Thermo Fisher Scientific Inc. (TMO) Earnings Call Transcript & Summary

Hello, and thank you to everyone attending our BPI Europe Presentation, Learnings from Perfusion Medium Development. I'm Christopher Brau, a Staff Scientist and Engineer at Thermo Fisher Scientific's Grand Island, New York site and Team Lead on our high-intensity perfusion CHO medium project, or HIP CHO, as we like to call it. We recently gave a presentation at BPI digital week that was more focused around the perfusion process itself, different types of applications and their benefits. And we'll touch on that again in a future knowledge culture seminar around September. Today, I'm going to focus more on the hurdles in development and medium testing necessary to support it. With 20 minutes and some aspects being trade secret, I'll have to keep things pretty high level. But I hope you'll find the discussion beneficial and feel encouraged to think of Thermo Fisher Scientific when looking at how to bring your molecules and production plans into the future. Let's begin. Perfusion involves continuously exchanging medium, usually expressed in units of vessel volumes per day, or VVD, with around 2 to 3 vessel volumes per day being pretty common. It also needs some kind of mechanism to retain cells and keep them from being lost in the removed spent medium. That mechanism consists either of some kind of settling method or a hollow fiber filter-based approach. It's generally desirable to have a lower medium exchange rate, both to make medium handling easier and also to reduce flow rate across your retention device, which may allow higher operating efficiency. The ratio of medium exchange rate to cells being maintained is usually referred to as CSPR, or cell-specific perfusion rate, usually expressed in PICO liters per cell per day, and gives you a loose idea of the medium exchange rate to viable cell density ratio needed during operation, effectively a VVD-to-VCD ratio. Although in practice, it's generally better to scale exchange rate based on viable cell volume, or VCV, as opposed to viable cell density. More on that later. In continuous perfusion and on rare occasion intensified fed-batch perfusion, a continuous bleed is used. This involves pulling medium directly from the vessel in order to remove cells so that we can maintain the culture health. It is usually desired to minimize cell bleed as a small bleed often doesn't make economical sense to clarify and route to downstream for further processing. Thus, any product in it is lost. Bleed rate is typically described as a percent bleed, which refers to the percent of VVD medium flow rate that is directed towards bleed instead of towards spent medium being removed. Bleed essentially maintains culture health in 2 ways. The obvious benefit of a cell bleed is that it can keep the viable cell density low enough to avoid excessive nutrient loss and waste buildup. Both of these can hurt productivity or cause a surge in death rate that might crash your run, and excessive waste buildup may follow your retention device faster. The other benefit of bleed is that it can push to a specific percent viability to maintain a consistent quality target. This happens partially because cell growth rate reduces as VCD increases and vice versa. At steady state, your VCD is constant, and your growth rate is directly balanced by death rate plus the cell removal rate from bleed. If we increase the bleed, the VCD will be reduced, and growth rate will increase until a new steady state is reached. At this higher constant growth rate, but the same or lower death rate, we get a higher percent viability. Thus, if you want to limit cell bleed but still maintain high VCD and high percent viability, what you really need to do is keep a low cell death rate as the VCD climbs higher. So ideally, if perfusion medium needs to support growth rate well in the seed train and maintain cell health as you climb the higher cell density and the cell growth rate slows down. We also need to consider bleed with regard to cell retention mechanism. Recall that for suspension cell culture perfusion, we're looking at either filter-based or settling-based methods to retain cells. But it's generally only practical to use settling-based methods on very small-scale perfusion models. When developing a medium in ways to test it, we need to think about that carefully because different cell retention mechanisms do not behave the same way, and that will impact how your culture behaves. One of those differences is the inherent bleed or percent of cells lost in the spent medium when using settling-based methods. All settling-based methods lose some cells in the spent medium. Filter-based methods do not. That means you need to be very careful of 2 different cell retention methods are used. For example, an intensified fed-batch run with a centrifuge generating a 10% natural bleed will have a slower rise in VCD and a lower but longer peak VCD with a slower drop-off. That same run with a filter-based method will generate a very different result. In the figure here, we have a percent bleed showing for 12 different Ambr 15 conditions running a perfusion model. The bleed is caused by a combination of a direct 3% cell loss due to daily culture sampling plus the cell loss to spent medium during centrifugation or natural cell loss in the spent media. The average of these conditions is around 6.5% with the manual operating nature causing quite a bit of variance. The bleed can be made more uniform by adding an additional bleed step after analyzing the cells lost in the spent medium in order to get a consistent total bleed amount. However, an immediate cell loss once per day has a much larger impact than a continuous cell loss. Just like investing $10 upfront, will return more than investing $10 spread out over time, or pounds or euros. To complicate things further, recall that cell growth slows down when VCD increases. So this jigsaw effect of suddenly reducing the VCD each day may have additional metabolic impacts versus a continuous perfusion operation. In fact, there's actually quite a lot of things wrong shared across most scale down perfusion models when compared with bench scale and higher perfusion, where you can use continuous fluid handling and the same retention mechanisms as you plan for pilot and production scale. Some of these can be mitigated better than others. Some of them are sneaky. Like the fact that you can find your gas mass transfer rate limited in spin tubes and deep well plates when working with more aggressive cell cultures and perfusion modeling. It took a bit of effort for us to develop a well-played perfusion model that doesn't restrict our better performing cell lines due to insufficient gas transfer. In the plot above, we have some early well plate studies on the 2 left-hand sets of plots, where we did a rapid blast of multiple maintained percent bleed conditions at 1 relative volume per day in 2 different cell lines. Cell line A, which is a bit lower performing, has somewhat of a spread between bleed conditions, though we can see it being a little limited at the lowest bleed rates. By comparison, cell line B, which is much more aggressive, becomes smashed together at multiple bleed rates and extremely limited, which is also reflected in the poor viability. To the right, shaded in gray, all conditions in here are run at 6% bleed with varying adjustments to optimize gas mass transfer KLA with varying degrees of success and the best options moving towards 100 million cells per ml versus the very stunted 40 million cells per ml that it was limited to previously. We'll still need to perform follow-up work to prove if cell line B is no longer limited in these newer models. You may have noticed I said relative volume in the well plate work. That's because another aspect that is very difficult to reconcile is an instantaneous medium exchange versus a continuous one. Performing a full volume exchange at a single point in time every day is not at all the same as performing a 1 VVD continuous medium exchange, where not all the waste is removed. Mathematically, we only remove about 63% of the medium after spinning down to better match waste and spent medium left in the vessel after 1 VVD continuous medium exchange. This isn't perfect, but it does get us considerably more representative versus a complete instantaneous medium exchange. From an operating perspective, a practical limiting factor is the manual effort required to handle these scale down models. It makes operating at higher than 1 relative volume per day troublesome if someone must come into the lab late at night, usually a different person than in the morning. So we're very hopeful about the idea of increasing automation by using passive settling in the Ambr 15. But in practice, with moderate and aggressive cell lines, if we settle long enough to keep the bleed manageable, it affected the cell line behavior due to oxygen depletion, among other things. But if we set our settling time to avoid excessive oxygen depletion, we get far too many cells bled off in the spent medium. In the figures above, you can see considerable penalty one of our cell lines saw when implementing passive settling compared to spinning down with a centrifuge. All of these challenges may seem overwhelming, but in practice, we just have to be careful and prove out where things get bent versus where they get broken so that we can make reasonable decisions with the data. So the scale down models are wrong. But if we're careful, they're still useful. For designing a perfusion medium or any medium really and for testing and comparing it, we don't need a perfect model. What we need is to generate an environment that challenges the medium in a similar enough manner that the rank order of our experiment results agree with more representative bench scale work and higher. As you can see, we've worked with spin tubes, flasks, well plates in the Ambr 15 in multiple approaches and compared those results against our bench scale work with actual continuous perfusion. The plot above shows optimized Ambr 15 spin down model results of a comparison between a few different medium conditions at 1 relative volume per day. In this case, we're looking at spent medium IgG concentration and can see that the green triangles condition starts off well, but then the performance falls away. You hope to catch this early on, but you need to prove out study state to be sure. And this is the case with the top condition in purple squares. It seems stable here, but will eventually drop in productivity when steady state is finally achieved. In this case, the second best condition, the Blue Diamond condition, actually ends up being the best performer, not perfect, but these test results narrowed us down to 2 candidates instead of 6, useful. After considering the needs of an upstream perfusion workflow as a whole, 2 primary goals were identified for HIP CHO medium design: ease of use and cost-effective performance. Towards this end, the target was a chemically defined animal origin-free protein-free formulation. The AGT format was chosen to support fast and simple reconstitution at scale up and a 12-month shelf life. The formulation needed to function both as a seed train medium and a production perfusion medium and be flexibly capable of supporting high cell densities and productivity at low 1 VVD medium exchange rate but also be able to run diluted, if needed, for quality purposes or in cases where a cell line might not like a rich medium and struggle with growth rate when operating at low VCD. All this while supporting cost-effective scale up and avoiding any components that could cause a sourcing risk. We started the development of this formulation by creating a panel of 10 perfusion media, a mixture of existing panel media, fees experimentally determined components and ratios. This panel was tested in-house via our early spin tube model and by external alpha testers with top candidates verified at bench scale. From there, a series of experiments were designed to further optimize the formulation using the spin down perfusion Ambr 15 model. In addition to this, [ Omex ] work was used to facilitate component selection and concentrations during the final design of the experiment phase. At this point, the formulation was ready for initial beta testing and verification at bench scale and was challenged against an array of cell lines, operating conditions and screening methods. Such as N-1 perfusion, where the HIP CHO medium exceeded VCD requirements a day earlier than expected versus the bank cell line set batch medium with all conditions at 1 VVD. Our concentrated fed batch where the University of Bielefeld explored ways to improve performance of a lower producing cell clone and in doing so, reached 120 million cells per milliliter at 1.2 VVD, reaching a 42x increased titer of over 1.9 grams per liter versus the originating batch process in medium at 46 milligrams per liter. Zurich University of Applied Sciences performed an extension of their N-1 application to estimate intensified fed-batch perfusion performance at 3 VVD, reaching 160 million cells per milliliter before the vessel was unable to maintain oxygen. By comparison, the best fed-batch performance of this cell line peaked at 18 million cells per ml, and it suggests an 8x titer improvement of the perfusion process over the fed batch. Continuous perfusion is the most demanding environment in regard to media. Here, we demonstrate the formulation's ability to support different concentrations and sustain a healthy cell line at only 1 VVD. This was operated in bench reactors and bleed was adjusted to actively target a 95% cell viability. Both conditions are seeded in diluted HIP CHO medium at only 66% concentration. The blue condition, which shows steady state behavior around 13 onward, used this diluted medium for the entire process. The gold condition switches to the standard medium concentration after perfusion starts. For the standard 100% condition, bleed was started too early on day 7 and run too high, causing the percent viability to stay too high at around 98%, and limiting the initial viable cell density. Eventually, the bleed rate was carefully reduced to bring percent viability to the steady state target of around 95% and maintained for steady state from day 30 to 42. Both continuous perfusion conditions did well, with 100% concentration at steady state reaching about 120 million cells per ml and a titer over 1.7 grams per liter each day. For comparative purposes, the best fed-batch run with this clone using fed-batch medium and feed reached 3.3 grams per liter after a 14-day run, while a simple batch run with the HIP CHO medium only reached 1.3 grams per liter. No one would look at those simple batch results and say, "Hey, that makes a good perfusion medium." In other words, you cannot run perfusion medium comparison in a simple batch and expect useful results. When comparing run condition behavior, it's important to consider cell size. The plot to the right takes no account cell size by showing viable cell volume in the solid lines and productivity per viable cell volume in the dashed lines, where the plot to the left shows traditional VCD in cell-specific productivity. When looking at productivity per viable cell volume, the 2 conditions are almost perfectly in alignment during the entire process, a very different picture than what we see when looking at the change in cell-specific productivity during the time where the cells were over bled in the gold condition. Also, we should bear in mind nutrient requirements generally scale with total cell mass being supported. So it makes more sense to scale medium flow rate based on viable cell volume, bearing in mind that a cell diameter increase from 13 microns to 14.5 microns causes a 1.4x increase in cell volume, which is a significant increase in nutrient demand. A second continuous perfusion run was carried out with the same clone. This run was over bled such that percent viability is maintained at 97% instead of around 95%. That little difference resulted in a steady state VCD of around 95 million per ml instead of 120, and a titer of around 1.6 grams per liter per day instead of 1.74. In terms of running a perfusion medium test long enough to reach steady state, recall that in the previous continuous perfusion test, the 66% concentration condition had the bleed line-up exactly right, and we got the steady state from about day 13 onward. However, the full concentration condition didn't reach steady state until about day 29, and the run shown here hasn't quite reached steady state yet by noting the change in viable cell volume in blue. Never assume steady state in a short test, define it by having to maintain quality parameters such as percent viability because this tends to translate across conditions well. We've discussed the importance of knowing your control target in normalized test conditions. Equally important is knowing your equipment limits and scaling parameters such as agitation, so that you can support higher cell demand and make your results more applicable. In regard to supplements, ones that build up in a fed-batch will be removed, so you should increase their concentration, whereas supplements that deplete early will be continually added back so it's often good to reduce concentration. And supplements that are absorbed by the cells but not metabolized should be used very sparingly so they don't overwhelm the cells. Always adapt your cell line to new medium conditions. We recommend parallel sequential adaptation with selective pressure maintained and at least 2 passages after you think growth and productivity are stable. And testing your clone CSPR in different medium conditions will help guide practical medium exchange rate and testing. Equally important is to know your cell line limits. Some clones handle high VCD poorly and don't have sufficient production stability for continuous perfusion. These were the top medium conditions from the optimized Ambr run shown previously. We can see at day 17, there's a slight percent viability drop as they come into study state, followed by the second best-performing medium condition coming out on top and maintaining while the top-performing conditions performance erodes away. Verify short-length testing and apply what you learn back to your scale down operations so that you can have confidence in your work and build from your results. I hope this high-level overview of some of the things we've learned while developing high-intensity perfusion CHO medium were both interesting and useful to you. For more information, you can follow up with myself in the Q&A section later on or e-mail our product manager or your Thermo Fisher contact. Have a great day and enjoy the rest of BPI Europe.