Study population and histology
From the archive of the Institute of Pathology, University Hospital Schleswig-Holstein, Kiel, we retrieved all samples of patients who had undergone oncologic resections of primary CRCs between 1995 and 2011. All tissue samples had been obtained from routine therapeutic surgeries. After fixation in neutral buffered formalin, all tissue specimens had been embedded in paraffin. Paraffin sections were subsequently cut and then stained with hematoxylin and eosin (H&E). The World Health Organization criteria were employed for histological classification. Board certified pathologists classified the tumor-node-metastasis stage according to the criteria of the union internationale contre le cancer (UICC; 7th edition) [10]. After study inclusion, all patient data were pseudonymized. Patients were excluded (1) if syn- or metachronous colon cancer was documented and (2) if the sample did not contain tumor cells.
Tissue microarray construction.
Tissue microarrays (TMA) were constructed from formalin-fixed and paraffin-embedded tissue samples as previously described [11]. H&E-stained tissue slides of each CRC sample were examined and three separate representative areas were selected randomly from the tumor area of the donor paraffin block(s) by a board-certified pathologist. A core was punched and transferred to the recipient paraffin block, thereby yielding three representative tissue cores per CRC patient within our TMAs. Successful transfer of tumor tissue was verified by H&E-staining of serial sections obtained from the TMAs.
Immunohistochemistry.
Paraffin sections were deparaffinized and boiled in EDTA buffer (pH 9.0) for 1 min at 125 °C. All tissue slides were washed with Tris-buffered saline (TBS) and then blocked with hydrogen peroxide block (Thermo Fisher Scientific) for 15 min. After washing with TBS and subsequent incubation with Ultra V Block (Thermo Fisher Scientific) for 5 min, all slides were incubated with the primary antibody. The incubation with the primary antibody was performed for 30 min at room temperature, followed by an incubation overnight at 4 °C. The IGF1-receptor β antibody (rabbit monoclonal; clone D406W; Cell Signaling Technologies, Danvers, USA) was used with a 1:50 dilution. The ImmPRESS reagent peroxidase universal anti-mouse/rabbit Ig-MP-7500 (Vector Laboratories, Burlingame CA, USA) served as the peroxidase conjugated secondary antibody. The ImmPact NovaRed peroxidase substrate SK-4805 Kit (Vector Laboratories, Burlingame CA, USA) was used for the visualization of immunoreactions. All tissue slides were counterstained with hematoxylin. Negative controls were generated by omission of the primary antibody (Fig. 1). Endometrium samples served as positive controls (Fig. 1).
Evaluation of IGF1 receptor immunostaining
At first, the entire series of 4497 TMA spots was screened to assess minimum and maximum staining intensities achieved with the staining protocol. Finally, a three-tired (0, 1+, 2+) scoring system of the staining intensity was considered to be appropriate and samples representing each staining intensity were selected as references for further assessment (Fig. 1). The evaluation distinguished between cytoplasmic (cCC-IGF1R) and membranous immunostaining (mCC-IGF1R) of the tumor cells. Subsequently, the whole study population was evaluated in depth. The three cores of each CRC specimen were treated as a single case. Subsequently, a modified HistoScore (HScore) was employed for the evaluation of IGF1R immunostaining. First the intensity of cytoplasmic and membranous IGF1R immunostaining, respectively, within tumor cells was evaluated and categorized as absent 0 (=no evidence of staining), weak (1+) and strong (2+). Secondly, the percentage of tumor cells with no (0), weak (1+), or strong (2+) immunostaining within each given tumor sample was estimated. The percentage of immunostained cells always added up to 100% according to the following formula: % (0) + % (1+) + % (2+) = 100% tumor cells. Subsequently, an HScore was calculated using the following formula: HScore = [0 x percentage of immunonegative tumor cells] + [1 x percentage of weakly stained tumor cells] + [2 x percentage of strongly stained tumor cells]. The maximum possible HScore was 200, if all tumor cells within a sample showed a strong (2+) immunostaining. The HScore served to improve the stratification of our samples, by separating more distinctively samples of low and of high immunostaining intensities. Finally, the cohort was split at the median HScore in high or low IGF1R expression. One person scored all samples (MP) and repeatedly compared the scoring with the study’s predetermined reference samples (Fig. 1) in order to decrease intra-observer variability. In the case of ambiguous immunostaining results, a second investigator (CR/SH) from the team was referred to and a consensus was reached. The scoring was reviewed on a random sample basis by a second investigator (CR/SH) in order to validate the consistency of the evaluation process.
Assessment of the insulin receptor status
The IR status was assessed as previously described [4]. In brief, a monoclonal anti-insulin-receptor antibody (rabbit, clone 4B8; Cell Signaling Technologies, Danvers, USA; dilution 1:50; manual immunostaining) was used for immunohistochemistry. IR expression in vessels and in tumor cells was evaluated. With respect to IR expression in tumor cells, immunostaining was classified as either being negative, if no staining was evident, or positive, if any immunostaining was present. IR expression in cancer vasculature was scored ranging from absent (0) to strong (3+). Vascular IR expression was categorized into absent (0), weak (1+), moderate (2+) and strong (3+). The IR expression data as such has been published elsewhere [4] and has now been correlated with the new IGF1R expression data of the present study. The comparative analysis of IGF1R- and IR-expression was based on the same TMA cores.
Assessment of DNA mismatch repair protein immunostaining
The expression of DNA mismatch repair proteins (MMR) MLH1, PMS2, MSH6 and MSH2 were assessed according to the algorithm suggested by Remo et al. [12] as previously described [4]. The algorithm is based upon the evaluation of nuclear staining within tumor cells. MMR deficient (dMMR) and MMR proficient CRCs were discriminated. Occasional cases of inconclusive MMR staining results were excluded, e.g. due to the absence of a positive internal control, or due to technical artifacts.
KRAS genotyping
For genotyping one representative tissue section and the corresponding paraffin block were chosen from the resection specimens. The tumor area was marked on the H&E-stained slide. The percentage of tumor tissue in the marked area and the relative amounts of the histoanatomical components of the tumor, i.e. tumor cells and desmoplastic stroma were estimated visually to guarantee a valid tumor cell content. Genomic DNA was then extracted from formalin-fixed and paraffin-embedded tissue with the QIAamp DNA mini kit (Qiagen, Hilden, Germany) following the manufacturer’s instructions. To ensure a tumor cell percentage of > 40% in the analyzed specimens the tissue sections were manually microdissected prior to DNA extraction. For mutational analysis of codons 12 and 13 of the KRAS gene a 179 bp fragment was amplified by polymerase chain reaction (PCR) using the primers 5′-AGGCCTGCTGAAAATGACTGAATA-3′ (sense) and 5′- CTGTATCAAAGAATGGTCCT GCAC-3′ (antisense).15 PCR products were purified using the Nucleospin Extract II kit (Macherey-Nagel, Düren, Germany) and both strands sequenced by dye terminator cycle sequencing (BigDye Terminator v1.1 Cycle Sequencing kit, Applied Biosystems, Darmstadt, Germany) with the primers used for PCR amplification. The sequencing products were analyzed on an ABI Prism 310 Genetic Analyzer (Applied Biosystems). The results were confirmed by pyrosequencing on a PyroMark Q24 instrument as described by Ogino et al. [13]. Mutational analyses of codon 61 of the KRAS gene were done by pyrosequencing. In brief, specific DNA fragments of the individual genes were amplified by PCR using the primers 5′-AATTGATGGAGAAACCTGTCTCTT-3′ and 5′-TCCTCATGTACTGGTCCCTCATT-3′ (KRAS, codon 61). The resulting PCR products were sequenced on a PyroMark Q24 instrument with the sequencing primers 5′-GGATATTCTCGACACAGC-3′ (KRAS, codon 61), The KRAS-genotyping had been certified by an external quality assurance program done by the German Society of Pathology and the Bundesverband Deutscher Pathologen e.V. (www.dgp-berlin.de).
Statistical analyses
For statistical analyses SPSS version 24.0 (IBM Corp., Armonk, NY, USA) was employed. Fisher’s exact test was used to test the correlation between non-ordinal clinicopathological patient characteristics and the mCC-IGF1R-status, or the cCC-IGF1R-status, respectively. Fisher’s exact test also served to test correlations between the IR status and the mCC-IGF1R-status, or the cCC-IGF1R-status. Variables of ordinal scale such as the T category, N category, UICC stage and tumor grading were tested with Kendall’s tau test. The Kaplan-Meier method was used to determine median survival with 95% intervals. The log-rank test was employed to test differences between median survivals. A p-value of ≤0.05 was defined to be significant. All p values are displayed without correction. We applied the Siemes (Benjamini-Hochberg) procedure to compensate a false discovery rate within the correlations. P-values having lost significance are marked.