Among a couple of very interesting posts received as comments in Ciencia Hoje blog (http://cienciahoje.uol.com.br/noticias/2013/04/polemica-a-mesa), I got a message from Vinicius Vilperte arguing that risk assessors should use omics to add valuable info to their risk assessments. Even more, he argues that Southern and Northern blots and many other conventional molecular info supplied by the applicants in risk assessment dossiers for the risk agencies are not very informative.
I would like to argue against most molecular data supplied in these dossiers. Indeed, in my opinion, which is coincident to that of other risk assessment specialists, molecular data do not significantly contribute to risk assessment. I had this feeling since long ago, but it was very much reinforced during a Hesi meeting I attended in Paris two years ago and the feeling was consolidated in the last year, when a couple of risk assessment specialists wrote the first guide on GMO environment risk assessment and I was among them. It was a long period of apprenticeship (which, in fact, never ends), but I am confident my feeling was right and will try to explain the main issues to the reader.
I kindly ask the reader to keep in mind that the GM plant in the market is not a regular event produced by a researcher in his lab: it is the result of a very stringent selection of thousands of candidates that need to behave EXACTLY as expected. Hundreds of different parameters are evaluated in dozens of different situations and the final ELITE EVENT has all the expected phenotype and no unexpected characteristics.
Let us begin with a sensible list of molecular data that can be generated by a reasonable cost
a) Sequence of the insert
b) Sequence of the flanking regions
c) Chromosome localization of the inserts by FISH
d) Chromosome localization of the inserts by Southern blot
e) Specific gene expression at the RNA level by Northern blot
f) Investigation on other RNAs by Northern blot
g) Insert stability by PCR analysis
h) Specific gene expression at the protein level by ELISA and Southern blot
We can add to the list some more expensive approaches.
a) Super SAG tags or microarray data on gene expression
b) 2D gels for protein profile studies
c) Maldi-TOF and other proteomic approaches to generate detailed protein profiles
d) Metabolic flow charts
e) Whole genome sequencing
f) Large scale direct RNA sequencing
The first list of data is frequently available in dossiers and papers. What does it tell us that could be useful for risk assessment?
· The exact insert sequence is important to evaluate the potential for the production of truncated peptides and many other problems derived from breaking and rejoining the genome and the insert. It is certainly informative.
· Sequencing the flanking region is important for PCR detection of the even, but does not bring useful information for the risk assessor. Many colleagues may argument that it is nice to precisely identify the insert location, since it may have disrupted important genes and created spurious ORFS. Yes, that may happen, but if such a thing happens and has detectable deleteriouys consequences, it will most possible lead to unstable expression of the transgene, changes in the regular protein/lipid/sugar profile, in the hundreds of agronomical parameters evaluated, etc. Unanticipated changes do occur, but any event having unexpected phenotypes is discarded in the selection process, as alerted above. There is presently not a single report of this kind of information as important for any specific risk assessment, from hundreds of elite events analyzed in commercial release assessments.
· Chromosome location (by both methods) and copy number are nice information, but are really not informative. The important information is insert sequence. If the event has more than one such copy, it is important to know all of them, not that much because of safety reasons, but to help on the interpretation of all experimental results using molecular biology. Similarly to what I said before, there is not a single evidence that copy number is important for biosafety and risk assessment.
· Gene expression at the RNA level may be useful as a tool to understand how the transgenic plant deals with the product and its eventual translation. But, like chromosome location, it is not important for risk assessment: a lot of intervening issues cross the road to protein expression, which is the important quest in most cases. Obviously, specific RNA expression may be important in iRNA-based GM plants, but even so we can´t generalize its usefulness as a bare info for risk assessment.
· The use of blots to detect other RNAs than the target ones is disputable, but if the researcher has the precise sequence of the insert, there is no reason to believe that unanticipated RNAs will be a source of problem. They may even exist, but if they do not change the many hundred parameters that are measured to ensure that the event elite is behaving as expected, they are not important for the risk assessor.
· Insert stability (its presence and expression) is obviously very important for the technology, but is absolutely useless for risk assessment. Data usually support the plea that the GMO is doing what is expected from it. What if the insert is eliminated? The GMO will not work properly, but it will not be more “dangerous” than the original event.
· Specific gene expression is a very important issue, especially if data is produced to demonstrate expression levels of the relevant proteins in different tissues, in different conditions. This is something that has consequences on an eventual deleterious impact of a GMO. However, not every protein should be investigated: data must be generated on the proteins encoded by the transgenes. A plant has hundreds of thousands of different proteins and it is nonsense to try to investigate changes in all of them. If any subset is selected, due to technical limitations, it is a bias that completely invalidates claims of “broad” expression analysis for risk assessment use: the discarded subsets can well have the “relevant” proteins.
Summarizing the analysis of the first block of techniques, both insert sequence and protein expression levels (sometimes RNA expression levels also) are important molecular data. Everything else is nice to know, but has very limited use in risk assessment.
Let us now discuss the omics block.
· Transcriptomics, either using SAGE, microarrays or other techniques, can produce an incredible amount of data in very short time and the costs are sinking fast. However fancy, this data is of very limited use for risk assessment and the main reason is: we don’t have large enough baseline data as to be able to come to any significant conclusion related to risks. Moreover, even if consistent changes of a certain group of genes are observed between transgenic and non-transgenic plants, it may and possibly will have no meaning for risk assessment unless a route to damage could be established for any of these genes and gene products or for the metabolic change, as a whole. This is a daunting challenge and we are VERY FAR FROM BEEN ABLE to use the info.
· The same can be said of proteomic data, with the aggravating circumstance that proteomics is still unable to produce data from a large set of proteins. As said above, if a selection is made, we may be missing the “dangerous” protein, left in the other subsets. Moreover, as for transcriptomics, a very limited info is available for crop baselines. And again, as for transcriptomics, even if we have some reproducible changes in protein profiles, it is very difficult, if not at all impossible, to correlate them with risk issues. The use of proteomics for risk assessment is an even larger challenge that the use of transcriptomics. I will come back to this point when I discuss Zolla´s results, below.
· Some people claim we should re-sequence the whole GMO genome and try to find any unexpected changes. The reader will be surprised to know that cells from a same organism can have slightly different genomes! It would be very difficult to know what was caused by the transformation and other post-transformation events and what was natural.
To conclude, none of these new omics have presently a clear use in risk assessment, even if their results are fancy and informative for researchers in different areas.
The reader could well ask: if molecular data are of little use, what are the main elements for a sound risk assessment?
As stated above, risk assessment is not a mysterious process: it is a structured process in five steps, starting with the definition of the context and ending up with a decision on risk acceptability. All the molecular data discussed above is just a small part of the context. Many other information is also important, maybe much more important: all phenotypic data are of paramount importance and most of it is not of molecular nature. The biology of the parental organism is also critical. It is therefore wrong to try to impose molecular approaches as vital, indisputable and essential for a good risk assessment. The formal proof that the present approach to a good risk assessment is valid is the fact that not a single reliable report is available on health or environment harms due to commercially released GMOs.
The figure below represents the five steps on risk assessment.
Figure 1: Five steps on risk assessment. The reader should have a close look on the five ellipses over the first step. Moreover, he should bear in mind that for the estimation of both probability of occurrence and magnitude of the associated harm a pathway to the harm must be established for every danger imagined. Following this approach it is possible to discard most imaginary dangers and concentrate on the relevant issues.
To conclude this discussion, I would like to emphasize that much information derived from lab experiments serves as warnings to the risk assessor, but by no means should be regarded as essential. Epigenetic modifications in gene expressions are a well know and much discussed phenomenon and it may change gene expression in GMOs, but the stringent selection of the elite event effectively eliminates any important change due to it. This is also true for most other gene modulation phenomena. One should never forget we are not dealing with lab experiments, but with long studied elite events.
I hope to have been clear enough as to be of some use to the reader, as an answer to Vinicius, who I suppose is now in the beautiful but cold Norway.