The Anthropomorphic Mouse

Transgenic mouse models clearly are here to stay: they have become the wave of the future, largely supplanting other forms of animal research.1 Yet we are beginning to see more stress on the limitations of research in mouse models of various diseases.2–6 For example, it has been noted that acute inflammation of diverse etiologies induces similar genomic responses among different human subjects, whereas data from mouse models correlate poorly with the human experience and with one another.2 In fact, the match between murine orthologs and results in human subjects is reportedly close to random.2 Similar concerns have been voiced with regard to autoinflammatory diseases,3 the natural history of pulmonary hypertension,4 some settings of bone loss,5 and cardiac arrhythmia models.6 As a result, we are coming to realize that the mouse may not be the ideal answer we hoped it might be in our search for molecular targets to effect better cures. Scientists work with genetically modified mice for a variety of reasons: on one hand they are highly adaptable to study in the laboratory and, as expensive as they are, cost far less than large animal models might. On the other hand, there is in society an unease regarding the ethics of animal research in general and large animal research in particular. The intent to avoid large animal research has a variety of roots, in part genuinely humane. But, in part, it may stem from our tendency as human beings to anthropomorphize: as evidence, check out any toy store or cartoon and look at the animals displayed or represented. Mice occupy a unique niche here. Walt Disney anthropomorphized them by dressing them in cute little outfits and naming them Mickey and Minnie. We've since improved on that methodology by knocking them in or knocking them out to give their progeny diseases which previously were ours alone, making them into tiny replicas of our worst afflictions or our vivid imaginations. Mary Shelley would've loved it—we've all become the scientist of her dreams.7 And because of the Promethean aspects of our methods, we've learned a lot. But a question arises: does what we’ve learned relate to human disease as it is or to human disease as we imagine it might be. A major application of knowledge derived from transgenic models is the identification of molecular targets for therapy. So much so that transgenic models have entered the mainstream in the field of pharmacology and in our own Journal. Of the 978 manuscripts listed in Medline as having been published in Journal of Cardiovascular Pharmacology (JCVP) since 2008, 5% have used genetic murine models, 46 of them knockouts and 5 overexpression. Forty-four have been targeted at the vascular system (68% atherosclerosis and the rest divided between vascular biology and hypertension) and 7 have been cardiac. My own field of cardiac arrhythmias has not been represented in genetic mouse models in JCVP, although there is a good deal of reference to murine transgenic models in the arrhythmia literature. Congenital long QT syndrome, atrial fibrillation, and sinoatrial node (SAN) dysfunction are 3 examples. For the purpose of limiting this discussion, I’ll focus on SAN dysfunction, using an article published in one of our sister journals as an example.8 I use it because it typifies both points I’d like to make: first it shows us the elegance in physiologic expression that can be achieved when working with a specific mouse model; second, it exemplifies the frustration that is engendered in trying to relate a specific genetic model to a human disease we are trying to understand. The crux of the publication is that by deleting the HCN1 gene (a member of the family responsible for the inward pacemaker current, If) in mice, a variant on SAN dysfunction is created.8 However, the HCN1 knockout had not originally been developed with heart disease in mind. Rather, the intent of those who made the mouse was to explore the potential role of cerebellar Purkinje fiber HCN1 in motor learning and memory deficits.9,10 No cardiac recordings were made in the original study, but it was noted that life expectancy in the HCN1 knockout mice was normal.9 The authors of the recent report8 understood that if HCN1 is functional in SAN, HCN1−/− mice might provide a useful model for understanding the cardiac impact of this gene, specifically in relation to SAN dysfunction. They have corroborated the normal life expectancy previously noted9 and shown as well that deletion of HCN1 is associated with sinus bradycardia, sinus pauses, and abnormal sinoatrial conduction.8 The problem we encounter in trying to relate the findings in HCN1−/− mice to clinical SAN dysfunction is not in the experiments reported but in (1) the fact that SAN dysfunction is a highly inclusive syndrome rather than a specific clinical disease entity and (2) trying to associate HCN1 deletion with a specific human cardiac disease or diseases. With regard to the clinical entity, SAN dysfunction is a cornucopia of ailments. Review of its history reveals the following descriptors: sick sinus syndrome, sluggish sinus node syndrome, lazy sinus node, inadequate sinus mechanism and sinoatrial syncope.11 Older texts don’t even highlight SAN dysfunction as a syndrome, rather they describe specific arrhythmias associated with SAN disease. Among these are excessive sinus bradycardia, sinus pauses or arrest, SAN exit block, chronic atrial tachyarrhythmias, alternating atrial bradyarrhythmia and tachyarrhythmia, and inappropriate responses of heart rate during exercise or stress.12 The arrhythmias described in HCN1−/− mice fall within this spectrum.8 Given the diverse electrocardiographic expression of SAN dysfunction, it is helpful to organize our thoughts in terms of the syndrome’s mechanistic subsets. These include SAN fibrosis, coronary artery disease, age (associated or not with fibrosis or coronary artery disease), and familial or nonfamilial disease. And this leads to the second concern: there have been no reports of an HCN1 association with human SAN disease. Described to date are a nonfamilial G6A polymorphism,13 families with ankyrinB gene variants,14 and 200-plus mutations in the cardiac sodium channel gene, SCN5A, more than 20 of which are associated with SAN dysfunction.15 An “overlap syndrome of cardiac sodium channelopathy” has been proposed, as SAN dysfunction shares clinical expression not only with Brugada’s syndrome but also with long QT syndrome type 3 and conduction system disease.16,17 Other variants on SAN dysfunction and SCN5A mutations are also in the literature, with symptoms ranging from minor to severe.18–24 Among HCN channels, only HCN4 has been clearly associated with SAN dysfunction in human subjects. HCN4 mutations in individuals or families have presented as benignly as asymptomatic sinus bradycardia and as devastatingly as recurrent syncope, bradycardia–tachycardia, and with accompanying ventricular arrhythmias.25–28 In contrast, HCN1, HCN2, Cav1.3, and Cav3.1 are associated with SAN dysfunction equivalents in animal models.8 They cannot move beyond the realm of potential contributors to human disease until data are available establishing a firm link between the models and specific human disease(s). The diversity of diseases that can be created in the mouse is a good application of Murphy’s law to genetically modified animals. But if we see no representation in patients of a disease we can create in mice, what does this mean? Maybe it never happens. Maybe we haven’t found it yet. Maybe it happens so rarely that it is not an important determinant of disease in general populations of patients. That doesn’t make it less interesting as science. It simply provides a public health context. Because no association has been found between familial or nonfamilial SAN dysfunction and HCN1 in humans, we might next ask what if any specific role HCN1 plays in human SAN physiology. Given that HCN1 deletion induces bradycardia in the mouse, and given its functional presence in mouse8 and in rabbit29 SAN, we might expect a role in other mammals. Although data on function of HCN1 variants are available in large animals, these demonstrate a contribution to pacemaker activity in overexpression models30,31; there are no data on knockouts. As for human hearts, HCN1 message has been identified in sinus node,32 but we have no information about protein levels or function. Hence, we await definitive information regarding expression and function, remaining “lost in translation” from mouse to man. In the desire to translate knowledge from animal to human, what do we see in a patient and what do we want to know about his/her disease entity that we might better understand by creating the disease in an animal model? Considering the mouse at face value, its normal sinus rate, 500–600 bpm, is 10 times that of many human subjects. The murine action potential derives from an ion channel spectrum identical to that in humans in channel identity only: the representation of channels in different regions of the heart and their biophysical properties, including kinetics, differ importantly. In light of this, how applicable are murine data on sinus bradycardia to the human? How generalizable are the mechanisms determining sinus bradycardia in mice to those in humans? In a broad mechanistic sense (i.e., ion channel populations and autonomic input), man and mouse occupy the same universe. But it is a far distance between solar systems. We can shorten the distance by referring to the closest thing we have in man to an HCN channel knockout and resultant bradycardia. This is pharmacologically determined: the tool is the If blocker, ivabradine, “an ‘open-channel' blocker of hHCN4, and a ‘closed-channel' blocker of mHCN1 channels.”33 When administered to human subjects, ivabradine decreases sinus rate by about 10%–15%.34,35 This modest rate decrease is reported as clinically therapeutic34,35 and is the same percent reduction found in HCN1−/− mice.8 However, the sinus bradycardia and sinus pauses are interpreted as pathological in HCN1−/− mice,8 whereas arrhythmias have not been a significant consequence of ivabradine administration in clinical trials.34,35 The outcomes in mouse and man likely reflect the influence of the multiple physiological contributors to SA node pacing, loss of any one (or perhaps more) of which can be sufficiently compensated by the others. But the outcomes also suggest that syndromes in which HCN4 and/or HCN1 are depleted in the absence of additional, accompanying pathologies might not be of arrhythmogenic consequence in afflicted human subjects. In other words, the knockout itself likely does not tell the entire story … whether there is an environment of supporting pathology or pathologies must be discerned. A useful guide to our thinking here was provided by Sturm and Mohler, albeit in another context: “In vivo functional studies performed in genetically engineered mice displayed a much more significant phenotype than what was observed with in vitro assays … [but] … how do we apply genetic variant information clinically, especially when the variant is novel?”6 In conclusion, the genetically modified mouse provides us with a burgeoning realm of opportunity and information, but we are learning from a variety of sources2–6 that it has been accompanied by excessive simplification. Seoka et al2 voice concern about direct translation from mouse models to human disease and suggest that, “because virtually every drug and drug candidate functions at the molecular level … [we should require] … molecular detail in the animal model studies indicating whether the model mimics or fails to mimic the molecular behavior of key genes, key pathways, or the genome-wide level thought to be important for the relevant human disease.”2 Although this may increase the expense of research in terms of time, effort, and dollars, the resultant increment in intelligent application of knowledge should more than outweigh the cost. ACKNOWLEDGMENTS The author thanks Drs. Penelope Boyden and Vadim Fedorov for helping to organize the author’s thoughts on SAN dysfunction, and to Drs. Ira Cohen and Tove Rosen for their critical reading of the text.