RA&RM of LMOs with Stacked Genes or Traits

Dear All

Thank you for all comments received so far.
The question was raised if it would be too early to prepare a guidance document.

I would like to encourage comments on this point and especially on the scope of a document because there are some conflicting views on this.

Thank you very much!

best regards 

Beatrix

I don’t think there is that much disagreement.
Probably all contributors agree that stacked genes that already have undergone a risk assessment as ‘single genes’ in the same genetic background, only offer a new challenge for risk assessment if their gene products interact in any way so that a new phenotype arises that has not yet been assessed in the previous risk assessments for the single genes.
If not, then there is no reason to perform a new risk assessment.

If a new risk assessment is needed then this risk assessment should be done according to the principles of the roadmap. As the roadmap is supposed to cover the risk assessment process of any GMO, it should provide enough guidance. After all, stacking genes is just a way of making a new GMO; you could also have brought the two genes into the plant in just one event, and then the fact  that these are formally not ‘stacked genes’ should not preclude that you are able to a risk assessment.

If you find new points to consider that are not yet taken into account in the roadmap, then these points to consider should be added to the roadmap, not to another document.

The guidance document on stacked genes could be short if we just make these principles clear: when is a new risk assessment needed, and if it is needed, the roadmap should provide enough guidance. Still, there would be a lot of merit in such a guidance document, just because it shows a consistent straightforward approach of a situation that is commonly seen as a difficult issue.

For the underpinning of the guidance with available literature, I agree with David that we might add a number of documents from different legislative systems. I think however, that we should not focus on the legislative part, but rather on the scientific underpinning of the approach that has been taken.

I agree with the comments of Hans-Jörg Bukh, but many of these issues would not come up in a document as I envisage here.

I think the recent interventions from Hans and others have helped clarify the direction this guidance document should take by highlighting the necessary relationship any quidance document should have with the Roadmap.  As noted by others, all "new events" will have been subject to a risk assessment, which should have been conducted in a manner harmonized with the Roadmap.  This leaves a potential gap that may be more of a policy challenge than a risk assessment challenge.  How a country views stacks may be addressed through their national policies, and may not necessarily have a relationship with science or the risk assessment.  Those Parties that choose to use a risk assessment approach may look to this guidance for help.

As we go forward with developing language stacks and describing how to appropriately assess the potential for interactions I feel we will need to take care to be accurate.  For example we should consider that the risk assessment focuses on the gene products rather than on gene interactions themselves.  In the case of a stacked Bt product, one would examine the potential for the Bt proteins to interact in a manner such that they must be tested in combination to adequately assess the potential for adverse effects (harm) to result.  In other stacks, this question may have been adequately addressed without additional data e.g., a stack of a herbicide tolerant trait with a Bt trait.

Because genes and genomes (regardless if they are transgenic) are dynamic and interact in some manner for functionality, assessing interactions at the genetic level is exceedingly complicated and the information has no obvious relevance to the risk assessment.  Our use of terms like genetic instability should be replaced with expression or phenotypic variability.  A risk assessor should look at this variability first and then determine whether it is relevant to the potential for an adverse effect to result i.e. there is evidence of interaction(s) relevant to the risk assessment. 

I strongly support the simplifications being suggested, and clarifying the link to both the Roadmap and existing national policies that cover stacks.

I may be slow in catching up with the current focus, however, I would like to confirm two things, which also associate with the comment buy Tom.

This message consist of two components:

1) Comments based on the plant breeding practices
2) Japanese experiences on  stacking genes in LMOs or transgene stacking


1) Comments based on the plant breeding practices
It is common plant breeding practices on stacking genes or combing genes with the trait(s) of interest. The safety assessment of the new varieties are product based approach, and phenotype is actually more important than the genes which are responsible for the trait(s). Also Attention to G X E is very basics of the plant breeding, and with following the practices, there is no major deviation on elements for RA needed on stacked genes.
2) J-BCH / GOJ: Japan has been examining a lot of cases of stacked genes associated with environmental biosafety associated with AIA. On the aspects over several dozen of scientists with very high qualification with diverse disciplines, examine the RA on stacked genes, and the general understanding from GOJ is, more on phenotype and product based RA.

The present discussion seems, in some part imaginary rather than cautious, and the fact from practices shall be referred more to consider practical suggestion to guiding document.

Kind regards,

Kazuo Watanabe

Thanks to all comments so far.

As Prof. Watanabe has mentioned, Japan has conducted risk assessments of genetically modified crops with stacked genes/traits. We assess environmental risks of LMOs on case-by-case basis.

In risk assessments of GM crops with stacked genes/traits, we think that phenotypic interaction is more important rather than genotype interaction because gene interaction/change do not always result in phenotype expression.

As I have read all comments, I have confidence and stressed that what important is “interaction”.
But I think that it is harder to determine whether or not there is interaction of gene/gene products in stacked LMOs, of which metabolic pathway is engineered.

Thus, it would be very useful and helpful for all assessors if general method is referred in guidance document/roadmap how to decide whether or not there is interaction.

Thanks,

Kazu.

I support Hans’ proposal to limit additional risk assessment to those stacks where their gene products interact in any way so that a new phenotype (new trait) arises that has not yet been assessed in the previous risk assessments for the single genes.  As Hans has proposed, then the risk assessment of that new phenotype would be done according to Annex III, with the roadmap as guidance.  Starting with a complete re-evaluation of the stacked product is not supportable because it ignores the fact that much is already known about the individual parental events.  A sound scientific approach, as called for in Annex III, would make use of pre-existing information as the basis for identifying new areas of concern.

The focus on phenotype makes sense because in the Protocol, we are focused on conservation and sustainable use of biodiversity.  Whatever is going on at the molecular level with regard to biochemical and molecular interactions (including such things as position effects, etc.) would only have an impact on our area of concern through the phenotype.  Thus, while molecular analysis might be useful to explain a new phenotype, one should not have to conduct a molecular analysis unless it is useful to understand the new phenotype for the purpose of risk assessment.

The preparation of a separate guidance document would be confusing and redundant with the elements already included in the roadmap.  In the case of the key elements currently in the draft outline, elements 2 and 3 are already covered by the roadmap.  Element 2 mentions compositional analysis, but that is primarily relevant for food safety.  Compositional analysis as a first line of assessment, before knowing whether a new phenotype or trait exists, does not add value to the risk assessment that is the subject of Annex III.  Furthermore, I question the rationale for element 1.  It seems to me that the assertion, “The combination of transgenic traits via cross breeding may change the organization of the inserted genes/gene fragments…”, should be supported by evidence of such things occurring in stacked events that already exist.  Speculative rationales should not be the basis for risk assessment guidance.  The second part of the rationale (influence on the regulation or expression of the transgenes) is covered under element 2 of the current outline, and therefore is also redundant with the roadmap.

Dear all,
unfortunately I have not been able to contribute to this very important discussion earlier. However, after reading comments by others I would agree with Hans: a short, separate (of course annexed to Roadmap) would be very helpful. This is a difficult issue that needs some quidance - as the discussions clearly show.

Marja Ruohonen-Lehto

I do agree with Hans and comments from others on the issue that stacked events only need another risk assessment if their gene products interact in a way that a new phenotype arises that has not yet been assessed in the previous risk assessments for the single genes.

I also agree with the notion that when a new risk assessment is needed then it should be done according to the principles of Annex III and hence according to the roadmap.

So the focus of this SWG, should be on identifying new points to consider that are not yet taken into account in the roadmap that will be needed to assess risks of LMOs with stacked events. Then these points to consider (if any)should be included in the roadmap, since the roadmap is supposed to serve as guidance for any LMO, including those with more than one new trait. This exercise will clarify that there is no need for another document.

Part of the process to identify the need for additional guidance could include to apply the roadmap, during its testing stage, to a particular case of a LMO that has stacked genes. With this we could fulfill two purposes: making sure that the objective of the roadmap as a guidance for risk assessment is accomplished, and that its range of application is adequate.Still if something is missing and we can determine that it would require further guidance, it could be added as a short document or annex to the roadmap, emphasizing the differences that the risk assessors need to take into account when dealing with LMOs with stacked traits.

Greenpeace welcomes the opportunity to comment on the AHTEG and, in particular the section on stacked traits

Indeed a guidance docuemnt is warrented as there is much deabate currently regarding the assessment of such GMOs (e.g. EFSA in the EU). It is important that it is recognised that assessment of such traits cannot be made solely on the basis that "A" is Ok and "B" is OK, therefore "A x B" must be OK. The interactions between the genetic inserts and traits (including implications to farming prcatices and biodiversity), their stability and loci are all important factors that must be considered. Hence, GMOs with stacked traits can only every be assessed individually.

Dear all,

So far I haven’t been able to comment on the draft outline for ‘stacked events’, and find the deadline for comments somewhat tight - I think it would be good to extend it for further discussion. 

In the following I will first comment on the outline directly (1) and in the latter part respond to comments already posted by others (2).


(1) Direct Comments on Draft Outline

I find the draft outline ‘stacked events’ – dated 9 June 2009 - very helpful as a basis to build on and do mostly support its content and structure.

The Introduction will obviously become more understandable once fleshed out more. But just to clarify: I am not sure what exactly is included or excluded within the term “stacking” in the second intro bullet at present – as I understand its intended meaning (and following our previous round of discussion) it is to be an all inclusive description of gene stacking.

Of course we were not to develop additional guidance to all possible manifestations of gene stacking, but –at least at present- only for those resulting from the cross-breading of already existing parental LMO lines and possibly the additional or secondary transformation of already established LMO plants. At present the description of our ‘Scope’ seems to be under the heading of ‘Definition’.  We should better move that section into the beginning of the ‘Risk assessment’ section and call it something like ‘Scope of Risk assessment guidance’.

This 'Definition' or rather 'Scope' paragraph still has some ambiguities and I would like to place the word ‘parental’ in its fourth line in square brackets until we have solved the exact scope for this particular guidance.

Concerning Introduction bullet point four: Will there be cross references to/from RA points below? … I am sure we will sort it out in the next round.

Risk Assessment section:

To reflect the current ‘definition/scope’, we need to alter the ‘rationale’ of point 1 to include the combination of transgenic traits via secondary transformation of a parental transgenic line, as the integrity of the first/parental transgene loci might have been altered due to the secondary transformation and transgene insertion.

I agree with the other points covered under 1, its rationale and points to consider, including the demonstration of maintenance of the flanking regions of the insert, as they are crucial for identification purposes.

Concerning point 2, it could include further explanations of the rationale behind the points to consider, if that would be helpful to a later reader for following the scientific argument and reasoning. I regard all the rationale and points to consider listed as crucial for risk assessment of stacked events and would like to see an additional aspect covered in the third bullet point of points to consider, namely: comparison of the stacked event LMO has to also be with its as-near-as-possible-isolines. In fact, choice/availability of the right unmodified comparator line presents a problem for the risk assessment of stacked events, which needs further considerations in the development of guidelines.

Concerning point 3, some points could do with strengthening, in particular indirect (cumulative, synergistic, combinatorial) effects both in general as well as in the context of changes in management procedures. Furthermore, it would be important to address stacked events in the context of soil, including soil organisms, soil fertility, availability of nutrients, pathogens, symbiotic interactions etc.

And should there not be a clear reference to (human) health somewhere? I thought we had agreed on that at the Montreal meeting?




(2) Responses to previous comments

It is of key importance to assess combinatorial (e.g. accumulative and synergistic) effects of stacked genes, and not to assume that stacked genes could be reduced to separate individual gene products and traits.

There seems to be a misconception of the existence of combinatorial and synergistic effects. In fact these combinatorial and synergistic effects do/can not only arise from the presence of the various transgenes, but also from the presence of the side effects or the original transformation in the parental line, including positional effects, transformation induced mutations and unintended impacts of regulatory and metabolic processes and pathways.

Whilst some unexpected and/or unintended effects might have been noticed and reported in the assessment of the (single) transgene parental line, they might have been regarded as not phenotypically, metabolically, biologically or statistically significant. Both such noticed or unnoticed and reported or unreported apparently statistically or biologically insignificant changes can add up or combine synergistically to significant changes as compared to the original non-transgenic lines. We should not fall in the trap of declaring the parental transgenic line as a clean slate starting line / reference point for future comparisons. Instead we have to take the unattended changes of these first modifications (whether known or unknown) into account, and test whether combined they cause a stronger or different effect.

Consequently the assessment of stacked events has to be different than the simple and theoretical addition of the intended parental transgenic traits. Whilst some accumulative effects might be anticipated, such as due to the new combination of two Bt genes, possible combinatorial effects should not be assumed but tested.

Thus, whether the combination of a Bt event and a herbicide tolerance event have an impact on mycorrhiza requires testing, assumptions in either direction cannot suffice as a replacement of actual scientific assessment. The combining of two events with previously no impact on earthworms might indeed lead to an impact on earthworms, whether this is due to combining two different toxic compounds (eg. Bt and herbicide), combining of two non-toxic compounds that synergetically act as toxins or the addition/doubling of a low level toxic compound to an effective dose level.

Obviously, combinatorial (e.g. cumulative or synergistic) effects are relevant for changes within the plant, as well as in terms of environmental and health impacts.



Some other points:

It is important to stay away from assumptions to the extent that we are able to recognise them and replace them with proper testing and assessments.

For example transgene and insert site stability require molecular testing. As stability is an issue in single gene events, it is even more so in multi-gene and stacked events.

Equally, gene silencing remains an issue to be tested and looked out for. No doubt there are a number of mechanisms that can give rise to gene silencing. One of the triggers for example can be promoter sequence homology. Many of the currently available single-gene events are making use of the same promoters (e.g. CaMV 35S).  Combining such promoters may result in the occurrence of gene silencing under various conditions. Whilst no gene silencing might have occurred in the parental or single gene event, the combining of transgenes and their promoters in a single plant could.

It was raised in Hans-Joerg Buhk’s comments that no such silencing due to promoters has been observed in current transgenic plants. I don’t think that this is a sufficient argument for assuming gene silencing does not occur, or for not assessing gene silencing in stacked events. In this context it has to be remembered that during the development of transgenic crops a large number of events are discarded, for example for failing to display the desired transgenic trait. There is no record what causes such failings, and how much of that is due to gene-silencing. Of course those few commercialized LMOs are unlikely to include events where gene-silencing occurred – but placed together with other transgenes and their promoter sequences within a rearranged/different genomic surrounding things might look different..

On another point, there was the claim that there is no fundamental difference between multi-gene stacked events and a single-construct-multi-gene event. I actually see at least two such differences. Firstly, stacked events will have the inserted genes distributed at different places within the genome, thus having the potential to give rise to different positional effects. Furthermore, the transgenes present in stacked events are results from different transformation occasions, thus transformation induced mutations will be distributed differently and with potentially different impacts and combinatorial effects as compared to a single construct single transformation event.
Secondly, there is a difference on the risk assessment level. In the envisaged multi-gene (single-construct) single-transformation event testing, analysis and risk assessment would  automatically include the combinatorial effects of: all the genes present, their products, the overlaps in their metabolic pathways, their unintended effects, the transformation induced effects. The RA of a plant with multiple genes introduced in one  step is by definition a RA of their synergetic effects. 
On the other hand, the assessments of the different single gene parental lines will not – when combined – contain any information or assessment of the combinatorial effects of the different events.
On a practical level, it can also be added that in the RA of a a single-construct multi-gene single-transformation LMO there will most definitely be an isogenic non-transgenic parental line available as direct comparator – this looks rather different in the case of stacked events. We will need to address the issue of comparators at a later stage.

Rather than going into details of other comments ..

There is a general comment I want to make. I fully disagree with the notion that only then should combinatorial effects be tested if there “expected” effects and “credible causal responsibility that underlies risk assessment and going along with this an openness to see the unexpected and/or unintended. Both assumption based and expectation based risk assessment does not serve our all responsibilities under the CBD and BSP well.


By the way, I am more than fine to use the CPB language and talk about LMOs rather than GMOs or transgenic plants. … In fact I think we have to use the CPB language. I would like to highlight though, that ‘transgenes’ are not defined as genes derived from other species but genes that are transported into the organism by means of modern biotechnology, transgenes are genes that are introduced via transformation events. A transgenic plant that has been transformed with a gene construct based on its own genes (i.e. gene of the same species) is still a transgenic plant containing transgenes. Wanting to call these “cis-genes” and thus attempting to exclude this category from the term “transgene” is most unhelpful and will only lead to misunderstandings, false assumptions and extra problems.

Richarda makes some very good points here that should be taken careful note of.

Extending and supporting Ricarda's intervention

“Risk Assessment section:

To reflect the current ‘definition/scope’, we need to alter the ‘rationale’ of point 1 to include the combination of transgenic traits via secondary transformation of a parental transgenic line, as the integrity of the first/parental transgene loci might have been altered due to the secondary transformation and transgene insertion.”

I agree.


”Concerning point 2, it could include further explanations of the rationale behind the points to consider, if that would be helpful to a later reader for following the scientific argument and reasoning. I regard all the rationale and points to consider listed as crucial for risk assessment of stacked events and would like to see an additional aspect covered in the third bullet point of points to consider, namely: comparison of the stacked event LMO has to also be with its as-near-as-possible-isolines. In fact, choice/availability of the right unmodified comparator line presents a problem for the risk assessment of stacked events, which needs further considerations in the development of guidelines.”

Ricarda’s point on the proper comparator is absolutely crucial. EFSA provides useful information on the appropriate comparators. "In the case of vegetatively propagated crops, comparative analyses should include the non-genetically modified isogenic variety used to generate the transgenic lines. In the case of crops that reproduce sexually, comparators would include appropriate information required in applications for GM plants and/or derived food and feed non-GM lines of comparable genetic background" (p. 22-23).

EFSA, (2006). Guidance Document of the Scientific Panel on Genetically Modified Organisms for the Risk Assessment of Genetically Modified Plants and Derived Food and Feed. EFSA J. 99, 1-100.

There is debate, however, on what constitutes isogenic or comparable. In the case of vegetatively reproducing plants, there is no reason that the regulator could not be provided with a conventional counterpart that was the absolute isogenic parental relative except for the rDNA. Increasingly often, however, developers are substituting siblings of the GM plant especially for assessments using sexually reproducing plants (which may also be derived from GM parents). Guidance on what constitutes a true isogenic comparison would be welcome.

This practice has been justified because of the many breeding steps between the first regenerated plant and the near commercial variety. While this may suit the developer who can then combine normal agronomic testing done for marketing purposes with safety testing, it is not necessary optimal for the assessment. Because many jurisdictions do not require each variety created by subsequent conventional breeding to be tested, allowing a determination of safety in one variety to extend to all subsequent varieties developed through breeding, there is no need for the developer to defer testing to this late pre-commercial stage. A request by the regulator for a comparison between a GM and a non-GM parent that are truly nearly isogenic is both scientifically valid and technically possible.


“Concerning point 3, some points could do with strengthening, in particular indirect (cumulative, synergistic, combinatorial)”

I think it is important that the language not be limited to synergistic. In medicine, this is a carefully defined term meaning that the effect of the combination is a certain factor greater that of either component on its own. We are not interested in arbitrary factors for the assessment, but on both qualitative and quantitative differences that might be relevant. To avoid confusion on whether a new trait meets the definition of synergy, the terms cumulative and combinatorial (or something to this effect) should be included.

“effects both in general as well as in the context of changes in management procedures. Furthermore, it would be important to address stacked events in the context of soil, including soil organisms, soil fertility, availability of nutrients, pathogens, symbiotic interactions etc.

And should there not be a clear reference to (human) health somewhere? I thought we had agreed on that at the Montreal meeting?”

I agree

”There seems to be a misconception of the existence of combinatorial and synergistic effects. In fact these combinatorial and synergistic effects do/can not only arise from the presence of the various transgenes, but also from the presence of the side effects or the original transformation in the parental line, including positional effects, transformation induced mutations and unintended impacts of regulatory and metabolic processes and pathways.”

This is of course well documented. For example, when Monsanto combined a modified DHDPS “high lysine corn” line by stacking with a post-transcriptional inhibition of zein translation, they found an unanticipated combinatorial effect (ie synergy). It was not at all clear that the effect was from just the two transgenes alone. The Huang et al study and that of others leads to the possibility that recombinants achieved their levels of lysine accumulation in part by compensatory changes in lysine catabolism (either decreased activity or failing to increase activity relative to conventional varieties), a factor that cannot be predicted in advance for novel hybrids.

Huang, S., Kruger, D. E., Frizzi, A., D’Ordine, R. L., Florida, C. A., Adams, W. R., Brown, W. E. and Luethy, M. H. (2005). High-lysine corn produced by the combination of enhanced lysine biosynthesis and reduced zein accumulation. Pl. Biotechnol. J. 3, 555–569.

Furthermore, the combinatorial effect may be at the level of the transcriptome (as in gene silencing, see below) or the phenotype (as in the example above with lysine corn). The phenotype affected may not be in the plant. As the Monsanto paper (Baum et al) using dsRNA as an insecticide demonstrated, it was probably the overproduction of the pre-siRNA in the plant that allowed it to saturate normal processing enzymes and survive to be differentially processed in the insect. Thus, a plant phenotype level evaluation would miss non-target effects if food transmitted dsRNA were the level of potential hazard. However, finding high levels of dsRNA of a particular sequence could alert the assessor to potential non-target effects.

Baum, J.A., Bogaert, T., Clinton, W., Heck, G.R., Feldmann, P., Ilagan, O., Johnson, S., Plaetinck, G., Munyikwa, T., Pleau, M., Vaughn, T. & Roberts, J. (2007). Control of coleopteran insect pests through RNA interference. Nat. Biotechnol. 25, 1322-1326.

”Consequently the assessment of stacked events has to be different than the simple and theoretical addition of the intended parental transgenic traits. Whilst some accumulative effects might be anticipated, such as due to the new combination of two Bt genes, possible combinatorial effects should not be assumed but tested.

Thus, whether the combination of a Bt event and a herbicide tolerance event have an impact on mycorrhiza requires testing, assumptions in either direction cannot suffice as a replacement of actual scientific assessment. The combining of two events with previously no impact on earthworms might indeed lead to an impact on earthworms, whether this is due to combining two different toxic compounds (eg. Bt and herbicide), combining of two non-toxic compounds that synergetically act as toxins or the addition/doubling of a low level toxic compound to an effective dose level.

Obviously, combinatorial (e.g. cumulative or synergistic) effects are relevant for changes within the plant, as well as in terms of environmental and health impacts.”

I agree.

”It was raised in Hans-Joerg Buhk’s comments that no such silencing due to promoters has been observed in current transgenic plants.”

It is a well-documented phenomenon that, for example, when two genes run by homologous promoters come to be together in the same plant cell, both genes may be silenced (De Schrijver, A. et al., 2007). This can result in inadvertent silencing of agronomic traits. “The cauliflower mosaic virus 35S (35S) promoter has been extensively used for the constitutive expression of transgenes in dicotyledonous plants. The repetitive use of the same promoter is known to induce transgene inactivation due to promoter homology” (p. 988 Bhullar, S. et al., 2003). In an elegant study, Al-Kaff et al. showed directly that infection of susceptible plants with the cauliflower mosaic virus (the source of the 35S promoter) can cause silencing of a herbicide tolerance transgene with a 35S promoter (Al-Kaff, N. S. et al., 2000).

Al-Kaff, N. S., Kreike, M. M., Covey, S. N., Pitcher, R., Page, A. M. & Dale, P. J. (2000). Plants rendered herbicide-susceptible by cauliflower mosaic virus-elicited suppression of a 35S promoter-regulated transgene. Nat. Biotechnol. 18, 995-999.

De Schrijver, A., Devos, Y., Van den Bulcke, M., Cadot, P., De Loose, M., Reheul, D. & Sneyers, M. (2007). Risk assessment of GM stacked events obtained from crosses between GM events. Trends Food Sci Technol 18, 101-109.

Whether this effect is actually monitored in the field I don’t know but I truly doubt it. In the field it is unlikely that all plants will simultaneously demonstrate the effect, but even if only a few percent do, it could be important for yield especially for subsistence farmers.

Reviewed in Heinemann, J. A. (2007). A typology of the effects of (trans)gene flow on the conservation and sustainable use of genetic resources.. Bsp35r1e. UN FAO. ftp://ftp.fao.org/ag/cgrfa/bsp/bsp35r1e.pdfE


”On another point, there was the claim that there is no fundamental difference between multi-gene stacked events and a single-construct-multi-gene event. I actually see at least two such differences. Firstly, stacked events will have the inserted genes distributed at different places within the genome, thus having the potential to give rise to different positional effects. Furthermore, the transgenes present in stacked events are results from different transformation occasions, thus transformation induced mutations will be distributed differently and with potentially different impacts and combinatorial effects as compared to a single construct single transformation event.

I agree

”Secondly, there is a difference on the risk assessment level. In the envisaged multi-gene (single-construct) single-transformation event testing, analysis and risk assessment would  automatically include the combinatorial effects of: all the genes present, their products, the overlaps in their metabolic pathways, their unintended effects, the transformation induced effects. The RA of a plant with multiple genes introduced in one  step is by definition a RA of their synergetic effects. 
On the other hand, the assessments of the different single gene parental lines will not – when combined – contain any information or assessment of the combinatorial effects of the different events.”

I agree

“I fully disagree with the notion that only then should combinatorial effects be tested if there “expected” effects and “credible causal responsibility that underlies risk assessment and going along with this an openness to see the unexpected and/or unintended. Both assumption based and expectation based risk assessment does not serve our all responsibilities under the CBD and BSP well.”

I absolutely agree with Ricarda's statement.

”I would like to highlight though, that ‘transgenes’ are not defined as genes derived from other species but genes that are transported into the organism by means of modern biotechnology”

In my opinion, we must stick to the Cartagena definition which is any in vitro modified nucleic acid…this includes RNA! and any attempt to create an inherited change with dsRNA. To redefine transgenes as only those sourced from other species is not in the spirit of the Protocol.

Reply to Jack Heinemann on the issue silencing due to homologous promoters (text with marks attached)

In my initial comment I have questioned the statement “Having the same or similar promoters may lead to gene silencing…” in the draft outline and asked for an example.

Jack Heinemann has responded as follows:
“It is a well-documented phenomenon that, for example, when two genes run by homologous promoters come to be together in the same plant cell, both genes may be silenced (De Schrijver, A. et al., 2007). This can result in inadvertent silencing of agronomic traits. “The cauliflower mosaic virus 35S (35S) promoter has been extensively used for the constitutive expression of transgenes in dicotyledonous plants. The repetitive use of the same promoter is known to induce transgene inactivation due to promoter homology” (p. 988 Bhullar, S. et al., 2003). In an elegant study, Al-Kaff et al. showed directly that infection of susceptible plants with the cauliflower mosaic virus (the source of the 35S promoter) can cause silencing of a herbicide tolerance transgene with a 35S promoter (Al-Kaff, N. S. et al., 2000).

Al-Kaff, N. S., Kreike, M. M., Covey, S. N., Pitcher, R., Page, A. M. & Dale, P. J. (2000). Plants rendered herbicide-susceptible by cauliflower mosaic virus-elicited suppression of a 35S promoter-regulated transgene. Nat. Biotechnol. 18, 995-999.

De Schrijver, A., Devos, Y., Van den Bulcke, M., Cadot, P., De Loose, M., Reheul, D. & Sneyers, M. (2007). Risk assessment of GM stacked events obtained from crosses between GM events. Trends Food Sci Technol 18, 101-109.“

I have checked the three literature references and came to the conclusion that my initial comment is still valid.

A. De Schrijver at al. (2007) is a Viewpoint Article not a Research Article with the view expressed:
In addition, one should take into consideration that in the case of GM stacked events, the combined presence of transgenes might influence expression. For example, gene silencing that involves transgene/transgene interactions might occur in case homologous DNA sequences, e.g. expression controlling elements, are brought together (Fagard & Vaucheret, 2000).

The article of M. Fagard and H. Vaucheret (2000) cited is a Review Article referring to a Research Article of Mette, M.E., von den Winden, J., Matzke, M.A., Matzke, A.J. (1999). Production of aberrant promoter transcripts contributes to methylation and silencing of unlinked homologous promoters in trans. EMBO J. 18: 241-248
Citations from that article:
To test the hypothesis that transcriptional silencing and methylation of target gene promoters could result from a trans-acting promoter RNA, we constructed a chimeric gene consisting of a nopaline synthase promoter (NOSpro) under the control of a cauliflower mosaic virus (CaMV) 35S promoter (35Spro) and used it to transform plants expressing an unmethylated NOSpro-nptII target gene. Transformed plants were analyzed for NOSpro RNA synthesis, and for activity and methylation of the NOSpronptII gene. Although production of a full-length, polyadenylated NOSpro RNA did not lead to inactivation or methylation of the target locus, the synthesis of nonpolyadenylated NOSpro transcripts from a 35Spro-NOSpro inverted repeat (IR) was associated with silencing and methylation of the NOSpro-nptII target gene. This suggests that aberrant promoter RNAs can mediate the methylation of unlinked homologous promoters in trans.
---
Here, we have tested whether a transcriptional trans-silencing locus can act by producing transcripts of promoter sequences. Full-size, polyadenylated NOSpro RNA, even when abundant, did not induce trans-silencing or methylation of the NOSpro-nptII target gene. Silencing and promoter methylation of NOSpro-driven target genes was observed, however, when non-polyadenylated NOSpro RNAs that deviated from the expected size were synthesized from an IR comprising 35Spro-NOSpro sequences. Decreasing NOSpro transcription by repressing the 35Spro partially alleviated silencing and reduced the methylation of the NOSpro-nptII target gene, indicating a role for aberrant NOSpro RNA in this trans-silencing phenomenon.

Conclusion:
Full-length transcript of the nos promoter (transcribed via the 35S promoter) did not lead to silencing of an other nos in trans. In one out of 9 transformants a rearrangement of the transforming DNA hat occurred during the process of transformation, resulting in inverted repeat structure of two elements both carrying the 35S promoter and the nos promoter, whereas the terminator sequenced was deleted. This transformant produced an aberrant nos promoter transcript capable of silencing a nos gene in trans. Silencing was mediated by an aberrant RNA and not by the homolog as such of the two copies of the nos promoter present.

S. Bhullar et al. (2003) mention in there introduction that homology-based silencing has been reported to occur extensively in transgenic plants and refer mainly to a second review (Vaucheret H. and Fagard M, (2001), transcriptional gene silencing in plant: targets, inducers and regulators, Trends Genet 17: 29-35).
This review refers to Mette M.F. et al. (2001), Transcriptional silencing and promoter methylation triggerd double stranded RNA. EMBO J. 19: 5194-5201.
With regard to promoter homology this article refers back to Mette et al. (1999) as mentioned above.

Al-Kaff et al. (2000) report a host defense response to CaMV infection that leads also to silencing of a recombinant gene under the control of the CaMV 35S promoter.
Citation:
Our experiments have shown that virus infection can lead to loss of herbicide tolerance in transgenic plants expressing the BAR gene. This effect is a consequence of a host defense response to CaMV infection causing inactivation of virus replication by gene silencing leading to silencing of a transgene homologous to the invading virus. From analysis of different transgenic lines, we have found no evidence that transgene position effects or copy number play a role in the virus-elicited suppression of BAR, NPT II, or GUS transgenes reported here or previously. Gene silencing in cruciferous hosts of CaMV can affect viral functions at the transcriptional and posttranscriptional levels, with a concomitant effect on transgenes sharing promoter or RNA homology with the virus.

Overall conclusion:
All these articles do not indicate that silencing may occur only on the basis of homology of promoters. However, I am aware that in the early days of gene silencing research sequence homology mediating direct DNA-DNA contact was discussed as a possible step that may lead to gene silencing

The discussion on RNAi-type silencing mechanisms is very interesting indeed. It is no surprise to me that some investigators will see silencing initiated when DNA sequences include little or no other known similarity other than in the promoter region and others will not. Indeed, we already published a paper in 2001 showing that most cells (in this case transgenic tobacco) transformed will die, but the majority of those that remain viable silence the transgene probably through RNAi. Among those transformants are ones also that don't. We and nobody to my knowledge knows with absolute certainty when you will or will not get silencing, even though we can say how it happens.

There is ample literature evidence to suggest that transgenes and genes run by common promoters (I am unfamiliar with one called nos, but certainly this applies to P35S) can be silenced. The surprise here is only that the regulatory sequence itself, which we don't normally associate with being transcribed, is sufficient for this phenomenon. Maybe it is and maybe it isn't. The amplification pathway for dsRNA (the active molecules in silencing) may lead to the generation of additional dsRNA molecules that eventually lead to methylation of the promoter. That mechanism is quite plausible from existing literature.

Therefore and finally, my conclusion is that molecular characterisation of novel stacking events is critical for case by case hazard identification leading to proper risk assessment.

I would like to offer support the interventions of Ricarda, and later Prof. Heinemann posted here. All very relevant points.

Particularly Ill mention one: I like the idea of "cumulative and/or combinatorial" rather than synergistic as a descriptor, for the reasons statd.

Clearly there is much plausible and prudent science to be done to support the RAs on stacked events.

I support several of the postings to this forum calling for a simplified guidance document on stacks and, in those situations where additional risk assessment is justified, an environmental risk assessment approach for stacked events that is consistent with the roadmap.  Case-by-case evaluation is required to determine whether the characteristics of the stacked event indicate a need for additional risk assessment and, if so, the relevant assessment endpoints.  After which, risk assessment may be performed according to the roadmap, and the details of this should not be incorporated in this guidance document.

It is important to provide clear definitions and direction to allow for a case-by-case evaluation of stacks.  For the purpose of this guidance, I support the definition of a “stack” as representing the progeny of two or more individual LMOs, each having undergone risk assessment, combined by conventional breeding.  Including “molecular” stacks within this definition, such as LMOs that have been retransformed or a single transformant with multiple traits, unnecessarily complicates this discussion within the AHTEG.  Regulatory practice to date indicates that these “molecular” stacks are already considered “new events”, and therefore subject to complete environmental risk assessment consistent with the principles in Annex III.  A separate guidance document on stacks is neither necessary nor applicable for “molecular” stacks.

The extensive risk assessments conducted on the parental LMOs provide the starting point to determine whether there is a valid hypothesis to support the potential for adverse environmental impact resulting from the combination of traits in the stack.  Any meaningful interaction between the traits may change the conclusions of the earlier risk assessment.  Depending on the nature of the traits combined and lack of meaningful interaction, risk assessments already conducted on the parental LMOs are applicable to the progeny stack.  For example, the lack of potential toxicity or allergenicity demonstrated for newly expressed proteins in parental LMOs is applicable to these proteins expressed in the progeny stack.  To disregard the completed risk assessments of the parental LMOs is to disregard previous science-based regulatory decisions.

Some commentators have pointed to alterations or rearrangements at a molecular-genetic level as a possible consequence of breeding individual LMOs to produce a stack.  These types of alterations may also occur as a result of crossing two conventional (non-LMO) plants or the crossing of an LMO with a conventional variety and are within the normal range of genetic variability that occurs during the breeding process.  These kinds of effects may often be more scientific curiosities with little practical relevance for environmental risk assessment, where phenotypic characteristics provide a much better foundation for predicting environmental safety than molecular analysis.  For stacks, the primary concern is whether there is a safety consequence arising from a meaningful interaction between co-expressed traits or a potentially altered phenotypic characteristic.  Even in cases where there has been some gain or loss of function as a result of conventional crossing, this phenotypic variability may be within the normal range and inconsequential with respect to potential adverse environmental impact.  The evaluation of meaningful interactions, those that would change the conclusions of the earlier environmental risk assessments, will always be case-by-case and should be based on sound scientific knowledge and principles, supported by relevant hypothesis-driven studies.

I very much agree with the last paragraph of Ms. Weber's comments.  We cannot be lured into viewing every biological phenomenon or molecular/biochemical detail is a risk.  Keeping in mind that our focus in these risk assessments is to "identify and evaluate the potential adverse effects of living modified organisms on the conservation and sustainable use of biological diversity in the likely potential receiving environment, taking also into account risks to human health" (Annex III), then whatever is going on with interactions at the genomic, biochemical or molecular level will manifest itself in the phenotype of the stacked LMO, at least for the adverse effects that we are supposed to be focused on.

I also wish to join those postings to this forum which supported a simplified guidance document on stacked transgenes. I believe extra risk assessment of plants with stacked transgenes/traits should only be restricted to exceptional situations where there is a "justifiable scientific concern". This should be in consistent with the road map and there should not be a separate guidance for this purpose. In particular I would like to support comments made by Ms. Weber (posting #1285) and Mr. Quemada (posting #1291).
Behzad Ghareyazie

I support the view of those suggesting no need at this point of a guidance document about stacked events.  In the Mexican experience on ERA they are considered as new events and we dealt with them according to the Anex III. As the Japanese experience, we also are more concern on the expression of the genes or phenotypic variability.
The element 1 of the draft give us just a characterization of the transgene, but the information that is used for the ERA are the elements 2 and 3, all included all ready on the Road map.
Best regards and thank you all for the draft and your comments.

Dear all,

Just some comments to a few points that keep coming up  (plus codex):

1) ‘Definition’ of stacked genes/traits vs. limiting ‘scope’
2) Definitions in general
3) Hazard identification as a basis for risk assessment
4) Codex and the requirement of molecular assessment of stacked genes/traits


(1) ‘Definition’ of stacked genes/traits vs. limiting ‘scope’:

Depending on whether we refer to general use of language of stacked genes/traits or limited use of language, we will end up with different descriptions of ‘stacked genes/traits” in the introduction. I felt it might be good to reflect the broad understanding in the introduction narrow it down to what we are intending to deal with under scope (currently disguised behind the ‘definition’ above ‘risk assessment’ (see my comments from 3 July - [#1241]).


(2) Definitions in general:

I agree with the many suggestions to stay within the language of the Cartagena Protocol where possible. I also agree that if we need to use terms not covered by the Protocol and that allow flexible interpretation, we will need to define them for the use within these additional guidelines. Maybe our SWG chair could offer some ‘Draft definitions’ in the next round or at such a stage when we know which terms we will actually need to use.


(3) Hazard identification as a basis for risk assessment:

I want to support Jack’s comments and in particular want to underline the importance of solid and comprehensive hazard assessment as the basis for a proper and reliable risk assessment. A hazard assessment – and I hope we all agree here - can’t be guess work or probability assumptions but actually requires knowing, i.e. clear and factual data and information. The only way to acquire such ‘knowing’ is molecular testing and assessments as a starting point.  ‘Unintended changes’ are often subdivided into predictable changes on one hand and unpredictable or unexpected on the other. Both require testing. In general, the basis for such unexpected changes may for example reside in transformation induced mutations, as mentioned previously (Wilson et al., 2004, 2006 and Latham et al. 2006) or in the gene sequences used or positional effects.

The communality of these effects is that they cannot be predicted. I refer for example to the publication by Zolla et al. (2008) “Proteomics as a Complementary Tool for Identifying Unintended Side Effects Occurring in Transgenic Maize Seeds As a Result of Genetic Modifications.“ >>attached. Looking at Mon810 in comparison to isolines and using proteomics as the investigative tool, Zolla et al. found that 43 proteins were up- or down-regulated in transgenic seeds with respect to their controls. They mention: “Interestingly, a newly expressed spot (SSP 6711) corresponding to 50 kDa gamma zein, a well-known allergenic protein,[16] has been detected. Moreover, as a major concern, a number of seed storage proteins (such as globulins and vicilin-like embryo storage proteins) exhibited truncated forms having molecular masses significantly lower than the native ones.” These findings of course bear relevance to the allergenicity and adjuvant effects found in transgenic bean engineered with a bean alpha-amylase inhibitor. (Prescott et al., 2005)

None of these effects could have been predicted, but only found by testing. To my knowledge there are no equivalent experiments investigating ‘stacked gene/trait’ LMOs/events. We can’t answer what the result would be of combining these unintended and upredictable effects of one parental LMO line with a different set of unintended and unpredictable effects of another parental LMO line. To answer any of this, the first step (after using breeding observations – see below) would be testing and gaining information and data from proteomics, metabolomics, etc.


(4) Codex and the requirement of molecular assessment of stacked genes/traits:

I read with interest Tom’s comments with reference to Codex. To my best knowledge, Codex does not use the terminology "weight of the evidence" approach in the main part of the plant guidelines referred to by Tom and commented by me below.

Tom [#1293] is of course correct that the appropriate paras for stacked genes/traits in the Codex plant document are Section 3, paras 14-17 (e.g. “Unintended Effects”).  However, I would argue that the Plant Guideline clearly implies that stacked GM traits should go through a thorough molecular characterization.

Concerning point 1 and the quote from para 14 – clearly para 14 includes stacked traits as GM plants that fall under the Codex guidelines:  “Unintended effects in recombinant-DNA plants may also arise through the insertion of DNA sequences and/or THEY MAY ARISE THROUGH THE SUBSEQUENT CONVENTIONAL BREEDING OF THE RECOMBINANT-DNA PLANT. Safety assessment should include data and information to reduce the possibility that a food derived from a rDNA plant would have an unexpected, adverse effect on human health.”  CAPs added. 

Point 3 that Tom makes (from para 16) is incomplete as the next sentence states, “Molecular biological and biochemical techniques can also be used to analyse potential changes at  the level of gene transcription and message translation that could lead to unintended effects.”  Thus, para 16 clearly points out the utility of molecular information. 

Concerning point 4, about the need to use a variety of techniques “that are typically observed by breeders in selecting new varieties for commercialization,” again this quote from para 17 is incomplete, and could therefore be misleading. Indeed that para continues saying “These observations by breeders provide a first screen for plants that exhibit unintended traits. New varieties that pass this screen are subjected to safety assessment as described in Sections 4 and 5.”   These last two sentences in para 17 explicitly say that the observations of plant breeding is a first screen and that any GM plants that make it through this screen should go through the full Codex safety assessment process, including a complete molecular characterization. 
If one follows on to Section 4, in particular paras 30-33, one can find the description of the kind of molecular data/characterisation that should go into a safety assessment.

In sum, it is evident that the Codex Plant Guideline does clearly apply to stacked genes and it lays out clearly that these should indeed go through a safety assessment, as para 14 clearly says that subsequent conventional breeding of a GM plant —which is exactly what is used to produce stacked genes/traits as we are dealing with in this guidance material— could produce unintended effects, while para 17 clearly says GM plants should be screened by breeders for unintended effects and any plant that survives such screening should subsequently go through a full safety assessment (including molecular assessement).


We might want to consider a full reference to the Codex Plant Guidelines and their sections relevant to stacked genes/traits.


Refs:

Prescott VE, Campbell PM, Moore A, Mattes J, Rothenberg ME, Foster PS, Higgins TJV and Hogan SP. Transgenic expression of bean a-amylase inhibitor in peas results in altered structure and immunogenicity. J Agricultural and Food Chemistry 2005, 53, 9023-30.

Zolla L, Rinalducci S, Antonioli P, Righetti PG. Proteomics as a complementary tool for identifying unintended side effects occurring in transgenic maize seeds as a result of genetic modification. J. Proteome Res 2008, 7, 1850-61.