cells on ERK/1/2, ICAM1 and TACSTD2 protein levels were investigated by Western Blot analysis in comparison to the corresponding non-treated and scrambled-treated cells. Cells were harvested 48 h after transfection. Representative images of two independent experiments are shown. GAPDH was used as loading control.
loss was found to be the cell surface protein ICAM1. Here we describe a role for DAPK in regulating ICAM1 and subsequently in interaction with ECM components of the tumor microenvironment. Indeed, we showed that the loss of DAPK1 and ICAM1 was associated with an upregulation of integrins that are specifically interacting with collagens. Since under DAPK1 loss both ICAM1 transcripts and ICAM1 protein are completely lost we believe in a transcriptional regulation mechanism. ICAM1 has been shown mainly to act in different cell types as a metastasis suppressor [51][52][53] . A lower number of ICAM-1 positive cells has been observed in metastasizing CRC 54 . Under conditions where ICAM1 is lost, the tumor microenvironment is remarkably rebuilt and an increased M2 polarization of macrophages has been observed 55 leading to maintenance of acute inflammation and an attenuated tissue repair 56 . There is still controversy about the role of ICAM1 in tumor metastasis since high ICAM1 expression has been described in advanced melanoma 57 . So ICAM1 might fulfill different functional properties dependent on the cellular subtype and the specific pathologic situation. DAPK1 has been shown to be inactivated by Netrin-1 58 to potentiate tumor-associated vessel formation in metastatic lung cancer 59 . Through modulating the tumor-associated vasculature DAPK might also be involved in shaping the tumor microenvironment.
We have identified a connection between DAPK1, ICAM1 and MAP kinase ERK1/2 in STRING analysis. DAPK1 is a phosphorylation substrate of ERK and vice versa. When DAPK1 is phosphorylated by ERK at Ser735 (interacting via its death domain) it holds ERK in the cytoplasm 31 , thus preventing its nuclear translocation. This reciprocal regulation leads to a positive feedback loop that promotes apoptosis. Thus, we speculate that endogenous DAPK1 loss could lead to an increase in nuclear shuttling of active pERK1/2. Indeed we observed more pERK1/2 protein in the nucleus in 2D immunofluorescence, in cell fractionation, and in vivo in CAM xenografts of DAPK1 ko clones. Since ERK has to translocate into the nucleus to regulate gene transcription, cell proliferation and differentiation 60 , we suggest that DAPK1 loss at the invasion front of CRC should have to do at least in part with an ERK1/2 triggered signaling cascade. Since inhibition of ERK signaling did not affect ICAM1 expression level we suggest that ERK signaling is not majorly involved in DAPK1-mediated ICAM1 expression.
In summary, we have successfully modeled the in vivo situation that DAPK1 is mostly lost at the tumor invasion front of CRC. We suggest a novel and even more comprehensive picture of DAPK1´s antimetastatic functions by giving the first time experimental data that it diminishes an effective tumor cell-ECM interaction. DAPK1 exerts its tumor suppressor function at least partly via suppressing the metastasis associated TACSDT2. The potential of the DAPK1 loss gene signature for metastasis prediction and therapy resistance should be further investigated.