Onconova is developing a late-stage clinical asset and has run up 108% over the last month. With the run up completed, Onconova represents a more risky bet with an upcoming binary event. In the meantime, a recently published debate on Onconova's primary asset, Rigosertib, raises questions about the drug's MOA.

Back in June, and what seems like a lifetime ago by 2020 standards, I was going through my watch list of approximately 300 companies in the biotech space and finishing my quarterly review of quarterly earnings reports. An unlikely candidate rose up to the surface after being left to the outer darkness of my extended watch list for several years. Onconova, a development stage Phase 3 biotech with an upcoming catalyst in MDS and an attractive valuation (which at the time fully diluted was around $90M with an enterprise value of around $60M). That valuation felt low given the point at which they were at in their lifecycle.

I was familiar with this company from its original IPO. When it debuted, I was drawn in because it fell squarely in my zone of interest with its development of Rigosertib. At the time, the MOA was described as an inhibitor of PI3 kinase, which should function to slow cellular proliferation in various cancer indications. At the time, I was very interested in the potential of coupling genomic analysis with targeted cancer therapeutics, so Rigosertib fit nicely into the toolbox that I envisioned being utilized in individualized cancer therapies. The molecule had been first characterized by Dr. Premkumar Reddy, a highly accomplished cancer researcher, who also happens to be the original founder of Onconova. Dr. Reddy continues to be involved with the company as a member of their board of directors.

As of their last quarterly report, Onconova had a cash position of $31 Million. They have guided that they will not need to raise additional capital before the readout of INSPIRE. Additionally, since their financing activities in November of 2019, $10.6 Million in additional capital has come into the coffers via the exercising of outstanding warrants. The company indicated that there are approximately 29 million warrants outstanding at the end of March 2020, and 80% of them were in-the-money as of May 13th. They guide that they have sufficient cash to fund operations through 3Q2021. This was an attractive feature in my evaluation of the company that there would likely be no capital raise before the release of pivotal clinical results. Onconova appeared ripe for a run up.

Before I get started on all of the background on the history of Rigosertib I need to give enormous credit to Jonathan Weissman at University of California San Francisco, who published work on this subject that was incredibly informative on the various efforts in elucidating the MOA of Rigosertib. I have not interviewed Dr. Weissman, but his published work was a great resource for the background in this article. Prior to finding his work, we took a significant position in the company at a cost average of $0.45 per share. But recently, after reviewing his work published on July 2, 2020, and performing a retrospective and comprehensive analysis of the work of others in this space, we have closed the entirety of our position.

Early studies with Rigosertib revealed its promise as a potent anticancer agent. Initially, it was hypothesized by Reddy et al that the drug worked on polo-like kinase1 (Plk1) and arrested cells during mitosis and induced apoptosis. Mitotic spindle abnormalities were identified in these early studies which were deemed to be downstream from the compounds inhibition of Plk1. Binding affinity for Rigosertib to Plk1 was found by Reddy's group to be strong with an IC50 of 9-10nM.

Soon after this publication, others questioned the affinity of Rigosertib (referred to in early studies as ON01910). Steegmaier et al was testing the affinity of a different drug, BI 2536, against Plk1 and found no appreciable affinity of Rigosertib against Plk1 up to a concentration of 30,000nM. For those unfamiliar with binding studies, the IC50 represents the concentration at which 50% of the target would be inhibited, or bound. A lower IC50 means that the binding is more efficient. For a "good drug" we like to see low nano molar affinity for its target ligand, and no appreciable affinity for other "off-target" molecules.

Steegmaier's study found no appreciable direct binding of Plk1 at concentrations 3000 times higher than the IC50 that was determined by Reddy. Instead, Steegmaier claimed that "The cellular effects caused by ON01910 resembled the phenotype that is caused by treatment with low doses of microtubule depolymerizing agents rather than Plk1 RNAi. Supporting this notion, depolymerization of microtubules was clearly visible in interphase cells when these were treated with higher doses of ON01910".

I have not seen data where Reddy or Onconova retracted the Plk1 claim, but eventually additional studies were carried out by Reddy's group and it was proposed that Rigosertib targeted PI3 kinase signaling. However, here again, independent groups were unable to confirm this and similar to Steegmaier, Maki-Jouppilia et al implied that the effects of Rigosertib were targeted at the mitotic process rather than a modulator of signal transduction.

After Onconova's IPO I was drawn to the company because of the indication being tested. And although the trial was with IV Rigosertib, I liked that the drug was orally bioavailable and that the side effect profile looked pretty good. At the time of their IPO, they were conducting a Phase 3 trial dubbed "ONTIME" for Myelodysplastic Syndrome. For patients who are not good candidates for a stem cell transplant, MDS was essentially a death sentence with limited treatment options and it represented a largely untapped field for an effective new agent. At the time of their IPO, the only approved therapies were the hypomethylating agents, Azacitidine and Decitabine.

Since then, of course, three new agents have been approved in MDS. These are Lenalidomide for transfusion dependent del(5q) MDS, Luspatercept-aamt, and Inqovi (an oral combination of decitabine and cedazurine for intermediate and high risk MDS). Still, MDS patients have a poor prognosis, so Onconova represented precisely the type of drug and company that I feel good investing in. The ONTIME trial enrolled 299 MDS patients that were treated 2:1 in the Rigo arm. However, despite demonstrating signs of efficacy, the results failed to reach statistical significance. Median overall survival was 8.2 months in the Rigo group and 5.9 months in the best supportive care group, but the P value was 0.33. I'll come back to this data later in this article. But the bottom line is that ONTIME failed.

In the aftermath of the trial's failure, they were able to identify a sub-group of patients for whom the drug looked more promising. In the 184 patients who had progressed on or failed previous treatment with HMAs, the median survival was 8.5 months vs. 4.7 months (p=0.022). And the study was able to build on the SAE profile for Rigosertib with Grade 3/4 hematologic and non-hematologic AE's in less than 7% and 3% of patients, respectively. Not bad when you consider that the overall health of this patient group was pretty fragile.

So with this promising subgroup analysis, the company vowed to redo a Phase 3 trial looking at patients who matched the subgroup. That was right about when I lost track of them. The share price languished, resulting in two reverse splits and compromised capital raises involving warrants. I'm never a fan of warrants, but they have a place and time in attracting investment when the potential reward may be a way off.

It would cost Onconova a great deal of capital and an indeterminable period of time to bring the trial to fruition. In the meantime, additional pre-clinical data were generated by the company around Rigosertib's MOA.

A few words about subgroups before I continue. When we conduct clinical trials, we are simply conducting an experiment in which we seek to answer a question. Easy enough, right? In the land of biostatistics, when we design a clinical trial to answer one question, we utilize assumptions in our hypothesis to power the trial with the appropriate number of patients to answer the question...that question, and only that question. Very often, when a trial is completed, we generate lots of other data and within this data we almost always find other bits of interesting information. Sometimes this information can guide new experiments that will answer new questions. However, it is a bad idea to take this ancillary information as gospel. I was down at FDA some time ago to hear deputy director and FDA legend Robert Temple speak on this topic. He recanted a few fascinating examples.

One of the most referenced examples of subgroup analysis gone awry was in the Anturane Reinfarction Trial in which FDA rejected sulfinpyrazone for the prevention of sudden death following heart attack. This study looked at 1600 patients with a recent MI and followed them for two years. There was identified in the study a 74% reduction in sudden death in a subgroup of the treatment arm and multiple groups for which the p values were well under 0.05. However, the data excluded some patients and when the pooled group was analyzed as originally designed, the treatment failed to meet the threshold of statistical significance. Such was the data manipulation with Anturane that FDA published their rationale in NEJM as an example for the industry in 1980 and as I listened to Dr. Temple speak about this example literally 40 years after the original FDA decision, I was reminded about how timeless some clinical failures are in instructing the future. Another useful retrospective analysis of the trial can be found here. Ultimately, sulfinpyrazone was not found to be efficacious in preventing cardiac arrest or secondary MI in post infarction patients.

Expressed another way, subgroup analysis is kind of like shooting at a barn and then drawing a bullseye around the bullet hole. You would be correct in concluding after-the-fact that you hit the bullseye, but if you shoot at the barn a second time, you'll probably hit the barn again, but not the bullseye. Statistical significance is customarily accepted as a P value equal to or less than 0.05 or a less than 5% probability of happening by chance. When you allow multiplicity to occur with multiple subgroups, the probability of identifying a false positive is directly in proportion to the number of subgroups analyzed. For example, if you look at 20 different subgroups, you have nearly a 100% chance of identifying a false positive. Generally, companies present a subgroup finding not as a false positive, but as a potentially promising finding (but they generally neglect to mention that they cut up the data forwards, backwards and sideways to come to the finding). Here's another great article on the topic for those that want further reading.

Getting back to Onconova and Rigosertib, the new Phase 3 trial of Rigosertib in MDS was named "INSPIRE" and began enrolling patients in December of 2015. This trial would initially enroll 225 patients, again 2:1 Rigo vs. "Physician's choice". The primary endpoint would be overall survival as it had been in ONTIME, but also look at the International Prognostic Scoring System - Revised (IPSS-R) in the Very High-Risk (VHR) Subgroup. After an interim analysis by the Independent Data Monitoring Committee in early 2018, the trial was expanded to 360 patients on a pre-planned sample size re-estimation.

In the meantime, additional pre-clinical data were generated by Reddy et al., this time identifying Rigosertib as a RAS mimetic. A key journal publication in 2016 co-authored by Reddy, laid out compelling evidence that Rigosertib exerted its anticancer effects by binding the RAS binding domain (RBD) of RAS effectors, like RAF. Binding studies were reported with a sub-1nM Kd for both b-RAF and c-RAF. This new proposed MOA would represent a significant discovery because mutations in various RAS effector genes are implicated in 20-30% of cancers. If you follow the sector, you probably know that the RAS-pathway has been a prime, yet elusive target for cancer therapeutics. Several companies have pursued K-RAS and other RAS candidates (Amgen and Mirati, in particular), but the failures and only partial successes over the years have caused some to call RAS an "undruggable" target.

Unfortunately, as it was with Plk1 and PI3k, the controversy around Rigo's MOA didn't stop there. Although the company still regularly references the 2016 article, shortly after the article's publication, a different group, Ritt et al., conducted experiments in which they found Rigosertib did not block interaction of RAF with RAS, again at concentrations far higher than those looked at by Reddy's group. They did find signs of JNK signaling, but they could not conclude with the experiments conducted if the effect was direct or indirect, but since binding of RAS was unaffected, they concluded that any effects observed on RAS signaling would be indirect and downstream from Rigosertib's actual MOA. The only reason I can speculate as to the stark difference in binding between the two groups is that Reddy's group tested Rigo binding against a shortened version of the RAS Binding Domain of b-RAF and c-RAF in a recombinant construct rather than the whole protein.

Meanwhile, Anang Shelat's research group at St. Judes published work in 2016 applying phenotypic grouping to 154 molecules, including Rigosertib. The study was conceived as a methodology in high throughput screening for novel drug candidates. As they ran the exercise, 78% of the molecules correctly matched to the phenotypic class previously identified for them. One of the notable failures from the study was Rigosertib which classified in the study as a microtubule disruptor rather than an inhibitor of intracellular signaling with Plk1. The study also referenced a structural analysis of another compound in the study, TL-77, which was a close analog of Rigosertib. It's MOA? Strong direct inhibition of tubulin polymerization, and no interaction with Plk1 or PI3k signaling pathways.

All of the controversy caught the eye of Jonathan Weissman, who was using a CRISPR-mediated chemical-genetic strategy for identifying molecular targets. Applying his method to Rigosertib pointed to a MOA as a microtubule-destabilizing agent, but then he went further to demonstrate direct tubulin binding in several cell lines, and also by showing Rigosertib's ability to inhibit microtubule growth in vitro in a concentration specific manner. He went on to identify the colchicine site of tubulin as the Rigosertib binding site by isolating and analyzing crystal structures of the complex formed by alpha and beta tubulin. Finally, he identified a beta tubulin mutant with modifications in the colchicine binding site that conferred resistance to Rigosertib. In my view, Weissman's work was pretty comprehensive.

Brief side note: Colchicine was originally derived from the autumn crocus in the 1800s and is still used for the treatment of gout. If you're an adept botanist, be careful, because too much is deadly because of its anti-mitotic effects. Many early therapies have their origins in flower extracts, as was the case with the Madagascar periwinkle. In the 1950s, researchers from Eli Lilly noticed that the flower extract significantly decreased the level of white blood cells in mice, which lead to the discovery of two mitotic disrupting molecules, Vincristine and Vinblastine, which are still widely used today globally.

The disagreement between Reddy and Weissman culminated in the journal Molecular Cell on July 2, 2020, with dual publications on the matter. Reddy's group published data that a contaminant impurity in commercially available Rigosertib, ON01500, was responsible for the tubulin binding rather than Rigosertib itself, and that in Pharmaceutical grade Rigosertib, the impurity is absent and therefore, does not bind Tubulin. However, in response to this study, Weissman's group repeated the analysis on Pharmaceutical grade Rigosertib and was able to repeat their previous results.

Not only did Weissman repeat the results of microtubule inhibition with pharmaceutical grade Rigosertib, but they also went on to refute Reddy's rebuttal to Weissman's original study nearly point by point. Rather than go through the rebuttal section by section, you should read the paper, and especially the discussion section for yourself here. While they don't specifically accuse Reddy or Onconova of fabricating data, they write, "We find it impossible to reconstruct what may have led to these conflicting results and thus choose not to speculate about the origins."

Whenever I invest in a drug-in-development, I always feel better when I can find independent research saying similar things about the drug and/or target. That was actually how I ran into the conflicting research because I was digging into the MOA so that I could learn more and write an article in support of the company (we were long at the time). But that's science, it doesn't always give you the answers you want. When you can couple unbiased and independent scientific literature to rationalize claims put out by companies on the FDA development path, you will always have a stronger probability of success. That just wasn't the case with Onconova, unfortunately, and I adjusted accordingly.

With Rigosertib, you essentially have the company and the founder behind the company saying one thing, and then the scientific community at large saying something else. That said, I'm certain Reddy and Onconova utilize collaborators and/or contractors that may have independently verified some or all of their data, but the truth remains that I cannot find any truly independent research published that agrees with the company's findings. Their publications and the science behind them appear to be well thought out. Unlike the independently published literature, the company, and presumably Reddy, have a financial interest in the success of Rigosertib. So it begs the question why the company is seemingly in denial of what I would characterize as comprehensive and compelling evidence from multiple research groups over nearly 15 years that Rigosertib is functioning as a microtubule-destabilizing agent.

I don't think this really changes the current INSPIRE trial. Obviously, that trial is now on autopilot until it reads out later in the year. But for everything else in the hopper for Rigosertib, and in particular, its recently announced investigation into K-RAS mutated lung adenocarcinoma, this changes everything. The rationale for the trial is that Rigosertib exerts a direct effect on RAS signaling, but the external evidence suggests that any effect on RAS is indirect and secondary to its primary MOA on tubulin. In other words, the rationale for this trial is likely misguided. In my view, the company needs to take a harder look at the external data generated about Rigo's MOA and potentially retool their entire clinical strategy if they can change their view at some point. This includes their hints of Rigosertib's potential utility in COVID-19, based on this. I don't think you could argue that Vincristine with its MOA in mitotic arrest would be a logical approach for COVID-19, a similar rationale would probably follow for Rigosertib.

It's worth remembering that subgroup analyses often result in false positives that are likely to fail when retested. On that basis, INSPIRE is more than likely going to fail. However, I think INSPIRE has a chance of reading out with a nominal survival benefit in MDS. Here is my rationale. In the ONTIME trial, there appeared to be a survival benefit that failed to achieve statistical significance. Recall overall survival was 8.2 months in the treatment group and 5.9 months in the BSC group. In the treatment experienced subgroup, the survival in the BSC group was lower and Rigo arm was slightly higher, and this pushed the subgroup P value below the threshold of 0.05. So without seeing the totality of the data, I think that the study didn't have so much of an efficacy problem, but rather a standard deviation problem where this original study included both treatment naive and treatment experienced patients. I suspect that this variety in the disease timeline for patients enrolled in ONTIME resulted in a greater standard deviation in the survival and that if they had picked a more specific group, they may have pulled off a win the first time around.

That said, 2.3 months, or even the 3.8 months in the subgroup, isn't a huge survival benefit, but it is possible that with a more refined patient population as they've built into the INSPIRE trial, that the deviation will be lower and they will achieve the P value needed to deem it successful. Plus, they have two shots on goal with the trial design looking at the total study population, but also the Very High Risk (VHR) sub-population. In summary, I estimate INSPIRE has a 40% chance of success. Not great, but not terrible odds if you're the gambling type. However, if you are invested thinking that this is a slam dunk, you should probably adjust accordingly.

If INSPIRE fails, I think Onconova has a difficult decision to make in terms of continuing to throw money at a drug that has failed two Phase 3's. In other words, if INSPIRE fails, I think they go to zero. If INSPIRE succeeds, the company will have lots of work to complete for an NDA. The share price will spike, however, if I worked for FDA, I can already imagine lots of additional questions I would want answered about Rigosertib. Their recent denial of the published literature may work for regaining Nasdaq compliance, but it's not going to work on the mother ship.

For me, the more concerning thing is the company's reluctance to embrace the external science that points to a MOA that differs from the company's public pronouncement that Rigosertib acts on the RAS pathway. After my review of the literature on this matter, I emailed Investor Relations at Onconova on July 22. My question was simple, does Onconova acknowledge or refute the Weissman study that was published alongside the company's study on July 2, 2020. The next day I received an email back that the company had put out a PR that essentially restated their position on the matter and addressed NONE of the criticism of their own study. I found their response lacking given the strong and cogent data presented by Weissman. A Latin legal principle comes to mind, "Falsus in uno, falsus in omnibus".

Like many these days, I enjoy reading the conversations and comments on StockTwits and Twitter (mostly as a quiet observer) for the various equities I follow. If nothing else, they're entertaining. When you look at Onconova specifically you'll find a population of 30,000+ investors that are "in the room". Many posters claim to have completed significant due diligence on the company, yet none of them have challenged the company on the vast body of published data questioning the assertions of the company regarding the MOA of Rigosertib. Moreover, I witnessed many investors claim that the July 23rd PR stimulated by my innocent question to Investor Relations was "Great news".

With professional sports on a seemingly indefinite pause, many retail investors who have never made direct investments in equities are throwing money in the ring. Call it a combination of boredom and nowhere else to spend your stimulus checks, I suppose. If you're looking for big gains, biotech can indeed be a point of differentiation in your portfolio, and nowhere are these gains (and losses) bigger than with the small caps. However, credible investors in the biotech space need to question everything and not simply rely on the claims of the company and/or other investors to decide where to put their money. We are in interesting times, folks, and the technical and momentum traders treat any shred of PR like free press, even if they don't understand the reason behind it. The Sheriff of Nottingham will take your gold soon enough if you fail to do your homework.

In summary, Onconova has enjoyed a significant appreciation in share value, but investors should tread carefully into the readout of INSPIRE and consider how the conflicting data around Rigosertib's MOA could potentially impact the overall value and potential of this drug.

Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours. 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.

Read more:
Onconova And The Conflict Lurking Beneath The Hype - Seeking Alpha

Comments are closed.