In your opinion, what additional data could be collected to improve your confidence in the MOA for 1,4-DX and liver tumors?

I would like to agree with and emphasize the DNA damage aspects pointed out by Expert 4. Moreover, the somatic mutation landscapes and nature of adaptive cell origins, including the interactive signals and events driving clonal origins of cell populations (healthy plus metabolically challenged livers) should be significant.  

In considering the possible mode of action for 1,4 DX  several data gaps should be addressed.  Specifically linking the treatment of 1,4 DX to liver cell proliferation, oxidative stress , and cytotoxicity in a dose response manner.  A simple short term in vivo study has been used for other compounds to address the liver mode of action and could be used to fill some of the data gaps.  Suggestion: 

Treat mice or rats (use the chronic bioassay strains) with 1,4 DX  (control and at least 4 doses including the tumorigenic dose and one dose higher and two doses lower) for 7 and 28 days continuous. At 7 and 28 day sampling times measure liver histopath and also cell proliferation (either via Brdu pumps or Ki67), measure liver serum enzymes, measure oxidative damage (lipid peroxidation,8OHdg), measure Cyp2E1 activity, secure liver RNA for future pathway analysis. These results will show a dose response correlation between cell proliferation, possible necrosis, oxidative stress, and metabolic saturation. These data may help resolve the ox stress versus metabolic saturation debate as the initial key event.   If necrosis, serum enzymes and cell proliferation occur at doses that induce oxidative stress but not metabolic saturation the oxidative stress should be considered the key event.   

First of all, I support the suggestion made by expert 5 although the experiment sounds like a big (and expensive) effort. Nevertheless, measuring several key parameters in parallel is crucial to understand better both the dose-response and the MoA. Secondly, I suggest to explore further possible modes of action by targeted experiments including cell culture assays, e.g., in hepatocytes. It would be very helpful to understand the molecular events leading to cytotoxicity.

I also support the recommendation of Expert 5 concerning the nature of study that may provide valuable additional information on the sequence and dose-response/temporal concordance of key events. 

I believe refined toxicokinetic studies would not be cost or time consuming to conduct (other than repeated dose inhalation and drinking water administration) in that validated and appropriately sensitive DX and HEAA analytical methods are readily available.  However, as noted by expert 6, such studies would go a long way in supporting a conclusion that whatever MoAs might be driving tumor outcomes, they are demonstrably specific only to dose regimens sufficient to challenge sufficiency of DX first-order toxicokinetics.  If such can be more robustly confirmed, any further MoA refinements are unlikely to better inform human tumor relevance, i.e., whatever is going on at such metabolically saturating conditions, it very likely involves a multiplicity of highly complex toxicological interactions for which a further characterization of which events are primary or associative will be challenging if not impossible to reliably identify.  However, what can be confidently established, however, is that those potentially complex events are restricted to saturating dose conditions that have no quantitative relevance to realistic human exposure scenarios.  

Expert 5 design will answer several questions in a single study and will also provide BMD data for several endpoints (ROS lesions, enzyme induction, histopathology assessment, cell proliferation) that can be used to identify PODs. A 28 design will not only allow an assessment of genotoxicity in bone marrow, but also in a target tissues (e.g. liver); an assessment of DNA damage and role of ROS in contribution to that damage by using ROS-lesion specific enzymes can be done. Secondly, a 28 day study will also allow an assessment of mutations by error-corrected DNA sequencing in wild type mice at doses that are known to cause tumors; this will put to rest whether or not there is a genotoxic/mutagenic MOA using time points and doses that are OECD compliant and useful for regulatory submissions 
Although the ETAP was made for "low information" chemicals it was based on several very well known prototypic  compounds using 5, 28, 90 day studies that cause a host of biological events that captured several MOAs.  Secondly the cost of a 28 day study and more importantly the "turn around time" to get useful data from a lab that has run these types of studies is around 6 - 9 months. I do disagree on ETAP philosophy to use the first dose that causes any change in gene expression. In the 28 day study design the investigators can use apical and MOA based transcriptomic changes to calculate BMD- these usually align quite nicely. 
With respect to dose saturation as a key event , as I stated early in my mind this happens in all OECD based testing based on the language written - the top dose is based on the fact that the next highest dose would be lethal or cause obvious adverse effects. Secondly, the current use of key characteristic of carcinogens is not chemical specific. I do agree that the high doses used that saturate metabolism results in the events e.g., ROS from Cyp2e1 leading to indirect genotoxicity or cell proliferation.
If designed properly a 28 day study and also include levels that approach worst case scenario human exposure levels to nail down NOELs,  support nonlinear threshold, and to demonstrate that the dose levels required to elicit biological effects as stated by Rev. 3 have no quantitative relationship to levels of exposures experienced by humans.

The 28 day design can provide not only traditional MOA-based endpoints but also transcriptomic data that can support the MOA for the observed biological effects. As importantly, these endpoints can be combined to derive BMD, POD and calculate a human equivalent dose with margins of safety for quantitative risk assessments.

In my view a MOA based 28 day study using several doses known to be NOEL and doses known to cause tumors is the best option. Several doses means around 8 doses to clearly define the shape of the dose response curves particularly at low doses and to leave no doubt what the NOEL and the BMD10 are. 

One note, a preliminary assessment of transcriptomic MOA can be done using existing FFPE tissues from an archived 28-day study. This can be very reliable source of RNA for gene expression assessment and at a fraction of the cost of doing a de novo study.