Improved Immunohistochemical Detection of Type 1 Insulin-Like Growth Factor Receptor in Human Tumors

Background: Insulin-like growth factors (IGFs) are known to play important roles in cancer biology, prompting evaluation of drugs targeting type 1 IGF receptor (IGF-1R). However, there is considerable lack of consensus in immunohistochemical (IHC) studies of IGF-1R in human tumors, confounding attempts to assess the predictive and prognostic significance of IGF-1R expression and subcellular localization. Likely sources of variation include use of different IGF-1R polyclonal antibodies and methods for IHC. Here, we aimed to develop a robust IGF-1R IHC protocol using a monoclonal antibody, suitable for use in formalin-fixed paraffin-embedded (FFPE) tissues. Methods: Using controls including samples of FFPE tissues and tumor cells of defined IGF-1R expression, we used IHC and western blotting to compare polyclonal antibody #3027 with monoclonals #9750 and #14534 (Cell Signaling Technology). Results: Compared with #3027, the monoclonals exhibited superior discrimination between IGF-1R-high and IGF-1R-deficient cells in manual IHC, signal generated by #9750 reflecting differences in IGF-1R expression detected by western blotting. In tissues, IGF-1R detected by #14534 was predominantly plasma membrane-associated, while #9750 detected IGF-1R in the plasma membranes, cytoplasm and nucleus of prostate and renal cancers, recapitulating appearances we described using previous lots of #3027, and reflecting subcellular localizations reported using other techniques. Use of #9750 and #14534 in an autostainer showed adequate differentiation of high vs low IGF-1R cells, but did not recapitulate appearances of manually-stained tissues. We provide a detailed protocol for the preferred manual method using #9750. Conclusion: Standardization of IGF-1R IHC will promote understanding of the role of IGF-1R in tumor biology, and its potential as a candidate prognostic and predictive biomarker.


Introduction
The contribution of insulin-like growth factors (IGFs) to cancer biology has been extensively studied in cell lines, revealing that IGFs activate type 1 IGF receptors (IGF-1Rs) to promote cell cycle progression, cell survival, motility and invasion [1,2]. These findings provoked interest in studying IGF-1R expression in clinical cancers, and in development of drugs that block IGF signaling. However, there has been striking variation in reported IGF-1R expression in tumors and normal tissues when detected by immunohistochemistry (IHC). For example IGF-1R was reported to be unchanged or down -regulated in prostate cancer compared with benign prostate [3,4], although since our report of IGF-1R up-regulation at the mRNA and protein level [5] most publications support up-regulation [6][7][8].
This lack of consensus also confounds attempts to interpret results of clinical trials of novel IGF inhibitory drugs. Early trials reported striking clinical responses to IGF-1R inhibition e.g., Ref. [9], but later trials showed very limited activity in unselected patients (reviewed in Ref. [10]. This raises the question as to whether tumor IGF-1R expression correlates with sensitivity to IGF-1R inhibition. Preclinical reports supporting such a link include studies in non-small cell lung cancer (NSCLC), breast and colorectal cancer, rhabdomyosarcoma, Ewing family tumours and neuroblastoma [11][12][13][14]. However, other preclinical and some clinical studies found IGF-1R expression not to associate with response to IGF-1R inhibition in breast cancer, NSCLC and sarcomas [15][16][17]. While it is possible that IGF-1R is predictive of response in some tumor types but not others, it is also likely that technical differences in IHC protocols contribute to the apparent variation in significance of IGF-1R expression. Notably, the study conducted by Schwartz and colleagues found no difference in clinical activity of IGF-1R antibody cixutumumab with temsirolimus in patients whose tumors were IGF-1R 'positive' or 'negative' by IHC using the automated Ventana platform [17]. This trial also assessed tumor IGF-1R in a subset of patients by western blotting of fresh tumor lysates, resulting in detection of IGF-1R in all 'IGF-1R negative' tumors that were tested.
Further complicating interpretation of IGF-1R IHC is the phenomenon of IGF-1R nuclear translocation, reported by several groups including ours [18][19][20][21]. In FFPE samples of human tumors, nuclear IGF-1R has been shown to associate with adverse prognosis in renal cancer [19], and with response to IGF-1R antibody in patients with sarcoma [21]. Lack of detection of nuclear IGF-1R prior to adoption of heat-based antigen retrieval [5,19] suggests that this also is influenced by technical variation in IHC methods.
These findings highlight the insensitivity of some IHC protocols, and the difficulties inherent in assessing the biological and clinical significance of IGF-1R expression and subcellular localization. Likely sources of variation include use of different antigen retrieval protocols and IGF-1R antibodies, and lot-to-lot variation in polyclonal antibodies. These technical factors may confound attempts to understand the role of IGF-1R in tumor biology, and the assessment of IGF-1R as a candidate prognostic and predictive biomarker. The aim here was to develop a robust IGF-1R IHC staining protocol that can be used to address these questions.

Materials and Methods
We used formalin-fixed paraffin-embedded (FFPE) human tumors that were surplus to diagnostic need, available in the Department of Cellular Pathology, Oxford University Hospitals NHS Foundation Step Protocol

Results and Discussion
In our initial study of IGF-1R expression in prostate cancer, we had used polyclonal antibody sc713 (Santa Cruz) [5]. In a later study [24], sc713 was insufficiently specific, and we re-optimized our IHC protocol to another polyclonal, #3027 (Cell Signaling Technology) [19,24]. This antibody performed acceptably in western blotting (e.g., Ref. [19,25], but we recently questioned the specificity of current lots of this antibody in IHC. Therefore, we tested currently available lots of #3027 in IHC using FFPE sections of MCF7 and SKUT-1 cells, which express high and low IGF-1R respectively. Using the manual IHC protocol, there was detectable IGF-1R staining in both MCF7 and SKUT-1 cells, signal in the latter suggesting lack of specificity ( Figure 1A). In comparison, monoclonal antibody #9750 showed greater differentiation between SKUT-1 and MCF7 cells. We also compared #3027 and #9750 by western blotting of cancer cell lysates ( Figure 1B). Probing of duplicate membranes confirmed that both antibodies detected immunoreactive IGF-1R of the predicted size: mature IGF-1R beta subunit of 98 kDa and IGF-1R pro-receptor of 220 kDa. We noted that #3027 generated stronger signal than #9750, paralleling differences seen in IHC ( Figure  1A). IGF-1R expression was relatively high in MCF7, lower but clearly   Figure 1D) but generated very weak staining in FFPE tissue (see Figure 2A) and is not recommended for future studies. detectable in A375M melanoma cells, and low/undetectable in SKUT-1, PC3 and A375M cells in which IGF-1R was depleted by siRNA transfection ( Figure 1B). These relative expression levels correspond to findings of our previous studies [19,25,26]. Comparable differences in signal were detected when we stained the control TMA with #9750 ( Figure 1C). The close parallels between the relative intensity of IGF-1R signal on western blot and IHC suggest that this antibody was sufficiently sensitive to detect differences in endogenous IGF-1R expression and changes induced by IGF-1R gene silencing. We then tested #9750 in parallel with a more recently-available IGF-1R rabbit monoclonal, #14534, comparing manual IHC with automated IHC on a Leica-Bond auto-stainer. Both antibodies showed clear discrimination between high IGF-1R MCF7 cells and IGF-1R-deficient SKUT-1 cells, with generally lower signal in the auto-stained samples ( Figure 1D).
Having confirmed the specificity of both monoclonal antibodies in control cells of known IGF-1R expression, we performed testing in FFPE samples of human cancer. Staining of prostate cancer using manual and automated IHC showed negligible signal using either antibody at 1:200 dilution, and detectable signal at 1:50 that was more intense in the epithelial than stromal components, as previously noted [5]. The two antibodies showed different staining patterns, with faint membrane and more marked intracellular staining by #9750, and pronounced membrane staining by #14534. In both cases, the staining was much fainter in the auto-stained slides (Figure 2A). Finally, we analysed IGF-1R expression and subcellular localization in clear cell renal cancer, in which we had detected prominent nuclear IGF-1R in our previous study [19]. At low power it was apparent that #9750 gave stronger staining than #14534 ( Figure 2B, left). As in prostate cancer (Figure 2A), #14534 showed prominent membrane IGF-1R, while #9750 showed membrane and cytoplasmic signal. In #9750-stained tumor there was clear detection of nuclear IGF-1R that showed cell to cell variation, recapitulating appearances generated by previous lots of #3027 in our prior study ( Figure 2B, upper center and right panels; compare Figure 4B panel H in Ref. [19]. In contrast nuclear IGF-1R was only faintly detected by #14534 (Figure 2B lower). Given that nuclear IGF-1R is detectable by a range of methods including subcellular fractionation, immunofluorescence, proximity ligation assay and chromatin immunoprecipitation, and is reported to have prognostic and predictive significance [18][19][20][21], we suggest that #9750 is preferable to #14534 when using IHC to study IGF-1R expression and subcellular localization in human tumors.
In conclusion, we developed a specific and sensitive manual IHC protocol using antibody #9750. This is a monoclonal antibody, ensuring consistent antibody supply in future work. Nuclear IGF-1R was only weakly detected by antibody #14534, and auto-staining did not recapitulate appearances of the manual #9750 protocol, likely reflecting differences in antigen retrieval and detection. We provide a detailed protocol for the preferred manual IGF-1R IHC method that we commend for future studies. Adoption of a uniform protocol should reduce inter-study variation and may help to clarify whether responses to anti-IGF drugs associate with IGF-1R expression or subcellular localization.