Thank you very much for the introduction. As James mentioned, I am at Gilead. I’m part of their protein therapeutics group in early R&D.

We have two high-throughput protein production workflows there, one a small-scale plate-based workflow, where we are using the Dynamic Devices’ VVP to aid in the transfections and purifications. And then we have a TubeSpin 50 mL Bioreactor-based production workflow. And what I’m going to be talking about today is some of our recent efforts using Dynamic Devices’ MagLynx instruments to automate purifications from those bioreactors.

So when I say these 50 mL bioreactors, what do I mean? They’re essentially a 50 mL Falcon tube that has a cap that’s vented to allow gas exchange, as well as a septa to allow easy addition of cells or reagents, media, whatever it may be, without having to decap and recap the tubes.

And so in our TubeSpin-based workflow, the goal is really we want to produce a couple milligrams of protein that’s purified and QC’d that can be used in downstream assays, whether that be for binder characterization, biophysical characterization, or functional analyses.

So in order to do this, we essentially do the transfections in these bioreactors, then allow them to express for about five days before prepping them for purification. And prepping them requires centrifuging them to pellet the cells and then filtering them to remove any extraneous debris so that it doesn’t clog the active systems that we’re using for purification. And for our standard monoclonal antibodies or symmetric molecules, we’ll use a 2D active method, where we first run them through a pro-A column followed by a desalting column to put them into a formulation buffer.

But one of the things we’ve noticed is when we went to 150, things worked decently well. But as we wanted to scale beyond, we hit a couple of pain points. And this became a real issue as we have the imperative to now scale to 300 per week, where we’re at about 150 per week now.

And those pain points are that the centrifugation and supernate clarification is incredibly laborious. It takes a lot of time and takes a lot of consumables. Not only do you have to centrifuge everything, then somebody has to pour it through a filter, run it, label it, and then hand it off to the group that’s doing the purifications. And then the purifications themselves are relatively time-consuming. This 2D method takes about 25 minutes per sample. So just 150 times 25 gets you to about two days of pure ÄKTA time, as well as there’s a variety of different species, FCs, and isotypes that were processed. And not all of them are compatible with the columns that we’re using that enable this fast flow rate. So if we move to a different FC that’s not as compatible, it has to be processed a different way, which further extends the time.

And then also with the active base purification is it’s not very scalable. If we want to go to 300, we essentially have to double the number of ÄKTAs we have and double the number of scientists we have who are running them. So, you know, it might work for 300, but if we think beyond scaling beyond that, it becomes a little more tenuous to justify.

Fortunately, from the small-scale workflow in my previous experience, I knew that there is a strategy that would alleviate these two pain points. And that would be if we could do magnetic bead-based purifications of our bioreactors.

One of the really nice things about the mag bead purifications is that it doesn’t require removal of the cells. Because you don’t have to worry about fouling the cells or them clogging, you can simply remove the media and the cells during that first step of the purification and then wash and go to your elution.

Additionally, while in theory you could do mag bead-based purifications in a sequential manner, there’s no reason to. There already exists great instruments and methods to do it in a parallel manner. We do have an instrument that’s designed for doing mag bead-based purification. It doesn’t meet what we are looking for. Its throughput is not good enough. It can only do about 40 to 60 per day. Additionally, it’s not fully automated. It will only do the cell and media removal and washes and add the elution buffer. Requires manual intervention to remove the eluted protein from the beads and neutralize if you are doing a pro-A elution.

And so, what we were really looking for is an instrument that could do a higher capacity up to 150 per day per instrument.

As well as fully automated, we want to be able to set everything on the deck, walk away, and come back with eluted protein. And preferably that would be in a plate, a 96-well plate. And that’s for logistics. It’s much easier to work with one or two 96-well plates as opposed to having to hand off 150 tubes.

Fortunately, at this time, through some conferences and talking with people, we had heard that Dynamic Devices was developing a new instrument called the MagLynx that were specifically designed for this. And so we were very fortunate to get to work with them and really kind of help test out this alpha instrument and show that it will work for our purifications from these bioreactors.

And so how this works is essentially it’s a standard LM900 deck. It has a number of iMagZs on it. Thanks for introducing those, that was great. As well as spaces to put your reservoirs and tips and destination 96-well plates.

And so we’re sort of showing you over here is just an overview of the deck. You can see there are also holders for the bioreactors. One thing we learned early on– shout out to Scott, who I think printed two or three different versions of these– is not every 50 mL Falcon tube is built the same. And so some of the original ones we had had slightly different geometries that didn’t quite work. And so Scott was able to 3D print some new ones that are fully compatible with the bioreactors we’re using.

And so in addition to all the labware on it, there’s also an MCPA-compatible head that is for aspiration mixing by sparging and dispensing of wash buffers. And then there is a 6-channel standard pipetting head that, along with the 24-well tips, can add elution buffers and transfer from the tubes to your destination plate after your elution has occurred.

And so, what I’m showing here is essentially the aspiration dispensing mixing head going down into tubes. And then in the bottom left are the iMagZs, which can be elevated and lowered. And so how this would work is we’d put a tube with magnetic beads into the holder on the nest. We would then elevate the iMagZ, which will pull all the beads in a circle around the bottom of the tube, leaving the very bottom free. We can then lower the aspiration tool, aspirate off all the cells and media. And then from these little needles up here, we can essentially dispense our wash buffer. And there’s a column switching valve, so we can use different wash buffers.

For our protein A elutions, we tend to use just PBS as our wash buffer. And it seems to be working well. But that is an option.

Yeah.

And so, from my previous experience with mag beads, one thing is I knew is that not every mag bead is created equal. And while they all have very high dynamic binding capacity, it really is a question of how long do you have to incubate your sample with the beads in order to get complete binding. So, as we were moving forward and developing a protocol to really put this through its paces, one of the first things we wanted to look at was, how did incubation time affect our ability to recover a protein?

In order to do this, we essentially just expressed a large amount of antibody and then took some of that and ran it through our standard purification, which is the filtration and then on the ÄKTA. And then we took the rest of it and split it. And some of it we incubated with mag beads for two hours, some we incubated with four hours, and some we incubated for 24 hours. And what you can see here is you see a pretty linear increase in the amount of protein that you’re recovering as the incubation time increases.

And you see all of them compare fairly favorably to the ÄKTA. But one thing to combine, it’s not quite an apples-to-apples comparison. It is a 2D step on the ÄKTA, which does involve some loss. However, when we go to our 24 hours, we are capturing about 70% more material than the ÄKTA. So even when you account for that loss, we would still be ahead.

And protein yield by itself is not enough. Quality makes a huge difference. If everything we’re capturing is aggregated, it doesn’t matter. No one’s going to want to use it, and this would be dead in the water. Fortunately, that is not the case at all. We ran all of these on analytical SEC, as well as by SDS Page, and looked at their endotoxin levels as well. And so what you can see here, I’m showing for their various bead incubation time points how their analytical SEC profiles compare. And they’re essentially indistinguishable from what was purified on the ÄKTA. Additionally, you can see from the SDS Page gel that the quality and purity is fairly similar. There are some small, very minimal contaminating species in the MagLynx that you don’t see in the ÄKTA. But they’re present at a very, very low percent, and they’re not a cause for concern.

Additionally, in terms of endotoxin, we do see some level of endotoxin, and it does start to accumulate. And so, what we realized is that by introducing weekly washes with sodium hydroxide, we’re able to control the endotoxin levels, and they don’t accumulate, and they stay below our thresholds for cutoffs.

Additionally, one of the things I’m not showing here, but we tested, is that the mag beads that we’re using can be regenerated and used again. So, in comparison to something like a resin-filled tip, which is often used once and then discarded, these can be regenerated multiple times and used, which is a pretty significant cost savings. And it was another big plus for the instrument.

As I alluded to before, we don’t just work with a human G4 in a body or a human G1. We do different species, different isotypes, and so we really wanted to know, could our mag beads work with all the different species and isotypes? And so, in order to test this, we essentially just spiked in some purified protein that we’d previously made into media with the magnetic beads and then ran it over the MagLynx. And what we saw is, yes, all of them do recover protein. We do see less recovery for mouse immunoglobulins as well as rabbit immunoglobulins, but hat’s pretty common for IgGs. There’s nothing surprising about that.

And also, while the bulk of the purifications we do are with things that have an FC on them, we do occasionally do his tag constructs as well. And we wanted to make sure that the MagLynx was compatible with nickel-NTA magnetic beads.

Fortunately, we have some constructs that have both an FC and a his tag on them. And so we can do a direct head-to-head comparison, as well as we have a number of different antigens that have his tags that we can look at. So again, just spike in experiments where we spike in the protein into media, incubate with the mag beads overnight, and then run on the MagLynx. And what we see is we are able to recover with both the nickel-NTA beads and the pro-A beads. You do see there is a loss of recovery with the nickel-NTA. And I strongly suspect, and we’re working on collecting the data to validate it, it’s because the protocol to this point has been optimized for pro-A elution.

This was sort of a very quick sanity check that it will work, but it’s not optimized in any way. And I suspect that as we spend a little more time and improve our nickel-NTA purification process, we’ll narrow that gap between percent recovery.

So having validated that, I mean, it works. We found a good incubation time, number of different species and isotypes it’s working for, as well as different bead types. Well, the last thing we were curious about was: what was the kind of reproducibility across the various nests?

And so there are eight nests, and each holder that lives in one of those nests can hold six samples. So that’s 48 per run.

And so, we wanted to see how did the percent recovery, how did our percent recovery or yield compare across all of those 48? So again, we just did a spike-in experiment with a large volume, allocated it across 48 tubes, and then ran it on the MagLynx. And it looked pretty good. We saw something in the middle that looked like it might be a systematic error. And so to dive into that in a little bit more detail, we then stratified the data. We have two ways of looking at it for potential systematic error. One of them is by perhaps one of those aspiration or dispensing needles is not as good as one of the other ones. And so we can look at all of those as a group. Or perhaps one of the deck locations is just not functioning quite as well as the other ones. However, when we do that, there’s nothing that quite stands out, so it doesn’t look like we’re seeing any kind of systematic error.

And with that, we were fairly comfortable with how the MagLynx was working. And really, the next step was to put it through its paces and see how it compared to the– sorry, how it compared to the ÄKTA 2D system.

And so in order to do that, we essentially took 48 molecules, 24 of them that are symmetric. And so that would be like an IgG or a VHH-Fc, and then 24 that are asymmetric. So these could be a mutine or a bispecific.

Expressed them in duplicate and ran them either over the MagLynx or over the ÄKTA.

And so what I’m showing you here is just the symmetric molecules. And we subdivided those as well, half of them getting a four-hour incubation time, half of them getting a 24-hour or overnight incubation time. And that was us deciding how we wanted to make our workflow, either adding our beads on the morning and then starting our purifications that afternoon on the MagLynx, or essentially adding the beads and then letting them go overnight and starting the next day.

And what we see is that with a four-hour incubation, you get pretty equivalent yields between the ÄKTA and the MagLynx. However, this is, again, not buffer exchange. So there would be some losses. Whereas when you move to an overnight incubation, again, you’re capturing more material with the MagLynx than the ÄKTA. And so even after you account for loss due to buffer exchange, it still put us in a fairly nice place.

We were also interested in how the quality looked like. So again, ran them all on analytical SEC. And what we saw is that the quality between the MagLynx and the ÄKTA compared very favorably. There were some samples that were slightly more monodispersed by the ÄKTA, some that were more monodispersed by the MagLynx. But all of them were within a couple percent, which is what we put as our threshold cutoff.

There’s a slightly different story for the asymmetric molecules. Here we see we’re capturing more at both 4 hours and 24 hours with the MagLynx.

However, when we go to look at them by SEC, there’s a pretty large discrepancy between the monodispersity for things coming off of the ÄKTA versus the MagLynx. And this is likely because, on the ÄKTA, we’re doing different types of washes that are designed to remove aggregated protein and misassembled protein. And we haven’t incorporated anything like that onto the MagLynx. It’s one of the next optimizations is to do multiple different buffer washes instead of just PBS, PBS, PBS.

However, for our TubeSpin workflow, 80% to 90% of the samples that we purify are symmetric molecules. And so, we are very comfortable in putting this forward and using it in production for everything that we’re making at TubeSpins that is symmetric.

And one of the big reasons for that is that it really does help us hit that 300 samples per week goal. And so here I’m showing your timeline of what it looks like for purifications for the ÄKTA, where you have to prep your filters on Friday. You have to filter everything and prepare it on Monday and hand it off. And then you can start your purifications on Tuesday, and they basically run through Tuesday, Wednesday, and some of Thursday. Whereas with the MagLynx, you add your beads, you purify the next day, and then you can buffer exchange. If you want to do 300, you just offset your transfections by a day, and you do a second run of MagLynx. And one of the big ones is the hands-on time. We’re spending about 10 hours doing all the work to prep things for purifications and the purifications. We can cut that in a third using the MagLynx.

And so this is just reiterating exactly what I told you before. And I just want to thank everyone at Gilead who helped out with this, as well as Scott and Toby, who were instrumental in getting this alpha instrument up and running and working with us to get a protocol and all the little hiccups that occurred along the way. Thank you guys very much.