Hey Dipon,

usually, investigation of primary samples would be much better than the use of cell lines. You have to keep in mind that all cancer cell lines change progressively during culture (copy number variations, culture-induced mutations, chromosome aberrations, epigenetic profile and so on...). Generally, every mutation providing evolutionary benefits to cultivated cells will be accumulated and may provide misleading data since its not easy to distiguish between "biomarkers" of original cancer species and biomarker just upregulated or changed during culture.

If the use of cell lines can not be ommited, you shoud investigate at leat 3-10 different cancer cell lines of the same origin (e.g. 10 different glioma lines)...

In such approach, full genome micro-arrays (healthly tissue as a control) may be useful. Potential biomarker cancidates can be further investigated at protein level using westen blotting or immunocytochemistry....

Good look,

Darius

Hey Dipon,

usually, investigation of primary samples would be much better than the use of cell lines. You have to keep in mind that all cancer cell lines change progressively during culture (copy number variations, culture-induced mutations, chromosome aberrations, epigenetic profile and so on...). Generally, every mutation providing evolutionary benefits to cultivated cells will be accumulated and may provide misleading data since its not easy to distiguish between "biomarkers" of original cancer species and biomarker just upregulated or changed during culture.

If the use of cell lines can not be ommited, you shoud investigate at leat 3-10 different cancer cell lines of the same origin (e.g. 10 different glioma lines)...

In such approach, full genome micro-arrays (healthly tissue as a control) may be useful. Potential biomarker cancidates can be further investigated at protein level using westen blotting or immunocytochemistry....

Good look,

Darius

Yes, agree. The first step should always be the search for differentially expressed (or differenrially modified, e.g. glycosylated etc.) proteins in clinical samples (vs. healthy tissue of the same origin) followed by in vitro validation.

Greets,

D.

Cell lines are un approach method of analyse not the objet !

cell lines are not appropriate to understand for biomarker analysis, It would required to see the significant differential alternation in cancer and normal cells. In 2D cell culture of particular cell lines the population quite changeable by their micro environment so, it is hardly to say the proteomic changes as biomaker for cancer. In such case primary tissues are primarily required for validation by various different techniques like IHC and blottings etc...

We have found that data from panels of cell lines can be very useful for discovering biomarkers in cancer, particularly if the panel is quite large, like those for breast cancer (N ~60). Differential expression of genes can be easier to identify in cell lines than in primary samples because the tumor microenvironment dilutes the mRNA from the cancer cells. Of course, all biomarker findings need to be validated with primary samples. In our experience, the cell line results have translated quite nicely to patient cohorts.

I don't agree with Jennifer. In vitro studies cannnot directly carry out on the patients. Physiology means control systems. There is no any external control system in the cell line. First, according to microarray results, to understand intracellular control mechanisms, shRNA and miRNA studies should do to the possible target gene. Then, in vivo studies should support these in vitro results. If you can develop a model for the target gene, maybe you can start to translate to patients cohorts. Unfortunately most of the molecular biologist ignore physiological and environmental factors. There are a lot of paper arrogating to recover the world.

By their very nature, cell lines are inherently different to actual tumor tissue. Firstly, most cell lines are artificially immortalized by some means (though there are exceptions, such as HeLa cells). Secondly, cell lines are passaged numerous times and as such likely acquire multiple additional mutations that favor their growth and make them genetically distinct as a clonal population. Thirdly, cells are cultured in a 2D system as a single clone, whereas tumors grow in a 3D microenvironment along with other cells types (e.g. stromal). For these reasons cell culture systems are artificial and, In my opinion, cannot be used as an exclusive source for biomarker discovery.

In addition to the reasons already given, cell lines growing in a dish are in a different microenvironment entirely from those same lines growing in an animal, or even from the original source tumor. But the real problem is people keep looking for the "magic biomarker," the one marker that will detect cancer with perfect fidelity. There is no such creature. Biomarkers exist within a complex matrix of protein expression. It is a system, but the system is a bit unpredictable because it is an interacting system. Thus, sensitivity and specificity are less than perfect. See my opinion piece in BMC Urology "Does the biomarker search paradigm need re-booting?"

In bladder cancer, there are two, main tracks by which cells become malignant. The first involves FGFR3 signaling, and the second involves p53-cell cycle. These two tracks can involve different mutations within the networks, so no one protein has either the sensitivity and specificity. What is needed is a small panel that covers the range of disease, but deriving such a sample is not simple nor intuitive.

If cell lines are used, their only possible use is as a starting point to derive candidates that are then validated against tissues. BTW, the usual starting point for biomarkers of analyzing a bunch of cancer specimens and finding a list of 20-100 markers, as was done with breast cancer, is not robuse for reasons given by Liat Ein-Dor and colleagues. The only advantage to cell lines is that only one cell type is involved. Growing them in Matrigel might yield something similar enough to a real tumor to stand a chance of identifying candidates for evaluation against real tissues if there were enough cell lines (I would guess 20). But, in general, my answer is not to expect much correspondence between cell lines grown on plastic and real tumors.

You can grow the cell lines and can try to get the biomarker but ofcoure its not a guarantee that they would be specific for certain type of cancer. However, you can use high through put screening such as LC-MS, ITRacQ to get the proteome of a cell line as well as you can create specific conditions where a particular cancer become more aggrasive phenotype. Once you get the proteome of a cell line under normal conditions, you can perform hypoxia and hypoxia reoxygenation to mimic more aggressive metastatic phenotype. You can compare all these conditions with your normal conditions. Hopefully, you will get some proteins upregulated under hypoxia and hypoxia reoxy conditions as compare with Normal.

In a different approach, you can treat the cells either with EGF or TGF beta if you are working with epithelial cell lines as once you add the growth factors after 48-72 hrs cells will undergo epithelial meshenchymal transion (EMT) which is again an indicator of metastasis. Once you get different condition in same cell lines you can again compare the proteome with LC-MS/ITRaQ.

Hope it will be helpful for you.

The serum proteome is so complex and has 10^12 proteins and most of proteins are common to all disease conditions. Hence you need to go for primary cell culture of tumor tissues, identify some proteins, and evaluate the same from serum of patients Vs Controls for validation, using LC-MS/MS. And you need to have large sample size of patients and controls for the study.

regards

siva

In the long run, we also need to consider within and between person differences in metabolism response, repair function, epigenetics, etc. before one or a few proteins can be considered to have diagnostic relevance. I agree with Sivasankar that sample size is crucially important and further suggest that candidate compounds must be subsequently validated in independent studies.

Along these lines - I'd like to solicit information from the community about concepts and data dealing with internal human variability. We have a lot of cross-sectional information from humans and longitudinal data from cell lines, but very little connecting the two. The cells (or animals) tend to be genetically pure and so do not give a genetic variance component in response.

So, does anyone have any thoughts?

Ioannidis' study concludes that the initial claims on the prognostic and/or predictive value of biomarkers are overstated: JAMA. 2011 Jun 1;305(21):2200-10. doi: 10.1001/jama.2011.713. (see also his video regarding hypothesis testing http://www.youtube.com/watch?v=hBNeuG10-ac, and the a priori assumption that one can have when e.g. identifying a biomarker!). I echo Joachim's responses with respect to the immense human variability. In order for a biomarker to make into clinical practice, we need to realize that most cell lines as well as rodent models are merely poor reflections of human disease in the general population. See also Begley and Ellis' commentary in Nature 2012 Mar 28;483(7391):531-3. doi: 10.1038/483531a. Whatever tool one wants to use to identify a biomarker, its true power must be rigorously validated in human samples in order to obtain the highest level of evidence. You will obtain the lowest level of evidence when using a series of cases of human samples with an indirect surrogate as endpoint such as disease-free survival (this accounts for >90% of all biomarker studies!). The highest level of evidence is obtained in randomized controlled clinical trials with (disease-specific) mortality as an endpoint. Most 'biomarkers' that are identified using cell lines do not pass these strict criteria.

YES!! Thank you Leon! This is a real problem that is overlooked in the scramble to screen many chemicals with "high throughput"; just because one sees a mode of action in a cell line that seems adverse does not necessarily mean anything adverse in vivo, nor is the inverse true.

Cell lines alone are totally useless for detecting cancer biomarkers and they can be very misleading...moreover a single biomarker (even if catched from a large population of primary tissues) is of very limited use in diagnosis given the discriminant ability must be almost maximal to be effectively of use , given the relative low a priori probability of the event (is a matter of Bayes theorem of probability).

I think we should make an effort for develping 'multidimensional markers' based on metabolomics in which an entire pattern is taken in consideration for diagnosis.

I agree that data from cell lines cannot be useful for biomarkers. Now, there are companies which do 7-10 day culture of biopsies from human tumors-such cultures are apparently useful for doing quick chemoresistance testing, and they help the oncologists in real-time. If this is true, then 3D cultures of at least some cell lines may yield meaningful data?? A systematic, comparative study of 3D cultures of known cell lines and primaries from the relevant stages of cancer biopsies needs to be done!

Short term cultures of fresh material is a good option, and 3D is better than 2D, but using estahblihsed cell lines gives you all the disadvantages as discused above.

Alessandro: One indeed needs to understand the interaction between biomarkers in a multidimensional setting as was shown for BRAF/EGFR/KRAS/PTEN/PIK3CA in colorectal cancer, e.g Clin Colorectal Cancer. 2010 Dec;9(5):274-81

Not to mention the precursor problem: What triggers cancer mode of action in the first place? That is, can we have a healthy cell culture that can be tested against new commercial chemicals to assess potential adverse outcomes before they (the chemicals) are produced in high volume and enter the environment.