Skip to main content


Regeneron’s Monoclonal Antibody Cocktail in Primates

I’m very glad to see some data emerging on the monoclonal antibody therapies that have been in the works. Regeneron  and Eli Lilly/Abcellera are two of the front-runners in this field, which could pick up some very much needed slack while the many vaccine trials are going on. And late yesterday a preprint emerged from the Regeneron effort, looking at challenge studies in animal models. Since we were just looking at the primate challenge studies in the various vaccine efforts the other day, this is timely indeed.

This team has a mixture of two monoclonal antibodies (REGN10933 and REGN10987). The isolation of these has already been described, as well as a study looking at their effectiveness versus Spike protein mutations. That latter paper illustrates some key points – when a patient develops their own antibody response to an infection like this one, there’s a wide variety of antibodies that get raised. Some of them are neutralizing by themselves, and some not, but that simultaneous defense by a number of different antibodies is definitely a good feature. By contrast, look at what happens if you try the one-antibody-only technique of treating a pathogen with an injected monoclonal antibody: you’ll hit the target hard, but at the same time you’re applying strong selection pressure for any viable mutants that can escape the attack. And that’s what Regeneron saw in their model systems with single antibody treatments – mutants that would normally have a low profile with no particular advantage suddenly had a chance to amplify if they escaped the binding of the monoclonal antibody. Thus the mixture of two monoclonals, which bind to non-overlapping regions of the Spike protein. Now a virus needs two separate escape mutations, which is extremely unlikely.

The paper shows two rhesus challenge experiments, each done by injecting the animals with 50 mg/kg of the mAb mixture (25 mpk of each antibody), followed three days later by exposure (both intranasal and intratracheal) to either 1×105 or 1×106 PFU (plaque-forming units) of coronavirus. Those doses, when compared to the ones we’ve been seeing for the vaccine challenge experiments, are bracketed on the low side and the high side of the range. For the lower dose (Figure 1B in the paper), it looks like subgenomic RNA (which is indicative of replicating virus) in the nasal swabs is pegged at the low limit of detection in all animals at all time points, and sgRNA in the lungs is only detectable in one animal at Day 1. In the higher-dose experiment (Figure 2B), sgRNA is detected (low level) in the nasal swabs at Day 1 and Day 2, but not thereafter. The animals were also checked via oral swab, and the sgRNA was also cleared very strongly from these samples. I should note that they also tried a much lower dose of the antibody cocktail (0.3 mpk) in that second experiment, which was ineffective.

There’s also a treatment experiment, where the animals are not pre-treated, but dosed at the higher challenge level and then after a day are given either 25 or 150 mpk of the antibody cocktail to see if it can deal with the infection once it gets going. The sgRNA data show significantly enhanced clearance of the virus for both doses – in fact, the two look pretty similar, except (oddly) the 150 mpk group shows higher viral titers on Day 1. Not sure what’s going on there, but it looks like the mAb cocktail would be an effective treatment as well as a prophylaxis. Lung histology on all the infected animals confirmed this – the signs of viral pneumonia were far milder in the treated animals than in the controls.

The paper goes on to examine a golden hamster model. That’s because although rhesus macaques are certainly a lot closer to human, they don’t really get hit hard by coronavirus symptoms. The virus infects them and replicates, so it’s a good model in that sense, but the animals don’t show any severe symptoms. The hamsters, on the other hand, have it rough: severe lung pathology, high viral load, and noticeable weight loss. The same sorts of prophylactic experiments on the hamsters showed that all the doses tried (0.5, 5, and 50 mg/kg) protected the animals from weight loss, with a strong decrease in viral sgRNA load in lung tissue at 5 and 50 mpk. A treatment experiment (as done with the monkeys) showed significant effects at the 5 and 50 mpk doses as well.

I was also glad to see an analysis of Spike protein mutants before and after antibody treatment in the monkeys. The good news is that nothing worrisome seems to have been raised up by the treatment – there were new mutant forms picked up, but these showed up in both the antibody-treated animals and in the placebo controls, which just shows you that the virus was indeed replicating in these animals (and throwing off mutations, as viruses do), but that the antibody treatment wasn’t causing anything new to be selected. There was also no sign of antibody-dependent enhancement in any of the treated animals.

What we don’t know from this study is the duration of prophylaxis: how long would someone be protected after a dose of the antibody cocktail? That really has to be answered in humans, though – it would be nice to get a number from a primate study, but you’d never be sure how well it would extrapolate to humans over such a longer time. But what this work demonstrate is certainly enough: the antibody cocktail appears to be effective both as a preventative and as a treatment, and it seems to have effects on viral replication that are in the same range as those seen in the vaccine studies reported to date (the more promising ones, at that). This is good news, and I am very eager to see human data.

54 comments on “Regeneron’s Monoclonal Antibody Cocktail in Primates”

  1. A Knowing Mess says:

    Potency may be a manufacturing issue if we actually need to hit 25 mpk of each antibody – 2 g per (80 kilo american) patient? Assuming a really good CHO cell clone (say, 8 g/L mAb production), and a standard 2000 L bioreactor, we’d need two 2000 L bioreactors to supply <8000 doses of the mixture. As a therapeutic for severe cases, Regeneron might be able to get away with it, but making enough for post-exposure prophylaxis or any sort of larger-scale use could be challenging..

    1. Joe Schmoe says:

      Regeneron is added 2x10k bioreactors in 2014 which was an additional 50% – so they have ~60k total now (assuming none was added). Should be plenty.

      1. TallDave says:


        As part of the agreement, Regeneron will amp up manufacturing to produce up to 1.6 million doses of the cocktail, dubbed REGN-COV2, as early as the end of summer, the company said.

        Because Regeneron is still investigating dosing for possible treatment and prevention uses, the drugmaker said the order covers anywhere between 70,000 to 300,000 treatment doses and 420,000 to 1.3 million preventive doses.

        1. A Knowing Mess says:

          I have faith in Regeneron’s manufacturing capabilities, but if they actually need to dose 25 mpk of each mAb, it’s not trivial to make 1.6 million doses in a hurry. Again, doing some back-of-the-envelope calculations, assuming 60kL reactor space + 8 g/L for each mAb + near 100% process efficiency, they’d need ~14 runs of the complete suite (with no bad lots). Assuming a 2 week production cycle + a half week for harvest & cleaning (probably pushing it for GMP manufacturing), they’d still need 35 weeks. I’m skeptical that they can deliver 1.6 million doses by the end of summer without either bringing on board substantially more capacity or dropping the human dose substantially below 25 mpk.

          That said, given the dosing requirements for other anti-viral mAbs for RSV and Ebola, I wouldn’t be surprised if Regeneron actually does need to dose 25 mpk of each mAb to see efficacy.

          1. Derek Lowe says:

            Exactly – that does seem to be around the right dose for an anti-infective application like this one. I hope someone’s at work on the distribution and the order of who gets dosed first, because those will certainly be issues, especially at first. The first one sounds like mostly Regeneron’s problem, and the second one – under a normally functioning system, which this isn’t, any more – would be more the CDC’s.

          2. TallDave says:

            seems a lot of partnerships underway to increase synthetic antibody production capacity, but yeah would expect primary focus would be ARDS treatment rather than prevention in the uninfected unless/until sufficient supply materializes after approval

            apparently it’s not clear precisely what anyone’s committed mAb production actually is at any point in time, but note mfrs are generally not waiting for approval

            of course even 100K doses could significantly reduce US deaths in the near term, if efficacy is as high as hoped

          3. TB Anaconda says:

            Regeneron have teamed up with Roch to triple production. Regen (45K L capacity) & Roche (100k L capacity).

    2. Marko says:

      With the $450 mil. kick-start they got from BARDA , I suspect they’ve procured a few extra bioreactors. Back in April , the REGN CSO said this :

      “…George D. Yancopoulos, MD, PhD, Regeneron co-founder, President, and Chief Scientific Officer, added that Regeneron is committing its Industrial Operations and Product Supply manufacturing facility in Rensselaer, NY toward manufacturing the antibody cocktail, “which on its own could supply hundreds of thousands, if not over the course of time, maybe even on the order of a million or so doses per month.”

      Assuming the treatment significantly reduced death rates in severe cases , even only tens of thousands of doses produced per month for therapeutics could have a big impact , considering we’re now running at about 30k deaths per month. Similarly , if prophylaxis was only used for high-risk populations ( like HCWs ), and if the ab lasts on the order of months as does natural IgG1 ( minimally ), it sounds like they may have more than sufficient capacity to make an impact.

      1. TallDave says:

        yep, and CDC midpoint of best estimates for new hospitalization seems to be around 10K/day

        allegedly Regeneron and Lilly both already have hundreds of thousands of doses available, the exact number depending on dosages which are TBD based on the ongoing trials

        so w/September approval, either or both together might conceivably be able to treat all serious US hospitalizations in October and onward, assuming they can ramp up, and that new hospitalizations are stable or falling

        the end of the pandemic

        1. Charles H. says:

          Re: the end of the pandemic

          Well, no. Since much of the spreading is by asymptomatic carriers, this won’t be the end of the pandemic, but it will make it a lot less serious. Unless that study of heart irregularities among mild cases is really a heads up about a serious problem.

          OTOH, it may also turn out that the problem with COVID is less the people that it kills, and more those it permanently injures. That’s a problem that seems to have been largely ignored, and I really have no idea how serious it is, but we’ll be finding out over time.

          1. TallDave says:

            technically yes, but if the risk of dying/injury falls by >90%, then the pandemic is probably over for most practical purposes… we’d be back in seasonal flu territory, even for strokes (untreated COVID risk seems to be about 8x higher)

            agree the low-symptom carriers are likely to cause cases to trickle in for some time, hopefully supplies are soon sufficient to being treating contacts as that would presumably push new infections down quite a bit faster

            also agree we are likely to find some really unpleasant long tails, so hopefully we can treat those mild cases as well

          2. confused says:

            Yes, the pandemic being over doesn’t require cases to go to zero; pandemic flu viruses (like H1N1) have continued to circulate as seasonal flu viruses even after the pandemic was officially over. The 2009 pandemic strain of H1N1 is still out there.

        2. Chris Phoenix says:

          There are serious long-lasting health effects in a large proportion of _all_ hospitalizations. If you want to have a healthy population 5 years from now, better treat them all.

          Also, better have a robust health system that can diagnose / test everyone with symptoms, so they get into the treatment pipeline in time. That means tests that come back in a day or two, not a week or two – and are accessible to people without health insurance.

          I think it’s a little early, especially in the U.S., to think we can treat our way out of this.

          1. TallDave says:

            yeah, a lot of marginal outcomes may hinge on exactly how many doses are actually available

            but no OECD or middle-income country should have much trouble administering synthetic antibodies to the infected once they’re available

            agree it’s too early to know anything for sure yet, but an effective treatment is a bridge to vaccination… soon human trial data will give some insight into whether the efficacy is high enough

    3. Anon says:

      No way is a sub nanomolar potent antibody (science paper) with intact Fc going to need to be dosed at 25mpk for prophylaxis. A dose of 1mpk, perhaps every few months is more likely (say 400 mg/year/patient). Rapid lung access for treatment might need IV dosing, but prophylaxis will not.

      1. Barry says:

        prophylaxis requires lung access as surely as treatment would.

        1. Anon says:

          15% partition from plasma to lung, 5% to the nasal mucous. Their published IC90s are very impressive – sub 1nM means 0.15 mg/L in lung tissue or 1 mg/L in plasma. Clearance of a typical mAb is about 0.3L/day. Dose to maintain1 mg/L average plasma concentration is only 3 mg/day or 100 mg/month. I’d say prophylaxis is on.

  2. loupgarous says:

    Thanks for sharing this, Derek!

  3. Marko says:

    I’d love to know if the Fc region is required for the demonstrated activity or not. I’m sure Regeneron must have looked at full IgG vs Fab or F(ab)2 fragments at some length in their research , but it sure would be nice to see what these macaque results would look like if the Fc region was clipped off. In other words , is this entirely about blockage of spike/ACE2 , or is there more going on ?

    Whatever , these results are encouraging re: antibody therapy vs CoV2.

  4. bacillus says:

    As “a knowing mess” indicates, this is going to be technically and economically useless for prophylaxis. Also, the therapeutic arm was only tested 1 day post-exposure which might be okay for a known laboratory incident, but potentially useless for someone who only shows signs of severe infection several days to weeks after exposure to the virus. As with the NHP vaccine trials you covered in an earlier post, nearly everyone is trying to rig their experiments by choosing either low dose virus challenges and/or small intervals between vaccination and challenge/treatment. This is problem throughout the field of vaccinology and no good will come of it. I can only hope that the FDA demands data on long-term benefits of any vaccines (e.g. 6-12 months out) before approving them. The eyes of the world are on these experiments and if none of the early hyperbole pans out in Phase 3, then expect a long-term deleterious effect on vaccination as a whole. As I stated in an earlier email elsewhere, we’ve only got one chance to get this right

    1. Cira Goettig says:

      Re 1 day post exposure being unrealistically early – the paper adressed that; the disease progression is faster in these models than in humans, (if I recall correctly, the entire disease duration was something like 5 days) and they are being treated around the time of peak viral load that way. In humans this would be occuring just as the symptoms start, so they think its a plausible model for getting treated right around the time of hospitalisation.

    2. confused says:

      Would it really be worth it to wait 6-12 months *after* the start of Phase III trials? With Phase III for the first vaccines starting just a week ago or so, that would mean late January to late July of next year. At this rate, I’d expect the epidemic in the US and Latin America at least to have largely burned out by then.

    3. Nope says:

      Nonsense. The 1 day post-exposure treatment is due to the nature of the animal models. The infection in monkeys is fast and short, unlike in humans. It’s like a 5-day duration. So treating after 1 day makes sense in this model. Treating after 5 days would be like treating a human after a month.

      This isn’t a vaccine, and nothing was rigged here.

  5. hse says:

    Somewhat off-topic, but does anyone know why Covid-19 appears to be so much less deadly in elderly Japan than in slightly younger Italy, especially since it’s unlikely that Italy’s cases are under-counted relative to Japan’s with regard to testing per capita? Perhaps genetic variation in virus or human population, or differences in elderly CVM health status or in mask/social distancing behavior?

    1. Marko says:

      Evidence has been accumulating that suggests that when you get infected while wearing a mask, you get a lower inoculum and less severe disease than in those not wearing masks. In Japan , wearing a mask has long been a fashion statement , while I suspect the Italians only started wearing them in earnest after the bulk of the deaths were already baked in.

    2. Guppy says:

      I’m guessing lower rates of co-morbidities such as obesity.
      Cardiovascular health overall is not the same, for instance patterns of stroke types are different for East vs West, with Westerners typically having more thrombotic vs hemorrhagic strokes (and more CVAs overall), while East Asians have more hemorrhagic strokes.

    3. 10 Fingers says:

      One other factor may be the multi-generational living/childcare situation that is common in Italy and thought to be a key component of disease spread there. I’m not sure how different that is from the situation in Japan, but I work closely with folks in northern Italy and they anecdotally confirm this aspect.

    4. Riah says:

      There is a paper that discusses exactly this: “Why does Japan have so few cases of COVID-19” (by Akiko Iwasaki of Yale whose webinar I attended last week presenting her new research on the immune response in severe COVID-19 and also sex differences. I would also recommend the rest of her research incidentally- very interesting).

      re Japan:

      1. Riah says:

        The paper I linked above doesn’t consider this further factor which, having looked into it, I now think is far more likely to be the main reason. T-cells are absolutely dependent on vitamin D to activate from the naive state

        It has also been demonstrated by copious research, summarised in this paper at section 5:

        that immunomodulation, cytokyne balance,1st line physical barriers, control of blood clotting and production of antivirals, cathelecidin and defensins, are also highly dependent on vitamin D.

        Japan’s elderly have the highest levels I have come across in elderly populations: 95% of active elderly have incredibly high levels of over 30ng/ml which are way higher than the woeful UK population average of less than 20ng/ml:

        Compare this to Northern Italy, the vitamin D blackspot of Europe where 76% of older women aged 60-80 had levels below 12 ng/ml!

  6. ezra abrams says:

    the knowledge here about mAb production is astonishing

    it maybe to $, but given the scale of the economic destruction, if the mAbs actually work, it would be *cheap* for a sane Fed Gov’t to invest, say, order of mag, 200 million dollars for production scale up.

    Heck, if there are any economists here, it might be cheap at 5x that

    1. Anon says:

      “it would be *cheap* for a sane Fed Gov’t to invest…”

      I think you answered your own question. Don’t see any rational leadership on the horizon.

      1. Hap says:

        The problem is that there aren’t enough rational voters, or at least it’s too easy to get people to forget rationality when they decide who to vote for. We can’t think ourselves out of a position we didn’t think ourselves into.

        1. TallDave says:

          scroll up, Feds already throwing bales of money at this

          1. J N says:

            No doubt at $75,000 per prophylactic regime it will be a bargain.

            Even $200,000.

            The latest trend in US government is to demand a vig for investment so this all sounds reasonable.

          2. TallDave says:

            don’t know the exact contract details, but from the article 1.6M doses at $450M would work out to $281 each, which the Feds would apparently cover

            if it works it’s a bargain from the latter’s perspective, given they also have a few trillion in GDP to consider 🙂

  7. ScientistSailor says:

    “Now a virus needs two separate escape mutations, which is extremely unlikely.”

    I have heard (in a talk by a reputable virologist) that the mutation rate multiplied by the error rate in a viral infection is such that each double mutation is sampled multiple times per day (at least for HCV). Hence the triple combinations.

    Does anyone here know if the same is true for coronavirus? Does Regeneron need a 3-mAb cocktail?

    1. John Wayne says:

      The speed with which a virus will find a way around a given therapy is related to:
      – The number of mutations needed to generate a viable virus
      – The error rate of replication at the time of therapeutic pressure
      – The total viral load at peak infection

      In the case of HCV, it has a 10 to the -4 error rate and a viral load of around 10 to the 11th (pulling this from memory). So a single mutation is sampled 10 to the 7th times, two mutations 10 to the 3 times; three really potent agents can shut it down.

      I have no idea what the numbers are for COVID

    2. Barry says:

      Coronaviruses are big in part because their RNA polymerase does have some proofreading. Covid’s error rate is less than HIV’s. I don’t know how much less.

    3. TallDave says:

      fortunately not all combinations of mutations are equally likely, so in theory with sufficient understanding of a virus one could deliberately select two antibodies whose targets are so unlikely to occur together that there’s basically no chance of a double escape mutation

      1. Hap says:

        Or if the mutations on their own are negative – if the mutations individually are detrimental to reproduction or survival, then it will be difficult for anything to develop both mutations because there won’t be enough viruses with one to develop the second. I don’t know if those mutations are actually observed in practice, though.

  8. TallDave says:

    oh great, I was just reading the biorxiv on this, was hoping we’d hear from you

    on the preventative note, Lilly yesterday announced they’re doing a Phase 3 in nursing homes in addition to the ongoing Phase 2

    emergency use authorization for these could potentially bend the death curve way down until the vaccines finish it off

    btw the Chinese Antibody Society tracker seems to be a fairly good resource, they list several others as well

    1. SEC says:

      “on the preventative note, Lilly yesterday announced they’re doing a Phase 3 in nursing homes in addition to the ongoing Phase 2”

      Along similar lines, Regeneron is expanding its trial to patients who aren’t sick enough to go to the hospital.

      I’m not in this line of work, but is this evidence that the ongoing trials in hospitals are going really well? Not going well? Impossible to tell?

      1. TallDave says:

        imho evidence of nothing

        1. SEC says:

          Thanks for the reply.

  9. Fonli says:

    On the issue of mixing two monoclonal antibodies, I wonder what happens in the case of vaccination. Do vaccines generate more than one antibody? Inactivated virus vaccines certainly do but what about mRNA, DNA, subunit platforms? If a vaccine elicits one single antibody escape mutants will also complicate things.

    1. BioPhD says:

      All vaccines (attenuated virus, mRNA, DNA) introduces an antigen, that will then raise a number of different antibodies against it, so this particular issue shouldn’t be too much of a problem. Even the mRNA or DNA vaccine produces a whole protein, or part of one, that contains multiple epitopes (regions that are recognized by antibodies).

  10. Liz Jenny says:

    Most people who get infected recover and don’t need intervention/-and certainly not an unaffordable one. Does the animal model reflect this heterogeneity in outcome that is seen in humans?
    Predicting who is at risk for severe disease would help assess vaccine efficacy in a relevant group.

  11. Scott Stewart says:

    I should know this but I don’t. If a) the mAb works and b) someone has a mAb on board and gets exposure, does the mAb typically prevent seroconversion and acquired immunity?

  12. Riah says:

    Thanks so much for this Derek. I somehow instinctively prefer the idea of monoclonal antibodies to a vaccine that is being rushed through so quickly. Re vaccines, I have concerns about safety testing (or lack thereof against true placebo), duration and quality of safety test monitoring (especially given the hot competition between the various contenders) and about the adjuvants used. What are the downsides, if any, to monoclonal antibodies (apart from the obvious one of them not being long lasting?)

    1. Riah says:

      The Oxford Recovery trial just announced a 4,000 patient RCT of Regeneron – for anyone interested

  13. JasonP says:

    NovaVax reports on its P1 results, includes some primates data too!

    Preprint is suppose to be out there on medRxiv.

    NVX-CoV2373 was well-tolerated and had a reassuring safety profile.
    Overall, the vaccine was well-tolerated and reactogenicity events were generally mild. Following Dose 1, tenderness and pain were the most frequent local symptoms and systemic events were individually less frequent with headache, fatigue and myalgia being reported most commonly. As expected, following Dose 2, greater reactogenicity was reported, although the majority of symptoms were reported as ≤ Grade 1. The average duration of events was < 2 days.

    Unsolicited adverse events were collected through 28 days after Dose 2. There were no severe (Grade 3) unsolicited adverse events, and the vast majority of adverse events were mild and deemed not related to vaccination. No serious adverse events (SAEs) were reported and safety follow-up continues.

    NVX-CoV2373 induced neutralization titers in 100% of participants; 5 µg adjuvanted dose group peak GMT: 3,906 (95% CI: 2,556; 5,970).
    All subjects developed anti-spike IgG antibodies after a single dose of vaccine, many of them also developing wild-type virus neutralizing antibody responses, and after Dose 2, 100% of participants developed wild-type virus neutralizing antibody responses. Both anti-spike IgG and viral neutralization responses compared favorably to responses from patients with clinically significant COVID‑19 disease. Importantly, the IgG antibody response was highly correlated with neutralization titers, demonstrating that a significant proportion of antibodies were functional.

    Matrix-M™ adjuvant induced robust polyfunctional CD4+ T cell responses.
    The adjuvant was dose-sparing, with the lower 5 µg dose of NVX‑CoV2373 performing comparably with the 25 µg dose. Cellular immune responses were measured in a subset of participants, and NVX‑CoV2373 induced antigen-specific polyfunctional CD4+ T cell responses with a strong bias toward the Th1 phenotype (IFN-g, IL-2, and TNF-a).

  14. An Old Chemist says:

    NIH to test COVID-19 antibodies in hospitalized patients—starting with Eli Lilly’s

  15. DrivingOnBy says:

    Cocktail? May I have Gin with that?
    Gin with quinine, of course.

  16. Will says:

    Cytodyn’s leronlimab is already kicking crap out of cv19 in ph 2 and 3 m2m and sc trials. They’re a small company not getting alot of recognition but they’re results are amazing with zero side effects. There are 65% fewer sae and ae in treatment than in soc placebo.

Leave a Reply

Your email address will not be published. Required fields are marked *

Time limit is exhausted. Please reload CAPTCHA.

This site uses Akismet to reduce spam. Learn how your comment data is processed.