Modeling Human Disease Phenotype in Model Organisms: “It’s only a model!”

Preface

A perspective by definition is a viewpoint. A viewpoint, like any other opinion, could be utterly erroneous. This Perspective is meant to be provocative but not to lessen the accomplishments of the scientific society as a whole or belittle any particular field of science or investigators. Scientific discoveries are typically incremental with various levels of increments. Often the significance of the discoveries remains unrecognized for many years if not decades, as was the case for the discovery of DNA by Friedrich Miescher in 1868 ¹. The significance of the discovery remained largely unrecognized for about 75 years, until simple and elegant experiments by Hershey and Chase showed that DNA and not protein, as was commonly perceived, was the genetic material ². Our shortcomings in recognizing the significance of the scientific discoveries should not deter us from Cartesian skepticism, which was pioneered by the Persian philosopher Ghazali and popularized by Rene Descartes’ “I doubt, therefore I think, therefore I am.”

An essential component of our academic society is the freedom to express viewpoints. Yet, personal opinions must not guide judgment on merits of the scientific discoveries and other peer-review matters. Science must be judged by the scientific standards of the time. It must not be judged by personal views. Scientific referees like all judges must be impartial and devoid of personal biases on rendering judgments. Accordingly, this viewpoint is simply that, a viewpoint. It is not indicative of author’s personal biases on any specific scientific discipline. The Perspective is aimed to raise doubts, as doubt is an incentive to truth.

Scene from “Monty Python and the Holy Grail”

Sir Lancelot: Look, my liege!

King Arthur: Camelot!”

Sir Galahad: Camelot!

Sir Lancelot: Camelot!

Patsy: It’s only a model!

King Arthur: Shh!

The scene might be germane to research laboratories that aim to construct the human disease phenotype in model organisms. There in the laboratory, the post-doctoral fellows, the principle investigator and the research assistant cast the characters of Knights, King Arthur and Patsy, the “naïve” and loyal assistant of King Arthur, respectively. Like the preconceived exuberance of the “Knights of the Round Table” upon sighting of a model of the “Castle of Camelot”, the over-enthusiastic interpretation of the laboratory data combined with unconscious biases could lead to contriving a model of a human disease that might only partially resemble the human phenotype but is far from recapitulating it. The primed excitement often mollifies the unpleasant reality that “it is only a model”, as Patsy would say and that no modeling is perfect.

The main reason for scientific research is to understand “how”, whether the “how” relates to the human diseases or the operation of the universe. Likewise, the primary reason to develop an animal model of a human disease is to gain insights into the pathogenesis of the phenotype with the ultimate goal of identifying therapeutic targets and diagnostic markers. The model organisms are instrumental in understanding of the “how” in biological processes, which ultimately relate to human health and diseases. Studies in C. elegans, one of the simplest multicellular eukaryotic model organisms, have led to numerous fundamental discoveries that have clear relevance to human diseases. Perhaps, the discovery of RNA interference by Andrew Fire and Craig Mello in C. elegans is a palpable utility of the model organisms – in terms of their relevance to human health and disease ³. The discovery, which was honored by the 2006 Nobel Prize in Physiology or Medicine, laid the foundation for findings of over 1,000 microRNAs in humans, which regulate various biological processes. No reasonable person would argue against the value of the model organisms in providing fundamental insights into biological processes relevant to phenotypes in humans (Table 1). However, to consider the phenotype in the C. elegans, as a true model of a human disease is simply overlooking more than 700 million years of evolution that separates humans from this most valuable multicellular eukaryote ⁴. Nobel laureates Sydney Brenner, who in 1965 proposed the C. elegans as a model metazoan for studying higher organisms and won the Nobel Prize for his pioneering work on developmental genetics of this nematode, shared an enlightening conversation that took place between him and a visiting scientist.

TABLE 1.

Selected scientific discoveries in model organisms that have been recognized by The Nobel Prize

Model Organism	Discovery	Nobel Laureates
C. elegans	“Genetic regulation of organ development and programmed cell death”	Sydney Brenner, H. Robert Horvitz, and John Sulston/(Physiology or Medicine, 2002)
C. elegans	“RNA interference - gene silencing by double-stranded RNA”	Andrew Fire and Craig Mello, (Physiology or Medicine, 2006)
Mouse	“Introducing specific gene modifications in mice by the use of embryonic stem cells”	Mario R. Capecchi, Sir Martin J. Evans, Oliver Smithies, (Medicine or Physiology, 2007)
Drosophila melanogaster	“Genetic control of early embryonic development”	Edward B. Lewis, Christiane Nüsslein-Volhard, Eric F. Wieschau (Medicine or Physiology, 1995)

The scene: Dr. Brenner’s office

Visiting scientist: By the way, where is your vivarium?

Dr. Brenner (walking to a window and pointing out to a building): There.

Visiting scientist: But, that is the hospital.

Dr. Brenner: That is right. That is where my vivarium is, the hospital, where the patients are.

Then, he continued: “We have generated so much data from mouse models that if we stop doing research today, we have enough material to analyze in the next decade!”. Dr. Brenner’s witty points should not be interpreted as refuting the indispensible value of model organisms in understanding vital biological processes but rather as an indication of inadequacies of the model organisms in truly representing the human phenotype. The shortcomings are equally relevant in our attempts to capture the human phenotype in cell models, whether isolated cardiac myocytes or induced pluripotent stem cells (iPS). It seems that we scientists have a trait for “irrational over-exuberance”, whether it relates to the gene therapy aura of 1990s, Dot.Com era of 2000s, or the current fascination with trying to model the human diseases in iPS. The iPS models are likely to advance our understanding of transcriptional regulation of cell fate, differentiation and various other important biological processes. However, it is surrealistic to consider iPS, generated through unnatural expression of selected transcription factors in culture environment, as a model of a complex phenotype, such as a human disease. The cells as many other culture models would display various genetic or genomics as well as endophenotypic changes that not only diminish but perhaps even abolish their utility as a model of human phenotype that they are hoped to recapitulate ^{5, 6}.

Evolutionary divergence of the humans and model organisms simply emphasizes the fact that no model organism, whether C. elegans, mouse or a higher primate, can truly reiterate the human phenotype. The modeling is particularly challenging for chronic human diseases wherein multiple stimuli – often modest in intensity – operate in conjunction with a very large number of other causative factors, which are typically not shared by the models organisms to induce the phenotype. The most commonly used model organism, namely the mouse, is separated from humans by about 75–100 million years of evolution and a newer fashionable model organism, the zebrafish by about 450 millions years ^{7, 8}. Even the model organisms closest to humans, namely African apes, which are separated from humans by about 5 million years, are not expected to fully recapitulate the human phenotype ⁹. The proponents have emphasized high level of genomic synteny (evolutionary conservation) between humans and models organisms to emphasize the robustness of these models for studying human diseases. However, synteny does not indicate genetic identity at the syntenic chromosomal regions. Chromosomal synteny may simply point to potential utility of the models in understanding basic biological principles but is not much relevant to modeling the human diseases. The genomes of the human and chimpanzee, our closest relative species, differ by about 35 million single nucleotide polymorphisms (SNPs), 5 million insertion/deletions (indels) and various chromosomal rearrangements ⁹. The phenotypic consequences of such differences in DNA sequence variants (DSVs) – although not adequately known – cannot be ignored. It merits reminding ourselves that intra-individuals differences in the DSVs is the essence of genetic studies of complex diseases in humans, whether the study is a Genome-wide association study (GWAS) or a whole-genome sequencing project ¹⁰. The complexity is beyond the mere presence of differences in DSVs, as the DSVs in the same genes could lead to different phenotypic consequences in different individuals. Phenotypic plasticity of DSVs is best illustrated for LMNA, which codes for Lamin A/C. LMNA mutations lead to at least a dozen of distinct phenotypes in humans, ranging from progeria to cardiomyopathy ¹¹. Furthermore, compounding the limitations of the model organisms in properly representing the human disease is the enormous diversity of the mankind, the magnitude of which has only recently been recognized ^12–14. Throughout the evolution new mutations have occurred and continue to occur at an estimate rate of about 10⁻⁸ per base pair, which equals to about 30 new mutations per each generation ¹³. During the last 10,000 years the rapid growth of human population has led to considerable expansion of the new alleles, which are, as would be expected, restricted to humans and not found in the model organisms. The majority of these new DSVs might not have significant phenotypic consequences but a considerable number of them would be expected to influence phenotypic expression of diseases. Moreover, considering that the evolutionary selective filtering takes time to apply, the new alleles are typically spared from such filtering and hence, are more likely to be pathogenic than the ancient alleles. Therefore, the human-restricted alleles would be expected to exert substantial effect sizes on the disease phenotype. This is in contrast to ancient alleles, which are shared between humans and model organisms, and have been subjected to evolutionary selective pressure in eliminating those with severe phenotypic effects. Consequently, genetic similarities between humans and model organisms are phenotypically less consequential than dissimilarities. Finally, genetic variability across strains of every model organism, as best documented by sequencing genomes of 15 mouse strains (http://www.niehs.nih.gov/research/resources/collab/crg/index.cfm), further compounds the efforts to model the human disease phenotype.

The emphasis on genetic differences between humans and model organisms is only one facet of the differences that restricts the utility of the model organisms in modeling the human phenotype. The disease phenotype is the consequence of stochastic and typically non-linear interactions among various constituents, which are comprised of complex intra-cellular and extra-cellular causal fields. The causal fields interact through complex networks that link all etiological constituents, such as DSVs, epigenetics, microRNAs, long non-coding RNAs, splice variants, post translational modifications of proteins, metabolites as well as the environmental factors including microbiome. Considering the presence of 23,500 protein-coding genes and over 4 million DSVs in each human genome; over 1000 microRNAs, several hundred metabolites, and a large number of yet-to-be characterized determinants, the complexity of the interactomes that ultimately determine the disease phenotype is limited only by one’s imagination ¹⁵. Simply based on the genetic differences among the model organisms and humans, let alone other causal fields, it is clear that human disease interactome would not be identical to interactomes that determine the phenotype in the model organisms. Hence, no model organism is expected to faithfully replicate the human disease phenotype. At best it might exhibit some resemblance to the components of the disease phenotype in humans (Table 2). Considering this inherent limitation, it will not be surprising but rather somewhat anticipated that the “translational” findings in the models organisms not to effectively translate to the human phenotype.

TABLE 2.

Partial resemblance and dissociation of the phenotype from the human disease phenotype in selected animal models of cardiomyopathies

Animal model	Expected disease model	Phenotypic similarity to human disease	Unsettled phenotypic similarity/dissimilarity	Phenotypic dissimilarity to human disease
β-myosin heavy chain –Q403 transgenic rabbit16–19	Hypertrophic cardiomyopathy	Hypertrophy, Myocyte disarray, Fibrosis, Preserved global systolic function but regional dysfunction	Reduced Ca+2 sensitivity of myofibrillar ATPase activity (Human phenotype unknown)	Higher incidence of systolic dysfunction in old animals, Results of pharmacological studies were not replicated in humans
Cardiac troponin T-Q92 transgenic mice20	Hypertrophic cardiomyopathy	Enhanced systolic function, Fibrosis	Enhanced Ca+2 sensitivity myofibrillar ATPase activity (Human phenotype unknown)	Smaller hearts, Smaller cardiac myocytes (No cardiac hypertrophy in transgenic mice)
Cardiac troponin T-W141 transgenic mice20	Dilated cardiomyopathy	Cardiac dilatation and dysfunction, Fibrosis	Reduced Ca+2 sensitivity myofibrillar ATPase activity (Human phenotype unknown)
Cardiac-restricted deletion of esmoplakin21	Arrhythmogenic right ventricular cardiomyopathy	Fibrosis, Excess adipocytes Cardiac dysfunction		Mild adiposis, Polymorphic as opposed to monomorphic ventricular tachycardia

The author apologizes for not including a large number of animal models of cardiomyopathies that others have generated and characterized. The purpose is to avoid potentially misinterpretation others’ data or criticizing them without detailed knowledge.

The judicious words of Sir William Osler, the father of modern medicine that no two humans have the same disease is best reflected by our inability to exploit the group data, whether clinical or genetic, to predict the phenotype in a single individual. Inter-individual differences in expression of a disease state is well appreciated and very much forms the essence of the population genetic studies. However, the scientific society has largely remained heedless to the limitations of the model organisms in correctly replicating the human disease phenotype. Perhaps, it is time to be reminded of Patsy’ comment to Kind Arthur and the Knights of the Round Table that “it is only a model.”