Uncoupling of EGFR–RAS signaling and nuclear localization of YBX1 in colorectal cancer

The transcription factor YBX1 can act as a mediator of signals transmitted via the EGFR–RAS–MAPK axis. YBX1 expression has been associated with tumor progression and prognosis in multiple types of cancer. Immunohistochemical studies have revealed dependency between YBX1 expression and individual EGFR family members. We analyzed YBX1 and EGFR family proteins in a colorectal cancer (CRC) cohort and provide functional analyses of YBX1 in the context of EGFR–RAS–MAPK signaling. Immunohistochemistry for YBX1 and EGFR family receptors with two antibodies for YBX1 and EGFR were performed and related to clinicopathological data. We employed Caco2 cells expressing an inducible KRASV12 gene to determine effects on localization and levels of YBX1. Mouse xenografts of Caco2-KRASV12 cells were used to determine YBX1 dynamics in a tissue context. The two different antibodies against YBX1 showed discordant immunohistochemical stainings in cell culture and clinical specimens. Expression of YBX1 and EGFR family members were not correlated in CRC. Analysis of Caco2 xenografts displayed again heterogeneity of YBX1 staining with both antibodies. Our results suggest that YBX1 is controlled via complex regulatory mechanisms involving tumor stroma interaction and signal transduction processes. Our study highlights that YBX1 antibodies have different specificities, advocating their use in a combined manner.

INTRODUCTION Y-Box-binding protein 1 (YBX1) is the most prominent member of the Y-Box-binding protein family, comprising of transcription factors binding to DNA sequences called Y-Boxes. 1-3 YBX1 has been associated with multiple cancerrelated processes such as DNA-repair, extracellular stress response, 4-7 transcriptional 4,8-10 and translational control 8,10-12 as well as cell proliferation. 3,13 YBX1 was suggested to be a prognostic clinical biomarker in different cancer types and correlated with poor prognosis in breast cancer, 14,15 lung cancer, 16,17 multiple myeloma, 18 osteosarcoma, 19 synovial sarcoma, 20 prostate cancer 21 and in ovarian cancer. 22 Recently, Woolley et al., 23 challenged the relevance of various prognostic data due to the fact that different YBX1 antibodies recognize specific but distinct epitopes and thereby provide incompatible information on YBX1 expression, nuclear or cytoplasmic localization.
YBX1 has also been singled out as an experimental therapeutic target. SiRNA-dependent knockdown unraveled a functional role of YBX1 in tumor invasion, proliferation and apoptosis. 18,24,25 YBX1 is involved in chemotherapy response. 18,19,26 Intracellular localization is essential for YBX1 function and a complex regulation underlies the translocation of YBX1 between cytoplasm and nucleus. Bargou et al. 27 reported that nuclear localization of YBX1 is associated with drug resistance in human breast cancer. In lung cancer, nuclear localization of YBX1 is correlated with EGFR and LRP (lung resistance protein) expression. 28 In colorectal cancer (CRC), YBX1 has been identified as a prominent mediator of malignant properties downstream of the EGFR-RAS-MAPK signaling cascade. 29 NFY/YBX1-binding sites were overrepresented among RAS/MAPK target genes, YBX1 binding was enhanced on a group of RAS/MAPK target genes in KRAS-transformed CRC cells and nuclear YBX1 staining was detected in lung metastasis of CRC. A subgroup of genes targeted via YBX1 in CRC had been identified earlier as YBX1 target genes in breast cancer. 30 Several levels of interaction exist between YBX1 and receptor tyrosine kinases of the EGFR family, consisting of EGFR (also known as ERBB1 and HER1) ERBB2 (also known as HER2), ERBB3 (HER3) and ERBB4 (HER4). In breast carcinoma cells, but not in CRC cells, YBX1 was found to bind to the promoter and act as a transcriptional activator of the EGFR gene. 31 YBX1 mediated resistance to anti-ERBB2 therapy via a complex, RSK-dependent mechanism 32 and prevents apoptosis in ERBB2-overexpressing breast cancer cells. 33 In contrast to the well-known link between YBX1 and EGFR in breast or lung cancer, there is little knowledge about the interaction of YBX1 and the EGFR family in CRC.
The aim of this study is to examine a potential prognostic correlation between YBX1 and/or EGFR family expression in a large colon carcinoma cohort. We applied two antibodies against different epitopes of the YBX1 protein (YBX1n 27 and YBX1c 3 ) and examined the staining patterns. We also investigated YBX1 expression and its dependency on RAS signaling in KRAS-inducible transgenic colon cancer cell lines and mice, to provide a rationale for the observed sub-cellular localization of YBX1 in CRC tissue.

Expression of YBX1 and EGFR family members is independent in CRC
We performed immunohistochemical staining for YBX1 in a cohort of 423 CRC samples using 2 different YBX1 antibodies directed against the C-terminal (YBX1c) and the N-terminal domain (YBX1n).
Assessing YBX1c antibody, we found weak cytoplasmic staining in 169 cases (39.9%), moderate staining in 76 cases (18.0%) and strong staining in 82 cases (19.4%). Ninety-six cases were negative (22.7%). We observed focal nuclear staining in six cases (1.4%; Figure 1a upper panel). With YBX1n antibody, we observed weak cytoplasmic staining in 234 cases (55.3%), moderate cytoplasmic staining in 42 cases (9.9%) and strong cytoplasmic staining in 17 cases (4.0%). One hundred thirty cases (30.8%) were negative. Focal nuclear staining by YBX1n was observed in four cases (1.0%; Figure 1a lower panel). These data indicate that antigen detection is stronger using the YBX1c antibody. In full section tumor tissue slides, we found a homogeneous staining result throughout the tumor. We did not detect differences in staining at the invasion front (Supplementary Figure S8). In total, the 2 YBX1 antibodies, despite being directed against the same protein, displayed discordant staining in 250 cases ( Figure 1b). Sixty-one cases showed a low-expression pattern and 32 cases showed a high-expression pattern detectable with YBX1c, whereas YBX1n did not detect the antigen in these cases. Fifty-two specimens showed a low-expression pattern and seven showed a high-expression pattern by staining with YBX1n, whereas YBX1c was negative in these cases. Nuclear YBX1 protein was equally detected by YBX1c and YBX1n in three specimens (30%). In one case, YBX1n staining showed a focal nuclear pattern, whereas YBX1c staining did not detect nuclear YBX1.
The EGFR receptor revealed a diffusely distributed cytoplasmic granular pattern (Figure 2a). Two hundred thirty six cases (55.8%) were weakly positive, 61 cases (14.4%) stained moderately and a strong cytoplasmic staining was observed in 19 cases (4.5%), while 107 cases (25.3%) were negative. The ERBB2 receptor displayed a regular distributed cytoplasmic staining, with 258 cases (61.0%) showing weak cytoplasmic staining, 63 cases (14.9%) stained moderately and 15 cases (3.5%) showed a strong cytoplasmic staining ( Figure 2a). No staining was observed in 87 cases (20.6%). C-terminal ERBB4 antibody showed a weak cytoplasmic signal in 219 cases (51.8%), moderate staining in 53 cases (12.5%) and strong staining in 3 cases (0.7%; Figure 2b). No staining was observed in 148 cases (35.0%). The N-terminal ERBB4 antibody showed weak cytoplasmic staining in 144 cases (34.0%), moderate cytoplasmic staining in 34 cases (8.0%) and strong cytoplasmic staining in 18 cases (4.3%). No staining was observed in 227 cases (53.7%; Figure 2b). The 2 ERBB4 antibodies displayed discordant staining in 266 cases (Supplementary Figure S2). One hundred fourteen specimens showed a low-expression pattern and 28 cases showed a high-expression pattern detectable by ERBB4c staining, whereas the ERBB4n antibody showed no staining in these cases. ERBB4n staining showed low antigen expression patterns in 56 cases and high-expression patterns in 7 specimens, whereas ERBB4c showed no staining in these cases. As expected, nuclei were never stained with any antibody directed against EGFR family members. Data for ERBB3 in CRC will be presented elsewhere (manuscript in preparation). Full section tumor tissue slides showed only faint and inconspicuous intra-tumor staining differences. ERBB4 antibodies showed more differences in 10% (ERBB4n) and 30% (ERBB4c) of the tumor tissue (Supplementary Figures S3 and S4).
We correlated the staining patterns for YBX1 and the EGFR family members using a χ 2 -test. Importantly, we could not detect a significant correlation between YBX1 expression and patterns obtained for any of the ERBB proteins, also irrespective of the intracellular localization of YBX1 (Table 1). Comprehensive assessment of discrepancy levels between the antibodies confirmed previous findings (Supplementary Figure S5). The vast majority of tumors showed only 1-5 of 12 score points expression difference between the antibodies. YBX1n and EGFR displayed the highest number of cases with concordant staining (n = 115/27, 2%), YBX1c and EGFR had the fewest number of cases with concordant staining (n = 49/11, 6%).
Statistical analyses of clinical parameters and survival analysis χ 2 -tests indicate that YBX1n positivity, but not YBX1c positivity, obtained in primary CRC reached a borderline significant negative correlation with the propensity to metastasize (Pearson-χ 2 : 4.09; P = 0.043), that is, a lower number of metastasized tumors were scored YBX1n positive than expected (residual value: − 6.4; Table 2A). Patients with YBX1n-positive tumors (independent of the localization) exhibited increased survival compared with patients with YBX1n-negative tumors (P = 0.016; Supplementary Figure S6C). A tendency (P = 0.45) is also visible on splitting tumors into low-and high-YBX1 expression subgroups (Supplementary Figure S6D). YBX1c staining did not show significant association with clinical parameters and Kaplan-Meier survival analysis showed no advantage in survival for YBX1c-negative tumors (P = 0.977), nor was there a significant difference between high-and low-expression subgroups (P = 0.922; Supplementary Figure S6A and B).
ERBB2 expression is statistically significantly correlated with tumor localization (P = 0.005), with a higher number of ERBB2-positive cases in the left colon. Further correlations between EGFR family members (ERBB1; ERBB2; ERBB4) with clinicopathologic data were not significant (Tables 2B and C). Multilinear regression analysis showed that expression patterns obtained with antibodies against YBX1c, YBX1n and the EGFR family proteins are not correlated with clinicopathologic data (Supplementary table S1).

EGFR/RAS signaling can modulate YBX1 localization in vivo in CRC cells and intestinal tissues
The immunohistochemical investigation revealed a low number of specimens with nuclear YBX1, although we have found nuclear YBX1 in a limited set of pulmonary metastases of CRC before. 27 We therefore investigated functional mechanisms contributing to differential YBX1 localization.
We used Caco2 CRC cells harboring an inducible KRAS V12 oncogene/green fluorescent protein (GFP) transgene and tested YBX1 expression via immunofluorescence. In non-induced cells, we observed strictly distinct staining patterns using YBX1c and YBX1n antibodies. The YBX1c antibody displayed a cytoplasmic perinuclear staining, while the YBX1n antibody stained the protein in the nucleus (Figure 3a). Following RAS induction, the YBX1n-positive signals first accumulated in a few nuclei ( Figure 3b) and after 96 h a strong and condensed nuclear YBX1n signal was detected, whereas the YBX1c signal remained perinuclear. Both, YBX1c and YBX1n staining intensity increased after KRAS V12 induction in a subset of cells. Other cells in direct neighborhood did not show increased staining despite efficient RAS activation (as judged by GFP fluorescence linked to KRAS).
We also tested the localization of YBX1 in the intestine of transgenic mice harboring an inducible KRAS V12 transgene. In the non-induced intestine, both antibodies showed robust YBX1 protein expression, which was cytoplasmic in the villus but cytoplasmic and nuclear in the crypt compartment. In contrast, 4 days following induction of the KRAS V12 transgene both antibodies displayed combined cytoplasmic and nuclear signals throughout the crypt-villus axis (Supplementary Figure S7). To test a potential increase in YBX1 total protein levels following RAS or EGFR activation, we performed western blot analysis of Caco2 cells Y-box-binding protein 1 in colorectal cancer F Roßner et al  Y-box-binding protein 1 in colorectal cancer F Roßner et al after KRAS V12 induction and after addition of the EGFR ligand TGFα. We also tested the impact of MEK inhibition. Transgenic RAS expression became visible 24 h following doxycycline-induction and was most prominent after 48 h ( Figure 4). Concomitant activation of MAPK signaling was evident via increased pERK levels, however, there was no change in YBX1c-positive or EGFRpositive cells. Likewise, treatment of the cells with TGFα-induced pERK levels ( Figure 4a, 30 min; 4 h), however, there was no effect on YBX1c staining. Similar results were obtained using the YBX1n antibody, albeit with weaker intensity due to a lower sensitivity of the antibody on western blots ( Figure 4b). These results indicate that nuclear localization of YBX1 is detectable prior to, but accentuated, following KRAS activation and are consistent with a preferential detection of nuclear YBX1 using the YBX1n antibody. The total level of YBX1 protein, however, is not increased following activation of the EGFR-RAS-MAPK axis.
The rather uniform localization of YBX1 in the nucleus of KRASinduced Caco2 cells in vitro using the YBX1n antibody was in contrast to our previous finding of low numbers of CRC specimens, in which nuclear YBX1 was detectable. We therefore asked whether the tumor microenvironment could modulate YBX1 expression. We determined YBX1 expression and localization in xenografts from KRAS V12 -inducible Caco2 in the absence and presence of doxycycline, yielding KRAS V12 -negative and -positive tumors. We found no significant difference in YBX1 and EGFR mRNA levels between KRAS V12 -negative and KRAS V12 -induced mouse xenograft samples (data not shown). At the protein level, we detected a moderate YBX1 staining in non-induced xenograft tumors. The YBX1c staining pattern was largely cytoplasmic in non-induced xenografts. Most interestingly, there was a clear nuclear staining detectable using the YBX1n antibody (such as tumor C3, Figure 5). Staining was stronger in xenografts derived after KRAS V12 induction as revealed with the two antibodies ( Figure 5). YBX1c staining remained mainly cytoplasmic, however, occasional focal nuclear staining was visible ( Figure 5, upper panel, inserts C2 and C3). YBX1n showed a more variable staining pattern within and between the individual mouse samples. The heterogeneity of YBX1 protein expression and spatial distribution observed with both antibodies was significantly higher in Caco2 xenografts as compared with the cell cultures. This indicates that in vivo, both YBX1 protein levels and intracellular localization are most likely controlled via additional stromal effects, beyond activation of KRAS.

DISCUSSION
With this study we aimed to scrutinize the role of YBX1 as a prognostic marker in a large cohort of CRCs, and to determine any functional connection between EGFR, RAS-MAPK signaling and YBX1 expression. We observed striking differences between YBX1 patterns detected either with YBX1n and YBX1c antibody (Figure 1b). It was found previously that YBX1 staining can depend on the choice of antibody. Woolley et al. 23 assigned these variations to different affinities to the target and affected by phosphorylation-induced conformational changes of the protein. [34][35][36][37] Yet, this theory was disproved by testing phosphorylated YBX1 in complexes immunoprecipitated with either antibodies. 23 We saw that differences in antigen staining patterns between YBX1c and YBX1n antibodies extended beyond clinical specimens, as they were also obvious in immunofluorescence analyses of Caco2 cells. The molecular events controlling YBX1 shuttling and sub-cellular localization have been delineated only incompletely until now. Raffetseder et al. 38 described that YBX1 nuclear localization is actively mediated via the splicing factor SRp30c/ SRSF9. Furthermore, Stein et al. 39 described a heat-shock-induced, rapid nuclear accumulation of YBX1-inducing MDR1 transcription, which declined again within 4 h following heat shock. Both examples indicate that cytoplasmic-nuclear shuttling of YBX1 is a dynamic process. Woolley et al. 23 demonstrated by immunoprecipitation and gel electrophoresis that the antibodies against YBX1 recognize the phosphorylated protein presumably in different complexes. The authors concluded that one of the YBX1 epitopes is likely to be masked within one of the complexes in the nucleus under conditions of immunohistochemical staining. Our analysis revealed several cases in which the YBX1c and YBX1n reacted discordantly on the same tumor under identical processing conditions (Figure 1b). This observation could indicate a tissue processing effect affecting the two epitopes in a different way and rendering either the nuclear or the cytoplasmic YBX1 inaccessible. Thus, a negative YBX1 staining with only one of the two antibodies does not indicate that YBX1 is not expressed or present at very low level, but rather suggests an intracellular molecular complex formation, which cannot be disregarded as an artifact of tissue processing. Elucidation of the functional relevance of these complexes will require detailed biochemical analysis.
In our human tumor cohort, we observed low numbers with focal nuclear staining of YBX1c and YBX1n. Regardless of the different results obtained with the antibodies employed, we conclude that nuclear localization of YBX1 is rare in primary CRC. Other reports also describe low percentages of nuclear YBX1 using immunohistochemistry in breast cancer cohorts 23,40 The overall (cytoplasmic plus nuclear) presence of YBX1 seemed sufficient for establishing prognostic relevance in several tumor entities. [41][42][43] A separate assessment of nuclear and cytoplasmic staining might not be necessary. With our analysis, we were not able to define a significant prognostic impact for YBX1 staining in primary colon cancer. The current results qualify our previous study, suggesting a trend (P = 0.076) towards reduced survival of patients with strong YBX1 cytoplasmic staining in an independent set of 118 CRC Y-box-binding protein 1 in colorectal cancer F Roßner et al samples and a significant correlation between high nuclear staining and reduced survival in pulmonary metastases (P = 0.005). 29 We suggest that correlations between nuclear YBX1 and patient outcome currently cannot be established in primary CRC possibly owing to the fact that YBX1 translocation to the nucleus is transient and considerably rare in non-metastatic colonic tissue specimens. Therefore, cytoplasmic YBX1 staining likewise does not represent a robust prognostic marker. Our results on tissue micro arrays (TMAs) are strengthened further, since staining of full section tumor tissues did not reveal significant expression heterogeneity, whether it is intratumoral or at the invasion front, for both YBX1 antibodies (Supplementary Figures S3 and S8). Immunohistochemical analysis of EGFR, ERBB2 and ERBB4 receptors revealed distinct patterns for the different receptors ( Figure 2). Full section tumor tissue showed only faint and defined intra-tumor discrepancies for ERBB4 in low percentages of tumor tissue, but eventually these are more likely fixation artefacts than true expression differences (Supplementary Figure S4). In contrast to findings in NSCLC and breast cancer, we could not define a statistically significant correlation between expression of YBX1 and EGFR family receptors in our large CRC cohort. 28,44,45 Specifically, neither a correlation between YBX1c/n with EGFR staining nor an impact onto EGFR in Caco2 cells was visible. While this indicates that YBX1 and EGFR are uncoupled in CRC, further functional testing is required to exclude this relationship, for example, at the level of receptor phosphorylation. While Fujii et al. 46 described ERBB2 dependency on nuclear YBX1 in breast cancer, another study stated that ERBB2 is not dependent on YBX1 expression in breast cancer. 41 Similar to the latter study, no correlation could be determined between YBX1 and ERBB2 in CRC in our cohort. With some borderline significant exceptions, we did not find compelling links between YBX1 and the EGFR family with clinicopathological parameters in CRC.
Further investigations of the mechanisms of YBX1 expression and localization in vitro and in vivo revealed that oncogenic KRAS induction can lead to a condensed expression of YBX1 in Caco2 cells in vitro and to an increased nuclear shuttling in intestinal epithelial cells in the mouse in vivo. These results provided compelling evidence that YBX1 activity is influenced and functionally tied to a strongly activated RAS/MEK/ERK signaling pathway in certain experimental settings. We found highly variable patterns of YBX1 protein localization in mouse xenografts derived from the same KRAS-inducible Caco2 cells used for in vitro experiments, whereas gene expression levels of YBX1 and EGFR seemed not to be dependent on RAS signaling. These important results suggest that YBX1 expression and localization is controlled by a complex and as yet underappreciated regulatory network, and may strongly depend on the microenvironment of tumor cells. It is important to notice that prognostic statements and conclusions regarding YBX1 cannot be easily made without appreciation of contextual molecular data, the histological setting and the antibodies applied as diagnostic tools.

Patient cohort
Our study was carried out on 423 cases of CRC paraffin-embedded tissue specimens (years 1995-2012). All patients provided informed consent at the Charité University Hospital for non-commercial and fully anonymous tissue use for research purposes. Two hundred and thirty five patients (55.6%) were male, 188 (44.4%) were female. Tumors were derived from different parts of the colon. Three hundred and three cases had a Grade 2 status (71.6%). 249 cases (58.9%) infiltrated the subserosa (pT3 stadium). In 40 cases (9.5%), the tumor was metastasized. KRAS mutation status was available for 380 samples. About 53 tumors were KRAS mutated (13.9%). BRAF mutation status was available for 58 samples and 5 tumors were BRAF mutated (8.6%) ( Table 3).

Immunohistochemistry
We manufactured TMAs from the study cohort tissue. To evaluate intra-tumor expression discrepancies and to thoroughly exclude a sampling bias, we also evaluated representative full section tumor tissues with high (score: 12) expression profile. TMAs were manufactured as described. 47 Three representative regions with vital tumor tissue within the donor block were chosen. TMAs, Caco2 cell line xenografts and mouse intestine sections were cut into 2-μm slices and placed on glass slides. Excessive paraffin was melted off in a microwave at 70°C. Tissue slides were incubated three times in 100% Xylol and descending ethanol series of 100, 96, 90 and 80% and twice in 70% ethanol each followed by aqua bidest washing. Antibody-dependent citrate-(1,97 mM, pH 6,0; Merck Millipore, Merck KGaA, Darmstadt, Germany) or EDTA-buffered (1,71 mM, pH 7,8; Sigma-Aldrich Corporation, St Louis, MO, USA) antigen-retrieval and cooling was performed according to the manufacturer's product guide followed by two aqua bidest/TBS-washing steps. Blocking solution was coated on HybriSlips (Sigma-Aldrich Corporation) and the slides were incubated. Blocking solution was removed with aqua bidest and TBS-washing solution and the primary antibody was applied overnight at 4°C. The YBX1 antibodies were previously described. 3,27 The appropriate dilutions for YBX1 antibodies 27 against the C-terminal epitope (YBX1c) and the N-terminal epitope (YBX1n) were tested in distinct experiments on tumor tissue and normal mucosa. EGFR (rabbit monoclonal EGF Receptor   chromogen substrate diluted in buffer solution was incubated on the slides and a proper staining degree was determined with light microscopy. Staining reaction was stopped within a water bath. Hematoxylin core staining solution was applied, followed by alcohol fixation and cover-slip. For negative controls the primary antibody was omitted (Supplementary Figure S1). Pictures were taken with an Olympus BX53 light microscope linked to a DP25 camera (Olympus Corporation, Tokyo, Japan).

Cell culture and immunofluorescence
We used a previously modified Caco2 cell line stably transfected with a doxycycline-inducible KRAS G12V , 48

Immunohistochemical evaluation
Immunohistochemical staining was evaluated independently by two pathologists (HB and FR). An immunoreactive score (IRS) 49

Mouse intestine samples and Caco2 cell-derived xenografts
Transgenic mice harboring oncogenes in the Gt(ROSA)26Sor locus were described previously. 50,51 For short-term KRASV12 induction in vivo, mice were provided doxycycline at 4 mg/ml in a 1% sucrose solution via the drinking water for 3-4 days. Induced and control mice were killed, and intestines (Jejunum/Ileum) were fixed for 12-24 h in a 4% formalin solution, before they were dehydrated and embedded in paraffin. For immunohistochemistry, 4 μm sections were used, as described above. Experiments were approved by Berlin authorities LAGeSo (G0185/09). For xenografts, 3x10 6 Caco2 cells stably transfected with a doxycyclineinducible KRAS G12V48 and suspended in a 1:1 mixture of matrigel and PBS were injected s.c. into NMRI nu/nu and NSG mice (6 mice per group). For transgene induction, 2 mg/ml doxycycline (in 5% sucrose) was administered ad libitum starting at day 1 after injection. Mice were observed for 35 days and maintained under sterile and controlled conditions (22°C, ca. 50% relative humidity, 12 h light-dark cycle, autoclaved food and bedding, acidified drinking water). Tumor diameters were determined 2 × weekly by caliper measurements and tumor volume was determined with the formula TV = (width 2 x length x 0.5); body weight was measured as a parameter for tolerability. Xenograft experiments were approved by Berlin authorities LAGeSo (A0452/08).

Statistical analysis
Data sets were administrated with Microsoft Access 2003, statistical analysis was performed using the SPSS 21.0 software (SPSS, Chicago, IL, USA). Survival analyses were calculated with Kaplan-Meier functions and Log-rank test. Linear regression model and χ 2 -test were used to describe the correlation between expression data and clinicopathological data. A non-parametric correlation test (Spearman-Rho) was applied to describe the correlations between expression data. All tests were two-sided and P-values o 0.05 were considered statistically significant. Abbreviation: MSI: microsatellite instability.
Y-box-binding protein 1 in colorectal cancer F Roßner et al