Seeking the Killer App

Immunohistochemistry (IHC) was the first companion diagnostic test for estrogen receptor (ER) and then human epidermal growth factor receptor 2 (Her2). Used in this context since the late 1990s, the tests have not changed much since then. Newer IHC tests, like those for EGFR (epidermal growth factor receptor) have more recently been introduced, but others, such as that for MET (mesenchymal epithelial transition factor) have been tried but failed, most likely due to lack of standardization, controls and reproducibility. And, there have been numerous efforts to quantify expression, including automated measurement of both traditional IHC and quantitative immunofluorescence. While these tests have been widely published, and even show significant improvements over subjective testing, levels of evidence or political factors have prevented any quantification from becoming “standard of care.”

In my view, this is a missed opportunity for the field of pathology. Arguably, the reason for this is that it is not necessary. That is, the quantification of protein expression on slides has not been “required” for any assay. Instead, we have been able to get by with semi-quantitative subjective measurements for ER and HER2, and with the appearance of newer drugs requiring companion diagnostic tests, the companies introducing them were afraid of asking for a measurement method that the majority of routine pathology IHC labs could not easily adopt. As a result, big pharma have been making deals with large diagnostic companies to develop companion diagnostics that can broadly deliver conventional IHC. Because of such deals, we may never know whether onartuzumab (anti-MET), for example, would have been more effective if targeted with an accurate, sensitive and specific IHC method.

Is this a lost opportunity? In my view, it is not, but it will require a “killer app”. That is, a test that explicitly requires careful measurement that requires performing on select patients for a highly effective drug. The killer app may be around the corner in the detection and measurement of specific protein domains, in a manner that requires quantification or at least standardization. The test would be specific HER2 domains that may be required for the drug trastuzumab emtansine (TDM1 or Kadcyla) (1).

In a recently published a paper, our research shows that different antibodies that bind to different domains of the HER2 molecule display variable expression in some breast cancer cases (2). We demonstrated that antibodies like CB11 that bind to the intracellular domain (ICD) are not equivalent to those that bind to the extracellular domain (ECD) like SP3. This is biologically sound because proteases can cleave the ECD of HER2 and release it into the serum (3). Also, a cytoplasmic domain may also be produced by alternative ribosomal start sites resulting in short HER2 molecules that have the kinase domain (in the ICD) but not the ECD. Our hypothesis was that patients that had only ICD could respond to the standard regimen of trastuzumab with chemotherapy, but would not respond to trastuzumab alone.  It was difficult to find a trial that offered this test, but we described an improved benefit corresponding to the presence of ECD in The Hellenic Cooperative Oncology Group (HeCOG) 10/05 – a trial in which patients received trastuzumab months after their chemotherapy.

Soon, there will be a test of TDM1 on patients in the absence of chemotherapy to try to spare patients the associated toxicity (the GEICAM [Spanish Breast Cancer Research Group] trial). We hypothesize that patients with insufficient ECD, will not benefit from TDM1. Therefore, we think that the killer app could be a quantitative measurement of ECD and ICD to find patients that have enough ECD (probably 70 percent of the cases) that can benefit from TDM1 versus those with predominantly ICD, who will still require chemotherapy, and perhaps other non-trastuzumab therapy (such as lapatinib, the small molecule HER2 kinase inhibitor). It is too early to tell whether this will be a killer app and if it will be required to prescribe TDM1. It will require a number of studies, and perhaps the development of a simpler application than that described in our first paper (2).

In my view, this may be the tip of the iceberg. Many transmembrane proteins are processed and detection of domain specific activity could go beyond HER2. The next candidate is PD-L1.  While the detection methods for PD-L1 are currently highly controversial and in development (4), it should be noted that both ICD and ECD antibodies for this molecule are commercially available and that we and others have suggested that cases show variable expression of PD-L1 as a function of the domain specificity of the antibody. This test also has potential to be the killer app.