Abstract
The rodent bioassay has been a mainstay of carcinogenic risk assessment for close to half a century; not that it has been without criticism in that time. As early as the 1990s, there were clearly established shortcomings with regard to relevance and extrapolation, but despite these concerns, this test continues relatively unchanged. As our understanding of the mechanisms of carcinogenesis and the availability of investigative techniques improve, there seems less reason than ever to persist with this exercise that provides little meaningful scientific data at great cost in terms of animal usage and pathologist time. This article highlights possible ways to gather and present data pertinent to carcinogenesis in man and suggests that it is the toxicological pathology community that should take the lead here in persuading regulators that there is great room for improvement in this particular aspect of regulatory pathology.
Introduction
In a recent article in Toxicologic Pathology, “A Data-Based Assessment of Alternative Strategies for Identification of Potential Human Cancer Hazards” (Boobis et al. 2009), the authors argued that biological effects seen in shorter term studies predict carcinogenicity and that such changes could be used to predict carcinogen activity. This could then be the basis of a new paradigm for hazard identification. Although in full agreement with the thrust of the recent article, I would argue that any new paradigm should encompass all available techniques. This is not a new type of proposal, and as early as 1993 Alastair Monro proposed a similar approach. With that in mind, I would like to reopen this particular debate and set the ball rolling for a discussion of an approach that includes a comprehensive package of focused short-term studies. The hope is that this will stimulate the toxicological pathology community to create the consensus that could perhaps form the basis of a new, and considerably overdue, paradigm.
Predicting Liver Carcinogenesis
Although I would question some of the choices made by the group in terms of organ systems, it is without question that the liver, particular the mouse liver, is the most common site for induced neoplasia in regulatory testing in lifetime studies. Prior to the article by Boobis et al. in 2009, several authors looked at early biological change (Carmichael et al. 1997; Elcombe et al. 2002). The broad conclusion from these groups has been that liver weight changes of more than 20%, hepatocyte hypertrophy, and p450 isoform induction per se all predict liver carcinogenesis. However, not all p450 enzyme induction is created equal. Chemicals that induce 1A1, 2B1/2, 4A1, and peroxisome proliferators all have a pretty much 100% strike rate in terms of inducing liver tumors. So in terms of a new paradigm for this particular aspect of carcinogenic potential identification in the rodent, I would suggest that a prediction of liver carcinogenesis should begin with a gene array to investigate up- and/or downregulation of p450 isoforms.
It should of course be borne in mind that isoforms such as 2B1/2 and 4A1 are expressed in rats but not in man, which would naturally influence any risk assessment. By comparison, 3A4 and 2D6 are expressed in man but not in the rat (Wauthier et al. 2007). Both of these isoforms are significant in terms of drug metabolism in man, and therefore induction or inhibition of these isoforms is certainly more of a risk to health in man, but given their absence in rat, inducers of these isoforms are likely have a much cleaner profile, unless alternative pathways of metabolism result in significant induction of another CYP family.
Predicting Lung Carcinogenesis
The lung as a target organ for carcinogenesis presents unique problems, not least because of the huge variation in strain sensitivity and route dependency. Mice models have been extensively used in the investigation of carcinogenesis induced by cigarette smoke and particulates (Hecht 2005). As a result of the large number of these types of studies, we have a shrewd idea as to how to investigate such exposures. In my view, the increased burden to human lung carcinogenesis that noninhalation exposure to NCEs represents is insignificant compared with the environmental influences. Thus, the suggested investigation for chemicals that are designed for the inhalation route, or likely to be inhaled in large amounts, is for an extended inhalation study in a sensitive strain (e.g., A/J or SWR mice).
Predicting Kidney Carcinogenesis
Of the major tissues, the kidney provides the most difficult challenge. A personal view of carcinogenesis in the kidney is that it appears to require a Goldilocks level of challenge to produce tumors. Potent nephrotoxins that damage the basement membrane appear not to act as kidney carcinogens, and it seems to be the permanently increased turnover caused by agents such as light hydrocarbons that seem to be just right. High reliability predictors of kidney carcinogenesis are hyaline droplets in the tubular epithelium (alpha 2 μ globulin) and indicators of high epithelial turnover (Elcombe et al. 2002; Hard et al. 1998). Agents that cause crystaluria generally induce neoplasia in the urinary bladder rather than the kidney (Cohen 1999), although hyperplasia of the urothelium in the kidney and ureters is often seen concomitantly. My suggested investigation for this type of hazard would be urinalysis and assessment of tubular epithelial turnover. In the event that crystaluria is present, investigation of the likelihood of this occurring at physiological pH in man could be assessed.
Immunosupression as a Cause of Carcinogenesis in Man
The issue of immunosupression is interesting. It has clearly been identified as a risk factor for carcinogenesis in man. Unsurprisingly, Boobis et al. (2009) found no evidence of induction of carcinogenesis in rodents secondary to immunosupression. Given the complexities of the immune status of a human population (i.e., a genetically diverse population that is not behind a barrier), I think we might safely assume that any chemical not obviously having an effect on tier I immune system parameters is not assessable in terms of human risk assessment. The best that can be done is probably to identify tier I immunotoxicants as having the potential to increase the suppressive burden on the immune system, which may have the effect of increasing carcinogenesis in some individuals. However, it should be questioned as to whether it is really valid to perform this risk assessment, vague as it is. Even with compounds that suppress the immune system quite powerfully, the effects of environmental chemicals and the influence of age and genetics mean that everyone will have a personal risk.
Endocrine Disruptors as Carcinogens
There is another well-established mechanism of carcinogenesis in man that requires consideration, namely, that which results from endocrine disruption. Although this is a broad and complex subject that probably cannot be investigated in its entirety as part of preclinical programs, there are areas that can be more readily addressed.
The most straightforward to deal with is the well-established liver-thyroid axis. There are several ways of approaching the tumor induction that results from depletion of T3 and T4. Direct measurement of T3, T4, and TSH is practical. Hypertrophy and hyperplasia of the follicular epithelium are also sensitive predictors of neoplasia. Looking at the liver end of this axis, one could examine induction of Cyp 2B1/2 on a gene array, as inducers of these isoforms also tend to induce the glucuronyl transferases responsible for the increased clearance of T4 (McClain et al. 1989).
The disruption of estrogen, progesterone, testosterone, and other endogenous steroids presents quite a different problem. Quite clearly there is a risk to the public at large from the presence of estrogen-like compounds in the environment. In recent years, the increase of prostate and testicular tumors alone speaks volumes for the reality of the risk associated with the disruption of these hormones. Equally clearly, there are assays such as uterotrophic and Hershberger assays and others recommended by EDSTAC in their tier I scheme that have the sensitivity required for identifying potential hazards (Yamada et al. 2000).
It is not only the direct effects of endocrine disruptors that warrant consideration but also those chemicals that affect endogenous steroid metabolism. In addition to being the isoform that metabolizes endogenous steroids, Cyp3A4 is also the isoform that metabolizes the largest number of small-molecule drugs currently on the market (Maurel 1996; Yamazaki et al. 1997; Waxman et al. 1988). Chemicals that can induce this isoform, or competitively bind to the active site and inhibit metabolism, may have a substantial effect on endogenous hormone levels.
Another area of concern is the additive nature of endocrine disruption. If considering the uterotrophic assay as an indicator of endocrine disruption, there is evidence that mixtures of chemicals have effects under conditions in which the individual components have no effect, although this is far from a simple relationship, and the addition of the effects of the individual components of the mixture does not provide a useful estimate of the activity of the whole. Also significant is that in terms of endocrine disruptors of vastly different potencies, the toxicity appears similar to that of the most potent component of the mixture (Tinwell and Ashby 2004).
With this in mind, I would suggest that any attempt to evaluate subtle changes in the reproductive tract as indicators of changes to sex hormones and subsequent inducers of reproductive tract tumors in man may be beyond the scope of animal studies. It is important to remember that the industrialization of the planet has resulted in our residing in a soup of endocrine disruptors, and by and large, small changes in our exposure to potential endocrine disruptors are not going to disturb the homeostasis that our physiology does its level best to maintain, at least not in a way that is readily quantifiable in animals.
Therefore, I would propose that we should concern ourselves with only obvious histopathological changes in the reproductive tract as being of concern in terms of human risk. Many chemicals have subtle effects on the reproductive tract, such as disruption or asynchronicity of the estrus cycle or atrophy of the secondary sex organs in male animals. In my experience, these types of changes do not predict neoplasia in carcinogenicity studies. Changes such as severe degeneration of the germinal epithelium or atrophy of the ovaries and uterus are those that typically result in the appearance of reproductive tract neoplasia in the rodent bioassay. That is not to say that small changes are irrelevant in the overall scheme of things, only that the increased risk that such changes present is unquantifiable and is part of a completely different question.
A Carcinogenicity Paradigm for the Next Decade
Putting all of this together, what could carcinogenicity assessment look like in the future? Well, as you may have guessed, there will be no 2-year bioassays any more. What there would be instead would be a carcinogenicity hazard prediction and risk assessment package. The package would consist of a set of studies and an expert report prepared by a suitably qualified pathologist. In addition to the generally recognized educational attributes that a pathologist requires, the pathologist would need to have a broad knowledge of toxicology as well as specialized knowledge of mechanisms and their impact on human risk assessment. This may well need to be acquired through specialized training.
The package would mostly consist of three parts: a standard set of genotoxicity studies, a receptor binding study, and a special in vivo study. For compounds delivered by the inhalation route, a 9-month inhalation study in a sensitive mouse strain would also be required. The standard in vivo study would be of 13 weeks' duration in the rat, comprising group sizes of twenty-five in four dose groups with three sacrifice points. After 7 days of dosing, the first subset of five of each sex would be sacrificed for genomics; essentially, this would be a CYP array, but others can probably suggest other parameters that could be added to the array. Full clinical pathology, urinalysis, organ weights, and a labeling index in the kidney would be the other recommended investigations for this time point. At the 28-day time point, a further ten + ten rats would be sampled along with a full histopathology list, organ weights, clinical pathology, urinalysis, and possibly follow-up of anything identified at the 7-day sacrifice. The final sacrifice at 13 weeks would have a reproductive tract focus and include routine histopathology on the primary and secondary sex organs along with measurement of estrogen, progesterone, and testosterone in the blood.
Armed with these data, the pathologist would write an expert report pulling all of the data together and giving an expert opinion regarding the likelihood of a set of data representing a carcinogenic hazard in the rodent and then putting that into a context of risk under the conditions of expected human exposure. This risk assessment should be couched in terms of pediatric and in utero exposure, the general population, and postmenopausal women.
As an example, let us consider a possible risk assessment for a potent CYP inducer such as phenobarbital. It is well known that in utero or perinatal exposure to phenobarbital can cause CYP isoform imprinting, even at therapeutic levels and in the absence of any clinical symptoms of overdose. This imprinting can program the liver to continuously overexpress constitutive CYPs. A lifetime of challenge by drug consumption and exposure to environmental insults can result in overexpression of inducible CYP isoforms, which may result in the creation of toxic metabolites, enhancing tumorigenesis and reducing life expectancy (Agrawal and Shapiro 2005; Gold et al. 1978). Thus, a risk assessment for this type of test article would identify the compound as a likely rodent liver carcinogen, with the pregnant mother and pediatric patient as possible at-risk populations, whereas the risk for the general population and postmenopausal women would be considered as minimal to nonexistent. As an aside, it is interesting to note that a published survey of the offspring of mothers administered phenobarbital during pregnancy reported that the treatment was associated with a 2.5 times higher number of deaths by early adulthood (Dessens et al. 1994).
Given the nature and subtlety of this type of interpretation, I think it is evident that it makes good sense to entrust this to an experienced pathologist rather than regulatory bodies, some of whom may have comparatively little experience. The role of the regulatory reviewers should be to cross-examine the expert witness, rather than formulate the case for the prosecution. In this respect, it is essential for the regulators to have confidence that the data presented are completely unbiased and without gloss. This may be difficult if these studies are conducted by the company developing the compound. To promote confidence, such work might be better undertaken either in government-sponsored laboratories or independent CROs with results presented simultaneously to regulatory agencies and the developer. In this development scenario, the regulatory submission would be replaced by the regulator and developer discussing the expert report of the pathologist. Although this proposal may appear calculated to cause a rush of consternation in boardrooms, I suspect that a realistic contextualization of risk will breed an atmosphere of trust in which the regulators will be more confident in their decisions, making it easier, and therefore faster, for them to make a rational decision. It is, after all, in the agency’s interest that good therapies are registered in a timely fashion, to benefit society as a whole.
As I sit in front of my microscope plowing through countless female rats with pituitary and mammary tumors, I cannot help but think that in these times of scarce pathology resources, this is a criminal waste of resources. Would the type of approach I am advocating be a better use of that resource? I would say unequivocally, yes. Would this guarantee to identify every potential human carcinogen? No, of course it would not, but then again, our current gold standard does not exactly inspire confidence either (Grisham 2006).
It is the community of toxicological pathologists that needs to take the lead here. I think the case we should be making as pathologists is that a well-designed package, independently interpreted by an expert in the field, will present a better assessment of true risk than a panel of nonexperts arguing the toss with a pharmaceutical company over the significance of an increase in a common tumor.
No doubt there are a couple of glaring omissions, and there are probably holes in my argument that you can drive a bus through, but then it is not really intended as a precise blueprint for where to go but rather as a starting point to stimulate what is really an overdue discussion on a better way of assessing the carcinogenic risk of NCEs.
Over to you.