Utility And Limitations Of Pre-Clinical Research: The Example Of Genetic Epilepsies

Introduction

When I explore promising new approaches to treating disease as part of my due diligence in the biotech space, I’m always keen on emerging academic findings. For instance, my interest in Marinus Pharmaceuticals’ (MRNS) Ganaxolone was bolstered by papers pointing towards the molecule’s usefulness in indications from CDKL5 over neonatal seizures to synergistic activity in status epilepticus when combined with diazepam.

Most recently, a fellow investor made me aware of this paper in Nature Communications, entitled Brain activity patterns in high-throughput electrophysiology screen predict both drug efficacies and side effects. Beneath the ambitious-sounding title is an ingenious exploration of anti-epileptic compounds by means of in-vivo testing in zebrafish larvae combined with advanced bioinformatics and verification through detailed behavioral observations. The authors mostly stem from MIT, which helps explain the extraordinarily thorough methodology underlying their work.

[Figure omitted: a zebrafish. Caption: “Not your typical lab rat. Image source: technology.org”]

After going over the paper a first time some days ago, I realized that it lends itself to a very helpful discussion on the limitations of academic research and preclinical work in the context of drug development. In a nutshell, while the methodology advanced here is extremely sophisticated and the outputs are highly intriguing, extrapolating these findings to predict successful drug development in human trials is not as straightforward as it may seem.

“Indeed, many widely used neurotherapeutics owe their efficacy to action at multiple molecular targets, while numerous promising compounds with potent activity in cell- or target-based assays have limited clinical utility due to off-target side effects.”

The authors try to tackle this prominent issue by devising of a more sophisticated preclinical methodology to measure the effect of neurotherapeutics than simple behavioral screening.

Augmenting behavioral analysis with brain activity measurements

“However, similar to in vitro assays, behavioral readouts in animal models often reduce complex neurological disorders to simple metrics that do not completely reflect underlying deficits and off-target effects. Direct high-content readouts of neural activity and brain activity patterns (BAPs) represent an attractive alternative to behavior-based screens as they may more accurately capture disease pathology, drug activity, and side effects.”

Two key terms that run like a red threat through the paper are brain activity patterns (‘BAPs’) and local field potentials (‘LFPs’). LFPs are “electrophysiological signals generated by the summed electric current flowing from multiple nearby neurons within a small volume of nervous tissue.” To assess whether neurotherapeutics have a positive effect on brain activity in the context of genetic epilepsies, the researchers make extensive use of LFP measurements into which they inject additional granularity by throwing bioinformatics into the mix:

“To detect effects of drugs on brain activity patterns and brain disorders, we developed algorithms that decompose LFP signals into independent subcomponents.”

Subsequently, the obtained data streams are assembled with the goal of observing changes in brain activity of the zebrafish larvae during experiments.

[Figure omitted: brain-activity maps for four language tasks. Caption: “Brain activity patterns during (clockwise from top left) reading, listening, speaking and thinking about words. Source: alz.org”]

Finding zebrafish analogues to Dravet Syndrome

To make this in-vivo assay possible in the first place, the authors draw on the knowledge of SCN1A gene expression and its analogue called scn1lab in zebrafish.

“SCN1A encodes the pore-forming alpha subunit of the NaV1.1 sodium channel and is widely expressed throughout the central nervous system. SCN1A mutations are linked to variety of childhood epilepsies in humans. […] Over 1200 SCN1A mutations have been identified to date and the most severe clinical phenotypes correlate with complete loss-of-function mutations or point mutations in critical residues of the pore region.”

Furthermore, “The zebrafish scn1lab gene (previously known as double indemnity or didy) encodes a voltage-gated sodium ion channel alpha subunit that is orthologous to the mammalian SCN alpha subfamily comprising SCN1A, SCN2A, SCN3A, and SCN9A26, 27. Mutations in SCN1A—and to a lesser extent SCN2A, SCN3A, and SCN9A—are associated with a variety of monogenic childhood epilepsies such as DS in humans.”

The authors verified this genetic hypothesis by exposing both the ‘mutants’ and ‘wild-type’ siblings to light stimuli. Only the mutants exhibited “seizure-like locomotor activity”.

Drug pre-screening

From a group of 154 drugs under consideration, the authors proceed to select 31 for in-depth evaluation based on a threshold criterion: “50% reduction in abnormal light-triggered locomotor activity” upon administration of the drug. Interestingly, agonists and positive allosteric modulators of the GABA-A receptor – including allopregnanolone and ganaxolone - were over-represented in this first selection.

“Our preliminary hits included compounds with a wide variety of targets and appeared to be substantially enriched for agonists and positive allosteric modulators of γ-aminobutyric acid (‘GABA’) receptors, particularly the GABAA receptor”

Allopregnanolone and ganaxolone are of course the two flagship molecules evaluated by Sage Therapeutics (SAGE) and Marinus Pharmaceuticals (MRNS).

Electrophysiology screening and bioinformatics

Having narrowed down their sample to about 31 drugs including allopregnanolone and ganaxolone, the authors proceed to the main selling point of this paper: an in-depth evaluation of each compound’s effect on mutant zebrafish larvae following induction of epileptic symptoms. Their approach consists of evaluating brain activity patterns with the help of advanced algorithms to establish ‘LFP complexity scores’: “In addition to the high-amplitude spikes that characterize seizure-like events, we observed that interictal LFP activity patterns in scn1lab mutants appear to be considerably less complex and more stereotypic than in sibling controls. We speculated that interictal pattern structure could provide another metric to evaluate the efficacy of neuroactive compounds. In order to assess this aspect of the scn1lab phenotype, we utilized independent component analysis (‘ICA’). ICA is an unsupervised analysis method for separating multivariate signals into independent subcomponents and is widely utilized for decomposition of EEG signals.”

As an aside; the guy who literally wrote the book on ICA is now a professor of machine learning at University College London - the sophistication of the underlying methodology is what really gives this paper an edge.

The authors go on to explain that the utility of this subcomponent analysis of neural activity is informed by comparing wild-type and mutant zebrafish larvae:

“When larvae are analyzed using our ICA approach, only a small number of ICs tend to dominate LFP traces from untreated scn1lab mutants (“low-complexity” LFPs), while traces from sibling controls comprise a far greater variety of ICs (“complex” LFPs; Fig. 3a).”

[Figure omitted: two panels from the paper. Caption: “Illustration of LFP complexity analysis on the left, ranking of compounds on the right-hand side. Source”]

Thus, the more brain activity patterns post-treatment of mutants resemble those of their non-mutant siblings, the higher the LFP complexity score awarded.

Based on this composite score, authors determine that 6 out of the 31 preliminary hits are “highly effective” at restoring LFP recordings to a “more wild-type state”. The compounds are:

• progesterone
• mifepristone
• pargyline
• promethazine
• fluoxetine
• allopregnanolone

By comparison, ganaxolone’s LFP score is closely aligned with diazepam, a first-line treatment for status epilepticus, yet both drugs are poorly ranked according to this LFP score. We will revert to the important question of what this all means for clinical development later on.

To corroborate their LFP score, the authors proceeded to an “in-depth behavioral assessment of all 31 preliminary hits using multiple independent metrics instead of a single-behavioral outcome”. On that basis, authors try to predict undesirable side-effects of the drugs:

“In this analysis, 19 out of 31 preliminary hits fall further from the wild-type end of the spectrum than untreated mutants, suggesting they alter relatively normal interictal behaviors and may have undesirable side effects that could give rise to false positives (e.g., sedation) or indicate other off-target concerns. We therefore flagged compounds ranked in the upper quartile of the interictal behavioral spectrum (i.e., those furthest from siblings) as abnormal.”

[Figure omitted: panels C and D of the paper’s behavioural analysis. Caption: “Section C: both allopregnanolone and ganaxolone are deemed to induce ‘abnormal’ behavior in the timeframes between seizure-like behavior (‘interictal’). Section D: Further breakdown of compounds based on behavioral assessment by the authors. Source”]

Discussion of findings

Only a handful of compounds pass the double test of this multivariate behavioral assessment combined with the previous evaluation of local field potentials:

• Pargyline
• Mifepristone
• Progesterone
• Promethazine
• Fluoxetine

This is an odd but intriguing list of ‘ideal’ drugs for the control of intractable seizures. Pargyline, now discontinued, was once prescribed for hypertension. I found this paper discussing pargyline as an effective pre-treatment in the case of rats genetically susceptible to intractable epilepsies – which corroborates the findings in zebrafish. Mifepristone is a drug used to induce abortions; again, not something you would expect to be effective in controlling intractable seizures. Seeing how well progesterone, another top contender, performed in these tests, we might have a ready explanation in the fact that “mifepristone’s relative binding affinity at the progesterone receptor is more than twice that of progesterone”. Source. While progesterone has been found to be effective in certain female seizure disorders, its use is severely limited by the fact that it is a sex hormone which inevitably leads to undesirable side effects. Next on our list is promethazine, a strong sedative. This is weird, considering how the authors are keen on eliminating drugs with sedative effects that might mask a lack of anti-seizure activity. Last but not least, fluoxetine better known as ‘Prozac’ is a classic SSRI antidepressant. There is evidence that SSRIs may help reduce seizures in other animal models and in some human Dravet patients.

Readers should now understand that this animal model, though certainly instructive, does not yield results that can be translated 1:1 to human subjects. Besides obvious limitations due to side effects that cannot be captured by this zebrafish assay such as progesterone’s hormonal side effects, other considerations in humans include dosing and long-term tolerability. Another important question is whether an ‘ideal’ compound really exists in the sense of restoring ‘normal’ brain activity in patients with genetically caused epilepsy.

The authors acknowledge that neurosteroids hold significant promise in real-life settings (lengthy citation ahead):

“One potentially interesting class of compounds picked up by our screen is a structurally related group of C21 (pregnane) steroids. Our chemical library contained three members of this class (allopregnanolone, ganaxolone, and progesterone), all of which were among the top hits in our preliminary behavioral screen. Progesterone subsequently tested positive for anti-seizure activity based on both its composite LFP score and deep behavioral phenotyping, while allopregnanalone (3α-hydroxy-5α-pregnan-20-one) appeared to be highly effective based on its LFP score but showed some behavioral side effects. Among our top hits, allopregnanolone was the most effective at improving the seizure component of the LFP score (seizure score = 0.93) but the least effective at improving the ICA complexity component (ICA score = 0.69; Supplementary Table 3). Interestingly, although allopregnanolone’s effect on interical metrics (temporal windows 6 and 7; Fig. 4c) caused us to flag it as abnormal in deep behavioral phenotyping, it still clustered with sibling controls (Fig. 4d) due to its efficacy at reducing abnormal light-triggered ictal behaviors (temporal windows 1–5). This outcome—strong anti-seizure activity in combination with abnormal locomotor activity and a relatively poor ICA score—is reminiscent of what we observed with high concentrations of diazepam and stiripentol (Supplementary Table 4). Taken together, these data suggest allopregnanolone may possess both anti-seizure activity and off-target side effects.

Allopregnanolone is an endogenous pregnane neurosteroid that is synthesized in vivo by 5α-reduction of progesterone. It has been shown to exhibit potent anticonvulsant effects in diverse animal models and clinically in patients with super-refractory status epilepticus and appears to be less subject to anticonvulsant tolerance than benzodiazepines. Unlike progesterone, allopregnanolone is not believed to have activity at nuclear steroid hormone receptors. Conversely, progesterone itself is not a GABAA receptor modulator, but it can be metabolized to allopregnanolone and other related compounds. The anti-seizure effects of progesterone may therefore occur through its conversion to allopregnanolone, as has been shown in other animal models. However, the time course of this effect is remarkably rapid based on our data, with light-induced seizure-like locomotor activity dropping to 31% of baseline within 45 min of application (Supplementary Data 1) and spontaneous LFP-detected seizure frequency decreasing to 27% of baseline in a comparable timeframe (Supplementary Table 3, seizure frequency, 45 min column). Based on our results, allopregnanolone and compounds with related pregnane skeletons warrant further investigation as therapeutics for DS.”

These are important acknowledgments as they at once put forth a partial explanation for the great score obtained by progesterone all the while underlining that allopregnanolone and related neurosteroids may feature significant advantages over benzodiazepines.

However, allopregnanolone has not been able to show efficacy in a recent PhIII trial conducted by Sage Therapeutics in super-refractory status epilepticus (‘SRSE’) and an additional issue with (long-term) administration of allopregnanolone is the potential effects of back-conversion into… progesterone. Marinus Pharmaceuticals’ Ganaxolone on the other hand is not supposed to back-convert into progesterone, allowing for long-term administration versus Sage’s ‘one off’ administration of allopregnanolone in recent depression trials. Most importantly, ganaxolone has proven to be efficacious in patients with severe genetic epilepsies such as CDKL5.

Concluding thoughts

While this is a particularly interesting and sophisticated piece of pre-clinical work which gives important clues with regards to potential treatments for genetic epilepsies, it doesn’t allow us to circumvent clinical trials in humans. Additionally, as demonstrated by allopregnanolone’s PhIII trial failure in SRSE, differences between seizure disorders and sub-populations make it difficult to extrapolate these findings in zebrafish onto other seizure disorders in humans.

The same is true for virtually any disease: pre-clinical testing is necessary and can be useful but never sufficient to establish a treatment’s efficacy in humans. As someone famously said, “we can cure any cancer… in the mouse.”

Disclosure: I am/we are long MRNS. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. I have no business relationship with any company whose stock is mentioned in this article.