VDP-98 [Funding]: ETTA Biotechnology

[Paolo]

One-liner: ETTA is an early-stage biotechnology company developing AI/ML-designed novel molecules for simultaneously targeting senescent cells, mTOR, and insulin signaling to extend lifespan and treat serious diseases, such as chronic kidney disease.

Project PI: Kyle Brewer

Longevity Dealflow WG team

• Senior reviewers: 1 entrepreneur, 1 pharma professional, 2 scientists, 1 VC
• Shepherd: Paolo Binetti
• Squad members: Tuan Dinh, Laurence Ion, Vivian Qu, Derek Underwood
• Sourced by: Lutz Kummer

Simple Summary

ETTA Biotechnology is using its proprietary artificial intelligence (AI) and machine learning (ML) platform to generate novel molecules to extend healthy mammalian lifespan and demonstrate treatment of chronic kidney disease in clinical trials. ETTA has a lead novel compound in hand verified to target senescent kidney cells, without toxicity to normal cells. ETTA has also discovered a FDA approved drug that can clear senescent kidney cells, which will be improved with the AI/ML platform due to some apparent toxicity to normal cells.

Problem

37 million people in the US with chronic kidney disease will eventually die as a result of disease progression, caused by the build up of senescent cells in their kidneys during aging. Without a transplant, chronic kidney disease leads to severe illness and death.

During aging, the kidney builds up senescent cells, leading to chronic inflammation, causing the kidneys to begin to perform poorly. As excess damage occurs from inflammation and senescent cells, fibrosis occurs in the kidney, which further exacerbates the decline of renal function. The anti-cancer drug ABT-263 (Navitoclax) was shown to clear senescent cells for a chronic kidney disease model in aged mice, leading to improved kidney function, reduced kidney fibrosis, and even regeneration of the kidney through cellular proliferation [1]. In addition, dasatinib plus quercetin has been shown to limit renal fibrosis [2]. However, many senolytics such as ABT-263 and dasatinib cannot be used safely in older, frail populations due to toxic side effects. Also, patients with chronic kidney disease do not deal with toxin challenges well, making the use of ABT-263, dasatinib, and similar drugs even more of a problem.

Solution

ETTA’s novel small molecule drugs act by a mechanism targeting inflammatory (mTOR) and insulin signaling (PI3K/AKT) pathways, allowing senescent cells, normally resistant to apoptosis, to undergo apoptosis. Clearing senescent cells in the kidney enables resumption of proper renal function and even kidney regeneration in aged mice [1]. In recent years, natural products, such as fisetin, have been discovered that can simultaneously clear senescent cells, reduce inflammation, inhibit insulin signaling, and extend mammalian lifespan. Furthermore, many of these compounds are present in low concentrations in common fruits and vegetables, offering great promise for safe treatments targeting kidney disease3 during aging that are also safe enough for millions of people to use regularly. However, these natural molecules have poor solubility, bioavailability, and half-life, which is the main hurdle to implementing these discoveries as therapeutics.

For preclinical development, ETTA has created a proprietary AI/ML platform, which combines currently available data from natural molecule-based senolytics to recognize and predict the most effective senotherapeutics in silico. In addition, 15 additional criteria related to safety, bioavailability, and metabolism are applied to choose the best possible drugs to maximize clinical success probability from the outset. ETTA has generated nearly 2000 proprietary small molecules that are predicted to outperform natural molecule senolytics.

After synthesizing our first candidates from the platform, we induced senescence in normal kidney cells in culture. We then incubated the cells with each compound for 48 hours. Of our first lead candidate molecules, we found ETTA1 can clear senescent cells (Fig. 1), while the total number of cells was unaffected. We found fisetin can also clear senescent cells, as published previously by our advisor Matt Yousefzedah [4], albeit not to the same extent as previously published, likely due to kidney cells being used here, while mouse embryonic fibroblasts and IMR90 fibroblasts were used in Dr. Yousefzedah’s publication. In addition, fisetin appears to not have the same selectivity or efficacy as ETTA1, which demonstrates the power of our platform.

ETTA2 and ETTA3 only trended toward lower senescence, likely because our machine learning training dataset currently contains only 108 molecules, so our results expectedly show prediction of senescent cell targeting-small molecules is imperfect, but practical. We expect to have even more accurate predictions as we increase the number of molecules we discover to deepen the training dataset. Noteworthy is that ETTA2 and ETTA3 do not affect total cell number, likely owing to the fact that safety is much easier to predict with AI/ML approaches due to thousands of small molecules in existence that have some safety and/or toxicity data.

As our platform is based on natural molecules, which have had success in safety in the clinic [5], while other senolytics have had safety problems [6, 7], we expect our platform to outperform competitors for patient safety. In addition, drugs are filtered out of the body by the kidney, so we can expect to use lower concentrations of our molecules compared to competitors going after other organ targets. The kidney acting as a drug filter also allows us to more specifically target the kidney compared to competitors focusing on other organ targets. We expect other tissues or organs to be exposed to the drug, as with any small molecule. This fact can also be beneficial because reducing senescent cells in other organs will decrease the inflammatory cytokines circulating in the blood that may also impair kidney function. As our approach and platform are designed to excel in safety, these off-target senolytic effects are expected to help, not hinder our efforts.

To further expand our pipeline, we looked for candidates that inhibited both mTOR and PI3K/AKT pathways like fisetin does, which is the hypothesized mechanism of action to clear senescent cells. We discovered an FDA approved small molecule drug that inhibits these pathways can indeed clear senescent cells (Fig. 1). This discovery is important as it shows FDA approval is a strong possibility. However, total cell number may be affected (p=0.058), so we will use our AI/ML platform to make novel improvements for this drug.

Fig. 1 | Evaluation of lead candidates to clear senescent cells

Senescence was induced in normal kidney cells, then cells were incubated with each novel compound or fisetin at 10 µM concentration, or with DMSO vehicle only. Cells were stained with C12FDG to identify SA-β-gal+ cells, then Hoechst, as previously described [4]. Results were compared for n=4-6 independent wells, evaluated using unpaired Student’s t-test relative to the DMSO control, and plotted as the mean±SEM. *p<0.05, **p<0.01

Opportunity

Chronic kidney disease is a $95 billion dollar market and affects 1 in 7 people. The risk of death from chronic kidney disease exponentially increases with increased age, as with other major age-related diseases. Despite this opportunity, the only treatments available are for controlling blood pressure and diabetes to manage kidney decline. In absence of a transplant, dialysis is the only way to filter blood due to the failure of the kidney to function normally. Decrease in inflammation and scar formation, both of which are important to treat chronic kidney disease, appeared to be the main physiological outcomes in our previous work with fisetin nanoparticles, which is why we prioritize this disease. ETTA’s treatments would be the first drugs that target the underlying cause of chronic kidney disease and would likely be taken by all patients.

Other oral senolytics have severe safety problems, such as navitoclax (ABT-263) having testicular toxicity, lymphopenia, and thrombocytopenia6, and dasatanib having side effects of pleural effusion (“water on the lungs”) and pulmonary arterial hypertension [7]. In addition, companies such as Unity Biotechnology chose to do local injections due to the toxicity of their compounds. Fisetin, on the other hand, has shown no severe or serious side-effects in 60 subjects during ongoing clinical trials [5]. Our compounds are AI/ML-designed derivatives of natural products such as fisetin, and at least one of our first lead candidates appears to have better senescent cell targeting than fisetin (Fig. 1). Other competitors in this area include Oisin Biotechnologies, Rubedo Life Sciences, and Cleara Biotech, but we cannot evaluate the safety of these approaches as they do not yet have relevant data from clinical studies in humans.

Relevance to longevity

ML/AI designed molecules from ETTA’s platform likely target 3 key features of aging simultaneously:

1. Senescence by clearing senescent cells.
2. Insulin signaling by inhibiting AKT in the insulin signaling pathway.
3. Inflammation by inhibiting mTOR (the target of rapamycin).

Clearing senescent cells can extend lifespan by 17-35% in mice when treated throughout life [8]. Fisetin, used in our ML training data set along with closely related natural molecules, was previously shown to clear senescent cells and extend lifespan by ~10% when given to aged 22-24-month-old mice [4]. Likewise, insulin signaling inhibition with caloric restriction and mTOR inhibition with rapamycin can extend lifespan by 10-35% [9] and 23-26% [10], respectively, when mice are treated throughout life. Fisetin has also been shown to be a dual inhibitor of the insulin signaling (PI3K/AKT) and mTOR signaling pathways [11].

IP Roadmap

Two patents filed

• Patent 1: Oral Fisetin Nanoparticles
• Patent 2: Topical Fisetin Nanoparticles

Upcoming patent

• AI/ML designed novel molecules to clear senescent cells and treat kidney disease, inflammation, fibrosis, and aging.

Experimental plan

Tissue Culture Screening and Lead Candidate Determination
ETTA will synthesize additional top novel small molecule candidates from our AI/ML platform based on predicted efficacy, safety, bioavailability, metabolism, and chemical diversity. We will test these novel compounds applied to senescent kidney cells and normal kidney cells over the course of 3 days. Molecules that clear senescent cells with a p-value <0.05 (senescent cells compared to normal cells) will be considered appropriate for further testing to narrow down to a lead candidate. These molecules will then be tested at different concentrations to determine the half maximal effective concentration (EC50) for both clearing senescent cells and normal cells. Novel molecules with the lowest EC50 for senescent cell clearance and the highest ratio of EC50 for normal cells to EC50 for senescent cells will be tested in vivo.

Lifespan Study in C. elegans
We will apply our novel lead small molecules to C. elegans to determine any effect on lifespan. As molecules in our machine learning dataset also feature mTOR inhibition and insulin signaling inhibition beyond just clearing senescent cells, C. elegans screening may determine effects of our novel compounds on these pathways, which are conserved from worms to humans. Any hits on lifespan extension will be repeated in a daf-2 background to elucidate if the lifespan extension is due to insulin signaling inhibition or another effect.

Kidney Disease Model in Aged Mice
Based on ETTA’s previous work with fisetin nanoparticles, we have chosen kidney disease as a preferred indication to examine in a preclinical model. We will use the unilateral renal ischemia-reperfusion model of kidney injury in 22-month-old mice as described previously for ABT-263 senolytic treatment1. The mice will be treated by oral gavage with water (negative control) or our top lead novel molecule from our AI/ML platform. Kidney function will be determined by blood urea nitrogen and cystatin C levels in serum. Serum cytokines will also be examined to determine inflammatory markers. Senescent cell burden will be determined by p21 staining. Fibrosis will be assessed by picrosirius red staining in kidney tissues slices, and renal cell proliferation will be determined by Ki67 staining. Clinical observations will also be made for the mice (body weight, appearance, and health evaluation scoring) to determine any overt physical effects of the treatment, which may also correlate with effects on physical aging.

Budget

• Medicinal chemistry: $25,000
• In vitro screening: $5,000
• In vivo lifespan in C. elegans: $5,000
• In vivo chronic kidney disease model in aged mice: $35,000
• Patent filing: $20,000
• Overhead (10%): $10,000
• Total: $100,000

Financing and VitaDAO funding terms

ETTA has raised $100K pre-seed funding and is now raising a $1.5M seed round using the recently obtained data with the first compounds we have discovered. ETTA also has soft commitments of $100-200K from other investors, as well as interest from additional investors.

VitaDAO would fund $100K under Molecule’s sponsored development agreement terms for the line items in the budget. The seed round funds will be used for further determination of the optimal drug formulation, drug dose, and drug safety in mice, as well as further development of the platform and additional lead molecules, such as the FDA approved drug candidate we have discovered to clear senescent kidney cells.

Team

Kyle Brewer
Kyle founded ETTA Biotechnology to create the most effective treatments targeting aging currently possible. The first therapies he designed have the potential to extend healthy human lifespan by 10-30 years if preclinical mouse studies apply similarly to humans. Kyle has raised $100,000 for ETTA Biotechnology thus far.
Kyle is a scientist and entrepreneur with >15 years of expertise in aging and biotechnology, with a special focus on delivery of small molecules, DNA, mRNA, and proteins to extend lifespan. Kyle was the second scientist at Rejuvenation Technologies, a Stanford startup out of Helen Blau’s lab dedicated to using mRNA nanoparticles to extend telomeres, where he learned how to run and manage an early biopharma startup.
Kyle researched synaptic vesicle exocytosis in collaboration with Nobel Prize winner Thomas Südhof during his PhD in biophysics at the University of Texas Southwestern. He also investigated how circulating proteins in the bloodstream can influence the aging brain during his postdoc at Stanford in the lab of Tony Wyss-Coray, collaborating with Nobel Prize winner Carolyn Bertozzi.

Lurong Pan
Lurong recently joined as ETTA co-founder. Lurong is an expert in using ML and AI to generate novel molecules that outperform existing real world molecules. She founded Ainnocence, a company geared towards this goal that develops small molecules and antibodies using an AI approach. She was also the Director of Computation Science at UAB. In addition, Lurong is a Senior Investigator/Advisor to the Global Health Drug Discovery Institute, a Gates Foundation funded effort to advance pharmaceutical research and translational medicine.

Advisors

• Matt Yousefzadeh. Assistant Professor, University of Minnesota, discovered the natural molecule fisetin can clear senescent cells and can extend lifespan.
• Sam Roosz. Chief Executive Officer, Crescendo Health, founder of Datavant, a unicorn company with focus on clinical data analytics. MBA from Stanford Business School.
• Colin Maraganore. Co-Founder and Principal of Abeja Ventures. Venture Fellow at Healthspan Capital. Early Stage Biotechnology Operations and Strategy Specialist. Previously the Head Of Business Development at Rejuvenation Technologies, Scientist at BioAGE, and Senior Research Associate at Vium.

Bibliography

1. Mylonas, K. J., O’Sullivan, E. D., Humphries, D., Baird, D. P., Docherty, M.-H., Neely, S. A., Krimpenfort, P. J., Melk, A., Schmitt, R., Ferreira-Gonzalez, S., Forbes, S. J., Hughes, J., & Ferenbach, D. A. (2021). Cellular senescence inhibits renal regeneration after injury in mice, with senolytic treatment promoting repair. In Science Translational Medicine (Vol. 13, Issue 594). American Association for the Advancement of Science (AAAS). https://doi.org/10.1126/scitranslmed.abb0203 - The senolytic navitoclax (ABT-263) restored kidney function and even led to renal regeneration in normal, aged mice.
2. Li, C., Shen, Y., Huang, L., Liu, C., & Wang, J. (2020). Senolytic therapy ameliorates renal fibrosis postacute kidney injury by alleviating renal senescence. In The FASEB Journal (Vol. 35, Issue 1). Wiley. https://doi.org/10.1096/fj.202001855rr - The senolytic cocktail of dasatinib plus quercetin limits kidney fibrosis (scarring).
3. Ren, Q., Guo, F., Tao, S., Huang, R., Ma, L., & Fu, P. (2020). Flavonoid fisetin alleviates kidney inflammation and apoptosis via inhibiting Src-mediated NF-κB p65 and MAPK signaling pathways in septic AKI mice. In Biomedicine & Pharmacotherapy (Vol. 122, p. 109772). Elsevier BV. Redirecting - Fisetin limits renal inflammation and apoptosis in a common kidney injury model.
4. Yousefzadeh, M. J., Zhu, Y., McGowan, S. J., Angelini, L., Fuhrmann-Stroissnigg, H., Xu, M., Ling, Y. Y., Melos, K. I., Pirtskhalava, T., Inman, C. L., McGuckian, C., Wade, E. A., Kato, J. I., Grassi, D., Wentworth, M., Burd, C. E., Arriaga, E. A., Ladiges, W. L., Tchkonia, T., … Niedernhofer, L. J. (2018). Fisetin is a senotherapeutic that extends health and lifespan. In EBioMedicine (Vol. 36, pp. 18–28). Elsevier BV. Redirecting - The natural molecule fisetin can clear senescent cells and extend the healthy lifespan of normal, 22-24-month-old mice.
5. Verdoorn, B. P., Evans, T. K., Hanson, G. J., Zhu, Y., Langhi Prata, L. G. P., Pignolo, R. J., Atkinson, E. J., Wissler‐Gerdes, E. O., Kuchel, G. A., Mannick, J. B., Kritchevsky, S. B., Khosla, S., Rizza, S. A., Walston, J. D., Musi, N., Lipsitz, L. A., Kiel, D. P., Yung, R., LeBrasseur, N. K., … Kirkland, J. L. (2021). Fisetin for COVID‐19 in skilled nursing facilities: Senolytic trials in the COVID era. In Journal of the American Geriatrics Society (Vol. 69, Issue 11, pp. 3023–3033). Wiley. https://doi.org/10.1111/jgs.17416 - Fisetin has shown no severe or serious side-effects in 60 subjects during ongoing clinical trials.
6. Xiong, H., Pradhan, R. S., Nada, A., Krivoshik, A. P., Holen, K. D., Rhodes, J. W., … Awni, W. M. (2014). Studying navitoclax, a targeted anticancer drug, in healthy volunteers–ethical considerations and risk/benefit assessments and management. Anticancer Research, 34(7), 3739–3746. https://ar.iiarjournals.org/content/34/7/3739.long - The senolytic navitoclax (ABT-263) has testicular toxicity, lymphopenia, and thrombocytopenia.
7. Nekoukar, Z., Moghimi, M., & Salehifar, E. (2021). A narrative review on adverse effects of dasatinib with a focus on pharmacotherapy of dasatinib-induced pulmonary toxicities. In BLOOD RESEARCH (Vol. 56, Issue 4, pp. 229–242). The Korean Society of Hematology. https://doi.org/10.5045/br.2021.2021117 - The senolytic dasatinib has pleural effusion (“water on the lungs”) and pulmonary arterial hypertension as harmful side-effects.
8. Baker, D. J., Childs, B. G., Durik, M., Wijers, M. E., Sieben, C. J., Zhong, J., A. Saltness, R., Jeganathan, K. B., Verzosa, G. C., Pezeshki, A., Khazaie, K., Miller, J. D., & van Deursen, J. M. (2016). Naturally occurring p16Ink4a-positive cells shorten healthy lifespan. In Nature (Vol. 530, Issue 7589, pp. 184–189). Springer Science and Business Media LLC. Naturally occurring p16Ink4a-positive cells shorten healthy lifespan | Nature - Clearing senescent cells extends lifespan of normal mice by 17-35%.
9. Acosta-Rodríguez, V., Rijo-Ferreira, F., Izumo, M., Xu, P., Wight-Carter, M., Green, C. B., & Takahashi, J. S. (2022). Circadian alignment of early onset caloric restriction promotes longevity in male C57BL/6J mice. In Science (Vol. 376, Issue 6598, pp. 1192–1202). American Association for the Advancement of Science (AAAS). https://doi.org/10.1126/science.abk0297 - Insulin signaling inhibition by caloric restriction increases lifespan in mice by 10-35%
10. Miller, R. A., Harrison, D. E., Astle, C. M., Fernandez, E., Flurkey, K., Han, M., Javors, M. A., Li, X., Nadon, N. L., Nelson, J. F., Pletcher, S., Salmon, A. B., Sharp, Z. D., Van Roekel, S., Winkleman, L., & Strong, R. (2014). Rapamycin‐mediated lifespan increase in mice is dose and sex dependent and metabolically distinct from dietary restriction. In Aging Cell (Vol. 13, Issue 3, pp. 468–477). Wiley. https://doi.org/10.1111/acel.12194 - Rapamycin inhibition of mTOR can increase lifespan in mice by 23-26%.
11. Adhami, V. M., Syed, D. N., Khan, N., & Mukhtar, H. (2012). Dietary flavonoid fisetin: A novel dual inhibitor of PI3K/Akt and mTOR for prostate cancer management. In Biochemical Pharmacology (Vol. 84, Issue 10, pp. 1277–1281). Elsevier BV. Redirecting - Fisetin is a dual insulin signaling and mTOR inhibitor.

Slide deck

The slide deck is at this link.

Strengths

1. ML/AI designed molecules may target 3 key features of aging: 1. Cell senescence, 2. Insulin signaling, 3. Inflammation.
2. Many of the molecules used in the machine learning training data set have been used safely in humans already.
3. The ML/AI platform predicts key features important for successful drug development including efficacy, safety, metabolism, and bioavailability.
4. A novel molecule generated by the ML/AI platform has been validated yet to clear senescent kidney cells in vitro.
5. Clear clinical target

Risks

1. The novel molecules generated by the ML/AI platform have not been validated yet to treat an in vivo mouse model of disease or aging (this goal is part of this proposal).
2. Traditional pharma approach of “one disease, one target, one drug” is different from this approach, which is why targeting senescence, insulin signaling, and mTOR simultaneously is possible.
3. Competition from other longevity biotech companies also going for senolytics
4. Early stage: need to build the team further.

Senior Review Digest

Qualitative assessment

All 5 reviewers recommend to advance this project for community feedback and provide the following evaluation summaries:

• “While I do like senescent cells as a target and also the indication, their internal in vitro data is not convincing to me, and their strategy to improve on the molecules is not well developed.”

• “Overall I like their approach to using AI/ML technologies to find interventions for longevity and age-related diseases. I am hopeful that they can provide further validation results after more experiments and iterations. They have identified a good disease as a go-to market strategy. I would also recommend providing milestone-based funding to the company.”

• “I appreciate the set-up of the platform that is being built by ETTA, but it seems of limited use to me as it is built on a single cell line, a single method of senescence induction and only ~100 compounds. The lead compound is in its embryonic stage and I am not convinced that this will lead to a drug in the future, simply because data to convince me otherwise is lacking.”

• “CDK is a good first target which is actively being pursued by several big pharma companies. Targeting senescent cells in CDK is a new approach and is in line with VitaDAOs mission, however the team still lacks convincing data that they have a very effective and safe senolytic, and they have much optimization to do. Overall I am positive on the project after the responses from the team.”

• “ETTA Biotechnology showcases a promising and innovative approach to biotechnology by leveraging AI/ML technologies to expedite the drug discovery process. The company’s focus on targeting chronic kidney disease (CKD) offers a significant market opportunity, given the high prevalence of this condition globally and the limited current therapeutic options. Additionally, ETTA’s broader platform approach for longevity extension indicates substantial growth potential should the technology prove successful.
  However, caution is advised given the company’s early-stage status and the limited evidence of its technology’s effectiveness, particularly in humans. While the integration of AI/ML presents significant opportunities, it also introduces a degree of uncertainty and dependence. The reliability of AI/ML methodologies, the safeguarding of IP, and the potential for unforeseen technical challenges warrant close monitoring. Furthermore, the inherently lengthy timeline of biotech R&D, coupled with the regulatory complexities associated with novel drug approval, makes the path to market and revenue generation a long one.
  Given these strengths and weaknesses, a cautious approach to investment is recommended. Funding should be contingent upon meticulous evaluation of progress and developments, with a special focus on the effectiveness and reliability of the AI/ML platform, and clarity around intellectual property rights.
  Continued close monitoring of ETTA Biotechnology is advised to ensure that both the technology and business operations are developing as planned, and that the company is addressing the identified risks effectively. Future funding decisions should be based on evidence of progress in these areas, as well as the company’s ability to navigate the intricate biotechnology landscape successfully.”

Quantitative Assessment

The reviewers have scored the proposal on different aspects including general conviction, on a scale of 1-5 (with 5 being the highest). Here are the average scores:

• Novelty & Impact: 3.4
• Feasibility & Data: 3.0
• Relevance: 4.0
• Science Team: 3.2
• Market Advantage: 3.0
• IP Potential: 3.6
• Conviction score: 3.4

• Agree
• Revisions Requested (Detail in Comments)
• Disagree

[bowtiedshrike]

Slide deck is behind a Google sign-in.

I have mixed feelings about this one.

Looks like the approach and budget are reasonable. Seems very mission-relevant, both in that it deals with longevity and appears to be in the valley of death. However, Fig 1 is terrible, and I use quality of the preliminary data as a proxy for likelihood of success.

Weaknesses I see include the preliminary data and AI/ML approach. For the preliminary data, Fig 1 has a lot of problems. First, ‘4-6 independent wells’ sounds like technical replicates, not biological replicates. How many different times was this assay performed? T test is inappropriate for multiple comparison-- need to use ANOVA instead, which might cost significance on the FDA-approved drug. Absolute cell numbers are needed instead of relative (and stats need to be performed on absolute numbers, not relative-- normalizing to 1 for DMSO control for everything will give false statistical significance if there’s no error in that group). Not clear how many fields of view were used to quantitate, nor how many cells per field were counted for each. There must not be many senescent cells if you can cut them by 40% with no change to total cells. Or is ETTA1 a B-gal inhibitor instead, and it just interferes with the assay instead of clearing senescent cells? Also not clear on E50, dosing, solubility, binding target, etc. Presumably some of that is in the slide deck?

I’m generally skeptical of AI/ML approaches in bioscience. Even beyond my normal skepticism, 108 molecules seems small for an AI/ML approach, especially when there are screens of 10-50,000+ small molecules available for purchase (and have been thrown at the mTor pathway). The negative data are likely as important as the positive data.

One thing that further dampens enthusiasm is it sounds like the AI/ML platform has not predicted alternatives to the FDA-approved drug yet. Would like to see a couple leads in hand for that to test how well the AI/ML platform works. For the de novo leads, it is unclear how many leads they plan to make/screen in total. Out of the top 3 candidates, 2 failed already. Is the plan 2 more? 10 more? 100 more?

It is unclear what expertise the team has with kidney disease or medicinal chemistry. I don’t see any first/last author papers in the references from the team, which is another expertise concern. Is this a new area they are moving into? Advisors are either early career faculty or non-scientists, which is also concerning. What if the syntheses for new leads are hard?

The IP is for fisetin delivery, yet there is nothing about fisetin delivery in this proposal or how that relates to this proposal. Will the company be distracted by advancing its fisentin delivery method and/or trying to split between formulating liposomes for fistetin delivery vs developing new compounds?

[kyle]

Thank you for your helpful input. Your points are reasonable, and I have been thinking through them as well, so I provide some context below.

Slide deck is behind a Google sign-in.

Slide deck changed to publicly available now.

Looks like the approach and budget are reasonable. Seems very mission-relevant, both in that it deals with longevity and appears to be in the valley of death. However, Fig 1 is terrible, and I use quality of the preliminary data as a proxy for likelihood of success.

The point of Fig. 1 was intended to be a verification of the platform. The difficulty in AI/ML approaches in drug design is that it is easy to generate molecules, but very difficult to have molecules that actually do what is intended. The results in Fig. 1 are very positive because even with a limited training dataset for 108 molecules, we were able to create a novel small molecule that could effectively target senescent cells with just 3 synthesized molecules. This approach is a reduction of 100 to 100,000 fold in molecules tested over traditional, real life screening approaches. More importantly, the total cell number is not affected, which is promising in terms of safety compared to the other molecules. Senolytics have had safety problems, which has been a main stumbling point for others, so this first data point on safety is critical.

Another important point is we have previously performed case studies with fisetin nanoparticles. We have clinical leads in terms of treating conditions of inflammation and scarring in dermatology case studies using topical fisetin nanoparticles, as well as doing so safely. We have seen safety with oral fisetin nanoparticles as well, with promising anecdotal effects on inflammation. We have had feedback that a novel molecule would be needed for investment and partnerships for more extensive clinical studies, which is why we have pursued these novel molecules. Success from the outset with a novel small molecule related to fisetin, which is potentially better than fisetin in this initial data, means we will have much higher chances of clinical success than other approaches with our knowledge from previous case studies. I did not discuss these fisetin nanoparticles case studies as we are focused completely on the novel molecules at ETTA.

Weaknesses I see include the preliminary data and AI/ML approach. For the preliminary data, Fig 1 has a lot of problems. First, ‘4-6 independent wells’ sounds like technical replicates, not biological replicates. How many different times was this assay performed? T test is inappropriate for multiple comparison-- need to use ANOVA instead, which might cost significance on the FDA-approved drug. Absolute cell numbers are needed instead of relative (and stats need to be performed on absolute numbers, not relative-- normalizing to 1 for DMSO control for everything will give false statistical significance if there’s no error in that group). Not clear how many fields of view were used to quantitate, nor how many cells per field were counted for each. There must not be many senescent cells if you can cut them by 40% with no change to total cells. Or is ETTA1 a B-gal inhibitor instead, and it just interferes with the assay instead of clearing senescent cells? Also not clear on E50, dosing, solubility, binding target, etc. Presumably some of that is in the slide deck?

The assay was performed by inducing senescence in cells with 20 µM etoposide. The initial cells were all from the same biological source, but there should be some biological variation in the DNA breaks caused by etoposide that leads to senescence in each well. The assay was performed twice: once with fibroblasts and once with kidney cells, but the fibroblasts had little induction of senescence.

I repeated the analysis using ANOVA as requested. The only difference was the number of senescent cells with fisetin treatment was no longer significant (p=0.147), while ETTA1 treatment (p=0.0051) and the FDA approved drug treatment (p=0.0115) remained significant. Normalized cell numbers are used because the number of cells in the field imaged in the center of the plate (1 large field imaged per well) can be variable, likely due to cells moving to the sides in some of the wells due to imperfections in the plate. We also chose to use total cell number rather than normal cell number because this approach was used previously for identifying fisetin as a senotherapeutic. We did observe a significant increase in the normal cell number for ETTA1, the FDA approved drug, and fisetin. This result fits with the hypothesis and senolytic in vivo results that removing senescent kidney cells leads to kidney cell regeneration and proliferation. This result is likely due to secreted senolytic factors inhibiting normal kidney cell proliferation. The increase in normal cells is also why the number of total cells did not change. As listed in the experimental plan, we will use p21 antibody staining to ensure our lead molecules are not just B-gal inhibitors in vivo. p16 and p21 reduction was seen in molecules from our training dataset, so we also expect genuine senescent cell reduction. The EC50 will be determined for all the lead molecules with the rest of the synthesized molecules after validation for direct comparison, as stated in the experimental plan. We expect we may have to use some formulation for solubility, which will build on using our previous work formulating fisetin nanoparticles. Our hypothesis is these molecules are dual inhibitors of mTOR and PI3K/AKT, similar fisetin and related molecules. The optimal dose will be determined in vivo.

I’m generally skeptical of AI/ML approaches in bioscience. Even beyond my normal skepticism, 108 molecules seems small for an AI/ML approach, especially when there are screens of 10-50,000+ small molecules available for purchase (and have been thrown at the mTor pathway). The negative data are likely as important as the positive data.

I was at the Stanford Drug Discovery Symposium 2 weeks ago, and now all major pharmaceutical companies are using AI/ML at every stage in development. These approaches are likely the reason we are seeing an increase from ~10% to ~30% in clinical trial success. I agree with your skepticism as I have previously shared similar thoughts on in silico approaches that simply did not work, but this fact has been changing quickly over the past few years.

We also thought the 108 molecules may be too small for a training data set, which is a reason the platform validation (Fig. 1) was important for us. We were able to successfully identify a hit in ETTA1, and as we find more hits, the training data set will become deeper. We can also take advantage of the findings from other publications to integrate into our training data as they emerge. We were not surprised we had 2 negative hits in ETTA2 and ETTA3 due to the limited training data, although they trended in the correct direction, which may be promising. For senolytics, the main concern has been safety, which is where AI approaches excel because safety data is available for thousands to hundreds of thousands of molecules, which is a main reason why we prefer an AI approach over a traditional screening approach.

One thing that further dampens enthusiasm is it sounds like the AI/ML platform has not predicted alternatives to the FDA-approved drug yet. Would like to see a couple leads in hand for that to test how well the AI/ML platform works. For the de novo leads, it is unclear how many leads they plan to make/screen in total. Out of the top 3 candidates, 2 failed already. Is the plan 2 more? 10 more? 100 more?

We are creating derivatives of the FDA approved drug in silico and retesting them on the platform to produce alternatives. We will create derivatives of ETTA1 in the same process to improve the results. We also have 10 additional candidates for synthesis, which should produce further hits.

It is unclear what expertise the team has with kidney disease or medicinal chemistry. I don’t see any first/last author papers in the references from the team, which is another expertise concern. Is this a new area they are moving into? Advisors are either early career faculty or non-scientists, which is also concerning. What if the syntheses for new leads are hard?

I worked on kidney decline and senescent cell identification at Rejuvenation Technologies (mRNA nanoparticles to extend telomeres) because telomere shortening is related to organ function and senescence. I also worked on the kidney briefly during my postdoc in the Wyss-Coray lab. Lurong has expertise in structure-based drug design for small molecules and excels at producing improved versions of existing molecules. We are actively recruiting experts in kidney disease and medchem. I am working with Enamine on synthesis, which has already confirmed synthesis routes for 10 of our top leads.

The IP is for fisetin delivery, yet there is nothing about fisetin delivery in this proposal or how that relates to this proposal. Will the company be distracted by advancing its fisetin delivery method and/or trying to split between formulating liposomes for fisetin delivery vs developing new compounds?

Our focus is on the novel molecules, which we will patent after animal studies to maximize the patent life. We will use our knowledge in fisetin nanoparticle formulation to assist in our novel molecule formulation. We previously used the fisetin nanoparticles as a proof of concept that this approach would be safe and potentially effective in humans, but our focus at ETTA is only on the novel molecules.

[bowtiedshrike]

@kyle

The assay was performed twice: once with fibroblasts and once with kidney cells, but the fibroblasts had little induction of senescence.

…This is an n of 1 experiment, then.

Are you doing stats on your technical repeats that have ~10% intraassay variation (even after normalizing to cell number)? You need 2 more independent repeats in the kidney cells before statistical analysis is valid. That’s also important to demonstrate you can reproduce it in your own hands.

If the number of cells varies so much that you have to normalize, it’s hard to say if your effect is from your drug, or random plating artifacts. How many senescent cells are you basing your results on?

What controls did you do to validate your etoposide treatment induced senescence rather than apoptosis? How many times did you do those controls? How did you rule out ETTA1 effects on apoptotic cells instead of senescent cells?

Do you have data validating ETTA1 targeting mTor and PI3K pathways, or is that assumed based on structure/AI?

[kyle]

Are you doing stats on your technical repeats that have ~10% intraassay variation (even after normalizing to cell number)? You need 2 more independent repeats in the kidney cells before statistical analysis is valid. That’s also important to demonstrate you can reproduce it in your own hands.

I agree. We will conduct repeats, add in comparisons with newly synthesized molecules, and perform EC50 measurements on potential hits for independent repeats. The repeats often have >10% intraassay variation. This variation is expected as they are not pure technical replicates and the data are be more noisy than other assays due to the low FITC fluorescence with C12FDG in our hands. I typically have 2 to 5% intraassay variation due to pipetting errors.

If the number of cells varies so much that you have to normalize, it’s hard to say if your effect is from your drug, or random plating artifacts. How many senescent cells are you basing your results on?

We detect an average of 88 senescent cells per well for the DMSO control.

What controls did you do to validate your etoposide treatment induced senescence rather than apoptosis? How many times did you do those controls? How did you rule out ETTA1 effects on apoptotic cells instead of senescent cells?

We used C12FDG, a substrate that fluoresces when cleaved by the SA-Beta-GAL activity in senescent cells. C12FDG does not fluoresce in apoptotic cells to our knowledge. We repeated C12FDG in all wells, but also used wells where either etoposide or C12FDG was not added as negative controls to confirm specific detection of C12FDG cleavage in senescent cells.

Do you have data validating ETTA1 targeting mTor and PI3K pathways, or is that assumed based on structure/AI?

We are planning to obtain this data with in vivo results. These data are nice to have and would be excellent to show, but they are not critical to clinical or commercial success. To assess mTOR inhibition, we will look at inflammatory cytokines after treatment, as well as phosphorylated mTOR and downstream phosphorylated S6K1 levels. To assess insulin signaling, we will perform glucose tolerance testing and determine levels of PI3K and phosphorylated AKT. Traditional structural attempts to target these pathways have been challenging, due to the complexity of the mTOR complex 1 and complex 2 and due to safety problems seen with other molecules targeting the insulin signaling pathway. The targeting of these pathways is hypothesized due to data from molecules in our training dataset, such fisetin being a dual inhibitor of mTOR and PI3K/Akt. The inhibition of these pathways is also the proposed mechanism of action (see slide 7 in the deck), which is why we think this activity is preserved in ETTA1.

[Paolo]

Hello, I have just edited the proposal with the outcome of the senior review. For those who have already voted, feel free to update your vote as needed. For those who have not yet voted, please kindly do so. The poll will close in one week from now, provided we have at least 10 votes on either side.

[Abhijit]

The proposal is interesting, and the senescence/senolytic approach is very well characterized in the aging field. I have mixed feeling about the approach and methodologies used.

1. What cells were used for the assay in figure-1? Human or mouse? Also, were they fibroblasts, epithelial cells, endothelial cells, or a mix of all? Just wondering what cells were used for these assays.

2. Does ETTA-1 kill senescent cells induced by other means, such as doxorubicin or irradiation, etc., or specific for only etoposide-induced senescence?

3. This has been asked before, but I do not see the answer. What controls are there for etoposide-induced apoptosis and senescence?

4. C12FDG measurements alone are not an ideal marker for senescent cells, and it is generally used in conjunction with other markers such as changes in p16, p21, IL-6, etc. Most of the cited paper also shows that because senescence is always characterized by multiple markers. I do not see any of the additional markers.

5. In the CKI-aged mouse model, p21 staining is the only marker that will be used (Experimental plan: Kidney Disease Model in Aged Mice). Are there any plans to see other markers of senescent cells, such as changes in p16, Lamin-B1, or other secreted cytokines, SA Beta gal, etc? This will at least attest to the senolytic activity of ETTA-1.

[bowtiedshrike]

Surprised that the comments were harsher than the quantitative score. If that was an NIH grant, those are the comments you see after your grant got triaged.