Please don't take Lumina's anticavity probiotic

It's unsafe and probably ineffective

Note: in terms of edits, please see this post for more details.

I like to think of the Bay Area intellectual culture as the equivalent of the Vogons’ in Hitchhiker’s Guide to the Galaxy. The Vogons, if you don’t remember, are an alien species who demolish Earth to build an interstellar highway. Similarly, Bay Area intellectuals tend to see some goal in the future that they want to get to and they make a straight line for it, tunneling through anything in their way. 

Sometimes this Bay Area tunneling produces great results and benefits a lot of people. Sometimes it produces terrible results and harms a lot of people. But, regardless of the quality of the results, it always leaves behind a hole.

“You see, your response of ‘You should probably do at least a little bit of research before confidently stating your opinion on a technical topic’ was a clear example of an ad hominem fallacy, which is explained here on page 495 of the Sequences. Allow me to read this whole chapter to you…”

Speaking of holes, in my weird, nerdy corner of the Internet, there’s recently been a hubbub around an anticavity probiotic, Lumina Probiotic, coming straight from the Bay Area. In the literal sense, this probiotic is meant to prevent holes (in teeth). In the more metaphorical sense, though, it fits in well with previous Bay Area fads, tunneling not only through the dentistry industry but also through the FDA approval process. You see, despite this being clearly a drug and thus under FDA/CDER purview, the founder of Lumina, Aaron Silverbook, has decided to sell this directly to the public as a cosmetic product without doing the requisite safety and efficacy tests.

In case you can’t tell by the title of the post, I think this is a terrible idea, as well as probably illegal. Unlike most people in the Bay Area, I think formalized safety and efficacy trials are a must for health products. In fact, I told Aaron Silverbook this when he asked me for my advice about his product last fall. No, the FDA is not a perfect institution, and yes, it can be improved in many, many ways. But the history of health product regulation is written in blood. There’s a reason why we need some sort of regulation, even if that form of regulation can be improved.

I’ve been annoyed for some months now to see Lumina sell their product directly to the public without FDA approval or even human trials. I’ve also been disappointed for some months to see how many people I respect (and some I don’t respect) have been using it/promoting it. Taking unapproved drugs is a bad idea, no matter what bloggers, Aella, Twitter guys, or conservative firebrands who get sick immediately after taking the unapproved drug tell you.

Despite my annoyance, I had resigned myself to just grumbling about Lumina to my friends. That is, until last week, when I randomly started looking more deeply into some of Lumina’s claims as part of an unrelated project. As I looked more deeply, I became convinced that many of Lumina’s claims are unfounded, and it is, in fact, simultaneously a more dangerous and less effective drug than Lumina or its boosters have been publicly saying that it is. So, I thought it’d be best to write a blog post and share my criticisms to stop people from getting sick from a bad drug.

Also, before I start, I did only think it fair to ask Aaron Silverbook for his comments on this blog post before publishing. So far, I haven’t heard back, but I will update this post if I do [editor’s note: lol this turned out interestingly]. With that in mind, let’s begin.

Background to Lumina Probiotics

Your mouth has a bunch of bacteria in it all the time. In a healthy mouth, these bacteria are useful insofar as they prevent other, worse bacteria from taking up residence in your mouth. However, these bacteria, like the bacteria in your gut or vagina¹ (assuming you have one) are only “good” insofar as your body can keep them in check, mostly through flushing the bacteria from your mouth to your stomach via swallowing, at which point they get destroyed by stomach acid.

Sometimes your body can’t keep these bacteria in check, like when defense mechanisms fail (e.g. persistent lack of saliva) or the mouth becomes more hospitable for the bacteria (e.g. excess of sugar). The bacteria start to overgrow. And when you give bacteria an inch, they take a mile, fortifying their position. The most important way they do that in the mouth is by forming biofilms on the teeth (dental plaque), which are “bacterial cities”. These biofilms are mutual aid pacts, in which bacteria share defenses, nutrients, and swap useful genes.

In the mouth, the biofilms also help the bacteria produce a more hospitable habitat for themselves. These bacteria, specifically Streptococcus mutans, Streptococcus sobrinus, and lactobacilli produce lactic acid through their metabolism. This not only changes the local environment to be a more preferable pH, but it also lets them create pits in teeth by demineralizing the teeth to live in. These pits, if they don’t get remineralized by the saliva, become cavities.

This is all pretty settled science. Brushing your teeth with a fluoride-enhanced toothpaste, like you’ve probably done your entire life, is meant to disrupt these biofilms and promote remineralization of these pits. The newer, less settled science mostly revolves around ways to combat these cavity-causing bacteria directly. There have been various attempts to create vaccines or antibiotics for these bacteria, none of which have been amazingly successful.

One inventor who’s tried repeatedly to address this issue is a guy named Jeffrey Hillman, who’s technically a dentist by training. Starting in the 80s and continuing through the 2000s, he tried to make an idea work where he would genetically engineer safe bacteria to outcompete the bacteria in the mouth that cause cavities and then permanently reside in the mouth. The result would be a mouth permanently full of safe bacteria that couldn’t possibly cause cavities and would beat up any dangerous bacteria that would.

Pictured: Jeffrey Hillman’s origin story. Any resemblance to Tim Burton’s 2005 remake of Willy Wonka is purely coincidental. Ignore the Warner Bros. copyright in the corner.

His ideas around this went through several iterations, a lot of money (definitely tens of millions and possibly more), and several companies, most notably Oragenics. After a number of rat trials (which weren’t entirely successful), one or two aborted human trials², and two bouts of self-experimentation, the idea died on the vine in the 2000s. Then, it was reborn in a less exciting way as ProBiora, a formulation of specially picked, non-genetically engineered bacteria that would also outcompete dangerous oral bacteria. This product then failed its one-and-only human trial (in periodontitis, not caries prevention, to be fair), an outcome Hillman apparently ignored. You can now buy ProBiora today at all participating retailers.

Fast-forward to last year, when rationalist Aaron Silverbook came across Hillman’s original work with the genetically modified bacteria. Aaron, based on his previous work as guy at a rationalist nonprofit, videogame producer, and Aella’s business manager, decided to recreate Hillman’s work³. First, he applied for funding from FTX. He got it, but then FTX collapsed. Then, he applied for funding from alternative rationalist funding source Manifund, got that, and failed to recreate Hillman’s work. However, Aaron declared mission success anyways in that he negotiated with Oragenics to acquire a sample of BCS3L-1, one of Hillman’s later strains⁴, in exchange for $50k and promise of royalties, although he didn’t get any intellectual property rights .

Aaron then went on an intellectual journey where he tried to figure out what exactly to do with this genetically modified bacteria. After all, he was faced with basically the same daunting FDA journey as Hillman, but without Hillman’s scientific background or financial resources. After talking to a bunch of people, including me, he eventually decided on a very rationalist, very Bay Area, very strange approach:

1. Sell the genetically modified bacteria as-is for a one time payment of $20,000 in a libertarian charter city in Honduras

2. Give a bunch of rationalist-adjacent celebrities free samples of the GMO bacteria as-is, including Scott Alexander, Aella (the porn star/escort/sex researcher who he’s the business manager for), Richard Hanania, Cremieux, and Bryan Caplan

3. Take preorders for $200 a piece from the general public

It’s worth noting that, regardless of what I think of this plan (i.e. it’s bad and maybe unethical), I’m pretty sure this plan is also illegal. While Lantern claims to be marketing this probiotic as a cosmetic, it is meant to prevent and cure tooth decay. According to the WHO, tooth decay is a disease. A product meant to cure and prevent a disease is a drug, and legally needs to go through the drug approval process. But, you know, whatever.

Anyhow, enough about Lumina as a company. Let’s talk about Lumina’s science.

So what is BCS3L-1 anyways?

It’s a little hard to say. Jeffrey Hillman created a bunch of different versions of his genetically modified bacteria with a confusing naming structure. But, as far as I can tell, BCS3L-1 was/is a genetically modified S. mutans bacteria with four main features:

1. It produced mutacin-1140, a naturally occurring antibiotic in the bacteriocin family.

2. It was resistant itself to mutacin-1140.

3. It produced alcohol instead of lactic acid.

4. It has a deleted comE gene, which was intended to prevent genetic transformation by wildtype, existing S. mutans.

The first 2 features are natural and occur with some frequency in S. mutans population. Specifically, there are at least 3 S. mutans groups which both secrete and are resistant to mutacin including the group that BCS3L-1 comes from.

The second 2 features are engineered. BCS3L-1 has its ldh gene replaced by an alcoholic dehydrogenase gene, and has, as mentioned, a deleted comE gene.

The lack of an ldh gene means that BCS3L-1 does not produce lactic acid. The alcoholic dehydrogenase gene means that it produces ethanol instead. I don’t think this is the safest way to replace lactic acid, although I don’t think this is the biggest problem with BCS3L-1⁵.

The lack of a comE gene is hard for me to evaluate. As of the early 2000s, when Hillman was creating BCS3L-1, deleting comE seemed like a definitive way to prevent BCS3L-1 from forming effective biofilms or taking up helpful genes (like, say, regaining the ldh gene). As of 2017, this picture is much more complicated. comE is involved in many different things, and some of its functions are redundant.

So, deletion of comE would definitely make it more difficult for BCS3L-1 to be transformed, but it would likely not make it impossible. It would definitely make it harder for BCS3L-1 to survive in the mouth and compete against wildtype S. mutans who are resistant to mutacin-1140. And it definitely would not make it impossible for BCS3L-1 to transform other S. mutans, like by autolysis, which is basically bacterial self-detonation. This releases bacterial DNA in the biofilm, so it could result in wildtype S. mutans getting some of the genes of BCS3L-1 (like resistance to mutacin 1140).

Putting this all together, putting BCS3L-1 in your mouth results in your teeth being colonized by whatever bacteria are resistant to mutacin-1140. Ideally, this would be some mixture of BCS3L-1 and harmless bacteria. However, if any other bacteria are in the mouth that are resistant to mutacin-1140 and can live on your teeth (e.g. naturally occurring mutacin-resitant S. mutans), they will become the dominant species in your mouth, as BCS3L-1 has several fitness-decreasing mutations. Or, it’s entirely possible that BCS3L-1 will start off by being the dominant species in your mouth, then, at some point, will just transform the existing S. mutans into mutacin-1140 resistant S. mutans, at which point these wildtype S. mutans will, again, become the dominant species in your mouth.

But let’s say BCS3L-1 becomes the dominant species in your mouth. What then? Are there any health risks? I’m glad you asked.

The health risks of BCS3L-1

When I talk about the health risks of BCS3L-1, there are really two categories of health risks to discuss: the health risks if BCS3L-1 works as intended, and the health risks if it doesn’t. You see, one of the main functions of the FDA is to not only outline the health risks of products working as intended, but to minimize the health risks of products not working as intended. Like, say, the health risks involved with the product being contaminated.

Category 1 health risks: the unknown health risks

BCS3L-1 is a probiotic, which is a bacteria that you ingest. This puts it in the same category as kombucha or yogurt. And, like kombucha or yogurt, improper manufacturing of BCS3L-1 could introduce any number of contaminants into the product, including foreign bacteria or fungi (like mold). These can make you incredibly sick or even kill you.

The FDA has strict manufacturing standards when it comes to probiotics, known as GMP standards⁶. These include making sure your factory is clean and free from contaminants, your protocol reliably produces safe probiotics and is followed closely by your employees, and your end product is regularly tested. These standards aren’t just paper standards, either. The FDA regularly goes and inspects factories to make sure they follow GMP standards. Because these standards are so strict, most smaller drug developers, like my companies, find it cost-prohibitive to manufacture drugs ourselves, and rely on contract factories to do so.

[editor’s note: removed large section here…]

Lumina likely aren’t following the Best Practices Guidelines for Probiotics, which require you to state how much of each strain in CFUs is in each batch that you send out on your packaging.

[editor’s note: and another section edited out]

Onto the next category: the known health risks.

Category 2: the known health risks of BCS3L-1

BCS3L-1, if you remember, produces two main byproducts by design: alcohol and mutacin-1140. Now the alcohol that BCS3L-1 produces can definitely be a health risk (see footnote 6), but I want to discuss mutacin-1140.

Mutacin-1140 is an antibiotic in the lantibiotic class. Oragenics has spent the past 20 years or so trying to develop it and related antibiotics commercially, most notably for C. difficile, an infection of the digestive tract. They’ve struggled in part because, although mutacin-1140 can be effective against C. difficile, it’s somewhat cytotoxic (i.e. is somewhat dangerous to the body) and caused a hypersensitivity reaction in a rat when given through IV at a high dose. It also has difficult pharmacokinetics, because it binds strongly to blood and plasma, meaning that it tends to go everywhere that blood goes, making it hard to target specific infections. 

On the plus side, it is very effective at surviving the digestive tract, in that it survives for greater than 240 minutes in simulated gastric fluid and for 72 minutes in simulated intestinal fluid. This stability in the stomach is probably why Oragenics keeps trying to develop it for infections of the digestive tract.

Now, given that information, think about the wisdom of infecting your mouth with a bacteria that is designed to continually produce mutacin-1140. You are continually producing an antibiotic in your mouth that:

1) Can be dangerous

2) Goes everywhere that blood goes

3) Is not inactivated by stomach acid

4) Kills other bacteria very effectively

At the very least, this is a great way to give yourself the digestive equivalent of continually taking antibiotics (i.e. diarrhea and indigestion). This also might be a good way to give yourself a hypersensitivity reaction like that poor rat. It’s hard to say, because making a safety equivalence between taking an IV antibiotic one time at a high dose and taking an antibiotic orally at a low dose for potentially decades is really difficult. This is why the FDA requires safety studies.

What I can say for sure is that this would be exceptionally dangerous for infants and immunocompromised people. Infants have died from hospital-grade probiotics before, and immunocompromised people have gotten seriously sick. That’s from normal, “healthy” probiotics. How do you think your infant (who does not yet have a fully colonized microbiome) will respond if you infect them with a bacterium that nukes all other bacteria in their system? Better hope you don’t kiss your baby or share food or drinks with them!

What I think Lumina should do next

I hope I’ve conclusively proven, at this point, that Lumina has messed up big time [editor’s note: if they hadn’t before, they definitely have now!]. They have put a lot of people at risk, including everyone who they’ve non-consensually infected by introducing a genetically modified bacteria into the wild. They messed with things they didn’t understand, put profit and “acceleration” before safety, and, in one of the worst sins that a rationalist can do, tunneled a hole through the Chesterton fence of the FDA⁷.

If Lumina had the good sense of Hillman (who, to be clear, I don’t think is a scientific saint either), they at least would have sold the version of the GMO bacteria that had a self-destruct button, AJ2M, which I think was the last strain Hillman created. That one was designed to [editor’s note: removed a possibly incorrect claim] require an exogenous amino acid, d-alanine, to function. If the d-alanine stopped being provided, the bacteria died, assuming it didn’t acquire any mutations in the meantime that let it keep surviving.

But Lumina didn’t do that, even though I and, I assume others, told them to do that. They sold the earlier version of the probiotic without a kill-switch, which means that the cat is out of the bag and is probably giving overly credulous rationalists diarrhea as we speak. So, at the very least, Lumina needs to:

1) Stop selling the probiotic and refund everyone who’s bought it so far

2) Do research into which antibiotics and what course of antibiotics will completely remove BCS3L-1

3) Message everyone who’s bought BCS3L-1 with information on how to test if they still have it and how to rid them themselves of it if they do

4) Fund people’s doctors’ visits to rid themselves of BCS3L-1

[editor’s note: 5) Stop trying to sue bloggers who ask questions about the safety and efficacy of their product

Or, you know, they could also do the typical Bay Area thing and keep tunneling along regardless of what damage they leave in their wake. After all, as every dentist knows, one person’s hole is another person’s business opportunity.

1. These links go out to previous writings I’ve done on the digestive microbiome and vaginal microbiome, respectively.

2. It’s really hard to figure out what happened with these human trials. Lumina and Hillman claim the FDA sank the human trials before they began by requiring them to be done on young healthy adults with dentures, but I doubt that. Oragenic’s SEC filings also cast doubt on that (see full comment thread). 

Based on my limited dealings with the FDA, my guess is that Hillman, inexperienced with the strange ways of the FDA, asked them, “Hey, I’ve got a crazy idea. How should I test it?” And the FDA responded, “With maximum caution.” Then Hillman said, “If I use maximum caution, I don’t think I can run this trial.” And the FDA responded, “Ok, then give us a better idea of how you want to test it and we’ll tell you if we’re ok with it.” Then Hillman did do that, they gave him the ok, and then Hillman’s trial failed and Hillman was probably still mad at the FDA.

The pro strategy, in case you’re wondering, is to come to the FDA with a really good idea of what you want to do first, and frame all questions to the FDA as, “We’ve done a ton of research and thought a lot about this. This is what we think the best way to run a trial is. Does the FDA agree?” Asking them to design the trial for you is asking them to give you maximum restrictions.

3. I don’t mean to throw stones about this. This is just literally what Aaron reported his qualifications as on his funding application to Manifund.

4. As I’ll discuss later, BCS3L-1 doesn’t appear to be the last strain that Hillman created, which was probably A2JM. 

5. Although this may mean that BCS3L-1 is a bad idea for people who can’t metabolize alcohol, like many East Asians. See this post for more details.

6. GMP stands for Good Manufacturing Protocol, which is actually just the generic name for all the protocols that people are required to follow for manufacturing any regulated product. The specific GMP standard you have to follow depends on the type of product you are trying to manufacture.

7. Chesterton’s fence: if you don’t get why the fence is there, don’t climb it. Similarly, if you don’t get why the FDA requires safety, efficacy, and manufacturing trials, don’t assume they’re unnecessary.