Just like back in January, I find that I’ve been accumulating a short list of interesting news updates from RNA therapeutics that collectively seem worth sharing but no individual one of which has risen to the level of being its own blog post. Of course, the biggest news story from the past few months was the halting of tominersen trials in Huntington’s disease, but as I noted in that post, we don’t yet know anything about what happened in that trial so there isn’t much to say. In the meantime, here are five recent developments I found interesting.
nusinersen vs. onasemnogene
In January, Biogen announced that it had dosed the first patient in a new Phase 4 trial (NCT04488133) of nusinersen, the Ionis/Biogen ASO for spinal muscular atrophy (SMA), in patients who had already received onasemnogene abeparvovec, Novartis/AveXis gene therapy for SMA. The open-label trial will recruit children who still have “potential for additional clinical improvement” despite receiving the one-time gene therapy shot. In other words, if the gene therapy didn’t cure you, can the ASO still help?
The trial will last two years, and so there are no data yet, and there won’t be for a long time. But this news story caught my eye because the very existence of this trial is significant: it will evaluate the efficacy of layering one rare disease therapy on top of another. SMA now has three approved drugs: an ASO, and a gene therapy, and a small molecule (risdiplam). While SMA is not cured, it’s a shining example for any untreatable rare disease hoping to have therapeutic options become available. The results of this new trial could help write the future in SMA: when a rare disease has more than one approved drug, can patients get to benefit from all of them, or do they have to choose just one?
Consider the case of Gaucher disease, which is caused by partial or total loss of function of the gene GBA, encoding β-glucocerebrosidase. Gaucher was the indication for the first-ever enzyme replacement therapy (ERT), imiglucerase, approved way back in 1994. Fast forward to today, there are five approved drugs for Gaucher disease. Imiglucerase and two other nearly identical ERTs (velaglucerase alfa and taliglucerase alfa) all supply the working β-glucocerebrosidase enzyme, while two small-molecule substrate reduction therapies (SRTs; miglustat and eliglustat) inhibit glucosylceramide synthase (UGCG), the enzyme upstream of the mutated enzyme in Gaucher. One of those SRTs, miglustat, even has generic competition since 2018. Gaucher disease is far from cured — in fact, none of these drugs addresses the neurological manifestations of the more severe subtypes — but to those of us working on untreatable diseases, having five drugs with two mechanisms of action seems like a lot to aspire to! Here’s the catch: all of these therapies cost hundreds of thousands of dollars, and it’s either/or — patients don’t generally take both an ERT and SRT. The prior authorization forms that insurers ask doctors to fill out in order to prescribe SRTs sometimes list “Failure or clinically significant adverse effects to enzyme replacement therapy” as a prerequisite, and miglustat’s FDA label even specifies it is only for patients “for whom enzyme replacement therapy is not a therapeutic option”. Why? It all comes down to the available evidence from clinical trials. In this case, trials comparing imiglucerase to either miglustat (NCT00319046) or eliglustat (NCT00943111), as monotherapies, showed non-inferiority [Cox 2012, Cox 2015], but a trial of SRT/ERT combination therapy has not yet read out (NCT03485677). If there’s no clinical evidence to suggest that combination therapy is better than monotherapy, and the drugs cost a lot, then of course insurers don’t want to pay for both.
In prion disease, I often hear people say that one drug is unlikely to be a cure, or a completely effective chronic treatment, and we’re likely to need drug cocktails. It’s easy to say in the abstract, but the Gaucher story is a good reminder that getting there in reality is a struggle. The best therapy might be A, B, or A and B. Which regimen patients actually take will be determined by clinical evidence, and that means comparative trials may go on long after a drug is approved. While it’s a long road, it’s heartening to see that in SMA, at least, there is appetite to go down this road.
FDA “N of 1” guidance
Milasen, an ASO designed to restore proper splicing for a single patient’s MFSD8 mutation, has become a heralded example of an “N of 1” therapy [Kim & Hu 2019]. A growing number of activated patients and families are striving to design analogous therapies for their own genetic mutations. These are likely to be developed on highly accelerated timelines, with relatively limited opportunity for rigorous human data collection. This creates a host of potentially rather difficult challenges for regulators tasked with considering the risk and benefit to patients [Woodcock & Marks 2019].
Early this year, FDA released a draft guidance on N of 1 therapies. The guidance only addresses Investigational New Drug (IND) submissions — permission to begin human dosing. IND is likely the key or only stage of interest for sponsors and advocates. Presumably, most N of 1 therapies won’t have a need for marketing approval, nor any opportunity to generate the kind of clinical data that could support a marketing application. In the case of milasen, the maintenance dose was 42 mg every 3 months, and the sole manufacturing run yielded 18 grams of compound — more than a lifetime supply for this one patient. Families, and physicians, entering into an N of 1 study may expect to simply remain in a “treatment IND” status indefinitely.
What data are required to get to IND? The guidance, above all, encourages early interaction with FDA, making clear that there still is no one-size-fits-all formula for N of 1. Take toxicology for instance. Milasen underwent just 74 days of preclinical toxicology in a single species, rats. Guidance for standard drug development pathways would require two species, one of which must be non-rodent, to be dosed for 6-9 months or, at minimum, however long the trial is slated to last. Will a single-species, 2.5-month barebones tox package become standard in N of 1? The new guidance describes formats for submitting nonclinical data, but does not spell out the minimum type and amount of data required. Presumably, it all depends. Regulators have shown remarkable flexibility so far in accommodating unconventional development paths for “N of 1” therapies, but boiling that flexibility down to a formula without regards to each program’s particulars has got to be an awfully big challenge. Update: On April 26, 2021, FDA released a draft guidance on nonclinical studies for N of 1 therapies which does answer some of these questions in greater detail: for example, it allows for a single species, 3-month toxicology package, only 2 weeks of which need be completed prior to the first human dose.
The new guidance received comments from n-Lorem, NORD, and some of the other parties you’d expect. It will be interesting to see how the final guidance evolves. But ultimately, it’s early days. My guess is it will take several more examples of N of 1 therapies to help work out what the standards and expectations will be.
ligand-conjugated ASO success in hereditary angioedema
In March, Ionis announced a remarkable success in its Phase 2 trial of an ASO hereditary angioedema (HAE). HAE is caused by loss of function of the gene SERPING1, which encodes a protein called C1-INH or C1-inhibitor. Without it, the contact system goes into overdrive, causing painful and life-threatening swelling attacks. The ASO IONIS-PKK-LRx targets the gene KLKB1, encoding the protein prekallikrein (PKK), to rein in the contact system.
Ionis had previously published promising preclinical data from this program [Ferrone 2019] as well as a compassionate use pilot study in two patients [Cohn 2020]. But the new press release caught my eye for the sheer magnitude of therapeutic benefit: a 97% reduction in swelling attacks from week 5 onward, with 92% of drug-treated participants having zero attacks.
In contrast to many of the other indications Ionis works on, there are actually several existing treatment options for HAE, including infusions of recombinant or plasma-derived C1-INH, and both monoclonal antibody and small molecule inhibitors of PKK [Cohn 2020]. But while the ASO may not be first-in-class, it has a potential to be best-in-class. The antibody, lanadelumab, reduced attacks by 87% [Banerji 2018], and the small molecule, berotralstat, reduced them by only about half [Ohsawa 2020].
This success is a validation of Ionis’s ligand-conjugated antisense (LICA) technology. This ASO is conjugated to GalNAc3 to promote liver uptake. last year’s compassionate use study had compared this ligand-conjugated version of the drug against an unconjugated version, and found that the LICA achieved comparable efficacy at an order of magnitude lower dose (80 mg every 3 weeks, versus up to 400 mg weekly) [Cohn 2020]. This mirrors a recent study in transthyretin amyloidosis which found that the ligand-conjugated version of Ionis’s approved drug for TTR, inotersen, was 30- to 50-fold more potent than the original drug [Viney 2021]. As of today, liver seems to be the main target for LICA, but a recent review discusses research into other peripheral tissues as well [Roberts 2020]. It remains to be seen whether LICA will eventually prove relevant to brain diseases.
Wave kills two Huntington’s disease leads
Not long after Ionis’s HTT ASO, tominersen, got halted in Phase 3, competitor Wave dropped two of its Huntington’s disease leads as well. Wave has been developing stereochemically pure ASOs on the hypothesis that these might be able to achieve greater potency, tolerability, or allele specificity than the stereorandom ASOs that Ionis and most other competitors use. As explained by HDBuzz, Wave’s candidates in Huntington’s disease were all intended to be allele-specific, targeting only mutant huntingtin.
Things already weren’t looking good for this program. The first trial readout in 2019 saw only a 12.4% reduction of the target pharmacodynamic biomarker: mutant huntingtin in CSF. (For comparison, tominersen lowered mutant huntingtin by 40% in Phase I [Tabrizi 2019]). Wave responded by adding a higher (32 mg) dose arm in the hopes of seeing deeper target engagement. But according to Wave’s press release, even at that 32 mg dose, the two HD drug candidates reduced mutant huntingtin by only 9.9 to 11.6%, a result that was not significantly different from placebo. Wave hasn’t totally divested from Huntington’s, and says they plan to launch a trial of a a third candidate, WVE-003, in 2021.
Wave has published data on the performance of other stereopure ASOs in the liver [Iwamoto 2017] and more recently in the retina [Byrne 2021], but to my knowledge, they haven’t published any preclinical data for the Huntington’s disease program — all I was able to find was a conference abstract [Dale 2020]. So we don’t really have any priors on which to predict whether the doses explored in this Phase I trial should or should not have acheived better target engagement. The only CNS indication for which I’ve seen any published preclinical data from Wave is the recent paper on C9orf72 [Liu 2021]. In that paper, they target a splice junction in order to specifically knock down the transcript isoform that is more commonly seen in patients with the pathogenic hexanucleotide repeat expansion — so they achieve some allele specificity without actually targeting the mutation. Analogously, their Huntington’s disease leads weren’t actually targeted at the CAG repeat expansion that causes Huntington’s disease, but rather at benign SNPs within HTT, with the strategy of pre-screening patients to identify those for whom the repeat expansion was in cis with the targeted SNP. Overall, it remains to be seen whether stereopure ASOs can achieve properties in the brain that stereorandom ASOs cannot.
atlas of ASOs in the brain
In January, Ionis published a long-awaited paper with a large bolus of data on the distribution and activity of ASOs in the brain [Jafar-Nejad & Powers 2021]. If you follow ASOs at conferences, you’ve probably already been seeing data from this paper for a couple of years now.
The most common concern we hear about ASOs for the CNS, especially from neurologists, is whether they have sufficient distribution to all parts of the brain. Do they really reach deep brain structures or is drug activity limited to the cortex and spinal cord? Of course, the ultimate answer to this question will have to come from humans — efficacy in diseases with deep brain structure involvement, and autopsy data showing drug distribution and target engagement. But human data are slow to accumulate. In the meantime, the new paper does probably the best job you could ask for of using animal data to try to answer this question. As in previous studies of ASOs in the periphery [Hung 2013], Ionis uses the ubiquitously expressed non-coding RNA Malat1 as a model target. They examine the accumulation, potency, and distribution of a potent Malat1 ASO in mice, rats, and monkeys. In Figure 5, they show that across a battery of 35 different brain regions in cynomolgus macaques, the ASO lowers the target to anywhere from ~10% to 75% of its normal levels. The single least-reached region is putamen, with about 75% residual Malat1, followed by globus pallidus at about 60%. Even in caudate, thalamus, and deep cerebellar nuclei, the ASO lowers its target to about 50% of normal levels. In the best-reached regions, including virtually all parts of cortex, the target is lowered to about 10% of normal levels. So, while ASO distribution may not be equal, some level of target engagement is achieved throughout the brain in monkeys. Non-human primate data are a lot more convincing than just rodent data in this regard, but a caveat, of course, is that even an adult cynomolgus monkey brain still weighs in at only about 74 g [Pardo 2012], versus ~1,300 g for humans [Svennerholm 1997].
The paper also models (both in the monkeys and across mouse and rat models) the relationship between drug concentration and drug activity. Long story short, ASO concentration is a good proxy for target engagement — the reason why there is not much knockdown in the monkey putamen is because there is not much drug there. Finally, there are some limited cell type distribution data, showing that the ASO is active across multiple cell types in the brain: neurons, oligodendrocytes, astrocytes, and microglia.
All told, while we will continue to watch new human data roll in, this atlas of ASO activity provides a good foundation on which to believe that ASOs are a relevant modality for a whole-brain disease like prion disease.