Modeling harmony: learning from preclinical coordination of COVID-19 countermeasures
How COVID-19 consortia are harmonizing preclinical testing of antivirals, mAbs and vaccines
The all-hands-on-deck COVID-19 response provides a glimpse of what’s possible when drug developers converge around solving a problem. One of the most important developments might be in preclinical research where consensus on harmonized models and protocols would enable apples-to-apples comparisons before candidates start costly journeys through the clinic.
Most of the attention has gone to clinical trials, where master protocols represent the vanguard of harmonization, allowing treatments to be more quickly and efficiently assessed than standard clinical trials (see “Master Protocols Present Systematic Path for COVID-19”).
But behind the repurposed compounds in the clinic lie nearly 400 therapeutic and vaccine programs in earlier stages of development, many of which are designed for COVID-19 and may deliver a higher level of efficacy than the first wave of clinical candidates.
For this pandemic, the urgency is to avoid the mess of preclinical models that differ by lab, and even within labs, and have a famously poor record of translating well in the clinic.
Instead, systematic lab experiments using the same protocols, readouts and controls will be needed to quickly home in on the most promising compounds to advance to human testing.
“Hundreds of companies and institutions want to help, but it’ll be a waste of money to invest in non-replicatable and non-industrializable data,” said Evotec SE (Xetra:EVT) CEO Werner Lanthaler. An even bigger concern is that without robust ways to evaluate and compare compounds, good candidates might be edged out or missed. “That’s the worst opportunity cost that we all can incur in this industry.”
In addition to the collective benefit, individual companies stand to gain from independent confirmation of compound performance, as well as collective pressure-testing of models, assays and correlates of efficacy.
The system-wide benefits should carry forward at least to the next outbreak, and ideally to drug development more broadly.
“This goes way beyond COVID-19.”
At least three large consortia have committed to systematically triaging preclinical COVID-19 candidates: COVID R&D, which includes R&D leaders from at least 15 major pharma companies; Accelerating COVID-19 Therapeutic Interventions and Vaccines (ACTIV), organized by NIH and FNIH; and the COVID-19 Therapeutics Accelerator (CTA), launched by the Bill & Melinda Gates Foundation, the Wellcome Trust and Novartis AG (NYSE:NVS; SIX:NOVN) (see “Sorting Out COVID-19 Consortia”).
Some of these programs are bulking up frameworks established during the HIV, Ebola or Zika epidemics.
“There was never a coordination to this level, where we would be using a master protocol and harmonizing as much as we can across centers with shared controls,” said Jay Rappaport, a leader of ACTIV’s non-human primate work and director and Chief Academic Officer of Tulane University’s National Primate Research Center.
“The structures, relationships, SOPs and coordination we’re building now will serve us very well going forward into the next epidemic,” he added.
A key question is whether harmonization programs driven by COVID-19 will be successful enough to spur replication in other disease areas, and whether getting drug developers to play along without the backdrop of a global pandemic will require gating access to resources such as funds, model systems or patient data.
“This goes way beyond COVID-19,” said Lanthaler. “Preclinical research needs to go beyond anecdotal observations into true translational data.”
Making harmonization routine will require careful attention to logistics; for example, working out when to blind data, and making sure to stay away from discussions of contracting or pricing that would create legal issues.
But a bigger hurdle is that the more uncertainty there is about the biology the harder it is to harmonize R&D.
“Very early in a process, there shouldn’t be harmonization, because nobody knows what to harmonize around.”
For COVID-19, the most straightforward area is in small molecules and mAbs that target the SARS-CoV-2 life cycle, which have a relatively narrow mechanistic scope. A harder task is to harmonize assays for vaccines, which mobilize host immunity against the virus via a process that is less crisply understood.
An even bigger hurdle is harmonizing assays for therapies that target host pathologies underlying severe COVID-19, such as cytokine storms and blood clots. For these aspects of the disease, relevant natural histories and molecular targets are still unclear, said Evotec EVP of infectious diseases Kara Carter.
“Very early in a process, there shouldn’t be harmonization, because nobody knows what to harmonize around,” said Mathai Mammen, global head of R&D for the Janssen Pharmaceutical Cos. unit of Johnson & Johnson (NYSE:JNJ) and coordinator of COVID R&D’s preventative vaccines working group.
Assembling the choir
A handful of companies, institutes and labs have emerged as go-to sources for harmonizing specific types of experiments. This center of excellence model could be a template for preclinical harmonization work beyond COVID-19.
The Rega Institute at KU Leuven and the Calibr drug development division of Scripps Research have each established high throughput screening capabilities to identify small molecules with antiviral activity against SARS-CoV-2 amidst thousands of repurposing and other candidates.
Evotec has a master service agreement with Rega to test compounds from COVID R&D, in which Lanthaler leads the preclinical repurposing group. Separately, the institute is partnering on COVID-19 research with J&J and Biomedical Advanced Research and Development Authority (BARDA) and has additional funding from the Gates Foundation.
The Calibr program is part of the Gates Foundation’s CTA. Calibr VP of Medicinal Chemistry Arnab Chatterjee said the institute develops antiviral activity assays; builds, screens and shares its ReFRAME library of over 14,000 compounds previously tested in humans; and screens candidates from other companies or institutions.
Other groups are also sharing core competencies; for example, the Novartis Institutes for BioMedical Research (NIBR) has established itself as a resource to characterize inhibitors of the SARS-CoV-2 3C-like protease (see “Pharmas Align Behind Crowdsourced Solution”).
“The structures, relationships, SOPs and coordination we’re building now will serve us very well going forward into the next epidemic.”
The Gates CTA is also funding the Coronavirus Immunotherapeutics Consortium (CoVIC), which is both serving as a clearinghouse for preclinical characterization of mAbs against SARS-CoV-2, and developing virus-targeting mAb cocktails with product profiles well-suited for low- and middle-income countries in terms of COGS, manufacturing and delivery. The consortium structure is based on a prior group that developed and characterized mAbs against Ebola.
Within CoVIC, different academic labs and CROs conduct assays corresponding to their expertise.
For example, structural studies characterizing the interaction between candidate mAbs and the spike protein are conducted in the lab of CoVIC leader and La Jolla Institute for Immunology professor Erica Ollman Saphire, while biophysical binding assays are performed in labs at Duke University and Carterra LSA.
Ollman Saphire said the consortium can screen on the order of a few hundred mAbs, noting that the targeted nature of mAb discovery means it can afford a lower-throughput approach than small molecule library screening. “Maybe each of these companies or academic labs made 500 antibodies and they’re sending us their 10 favorites,” she said.
On the vaccine front, Mammen is encouraging companies within the COVID R&D consortium to share data on immune responses to their vaccine candidates. He also wants them to conduct harmonized neutralizing antibody assays, either by running them in the same lab, or by having different labs run the same protocol and incorporate multiple standards that can be mapped back to each other.
He argues that even companies with programs already in the clinic would benefit from a larger corpus of harmonized data that could help answer questions such as how much and what kinds of antibodies are correlated with protection. That information would allow companies to reduce dosing and make better use of their manufacturing capacity.
Under ACTIV’s vaccine program, the seven NIH-sponsored non-human primate centers are coordinating master protocols to test multiple candidates in parallel, with the goal of directly comparing results while minimizing the number of animals used.
Rappaport told BioCentury that the group expects guidance from ACTIV on what vaccine candidates to test this fall; it’s also considering additional candidates for the study based on performance in small animal models, or on how well a particular vaccine platform has performed in the past.
For antivirals, it may boil down to a handful of assays applied on different compound types.
Evotec’s program with the Rega Institute first screens compounds for their ability to prevent SARS-CoV-2 from killing green fluorescent protein (GFP)-expressing Vero cells, an African green monkey-derived kidney epithelial cell line that can’t produce interferons, which makes it more susceptible to viral infections.
Another high-throughput antiviral screening assay from Sanford Burnham Prebys Medical Discovery Institute also uses Vero cells, but quantifies cell killing and rescue by measuring cellular ATP levels.
The institute published an April biorxiv preprint in collaboration with Calibr researchers and others in which the assay was used to screen about 12,000 compounds, including candidates from the ReFRAME library.
Calibr published its own primary screening strategy in a biorxiv preprint on June 16, which uses human HeLa cells engineered to express ACE2, the host receptor that the SARS-CoV-2 spike protein binds to mediate viral entry into cells. The assay measures cell viability via a fluorescent dye and also quantifies viral entry, using antibodies from convalescent patient plasma to bind the virus, and then labeling those antibodies with a fluorescent signal.
Chatterjee said the dual readout “allows us to have really good signal to noise,” giving Calibr a more fine-grained quantification of a compound’s potency, and to tease out synergistic effects with lower doses of remdesivir. Based on the assay, the institute narrowed the list to around 70 repurposed compounds, and nominated 26 for further investigation based on known PK and safety profiles in humans.
Both Chatterjee and Evotec’s Carter said human airway epithelial cells are a useful secondary model system for triaging antiviral hits from primary screens. Other emerging models for testing antiviral compounds include human-derived organoid and lung-on-a-chip platforms.
The relevance of animal models for understanding the activity of antiviral compounds is still unclear. Chatterjee said the combination of in vitro antiviral activity and past human PK and safety data would be enough to get a compound into the clinic. “We’d love to have that animal pharmacology piece in between, but if it doesn’t come in, we don’t see that as rate-limiting.”
Agents of immunity
Immunity is emerging as one of the big stories in COVID-19, with the unfolding pathology centering on how the virus interacts with the immune system in the early stage of the disease, and then fires it up in the late stage to lead to acute respiratory distress syndrome (ARDS), often the cause of death.
A key focus among drug developers is to create mAbs or vaccines that arrest or prevent the disease in its earliest stages.
Assaying the two types of candidates in parallel enables researchers to identify patterns that uncover the way SARS-CoV-2 interacts with the immune system, creating a knowledge base that could inform development of next-generation compounds.
Ollman Saphire said CoVIC evaluates how different mAbs bind to multiple versions of the spike protein -- including those corresponding to different virus strains, isolated protein domains, and soluble versus membrane-bound forms of the receptor. It then uses high-resolution epitope binning to determine each mAb’s binding footprint.
She said that enables her team to find combinations of mAbs that could prevent the emergence of resistance by covering complementary viral epitopes, and to have well-characterized alternative candidates on deck in case the virus mutates in a way that eliminates a first-wave mAb’s binding site.
The consortium will also measure in vitro neutralization of pseudoviruses expressing the spike protein and SARS-CoV-2, and in vivo protection in Syrian golden hamsters, mice expressing a human ACE2 transgene, and aged wild-type mice. CoVIC will also study interactions between the Fc domain of mAbs and immune cells, and screen out antibodies that enhance viral cell entry.
According to Mammen, hamsters and rhesus macaques are emerging as the most promising models for vaccination, and those models are starting to be run in convergent ways. “I would encourage anyone testing vaccines going forward to be using those models, but I couldn’t have said that a month or two ago.”
Rappaport agreed that rhesus macaques are the model of choice for testing vaccines, partly because they exhibit a mild version of COVID-19 that resembles the common human form of the disease. They are also more widely available than similar models such as cynomolgus monkey.
He said the non-human primate centers in the ACTIV consortium will harmonize granular details of their macaque studies, including the time points for sample collection, serological virus neutralization assays, protocols for virus challenge, and flow cytometry-based surveys of B cell, T cell and innate immune responses to vaccine candidates, down to the specific fluorophores used to measure proteins on cells.
Mammen said companies would benefit from going beyond the serological assays that are traditionally the focus of vaccine studies to explore T cell responses, including sophisticated assays that have become mainstays of immuno-oncology. The goal is to determine the link between T cell responses and the potency and duration of a vaccine’s protection.
At least two companies are developing T cell-based tests to monitor patient infections and inform vaccine design (see “T cell Tests Aim for Broader Picture of COVID-19 Immunity”).
New disease models are also coming online for severe COVID-19, which could eventually offer a way to harmonize testing of compounds that target host pathology.
In a June 19 biorxiv preprint, Rappaport’s team presented data suggesting aged African green monkeys could serve as a model of ARDS and cytokine storm, which arise in severe COVID-19.
Another recently published animal model of severe COVID-19 involves adenoviral vector delivery of a human ACE2 trasngene into the lungs of mice; the model was described in Cell June 10.
ACE2 - Angiotensin-converting enzyme 2
SARS-CoV-2 3CLpro (SARS-CoV-2 NSP5; SARS-CoV-2 Mpro; SARS-CoV-2 main protease) - SARS-CoV-2 3C-like protease
SARS-CoV-2 S - SARS-CoV-2 spike protein