Epidemiologically speaking, we are all bats now.

(Prev article #1 of 5) Note: this is the 2nd of 5 articles in our series on flavonoids. (Next article #3 of 5)

Leschenault’s Rousette bat, of Chiroptera Pteropodidae

Update Jan 2021: This article documents what to date remains an inconclusive result based on the limited data available in the literature.

So it’s been an intense week of studying and research on hesperidin and COVID-19. But in the process of trying to find longitudinal datasets for demonstrating a de-facto in vivo efficacy test in order to support *actual* in vivo testing, we offer a new, hitherto unexplored hypothesis for why bats are a reservoir for coronaviruses.

Basically, in China’s Yunnan province where most researchers acknowledge this coronavirus likely originally came from, you have a lot of bats, and a lot of intermediary hosts that were shown to host coronavirus after the 2003 SARS epidemic. These are your civets, racoon dogs, and similar tropical forest-dwelling mammalian creatures. The bats, however, are 1. frugivores —( Yup, nothing but fruits and a couple leaves here in there in their diet) and 2. insectivores (insects also eat fruit as transient members of the food chain on up to bats) . Given that hesperidin is citrus fruit-derived, we began looking for source data that could shed light.

Where did we find the data? Well, in 2009–2016, a team including famous ‘bat woman’ Shi Zheng-Li herself catalogued and published all the strains of betacoronavirus they could isolate from multiple species of fruit bats in southern China. One species figured prominently in their study’s resulting viral clade catalogue — the Leschenault’s Rousette (pictured above).

Given our hesperidin focus, we were interested in what these bats were eating — surely no researchers would have squandered their valuable hours studying such a mundane thing? Luckily, some did. And that was the critical dataset we needed for this meta-analysis.

Although the tab-by-tab sheet progression of the data sourcing and logic is right here (since the core data was necessarily tabular so . . . spreadsheet) — with this background already provided you can happily skip it — here’s the central graphic from it:

Species hops on depletion uniquely of hesperidin-containing fruits

In these forests, hesperidin-containing fruits effectively start falling ‘out of season’ in September, and are totally gone by the end of October. That is to say, all of the mammals that participate in the hesperidin food chain simply no longer have it in their systems by the end of October. (In humans at least hesperidin has a metabolic half-life of 24 hrs).

And it turns out that humans aren’t exempt from the lack of availability of hesperidin-containing fruits. Why should they? I’ve visited Yunnan firsthand as well as its preserve Xinshuangban’na where the bat fruit study took place. It’s not like Beijing or Shenzhen — there are no major supermarkets there for the local wildlife traders to pull down packaged fruit juice off the shelf; — outside of the basic staples supplied by small sundries shops, locals gonna eat what locals gonna find around them.

In both COVID-19 and the 2003 SARS epidemic, the species-jump to humans took place in early November of their respective years. When was hesperidin out of the ecosystem completely? Early November. Not September, not December, but early November. And this isn’t just ordinarily seasonality at play which correlates with all sorts of phenomena on Earth — because the ecosystem still has plenty of access to hesperidin-free (update 2 July: low-flavonoid content rather than just hesperidin-free— e.g ficus racemosa just 1% quercetin-equivalent by dry weight) figs well into and past October.

And so, wham, you have the beginnings of an evidentiary framework for understanding the seasonality of coronaviruses.

When we look for examples of seasonality from the literature, we do indeed find it in the same famous study that identified the SARS cave. And on the ecosystem-wide fruit availability side, it’s clear that much of Yunnan fruit agriculture has a binary seasonality characteristic — a wet season when there is fruiting, and a dry season when there is not. Taking local natural mango cultivation as a proxy for general fruit availability and adjoining with the bat viral load seasonality data from Hu et. al, we have:

which when combined graphically yield:

A clear anticorrelation of fruit availability to the ecosystem and bat viral loads. Note that the bats in Hu et al are horseshoe bats — and so insectivorous. How then could they possibly be affected by fruit availability? Simple. Insects eat fruit, and when you look at what compounds are available to an animal, you look at the entire food chain leading to them, not just their proximate food sources.

It might be posed that two phenomena in nature featuring correlations to each other (the viral load and fruit availability) could simply be chalked up to nondescript seasonality with many possible unrelated confounding factors. However, note that we hypothesized in advance about viral load seasonality vs fruit consumption based on the more tenuous results from the Xishuangbanna study. Having found the result we were looking for only a posteriori, yields it as a non-zero evidentiary signal.

Update 2 Jul 2020:

A lot of in silico studies have taken place since the first version of this article was published. We and others have identified more flavonoids with inhibitory efficacy as protease inhibitors comparable in strength to hesperidin. So we need not restrict the above ecosystem analysis to hesperidin only, but largely any flavonoid.

The ‘flavonoid hypothesis’ would open up a lot of new avenues in zoonotic epidemiological research enquiry, even while it starts to resolve long-standing problems in betacoroanivirus virology (the super-family of the viruses plaguing us now and in 2003) — specifically, why is there an animal reservoir that the bats and other animals don’t develop herd immunity to? That should ordinarily happen if we’re talking about a semi-closed fauna ecosystem. The flavonoid hypothesis offers a resolution to this question — by acting as an antiviral, it is acting as a ‘crutch’ to the bats’ and small mammal’s immune systems. Indeed, one retrospective paper cataloguing viral reconnaissance after the 2003 SARS epidemic identified a civet that tested positive for coronavirus, but tested negative for coronavirus antibodies — a suggestion that the civet’s immune system couldn’t be bothered with launching a counterattack to the coronavirus infection, rather letting flavonoids in the local fruit eaten by civets (mangoes, in that species’ local case), fulfill its role.

And so the coronaviruses recirculate in the tropical forests’ bat and mammalian populations, going on millions of years, with no sign of stopping anytime soon.

Verification of this model will open up a particularly dicey question for public health policy in the wake of COVID-19. If this model is verified, then sure, we all keep our OJ, mango juice, and other flavonoid supplements up, and so keep the virus at bay, just like it is in the bat and civet populations. But then we won’t develop immunity, particularly herd immunity to it. Because, a year from now when the vaccines come out and only 10% of people actually get a vaccine that costs some money and a trip to the hospital, then for those who just kept drinking prophylactic orange juice, once they stop, they’ll still be vulnerable to coronavirus transmission from others who similarly fell lax on their hesp upkeep — — just like the bats in the coronavirus-season. And so that is why, epidemiologically speaking, we are all bats now, and we better figure out what we’re going to do about that.

Maybe, it will be possible for pharmacologist / virologist teams to determine how much flavonoid concentration will be enough to keep the virus at bay, but not enough to keep the immune system from developing its antibody response. Then flavonoids (and lack thereof) will have effectively been both the originating enabler of, and the treatment for, the COVID-19 epidemiological outbreak.

Update 31 Dec 2020:

Since the initial posting, I’ve been able to deep-dive into bat cov seasonality in China by scouring the bat cov research literature out of China (as well as Southeast Asia). The primary finding is:

  • The above bat cov seasonality data is too sparse to draw conclusions from. What’s more, the totality of available data (whose comprehensiveness was corroborated against that independently collected by another individual’s dataset collecting similar data albeit for different reasons) is still too sparse for any seasonality-related conclusions.

I had divided up the available bat sampling data for the region into 2 dimensions — insectivorous bats vs frugivorous bats, and tropical vs subtropical. This is because

  1. the provinces of interest for bat cov recombination in China are primarily characterized by these two climate types.
  2. In the above outlined food-borne coronavirus viral load reducing model, insectivorous and frugivorous bats are expected to have somewhat different seasonality of infection rates which we’re treating here as a proxy to viral load (which comes with its own caveats). Remember that in the insectivorous case, bats eat insects that eat plant matter (including flavonoids but also a lot of other compounds) outlined above.

In practice, one finds that frugivorous bats are more commonly identified & sampled in tropical regions and insectivorous were identified & sampled more in subtropical regions. So the two charts of relevance are shown below (the remaining two quadrants would have little to no data to show here).

Insectivorous bat infection positivity rate, Yunnan, Guizhou, Guangxi subtropical regions
Frugivorous bat tropical regions.

The richest data available is for insectivorous bats in the subtropical region. But considering there are no sampling points from Jan-Mar, nor June, nor Nov -Dec, (to say nothing of having same across multiple years ideally), then the available data in the bat sampling literature to date is simply inconclusive (i.e. neither sufficiently supportive nor sufficiently contradictory) as to whether there is seasonality in bat coronavirus positivity rates that could be connected with fruiting seasons (or to any other phenomena for that matter).

Expanding the lit review to Thailand, once finds more comprehensively collected bat cov data, but still too sparse to draw conclusions — for example data taken weekly over the course of 1 year might not reflect well a subsequent year’s data collection if it would ever be performed.

Frugivorous bat cov seasonality in Thailand.
Insectivorous bat cov seasonality in Thailand

So more source data of the bat fecal sampling from caves variety needs to be generated by bat reservoir researchers, period.

For any bat researchers (and their funders!) reading this, I would pose these requests, whether studying bat cov’s or other of the many viruses that bats host:

  1. Prioritize data collection scheduling & protocol toward seasonality detection of viral load, be it annual to super-annual or otherwise. It’s not all about phylogeny-mapping — especially considering that nsp’s have less selection pressure on them than spikes. Any-cause general seasonality is a worthwhile feature of bat viral load detection to study in its own right.
  2. Please avoid pooling samples (!). 1 bat at 1 time = 1 sample. Get more funding if necessary to support any additional effort / supplies / testing capability required.
  3. (NB: The above hypothesis doesn’t at all need to bought-into to recognize that the conditions underlying it are in any case, and at all times to varying degrees, present and should be taken into account during sample collection and assaying:
    Try to unify around a constant and robust methodology in PCR testing — Fecal samples will contain polyphenolic compounds, typically from plant sources upstream in the food chain. Polyphenolic compounds (such as the flavonoids focused on in this post but also many more such as tannins) are also known to interfere with PCR assay results (see linked comment section), and there are ways of controlling for this such as a parallel PCR of a control amplicon placed in the same fecal matter.

Source data and analysis for this post including all the 24 papers used to generate the source data: https://docs.google.com/spreadsheets/d/1jw3BJyE22ZHOfHys9clr7RdGQuGrIry6OEgqx6Vdn2o/edit#gid=206174651
If you use the processed data, kindly credit by cit this post (31/12/2020 ) in addition to the underlying researchers’ collection data papers.

Coming from a multidisciplinary MIT technical background and startups in San Francisco & SE Asia, Rick leads @EMSKEPhyto . linkedin.com/in/rick-phyto