Drugbaron Blog

January 8, 2021 no comments

Why Sarepta’s most recent failure in DMD was entirely predictable

Yesterday, Sarpeta (NASDAQ: $SRPT) announced that its gene therapy for Duchenne Muscular Dystrophy failed to improve muscle function in a study of 40 boys, despite achieving impressive expression of micro-dystrophin (at least in the short term).  However, you spin it – and this company is world-class at spinning a story – that represents a serious blow.

Sarepta is no stranger to controversy, following the decision of the US regulator to approve the company’s earlier product, eteplirsen, which induces exon-skipping to improve expression of dystrophin protein in individuals who carry the mutant gene, as Exondys 51 back in 2016.  That decision was controversial precisely because the company offered no robust evidence of clinically-relevant efficacy, and even the data showing increased dystrophin expression was unimpressive.  But backed by the vociferous, if largely unscientific, support of patients and their families, the FDA took the unusual step of granting approval for an agent that, on the balance of probability, was likely to be no more effective than placebo, on the proviso that the company provided post-marketing evidence to support the efficacy claims.

Four years later, and more than a billion dollars of total revenues from Exondys 51 later, no such data has ever been presented – and now we learn that a “next-generation” strategy for increasing dystrophin, using a gene-therapy approach, increased dystrophin levels to a much greater extent than does Exondys 51, but still yields no significant clinical benefit.

The study is relatively small, and cannot exclude the possibility of a small benefit.  But a consistent picture is clearly emerging: boosting dystrophin levels, even substantially, does not offer the curative potential that Sarepta and others have been promising.

On the face of it, that looks like a surprise.  Here is a disease that is caused by the lack of functional dystrophin in the muscles of these boys.  So surely one would expect raising levels to improve function…  Sarepta scored massive investments on that simple hypothesis.  Only it was always likely to fail.

To understand why, we need to look at the biology of skeletal muscle degeneration in DMD.  But before doing so, its worth reading a few paragraphs from a Commentary on gene therapy that DrugBaron wrote in 2001, marking the first publication of the completed human genome sequence, just to remind us that the central flaw in the Sarepta story was well-known two decades ago.  Anyone still backing this particular horse thoroughly deserved to lose half their money when the US stock market opened Friday morning.

The underlying assumption of most gene therapy procedures, irrespective of the disease being targeted, is that restoration of a wild-type DNA sequence will alone be sufficient to normalise the phenotype of the individual.  This assumption has been supported by the concept that the DNA encodes every component of the cell as well as when it should be produced and in what quantity.  Surely, then, a correct set of DNA plans will ensure a correctly functioning cell?

Unfortunately, this assumption is patently invalid: a liver cell and a neuron from the same individual share an identical DNA sequence but perform very different functions.  The phenotype of a cell depends not only on its particular DNA sequence but also the current concentrations and locations of all the different protein (and possibly even non-protein) components.  A liver cell may differ from a neuron in part because it contains a different array of transcription factor proteins.  Large as it is, the human genome databank is likely only to contain a tiny fraction of all the information needed to make a human cell.  In principle, we could use this information to make the DNA itself as well as all the protein component to put in a ‘synthetic cell’, but we would not know how much of each component to add nor where in our cell to put them.  A cell is such a complex system, that the overwhelming majority of combinations of the starting materials would fail to function as a stable cell.  Our ‘synthetic cell’ would simply degenerate despite containing a wild-type DNA sequence and all the correct components.

Once we consider the whole array of post-genomic information that is present in even the simplest cell, it becomes immediately clear why cloning is possible but de novo creation of a living cell is not.  Should some of the proteins in a cell be phosphorylated, glycosylated or proteolytically processed?  We would need to consider acetylation of histones, farnesylation of G-proteins, subcellular localisation of transcription factors, polymerisation of actin and tubulin and so on.  No amount of interrogation of the human genome sequence will answer these questions.  Cloning Dolly the Sheep involved transfer of DNA into an oocyte: a specialised cell set-up to convert the information in the DNA into a whole organism.  To have created Dolly from scratch would have required much more information.  The DNA tells you how to make each protein component of the cell, and how to change the levels of the protein: it does not tell you the absolute amount of the protein that should be present under any given circumstances.  Thus, it is not unreasonable to assume that the entire human genome contains only a minute proportion of the information density necessary to create a human cell from isolated molecules.

What are the implications of these observations for the future of gene therapy?

It suggests that restoring the wild-type DNA sequence after the disease phenotype has manifested itself will, in most cases, not reverse the phenotype.  Worse still, it may not even prevent further progression of the disease.  Consider a typical monogenic disorder which might be considered for gene therapy in the future such as Duchenne Muscular Dystrophy.  Mutations in the dystrophin gene mildly affect the function of skeletal muscle so that shortly after birth the muscle is able to function almost as well as wild-type muscle.  Over a period of years, however, the response to this defect is for accumulation of extracellular matrix around the muscle fibres.  After some time, the loss of muscle function is more a result of the presence of this fibrosis than the original mutation in the dystrophin gene.  This leads to a vicious cycle in which further loss of muscle function leads to more matrix deposition and eventually to the complete failure of the muscle.  To intervene with gene therapy once the fibrotic cycle has begun will be largely ineffective: the disease has progressed to the point where the mutation in dystrophin (although the initial trigger) is no longer the driving force of the disease

Duchenne Muscular Dystrophy may be the rule rather than the exception: cystic fibrosis, hypertrophic cardiomyopathy and Marfan syndrome may all be similar.  Once the protein distribution pattern of a cell or organ has been disrupted beyond a certain point even the wild-type genome may not have the wherewithal to correct it.  This hypothesis is graphically illustrated by the process of ageing.  Much of the degeneration in tissue architecture and function may not result from accumulated genetic mutation, but from accumulated damage to the protein distribution.  Wrinkled skin results from excess cross-linking of the dermal collagen through formation of advanced glycation end-products (AGEs).  Cataracts result from disorganisation of the crystallin proteins in the lens.  Reduced kidney function results from damage to the glomerular basement membrane.  All this degeneration could plausibly occur in the presence of an entirely normal genomic DNA sequence

Is this an argument for halting research into gene therapy?  Certainly not.  Instead it may provide some guidelines for devising successful gene therapy strategies.  Some diseases may be amenable to gene therapy even when the phenotype has become established: for example, repair of the gene encoding cerebrosidase may ameliorate the symptoms in Gauchier’s disease and restoration of a functional adenosine deaminase gene may indeed prevent the symptoms of ADA deficiency.  The important thing is to ascertain whether the absence of the trigger protein remains the driving force behind the pathology at the time when gene therapy is contemplated.  The second issue is therefore one of timing.  The earlier the gene therapy intervention can be made, the more likely it is to be successful.  Unfortunately, this is at odds with the demands of safety monitoring for a new clinical technique.  As with most new therapies, the benefit to risk ratio is greatest for those most severely affected with the disease being targeted.  While treating these individuals may successfully demonstrate the safety of the technique, it may dramatically underestimate the efficacy.  We must guard against taking an overly negative overview of the efficacy of gene therapy if the first few clinical studies produce surprisingly little clinical benefit.
[Quoted from Grainger DJ (2001) Quarterly Journal of Medicine 94:337-9]

To summarise, DMD is a particularly bad choice for gene therapy because we already know that low expression of the gene product causes only a very mild impairment of muscle function (observed in very young children).  The progressive nature of the disease is caused by the phenotypic response to that initial weakness.

Even at a molecular level, the biological mechanism of this response is relatively-well understood.  Strain in the muscle tissue drives expression of TGFb1 (the canonical member of the TGF-beta superfamily), a cytokine that drives a co-ordinated expression of intracellular and extracellular muscle proteins, including the major contractile proteins, actin and myosin, and the key extracellular components collagen and fibronectin.  These increase the contractile capacity of the muscle, and in the healthy individual yield the familiar increase in muscle bulk (hypertrophy) that we observe with regular exercise (and exactly the reverse that occurs after long periods of inactivity, such as bed rest during illness).

But if the muscle has a genetic defect that reduces tractive effort for a given amount of contractile apparatus (something we also see in Hypertrophic Cardiomyopathies, where a genetic mutation in cardiac myosin very mildly reduces contractile function, exactly as with the dystrophin mutation in DMD), then higher levels of TGF-beta are needed to achieve a given level of function.  And over time something bad happens.

The level of extracellular matrix is much higher than it would be in a healthy individual, and begins to disrupt the perfect alignment of the muscle cells.  Beyond a certain point, this misalignment actually decreases the contractile force rather than increasing it (because a smaller and smaller component of the contractile effort of each cell is contributed in the axis of shortening of the whole muscle).  Now a positive feedback loop is set in train, where declining function triggers yet more TGFb1, more extracellular matrix, and a further decline in function.

That which began as a mild defect due to the mutation in dystrophin has now been magnified through a biological pathway that is completely independent of dystrophin.

Remember that this is biology uncovered, in part in DrugBaron’s laboratory at Cambridge University, in the 1990s.  And it flashes a very strong warning that increasing dystrophin expression is likely to have little or no effect on the progressive muscle weakness in DMD.  In short it predicts exactly what we have seen with both Exondys 51 and now with a micro-dystrophin gene therapy.

Intriguingly, Sarepta suggest that in a post hoc analysis the younger patients (4-5 years old) may have shown a significant (if relatively minor) improvement with the micro-dystrophin gene therapy, while the older boys (6-7 years old) actually did numerically worse than placebo.  The biology suggests that the earlier you intervene the greater the possibility of benefit, so this may hold out some hope that if done even earlier micro-dystrophin gene therapy could still provide limited benefit, but as with the prior Exondys 51 data, it is just as likely that splitting the data by age yields only a mirage of efficacy that will not survive replication.

All this should be chastening to Sarepta management, their investors and even Janet Woodcock who decided to approve Exondys 51 in the absence of evidence of clinical benefit.  What we are observing in these failed trials is not bad luck, but something entirely predictable.  And by pursuing these flawed hypotheses, we are doing patients a disservice – no matter how strongly they might argue in support of the research.  Resources are drawn away from approaches that may actually stand a chance to work (such as focusing on anti-fibrotic approaches), and precious healthcare dollars are being spent on snake-oil rather than drugs that can make a difference.

But the last point has to be the same as in the original Commentary DrugBaron wrote in 2001: none of this is an argument against gene therapy.  In contrast to 2001, in 2021 we know that gene therapy can work, for example in haemophilia, where replacement of the mutant clotting factor is sufficient to completely reverse the phenotype.  Instead, it is a re-iteration of the warning that success with gene therapy needs careful selection of the target, careful interrogation of the biology to be sure that the initial trigger for the phenotype remains the driving force for disease progression at the time when intervention is contemplated, and above all a plea to listen to what the data is telling us.  

4.4 20 votes
Article Rating

Yearly Archive

Would love your thoughts, please comment.x