Current methods of drug development are costly, slow and inefficient and result in more failures than success. For rare diseases, drug development is even more challenging because of the small cadre of researchers working on each disease, heterogeneity of the clinical manifestations, limited availability of patient cohort and the limited marketability of the drugs even when they are successful. Moreover, although rare diseases by definition affect comparatively small numbers of patients, there are thousands of rare genetic diseases, and the advent of high-throughput DNA sequencing will expand knowledge of even more rare conditions. Therefore, innovative approaches to drug development are clearly needed to expedite the efficient clinical testing of potentially effective products for these rare conditions.

In contrast to the roughly 7,000 rare diseases, the number of underlying disease mechanisms and pathways is probably much smaller; indeed, some molecular mechanisms are shared in common across multiple diseases. Therefore, drugs that target molecular pathways common to multiple diseases can, in principle, be used to treat more than one disease, even if the clinical manifestations of those diseases are very different. The current approach to clinical trials, however, fails to address this feature, because one drug is often tested in patients suffering from one disease. A more efficient approach to clinical development in the era of molecularly targeted drugs would be to group patients by molecular etiology, rather than traditional disease classifications based on clinical manifestations, for the purpose of a clinical trial. Doing so would probably make clinical trials more efficient and facilitate testing of rational therapies in as many patients who might benefit from them as possible. Here we focus on two examples of pathologic mechanisms spanning multiple diseases to illustrate this concept and discuss some of the practical implications of this new approach.

Premature termination and protein folding in rare disease

Nonsense mutations resulting in premature termination codons (PTCs) are a common molecular mechanism underlying multiple genetic diseases. It has been estimated that roughly 11% of disease-causing mutations are nonsense mutations1. Following early studies in bacteria showing that aminoglycoside antibiotics can allow some read-through of termination codons, several efforts have focused on developing 'read-through' drugs for the treatment of disease resulting from PTCs. Here, we focus on PTC read-through compounds as a class of drug that acts on a mechanism common to multiple diseases.

The best known of these compounds is ataluren (PTC-124)2. This particular drug has been the subject of controversy, in part because of issues related to the assay used to develop it originally3,4, although it has shown efficacy in some clinical studies and particularly in cystic fibrosis5. It should be stressed that although ataluren is in clinical trials, it is not approved for the treatment of any disease. Importantly, however, other PTC read-through compounds that act in a similar manner, and which may turn out to be more effective than ataluren, are also in preclinical development6,7,8.

The actual amount of functional protein produced by a given read-through drug will depend on many factors, including the sequence of PTC and surrounding sequence9,10,11, the identity of the amino acid incorporated in place of the PTC, the ability of the drug to stabilize the mutant mRNA6 and the extent to which the mRNA is subject to nonsense-mediated decay (NMD)12. It should be noted that a small-molecule inhibitor of NMD has been developed and has been recently shown to enhance the activity of PTC read-through drugs13.

Another pathologic mechanism that is common to multiple diseases is abnormal protein folding14. In genetic diseases, abnormal protein folding typically results from mutations that alter the amino acid sequence of proteins. Abnormalities in protein folding can result in either recessive (loss-of-function) or dominant (gain-of-function) diseases. In recessive diseases, misfolded proteins can be retained in the endoplasmic reticulum (ER), in which case they do not reach sites in the cell where they are normally active, resulting in disease. A well-known example of this mechanism is in cystic fibrosis and is caused by the cystic fibrosis transmembrane conductance regulator gene (CFTR) delF508 mutation15. In addition, there are multiple examples of loss of function in lysosomal storage diseases (LSDs)16. In other cases, misfolded proteins can form toxic aggregates17.

Cells have a variety of mechanisms to maintain the correct folding of proteins, which are collectively referred to as the proteostasis network. In principle, manipulation of the proteostasis network is an attractive mechanism for the treatment of multiple diseases related to protein folding18. There are several published examples in which a single drug has been shown to substantially improve misfolded proteins in cells from more than one disease, demonstrating that a drug could be effective in multiple diseases19,20,21. As noted in this journal22, a small biotech company has been awarded a patent by the European Patent Office for heat-shock protein 70 to treat LSDs as a group, rather than to treat an individual disease, thereby demonstrating the plausibility of this approach.

Abnormal protein folding and PTC mutations are used here as two examples to illustrate the concept and value of grouping by molecular etiology. For the purpose of example, we assume two hypothetical drugs: a PTC read-through compound that has been shown to be effective in stimulating read-through at more than one PTC mutation, and a protein-folding drug that reduces the abnormal folding of more than one misfolded protein. We assume that both drugs have acceptable toxicology profiles and are ready for clinical trials. What is the most efficient way to test such compounds in patients, with different clinical diseases, who might benefit from them?

Grouping patients by molecular etiology

For simplicity, we consider three diseases: cystic fibrosis, Gaucher disease and Tay-Sachs disease. In all three disorders, a subset of patients have nonsense mutations and another subset have missense mutations that result in abnormal protein folding23,24,25. According to the current approach based on traditional clinical definition of disease, six trials would be needed to test the read-through drug and the folding drug (Fig. 1a). However, because the drugs under study are targeted to the molecular defects common to patients with different clinical diseases, the appropriate patient population should be all those with the relevant mutations. Therefore, if we group patients by molecular etiology, rather than by classic disease, only two trials would be necessary—one for the read-through drug and one for the protein-folding drug (Fig. 1b).

Figure 1: Benefits of grouping by molecular etiology.
figure 1

(a,b) Schematic representations of cystic fibrosis, Tay-Sachs and Gaucher disease are shown. In each disease, multiple patients have PTC mutations, protein-folding mutations or other mutations. According to the current approach (a), testing a PTC read-through drug and a protein-folding drug would involve six different trials. In contrast, if patients are grouped on the basis of molecular etiology (b), only two trials would be needed.

Full size image

Although simple in principle, the idea of grouping across disease by molecular etiology raises several issues related to designing clinical trials. First, a collaborative effort would be necessary, involving clinicians with expertise in each of the traditional diseases being studied in the trial. This expertise is essential because clinical trial end points would have to be defined for all the diseases studies, as would be the case in a standard clinical trial. Such end points would be based on the known clinical features and progression of the diseases, as well as appropriate biomarkers if available. For rare diseases, the lack of accurate knowledge about disease progression is often a limiting factor, emphasizing the need for adequate natural history studies.

In addition to standard inclusion criteria, for a drug trial based on molecular etiology it would also be necessary to demonstrate that the drug corrects the molecular abnormality in the cells of the patient. In the case of a misfolded protein, this would mean demonstrating that the drug substantially reduces the amount of misfolded protein in a cell culture model. For diseases associated with nonsense mutations, it would seem necessary to show that a patient has at least one such mutation and that the drug to be tested substantially increases read-through of the PTC to produce full-length protein in the patients' cells. Where available, genotype-phenotype data providing information on the amount of full-length protein necessary to restore function should also be considered. Testing drugs could be done in vitro, using patient-derived cells7. Optimally, induced pluripotent stem cells differentiated into the affected cell type might be used for such studies. An efficient strategy might be to distribute compound to different laboratories that are working with patient-derived cells for in vitro testing.

In terms of the practical reality of conducting a clinical trial, a reasonable setting might be an academic medical center, a rare-disease center of excellence or an existing consortium in which multiple investigators are studying different rare genetic diseases. Such organizational structures would facilitate interaction among the investigators involved, standardization of data collection and analysis and reporting of adverse events.

An illustrative example of a consortium approach to grouping by molecular etiology, albeit in the context of cancer, is the recent clinical trial of Xalkori (crizotinib) in pediatric cancer26 (http://www.clinicaltrials.gov/ct2/show/NCT00939770/). The study was carried out by a preexisting consortium, the Children's Oncology Group (http://www.childrensoncologygroup.org/). The trial included several anatomically and clinically distinct types of pediatric cancers (anaplastic large-cell lymphoma, inflammatory myofibroblastic tumor and non-small–cell lung cancer and neuroblastoma). The patients selected for the trial were grouped on the basis of activating mutations in the gene ALK (encoding anaplastic lymphoma kinase) as the underlying molecular etiology. Xalkori is an ALK inhibitor and had shown efficacy in individual trials of the different cancers with activating ALK mutations26. Notably, because different types of cancers were included in this trial, multiple clinical end points were defined and used, establishing a precedent for studying multiple rare diseases in a single trial.

Implications of grouping by molecular etiology for trials

The idea of grouping patients by underlying molecular etiology rather than the traditional concept of disease-based clinical phenotype may at first seem quite different from the current clinical practice. Even so, the molecular taxonomy of disease is already gaining acceptance (see http://www.nap.edu/catalog.php?record_id=13284), and, in our view, is likely to become more common in the future. For many genetic diseases, the current nomenclature of disease is based on historical considerations and is often confusing. In many cases, the same name may refer to a collection of genetically distinct conditions (for example, Usher syndromes or Charcot-Marie-Tooth disease). Diseases can also be named after biochemical abnormalities, but the same biochemical defect can result from very different underlying molecular defects (Fig. 1). In contrast, a molecular defect caused by a mutation in a specific gene, such as a PTC or misfolded protein, is unambiguous and objectively measureable, and therefore a better foundation on which to group patients for drug development and clinical trials.

The major advantage of this strategy is that it represents an efficient way to determine which patients may benefit from a drug. The amount of normally folded protein produced by a protein-folding drug will probably vary depending on other aspects of the proteostasis network, as well as the amount of other proteins being synthesized in the cell. Similarly, as discussed above, the actual amount of functional protein produced by a given read-through drug will depend on many factors. Indeed, a particular read-through compound might have no effect for some PTC mutations6,9. Even so, the process we propose here includes validation of the effectiveness of the read-through drug in the patient cells as an enrollment criterion. Therefore, if the PTC read-though drug is not effective for an individual patient's PTC mutation, perhaps because of an unfavorable surrounding sequence, then the patient would not be enrolled in the actual trial of that drug. The overall goal of our proposal is to outline the most efficient process to identify those patients for whom a PTC read-through drug or a protein-folding drug could have clinical benefit, and test the drug specifically in those patients.

A separate and more important question is the amount of full-length protein needed to produce a clinical benefit. A priori, it seems likely that this will depend on many factors, such as the protein's function(s), inherent stability and affinity for its cellular targets. However, the most important determinant of whether either type of drug is of clinical benefit may be the pathophysiology of the disease. Duchenne muscular dystrophy (DMD) is caused by mutations in the gene encoding dystrophin, a large structural protein that has an important role in muscle function. Available information indicates that ataluren treatment results in some clinical benefit for patients with DMD, although not to the extent of the originally proposed clinical end point27. Similarly, recent results using splice-switching oligonucleotides in DMD have also not met expectations28. In contrast, for other genetic diseases such as LSDs or DNA-repair disorders6,29,30, the mutations affect enzymes that prevent the accumulation of a toxic metabolic by-product. In these conditions, disease progression can take years as the toxic material accumulates. In such diseases, a small increase in active enzyme that substantially decreases the rate of toxin accumulation could in turn substantially slow the rate of disease progression. Indeed, a recent study demonstrated that PTC read-though drugs, in concert with an NMD inhibitor, were able to reduce the levels of toxic glycosaminoglycans to near wild-type levels in a mouse model of mucopolysaccharidosis resulting from a PTC mutation13.

Another issue that can influence drug efficacy is penetration into different tissues and cell types. Here again, testing drugs on the basis of shared molecular etiology in multiple rare diseases affecting different tissues and cell types could provide valuable information, which may benefit subsequent drug-development efforts. For these reasons, as well as the fact that the pathophysiology of many rare diseases is simply not well understood, it seems very difficult to predict, a priori, diseases in which a given PTC read-through drug or protein-folding drug will be clinically effective. Therefore, testing these types of drugs in as many patients as possible is important, not only to maximize the number of individuals that may experience clinical benefit but also to learn more about the relationship between pathophysiology and drug response.

One important long-term implication of grouping by molecular etiology on drug development should be mentioned. At present, drug development is focused on one disease at a time. However, as the concept of grouping by molecular etiology gains acceptance, this will in turn encourage drug-development efforts on mechanisms that are shared across diseases. A recent example of this approach is the discovery that d-tocopherol can reduce pathological phenotypes in patient fibroblasts from multiple LSDs31.

Other possibilities for grouping by molecular etiology

We chose drugs related to read-though of PTCs and protein misfolding to illustrate the concept of grouping by molecular etiology because of the numbers of diseases caused by such alterations. Even so, the basic principle could be extended to other types of mutations, such as those that affect RNA splicing. Two groups have identified small molecules that affect splicing of the gene IKBKAP (encoding inhibitor of k-light polypeptide gene enhancer in B cells) in familial dysautonomia32,33, and one of these compounds, kinetin, affects the splicing of a subset of other genes as well34. Although there are several different types of splicing mutations35, further research into the molecular targets of kinetin36 and other drugs that affect RNA splicing37 can identify protein targets that carry out specific steps of the splicing reaction. Such information could be used to identify and group patients for trials of drugs that target specific types of splicing mutations.

Several genetic diseases of the nervous system are caused by transcriptional deregulation, which can result from epigenetic gene silencing (for example, Friedreich's ataxia), or mutations in transcriptional regulators (for example, Rubinstein-Taybi syndrome). Grouping these as transcriptional-regulation disorders could facilitate the testing of drugs, such as histone deacetylase inhibitors, that are promising therapeutic candidates in several of these diseases38.

Grouping by molecular etiology could also be applied for drugs that act on pathogenic proteins. For example, several neurodegenerative diseases show prominent tau pathology (tau hyperphosphorylation and aggregation resulting in neurofibrillary tangles) and are collectively referred to as tauopathies. At least 85 phosphorylation sites are known in tau, and devising strategies to modulate the various kinases for tau phosphorylation is an active area of therapy development39. Other '-opathies', such as ubiquitinopathies40, are also potential candidates for drug development involving grouping by molecular etiology. Finally, several degradation pathways converge at the level of the lysosome, including endocytosis, phagocytosis and autophagy41. Lysosomal dysfunction is a common hallmark of many neurodegenerative conditions, such as Alzheimer's and Parkinson's diseases as well as Lafora and Pompe diseases; thus, therapies aimed at restoring lysosomal function may be beneficial for a host of disorders40.

Conclusions

Although there are thousands of Mendelian disorders, they can be classified into a smaller number of treatment groups on the basis of the molecular etiology resulting from the mutation. The good news is that novel therapies that are potentially applicable to many patients with different diseases are in development. Testing these drugs, in a timely and efficient manner, in patients who might benefit from them is a major practical hurdle. Grouping patients by molecular etiology for therapeutics testing has multiple advantages. For the biotech industry, larger groups of patients increase the financial incentive for drug development and may also spur competition for this patient population. And, for patients with rare diseases, grouping by molecular etiology should facilitate access to clinical trials of drugs that offer rational therapeutic potential, regardless of how rare a particular disease is or how little we understand about its pathophysiology.

The standard approach to therapeutics development in genetic diseases is based on targeting the biochemical defect resulting from the mutation. There is no doubt that this approach has been successful in some genetic diseases (for example, cystinosis42 and N-acetylglutamate synthetase deficiency43). Drug repurposing represents another attractive strategy for rare diseases, and it is being actively pursued44. It is important to stress that grouping by molecular etiology is not intended to replace either of these approaches. Instead, the approach we propose will probably be most useful for diseases in which biochemical pathway–targeted therapeutics are not available, either because the biochemical defect does not result in a 'druggable' target or simply because there is a lack of knowledge about the pathophysiology, which is a common problem in rare diseases. For individuals with such diseases, grouping by molecular etiology can provide an accelerated route to clinical trials of therapeutics targeted to specific types of mutations.

We realize that our proposal represents a different way of thinking about therapeutics development and testing for rare diseases and that it raises many unresolved, complex issues.As such, we view this proposal as an effort to begin a dialog about these issues with rare-disease clinical researchers, patient advocates and industry. In the long term, we look forward to the day when no patient is denied access to a rational therapy simply because their disease is too rare to treat.