Will general antiviral protocols always be science fiction?
Society relies entirely on medicinal chemists to pre-select new drug candidates for clinical trialing. Could they be overlooking a selectivity mechanism for antivirals hiding in plain sight?
(Short on time? Head to the in-a-nutshell summary at the bottom of the article☺)
December 17th, 2032:
Following an exhausting red-eye flight across the country to visit with family over the winter holiday, you get off the plane and amble over to the terminal’s rideshare pickup area. A couple days later, you’re coughing mildly. You go to the drugstore. They swab you and sequence whatever bug you got. It turns out to be this year’s strain of influenza-A virus — the flu. Your viral load was subsequently quantified from the same swab overnight; The results come back the next day: Just 10⁴ (ten thousand) viral RNA copies per mL, nothing so strong (yet). The pharmacist asks if you’d like take-home treatment. Eager to avoid a week being laid up alone instead of time with the family, you gladly accept.
The pharmacist looks up which among the several hundred pre-cleared flavonoids (commonly-encountered compounds from plants) were most recently verified for effective antiviral activity against this year’s rhinovirus strain. This thanks to a newly expedited approval process ratified by the FDA’s board, where supplements with a long-familiar and highly tolerant human safety profile are verified for efficacy in animals and clinical trial volunteers who caught the same bug just months before you did.
She swabs you once more before handing you a bottle full of capsules of the relevant flavonoid, a pack of disposable vials of sample buffer, long cotton swabs, and a QR code for an app to download that will beep a reminder every four hours to take your next dose. Per the pharmacist’s instructions, you begin your flavonoid therapy while producing a fresh swab sample every day. By day 3, your symptoms no longer persist. You drop the samples off at the pharmacist who emails you a report the next day.
‘Your 2nd sample before starting the therapy was 10⁶ (one million) viral RNA copies per mL — a growing viral load. Your subsequent 3 home samples after commencing therapy clocked in at 10⁴ (ten thousand), 10³ (one thousand), and 10² (one hundred) copies per mL, consecutively. Your viral load is successfully receding to undetectable levels. Please continue your therapy to completion.’
To avoid experiencing the oft-reported viral rebound, you continue taking your flavonoid therapy for the rest of the week. Afterward you go on to enjoy a rejuvenating Christmas dinner with family.
Of course this scenario remains a distant fantasy; Even as antibiotics are easily administered today against infections from a wide range of bacterial species, we still live in a world sorely lacking effective protocols against common virus species. Rare exceptions to this reality are rightfully celebrated as well-known feats of enormously hard-won scientific investment: Paxlovid for COVID-19, inhibitors for HIV and hepatitis-C spanning years of R&D, trials, and agency approvals. Could the fictional account describing an off-the-shelf general antiviral protocol ever become a reality? To answer that, we’ll need some context:
Whenever headlines pop announcing ‘researchers find molecule X inhibits the coronavirus’, contrarians will reflexively point to the famous 2013 “petri dish” comic from XKCD. Because it vividly reminds us that not only must a compound make its way through the body’s metabolism to the inside of cells where it can disrupt the virus’ replicating, but it also must avoid disrupting healthy cells’ functioning. — this is called selectivity. The more selective a drug is, the more easily it is tolerated while still doing its job.
Flavonoids (nothing to do with flavor)
Before the pandemic, several research papers reported flavonoid inhibition of the SARS (from 2003) virus component called the main protease. In 2021, appreciating the structural similarity of the SARS protease to the pandemic’s SARS-CoV-2 protease, several stakeholders and I had set out to initiate a covid clinical trial leveraging a commonly consumed flavonoid derived from plants.
What is a flavonoid? Flavonoids are the most diverse of the three classes of natural compounds known as polyphenols. Flavonoids are produced in most plants and are widely acknowledged to serve an antimicrobial role among others. But over the past 30 years the plant biology literature has begun to recognize that flavonoids form a key part of plants’ defenses against plant viruses. We consume small quantities of different flavonoids every day when we eat fruits and vegetables. One tends to see the same flavonoids pop up in every paper researching them — diosmetin, hesperetin, quercetin, luteolin — yet there are a couple thousand flavonoid molecules catalogued.
While drafting our trial protocol for covid, I had to elucidate a crucial lingering unknown: How large should the dose be? Scouring flavonoid pharmacology papers to answer that question yielded a finding that begged attention. A more compelling opportunity auguring a much wider scope of impact than covid therapeutics alone was emerging: Many papers from the flavonoid pharmacology literature formed a picture that, evidently, there was a subtle yet promising selectivity mechanism in all the mammal species studied, including humans. If the mechanism was verified for the antiviral context, its value wouldn’t be restricted to just coronavirus work, but would apply to infections from many different families of viruses provided they met certain typical criteria.
While there were a couple of reviews of this mechanism’s literature base in recent years, no one had yet revisited & updated it for the pandemic context. Rather than add to the already substantial base of plant compound trial candidates for covid, I thought it would be more valuable to deep-dive into this putative selectivity mechanism. The goal? I wanted to rigorously map out & illustrate how each of the relevant citations supported the several legs that the mechanism stands on. It might make it easier for plant therapeutics researchers to justify trialing some of the plant compounds they investigate.
I brought my advisor in and we wrote a review manuscript in late 2021. In February of 2022, the preprint server chemrxiv accepted it for publication. The manuscript is now in peer review at a well-known pharmacology journal. So what’s it all about?
A sticky situation
Polyphenols have a reputation among medicinal chemists for being “sticky”. One med chemist even uses a more colorful metaphor to describe their interaction with proteins: “dog poop sticks to a blanket — any blanket”. My own longitudinal analysis of researchers’ bioassay contributions to the open public chemicals database PubChem (see the Appendix below) demonstrates that polyphenols’ reputation for promiscuous binding is indeed well-earned. Polyphenols are easily found to bind to between 5%-30% of all targets they are screened against — that’s a whole lot of distinct protein & enzyme species. That reputation however is then cited as a reason that polyphenols should be ignored from consideration as therapeutic leads for clinical investigation for any particular condition. Noted opinion-leader for industry med chemists, Derek Lowe, writes of the most frequently studied flavonoid, quercetin:
“It’s just that it [quercetin] shows up as a hit in so many assays, and has been linked to so many diseases and conditions. There are quite a few compounds in that category [polyphenols], and people have been burned by them in the past trying to get something to work. … Is one compound good for all of those? The odds are very much against it, and when you see so many varied results, you start to wonder if it’s really much good for anything in particular.”
The question is, does his conclusion follow from the premise? Are polyphenols really just promiscuous red herrings to be dismissed out of hand? To answer this question, we‘re going to look under the hood:
Hydroxyls are polar, exposing a hydrogen atom to form a hydrogen bond with electronegative atoms in protein/enzyme targets they happen to encounter. Hydrogen bonds are often employed in the med chemist’s toolbox to stick a ligand (a candidate drug molecule) to a target (an enzyme or protein of interest).
(Just one example of what some med chemists think may be happening when aggregating molecules such as polyphenols bind promiscuously to enzymes — the enzymes essentially stick, like lint to a ball of yarn, onto these accumulated balls of polyphenols bound together):
That’s not to say that aggregation is definitely the promiscuous binding mechanism that a polyphenol could exhibit, but it is one such observed mechanism. Another, for example, could be a compound with many rotatable bonds that can easily conform to a protein target’s surface, sticking at relevant points-of-contact thanks to its several hydroxyl groups.
Promiscuous, but selectively so
A toolbox of promiscuous compounds leaves us in a precarious position regarding polyphenols as antiviral candidates. Are they sticky and bind to everything indiscriminately? Or could they be sufficiently selective to avoid causing excessive trouble elsewhere in the body while hitting our intended targets of interest?
That takes us to the curious part. It turns out that, since the early 2000s, a widely dispersed band of researchers spanning Japan and Europe unexpectedly began to hit upon a mechanism that exhibited selectivity of all the polyphenolic molecules they investigated. The first researcher to strike upon this mechanism was Kayoko Shimoi, PhD at the University of Shizuoka in Japan. Like many natural product researchers, she was studying intestinal absorption of her polyphenol of choice, luteolin, in rats.
As is typical with other polyphenols, she verified that luteolin would see a glucuronic acid group added to it by the rat’s liver to prepare it for rapid elimination from the rat’s body. Pharmacologists generally recognize that once tagged with a glucuronide group, a molecule becomes so polar it becomes effectively impossible to passively enter cell membranes — a critical barrier to overcome for any would-be antiviral. This is because antiviral compounds must infiltrate through cellular membranes in order to act upon viruses replicating inside the cell.
However if the rats were subjected to inflammation, like in the body’s natural immune response to having proteins from bacterial cell walls injected into them, the luteolin would lose the glucuronic acid component, becoming reverted to its original & more potent, cell-penetrating form.
Taking the example she studied of the conversion of the flavonoid luteolin → luteolin glucuronide…
…you can see the glucuronic acid in the top-left of the right-side image replacing one of the hydrogens from the left-side image (the luteolin core molecule, aka the aglycone).
Other researchers followed up Shimoi’s work by studying similar polyphenolic flavonoids like quercetin. All in all, we catalogued 20 papers (and several reviews) that cumulatively added support toward verifying that this deglucuronidation-through-inflammation mechanism is in fact a real phenomenon.
As later researchers would go on to uncover, the secret star of this show is an enzyme each of us produces called β-glucuronidase. Specifically, the identification that common white blood cells called neutrophils and macrophages accrete to sites of inflammation, and dump payloads of β-glucuronidase uniquely at those sites. The β-glucuronidase inevitably runs up against the glucuronidated flavonoid, ejects the glucuronic acid part, and voila, the flavonoid aglycone can suddenly passively enter cell membranes. Now it’s in a position to promiscuously inhibit many of the virus particles it encounters inside the cells. At the same time, the aglycone can just-as-promiscuously inhibit the infected cell’s ordinary enzymes, like those used for respiration and transcription. This is desirable considering the cell has already been co-opted by the virus to produce more virions. And because the serum flavonoid remains glucuronidated outside the inflamed site, it leaves the healthy cells happily intact. In other words, selectivity at its finest.
While the entire mechanism put-together formally looks like this,
…it’s more easily comprehended graphically:
The graphic shows how quercetin loses its glucoside through the intestinal lining, going on to accept a glucuronic acid group before the liver releases it into the bloodstream. From there, in the case of a locally inflamed site such as from a viral infection, the glucuronide inevitably encounters the enzyme β-glucuronidase, losing its glucuronide and entering the infected cell’s membrane.
This chart here in pastel colors shows how the evidence base is supported by the literature (and for researchers in the field, probably serves as the paper’s most actionable component altogether):
It graphically maps out the citations in the Shimoi mechanism literature base, showing precisely which of the four legs of the hypothesis, (C, D, E, F, or combinations of these like CDE), each citation supports.
So perhaps there’s a mechanism — what can we do with it that we don’t know about already?
A shortlist of polyphenolic antiviral candidates
First and foremost, verification of the Shimoi selectivity mechanism reinforces the value of mining the extensive polyphenolic antiviral in vitro literature base to provide candidates for effective, well-tolerated antivirals. Researchers in polyphenolic antivirals are most likely correct in studying the aglycone in vitro as an antiviral rather than its glucuronide, so long as the virus under study induces the inflammatory response in vivo.
When researchers do antiviral screens in the lab with polyphenols (or with any other molecule for that matter), a critical step they inevitably undertake is to determine the IC50 (inhibitory concentration to reduce viral replication by 50%). Our survey of whole virus replication inhibition studies based on flavonoids yields this table of IC50 values:
And although there are many more studies that look at inhibiting the action of isolated components of a virus, such as a protease, polymerase, or spike protein, we disincluded them in this table. This is because we wanted to provide insight into the effects of all viral proteins & enzymes being inhibited by each study’s molecule, including any additional reduction in viral replication provided by inhibiting cellular processes that the virus had co-opted in infected cells.
In the laboratory, future researchers could expand this table to cover additional known viruses and viruses that emerge in future epidemics. In promising cases, they could go on to viral challenge studies in animal models (employing adequate biosafety protocols) & ultimately trials in humans who have caught the bug.
There’s another encouraging implication of Shimoi’s mechanism. While researchers produce IC50 values from laboratory antiviral screens, a critical step they also inevitably undertake is to determine the CC50 (the cytotoxic concentration, or cell-toxic dose that will kill 50% of healthy cells so administered) value. The higher that CC50 is a multiple of IC50, the better. 10X is a good starting point, and we would of course love to see 50X+ and beyond. That way, we can safely target trialing higher doses toward the IC90 or IC99 end of things (90% and 99% viral replication inhibition, respectively) without running headlong into the CC50.
This selectivity mechanism tells us that for polyphenol screens, the routinely determined CC50 value for the flavonoid aglycone is practically irrelevant to provide. Rather, the CC50 value of the glucuronide — not its corresponding aglycone — is more useful because that’s the form of the molecule that healthy cells will be exposed to. And happily, we have every reason to expect that the glucuronide’s CC50 value is much, much higher than the aglycone’s CC50, since the glucuronide can’t even pass through cell membranes into cells. This means we anticipate a much larger therapeutic window at the whole organism (animal & human) level to explore a therapeutic dose within than we would have otherwise inferred based on the aglycone CC50 alone.
Pharmacologically relevant polyphenol serum concentrations are challenging to achieve. Researchers are going to want to trial at high doses using animal models. The human equivalent dose of a couple thousand milligrams every four hours wouldn’t be an unexpected regimen, for example. A typical polyphenol’s time profile in serum resembles the below:
Note the sharp initial spike and fast decay. Clinicians will want a wide therapeutic window to work with so that the initial spike can be safely accommodated while maintaining a more prolonged IC90–99 concentration (at roughly the “shoulder” concentration in green below the spike) till it’s time for the next dose. A comfortably high glucuronide CC50 value, as indicated by the Shimoi mechanism, provides quantitative justification for more potent trial doses than those ordinarily administered in humans for other conditions (yet happily still far, far below TD50 toxicity values that are inferable from historical animal studies).
We can only speculate on why the Shimoi mechanism exists. And to be sure, accepting its very existence should require further end-to-end verification in multiple animal species, humans included. Perhaps it is simply incidental that mammalian metabolism relates otherwise independent phenomena together.
Personally, I suspect a unified evolutionary basis. We mammals go back a long time, and our common ancestors who made it through evolutionary bottlenecks like the K-T extinction were small, rodent-sized creatures. Herbivores among them would unavoidably have been munching down therapeutic doses of polyphenols in the plants they ate. When lucky individual immune systems evolved the mechanism (strictly speaking, when their white blood cells began to express β-glucuronidase — all the other mechanism components were already present) they would have found themselves more robust to contemporary virus infections — — and natural selection would have run its course. Maybe follow-on studies of the mechanism will ascertain that it’s not restricted to the Mammalia class, making it either much older or a case of convergent evolution.
Now focus turns toward building awareness for studies investigating therapies that exploit the putative Shimoi selectivity mechanism.
Facing would-be sponsors of any natural compound’s clinical trial is the challenge that natural products aren’t patentable in the US. In Europe, one can patent a natural compound for a particular illness if the results are verified, but then it’s a single-payer system for purchasing your offering — limiting upside. Since these are the markets pharma investors pay attention to the most, they will perceive a limited pay-off yet all of the financial risk & lead time that comes with running pharma studies & clinical trials. (If someone from the natural supplements industry would like to correct me, it would be welcome! :) ). Therefore my intuition is that it would have to be a philanthropic endeavor to support studies.
Suppose clinical trials repeatably show reduced viral load in animal and human clinical trials during a regimen with particular flavonoids. In that case, patients can reasonably be expected to become less sick — or even remain asymptomatic — until they clear their infection completely. The win at the end of it all would be a shortlist of well-tolerated plant compounds that can be indicated for viral infections emerging from a wide repertoire of unique virus species.
Then maybe the futuristic pharmacy scenario could one day become a reality.
If you’d like to encourage these organizing efforts toward driving this line of research past the finish line, you can Buy Me A Coffee. (To minimize conflicts-of-interest, I’ll return any donations from potential recipients of funds, including their institutional / professional advisory orgs). Progress updates are trackable at @EMSKEphyto on Twitter.
In a nutshell:
There may be a class of compounds readily accessible from the plant kingdom that can inhibit viral replication (much as paxlovid does for SARS-CoV-2), with very generous safety profiles owing to a selectivity mechanism characterized over the past two decades. Verification of several polyphenols exploiting the mechanism to show safe antiviral effect in infections from several different virus species would provide a strategic bank to draw from for future pandemics. During a future pandemic of an uncharacterized virus, so long as it induces an inflammatory response in human hosts, then the strategic bank of polyphenols of known safety profiles could be quickly assayed and trialed in in vivo studies to determine which among them is most effective for reducing that particular virus’ viral load in infected patients. Reduction of severe outcomes and faster recovery from infections could be reasonably anticipated.
This piece advocates for supporting a particular line of clinical research and nothing more. Emphatically, no one should read this and consider self-administering discussed compounds for the purposes described. More than just boilerplate, there are relevant compounds and corner cases I’ve not covered here that my conscience would really, sincerely prefer readers not do trial-and-error on.
The author would like to thank everyone acknowledged in the manuscript, stakeholders & supporters of last year’s proposed clinical trial, M. Hu PhD for technical consultation, and D. Stein, S. Molnar PhD, Dr. Angela Reiersen MD-PhD, S. Ferguson PhD, Yu Wai Chen PhD, and K. Spelman PhD for reviewing article drafts & encouraging feedback.
About the Author
Rick Sheridan cheerleads for researchers & clinicians laboring in underrecognized lines of research. He has served as an instructor on in silico modeling & baselining techniques for a plant research institute, and led drafting of the FLAVOCOV clinical trial protocol. By day he runs a venture in the agricultural industry.
Appendix for mol bio folks
- Here are more examples of flavonoid glucuronides from the pharmacology literature studied. You can note their similarities to the luteolin figure above — can you spot the glucuronic acid structure(s) in each?:
- As alluded to earlier, we also collected data from the literature and did original analysis on it to verify the med chemists’ premise. Yet, since the data collection was only secondary to the article’s central thesis (and anyways was in happy agreement with the preliminary aspect of the med chemists’ reasoning), we decided to remove that section and maintain as a pure review.
Of course, because this is a Medium post and not a paper for peer-review, I can provide all the original work I want☺ Here is that data:
- By sifting through several hundred thousand in vitro screens submitted to Pubchem (specifically the Pubchem bioassays database, as very helpfully pre-sorted by University of Bonn medicinal chemistry researchers), this data shows that polyphenols do indeed bind to targets promiscuously — much more so than natural product molecules that lack phenolic groups. (Interestingly, even single phenol molecules were as or more promiscuous on average as polyphenols). So the med chemists’ suspicions about promiscuity are actually well borne out here.
- Additional Phase II metabolic pathways for polyphenols have been documented like sulfation & methylation. Glucuronidation remains the dominant metabolic process however.
- Per the license terms, edits to the XKCD image are described: “cancer cells” struckthrough and changed to “infected”.