You are here:
Letter to the editor of TAAP, in response to letter from Anders et al.
Citation:
EVANS, M. V. AND J. C. CALDWELL. Letter to the editor of TAAP, in response to letter from Anders et al. TOXICOLOGY AND APPLIED PHARMACOLOGY. Academic Press Incorporated, Orlando, FL, 248(1):65-67, (2010).
Impact/Purpose:
We would like to address the letter to the editor submitted by Anders et al. regarding the substantive issues raised regarding our paper "Evaluation of two different metabolic hypotheses for dichloromethane toxicity using physiologically based pharmacokinetic modeling of in vivo gas uptake data exposure in female B6C3F1 mice" (Toxicol. Appl. Pharmacol., 244, 280-290, 2010).
Description:
To the Editor, Toxicology and Applied Pharmacology: We would like to address the letter to the editor submitted by Anders et al. regarding the substantive issues raised regarding our paper "Evaluation of two different metabolic hypotheses for dichloromethane toxicity using physiologically based pharmacokinetic modeling of in vivo gas uptake data exposure infemale B6C3F1 mice" (Toxicol. Appl. Pharmacol., 244, 280-290, 2010). Indeed, we present a hypothesis generating analysis of the existing metabolism data that previously was assigned to the glutathione transferase (GST) pathway but fits equally well with a two-site metabolism of DCM hypothesis by CYP2EI. Anders et al. appear to have misinterpreted our results by stating that we dismiss the role of GSTs in DCM metabolism. In fact, as part of our analyses we used PBPK modeling of the same gas uptake data to fit both hypotheses: one that uses the well-established paradigm which includes GST metabolism of DCM at higher concentrations, and another that includes multiple sites for binding and metabolism of DCM within CYP2E1 that is able to metabolize DCM at lower and higher concentrations. Since the two-site model also fits the biphasic rate of DCM metabolism demonstrated by in vivo gas uptake data in rodents, we are suggesting that this alternate hypothesis should be considered. The Andersen et al. (1987) modeling of DCM was from the start based on statistical fitting of a multi-parameter metabolism model to overall metabolism data -not on having pathway specific metabolism measurements. Examination of statistical fits of other model forms fits well within the paradigm those authors employed. Our reexamination of historical data suggests that evidence (in vivo and in vitro) is not abundant that the GST pathway would be able to compete with CYP2EI at physiologic concentrations of DCM in vivo. Reitz et al. (1989) published in vitro values for metabolism of DCM by GST that were obtained at very high concentrations and were magnitudes higher than those obtained using the same metabolism parameters and in vivo data via PBPK modeling. The in vitro to in vivo extrapolation factor difference has not been explained satisfactorily. In addition, there are datasets that contradict the hypothesis that GST is the only mode ofaction (MOA) leading to DNA damage (Landi et al., 2003). As pointed out in our paper, the concentrations of DCM needed for GST activity that have been demonstrated in vitro are much larger than the highest physiologic concentrations that our PBPK models predict of the parent compound entering the liver. We agree that further experimentation is needed to examine the issues we have raised. We will address specific numbered comments from Anders et al. below: Comment I: Others have found that the prevailing hypothesis that describes DCM metabolism does not fit the experimental data. As we point out in our paper and in keeping with the current theory, Landi et al. (2003) were expecting GST status to be related to DNA damage in human cultures but found that it was not. Our examination of the Reitz et al. (1989) paper revealed that the difference between concentrations needed in vitro for GST activity that are in the millimolar range and those that would be predicted in PBPK modeling (i.e., in the micromolar range) were noted by the authors at the time (i.e. two orders of magnitude difference in Km). The assumption ofa GST pathway for DCM metabolism to only be responsible for carcinogenic effects ofthe compound has led to predictions of a lower potential risk to humans due to differences in affinity between rodents and humans for this pathway (Andersen et al., 1987; David et al., 2006). Rather than being precautionary, that assumption has been used to decrease extrapolated human risk from animal data. Our analyses points out the uncertainty in that assumption and we agree with Anders et al. that further study should be done before assumptions of decreased human risk from DCM exposure are accepted. Comment 2: The suggestion that we are criticizing the 2El/GST model is inaccurate. Our paper does not state that the GST pathway for DCM metabolism does not exist. We do not agree we have misinterpreted the findings of Watanabe and Guengerich (2006), and in fact adapted a figure from that paper showing that oxidative metabolism leading to DCM metabolism through GSH is not accurate and that the majority of missing CO2 assigned to that pathway is incorrect. We do not dispute that GST metabolism of the parent compound occurs but that the ability of this pathway to compete with CYP2E1 is doubtful. Certainly if enough DCM is applied in vitro to a limit that saturates the solution or volatilizes from it, GST can metabolize DCM. However, as we point out, these are not physiologic conditions. As to why no DNA adducts were found by Watanabe et al. (2007), we point out that such low concentrations of labeled DCM were used in vivo, that GST would have little ability to compete with CYP2E1 for substrate. The authors ofthe study pointoutthe limitation ofthe lowconcentration of radiolabel and we do not disagree that if higher concentrations were used, the study would have been more definitive. Comment 3: Anders et al. appear to assume that there is little support for non-hyperbolic kinetics which led to the two-site metabolism hypothesis. A review by Atkins (2005) states that evidence for non-hyperbolic kinetics has been available since the 1980's and has been largely ignored. Atkins also summarizes the work by Korzekwa et al.( 1998) which describes mathematically different cases of atypical kinetics (non Michaelis Menten), including the equation used in Evans and Caldwell for two-site binding. Atkins states: "From a historical perspective, it is interesting that non-hyperbolic CYP kinetics were documented as early as the 1980s (12-14), but this received little attention. Subsequently, Korzekwa et al. (15, 16) provided thoughtful accounts of the relationship between atypical kinetics observed with CYP3A4 and the possibility that multiple ligands could occupy the active site simultaneously. Today, it is widely accepted that several CYP isoforms, including 3A4, 1A2, 2E1, 2D6, and 2C9, can exhibit non-artefactual atypical kinetics in vitro. Furthermore, it is highly likely that the kinetic behavior is related in some cases to simultaneous binding of multiple ligands to a single active site, as elaborated here." Evans and Caldwell attempted to make use of published gas uptake data to examine the possibility of multiple-ligand binding kinetics for CYP2E1 as it metabolizes DCM in vivo. In regard to the other publications cited by Anders et al. regarding the state of this aspect of the science (i.e., Collom et al., 2008; Chowdhury et al., 2010) we do not see evidence in these papers to discount the two-site hypothesis for DCM metabolism by CYP2E1, and it was beyond the scope of our paper to present comments on these papers and to discuss further the review by Atkins. The purpose of the Evans and Caldwell (2010) paper was to test the hypothesis that atypical kinetics was possible for DCM metabolism and could be described by the data as an alternate hypothesis. Again, we agree that more research is necessary to answer the questions presented in our paper. Given that we do refer to Eadie-Hofstee plots, we would like to point out that the equations used to describe the portion of DCM metabolism that has been ascribed to GST does not conform to the traditional Michaelis-Menten equation. The 2El/GST PBPK model adds a linear term that describes GST metabolism that is in addition to saturable (Michaelis-Menten) or typical kinetics. Evans and Caldwell performed analysis to test three possible hypothesis for fit: (1) Michaelis-Menten alone (data not shown), (2) Michaelis-Menten with an added linear term (2El/GST model), and (3) the two-site metabolic rate equations. The correlation coefficients for all three cases were similar. We were not able to distinguish between any of the three models. The correlation coefficient for Michaelis-Menten alone was very similar to the Michaelis-Menten equation with an added linear term (2El/GST case). The question of our ability to distinguish between the models raised by Anders is also relevant to their ability to uniquely quantify the GST pathway which used the original PBPK modeling as its basis, since this equation is not very different from two other similar cases. Comment 4: We do not agree with the assertion that "There are already extensive results that are not consistent with a two-site oxidation model." As stated above by Atkins, multiple site binding is an accepted paradigm for many CYP450s, including CYP2EI. In the original PBPK modeling, the second or linear phase of metabolism was assumed to occur from GST metabolism of DCM. We note that an assumption of a two-site metabolism fits the deuterated as well as non-deuterated DCM. We also examined intravenous DCM data and enzyme inhibition data, as noted in the discussion. We feel that a strength of our paper is our examination of the same data using the same tools that have been used to support the previous hypothesis and determine whether it fits another hypothesis as well. It does. Consequently, differing types of experiments are needed to determine which hypothesis is correct. We are aware that Gargas et al. examined bromide production for brominated dihalomethanes. However, we suggest that what is really needed is accurate in vivo time-course measurements of CO that correspond to the total metabolism curves. The descriptions of those experiments are beyond the scope of our hypothesis driven paper. In regard to whether the two-site model is needed to accurately model metabolism for other compounds, even if the enzyme has two sites, this feature need not come into play quantitatively in the observed range of concentrations for all chemicals studied. Similarly, the GST pathway would be expected to playa greater role in metabolism at physiologic concentrations for substrates for which it has a much greater affinity. We agree that biological support needs to be established in conjunction with examinations of this topic through PBPK modeling and thus, our interest in papers such as Landi et al (2003). Comment 5: Part of the assumptions regarding the carcinogenic MOA of DCM being related to GST metabolism is that mutagenic activity from such metabolism results in cancer induction. As demonstrated by Landi et al. (2003), DNA damage in human tissues is not dependent on the GST pathway of DCM metabolism. The lack of carcinogenic activity in any bioassay and the difference between an inhalation and oral route of exposure should not be examined only by the simple sums of total metabolism. Many factors limit the extrapolation between these two routes. The simple assumption that only the nature of the metabolic products and no other event occurring at a high DCM concentration is responsible for positive tumor responses, and the question of whether the GST pathway is capable of DCM metabolism in vivo and therefore responsible for the creation of such metabolic products at such concentrations need to be tested. Indeed, we point out that the suggestion by Anders et al., that RISK2L represents GST metabolism of DCM and that at physiologic concentrations this enzyme is capable of metabolizing DCM, is an assumption. Our PBPK modeling predictions of the DCM concentrations going into the liver from extrapolated inhalation data do not indicate that this concentration is high enough for GST to be responsible for this activity as indicated by in vitro DCM metabolism data. As for whether the existing hypothesis involving GST metabolism comports with the DCM database as a whole, it fails to predict the mammary tumors and brain tumors associated with DCM exposure in rats. Certainly more study on the MOA(s) of action of DCM is warranted. Comment 6: We disagree that the information regarding DCM metabolism only supports the two-pathway model. Given that much support for the two-pathway hypothesis was based on the interpretations and assumptions regarding PBPK modeling predictions, we used that same data to evaluate the two-site model. To provide a "suite of straightforward studies" to study the two-site model but "simultaneously weaken evidence for the current DCM bioactivation" hypothesis is beyond the scope of this paper. We note that much of the older data regarding in vitro metabolism has been conducted at high concentrations. It would be advantageous to examine the abilities of these two enzyme systems to compete with one another at more physiologic concentrations. We acknowledge that the finding of two sites in CYP2EI is only emerging and further study of DCM and brominated analogues would be of great usefulness. We do not agree that CO2 has been a reliable indicator of the GST pathway using radiolabeled experiments. Comment 7: We are not rejecting or criticizing the DCM PBPK models themselves that have been developed using assumptions of 2EI/GST metabolism of DCM. We have employed these models for our own analyses and have used similar assumptions of two pathway metabolism for comparative purposes. There was no "ease" or "casualness" in our examination of the evidence for DCM and the analyses we did. We are pointing out that the two-pathway model is based on the assumption that CYP2E1 has only one active site with very tight affmity. However, CYP2EI is versatile in that it can metabolize both very small and very large molecules. We note current literature that describes hinges within the molecule that can open alternate pathways that accommodate the larger molecules. This is the same concept that we are applying to DCM as an alternate hypothesis. We advocate an examination of the data and acknowledgement of the uncertainties regarding the specificity of support of DCM data for the existing hypothesis. We encourage further examination and development of data for DCM MOA implications and assumptions regarding the dilution of human risk from exposure to this chemical. Rather than supporting a policy-driven approach, we are advocating a sciencebased approach. Such an approach includes acknowledging the uncertainties and discrepancies between the data and the existing two-pathway hypothesis, the non-specificity of the experimental data to distinguish between two-site and two-pathway PBPK modeling of DCM metabolism in vivo, and the need for further examination of the two-site hypothesis as a MOA of DCM-induced effects. Comment 8: We disagree with Anders et al. that our work is not of high quality. The present work meticulously uses the same models, approaches, and data as prior work by the Anders et al. to which we are replying, but incorporates newer information about CYP2El and reliance on the GST pathway for DNA damage to examine whether the original hypothesis based on such analyses is correct. New data and approaches and the recognition of inconsistencies between existing data and an existing hypothesis necessitates revisiting or modifying that hypothesis as part of the scientific process. This science-based process and objectiveness is part of the risk assessment paradigm and includes constant feedback and improvement of models which incorporate new biological concepts. We also agree with Anders et al. that, as time goes on, new lenses are created within scientific thought that allow for revisions on previous hypotheses and concepts. This is precisely the goal of our paper. The Anders et al. letter appears to confuse separate issues in regard to nonlinearity of responses and what we have discussed and provided analyses of in our paper. We presented information that showed that the two-phase metabolism curves for a two-pathway or two-site PBPK model fit the gas uptake inhalation data for DCM metabolism in mice equally well and we discussed the lack of the ability of the GST pathway to compete with CYP2El at physiological concentrations used in the studies. We were clear that in the two-site model one pathway does not shut off while another turns on and that both metabolic activities contribute to the metabolism curve in the dose-response continuum. In regard to whether threshold of response exists and whether mathematical modeling of experimental data can be used to distinguish linear and nonlinear (i.e. threshold) dose-response curves at low exposure levels, we refer to the work by Crump et al. This work was discussed most recently at a presentation by Crump at the Annual Society of Toxicology meeting in 2010. We were very careful in the discussion of our paper to distinguish this concept from the ones we discussed. Briefly, Crump et al. (1976) noted that if exposure to an external agent causes a response that acts with additivity to an ongoing process, then the response will be linear at low dose rather than starting at a zero response rate. More recently, Crump and Hoel (2010) examined the question of whether experimental data can prove/disprove low-dose linearity. For instance, Crump and Hoel (2010) showed through a variety of analyses that (1) dose-response data can be consistent with both a threshold and low-dose linear response; (2) even if a threshold is assumed to exist, there is generally no objective way to bound its value of the threshold; (3) as a consequence, upper bounds on low dose risk were still linear even when calculated under a model form that assumes that some (undefined) threshold exists. In addition, Crump et al. (2009)'s reported a lack of rigorous scientific justification for using the (sub-linear) log-normal distribution for lowdose extrapolation at the population level. Furthermore, Crump et al. (2010) suggest that biologically-based dose-response modeling, while having many important scientific uses, is unlikely to contribute substantially to the problem of low-dose extrapolation. Again we refer to the work of Crump et al (2009; 2010) and Crump and Hoel (2010) for further discussion of this topic which was not the topic of our paper. In summary, we thank the editors and authors of this letter for their thoughtful comments. We are convinced that further research is needed to address the questions posed by our publication. Disclaimer: This response has been reviewed by the U.S. Environmental Protection Agency and approved for publication. The views expressed in this manuscript are those of the authors and do not necessarily reflect the views or policies ofthe U.S. Environmental Protection Agency.