Applying science to green tea

{NOTE: Here is the post that was delayed last week due to my announcement of arson at the Holocaust History Project.}

It occurs to me that I haven't done much straight science blogging lately. Yes, debunking pseudoscience and quackery is fun, useful, and has the potential to educate people about how science is misused, but this is ScienceBlogs. Since arriving here four weeks ago, I haven't fulfilled my quota of science blogging, and it's time to remedy that.

Fortunately, while perusing a recent issue of Cancer Research, I found just the ticket, something that would let me discuss science and still stay related to one of the main themes of this blog, alternative medicine.

The paper? A Gene Expression Signature that Can Predict Green Tea Exposure and Chemopreventive Efficacy of Lung Cancer in Mice, out of Washington University, Japan, and the Chemoprevention Branch at the National Cancer Institute. Here's the abstract:

Green tea has been shown to be a potent chemopreventive agent against lung tumorigenesis in animal models. Previously, we found that treatment of A/J mice with either green tea (0.6% in water) or a defined green tea catechin extract (polyphenon E; 2.0 g/kg in diet) inhibited lung tumor tumorigenesis. Here, we described expression profiling of lung tissues derived from these studies to determine the gene expression signature that can predict the exposure and efficacy of green tea in mice. We first profiled global gene expressions in normal lungs versus lung tumors to determine genes which might be associated with the tumorigenic process (TUM genes). Gene expression in control tumors and green tea-treated tumors (either green tea or polyphenon E) were compared to determine those TUM genes whose expression levels in green tea-treated tumors returned to levels seen in normal lungs. We established a 17-gene expression profile specific for exposure to effective doses of either green tea or polyphenon E. This gene expression signature was altered both in normal lungs and lung adenomas when mice were exposed to green tea or polyphenon E. These experiments identified patterns of gene expressions that both offer clues for green tea's potential mechanisms of action and provide a molecular signature specific for green tea exposure.

You can see why it caught my eye. I'm a sucker for an interesting gene expression profiling paper.

Basically, the investigators started by looking at mouse models of lung cancer induced by different chemical carcinogens and asking whether green tea or the major constituent of green tea thought to be responsible for its chemoprevention properties, polyphenon E, could prevent these tumors. This process is known as chemoprevention. Next, they looked at gene expression profiling of the tumors in order to determine whether a "signature" could be derived that would identify which tumors are more likely to respond to chemoprevention by green tea.

Gene expression profiling uses what's known as oligonucleotide microarrays. These microarrays contain sequences for thousands of genes, and can be used to survey differences in gene expression over entire genomes. The result is an array with squares or dots ranging from red to green, each representing a single gene, with red generally meaning higher gene expression than in the control and green meaning lower expression. It's a powerful tool. (Sadly, it's also still a fairly expensive tool; otherwise I'd be doing a lot more gene expression profiling in my lab.) Basically, using this technique, it's possible to measure changes in the level of the mRNAs for every gene in the genome. To analyze this data requires sophisticated computing algorithms, such as cluster analysis. More recently, software has been developed that allows one to identify intracellular signaling pathways that are turned on or off based on the pattern of genes that are turned on or off. One such tool is GeneMAPP, which is what the authors used to identify the pathways altered by exposure to green tea or polyphenon E.

The authors first compared gene expression profiles of lung cancer versus normal lung in order to identify genes that might be associated with the tumorigenic process. They found a large number of such genes (>2,000) that were differentially expressed. The changes noted, as in most cancers, were in cell signaling pathways related to transcription and cell cycle/proliferation. Also activated were pathways involved in inflammation, which is not surprising, given the relationship between chronic inflammation and tumorigenesis.

They then looked at the effect of green tea and polyphenon E on these tumors, they found that both significantly decreased the number of tumors caused by the carcinogenic chemicals. They further found that there was a subset of genes in the tumors in the animals receiving green tea that were reversed towards levels found in normal lungs, suggesting that they might be involved in the chemopreventative effect of green tea. For example, certain cell cycle genes that were turned on in cancer were turned off by green tea, and vice-versa. Ultimately they were able to narrow the data down to a 17 gene signature that could reliably identify both lung tumors and normal lung tissue that had been exposed to green tea.The usefulness of this result is that there are a number of clinical studies out there that are looking at green tea and green tea-derived compounds for their chemopreventative activities in a variety of cancers. it would be useful to have a gene expression profile that can be used to determine whether the tea or compounds was having a measurable biological effect.

One thing bugged me about this paper, however. Maybe I misinterpreted it, or maybe it's just my background in clinical trials, where gene expression signature predictors mean signatures that predict response to treatment or prognosis, but to me the title implied that the authors were going to present data deriving a gene expression signature that could predict whether green tea would have an effect preventing lung tumors due to the chemical treatment. True, this would have been difficult to do rigorously, because it would have involved taking a multiple lung tissue samples from mice, first a pretreatment sample and then a sample after treatment with green tea in order to see if they showed changes in this 17 gene signature consistent with activity of green tea. Then the treatment with chemical carcinogens could begin, followed by green tea, to see if the presence of the 17 gene signature in response to treatment with green tea predicted which mice would demonstrate better chemoprevention with green tea. Logistically, this would have been a hell of an experiment to do. I know what they authors mean when they say that their gene expression signature predicts an effective dose of green tea. Basically, they mean that their expression profile can discriminate between untreated lung tissue and tissue of mice treated with an effective dose of green tea; i.e., predict which sample was treated and what sample was not without knowing which is which.

But this is mainly a nitpick about the title, which seemed to promise more than the paper actually delivered. (This should be a lesson to budding scientists to be careful about their titles.) Overall, it's a rather cool study showing that, contrary to the claims of some alties, it is quite possible to apply state-of-the-art science to alternative medicine claims.

Hmm. That wasn't a particularly insolent post, respectful or otherwise. Where are some creationists when you need them?


More like this

Tuesday - let's take a look at 4 out of 7 PLoS journals. As always, you should rate the articles, post notes and comments and send trackbacks when you blog about the papers. You can now also easily place articles on various social services (CiteULike, Mendeley, Connotea, Stumbleupon, Facebook and…
DUSP6! MMD! STAT1! ERBB3! LCK! Was ist diese? CB radio patter? Pilot to co-pilot chatter? 3rd-and-long huddle banter? Of course not - that would be too easy. These cryptograms represent five different genes found within non-small cell lung cancers (NSCLC) that predict the both the relapse-free…
Well, we've gone over why tea supplements aren't a good idea, but the question of chemoprevention by drinking tea is still up in the air. I thought it would be interesting to lay out what's been done. Let's start with the basics: How would tea inhibit carcinogensis? No idea. There are dozens of…
We dwell a lot on the many unknowns about the bird flu H5N1 virus. What could make it easily transmissible between people? What determines what host it infects? What makes it so virulent? With the threat of a pandemic looming it can sometimes seem the virus is an especially, maybe even uniquely,…

These microarrays contain sequences for thousands of genes, and can be used to survey differences in gene expression over entire genomes. The result is an array with squares or dots ranging from red to green, each representing a single gene, with red generally meaning higher gene expression than in the control and green meaning lower expression. It's a powerful tool.

And well worth considering a couple of key deficiencies as well. The first thing that should be noted with any microarray analysis is they are very rarely final. For example, in this study they found 17 genes that had their gene expression altered by 'treatment' against 'non-treatment', which seems good but bear in mind certain caveats. Firstly, a 'change in gene expression' is just that, we could be talking about a very minimal difference between the changed/non-changed expression.

In fact, the alteration of a certain proteins expression may not be a lot and may not be biologically significant. This is even if the expression originally appears to be statistically significant. As a result, a microarray study can only really give you a clue as to where to start looking. The use of more rigorous molecular techniques that tell you the actual amounts of expression are more conclusive.

The use of more rigorous molecular techniques that tell you the actual amounts of expression are more conclusive.

And even that is often not as conclusive as many would like to believe. The fact is, gene expression is several steps removed from phenotype. A change in gene expression may not correlate well with a change in the expressed protein concentration in the cell, due to translational and post-translational regulation, protein turnover regulation, etc.

But it's still neat, and a good example of how science can be applied to altie medicine.

to me the title implied that the authors were going to present data deriving a gene expression signature that could predict whether green tea would have an effect preventing lung tumors due to the chemical treatment

I had the same response to the title. Good to know it's just a bad title, and not a fundamental divergence in language between physics and biology.

And yeah, I would have loved to see the study that was implied by the title, but hey, still cool.

More green tea, please!

By Johnny Vector (not verified) on 13 Mar 2006 #permalink

" this study they found 17 genes that had their gene expression altered by 'treatment' against 'non-treatment'..."

I'm a layman, but what about 'treatment with green-tea' against 'treatment with other stuff'?

Joseph: You're spot on about interpreting fold-change in signals with micro-arrays. Having done a lot of work with them myself, essentially one throws out anything less than 2-fold change as insignificant (as it can't be clearly isolated from the unavoidable background noise).

mighty Orac: Without reading the paper, I'm going to hazard a guess that the researchers used an in-bred strain of mice, perhaps two or three, which were originally bred solely for genetic uniformity. It's a fairly well-accepted (and generally safe) assumption that if X number of mice have responded in a certain way, then Y number of mice of that same strain are going to respond the same way. Yes, there are caveats (the research facility housing conditions being a primary one), but insofar as regards to this paper, the title is an accurate one--if limited to the strain(s) of mice used. Now, if they started cross-breeding the strains to tease out which particular genetic loci were responsible, THAT would be an interesting paper. However, it's incredibly difficult and time-consuming work.

And as a silly note, your favorite: Liberator or Scorpio?

I forgot to mention:

Occasionally microarrays don't catch gene regulation that is known to occur from other causes. For example, I'm studying a gene that upregulates another gene by more than three-fold by Northern and Western blots. However, on the microarray, the upregulation was only 1.4-fold.

I certainly agree that a gene expression change is not neccessarily meaningful, and frequently is just a place to start. But I would say that a gene expression change is a phenotype in ittself and if you have an expression signature, that can be a phenotype to study in itsself.

I may be biased, but for my money (paltry though it may be) you get the more out of microarry experiments than most other omics. And when someone has done a really well controlled large experiment the possibilities are endless (you mouse guys do know about Genevestigator don't you?).

It is really expensive, but of course it should start to get cheaper. From what I here out of Affy, I think the cost per hyb will come down significantly within a couple years.


ps. I'm sure I am speaking to the choir, but using RMA/GCRMA instead of the standard affy processing can improve the accuracy and reliability of your results.

Great review of the paper. I should've known that Ming You was the senior author - he's a fabulous scientist and a leader in lung cancer biology. The concerns of JM O'D and others notwithstanding, I have three points to add:

1. Neither of the green tea products were well-characterized chemically beyond noting the concentrations of EGCG. I suspect that the Japanese company providing the polyphenon E has very detailed analyses of chemical composition, but this is the major oversight in far too many studies of natural product extracts: i.e., if you get X results, what guarantee do you have that repeating the experiment later with the same product will produce the same results? What good is all the microarray and/or protein expression data if polyphenon E produced next year or by a different company lacks critical compounds to produce the same effects.

For example, the failed NEJM trial of Echinacea in experimental rhinovirus infection contained some 22 pages of supplementary data on the chemical composition of each Echinacea extract used, so we at least know the levels of compounds that are ineffective. I'd really like to see journal editors and reviewers start requiring thorough analysis of such multicomponent mixtures before one gets too deep into the expensive biological studies.

2. Interesting that DNA topoisomerase I was one of the overexpressed genes in tumors suppressed by the natural products. Orac know this (among many other things), but irinotecan (Camptosar) is a commonly-used lung cancer chemo drug that targets topo I, converting it into an intracellular poison by stabilizing protein-linked strand breaks that cause replication fork collisions. But this leads to a question...

3. Do the green tea products also transiently suppress topo I expression in tumors such that they might confer irinotecan-resistance if used during chemo?

Way cool. Certainly it's not the last word on green tea but it's convinced me it's worth spending a couple bucks, and it'll make a great launchpad for future work.

Regarding chemical compositions, I assume the use of both a single-compound extract, and "whole" tea, was meant to cover a smaller question: Is there anything *else* in the tea that we ought to be looking at?

My own big question is, just how solidly are the genetics of the "cancer-prone mice" connected to the development of environmentally-caused cancers in humans?

By David Harmon (not verified) on 14 Mar 2006 #permalink

And let's not forget: Green tea tastes great and doesn't stain your teeth or give you bad breath like coffee and black tea do. Too bad it is not more generally available in the USA and when available often adulterated with jasmine.