Fig 2: ESRP RQ expression as T/N ratios of CRC patients were grouped by micro-satellite status and tumor stageESRP1and ESRP2 closed and open circles, respectively. Median with interquartile range is indicated by error bars. Wilcoxon signed rank test of actual median against theoretical median of 1 resulted P values for groups as indicated below graph. A. msi and mss groups analyzed by Mann Whitney test resulted P values 0.05-0.001 (**). B. Tumor stage I-IV groups analyzed by Kruskal-Wallis test with Dunn's multiple comparison test resulted no significant difference of the medians between tumor stages.

Fig 3: Binding of ESRP2 to rs152731-C and rs152731-T in EMSA and co-immunoprecipitation. (A) EMSA experiment using nuclear extract proteins derived from 16HBE14o. Signals marked “HMW” (high molecular weight) encompass different entities that can be differentiated by comparing the lane with anti-ESRP2-Ab to the control lane using IgG. Presence of ESRP2 in these HMW multi-protein complexes is deduced from the increased signal in the presence of an anti-ESRP2-antibody in comparison to the neighboring IgG control lane. Nonspec.: nonspecific signal, observed similarly with both, IgG control and anti-ESRP2-Ab. ESRP2/rs152731: signal observed predominantly in the presence of anti-ESRP2-Ab. Mixed: protein–DNA-complexes partially containing ESRP2 and partially nonspecific. The number of independent experiments is N = 6 whereby nuclear extracts were from 16HBE14o (N = 5) and T84 (N = 1). The blot shown is an exemplificative blot out of the 6 performed replicates (source data is provided with this manuscript). (B) Signals obtained with rs152731 probes and an antibody directed against ESRP2 were compared to signals obtained when IgG was used instead of the specific antibody. Data from such paired samples are shown from six independent experiments that differ concerning nuclear extract and electrophoresis conditions. Signals labelled “ESRP2/rs152731” in Fig. 6A. Could be distinguished clearly from the DNA–protein-complex labelled “mixed” in three experiments. Normalized data obtained for rs152731-C and rs152731-T probes were compared by Mann–Whitney U test whereby signals for high molecular weight complexes observed for rs152731-T were stronger (p = 0.013). (C) 150 µg nuclear extract from T84 cells was incubated with 500 nmol dsDNA probes for SCNN1B SNP rs152731 allele C and SCNN1B SNP rs152731 allele T in an EMSA experiment of a total reaction volume of 40 µl for 3 h at 37 °C. Protein–DNA-complexes were captured with an antibody directed against biotin (ab19221, Abcam, Cambridge, UK), or with IgG as a negative control (indicated by “IP”), and precipitated using protein-G-agarose-beads at 4 °C for 16 h. The supernatant was discarded and proteins were dissociated from agarose beads using Laemmli buffer in two elution steps (indicated by “Elu.” 1st and 2nd) of 30 µl each (10 min, 50 °C), loaded on a denaturing 11% polyacrylamide gel and transferred to a protean supported membrane with 20 µM pore size by tank-blot for 50 min at 100 V. ESRP2 immunoreactive bands were visualized with an antibody directed against ESRP2 (ab113486, Abcam, Cambridge, UK). A lane loaded with nuclear extract only was used as a control to identify ESRP2 (predicted molecular weight: 77 kDa). The number of independent experiments is N = 3 whereby nuclear extracts were from T84. The blot shown is an exemplificative blot out of the 3 performed replicates (source data is provided with this manuscript). (D) In the example shown in (C), ESRP2 signals derived from the probe with rs152731 allele C correspond to 85% of the signal observed for rs152731 allele T. Two replicate experiments, not shown here, have demonstrated 0% and 65% intensity on the allele rs152731-C, respectively. While the C-to-T-ratios observed in three co-IP experiments were highly variable, these three experiments consistently indicate that rs152731-C binds ESRP2 less tightly than rs152731-T.

Fig 4: ESRP1 and ESRP2 expression in CRCA. ESRP1 (closed circles) and ESRP2 (open circles) transcript expression in relation to ribosomal 36B4 as reference gene and to a reference tissue shown as scatter plot of all CRC cases. Relative quantity (RQ) levels for matched T and N tissue are different by P values <0.0001 (***) tested by Wilcoxon signed rank test. Median RQ values of T and N for ESRP1 were 82.6 and 163, and for ESRP2 142 and 354. Median with interquartile range is indicated by error bars. B. ESRP1 and ESRP2 expression correlate in T and N CRC tissues. Relative quantity (RQ) levels for T and N CRC tissues were log transformed to yield normal distribution of data and analyzed for linear correlation with Pearson method (P values of <0.0001). Linear regression line with 95% Confidence Intervals (dashed) and Pearson correlation coefficients R square are indicated. C. Ratios of RQ levels for matched T and N tissue (T/N) of transcript expression are shown as scatter plot. Wilcoxon signed rank test against theoretical median of 1 resulted P values of 0.076 and 0.009 for ESRP1 and ESPR2, respectively. Cases studied by IHC with lowest ESRP1 values are indicated by triangles and numbers. Median with interquartile range is indicated by error bars. D. Representative IHC of CRC cases with minor transcript expression in tumor tissue case #59 showed in normal colon epithelium cells (closed arrow) a stronger ESRP1 expression in the nuclei compared to adjacent invasive carcinoma cells (open arrow). Bars indicate 100μm.

Fig 5: Proteins specific for rs152730, rs152731 and 152744 and abundantly present components of the spliceosome identified by EMSA-PSeq. Numbers correspond to MASCOT scores obtained for protein sequencing (primary data provided in: raw data xls supplement R3) which is a probability score that describes – 10LOG10(P), with P being the probability that the detected protein represents a significant match in this data set. ESRP2 was captured on rs152731 probes with a MASCOT score of 119 (C-allele, p = 10–12) and 92 (T-allele, p = 10–9), respectively. Data is shown for proteins specific for one SNP (top 40 rows), proteins that exhibit partial specificity (spliceosomal proteins) and proteins that were unspecifically found in the majority of probe-derived samples but were absent from the majority of negative controls (hnRNPs). Proteins specific for one probe or one SNP which were identified from the dataset derived for 25 EMSA-PSeq samples based on 1. Their absence from empty negative control samples 2. Their absence from EMSA-PSeq samples obtained with probes for other analyzed SNPs 3. Their expression levels of less than 200 ppm according to pax-db and 4. Their annotated capabilities to interact with nucleic acids (see SupplFig 3 for details). Data on ESRP1 is provided for comparison as both, ESRP1 and ESRP2 were interrogated in an association study (see Fig. 5). As many RNA-binding proteins were found among the proteins identified as SNP-specific, and as an alternative SCNN1B transcript has been observed, EMSA-PSeq data was systematically screened for constituents of the spliceosome. Exp. ppm denotes expression values retrieved from the Protein Abundances Across Organisms-database pax-db61 for the data set “colon integrated”, except for EYA3, BAZ2B, DIEXF, PRMT6, TLX2, UTP3, LSM3, LSM5, RP9 [Whole organism, (PeptideAtlas, Aug 2014)] and TLX2 [Whole organism (Integrated)]. Samples #26, #27 and #28 have been derived from denaturing SDS-polyacrylamide gels, hence these lack the multiprotein complexes observed in all other samples derived from native polyacrylamide electrophoresis. Details on experimental conditions are described in SupplTab 9.