Chapter 2: Total Cell Kill/Cell Death Assays

As opposed to measuring cell proliferation, there is a closely-related family of assays based on the concept of total cell kill, or, in other words, cell death occurring in the entire population of tumor cells (as opposed to only in a small fraction of the tumor cells, such as the proliferating fraction or clonogenic fraction) [15-18]. The concepts underlying cell death assays are relatively simple, even though the technical features and data interpretation can be very complex. There has been considerable work based on these assays reported during the past 15 years. This body of work is not currently well appreciated among clinical oncologists, and the remainder of this review will focus on the cell death assays.

The basic technology concepts are straightforward. A fresh specimen is obtained from a viable neoplasm. The specimen is most often a surgical specimen from a viable solid tumor. Less often, it is a malignant effusion, bone marrow, or peripheral blood specimen containing "tumor" cells (a word used to describe cells from either a solid or hematologic neoplasm). These cells are isolated and then cultured in the continuous presence or absence of drugs, most often for 3 to 7 days. At the end of the culture period, a measurement is made of cell injury, which correlates directly with cell death. There is evidence that the majority of available anticancer drugs may work through a mechanism of causing sufficient damage to trigger so-called programmed cell death, or apoptosis [3,4].

Although there are methods for specifically measuring apoptosis, per se, there are practical difficulties in applying these methods to mixed (and clumpy) populations of tumor cells and normal cells. Thus, more general measurements of cell death have been applied. These include: (1) delayed loss of cell membrane integrity (which has been found to be a useful surrogate for apoptosis), as measured by differential staining in the DISC assay method, which allows selective drug effects against tumor cells to be recognized in a mixed population of tumor and normal cells [10,19], (2) loss of mitochondrial Krebs cycle activity, as measured in the MTT assay [20], (3) loss of cellular ATP, as measured in the ATP assay [21-23], and (4) loss of cell membrane esterase activity and cell membrane integrity, as measured by the fluorescein diacetate assay [24-26].

It is very important to realize that all of the above 4 endpoints can and do, in most cases, produce valid and reliable measurements of cell death, which correlate very well with each other on direct comparisons of the different methods [20,25-36]. This should not be surprising, any more than should the fact that auscultating heart sounds, observing spontaneous breathing, palpating a carotid pulse, measuring core body temperature, and recording an electroencelphalogram or electrocardiogram are all good and reliable methods of determining patient death.

We have performed direct correlations between the DISC and MTT assays in approximately 4,000 fresh human tumor specimens, testing an average of 15 drugs per specimen at two different concentrations. Thus, we have approximately 120,000 direct comparisons between DISC (membrane integrity) and MTT (mitochondrial Krebs cycle activity) endpoints in fresh human tumor specimens. The overall correlation coefficient between these endpoints in specimens containing > 60% tumor cells is 0.85 (These data do not include assays on 5FU, which, for biological reasons, may be tested in the MTT assay but not the DISC assay. These data also do not include assays for paclitaxel and docetaxel, which, for different biological reasons, may be tested in the DISC assay but not the MTT assay).

The above studies, demonstrating the comparability of results with the 4 different cell death endpoints, are important for the following reason. For perfectly understandable reasons, clinical studies correlating assay results with clinical outcome are very difficult to perform. The literature in this field may be characterized as including a great many small studies, but no big studies. Additionally, different investigators have favored different cell death endpoints, depending on the laboratory and clinical situation.

For example, the DISC assay is extremely labor intensive, and requires expertise in recognizing and counting tumor cells using a microscope, but it may be applied to specimens containing a heterogeneous mixture of tumor cells and normal cells. MTT, ATP, and FDA endpoints use semi-automated instrument readouts, but can only be applied to specimens which are relatively homogeneous for tumor cells. In addition, there are a number of additional reasons why one type of cell death endpoint may be advantageous in a given tumor specimen and why laboratories may apply several different cell death endpoints in the testing of a single specimen.

It should be noted that, historically, the DISC assay studies of the early 1980s provided the prototype for later studies of the other cell death endpoints. When the MTT endpoint was first introduced in the late 1980s, the first published studies compared the MTT results to the DISC results, with culture conditions and drug exposures being otherwise identical [20,27,29,31]. Many laboratories have preferred the MTT endpoint (and later the ATP and FDA endpoints), because of the difficulty in manually scoring the DISC assay microscope slides. But what is important is that each of the above cell death endpoints do give essentially the same results (except in the case of isolated drugs, such as taxanes and 5FU). Thus, it is entirely reasonable and proper to consider as a whole the clinical validation data which has been published using the above 4 endpoints.

The second point to understand is that cell death assays are not intended to be scale models of chemotherapy in the patient. The DISC assay was designed to address the major practical problems with the popular clonogenic assays of the late-70s/early-80s. Chief among these problems were (1) low evaluability rates and (2) uncertainty of what was being measured in individual assays (true tumor cell colonies, arising from clonogenic cell growth versus artifactual colonies arising from cell aggregation). Unlike the case with the clonogenic assays, there was no attempt to model in vivo pharmacokinetics (i.e. no attempt to utilize clinically-achievable drug concentrations or to determine something analogous to an anti-bacterial minimal inhibitory concentration). Instead, the assay conditions were rigorously fixed, with respect to culture media and drug exposure time (the latter being, most typically, 96 hours). Drugs were first tested in training set assays to determine the drug concentration which gave the widest scatter of results (mathematically defined as the greatest standard deviation). The hypothesis to be tested with clinical correlations was a very simple one - that above-average drug effects in the assays would correlate with above-average drug effects in the patient, as measured by both response rates and patient survival.

Return to HTAJ Home Page
Return to Title/Abstract/Chapter Table of Contents
Return to Chapter 1: Cell Proliferation Assays
Proceed to Chapter 3: Correlations between assay results and chemotherapy response