Coronin 2A (CRN5) expression is associated with colorectal adenoma-adenocarcinoma sequence and oncogenic signalling

Coronin proteins are known as regulators of actin-based cellular processes, and some of them are associated with the malignant progression of human cancer. Here, we show that expression of coronin 2A is up-regulated in human colon carcinoma. This study included 26 human colon tumour specimens and 9 normal controls. Expression and localisation of coronin 2A was studied by immunohistochemistry, immunofluorescence imaging, cell fractionation, and immunoblotting. Functional roles of coronin 2A were analysed by over-expression and knock-down of the protein. Protein interactions were studied by co-immunoprecipitation and pull-down experiments, mass spectrometry analyses, and in vitro kinase and methylation assays. Histopathological investigation revealed that the expression of coronin 2A in colon tumour cells is up-regulated during the adenoma-adenocarcinoma progression. At the subcellular level, coronin 2A localised to multiple compartments, i.e. F-actin stress fibres, the front of lamellipodia, focal adhesions, and the nuclei. Over-expression of coronin 2A led to a reduction of F-actin stress fibres and elevated cell migration velocity. We identified two novel direct coronin 2A interaction partners. The interaction of coronin 2A with MAPK14 (mitogen activated protein kinase 14 or MAP kinase p38α) led to phosphorylation of coronin 2A and also to activation of the MAPK14 pathway. Moreover, coronin 2A interacted with PRMT5 (protein arginine N-methyltransferase 5), which modulates the sensitivity of tumour cells to TRAIL-induced cell death. We show that increased expression of coronin 2A is associated with the malignant phenotype of human colon carcinoma. Moreover, we linked coronin 2A to MAPK14 and PRMT5 signalling pathways involved in tumour progression.


Background
Colon cancer is one of the most frequent cancers [1] and its progression is classified into five stages, in which in stage 0 the cancer involves only the mucosa and in stage IV the cancer has spread to a distant organ or set of distant lymph nodes [2]. The majority of colon cancers arise from pre-malignant adenomas. Although new screening methods and early diagnosis have increased survival rates in the past few years, the mortality rates of patients diagnosed with one of the later stages are still very high [3]. Once colon cancer has reached disease stage IV the five-year survival rate drops to 8.1 % [4]. It is therefore of major interest to identify cellular pathways involved in the migration and invasion of colon tumour cells.
Coronin proteins belong to the superfamily of eukaryoticspecific WD40-repeat domain proteins [5]. They play important roles in the regulation of F-actin dynamics in numerous cellular processes including the migration and invasion of tumour cells [6]. Phylogenetic analyses of the coronin family of proteins defined seventeen coronin subfamilies including seven paralogs in mammals [7][8][9]. Mammalian coronin 2A (synonyms are coronin 4, ClipinB, IR10, and CRN5; the latter is used in this study) is a member of the 'short' coronin subfamily containing a single WD40-repeat domain, which adopts the fold of a seven bladed β-propeller [10]. Compared to other well-characterized coronins less information is available about CRN5. In rat mammary adenocarcinoma cells a knock-down of CRN5 led to reduced cell migration velocity and increased size, decreased number, and decreased disassembly of focal-adhesions. Based on the observations that CRN5 interacts with the cofilin-activating phosphatase Slingshot-1 L and knock-down of CRN5 increases the amount of phospho-cofilin, CRN5 has been implicated in the regulation of the focal adhesion turnover rate [11]. In addition, CRN5 has been identified as a component of the nuclear receptor co-repressor (NCoR) complex [12] with a function as an NCoR exchange factor [13]. Here, the interaction of CRN5 via a SIM-motif (small ubiquitin-like modifier (SUMO) 2/3 interacting motif) located in its coiled coil region with SUMOylated liver X receptors (LXRs) prevents NCoR clearance from target gene promoters. In absence of SUMOylated LXRs CRN5 binds to oligomeric nuclear actin enabling NCoR clearance and de-repression of Toll-like receptor-induced inflammatory response genes in macrophages [13].
The protein kinase MAPK14 is stimulated by proinflammatory signals and environmental stresses such as heat shock, irradiation, and ultraviolet light leading to its activation via phosphorylation at Thr180 and Tyr182 by upstream MAPK kinases 3 and 6 [14]. An involvement of MAPK14 in colorectal cancer has been shown in Apc Min colorectal cancer mice, which showed a significant reduction in tumour size when they were treated with the MAPK14 inhibitor SB202190 after azoxymethane induction of colon tumours [15]. However, ablation of MAPK14 in the epithelial cells of the digestive tract of another mouse strain caused development of significantly more tumours [16]. In several human colon cancer cell lines the inhibition of MAPK14 by SB202190 induced growth arrest and autophagic cell death [17].
Protein arginine methyltransferases are important regulators of chromatin structure and gene expression, but are also involved in other cellular processes. PRMT5, a type II protein arginine methyltransferase enzyme, turned out to play a role in malignant transformation [18]. Moreover, PRMT5 functions in growth-promoting and prosurvival signalling pathways. Here, it has been reported that a knock-down of PRMT5 restored the sensitivity of several tumour cell lines to TRAIL-induced cell death [19].
In the present study, we demonstrate that the expression level of CRN5 is associated with the malignant progression of colon carcinoma, and that CRN5 over-expression led to elevated tumour cell migration velocity as well as altered MAPK14 and PRMT5 signalling pathways.

Immunohistochemistry
Tissue sections of paraffin-embedded human colon specimens comprised six low grade adenoma, five high grade adenoma, seven adenocarcinoma, five liver metastases, three lymph node metastases, six normal colon controls, and three normal (non-affected) lymph nodes. The normal colon controls were prepared from adjacent normal tissue of the surgical tissue specimens. After epitope retrieval using pre-heated sodium citrate buffer, the sections were immunolabelled with the mouse monoclonal antibody K77-578-1 directed against human CRN5. This was followed by blocking of biotin using the Avidin-Biotin Blocking Kit (Vector Laboratories). The primary antibody was detected by a biotin-conjugated rabbit anti-mouse secondary antibody (1:200 for 30 min; DAKO). Subsequently, avidin-conjugated horseradish peroxidase (Vector Laboratories) was applied following the manufacturer's instructions and visualized by 10 min 3-amino-9-ethylcarbazole (AEC) chromogen (DAKO) incubation. Sections were then counterstained with haematoxylin (DAKO), dehydrated, and mounted. The use of these human tissue specimens for protein expression studies was approved by the local ethics committee of the University Hospital of Cologne, informed consent was obtained from the patients, and the specimens were procured from the centralized Biobank of the Center for Integrated Oncology (CIO) Cologne-Bonn (http://www.cio-koeln-bonn.de/en/medical-professionals/ biobank/).

Immunoblotting and immunofluorescence imaging
For immunoblotting, protein samples were separated by 10 %, 12 % or 15 % SDS-PAGE under reducing conditions and proteins were transferred onto PVDF membranes by the wet-blotting method [23].
For immunofluorescence imaging cells were grown on 12 mm coverslips, washed with PBS, fixed in 4 % paraformaldehyde/PBS for 20 min at room temperature, permeabilized with 0.5 % Triton X-100 or 0.2 % saponin in PBS (identical results with both permeabilisation protocols), incubated in 0.15 % glycine/PBS for 10 min, washed three times with 0.05 % Triton X-100 or 0.02 % saponin in PBS, and blocked in 1 % FBS/PBS for 30 min. Incubation with the primary antibody was done in 1 % FBS/PBS for two hours, followed by three washes with PBS for five minutes each, incubation with secondary antibody AlexaFluor-488 or -568 diluted 1:500 in 1 % FBS/PBS for one hour, and three more washes with PBS for five minutes each. Finally, coverslips were rinsed in water and embedded in Gelvatol [24]. In order to visualise CRN5 in nuclei, cells were treated with 0.05 % Trypsin for 10 min for epitope retrieval [25] before proceeding to blocking in 1 % FBS/PBS.

Cell migration assays
Cell migration was measured using μ-slide 8-well chambers with silicone culture-inserts (ibidi #80209, #80826). 40,000 cells were seeded in each chamber and grown for 24 h. After removal of the culture-inserts, which leaves a 500 μm cell-free gap, images were taken every 15 min for a period of 20 h. Single cell tracking was done using the "Manual Tracking" (Fabrice Cordelières, Institut Curie, Orsay, France) plugin and cell path were analysed using the "Chemotaxis and Migration Tool" (ibidi) plugin for ImageJ (NIH). Directionality is calculated by the Euclidian divided by the accumulated distance.

Cell fractionation
Purification of nuclei was performed according to [26]. In brief, washed cell pellets were re-suspended in ice-cold PBS containing 0.1 % NP-40, and, after lysis, separated by centrifugation into a cytosolic supernatant and a pellet of intact nuclei; the latter was washed and centrifuged a second time before further use.
Lentiviral transduction of CT26 and U373 cells was performed as described in [27]. In brief, each plasmid of interest was mixed with three lentiviral packaging plasmids and co-transfected into 293TN cells. Resulting lentiviruses were harvested and used to transduce the target CT26 and U373 cells. After transduction CT26 and U373 cells were further enriched by addition of 9.00 μg/ml and 0.75 μg/ml puromycine to the growth medium, respectively.

Microscopy
Immunohistochemistry images were recorded with a Olympus microscope equipped with the photoP software, immunofluorescence images with a Leica TCS SP5/AOBS/tandem scanning system with emission detection in sequential mode equipped with the Leica LAS-AF software (v. 2.6.0.7266), and cell migration images with a Leica DMI6000B TIRF MC system equipped with software LAS-AF (v. 2.0.2.2038) and Hamamatsu EM-CCD camera C9100-02 in bright field (TL-PH) mode.
For pull-down assays glutathione-beads carrying GST-CRN5 or GST as baits were incubated in 50 mM Tris/ HCl pH 8.0, 150 mM NaCl, 0.5 % NP-40 with either TEV-protease tag-cleaved MAPK14 or purified His6-PRMT5 as prey, washed three times with the same buffer and analysed by SDS-PAGE.
For CRN5 co-immunoprecipitation and detection of novel binding partners U373 cells stably over-expressing GFP-CRN5 were used with GFP-antibody coated microbeads as described above. Washing steps were done with 50 mM Tris HCl pH 8.0, 150 mM NaCl, 0.1 % Triton X-100, and proteins were eluted with pre-heated SDS-PAGE sample buffer.
Mass spectrometry was performed as described in [29]. In brief, proteins were separated by SDS-PAGE, stained with Coomassie brilliant blue, and bands were excised and proteins were identified after in-gel digestion with V8 or trypsin protease by liquid chromatography tandem-mass spectrometry employing a HCT ETD II iontrap mass spectrometer equipped with a nano ESI source (Bruker Daltonics, Bremen, Germany). Putative phosphopeptides (neutral loss of H 3 PO 4 ) reported by Mascot searches ("Phosphorylation (ST)" was used as variable modification) were verified by manual inspection of the corresponding MS2 spectra.
Analysis of the MAPK14 signalling pathway 200,000 HEK293 cells over-expressing GFP-CRN5 or GFP were seeded in 6-well plates, grown overnight, exposed to heat shock at 42°C for 0, 15, and 30 min according to [30], washed with PBS, and finally lysed in SDS-PAGE sample buffer for immunoblotting analyses. Cells were counted using a TC10TM automated cell counter (Bio-Rad Laboratories GmbH); GAPDH was used as a loading control.
In vitro methylation assay

CRN5 expression is up-regulated in colon carcinoma
Expression of CRN5 in human colon tumours was investigated by histopathological analyses using our mouse monoclonal antibody K77-578-1. In normal colon tissue neither epithelial nor stroma cells did express CRN5, while tissue-residing lymphocytes showed a strong cytosolic localisation of CRN5 ( Fig. 1a; intense reddish stain marked by double-arrows). In low grade adenoma few epithelial cells were detected, which expressed CRN5 in the cytosol. Upon further de-differentiation into high grade adenoma virtually all tumour epithelial cells were positive for CRN5 (Fig. 1b). In adenocarcinoma, CRN5 was additionally present in the nuclei of a subset of the tumour epithelial cells. Moreover, fibroblasts of the tumour stroma were also positive for CRN5 (Fig. 1c). This pattern of CRN5 expression in both tumour epithelial and stroma cells was also visible in liver and lymph node metastases (Fig. 1d, e). Semi-quantitative analysis showed that the tumour epithelial cells in all tissue specimens of high grade adenoma, carcinoma and metastases were CRN5 positive. In contrast, only in one third of the low grade adenomas CRN5 expression was found, while normal colon epithelial cells were negative for CRN5 (Additional file 1: Figure S1A). CRN5-positive stroma cells were restricted to colon carcinoma and metastases. Presence of CRN5 in nuclei of tumour epithelial cells was only evident in carcinoma and metastases (Additional file 1: Figure S1B). Two antibodies, our mouse mAb K77-578-1 and the commercially available rabbit pAb L-20, were used for immunoblotting in order to compare the expression levels of CRN5 in colon tissue specimens. Both antibodies showed that the CRN5 signal intensity did not increase in colon carcinoma as compared to normal human colon tissue (Additional file 2: Figure S2A, B, left parts). This finding did, however, not contradict the CRN5 expression pattern seen by immunohistochemistry, as in the presence of an already strong expression of CRN5 in tissue-residing lymphocytes in the normal colon tissue (Fig. 1a, left image, double-arrow), the additional expression of CRN5 in colon epithelial tumour cells did not lead to an obvious increase of the overall CRN5 signal intensity in immunoblotting. Both antibodies detected CRN5 predominantly as a band of 48 kDa in normal colon and colon carcinoma tissues (Additional file 2: Figure S2A, B, left parts). This band was the exclusive signal of CRN5 in the WiDr and DLD-1 human colon carcinoma cell lines (Additional file 2: Figure S2A, B, right parts); and also murine CT26 and CMT93 and rat DHD/K12/TRb colon carcinoma cell lines solely showed this 48 kDa band (data not shown). In murine colon and non-colon cell lines (Additional file 2: Figure S2C, D), however, our own mouse mAb K77-578-1 and the three commercially available antibodies, i.e. goat pAb L-20, rabbit pAb H-110 (not shown), and mouse mAb A-3, detected CRN5 at the expected position of 56 kDa (calculated molecular mass 59.8 kDa). Specificity of our mAb K77-578-1 was further verified by detection of a GFP-CRN5 fusion protein (Additional file 2: Figure S2D).

CRN5 is present in nuclei, at F-actin stress fibres, and the front of lamellipodia
To address the re-localisation of CRN5 into the nucleus, which we had seen in the colonic tissue specimens, we used CT26 mouse colon carcinoma cells. A nuclear localisation of endogenous CRN5 was also detected in these cells after epitope retrieval (Fig. 2a). Nuclear CRN5 was further verified by fractionation experiments using HeLa cells, which could easily be transfected to over-express FLAG-CRN5 (Fig. 2b). We also studied the cytosolic localisation of CRN5 in more detail, as a previous study (See figure on previous page.) Fig. 1 CRN5 is over-expressed in human colon cancer and is present in the nucleus of a subset of tumour cells. a Transverse section of normal human colon tissue. Colon epithelium cells (arrowheads) and fibroblasts of the stroma (arrow) did not express CRN5, while tissue-resident lymphocytes exhibited a marked expression of CRN5 (double-arrows). b Section of an adenoma with a low grade (lower right region) and a high grade part (upper left region). In the more benign part only single epithelial cells were faintly positive for CRN5 (white arrowhead). In contrast, all epithelial cells of the high grade part showed a strong cytosolic stain of CRN5 (arrowheads). In both parts stroma fibroblasts were CRN5-negative (arrow), and lymphocytes displayed the marked CRN5 signal (double-arrows). c Section of an adenocarcinoma with virtually all tumour epithelial cells showing a strong cytosolic CRN5 expression (arrowhead). Furthermore, a signal for CRN5 was present in the nuclei of tumour epithelial cells (double-arrowheads). In addition to the lymphocytes (double-arrow), stroma fibroblasts (arrow) were CRN5-positive in the adenocarcinoma. d, e Sections of liver and lymph node metastases, where a strong CRN5 signal intensity was visible in the cytoplasm (arrowheads) and nuclei (double-arrowheads) of the tumour epithelial cells. Tumour stroma cells (arrows) and lymphocytes (double-arrows) also showed CRN5 expression. The images at higher magnification (right column) were taken from the same section except for the high grade adenoma, which was from another tissue specimen suggested the exclusive presence of CRN5-GFP at focal adhesions [11]. We found that the localisation of CRN5 was strongly dependent on the cell type. While GFP-CRN5 indeed co-localised with vinculin at focal adhesions in HeLa human cervical cancer cells (data not shown), it was also present at other actin structures, for example at F-actin stress fibres in A7r5 rat smooth muscle cells and the front of lamellipodia in U373 human glioblastoma cells (Fig. 2c, d). Moreover, we noted that stable overexpression of GFP-CRN5 in U373 cells almost completely suppressed F-actin stress fibres and caused the formation of multiple long extensions enriched in GFP-CRN5 at the rear of the cells (Fig. 2e). These effects were specific for CRN5, as they could not be observed in U373 cells over-expressing the coronin paralog CRN2 (Fig. 2f).

CRN5 expression level correlates with cell migration velocity
To further analyse the role of CRN5 in tumour cell migration, we aimed at knocking down CRN5 in cell lines derived from human colon carcinoma. We used a lentiviral transduction protocol to express shRNA oligos directed against CRN5 in human WiDr and DLD-1 cells. We tested 12 different shRNA oligos, but none of them had a significant effect on the CRN5 expression level, although transduction and expression of a non-target shRNA worked. Two of the shRNAs we used targeted sequence stretches of CRN5 paralogous to sequences successfully used for knock-down of CRN2 [27]. Using a commercially available siRNA pool (Thermo Scientific Dharmacon, containing siRNAs J-011420-05, J-011420-06, J-011420-07, J-011420-08) also did not lead to a CRN5 knock-down. We therefore changed to murine and rat cell lines and employed murine CT26 and CMT93 and rat DHD/K12KTRb colon carcinoma cells for further tests of CRN5-specific shRNAs. Using a shRNA that targets the sequence GGAACGTCTTGGA CATCAT, which is comparable to the one used for RAW264.7 murine macrophages [13] and identical, but independently designed, to the one used for MTLn3 rat mammary adenocarcinoma cells [11], we finally could generate a knock-down of CRN5, however, only in CT26 cells. This CRN5-specific shRNA led to a moderate reduction of CRN5 expression of approximately 40 % both at the mRNA and protein level (Fig. 3a). These cells were subjected to cell migration and in vitro wound healing assays. Analysis of the CT26 non-target shRNA control cells showed a mean migration velocity of 32.1 μm/h and a directionality of 0.66. Knock-down of CRN5 resulted in a significantly reduced migration velocity of 24.1 μm/h and an increased directionality of 0.74. In addition, the overall gap closure was decreased in the knock-down cells, but did not reach statistical significance (Fig. 3c).
When we next tried to over-express CRN5 in the above cell lines, we encountered similar difficulties. Neither by means of transfection nor lentiviral transduction we were able to generate any colon carcinoma cell line with stable over-expression of GFP-CRN5. However, U373 human glioblastoma cells, which we used as a control for the lentiviral transduction procedure, eventually showed a stable GFP-CRN5 over-expression (Fig. 3b). Analysis of these cells showed that U373 cells expressing GFP for control migrated with a mean velocity of 11.9 μm/h and a directionality of 0.8. In contrast, GFP-CRN5 over-expressing U373 cells migrated with an increased velocity of 17.3 μm/h and a decreased directionality of 0.5; the overall gap closure was not significantly changed (Fig. 3d). Although CT26 and U373 cells displayed different basal values for migration, the overall effect of a higher CRN5 expression level was a faster cell migration in conjunction with reduced directionality, while a decrease in CRN5 expression had the opposite effect.

MAPK14 phosphorylates CRN5 at S423
We performed co-immunoprecipitation and pull-down experiments in order to identify novel, tumour-related post-translational modifications and binding partners of CRN5. Mass spectrometry analyses of GFP-CRN5 immunoprecipitated from U373 cell lysates revealed phosphorylation of serine 423 (Fig. 4a). This residue is located within the C-terminal extension of CRN5 and is part of a MAPK14 target consensus motif [31]. To investigate a possible direct interaction between both proteins, pulldown assays were carried out with recombinant purified (See figure on previous page.) Fig. 2 CRN5 localises to the nucleus, F-actin stress fibres, and lamellipodia. a Cytosolic and nuclear (arrows) localisation of CRN5 in CT26 mouse colon carcinoma cells. Detection was done with pAb H-110. Cells were fixed with 4 % paraformaldehyde, permeabilised with Triton X-100, and treated with 0.05 % Trypsin for 10 min for epitope retrieval according to [25]. Secondary antibody, goat anti-mouse AlexaFluor 488; nuclei were visualized with DAPI. b Nuclei of HeLa cells over-expressing FLAG-tagged CRN5 were isolated, and presence of CRN5 was determined by pAb H-110 immunoblotting. For control, the same membrane was labelled with lamin B1 (nuclear envelope) and α-tubulin (cytoplasm) antibodies. c In A7r5 aortic smooth muscle cells GFP-CRN5 localised to F-actin stress fibres; arrows label a cell-cell contact. F-actin was visualised using TRITC-phalloidin. d In U373 human glioblastoma cells GFP-CRN5 co-localised with the Arp2/3 complex (anti-p34 antibody) at the front of lamellipodia (arrows). e, f U373 human glioblastoma cells over-expressing GFP-CRN5 or GFP-CRN2. While GFP-CRN2 was strongly enriched at all F-actin structures including stress fibres, the presence of GFP-CRN5 led to depletion of F-actin stress fibres (asterisk) and induced extensions at the rear of the cells (arrowheads) MAPK14 isolated either from E. coli BL21 or HEK293 cells in conjunction with bacterially expressed GST-CRN5. Note that MAPK14 derived from mammalian cells is phosphorylated and active, whereas the bacterially expressed protein is non-phosphorylated and enzymatically inactive (Fig. 4b). GST-CRN5 was pre-incubated with CRN5 mAb K77-578-1 coupled to Protein G-coated magnetic microbeads and then incubated with GST-MAPK14 or GST for control, both purified from HEK293 cells (Fig. 4c). Similarly, MAPK14 purified from E. coli was incubated with GST-CRN5 or GST coupled to glutathionesepharose beads (Fig. 4d). These experiments showed that CRN5 directly binds to both non-phosphorylated and phosphorylated MAPK14. At the cellular level, we observed a co-localisation of both proteins at the front of lamellipodia (Fig. 4e). Finally, we performed in vitro kinase assays to verify that CRN5 is a substrate of MAPK14. Incubation of both proteins led to a time-dependent increase of the CRN5 phosphorylation signal. Specificity of the reaction was confirmed using the ATP-competitive MAPK14 kinase inhibitor SB202190, which markedly suppressed CRN5 phosphorylation (Fig. 4f).

Over-expression of CRN5 activates the MAPK14 signalling pathway
We next investigated whether CRN5 plays a role in MAPK14 activation. For this purpose we used a set of antibodies to determine the amounts of total and phosphorylated MAPK14, downstream kinases, and substrates. We used HEK293 cells for these experiments, because heat shock treatment of the control cells expressing GFP resulted in a clear, time-dependent increase of MAPK14 phosphorylation. In contrast, HEK293 cells that over-expressed GFP-CRN5 already showed a two-fold increased MAPK14 phosphorylation at basal conditions with no further increase upon heat shock stimulation, and the total MAPK14 content did not change (Fig. 5a). Furthermore, the downstream kinase MAPKAPK2 as well as the cytosolic (Hsp27) and nuclear (activating transcription factor (ATF2) p53) substrates of MAPK14 showed increased phosphorylation levels in the GFP-CRN5 overexpressing cells (Fig. 5b).

CRN5 is a substrate of PRMT5 and increases TRAIL receptor sensitivity
Our co-immunoprecipitation and pull-down experiments also led to the identification of the 72 kDa protein PRMT5 (protein arginine N-methyltransferase 5) as a novel, direct CRN5 binding partner (Additional file 3: Figure S3A, B). We performed 3 H-labelled S-adenosylmethionine in vitro methylation assays and found that CRN5 is a substrate of PRMT5 (Additional file 3: Figure S3C). In a next step, we investigated the effect of CRN5 on TRAIL (TNF-Related Apoptosis-Inducing Ligand)-induced cell death, as a previous study had identified PRMT5 as a novel TRAIL receptor-binding protein, whose knock-down restored the sensitivity of several tumour cell lines to TRAIL-induced cell death [19]. We used HeLa cells, a well-established TRAIL-resistant cell line [32], to investigate the effect of CRN5 over-expression with and without additional PRMT5 knock-down on the TRAIL receptor pathway (Additional file 3: Figure S3D; note, that HeLa cells virtually lack endogenous CRN5). Knock-down of PRMT5 had no effect on cell proliferation in absence of TRAIL, while expression of FLAG-CRN5 increased cell proliferation by approximately 20 % (Additional file 3: Figure S3E, dark columns). Although in our experiments the addition of TRAIL ligand to HeLa cells control-transfected with empty vectors already caused a 40 % reduction of the cell number, knock-down of PRMT5 resulted in a further 10 % decrease (Additional file 3: Figure S3E, light columns left and right panels). Both, over-expression of FLAG-CRN5 alone and in combination with PRMT5 knockdown reduced the cell number by approximately 20 % (Additional file 3: Figure S3E, light columns right panel).

Discussion
Our CRN5 immunohistochemistry analysis revealed an increase of the CRN5 expression level in colon tumour cells during adenoma-adenocarcinoma progression. In addition, our cell migration experiments showed that a higher CRN5 expression level causes an increased tumour cell migration velocity. Although this is the first report on a putative function of CRN5 in the malignant progression of cancer, it seems to be a general pathophysiological feature of coronin proteins. Previous studies (See figure on previous page.) Fig. 3 Expression of CRN5 increases cell migration velocity. a Lentivirus-mediated expression of a CRN5-specific shRNA caused an approximately 40 % reduction of both the CRN5 mRNA and protein levels in CT26 mouse colon carcinoma cells. A representative immunoblot and agarose gel as well as corresponding densitometry values are shown. CT26 control cells expressed a non-target (nt) shRNA. b Lentivirus-mediated GFP-CRN5 over-expression in U373 human glioblastoma cells led to a 3.8-fold increase of the total CRN5 protein level as compared to control cells expressing only GFP. A representative immunoblot as well as densitometry values are shown. c CT26 cells expressing the non-target or CRN5-specific shRNA (knock-down, kd) and (d) U373 cells over-expressing GFP or GFP-CRN5 were used for analysis of cell migration employing in vitro wound healing assays. Plot graphs show two-dimensional migration patterns of individual cells. Column charts indicate mean values and standard deviations of migration velocity and directionality (CT26 cell lines, n = 96 cells each; U373 cell lines, n = 150 cells each) as well as wound closure (CT26 cell lines, n = 6 gaps each; U373 cell lines, n = 10 gaps each) reported an involvement of the paralog CRN2 (synonyms are: Coro1C, coronin 3, hCRNN4) in multiple types of cancer. In human diffuse gliomas the number of CRN2 positive tumour cells was correlated with the malignant phenotype of diffuse gliomas [27], and an increased CRN2 expression level promoted glioblastoma cell invasion [28]. Over-expression of CRN2 also enhanced cell migration in BEL-7402 human hepatocellular carcinoma [33] and in MKN28-NM gastric cancer cells [34]. Other studies showed increased CRN2 protein expression related to progression of dysplastic naevi [35], the clinical progression and pulmonary metastasis of hepatocellular carcinoma [36], the pathogenesis of primary effusion lymphoma [37], and the presence of lymph node metastases in gastric cancer [34]. Increased CRN2 mRNA levels were correlated to further tumour types with poor clinical prognosis [38][39][40][41]. For another coronin paralog, CRN4 (synonyms are: coronin 1A, coronin 1, ClipinA, Taco, p57), published data indicated increased expression levels in breast cancer [42][43][44].
A role of coronin proteins in tumour progression is not surprising, as basic characteristics of various coronin proteins are their interaction with actin polymers as well as the regulation of actin cytoskeleton dynamics [6]. Generally, the actin cytoskeleton is highly important for the processes of tumour cell migration and invasion [45]. In this respect, we found that CRN5 not only localises to F-actin enriched cell compartments including Factin stress fibres, the front of lamellipodia and focal adhesions, but also modulates the actin cytoskeleton by reducing the number of F-actin stress fibres leading to an elevated cell migration velocity. However, coronin proteins do not only regulate actin dynamics via direct interaction with F-actin, but also by modulation of related signalling pathways [46][47][48][49][50]. The finding that CRN4 is involved in calcium homeostasis [51,52] and CRN5 in the expression of inflammatory response genes via NCoR complex clearance in immune cells [13] fulfils the expectation that the β-propeller fold of coronin proteins serves as platform for multiple proteinprotein interactions. We therefore have searched for novel and tumour related CRN5 binding partners by coimmunoprecipitation experiments, mass spectrometry analyses, and pull-down assays. Amongst several newly identified proteins MAPK14 and PRMT5 were related to tumourigenesis. Pharmacological inhibition of MAPK14 inhibited growth of colon cancer cell lines [17] and colon tumour growth in mice [15]. In our experiments the interaction of CRN5 with MAPK14 led to phosphorylation of CRN5. In addition, cells that over-expressed CRN5 showed an activation of the MAPK14 pathway suggesting an involvement of CRN5 in tumour progression via the MAPK14 signalling pathway. With regard to PRMT5 it has been reported that a knock-down of this methyl transferase restored the sensitivity of several tumour cell lines to TRAIL-induced cell death [19]. We found that CRN5 is methylated by PRMT5 and that overexpression of CRN5 also increased the sensitivity to TRAIL-induced cell death. This latter, rather antioncogenic effect might become relevant in conjunction with cancer treatment.
Taken together, our present work links the expression level of CRN5 with the malignant progression of colon cancer. Further, our results suggest that CRN5 affects tumour-related cell functions on several levels, directly on the actin cytoskeleton and more indirectly via modulation of the MAPK14 and PRMT5 signalling pathways.
(See figure on previous page.) Fig. 4 MAPK14 phosphorylates CRN5 at serine 423. a GFP-CRN5 was immunoprecipitated from lysates of U373 cells using mAb K77-578-1. The GFP-CRN5 band at 83 kDa was cut from the Coomassie stained gel, subjected to mass spectrometry, and a phosphorylation was identified at serine 423 (highlighted in black). Underlined, MAPK14 phosphorylation consensus target motif. Amino acid sequences highlighted in grey indicate the sequence coverage by mass spectrometry. b GST-MAPK14 expressed in E. coli BL21 and HEK293 cells. MAPK14 derived from the latter is phosphorylated at Thr180 and Tyr182 as determined by phospho-MAPK14 pAb D3F9, and is visible at a slightly higher molecular weight than the non-phosphorylated protein derived from the bacteria. c Co-immunoprecipitation studies were carried out using the CRN5 mAb K77-578-1 coupled to Protein G-coated magnetic microbeads (Miltenyi), which were first incubated with GST-CRN5 purified from E. coli BL21 and afterwards incubated with either GST-MAPK14 purified from HEK293 cells or GST purified from E. coli BL21 for control. Analysis was done by GST rabbit pAb [21] immunoblotting. d A pull-down assay was performed using GST-CRN5 or GST coated glutathione-sepharose beads together with non-tagged MAPK14. All three recombinant proteins were purified from E. coli BL21 as GST-fusion proteins; from MAPK14 the tag was cleaved by TEV-protease. Analysis was performed using CRN5 mAb K77-578-1, MAPK14 rabbit pAb, and the GST pAb. c, d For illustration purposes the order of lanes from the original immunoblots were digitally re-arranged to omit dispensable lanes. e Immunofluorescence images show an enrichment and co-localization of CRN5 and MAPK14 at the front of lamellipodia (arrows). U373 cells stably over-expressing GFP-CRN5 were methanol fixed; detection of MAPK14 was done with pAb #9212 and secondary antibody goat anti-mouse AlexaFluor 568. f In vitro kinase assay employing full-length GFP-CRN5 and GST-tagged MAPK14 purified from HEK293 cells. Time course (10, 20, 60, 90 min) of phosphate incorporation from [γ-32 P]ATP into GFP-CRN5. As controls the selective, ATP-competitive MAPK14 kinase inhibitor SB202190 was used, and samples lacking GST-MAPK14 or GFP-CRN5. Left panel, autoradiograph ( 32 P); right panel, corresponding Coomassie brilliant blue stained gel. Note, that the part of the autoradiograph showing the time-dependent phosphorylation of GFP-CRN5 is presented with increased contrast settings (dotted rectangle), because of the high background signals from MAPK14 autophosphorylation

Conclusions
This is the first time it has been reported that increased expression of coronin 2A/CRN5 is associated with the colorectal adenoma-adenocarcinoma progression. Over-expression of CRN5 does not only induce cytoskeletal alterations, but also changes of the MAPK14 and PRMT5 signalling pathways. The present study adds further evidence that members of the coronin family of actin regulatory proteins play important roles in cancer progression.  5 Over-expression of CRN5 activates the MAPK14 signalling pathway. a HEK293 cells stably over-expressing GFP for control (left upper panel) or GFP-CRN5 (right upper panel) were exposed to heat shock (0, 15, 30 min at 42°C). Levels of Thr180/Tyr182-phosphorylated MAPK14 (pp38α) were determined by immunoblotting. Total MAPK14 (p38α, time point 0 min/without heat shock) and GAPDH of the heat shock experiment are shown for control (right lower panel). Densitometric analysis from five independent assays at time point 0 without heat shock (column chart with mean values and standard deviations; values were normalised to the GAPDH signals and indexed to GFP cells) revealed a two-fold increase of MAPK14 phosphorylation in cells expressing GFP-CRN5. b Analysis of the phosphorylation level of selected substrates downstream of MAPK14 revealed an activation of the MAPK14 pathway in GFP-CRN5 expressing HEK293 cells (Thr334-phosphorylated MAPKAPK2, Ser82-phosphorylated Hsp27, Thr71-phosphorylated ATF2, Ser392-phosphorylated p53). Loading control, see GAPDH in (a, right lower panel)