With last week’s big cancer vaccine news — Moderna’s positive phase 3 trial for a personalized cancer vaccine in melanoma — I took the opportunity to sit down with an actual cancer vaccine scientist to ask all of my questions.
Dr. Tiffany Ma is a senior scientist at Infinitopes, a spin-out from Oxford University. She helped make one of their current vaccines, for esophageal cancer, which they just started a clinical trial for, with the first patient dosed last month!
This episode is brought to you by my partners at Consensus, the AI agent for literature research and medical questions. Get a free trial with my link!
Moderna’s positive phase 3 trial in melanoma
Moderna (in collaboration with Merck) last week press-released the phase 3 trial data for their mRNA cancer vaccine, in resectable high-risk melanoma. Patients had melanoma removed by surgery, then Moderna built personalized vaccines based on that specific patient’s mutations — the neoantigens. The vaccine was given alongside the immune reactivating antibody, pembrolizumab.
It’s just a press release — with no actual data yet — but the vaccine improved metastasis-free and recurrence-free survival. There’s no data on overall survival yet either, which is of course the most important thing. On the news, Moderna’s stock price had the biggest one-day rise in S&P 500 history — about $50 billion in market cap…
The cancer-vaccine landscape
Globally, there are 241 vaccines in phase 1 clinical trials; 220 in phase 2; 33 in phase 3, and 11 approved globally (two in the US). So, although it’s true that they haven’t become fully standard of care, there has been some success — and it’s an interesting little stat that there are two approved in dogs.
Tiffany Ma:
Just to say the ones that are approved aren’t really true neoantigen cancer vaccines. The first one is BCG, in bladder cancer, which is just a weakened bacteria that acts as an immunostimulant. And the second one is a cell therapy for prostate cancer, a dendritic cell vaccine. Which is where you just take out the dendritic cells from the patients and then retrain them and then inject them back into the patient. There hasn’t been mRNA vaccine success, peptide vaccine success, or viral vector vaccine success before.
So Moderna’s news was such a milestone for the cancer vaccines world, because this is the first neoantigen based vaccine to get to this stage of success in a phase three trial.
How does Moderna’s cancer vaccine work?
I thought we should go back to some basics, for those of us not in the biotech world. There are three phases of clinical trials. Phase one is usually for safety, phase two is for efficacy, and phase three is for seeing whether or not your drug beats the current gold standard therapy that’s being used in the clinic. And cancer vaccines in general, it’s a drug that teaches the immune system what a cancer cell looks like, so that the immune system can recognize the cancer cells and kill them.
Vaccines educate your immune system, whether it’s a flu jab or a COVID vaccine, and then your immune system goes around your body like the police, and kills them. And this part is a bit sciencey but I think it’s important for people to understand how cancer vaccines work, and that’s the mechanism of HLA molecules presenting these little peptides on the surface of your cell.
You can think of your cell as a house with windows, and the windows are your HLA molecules, and their job is to show this flag with patterns and wave it to the outside world. And then your immune cells, your T cells come by like policemen and check the badge, check that flag to see whether or not you’re a healthy cell. And if you happen to be a cell that’s infected by a virus or you happen to be a cancerous cell, then the T cell would be able to tell by reading that flag, the patterns of the flag specifically that you’re showing on the window.
The vaccine is teaching T cells: this is what flag you should be looking for, and then once it comes across a cancer cell then it will know to destroy it.
Immune checkpoint blockade — pembrolizumab — unleashes T cells into a immune system which is already trained by the cancer vaccine. So they work together, they’re synergistic. And this is what Moderna used in their phase three trial.
Why have cancer vaccines struggled in the past?
Why they’ve had a checkered history is because of a whole host of problems.
The first problem is target selection. So, finding the right target is actually quite tough and it wasn’t until recent RNA, DNA and protein sequencing technologies that they got really robust that we were able to find really good targets. What makes a good target is something that is not expressed on your healthy cells and something that’s expressed only on tumor cells. And if you pick something that’s even expressed at a relatively low level on your healthy cell, you can imagine there’s going to be side effects.
The second issue is tumor heterogeneity. Tumors mutate very quickly, especially some types over others. And that means even though your vaccine might work on this part of the tumor, it might not work on another part of the tumor. They might have different clones or different populations of tumor cells.
The third problem is immune escape. Tumor cells are very clever and they know how to hide from the T cells, the policemen of your body, in multiple ways. Losing antigen presentation is one of them and expressing cells that suppress T cell infiltration is another one of them.
The fourth is clinical trial design. You want to give the vaccines to people who have a working immune system and, unfortunately, if you give it to people who have late-stage metastatic cancer their immune system might not be able to put up as good of a fight anymore.
The fifth is manufacturing — expensive and hard to scale — and, again not until recent years that we got the tech to do it at nearly an affordable level.
And the last one is delivery, which is something probably Moderna’s proud of — getting the actual vaccine into the cell can be quite tough as well.
How do you overcome those challenges?
Moderna’s is a personalized vaccine i.e. each patient gets their own bespoke vaccine. So for that you have to wait until the patient’s tumor is resected. After the surgeon takes it out, you then take it to the lab and you do deep sequencing. You do DNA sequencing and RNA sequencing to identify the mutations that the tumor has and you use these fancy computational models, obviously some of them involving machine learning.
Then you find these targets that we call neoantigens. And if you are Moderna then you’re making an mRNA vaccine with up to 34 targets strung together so then they can be translated by the cell once they get in. The whole process takes like 6 to 12 weeks.
I guess that overcomes the target problem, because you’re building it based on the neoantigens from that person’s tumor?
That’s only half true I will say. Because yeah the good thing is it’s personalized. So you might find things that are specific to your tumor and not to another person’s tumor.
But unfortunately mother nature doesn’t turn every mutation into a peptide that’s then presented on the surface of your cell — actually less than 1% of those DNA mutations are eventually expressed as peptides on your cell. So they’re not actually the most precise way of selecting targets — but it is efficient, because sequencing technologies have gotten so good and so cheap and scalable now that we’ve really mastered next-gen sequencing, you can do it for each patient. Moderna are predicting from DNA and RNA level what might be there on the peptide level.
In order to go straight to what’s being expressed on the surface of the cells, you need to do mass spectrometry, and specifically a technique called immunopeptidomics, which is what Infinitopes does and that’s why we call it our precision immunomics platform — because we don’t predict, we measure what’s actually being expressed on the surface of the cells by MHC I.
You mentioned that Moderna are leaving ~99% of the potential targets on the table using their neoantigen approach?
That’s what I would argue, whereas the Infinitopes approach takes all of that into consideration. The technology has limits, but yes we basically see as much as we can detect using the current mass spec technologies, on the surface of the cell. So we start from what’s actually there instead of doing guesswork and computational work.
But the caveat is mass specs are a lot harder to operate than DNA and RNA sequencers. So that’s our moat. If you can find someone who’s really good at mass spec and immunopeptidomics, you can build a cancer vaccines company out of it and that’s what the Infinitopes co-founders did.
The second thing you mentioned was heterogeneity of the tumor and how that can affect things. From the Moderna point of view, I can imagine that the 34 neoantigens they load into the vaccine addresses that issue?
The jury is still out on that one in the cancer vaccines world. Because if you treat your T cells like a person, you can’t distract the person with too many targets, but at the same time, you want more shots on goal. So if your target number one isn’t actually expressed on the surface of the cell, then you’ve got like the 33 other targets and you pray that at least one or two of them might actually be there. But there’s a maximum limited by manufacturing and by the biology — you can’t just put a huge string of epitope in there.
The third one you said was immune escape. Does adding pembrolizumab — anti-PD-1 — address that?
Yeah, exactly. It’s definitely very synergistic especially in melanoma which is the low-hanging fruit indication for cancer vaccines just because melanoma is famously known for having a high tumor mutational burden, which means they have lots of mutations for you to find — so it’s like the level one of cancer vaccines.
On trial design, so you said that previously we were putting vaccines into late-stage metastatic patients and of course with the melanoma trial from yesterday as well as your trial, it is in patients with resectable disease, right?
Yeah. It is in resectable disease, which is a really important consideration because at that point we assume that the patient still has a working immune system but then when you try to teach the immune system it will actually be effective. Usually at that stage, the point is not to prevent people from getting cancer ever, but it’s to prevent these leftover bits of cancer cells that your surgeon couldn’t resect probably because they don’t want to resect too much of your tissue right or they couldn’t spot it.
And those bits that are left behind are usually what causes relapse and recurrence. So the cancer vaccine does its job by teaching the immune cells. So then after the surgery the T cells do their job of killing the leftover cancer cells that the surgeon didn’t remove.
You mentioned the moat of Infinitopes was in the mass spec platform and in measuring rather than predicting.
Yeah that’s one of our moats. Our delivery platform is actually a moat as well.
It’s an engineered viral vector, that can be very immunogenic as well and cause a very long-lasting T cell response, which you want in the body. You want it to be there months or even years after you’ve had the vaccine. So then if a little cancer cell pops up and it was hiding, then your T cells are like, “Oh, you’re dead.”
And the delivery platform for Moderna is mRNA, right?
Yeah, exactly. And a lipid nanoparticle. They usually go hand in hand, but they’re very short-lived, which might mean you need multiple shots of the vaccine, whereas with viral vector, you may not need so many shots. So that adds up in terms of how costly it would be, how many times you would require the patient to go to the clinic. And the Moderna one is given nine times.
But the good thing about mRNA is you don’t even have to get to the nucleus. You just translate it into the target. The bad thing is it’s also exposed to all these things in a cell that can get rid of it, metabolize it very quickly. So you need way more shots of it.
And the other thing you said was on manufacturing.
There’s obviously a big difference between Moderna’s manufacturing versus Infinitopes’ manufacturing. Because for Infinitopes, whereas Moderna gives it in an adjuvant setting i.e. after the tumor has been resected, because you need 6-12 weeks to make the vaccine before you can give it back to the same patient. Infinitopes is making an off-the-shelf solution, which would in theory be a lot cheaper. We are finding targets that are shared between many different patients but they’re very precise because as I said we’re not predicting.
We did our discovery work by sourcing many different tumor samples from the same cancer type, and then we see what’s being shared — and we’re going to be really confident from our very precise and very in-depth mass spec combined with very cutting edge computational techniques to find those targets.
And a lot of those targets, they’re very interesting — there are different sources of targets. They could come from viral sources — famous ones like HPV, but it’s not just the canonical genome. There are people now looking at the dark genome, which is actually 99% of our DNA that doesn’t code for the classic proteins in our body. More and more companies in the cancer vaccine space are finding — wow — some of these dark genome targets are good cancer vaccine targets.
I’m going to ask you about the dark genome later but… I can definitely see that off-the-shelf is probably cheaper and faster. But my criticism would be that on measure not predict, I see that you’re measuring the antigens on the MHC, but I would argue that you’re measuring the antigens that are on the MHC of different patients — the test patients, and not the patients you’re actually giving the vaccine to.
We’re also going to measure the antigens on the patients that we’re giving the vaccine to. So, in case they don’t get a response, we can measure their immunopeptidome and be like, “Oh, it wasn’t there in the first place and that’s why it didn’t work.” But we designed our vaccine in such a way that it’s HLA-matched. So, the patient has to have the right HLA type, before they get our vaccine, to maximize the chances of success.
So, if you select for the HLA type subtype then you increase the chance of success? But the other thing that you do is you subdivide the market. You make the market smaller, because there’s fewer patients that have that HLA subtype. Whereas with the Moderna one, it’s not HLA matched, it’s for everyone.
It’s just not scalable. You have to make one vaccine per person. So there’s no right or wrong way. I think we’re all trying to move towards the same goal of making the right vaccine for the right patients, and that might mean their disease profile or whether or not they have deep pockets and healthcare system. And there’s always trade-off for every choice on the targeting, on the delivery system, etc.
I was super curious. When the news came in yesterday from Moderna, how was the reaction at Infinitopes?
I think we love that honestly. Even though they’re technically a competitor, we think they’re doing a great thing for the field because now for cancer vaccines people, it’s no longer a question of do cancer vaccines work, it’s how do we make them even better? And which ones are the right cancer vaccines for the right patients? So, it’s exciting times.
Infinitopes’ approach
Infinitopes’ mantra is right targets, right vectors, right patients, right time. Could you talk us through each of those and what it means?
Targets, like I mentioned, you can have a very focused approach. You could be a company that just looks for viral elements etc, or you can cast a wide net and just look for as many differences as you can between normal cells and cancer cells.
Or you can do the Moderna approach which is finding targets that are just so specific to one patient. So you can pick your approach and then obviously in the within targets you also have the discovery process of those targets. Are you doing RNA or DNA sequencing? Are you using mass spec? What computational techniques are you using? So that’s the first part.
Second part, right vectors and that goes back to is it a dendritic cell vaccine, peptide vaccine, DNA, mRNA, viral vector, and the differences between those.
And then right patients, so personalized versus off-the-shelf and are these early stage patients or late-stage patients?
And finally, right time — when in their treatment regimen are you giving them the vaccine? So is it preventative like we’re seeing with some of the Lynch syndrome vaccines that are being developed or BRCA vaccines or is it neoadjuvant, adjuvant or even metastatic in the future?
In that context, I wonder if you could tell us about the trial that you guys are running, how it works and when the vaccine is getting given.
Our trial is a really exciting phase 1/2a. So, we’re doing phase one and a bit of the two at the same time. It’s a 60 patient trial in esophageal adenocarcinoma. We have eight patients for the safety dosing parts and then 52 for the 2a part.
It’s going to be given before the surgery and after the surgery. Usually, esophageal adenocarcinoma patients get chemo-surgery-chemo. So here we’re doing chemo, Infinitopes vaccine, surgery, and then Infinitopes vaccine (with or without immunotherapy). And then the readouts — the best readouts are overall survival or recurrence-free survival. But in the meantime, we can look at other readouts like circulating tumor DNA, which can be measured by companies like Personalis, who are very famous for having very high sensitivity for detecting leftover presence of cancer cells. This is what’s called minimal residual disease (MRD) and you need to be able to detect such small presence of circulating tumor DNA — the DNA that’s leaked out from cancer cells — to be able to tell the patient hey you need to come back to the clinic.
And that’s one of the issues of the trial, right? Because it takes a long time for recurrence to happen sometimes. So you might have to run the trial for a very very very long time. But if you look at the ctDNA, then you can see the recurrence on the molecular level on the DNA level much sooner.
Exactly. So people use it as a proxy for whether or not their drug is working. But the time point really matters as well because sometimes there’s a bit controversy about the ctDNA among scientists because the moment you attack the tumor, you’re going to see a sudden increase in circulating tumor DNA because you’ve just killed tumor cells and they burst. But then hopefully after the initial increase, you see a decrease again. It depends on the kinetics of the drug and the patient.
The dark genome
You mentioned earlier the dark genome and the 99% of the genome which is not coding, and how that can help you build better vaccines.
Just to go back to basics, the coding part of the genome just refers to genes in your DNA that eventually become proteins.
The classical 20,000 genes that we call the coding parts of the genome. But the rest of the DNA, for many years, scientists have called it “junk DNA” because they have no idea what it’s there for. But recently (and when I say recently I mean like the past 10 20 years) scientists have found that it performs really important regulatory functions in the cell.
It could be these regulatory RNAs like circular RNAs can come from the dark genome. There are microproteins that can come from the dark genome. And these regions they sometimes produce peptides that end up as neoantigens on the surface of your cell. They can come from alternative splicing, certain retroviral elements or even like when the ribosome makes mistakes and slips when it’s reading the mRNA that can also produce neoantigens from the dark genome. So, it’s basically anything that’s not defined classically as a coding, protein part of the DNA.
So the mass spec approach is a better, more sensitive approach to to pick that up?
I would argue yes. But you also want to match it with other types of sequencing, which is what I’m sure all the companies are doing.
And there’s this concept of which we didn’t mention actually — tumor-associated versus tumor-specific antigens?
Tumor-associated antigens have been found in the past they’re basically proteins or targets that are expressed by healthy cells but usually at a low level, but they’re overexpressed by cancer cells which is why it’s just associated but it’s not specific to cancer cells — like HER2. Whereas tumor-specific ones are literally only expressed by cancer cells.
And if we are able to characterize the dark genome, can we find more tumor-specific antigens and therefore make better vaccines?
Yeah, perhaps there are some really interesting studies that found some tumor antigens that came from the unmutated genomic sequences, but they’re from the non-coding part. It’s just that they’re now being aberrantly expressed into proteins and then peptides. So, it doesn’t mean that looking at mutations will get you the best targets. Sometimes you have to look at perhaps what’s being expressed that’s not normally expressed.
So there are lots of creative ways to think about it, and each company has their own spin on it.
The future
One of the ideas for the future: vaccines for prevention?
There are trials ongoing for cancer prevention in patients with Lynch syndrome or BRCA mutations, but there are also broader ideas — vaccinating pre-cancerous states, like colon polyps, pancreatic cysts or Barrett’s esophagus.
I think that’s definitely part of the future. The only difficult thing about them is you have to wait for so long, to see the effectiveness of these vaccines, which is usually why they’re being led by academics and not by companies.
How do we actually, practically, do cancer vaccine trials for prevention?
I think you have to enrich for the very high-risk patients — at least initially — and then you’re going to have to do things like surrogates and ctDNA, and maybe even look at the CD8s. And you can also look at whether or not the T cells themselves — the expanded pool — whether or not they have the receptor that’s specific to your cancer vaccine target. You can do TCR sequencing.
The only issue then you’d have to prove to the regulators that that does indeed correlate with reduced progression to cancer, which is then going to be a little bit difficult. Only time can tell unfortunately, which is why these things take so long.
Another idea for the future: vaccines against immune-cold and metastatic cancers.
For the immune-hot cancer types — melanoma, mismatch repair-deficient colon cancer, lung cancer — I think cancer vaccines are very likely to work alongside immune-stimulating therapy. But you have to think about therapeutic vaccines for some of the diseases which are not so immune-hot — but where they have serious unmet needs.
Like glioblastoma and pancreas cancer — I think those are the going to be the challenges for cancer vaccines. I had a look at the indications where most of the cancer vaccines are being trialled, and it goes — melanoma, lung cancer, glioblastoma — so there’s a lot of attention for cancer vaccines in glioblastoma, which I think would be really good because it’s totally unmet as as it stands by therapies.
Some creative ideas are to vaccinate against the population of cells that’s making the tumor immune-cold — like a specific T cell subtype subgroup, like Tregs.
And I think eventually what excites me is not just preventing cancer from coming back. It’s if we’re able to administer a cancer vaccine to a metastatic patient and make it work. That’ll be amazing. Especially in an originally immune-cold setting.
Are you guys working on that?
Oh, sure.
Thank you very much to Dr. Tiffany Ma for the illuminating conversation!
Personalized neoantigen cancer vaccines represent a meaningful advancement over traditional tumor-associated antigen approaches. The translational challenge is the turnaround time from tumor biopsy to individualized mRNA vaccine synthesis. That window must be short enough to be clinically useful given disease progression timelines in aggressive solid tumor indications.
I’m looking forward to see what comes from BriaCell and can any of this research apply to other cancers. BTW: after the decipher test follow-up treatments my 3 month PSA has halved. Hope to keep on a downward trend.