It’s the vascular bed of the eye. And that includes the choroid, the ciliary body, and the visible part of the uvea that you can see is the iris. So often, in the front of the eye, we call this iritis, but technically these are all uveitis. Uveitis is not that uncommon. The incidence is about 50 per 100,000 and the prevalence about 100 per 100,000. So at any given time, there’s about 300,000 people in the US who have active uveitis, means that about 1.2 million people in this country alive today will have this at some point in their life.
And it can be a royal pain in the neck, but it can also be a blinding eye disease. And depending on which list you read, it’s either the fifth or sixth leading cause of blindness in the US, so not a trivial problem. Now, when we talk about uveitis, people say, well, what causes it? What do we do about it? And it’s really not a disease entity unto itself. It’s a descriptive term. It’s any intraocular inflammation. And so it’s a very heterogeneous collection of individual diseases, all of which share a final common pathway of leukocyte infiltration into the eye. So the way you know someone has uveitis is you see an inflammatory infiltrate in their eye. But that can be because of an autoimmune process, an autoinflammatory process, or occult infection.
And I’m going to spend most of the talk covering occult infection. When we think about autoimmune diseases, the canonical one is a disease called sympathetic ophthalmia. This was an unfortunate gentleman who took a BB to the eye, which basically ruptured this eye. This is what we call phthisis. So that’s an opacified cornea, and that eye really has no visual potential. The retina is detached. But unfortunately, if the eye is not enucleated and significant amounts of uvea are exposed to the immune system, because of the ocular immune privilege, because the eye is a privileged space, the thymus never really got a good shot at the unique intraocular antigens. And you end up with this terrible disease called sympathetic ophthalmia where the good eye becomes subject to a chronic autoimmune attack. And this is why, when we do see people who have severe injury to one eye, we frequently will enucleate that eye within two weeks of the injury, and that lessens the risk of sympathetic.
Exactly what the antigens are is not known at this point. And it’s a rare condition, only occurs in about one in 1,000 ruptured globes, but it’s catastrophic because it’s your only good eye. This is actually what Louis Braille suffered from and went blind from. He had had an injury in the early 1800s as a child in one eye and then lost his vision in the other eye to sympathetic ophthalmia. Most of the diseases that we know are autoimmune are fairly rare syndromes. And they’re really interesting diseases. Vogt Koyanagi Harada disease, which I doubt any of you have ever heard of, is a disease that features tinnitus, uveitis, and vitiligo. So you get patches of skin depigmented and ringing in the ears, and you go blind. And it’s because of an autoimmune reaction to melanin, which is found in your skin, your inner ear, and your eyes. Birdshot choroiditis, another interesting, very rare disease– this is a picture of birdshot down here– little yellow spots in the back of the eye.
It’s a retinitis. And this disease has among the highest HLA associations of any disease. Essentially 100% of patients with birdshot are HLA-A29 positive. And again, the antigen is not known, but it’s a bilaterally symmetric disease that’s highly HLA linked. And another interesting autoimmune disease is tubulointerstitial uveitis and tubulointerstitial nephritis and uveitis syndrome, or TINU, which, again, is incredibly highly linked to an HLA-DR locus, about relative risk of 150 and causes, mostly in kids, an acute interstitial nephritis and an acute uveitis. So there are some pretty interesting autoimmune diseases, but I would say this is the vast minority of patients that I see clinically. Autoinflammatory causes are also really interesting and perhaps even rarer depending on how you lump things. The only really canonical autoinflammatory disease that we know uses this mechanism is a disease called Blau syndrome, and this is one that you find in kids, again. It’s a bilateral chronic granulomatous uveitis that’s usually associated with erythema nodosum in the skin and, often, pulmonary nodules. And it looks a lot like sarcoidosis. The disease is due mutation in NOD2/CARD15, which is involved in toll receptor signaling and is in the activation pathway between TLRs and NF-kappa B activation.
So mutation there basically causes a persistent response to normal flora, which results in granulomatous inflammation. But what I really want to talk to you today about are the occult infections. And this is where I think lab medicine intersects a bit with what we do in uveitis. It’s a little bit mind boggling to think how brief in the history of human endeavors the knowledge of the microbial world is. Anyone know what this is a picture of? SUBJECT 1: Van Leeuwenhoek’s microscope. RUSSELL VAN GELDER: Exactly. It’s van Leeuwenhoek’s microscope. So the senior faculty were around when this was in vogue.
So and anyone know what van Leeuwenhoek did for a living? He was not a microbiologist. SUBJECT 2: Textiles. RUSSELL VAN GELDER: He was a textile guy. He made drapes. And he felt he was getting ripped off on his fabric, that the thread counts on his fabric was not up to snuff. And he was an amateur glass blower, and so he made this little device with a little molten spot of glass. You can barely see it in this picture. Right up here, there’s just a tiny globule of glass in the middle of this metal thing. And if he held his eye close enough to it, he basically got a simple microscope. He’d put the thread that he was looking at here on the screw and then raise it up into the line of sight, and that was the microscope. But what he found was not thread count, necessarily. He drew pictures of these things, which he called animalcules, and these animalcules looked a lot like fungi to most of us, and, of course, he could also see some hints of bacteria.
And it was this whole kingdom, this whole world that had opened up because of this the silly little device that he had made. Progress being slow, of course, microbiology kind of stalled out for about 200 years until the mid 1800s. Anyone know what this is as a culture medium? SUBJECT 3: Potato. RUSSELL VAN GELDER: Right. It’s anthrax growing on a potato, which is how Koch isolated anthrax.
So he did not have the luxury of agar plates. Agar plates were actually invented by one of his fellows. They were trying to make something besides potatoes that they could grow cultures on, and they were using gelatin. And the problem of gelatin is, of course, that there’s lots of gelatinases in bacteria, and so is dissolving the plates. And one of the spouses of one of his fellows was a baker and said, oh, you should try seaweed extract.
It works really nicely, makes a thing that looks like gelatin, but it isn’t. Of course, they had no idea it was polysaccharides or anything like that. But they said, OK, we’ll try that. We’ll use agar, and hence the agar plate was born, and that was in about 1880. And here we are today in 2019 and we’re still using agar plates. In fact, this could have been a Koch plate. And viewed in the future, I think when people look back at this era at mixtures of seaweed extract and sheep’s blood, or brain-heart infusion, or some of the things that we use that seem more out of Macbeth than out of modern medicine, they’ll kind of scratch their heads and say, well, I guess that worked.
It’s interesting. But really, was that optimal? But this is still state of the art for a lot of microbiology. But I’m bringing coals to Newcastle here. So all of you are familiar with PCR. Don’t need to review the background here except to point out Gobind Khorana’s contributions to PCR just for history. I think Kary Mullis won the prize for reducing PCR to practice. But if you go back in the literature, Khorana actually described PCR pretty accurately in a 1969 paper. But of course, he had no thermal stable polymerase to actually do this. Khorana, of course, famous for solving the genetic code, but also became a vision researcher in sort of the second phase of his career and was involved in purification of rhodopsin.
So what has PCR taught us about uveitis? Well, we have at least three diseases that were idiopathic uveitis. And these are all pretty rare, but they’re interesting. This first one is called Fuchs heterochromic iridocyclitis. And this is a really interesting chronic uveitis that this woman has where basically the uveitis eats away the iris over time and changes the eye color. And it’s always unilateral. This is the affected eye in this patient. It’s a triad of relatively asymptomatic uveitis cataract and glaucoma elevated pressure. And for years, people thought, oh, there must be some missing pathogen here because it’s always unilateral and it’s doing funny things to the iris. Subsequent PCR and antibody-based work actually revealed that this is a chronic rubella infection. And it’s childhood or congenitally acquired rubella that gets into the lens of the eye, which can serve as a safe harbor for this. And in about 30% of patients with this condition, if you tap the eye and run RT-PCR, you will find rubella virus. And all patients with this have high titers of intraocular rubella antibodies. A similar disease called Posner-Schlossman, which looks very much like this except the iris doesn’t turn as light but is associated with much more episodic uveitis and high eye pressures, turns out to be due to cytomegalovirus, but in an immunocompetent host.
So this is not CMV retinitis. This is actually CMV in the normal host. Again, the eye being immune privileged can provide a safe harbor for the virus to replicate and persist. And so people get episodic reactivation of CMV in an eye, and that’s Posner-Schlossman. And then PCR has also led us to realize that HHV-6, Human Herpes Virus 6, is associated with certain forms of bilateral panuveitis. So we’ve had a few small successes in identifying occult pathogens with specific uveitic diseases. But I want to share two cases with you that got my interest up in this area when I was at Wash U in St.
Louis. First one was this gentleman who came to see me having gotten something in his eye mowing his lawn about three months prior, so just felt something in his eye and then it went downhill from there. He saw an optometrist, an ophthalmologist, and then a cornea specialist. And by the time he got referred to us, he’d had two corneal biopsies, which were both negative for anything, no bacteria, no fungi. It was thought maybe he had a parasite or a protozoan, like acanthamoeba. He’d been treated for acanthamoeba, which is basically using pool disinfectant PHMB, and that didn’t work. So he came to us. And at that point, his cornea looked pretty bad. Hard to see here, but he’s kind of caved in on this side on his cornea.
So he basically needed a cornea transplant urgently. I sent him to my colleague, Tony [INAUDIBLE],, and asked Tony if he would share with me a piece of the cornea and that schmutz in the front of the eye, which he did. And I ran– this was in about 2004– ran 16S and 28S PCR. And lo and behold, the 28S band came up positive. The first band is from his cornea. The second is from the material in the anterior chamber. Third’s positive control. Fourth is negative control. So this was the old days.
So I cut the bands out and sequenced them and asked Tony, in the meantime– it took about a week to do this– how’s the patient doing? And he says, oh, the cornea transplant looks great, a little inflammation. I’m sending him back to you. So that’s a hypopyon. The eye is 50% full of white blood cells there. So I talked with the patient and I said, look, it’s clearly a fungus. We got the sequence back, and it didn’t match anything in the database. It was about an 88% match on the 28S to fusarium, but it wasn’t fusarium.
In the meantime, the pathology came back negative, silver stain, GMS, all negative on the cornea. So I said, I think that this is fungal. I can’t really prove it, but it sure looks like it. Why don’t we treat you like this is fungus. So we put him on Itraconazole and Natamycin, and two weeks later he looked like this. And when I left St. Louis four years later, he was 20/25 in that eye, seeing well and the infection was cleared. So this kind of got my interest up in occult pathogens. The next case didn’t end as happily.
This was a cardiothoracic surgeon, also in St. Louis, who had had a bone marrow transplant for myelodysplastic syndrome and was out fishing, cut his arm, got some water in it, and got a cellulitis. He was on, I think, Mycophenolate and Tacrolimus and was admitted to the hospital for cellulitis and started to go downhill. So he got transferred to the ICU and was complaining about floaters in his vision and decreased vision. So our resident service saw him. And this is a not great picture of the back of his retina because it was taken with a portable camera.
But you can see this little yellow dot here– there’s a couple of them– and this general haze, and that’s what we would call a multifocal choroiditis, multiple little choroiditis spots plus a vitritus. So we asked the ID service, well, what’s growing in his blood cultures? And they said, well, nothing’s really growing, but there’s something kind of weird here, and we’re sending the smear out for evaluation. So it was these little red spherules. And I think they sent it to AFIP, and AFIP came back and said, we’re not quite sure, but we think it might be prototheca wickerhamii. So how many of you have heard of prototheca wickerhamii? Yeah, this is a great audience. I’ve asked that like 20 times in ophthalmology audiences and no one ever, right? But this is coals to Newcastle. So prototheca wickerhamii is a non-photosynthetic algae. This is a really bad thing to be growing in your blood. And unfortunately, he succumbed to the infection. His wife was very kind and let us take the eyes at autopsy.
Mort Smith, our ocular pathologist, splayed them open. We punched out those choroiditis lesions. I made PCR primers to prototheca wickerhamii, and there it is– so positive for prototheca wickerhamii in the eye. Now, the scary thing about this case, other than what exactly you can catch from Missouri fishing streams, is I see a lot of, lot of patients with multifocal choroiditis that looks like this, and in a million years, I wouldn’t think, oh, that’s probably algae in the eye, right? So it just takes one or two cases like this and you start to become a conspiracy theorist and think, well, how much of what I see that I call idiopathic is really an occult infection and I have no idea what I’m looking for? So that kind of motivated us in the lab to try to develop some techniques for looking for occult pathogens.
Now, of course, again, coals to Newcastle, the revolution in deep sequencing has made a lot of approaches to this feasible that even five or 10 years ago would have been science fiction. And we’re fortunate in my lab. We have an aluminum iSeq. So we can play with this sometimes very quickly and try things out. All of you are familiar with Moore’s law, which basically says computing power doubles and price halves every two years, and that’s pretty much held for three decades. For sequencing, we’re on Moore’s law squared at this point.
The cost of sequencing since 2008 has really dropped at the square of Moore’s law, which, since that’s already a log law, is really impressive. And as you all know, that first human genome, which cost over $100 million, now is reduced to in the hundreds of dollars, if that. So we’ve applied three techniques to trying to find pathogens. First is just 16S metagenomics. For reasons I’m going to show you, we rarely use this. This is a really bad technique if you’re dealing with a relatively clean sample because, frankly, 16S makes stuff up.
It’s just too easy to get contaminants that are not meaningful. So we do very little 16S metagenomics. Wonderful for the GI system where you have a rich environment, not so great in the eye. The second technique, which we used for a number of years, we call Biome Representational in Silico Karyotyping, and I’ll explain that in a minute. And more recently, we’ve just gone to whole genome pathogen metagenomics. So 16S you’re all familiar with. The basic idea is that there’s conserved sequences in the 16S . Ribosome you get an amplification product that allows you to see pretty much all bacteria if you sequence it and put it on a high throughput platform. The Biome Representational in Silico Karyotyping was really a technique that we started developing at Wash U with Elaine Mardis. And when we moved in 2008, Jay Shendure was very kind to work with us on this. And the basic idea here was in the days when it was still too expensive to do whole shotgun metagenomics, could we do a representational method that approximated that? And the analogy here is let’s say you have an Oxford English Dictionary, which is your biopsy sample, and you think that there’s some French contaminating your English dictionary that is foreign DNA.
You could read the whole dictionary– that is computationally and timewise a daunting task– or you can pull out every 50th page and read it. And if you find French, you’re done, right? And it turns out, if you don’t find French, statistically you can put a pretty accurate upper limit on how much contamination there may be in your dictionary, and it’s much less than one per every 50 page for statistical arguments. So the way this technique works is it takes advantage of these type 2S restriction endonucleases. In particular, we use the one BsaX1. So BsaX1 recognizes this AC-5N-CTCC sequence. But unlike a type 1 endonuclease that would cut inside that sequence, the type 2S and 2B enzymes cut outside. And BsaX1 actually cuts a big chunk outside. It cuts 33 base pairs around that recognition sequence, leaving two random three base pair overhangs. So what we did was we basically digest with this enzyme. We isolate the 33 base pair fragment. And then we ligate on the Illumina adapters directly, which makes it very easy, then, to do high-throughput sequencing on these.
And then we do the same trick Illumina does to get the asymmetric adapter ligation. That is, we use one biotinylated adapter and one non. And then we amplify that and sequence it directly. So what does this get us? Well, the six base pair recognition sequence, AC-5N-CTCC, occurs once every four to the six base pairs, or 4,096 base pairs. And we’re cutting out 33 base pairs. We kind of throw away the two, three base pair overhangs because we don’t know that the ligation is perfect.
And so we get 27 base pairs, basically, out of every 4,000. And so that ends up being about 0.7% of the genome is released, on average, by this restriction enzyme, and then we cut it out and sequence that. One of the advantages of this technique is that it is, in terms of bioinformatics, it’s extraordinarily efficient because we can create a hash table of every BsaX1 site in GenBank, which we’ve done. And so once we get the sequences off of the Illumina, it takes us less than 10 minutes to run through a billion base pairs’ worth of sequence and map every single sequence to human, or if it’s non-human known, or if it’s unknown, put it in its own pile. We call it karyotyping because, as a side effect, we get a full karyotype of the human genome out of this very rapidly, as well. So it’s a very useful technique. It’s becoming less useful as whole genome amplification has become more affordable.
But we still use it, and I’ll show you an example where it showed us something that we did not see with whole genome. And of course, whole genome is just basically fractionate the DNA and sequence it all and then put it through, crack under another pipeline that’s designed to detect metagenomic sequences. So it’s almost like having a new microscope. And what I want to show you in the rest of the talk are five little vignettes of what we’ve learned with these techniques.
So here’s our new microscope. Let’s take a look. First question we want to ask is, what’s on the normal eye? What actually– does the eye have its own microbiome? Is it like the gut? Do we have this little colony that lives in our conjunctiva? You would think that this is a question that would have been answered a long time ago. But in fact, if you read the literature, this question has been debated for over 100 years, and it’s still not resolved. Does the normal conjunctiva have any flora? The eye surface has some innate antibacterial mechanisms. In particular, there’s a lot of lycozyme in tears, which will reduce bacterial load, and there are antibacterial peptides that are secreted by the conjunctiva. So it’s been known for a long time that the conjunctiva is relatively paucibacterial. But is it completely sterile in the normal case or is there a little colonization? So to look at this, Thuy Doan, who was a fellow resident and fellow with us and is now assistant professor at UCSF, and my colleagues did a study where we took 100 normal subjects and we swabbed all four surfaces of their choroiditis, that is, the palpebral, which is the lid side, and the bulbar, which is the eye side, and then did standard culture.
We did BRiSK and we did 16S metagenomics. What we found by culture was basically what other people have found, which is that about 20% to 30% of the samples yielded no growth. The remainder really only showed four organisms in any kind of prevalence and three of them at higher prevalence, and that was coag negative staph, staph epidermidis primarily, propionibacteria, diphtheroid, corynebacteria, and strep, and mostly strep viridans. And those are the four things that we found fairly routinely. There’s lots of other stuff that show up in a case or two, bacillus, Neisseria, lactobacillus, E. coli, et cetera. But for the most part, the only things that are found in double digit sorts of prevalence are those four bacteria. Now, when we did 16S metagenomics, we actually replicated a result that was published in 2011, which found 400 genera on an eye swab with metagenomics. That’s impressive given that we only had four total bacteria growing that we could identify and maybe another 10 that we ever identified.
Getting 400 seemed like a lot, but maybe it’s like van Leeuwenhoek all over again, this whole new world and there’s all these microbes that you can’t culture or whatever. But we decided to go back and look at this a little bit more carefully, and we did quantitative 16S PCR, qPCR. It normalized to actin NS. Well, how many bacterial genomes are there actually on a swab when we take it off the eye? And for controls, we did the inside of the mouth, the buccal mucosum. We did the skin of the lower eyelid. And what we found was interesting, that if we normalized a human actin, the skin actually has about 22 bacterial genomes per human genome on a swab. It doesn’t mean that that’s what’s living there.
It means it’s what comes off on a swab. And the cheek was actually a little less bacterial. I would’ve guessed it was more, but we got about 12. But when we looked at the conjunctiva, it flipped and we only got 0.02 bacteria per human genome. So when we calculated the human bacterial load on the swab, it turned out that this corresponded to about 50 bacteria per swab. So it’s very hard to imagine how you get 400 genera on a swab when there’s 50 bacteria on it. And that’s why I say 16S metagenomics does not work well in the paucibacterial environment. You pick up, I think, environmental contaminants, and I think it kind of makes stuff up. It takes small areas that are amplifiable, and because of sequencing errors, you end up with what looks like a very rich environment that I don’t think is there.
So we’ve taken this to be paucibacterial. We’ve confirmed this in a number of other studies that there’s relatively few bacteria, but there are bacteria. So what are the bacteria? Well, if we look at the things that showed up in more than 1% of the samples, it’s our four friends again. Its corynebacteria, propionibacteria, strep, and staph. So it’s the exact same stuff. And again, we find a few lactobacillus or Neisseria or other things, but the vast majority are just those four. How do we know that those four are really there of the things that we looked for? We did this heuristic calculation based on the difference between environmental swabs and our conjunctival swabs over 100– actually, 400 samples, and basically did a p value on whether we would find this by chance. And you can see the p values on the heuristic are in the order of 10 to the minus 48th. So we’re pretty sure they’re actually there for those bacteria. Is that niche unique? So principal component analysis has its advantages and disadvantages. This is not a validated set in that this is the set that we did the PCA on.
But we could distinguish a population that was conjunctival, which is in gray here, from the skin, which is in blue, the cheek, which is in red, and the environment, which is in black. But you’ll notice that the PCA, the principal component, really is on the continuum between the skin and the environment, which is kind of what you’d think. So it’s not like there’s a whole different microbial world on the conjunctiva. It’s related to both skin and environment. And if you gave me a sample on here and asked me to characterize it, I’d be about 90% accurate if I didn’t know which one it was. So I think there is a unique environment there, but it’s not bugs that you don’t find elsewhere. So this was all pretty disappointing. Basically our foyer into finding the whole brave new world of the conjunctiva didn’t really yield that much.
But what we did find was through BRiSK, and it was surprising, which is that the ocular surface actually has a resonant micro virome. So BRiSK is agnostic to the DNA source. We can find bacterial DNA, viral. Basically any kind of DNA will show up there. And when we ran through our samples– I realize it’s small, so I’ll enlarge it a little bit. There’s our propionibacteria, our corynebacteria, strep, staph again. But the arrows show three viruses that we found in substantial numbers of samples. And that’s a virus called torque teno virus, which is an anellovirus that I’ll say more about at the end, Merkel cell polyomavirus, which we found in a fair number of samples, and HPV, Human Papilloma Virus.
So the TTV, the torque teno virus, is shown down in the corner here for validation. This is just looking at 10 samples, five where we found it, five where we didn’t. And indeed, we could confirm that the TTV was there by direct PCR. So I’ll have more to say about TTV toward the end of the talk. But for those of you not familiar with this little virus, it’s tiny. It’s 3.8 kb. No one really knows what it does.
There’s no good in vitro system for replicating this virus. It’s nearly ubiquitous. Zero positivity rates are north of 90%. And of course, with any new virus, it’s been associated with everything, chronic fatigue syndrome, other chronic inflammatory diseases, but has not been really definitively linked with any of these conditions. It is sort of a normal commensal of the serum. And you can follow– you can track immunosuppression in patients by their TTV load in their serum, which is really interesting. So it appears that the immune system does keep this in check for whatever reason. But when it’s not in check, it’s not clear that it causes a whole lot of disease. We went ahead and confirmed the presence of this in 58 swabs, and we found it on 22. So this is not an uncommon virus to find on the ocular surface.
Merkel cell polyoma– very similar. I think many of you are more familiar with this. Paul Nghiem has done quite a bit of research here on Merkel cell polyoma for its role in Merkel cell tumors, Merkel cell carcinoma. But it’s also a relatively ubiquitous tiny virus, but it’s in the polyomavirus family. So it has a large T antigen, and it has oncogenic potential. But for the most part, it just is found as a commensal, not so much in the blood, but certainly on the skin. And again, almost everyone has been exposed to this virus. We found something interesting with this virus. I won’t go into the study in detail. But working with Chris Chambers, one of our oculoplastic surgeons, we wanted to look at whether the microbiome and virome of the eye surface changed when there’s no eye. So he does enucleations on patients who have tumors and other reasons for their eye to be removed. And when we remove the eye, we restore the conjunctiva, but there’s no longer a cornea there. It’s an acrylic implant underneath the conjunctiva.
So we took swabs of these patients, of their healthy eye and their enucleated conjunctiva, and ran PCR for the various things that we’d been finding on the surface. And interestingly, if you’ve had your eye out, you shed Merkel cell from that eye. So we found Merkel polyomavirus in pretty much every sample of the fellow of the enucleated eye, and in the same patients, in the majority of them, we did not find it in the healthy eye. So interesting finding– don’t know what it means. But this is a pretty ubiquitous virus in humans. And then HPV– a little plug for those of you doing metagenomic work. My colleagues Aaron and Cecilia Lee kind of tackled the Kraken problem.
Those of you who use Kraken as your pipeline for looking for metagenomic sequences know it has its limitations. It’s not super efficient, and it often pulls up stuff that’s not necessarily correct. So they devised a new pipeline that’s really interesting. It uses the Google search algorithm, basically, to create hash tables of chopped up 30-mers of your sequence and can then assign pathogen sequences very quickly that way. And using that, we were able to fully reconstruct this HPV that we found on an eye surface, which is shown here. So there are 7,000 bases of HPV and obviously very deep coverage on this, more than 30x coverage. So we’re pretty sure we got the whole virus. But when we ran it through GenBank, it did not fully match any known HPV. It had bits and pieces. Now, whether that’s a recombinant, I don’t know. But the tiling really looked like it’s one virus. And so our suspicion is that this is a new variant on the eye surface, but we haven’t confirmed that to date.
All right. So that’s the normal surface. Let’s talk a little bit about what we find in disease states, and I’ll give you three examples of this. First one is microbial keratitis. This is a picture of a patient with a corneal ulcer. This is still a pretty common clinical problem. You can see the ulcer here. Basically that’s an infiltrate. The epithelium has been eroded, and you have infectious keratitis with an immune reaction around it. Very serious problem. Obviously, if that’s in your visual axis, you’re not going to see well out of that eye. Still a leading reason for us to do cornea transplants. In the US, the main reason people get these is they abuse their contact lenses, sleep in them, wash them in tap water, things like that. Worldwide this is a huge problem, and there’s over 1 and 1/2 million cases of blindness a year worldwide due to corneal ulcers, many of them in India and many of them due to fungi.
Thankfully, in this part of the world, we don’t see a lot of fungal ulcers. So we wanted to know if PCR– so one of the surprises in this condition is the patient shows up like this in the emergency room. We take a sterile spatula. We go right here where we know it’s infected. We scrape it. We put it on our sheep’s blood and agar mix, and nothing grows. 40% of the time, these are culture negative, cultured Gram stain negative. Really weird because it’s infected.
There’s no question this is an infectious process. So we wanted to see if PCR would help us figure out what these pathogens were. This is an older study that I did when we were in St. Louis. Elma Kim, who was a medical student at the time with us, was first author. And we wanted to test 100 corneal ulcers as to whether they had infectious PCR evidence of pathogen or not. At Barnes Hospital, which is a 1,400-bed hospital in St. Louis, it would have taken us two years to collect 100 corneal ulcers. We got about one a week. And that was going to take too long. So I talked to my friends Jack Whitcher and Tom Lietman at UCSF who had a collaboration with the Aravind Eye Hospital in Madurai, India.
Aravind is an amazing facility. It’s an 800-bed hospital, so the size of Harborview and UW hospitals put together for eyes and only eyes. And they see about 50 corneal ulcers a week, not a year, coming into their ED. So I sent Elma over with the UCSF team who’s doing a study there, and I told her, come back when you have 100 corneal ulcers. I really thought it would take her a month. Two weeks, she emails me and says, I have 100 corneal ulcers and really bad diarrhea. Can I come home now? I said, yes, you can come home. So she brought 108 ulcers home. This is their operating room, by the way. When they do cataract surgery, there’s five patients in there at a time, each– or six, actually. You can’t see the sixth one here. Each surgeon is doing surgery on two tables. So they’re operating on one while they’re swapping the other patient in and out.
And the surgeon just does this loop all day where does a cataract surgery, turns around, rescrubs, rotates, does the next surgery while they move the other patient out, turns around, scrubs. They’re very efficient, and their outcomes are actually a little better than ours in terms of complications. It’s an amazing place. So what’d we find with PCR in these cases? The answer was that all of them showed either– nearly all of them showed either bacterial or fungal bands by PCR, but not both, and that’s important.
So these do not look to be commensal because we found either one or the other. Most of them in India or 3/4 of them are fungal, 2/3 to 3/4. Only one sample, 107, was positive for both a bacteria and a fungus. When they were culture positive, which was, in our study, about 55% of the time, the PCR agreed 95% of the time with the cultured organism. So we’re pretty sure that we’re capturing the pathogen. So what did we find in the culture negatives? Well, there were 46 of them altogether in the study. 17 of them amplified bacteria. 29 of them amplified fungus.
The bacterial ones were not that interesting. So we found stuff that we usually culture, corynebacteria, strep, pneumo, pseudomonas, staph epi. These are all common causes of corneal ulcers. But we did find two, quote, “uncultured bacteria” in the database. So maybe two out of 17 of those were new and interesting things. The fungal side was a lot more interesting. So out of the 29 fungal, a little over half of them, 17, were either fusarium or aspergillus or both. But the rest of them were stuff that we just don’t see very often– Sordaria, Phythium, Botryodiplodia. If you go to the literature, all of these have been associated in one or two case studies with a corneal ulcer, but no big series on any of these.
So altogether, if you take the ones in orange and red here, which are either weird bugs that didn’t grow or uncultured fungi or bacteria, that ended up accounting for about a quarter of the culture negatives were bugs that we wouldn’t have thought of or necessarily detected. So we do think that there is this little world of stuff that doesn’t grow well that shows up. In India, this would be a very useful technique because obviously treatment of fungal and bacterial ulcers is quite different, and you can’t tell them apart just looking at them. So a quick point of service test that said fungal or bacterial worldwide would be hugely helpful in endemic parts of the world where corneal ulcer is endemic.
Fourth lesson– this is actually something that I’m guessing about a third of you in the audience have– you may not know it– blepharitis. So blepharitis is a condition where you basically get a scaly excrescence of your eyelashes. Bacteria grow in the roots of the lashes or in the meibomian glands, tend to secrete some toxins. Those get into your eye and give you a chronic feeling of grit, foreign body sensation, a little redness, itchiness of your eyes. And it’s due to this condition. This is exacerbated because a lot of people, probably 30% to 40%, also have demodex, the mite, living in their hair follicles. I didn’t put my demodex picture in here, but it’s really cool. These alien-looking skin mites actually live in your hair follicles and they’re thought to contribute to blepharitis.
So blepharitis is the bane of ophthalmologists’ existence because we don’t have good treatment for it, it drives patients nuts, and it doesn’t blind anybody. So people take up a lot of chair time with their blepharitis, and you want to help them, but you can’t really help them, and they’re not going to go blind, and the next patient sitting in the waiting room is going to go blind. So if we had a cure for blepharitis, everyone would be happier. Why don’t we have a cure for something this simple? Well, I had the opportunity to work with a company, NovaBay that I mentioned earlier, who designed a drug that was intended to treat, initially, adenoviral conjunctivitis, but ultimately they kind of pivoted it to blepharitis. And it’s an interesting drug because it’s not a drug. The drug that they sell, which is called Avenova, is actually preserved saline. That’s the product. But it’s preserved with 0.01% hypochlorous acid, which basically, when it’s applied to the skin, converts to hypochloric acid for about 20 seconds and then is oxidized away into harmless stuff.
So it’s actually a really good antiseptic for the eye. In fact, in their studies, it’s comparable to povidone iodine in terms of its killing power on bacteria. And it has a little bit of antiviral activity, as well. So they wanted to test it on blepharitis as a lid scrub. And they did a very small pilot study where they took three patients and treated them with these scrubs where they took them for a week and then two patients with, I guess, non-preserved saline– that’s the control– and they wanted to know what changed on the flora of the skin.
So they sent us the swabs that they did. In the study, they swabbed people before they used it, 20 minutes after their first scrub, a week after using it, and then a week after they stopped using it. And what they wanted to see, I think, was that all the bacteria went away and the patients were happy. Well, patients were happy, actually, the results on patient comfort. Whether it’s placebo or not is hard to say, but they do feel better with this. But when we looked by PCR– and this was done with BRiSK– we didn’t really find a huge change. So this is just pie graphs here. Blue is staphylococcus in the subject, which is number 16. Left eye is treated. Right eye is untreated. And basically there’s not a whole lot of change there. It’s staph, staph, staph, staph. There’s a little bit of corynebacteria and a little bit of P. acnes, but it didn’t really change. Now, of course, BRiSK or any PCR can’t tell live from dead. We do know that this stuff kills four logs of bacteria when it goes on the eye. But we call this the mowing the lawn effect, which is you kill all your bacteria in the morning and then by night they’re all back.
It’s the same people. It’s the same lawn. You’ve just mowed the lawn. So it didn’t really seem to work. Here’s another patient. Same type of result, but the weird thing is that the software just colors the dominant organism blue. And now it’s corynebacteria in this patient. And in fact, there’s no staph to be seen, or very little. I think one sample here had a little bit of staph. And that was a surprise to us. I mean, why does this patient have all corynebacteria and the other one have all staph? And all of the results were kind of weird that way. So we decided to just validate this by direct PCR and see what’s going on.
So these are the five subjects, and this is their actin PCR. And you can see, we got good recovery on almost all the swabs. The key to these experiments is it’s the left, right eye, and visit one before they scrubbed, visit two is 20 minutes after their first scrub, visit three is a week after scrubbing, visit four is a week after they stop scrubbing. So that’s the key here. And this is the staphylococcal result. And that patient I showed you was number 16. And you can see, indeed, that confirms its staph, staph, staph, staph. But if you look at 12, who is the second one I showed you, indeed, there’s very little staph there. And same with 13, very little staph. But 14, 15, 16, lots of staph. What about the corynebacteria? Well, here’s 12, who I showed you before, lots of corynebacteria. Same with 14 and same with 15. Again, not much in 13, not much in 16. What about strep? Well, strep we find in 12, 15, and 16. And for amusement, we thought we’d run our torque teno because we know torque teno’s on the eye surface, and we find it here in 12, 14, and 16. So you need a scorecard to keep track of all of this.
Here’s the scorecard. And the answer is to each his own or her own. There’s no consistent patterns here. Some people are corynebacteria dominant. Some are staph. We find some who have both corynebacteria and staph. We have some who have no corynebacteria, but show strep. No patterns here. Everyone’s unique, and I think this is why we have such a hard time treating this condition is it’s not one condition, it’s a whole bunch of different conditions with different bacteria that probably respond differently to different treatments.
All right. Last lesson and this is probably the most interesting one and the most serious one. So this is an eye that looks like that first eye I showed you with uveitis, but this is a particularly bad form of uveitis. This is post-operative endophthalmitis. This is someone who had cataract surgery and then 48 hours later called in complaining their vision had suddenly dropped and the eye was red and painful. Thankfully, this is extremely rare. In the current cataract surgery, the rate of this condition is 0.03%, three per 10,000. So it’s rare we see this. But when we do, it’s a serious problem. The incidence of endophthalmitis, despite it being very rare nowadays, is actually going up, and that’s because, in the treatment of macular degeneration, we have to inject drugs directly into the vitreous of the eye, the anti-VEGF drugs, and so there’s actually more intravitreal injections being done now annually than cataract surgeries. And there’s already 3.3 million cataract surgeries a year done in the US. There’s about five million injections a year now done for macular degeneration. So even in a 1 per 10,000 risk, when you’re doing 10 million surgeries, it ends up you have thousands of cases of this.
So it’s a serious condition. So we published a study a couple of years ago where we did our BRiSK analysis on these endophthalmitis cases. Now, just like with the corneal ulcer, we have a problem in that we know the eye is infected, and yet, if we tap the eye, acutely infect it, and immediately culture it or Gram stain it, about half the time we get nothing, 40% to 50%. Are these infected? Is it the same story as the cornea? Are we just not detecting the pathogen? So collaborating with my friend and colleague at Wills Eye Hospital in Philadelphia, Sunir Garg and his group, we took 21 consecutive endophthalmitis cases that presented to retina specialists who treat this, and we did culture 16S metagenomics in BRiSK. And 11 of these were culture positive, 10 were culture negative. For the ones that were culture positive, for the most part, both 16S and BRiSK recapitulated the organism. So there were seven of these that grew staph epi. Five of them we detected molecularly by either 16S or by BRiSK, and they all agreed that it was staph epi.
Two of them we didn’t, actually, which was interesting. And I don’t– it’s challenging to know, with staph epi, whether, when you just go in and sample, you’ve gotten a little bit of the stuff that’s on the conj or was it really the stuff in the syringe. The streps were interesting. So the streps came back with the lab medicine microbiology strep intermedius and viridans that are descriptive but not really genera or species-level descriptions. And again, molecularly, we confirmed all three of them as strep, but as strep gordonii, agalactiae, or mitis, which were the true bugs in each of those three cases.
We had one where the lab called it light growth strep. We didn’t find anything molecularly in that one. The Moraxella case we all agreed on. Interestingly, the lab came back with one Prevotella. But molecularly, both molecular mechanisms showed it was not Prevotella. It was a strep species that came up positive. 16S did not find any detectable pathogenic organisms in the remaining cases, and BRiSK found one that might have had a tag or two for pseudomonas and one that had a strep. But they were very, very rare. So we went back and did our quantitative 16S analysis again. And what I’m showing you is the raw results here of the 16S normalize to actin and then the quantitative PCR that was done separately. And we take two bacteria, two genomes, as sort of our background level because you can’t go into an eye completely sterilely.
Even when you’ve done povidone iodine, there’s dead bacteria on the surface. What do we find? Well, if we take two as our limit, every culture positive was above the line. Every culture negative was below the line. So it means that the culture negatives really are bacteria negative or pretty close to it. It’s not that there’s this whole world of unculturable bacteria in these cases. We just didn’t find bacteria there. So what do we find? Well when we did BRiSK, we found a ton of tags for our friend torque teno, and not a few in some of these cases, but a lot in some of these cases. So this is number of tags recovered, which is– because there’s only one or two tags per genome, it’s a rough measure of the number of genomes recovered. And we were getting 1,000 and, in one case, almost 10,000 tags back.
So this is not a small amount of virus. This is a lot of tags. We found torque teno in every culture negative case. And we found it in half the culture positive cases. But we had seven controls of eyes that underwent vitrectomy surgery for macular holes and diabetes, and in none of those did we find torque teno. So it was only found in the infected eyes. We’ve subsequently done a larger study with outcomes. We didn’t have outcomes in this study. And in the larger study, we find torque teno present in 21 of the 33 vitreous samples, so about 63%. And interestingly, we find our friend Merkel cell polyomavirus virus in about 25% of these samples, although rarely both, usually one or the other. Here’s the one sample where we clearly found both viruses in the same sample. Now, I do not want anyone to leave here thinking torque teno is a cause of endophthalmitis. We have no data to support that. We don’t know if it gets in the eye and replicates, if it’s coming out of the serum because the eye is very inflamed, if it’s being carried by white blood cells.
No idea, but it’s a biomarker. So we decided to look at outcomes in the study. And the first outcome we looked at is how people did for vision. Now, vision here is listed as logMAR, which is logged minimal angle of resolution, which is a standardized way of showing vision. 0.1 is a normal logMAR. So 20/20 vision is 0.1 on the scale. A one is 2200 vision, which is legal blindness. And a two is really bad. That’s can you count fingers or see a hand moving? So most of these patients come in with a logMAR pretty close to two. And many of them get a lot better. So you can see that the average outcome in these two groups ends up being about 0.7, which is a visual acuity of about 20/70, which is almost legal driving vision.
And I certainly have patients whose vision is worse than that who are on the road. So be careful when you’re out there. But it’s a fair outcome, not a great outcome, but it’s fair. And those were the ones who either had staph epi or were culture negative. They had about the same outcomes. But the ones that were culture positive for everything else, the strep, the Moraxella, they did not do so well. So they ended up with vision worse than 2200, which is legal blindness.
So we do know that what bug you have in the eye matters. And I’ll come back to that in the slider, too, when I have my list for you guys. But the interesting thing is when we look just by TTV, torque teno virus. What we found is that the patients who presented with really bad vision were the ones who tended to have the torque teno virus, which, again, doesn’t say whether it’s causative or whether it’s just reactive to patients that have more inflammation.
But they came in with much worse vision, and they ended up with much worse vision. So the group that was torque teno negative actually ended up with visual acuities that were in the 20/40 range, which is pretty– most people’s cataracts come out when they’re 20/40. So they don’t do too badly. But the ones who were torque teno positive end up with a much poorer vision on their outcome. So at least we do consider this a biomarker, potentially, for risk stratification. We thought we had a clever result here. And I was at a meeting with one of my colleagues, Todd Margolis, who’s now the chair at Wash U of Ophthalmology, and I said, we found this interesting virus in endophthalmitis called torque teno virus. I didn’t expect that. And he says, that anello virus? And I said, yeah, I think it’s an anello virus. He said, didn’t I just publish that? And he did, and I’d missed it. So there’s this really unusual uveitis in Nepal called seasonal hyperacute panuveitis, and it looks just like endophthalmitis. It happens in kids every fall. And it’s coincident with the eclosion of this moth that has this giant caterpillar that tends to crawl on people’s faces and really likes hanging out and drinking their tears.
So the thought is that these little moth hairs, these little caterpillar hairs penetrate the eye. This happens with tarantulas, also, by the way. Don’t get a tarantula as a pet because their little hairs get into your eye and can cause a bad uveitis. But it never cultures anything. It’s always sterile. And Todd and George [INAUDIBLE] in Holland analyzed the fluid from these kids.
And what did they find? 90% of them had torque teno virus. And they looked in a cohort of endophthalmitis cases and found 50% very close to what we found, as well. So really the credit for this goes to George and Todd’s groups. But it suggests that this is a worldwide phenomenon. So some lessons learned from our foyer into deep sequencing– first, the conjunctiva turns out to be an interesting surface that’s paucibacterial. But it’s got a resident virome, and that’s not something we think about. But in chronic inflammatory disease, which is a lot of what we deal with on the eye surface, it kind of makes sense. If you’re shedding virus all the time and fighting it, doesn’t that give rise to chronic inflammation? And maybe that’s some of the dry eye, blepharitis stuff like that that we see in these patients.
When we see keratitis and it’s culture negative, PCR says usually interesting bugs, or often interesting bugs, whether it’s fungus or bacteria, but there’s something there. The blepharitis story is that everyone’s different, and what looks like one disease by just phenotype ends up having lots of different microbes associated with it. Maybe we need to do some personalized medicine here to say yours looks viral, yours looks bacterial. I didn’t tell you the EKC story today. That’s a similar story for conjunctivitis where it turns out that what we thought was all adenoviral is not, and there’s other things responsible there.
But the culture negative endophthalmitis, unlike the keratitis, is actually bacterial negative in most of these cases. But surprisingly, we find this virus in the eye that we did not expect, and it seems to be a biomarker for presentation and outcome. So I put together a little wish list for working together with lab medicine. This is, I think, what Jeff was alluding to. He gave a wonderful talk for the chairs at MSEC on your department, and it’s amazing the work that’s going on here and the innovation. And so, from our perspective, there’s some stuff we would love to innovate around. The main one is the eye is small and the fluids we get out of the eye are tiny small. If I tap an eye, I’m lucky if I get 50 microliters to 100 microliters. You really don’t want to suck 5 milliliters out of an eyeball because there’s not much left when you do that. So we need micro-scale analyses in order to do things that, for everyone else, is easy, like serology.
I would love to know if someone has toxoplasmosis antibodies, anti-toxoplasma antibodies or anti HSV or VZV in certain cases of inflammatory eye disease. But we can’t analyze small samples well enough to do that. That would be a godsend if we could do micro analysis for serology. Cytokine analysis– again, a lot of the stuff that we’re looking at, we don’t know if it’s infectious, autoinflammatory, autoimmune, or even a masquerade tumor. IL6, IL10 levels, for example, completely different from a tumor in the eye and uveitis. So IL6 will be sky high in uveitis, IL10 sky high in a lymphoma. We would love to be able to do a 25 or 50 microliter cytokine analysis on those samples. The deep sequencing stuff that I’ve shown here, I think, is great.
It’s all under research protocol. Obviously can’t use it clinically. It would be wonderful if we could do deep sequencing using the techniques and get it under a CLIA license so we could actually use this stuff clinically. The second wish list is the endophthalmitis. Again, many of these cases are negative, but if they’re either PCR negative or staph epi, we don’t worry as much about them. They’re going to do OK, for the most part, with just injecting antibiotics into their eye. The ones I worry about are the streps and the pseudomonases and those that can really completely destroy an eye within a day or two.
And we would want to take those patients to the operating room and clear all their jelly out, do a vitrectomy, just bathe the eye in antibiotics and treat it like an abscess. But I don’t know who’s who when they present, and they all look the same when they present. So a rapid point of service diagnostic for staph epi or a suite of bacteria would be, again, a godsend for us.
So that’s an area we’d love to work with you guys to help develop that kind of a test. And then what we’d really love is one or more people who are fascinated by tiny things, like tiny fluid volumes, and would want to be a point of contact with our department. So if you’re interested, please talk to me. We have lots of interesting projects, and we’d love to work with you. So we’ve come a long way from van Leeuwenhoek to Illumina. We still have a long way to go. But I think it’s an interesting era right now where we can re-examine many of these diseases with tools that just weren’t available 10 years ago and find some interesting things in terms of the microbiology.
Lots of folks contributed to this work. I want to particularly highlight Thuy Doan who was our resident and fellow who did the ocular surface microbiome– she now has a wonderful group at UCSF working with Joe DeRisi, continuing the work in deep sequencing and eye diseases– my colleague Cecilia and Aaron Lee who are both assistant professors in the department who have done much of the bioinformatics work that I’ve shown– and Cecilia really has done all of the endophthalmitis project– and then a number of medical students, UW medical students, who contributed to this, including Dallin Anderson and Michael Gutowski, both of whom are now ophthalmology residents who did other aspects of this work.
Lots of great collaborators, as well. Jay Shendure has been wonderful to work with on the representational deep sequencing, colleagues at Wills, Bascom, which is University of Miami and UCSF. And of course, none of this happens without funding. I’m very fortunate to have good NIH funding and also some foundation in philanthropy funding for this work, as well. So I appreciate the opportunity to present, and I’d be happy to take some questions. Thank you. MARK: Thank you for a great talk. What’s known about local production of antibodies versus systemic and ratios? How can we help you in that regard, and what can you tell us about it? RUSSELL VAN GELDER: Great question. Mark is the only person here that I really work directly with because we share some patients together with autoimmune diseases that affect the eye. That’s been pretty well studied. And if you sample intraocular fluid in a case where you know the eye is infected, for example, toxoplasmosis where you can see the scar and reactivate it, if you normalize to either IGG, total IGG, or you pick a ubiquitous antibody, like an anti-mumps or something like that, the ratio of intraocular antibody titer to the normalization relative to serum to the normalization is always above three.
So if you draw a line at three-fold higher antibody production, you basically get a 100% area under the curve for local production. So it’s a very useful diagnostic test for us, and it’s been used, in addition to toxoplasmosis, for HSV and VZV, and CMV infections in the eye. Obviously, AIDS patients it kind of falls apart because their immune responses are weaker. But in general, it’s a very useful technique. There’s a lab in Rotterdam that provides the service for the EU, Aniki Rothova’s lab. And so in Europe, they do this test routinely. And in fact, they do it in preference to PCR, in many cases, because often there’s a transient bump in viral load, but as the immune system comes in, it becomes harder to detect for the viruses. MARK: And if I could, the related question is are there B cells and plasma cells, then, within the eye that are making those antibodies? RUSSELL VAN GELDER: Yeah, there are. They’re rare, but they can be found. If you actually do flow cytometry on a patient with an active uveitis, you find a small CD19 positive or B220 20 positive population.
To date, no one’s done the single-cell cloning, which I would love to do in some of these idiopathic diseases. And if anyone’s doing single cell B cell cloning and reconstitution, like the Don Gilden type thing in MS or SSPE, we would love to work with you on that because that’s something we’ve been trying to do for a long time, or hoping to do.
MARK: Sean. SEAN: That was really interesting talk. I’m interested in this cutoff between background and true positives. So you said in your microbiome studies, you’re getting 50 bacteria, let’s say, off the surface of the eye. And so there are plenty of PCR tests that would pick that up, and yet it wouldn’t be pathogenic presumably. So when you go to a keratitis, is there a big jump? RUSSELL VAN GELDER: Yeah. That’s a great question. The answer is, yes, there is. There’s a huge increase in the number of bacteria that are recoverable, and it becomes dominant bacteria, like 1,000 to one over human cells in an active case where it can be recovered. So that’s why we didn’t have– well, we did that study, the PCR study in India, before we were doing a lot of deep sequencing. So that was manual. In the ancient days, we actually, I think, took 12 clones of each 16S product and sequenced each of them from each of those 100. And all of them agreed on one organism. So you don’t find this proliferation of organisms and it’s dominant to one.
So we’re pretty sure that, in those cases, it’s there. It’s just not growing on the culture plate. And whether it’s the immune system has already killed them off by the time you see them and they’re not viable or what is hard to know. But the flip side is it is true that the PCR techniques are not great for speciation when you’re in this paucibacterial environment. I would guess that would apply to CSF, too, that if you did sort of wholesale 16S metagenomics on lumbar punctures, you would get the same type of results that we’re seeing here, which is a lot of bugs that you just scratch your head and go, there shouldn’t be rhodococcus in a CSF. [MUSIC PLAYING] .
As found on Youtube