Interview: Interview with David Mangelsdorf for Personalized Medicine

▪ With a background in biochemistry, what led to your research interest in personalized medicine?

I was influenced by the poor relative efficacy of current cancer drugs (particularly in lung cancer) and the amount of money the pharmaceutical companies make on these drugs, which are only effective in a small number of patients.

My interest is in nuclear receptors (NRs) and cancer, I have been working on NRs my whole life as a scientist; as a graduate student I worked on the vitamin D receptor, and at the Salk Institute, where I completed my postdoctoral studies, my work was completely immersed in understanding how these receptors work and what their ligands were.

When I started to consider my independent laboratory options after my postdoctoral work, I was recruited by the Nobel Prize winner Alfred Gilman (at the University of Texas Southwestern in the Department of Pharmacology). I discovered that I really liked the place on my interview and decided it would be an ideal environment to start my independent career. I moved here in 1993 to do just that and my initial work was focused on discovering the ligands for the orphan receptors of the NR family – and this is what I built the early part of my career on.

It was here that I started to get interested in the receptor family as a whole as a possible way of understanding and even predicting how biology works in both physiologic and pathophysiological environments – under normal and disease conditions – this was really a prelude to the work that was carried out in this recent study.

▪ What were you expecting to find in your recent study on predicting non-small-cell lung cancer patient outcome?

Nuclear receptors are transcription factors, and as such are responsible for the turning on and off of genes in cells. Previous studies had already looked at gene expression and asked the question, how is this process being regulated? Well NRs are regulators of that process; many of them are already known to be involved in processes such as differentiation and growth – two of the major drivers of cancer. So the question then became, can studying not just one of these receptors, as had been done before, but rather the entire family, of which there are 48 members, glean some information that might be useful in understanding the process of the disease and its genesis and maintenance, and from that determine whether or not it is possible to identify any of these receptors or their targets as new therapeutic targets?

To begin with we were not really sure what we would find from this study, we were not expecting a prognostic signature or to discover any therapeutic benefit – so we approached the study as a survey, hoping to get information from just looking at this gene family.

▪ What types of techniques were employed in your recent study?

In our preliminary study we used quantitative real-time PCR to measure the expression of all 48 NRs in different cancer cell lines derived from non-small-cell lung cancer patients. Essentially we were trying to understand if the expression of NRs as a whole could provide a biomarker that might be of benefit to understanding and even treating the disease.

The patient sample that we used was in collaboration with the MD Anderson Cancer Centre in Houston. Dr Ignacio Wistuba was able to provide us with a patient sample set that was rather unique in that these were carefully resected tumors from non-small-cell lung cancer patients – he actually went in and dissected the tumor cells from the normal tissue in the lung – so we had a sample of both tumors from the lung and normal nearby material.

That in fact turned out to be quite revealing in the studies we carried out later on as it is quite rare to have a pure homogenous population of tumor cells and also, from the same patient, have some of their normal cells. We surveyed these very well-annotated and homogenous samples using high-throughput quantitative real-time PCR rather than typical microarray because microarray technology is not quantitative, it gives you a relational amount and its ability to tell you how much RNA is present is very weak in comparison to quantitative real-time PCR, which gives a very robust read.

Combining this powerful technique with our very good sample set allowed us to go in and interrogate NR expression. What we found first of all was that we could differentiate tumor tissue from the normal tissue. Second, we carried out an unsupervised cluster analysis to define what the meaning of the signature might be, and this gave us some interesting answers. First, it did a very good job of simple prognosis in comparison to what is already in the literature using whole-genome expression, that is, of predicting how long a patient would live, especially in stage one cancer cases.

Next we started to look more in-depth at what the sample set was telling us. We did this by designing cross-validation studies in a larger sample set using published studies. Here we did not have the luxury of using quantitative real-time PCR as the previous studies had been derived using microarray, so initially we were not very confident that we would collate any useful results, but the surprise was that we did, and they had statistically significant p-values. So from this dataset we were able to validate our findings in probably one of the largest sample sets available.

▪ How did your study differ from previous studies?

Previous biomarker studies have used the whole-genome approach. These were open-ended studies that look at the expression of all genes and try to find a common signature that might be used as a biomarker. These biomarker studies are designed to give three pieces of information: first, prognostic or diagnostic information – so what type of cancer somebody has and how long they are expected to live; second, the study may reveal novel therapeutic targets; and third, provide information on how the cancer started and progressed, basically understanding the biology of disease.

There have been many such studies performed over the last 10 years, and what is remarkable is that – especially in lung cancer – very few of them have been used in a clinical setting. You look back at some of these important papers on gene signatures and fail to see any that provide any useful information despite being published in high-profile places. What I wanted to do was something that would be of potential clinical use.

Another great failing of these papers was that although they provided a prognostic marker (how long a patient would live with an early stage of the disease), which might be useful for guiding current therapy, none of them revealed novel therapeutic targets or mechanistic information.

▪ What are the therapeutic implications of the discovery of this predictive gene signature?

There are four potential impacts of this work: first, the early diagnosis of a tumor and its type. In lung cancer, the earlier the diagnosis can be made, the better the prognosis. Second, prognosis of early-stage disease. Determining how aggressive a disease will be permits better therapeutic planning of how to treat the disease. Third, identification of new drug targets. A major failing of prior gene signatures has been the lack of new therapeutic targets that have been discovered. NRs are themselves well-known drug targets (many are already used in cancer). Fourth, insight into the mechanistic basis of the genesis, progression and maintenance of the disease. Another major failing of prior signatures has been the paucity of new information that has come from these studies that might lead to an understanding of how the disease started, why it progresses as it does, and how it is able to elude the body’s natural defenses. The NRs represent well-characterized factors for which much information is known about how they drive different physiologic and pathophysiologic processes, including cancer.

Our plan is for a more clinically oriented future. We want to see tumor profiling of every patient that comes in to the clinic and the use of that information as an approach to personalize medicine. We want to see a break from the current strategy of treating disease, which involves giving patients a set drug or a cocktail of drugs if they are diagnosed with a certain type of cancer. There are some markers that are used, for example, the presence of an EGFR mutation (to guide certain therapies); however in general, everybody gets the same drug mix.

The idea is that these drugs sometimes work very well, but the percentage of patients for which this is true is in fact very small, making the overall outcome of the drug very poor. As a result, the question now is; how do you find who is going to respond well to a drug and who is not? That is what biomarkers are for. Every person who comes in will have a unique signature that will dictate if they need to be treated and in what specific way, not everybody will get the same drug regimen. Furthermore, since these are drug targets, not everybody will have the same number of receptors or types of receptor in their tumor, so knowing which receptor is there could lead to targeting it by using a compound to turn it on or off.

This is already a well-established principle in other cancers, for example, in breast cancer, estrogen receptor (ER)-positive breast cancer is treated with the antiestrogen tamoxifen. This is a given and it is very effective; however, you treat the patient knowing that the ER is there. Lung cancer on the other hand does not normally express ERs, but surprisingly, a certain number of patients, and this is gender neutral, randomly express levels of the growth stimulator. The question is obvious; is the ER somehow affecting the cancer? The idea would then be that, as in breast cancer, you look at the patient population and see if they have ERs or not and treat them with tamoxifen accordingly. In addition, there is anecdotal literature that antiestrogen therapy actually works in a certain number of patients with lung cancer, and one potential reason could be that it expresses ER. Not every lung cancer patient should get that drug, but perhaps the ones shown to have the receptor should get it, and would respond. Furthermore, you could combine it with known therapies that work. The ability to combine therapies is where the real power will stem from when coming up with novel therapeutics to fight these diseases.

▪ How do you intend to proceed with your research to determine how NR proteins actually facilitate tumor development and growth?

We have a number of preclinical studies planned in lung cancer cell lines and rodent models of the disease to investigate the consequence of activating or deactivating specific NRs in these models.

One of the interesting things to come from this study is the discovery that certain NRs are the primary drivers of the entire signature. So although we get a signature based on the expression of all 48, not all 48 are really required to give that information, you could distill it down to just one receptor, for example, the progesterone receptor or the orphan NR, short heterodimer partner. The latter is in fact the best single gene marker that we know of that is able to predict and give prognostic information. So the question is now why? What is the meaning of having that protein there? Is it in someway affecting the cancer or is it just a marker? One way of understanding the relationship is by modulating the expression of that receptor inside the cancer, and for that we turn to the preclinical rodent model. This involves inducing cancer in the rodent. So in the case of lung cancer, we take human lung cancer cells that either express the receptor or do not and put them in and then modulate the expression of the receptor in the tissue, that is, if we knock the receptor out, how will it affect the tumor? The prediction with the short heterodimer partner receptor is that since it is a positive indicator for outcome, when it is present, patients will have a better prognosis. So we predict that when we knock it down in a tumor that has it, the tumor will become more aggressive. Alternatively, when we express it in a tumor, that tumor will shrink, not grow as fast, or even stop growing altogether. Then we can start asking what is the receptor doing/affecting in the cell and identify its downstream targets that we could also manipulate – that is the goal of the future work.

▪ In your opinion, what are the biggest scientific challenges facing pharmacologists today with regards to personalized medicine?

Identification of individual patients that will respond to specific drugs is the biggest challenge. Second is the discovery of new drugs or combinations of drugs that might be used on a patient-by-patient basis. The current strategy of treating the majority of patients with the same drug regimen has not been successful, and although drug companies may not like to hear that, I believe individualized medicine is the true way to fight cancer therapeutically.

In addition, a common challenge in translational or clinical research is coming up with proper protocols and gaining institutional review board approval, studies have to be carefully designed so that they give the right kind of information with the correct statistical analysis. You have got to be careful not to fool yourself into believing something is working when it is not!

On the basic preclinical research side, the challenge will be to design the right models in rodents that predict what will happen in a human.

▪ What other personalized medicine projects are you currently involved in?

Breast cancer and metabolic disease.

▪ Where do you think your research interests will be focused in the next 5 years?

One area will be in the clinic. Hopefully, we will be at a stage where our work will be used in a clinical setting to begin realizing the personalized medicine approach. What we would like to do very quickly is to start clinical trials using these receptors in patients that have the disease.

Other areas that the laboratory are focused on are coming up with novel approaches to target metabolic diseases and parasite diseases.