Open letter to participants in NCI-Sponsored Ovarian Cancer State of the Science Meeting

Bethesda, MD September 16-17, 2005

(from Larry Weisenthal)

At the afternoon session (Sept 16), the following question was asked: "Who needs Taxol?"

That's actually fairly easy to answer.  Ideally, it should be on the basis of currently available, fresh tumor cell culture drug resistance testing (CCDRT).  In the absence of CCDRT, previously-untreated patients with poorly-differentiated tumors should be treated with single agent carboplatin, while patients with moderate (or moderate-well) differentiated tumors should be treated with a regimen which includes one or more non-platinum drugs.  The data supporting this are presented elsewhere on this website (click  here and
here).

At the morning session (Sept. 16), I took note of the following problems:

1. Importance of improving treatment outcomes.
2. Accrual of both patients and tumor specimens to clinical trials involving translational research.
3. Problems with pathologists sharing tissues with an outside institution
4. Expense of equipping smaller, community hospitals for snap freezing tissue using established protocols.
5. Challenge of combined therapy approaches (traditional cytotoxics plus new agents)
6. The need for gold standards with which to discover which molecular endpoints are predictive of treatment outcomes.

All of the above problems could be improved by incorporating "open source" cell culture drug resistance ("chemosensitivity") testing in treatment planning.

Important caveat: In the real world, patients readily grasp the potential advantages in individualizing their drug selection.  A cancer patient can and will bend the clinicians and pathologists of even the most "powerful" institution to ensure that this gets done.

The controversial history of cell culture drug resistance ("chemosensitivity") testing is detailed on this website.  It is very complicated (I compare the level of complexity of knowledge required to read and understand the literature, principles, and status of this testing as similar to that required to gain expertise in electrocardiogram reading.  i.e. it isn't simple, but it is achievable to the motivated physician).

For the time being, I'll focus on only one very simple, yet very important application.

Problem: To determine if ovarian cancer is platinum resistant or platinum sensitive PRIOR to treatment with platinum.

Proposed clinical trial:

Obtain fresh tumor biopsies from 30 patients with previously-untreated ovarian cancer and from 30 patients whose disease meets the definition of platinum-resistant ovarian cancer.

Code the specimens to ensure blindedness. Send specimens to whatever laboratories wish and are able to participate in the exercise.  The laboratory(ies) will not know whether each individual specimen is from a previously-untreated patient (75% of whom may be expected to respond clinically to cisplatin or carboplatin) or from a patient with a clinically "platinum-resistant" tumor (25% of whom may be expected to respond clinically to cisplatin or carboplatin).

If a laboratory can be proven to show meaningful differences between the untreated versus "platinum-resistant" cohorts of patients, this would then prove the biological relevance of the assay technology(ies) in question, as well as illustrate how practical it is in a real world setting in the year 2005 to send out tissue for "culture and sensitivity" testing.  Presuming that the participating laboratory(ies) used "open source" (non-proprietary) technology(ies), this would open the way for a central laboratory to serve as (1) a clinical laboratory for a variety of obvious follow-up clinical trials (e.g. untreated patients randomized between platinum/taxane in the "empiric therapy" arm versus an individually tailored arm in which patients would receive single agent carboplatin (if assay "sensitive") or else in vitro best non-platinum therapy (if assay "resistant" to platinum) and (2) a central clearinghouse/repository for "leftover" tumor specimen in excess of the needs of the cell culture assay (nb. the newer methodologies are much more "frugal" with respect to the use of tumor cells than the older methodologies; in most cases with ovarian cancer specimens, tumor cell yield is sufficient to supply the needs of both the clinical cell culture assay and leaving resisdual cells for archival storage).

Note that "snap freezing" is disadvantageous, in that normal tissue cells and non-viable cells are frozen, along with the tumor cells. When viable tissue is FedExed in CO2-independent transport media, viable tumor cells my be selectively enriched with great efficiency from "contaminating" normal cells and non-viable cells.

Selecting for tumor cells is very important for an assay like the MTT assay, ATP assay, FDA assay, and caspase 3/7 assay, all of which use an endpoint which can't discriminate between tumor cells and normal cells (in contradistinction to the DISC assay, in which tumor-specific drug effects can be distinguished in a mixed cell population).  Serum free medium and anchorage independent growing conditions will prevent the _growth_ of normal cells but in most short term cell death assays we are not measuring cell growth but rather cell survival.  The selective advantage of anchorage independence and serum free status is much less in the case of short term cell survival than it is for actual cell growth.  One of my many pet peeves about this literature is the cavalier use of the term "growth."  Thus, it is important to "purify" tumor cell clusters prior to plating them in short term culture.
 
Here are what our own data show (representative data shown for specimens with an official histopathologic diagnosis of poory-differentiated ovarian cancer):

Treatment cohort (in ovarian cancer)
LC50
P2 (compared to "Relapsed < 6 months" cohort, below)
Previously Untreated (n=88)
1.53 ug/ml
=0.00015
Relapsed > 6 months post platinum (n=45)
1.82 ug/ml
=0.0015
Relapsed < 6 months post platinum (n=52)
2.43 ug/ml
(60% increase compared to untreated)

Comments: Methodology and patient survival correlations are presented in detail elsewhere on this website).  This isn't a scale model of chemotherapy in the patient. The LC50 value should not be used in the same fashion as an anti-microbial "MIC" level, but the relative increase in effective concentration is probably of clinical relevance (i.e. I think it would take about a 60% increase in platinum dosage to achieve in the average "platinum resistant" patient what could be achieved in the average untreated patient).

Some selected technical details follow:

We've received a total of 946 specimens of ovarian cancer, fallopian tube cancer, primary peritoneal papillary serous adenocarcinoma, and papillary serous endometrial adenocarcinoma.  All of these are now considered to be essentially the same general "disease," from the standpoint of biological behavior, response to treatment, and prognosis in the Stage III/Stage IV setting.  Most of these were simply diagnosed as ovarian cancer.
 
Only 170 of these specimens had a post-culture to pre-culture ratio of viable tumor cells which exceeded 1.01 (the definition of cell "growth.").  The median ratio was 0.85 and the mean ratio was 0.81.  So you see, most of our specimens are not "growing."  Of our 946 specimens, most of the time the sources of cells was a solid tumor biopsy or resection.  In the case of the ascites fluid or pleural fluid or pericardial fluid specimens (n=170), the median post-culture/pre-culture viable tumor cell ratio was 1.0; however, there was a distinct INCREASE (as opposed to simply 100% maintenance of viability) in only 40 of 170 specimens.
 
In the case of all 946 specimens, at the end of the culture the median percent tumor cells (compared to normal cells) was 90% tumor cells, compared to 10% normal cells.  The corresponding means were 83% and 17% (meaning that some specimens had a whole lot of normal cells at the end).  In the case of the fluid specimens, the median percent tumor cells at the end of culture was 90%, but the mean was only 73%, meaning that some of the cultures had a very low percent tumor cells (usually the "offending" normal cells in these latter cases were mesothelial cells).
 
We culture our cells in anchorage independent, polypropylene microwells.  I'm pretty sure that I was the first investigator to report that polypropylene (as opposed to polystyrene) provided anchorage independent growth conditions, in my publications of the early 1980s.  Maybe someone "scooped" me on this; I'm not aware of it and, anyway, it's not all that big a deal.  We do a lot of work pre-culture to obtain relatively "pure" populations of tumor cells.  One of our bread and butter methods was invented by my wife.  It's something we call "quick spins."  Basically, mix up the cells in a centrifuge tube, turn on the centrifuge (standard large benchtop Beckman TJ6 or Allegra6) and barely let it get up to 1000-1500 RPM (depending on the specimen), then immediately turn it off.  The three dimensional cell clusters go to the bottom; single cells, including dead cells and debris, stay in the supernatant.  Tumor cells are slightly more dense than are mesothelial cells.  With repeated 'quick spins' (sometimes up to 20 of them), you get a selective enrichment for tumor cells.  We monitor this process by making Cytospins along the way.
 
At the end of the assay, we have many endpoints at our disposal.  In cases where there are many normal cells, the only reliable endpoint is the DISC assay (which Robert Nagourney has re-named the "Ex Vivo Apoptosis Assay" and which Andrew Bosanquet has re-named the "TRAK" assay).  We always run at least one and usually two or even three additional endpoints.  I prefer the MTT, as it gives a very adequate signal in the typical ovarian specimen, and it is more specific for tumor cells than other metabolic endpoints, including the ATP.  We also use resazurin and caspase 3/7 (using the Promega Glo endpoint) for different purposes in different situations, as well as the ATP (chiefly in situations where we have a very pure population of tumor cells, but a very low cell yield).
 
I don't know of a biological advantage in testing "pure" tumor cells.  Robert Hoffman makes a major point about maintaining the native tumor architecture, including making sure that all  the normal cells are present.  So he claims it's a biologically better model to have all the normal cells present.  But his published correlations are just as good or bad as everyone else's (patients treated with assay-active drugs much more likely to respond and live longer).  The only advantage in selectively culturing tumor cells is if you use an endpoint which can't discriminate between normal cells and tumor cells.
 
Everyone in this field likes to claim that his/her method is the best.   I'm pretty sure that I'm about the only one working in this field (well, maybe Larssen and Nygren in Uppsala; Kaspers and Peters in Amsterdam also) who is happy to admit that all of the cell death endpoints can be made to give equally valid information, if they are done properly.  The choice of which to use depends mostly an a variety of technical factors relating to type of specimen.
 
I'm also unaware that there is any advantage at all in using serum free culture conditions, beyond providing selectivity for tumor cell growth (true growth, as opposed to merely survival in short term culture, where the selectiveness of serum-free culture is of much lesser magnitude).  One could make the case that having serum in the medium makes it much more "physiological," as many drugs bind to albumin and other proteins "naturally" present in the bloodstream (and often in malignant ascites fluid).  Without serum, there is less binding, and lower in vitro concentrations of some drugs need to be used to provide a proper "scatter" of results.  But this is just a matter of calibrating drug concentrations for the given assay conditions.  Everyone does it (calibrates his/her own assay for differing conditions of culture).
 
In short, I think that we (who work in this field) are all good guys and that none of us are selective technical geniuses.  The toweringly impressive achievement of Dr. Ian Cree was in actually completing a prospective randomized trial (discussed elsewhere on this website).  Now if we all could just figure out a way to build upon his pioneering work in this regard to make some real progress.  Alas, it's not going to be easy...at least here in the USA.
 
- Larry Weisenthal
Written (hastily) in hotel room in Bethesda, MD. Sept. 16, 2005