The aim of this study was to identify the minimum panel of membrane markers that may be suitable for detection of invasive breast cancer by molecular imaging. In order to determine this combination, we stained TMAs consisting of 483 clinical specimens of invasive breast cancer by immunohistochemistry. Based on the expression profiles in the normal breast tissue, we defined highly tumor-specific (no or low staining of the normal breast tissue) and less tumor-specific (moderate or high staining of the normal breast tissue) membrane targets. We found that the expression of highly tumor-specific targets (HER2, EGFR, GLUT1, CAIX, IGF1-R, and MET) is quite dependent on the tumor histology and molecular subtype: ductal cancers and in particular the basal/TN and HER2-driven subtypes express more frequently highly tumor-specific membrane targets than lobular cancers.
Because the individual tumor-specific markers are clearly not sensitive enough, application of a tumor-specific panel of tracers is required to detect all types of breast cancer. A panel of tumor-specific markers (GLUT1, EGFR, HER2, IGF1-R, MET, and CAIX) was in the present study able to 'detect' 45.5% of all cancers and 55.6% of ductal cancers. For lobular cancers and low-grade tumors, the panel was not very suitable because with detection rates of 18.3% and 22.5%, respectively. Addition of less tumor-specific markers theoretically increased the detection rate to 98.1% using MUC1, but of the less tumor specific markers only CD44v6 met the desired 3-fold tumor-to-normal tissue ratio measured by image analysis. When adding CD44v6 to the panel, 80.1% of all cancers could be ‘detected’ with at least one marker in a panel consisting of HER2, GLUT1, EGFR, IGF1-R, and CD44v6. CAIX and MET had no additional effect on the sensitivity of the panel once CD44v6 had been included.
Our estimation of positivity of breast cancers for our panel may have been conservative since we have been very stringent in calling expression positive, explaining why our rates of expression for GLUT1, CAIX, EGFR, MET, TfR, CAXII, and Mammaglobin are on the lower side compared to the literature [22
]. Tumors with 1+ membrane staining were consistently considered negative as we expect that this level of staining provides insufficient signal-to-noise, but only in vivo
studies can confirm this. Moreover, quantification of expression levels based on image analysis of immunohistochemical stainings may be hampered by the non-linear amplification of the signal during immunohistochemistry. For that reason we applied immunofluorescence of directly labeled antibodies for more reliable quantitation of protein expression. Tumor-to-normal ratios above 3 where only obtained when tumors are scored as DAKO 2+ or 3+ membranous staining. This justifies the predefined thresholds for calling tumors positive. Furthermore, cytoplasmic staining was ignored as imaging antibodies will not be easily internalized and will have to bind to receptors on the outside of the cancer cells. Lastly, using TMAs may have resulted in slight underestimation of GLUT1 and CAIX expression, because the expression is usually limited to hypoxic areas within the tumor [44
Adding further candidate tumor markers may enable to improve the results of our panel of membrane related markers. For instance, biomarkers that are specifically expressed in the stroma of breast cancers like growth factors (e.g. VEGF) may be valuable.
This study provides information on the expression levels of membrane bound targets for imaging using paraffin embedded material of invasive breast cancers. To be suitable for breast cancer detection or screening, multiple steps have to be taken before tracer development and testing in (pre)clinical trials results in treatment of patients. However, the present study elucidates which targets might be most suitable based on the expression in cancer vs. normal breast tissue. One of the current challenges is specific detection of lobular breast cancers and DCIS, because these lesions are difficult to detect by mammography. DCIS was beyond the scope of the current paper, but for detection of lobular breast cancer CD44v6 is potentially quite useful.
Next to expression of target proteins, tumor perfusion and penetration of the tracer into the tumor could influence the signal for imaging. Further, affinity after labeling and half-life of the tracer in the human body determine the tumor-to-background ratio and thus the applicability of a tracer in a clinical setting. Based on preclinical studies using NIRF labeled trastuzumab and bevacizumab, the maximal tumor-to-background ratio was obtained 6
days post injection [33
]. Optimizing this by reducing the half-life of the tracer would be beneficial for clinical practice.
The present study underlines that no single membrane marker probe is likely to detect all breast cancers by molecular imaging, and that a panel of least five probes may be required. So far, experience is however limited to maximally two different tracers at once. Barrett et al. [46
] showed that two antibodies allowed to identify differences in tumor expression of HER2 and EGFR in vivo
. When aiming to be just discriminative between tumor and normal, a panel of markers can be injected with the same probe attached to simplify imaging. Feasibility and toxicity of injecting a panel of markers require further in vivo
experiments in mouse models.