Overcome Exosome Production Challenges

Marcus Osei: What was the reasoning behind performing cultures under hypoxic conditions?

David Haylock: It’s well known that conditions for culturing cells can change exosome production and, also, modify the molecular cargo of EVs (extracellular vesicles) and exosomes. So, culture under low oxygen conditions can lead to direct regulation of specific genes and pathways, for example, the Hif1-alpha gene. And by doing this, you could potentially alter the molecular cargo of the exosomes to enhance their functional properties. And, in this case, under hypoxic conditions the strategy there would be to produce exosomes with enhanced proangiogenic properties as a consequence of activating that Hif1-alpha pathway. There may be other pathways that are activated, but that’s the general concept. By manipulating the culture conditions, you can modulate cell growth, exosome production and exosome cargo.

MO: What is the exosome’s max concentration achieved through Pall sterilizing filters?

Roberto Ciboldi: We can definitely cover a wide range of concentration, as David said during this presentation today, concentration is mainly based on the iCELLis® bioreactor productivity, but we can achieve over 1010, 1011 even 1013 EVs per mL without any problem on our sterile filters and maintain the transmission rate.

MO: Could you advise some typical analytical technologies to evaluate the quality of exosome?

DH: I think the key word in that question is the quality of exosomes, so we have at our disposal a number of analytical techniques which can investigate the physical characteristics and, maybe, the molecular characteristics of the exosomes. And I think many people, by inference, just assume that that infers quality. 

I think the important thing there is that the techniques you need to use to address quality are more functional assays. So, we need to be using a suite of appropriate potency assays or assays which really test the ability of the exosomes to bind to a target cell, release content and affect a phenotypical functional change in the cell. I think that’s really important, as part of your analysis that we use, we have developed a number of potency assays to inform us about angiogenic potential. For example, the ability to promote endothelial cell proliferation or induce endothelial cell tube formation or sprouting. So, I think those sorts of assays and technologies are really important. 

In addition to that, you can use a number of other more advanced techniques to give you information about the exosome subsets within a population of vesicles.  We have a feeling it’s moving us to use quite sophisticated flow cytometry-based methods, which can determine, in real time, the proportion of exosomes that might express particular cell surface markers, for example, the tetraspanins or our other tags that are considered to be hallmark markers. By doing that, you could define that a population, according to the number that are CD9 positive or C83 positive and the like, so I think that that sort of more sophisticated approach is going to be very useful. But, ultimately, the quality needs to be addressed by functional assays.

RC: I fully agree with David. We definitely never trust a single technique. We use different techniques in combination for different steps and, for sure, as I spoke during my presentation, there are those standard techniques, for some specific steps that need to be evaluated. But, overall, I fully agree with David and use different combinations of techniques to give you a comprehensive number where you have to rely on that.

MO: You started to use human platelet lysate as a supplement in the bioreactor. Did you switch to serum-free condition finally? Which medium would you recommend that can be handled in downstream processing (DSP).

DH: So, just let me clarify, we used at least three different media types for the first part of the culture in the iCELLis Nano bioreactor and those did contain serum, animal serum, foetal bovine serum, and/or a proportion of human platelet lysate as a supplement. So, we are well aware of the issues regarding vesicles that might be present in human platelet lysate and/or serum so, for that reason, we have switched into a serum-free media at the end of the cultures, prior to harvesting, for concentration of exosomes. 

In answer to the question, which media would you recommend for DSP, I think it has to have a number of properties. Clearly, it needs to support the ongoing growth and well-being, if I can use that term, of the cells. It needs to keep them happy and support cell proliferation and exosome production but, equally and maybe more importantly, ideally low protein, so free of serum and also completely free of any other vesicle that might interfere with the downstream process and might contaminate the final exosome product that’s cell derived. So, I think there are a number of media that could be used for that. I did mention the Rooster EV Collect, I think that’s one of the commercial media that’s now available, that is being tested by a number of people, particularly for MSC  (Mesenchymal Stem Cell) culture and exosome production from those cells. I think there’s some choice out there to be had, but as long as you have those characteristics, I think it serves the ultimate purpose. 

RC: As David mentioned before, the most difficult contaminants to be purified are other vesicles, other sub-population vesicles that you don’t want in your process. So, the best way is not to grow different sub-population vesicles in the bioreactor first, so you optimize the bioreactor. Then, through the Mustang® membrane, so chromatography, but then also during the tangential flow filtration (TFF) steps, we can definitely reduce the amount of contaminants by adopting different process parameters in order to drastically reduce other EVs subpopulations as well as protein concentration. And that’s proven from the product characterization.

MO: What are the problematic impurities and their level in the current purification process for clinical phase?

DH: I can’t give you an answer based on our experience, because we have not done enough work to assess this, but it’s a very important consideration. I think one of the most important contaminants, apart from other vesicles which are not cell derived, would be host cell DNA or histone DNA complexes which, if it co-fractionates with the final product, can lead to immune response. So, I’d be very concerned about that. We are aware that those entities, these histone complexes, host cell DNA complexes also cause problems in the downstream process. They can clog membranes and resins and they can also bind to exosomes and reduce the efficiency of the process, so I think they’re important for both the clinical aspects, but also the processing aspects.

RC: I can only add, definitely on the market, we can clearly see a correlation in contaminants between exosome producers and other gene therapy markers, like adeno-associated or AAV. So they are very close, as a process, let’s say, and also from a contaminants point of view, from the downstream side.

MO: Do you think a Benzonase♦ step should be implemented and where, in the process steps you presented, would you do that?

(♦Benzonase is trademark of Merck KGaA Corporation)

DH: Benzonase is an example of a nuclease enzyme, an enzyme which will basically break down DNA or DNA complexes. And it goes to the earlier question about what are the contaminants, so there are some procedures that have now been established where an enzyme, like Benzonase, can be applied. There is a commercial enzyme marketed by another company for that particular purpose. I think, ideally, I would make the decision based on knowledge of the nature of the starting cell culture supernatant and the substrate that you’re going to process. The informal way to do it would be to do an analysis of that and decide whether there’s sufficient material that requires enzymatic digestion before you put it into the process. As a rule of thumb, you might use Benzonase or another enzyme up front before a TFF process or a chromatography step, but I think it’s certainly worthwhile considering.

RC: I agree with David even if in application perspective, we can also use the combination of TFF and Mustang Q to remove HC DNA and DNA complexes without the addition of Benzonase. It depends on the amount of nucleic acids present in the process fluid and the performance of existing process steps in terms of impurity removal efficiency.

MO: How did you separate exosome and similar things in the chromatography step?

RC: It depends on what the similar things are first, so identifying your contaminants is always the best method to proceed in process development. We can share insights to deliver effective Exosomes vs contaminants separation on our Mustang Q/S membranes, first we suggest to reduce as much as you can the loading conductivity to enhance Mustang Q exosome binding capacity maintaining the pH one point over the Exosomes isoelectric point if known. If you are developing chromatography for the first time in your process we suggest to adopt a gradient method during the elution step to better identify which species will be eluted in the different fractions, different characterization methods are highly suggested during this process to improve EVs subpopulation identification. Last but not least, a reduction in flow rate between 2.5 to 5 CV/min could improve the operating binding capacity and resolution. 

DH: I’m just intrigued. We must get our hands on some of the Mustang Q materials to try, Roberto.

MO: How did you analyze and quantify initial seeding efficiency and perform cell counts without a nucleocounter?

DH: So, measurement of seeding efficiency. It’s an arithmetical consideration, so if you’re loading X number of cells, for example, let’s use 100 million cells into the vessel. Upon initial seeding, there’s a recirculation step immediately and the cells are basically passed through the vessel to contact the fibers. A sample is taken two hours after that, so this is a sample of the media and one measures the number of cells that are in solution in the media. And, for example, if there were 20 million cells in total, present in that media, you would then calculate that you have 80 million attached to the fibers in the vessel and therefore the seeding efficiency would be 80%. So, that’s the way that we have assessed seeding efficiency. There’s no direct measure, because you cannot count every cell that’s bound to a strip within the vessel, so this is an indirect way of doing it, but it’s proved to be useful and gave us some good guidance. And we’re able to measure differences in seeding efficiency in the course of these experiments and, as I mentioned, the final experiment in the large surface area vessel gave us some remarkably high seeding efficiency when calculated using that methodology. 

MO: May you share the advantage/disadvantage of using cassette or hollow fiber for TFF step of exosome process?

RC: The main cassette advantages are mainly but not limited to a lower footprint, lower hold-up volume, lower product loss due to a reduced wall shear rate during operational, higher product recovery and relatively higher permeate fluxes known with cassettes than HF. So, if it is a downstream (UF/DF) step, cassettes should be a preferred choice of TFF format. 

MO: In your quality control experiments for exosome cargo, I anticipate that it is possible to detect differences in exosome cargo between static monolayer flask cultures and various '3D' culture and media flow, mechanical signalling environments etc. Did you see differences like that between the control monolayer flask exosomes and the iCELLis bioreactor exosomes?

DH: We haven’t actually completed the analysis of that, but it’s a point well taken. We do have quite a lot of information on the molecular cargo of the exosomes from the cells that are cultured on flat surfaces in the flask cultures. But we have not completed the analysis of the exosomes that have been cultured in the iCELLis Nano bioreactor as a direct comparison, but that’s something that needs to be done and that information is very, very important. It’ll give us a good indication as to whether there are changes in the biogenesis of the exosomes and the way they load cargo. That is certainly something that we intend to do in great detail in the near future.

MO: Did you try three Mustang Q XT140 capsules in series in the purification of chromatography step and what is the problem?

RC: That’s a specific one. Generically speaking, no, we didn’t try to run three in series yet. It depends on the system that you’ve got, if the system allows you to proceed in that way, because it’s less time consuming, it’s preferred, but the majority of our customers prefer to stay with the one Mustang capsule at a time.  Adopting multiple Mustang Q in series is not different from a continuous multicolumn chromatography approach.   In order to expand the question including membrane chromatography against adsorbent chromatography is also profitable allowing to reduce time to market generating robust and scalable result , so why not take that advantage? 

MO: Could you also advise your plan to install iCELLis 500+ bioreactor?

DH: As I indicated, the iCELLis technology looks very promising. We have been engaged with Pall and talking about the possibility of using the larger iCELLis 500+ bioreactor for manufacture. It’s certainly a candidate for us to use in our process. The final decision’s going to be one based on cost of goods and economics and just these considerations around scale and costs, so we’ll work through that. It’s certainly a strong candidate. One of the key things that’s going to impact that decision are the technologies currently being used for productions of virus and other materials for clinical use, so it’s a GMP standard piece of equipment that’s used within the sector, so that positions it really well for strong consideration.

MO: What do you think about the interest of using computational fluid dynamics (CFD) simulation in order to perform scale up of the process?

DH: I think that’s fantastic. I think that’s a great idea. We are particularly interested to see what the influence of shear stress is on different cell types and using CFD  would be a very useful tool to give us some insights around that. I’d be interested in working with any parties that have some appropriate models in place and some algorithms that could enable us to calculate shear stress under particular flows in stress situations. And then we can do the necessary biological examination of the cells and the exosomes to determine what we can use to our advantage. That would be great.

MO: Which is recommended, in series or parallel Mustang capsule connection?

RC: It is definitely better in parallel than in series, because in series you get the loading from the previous Mustang. It depends, if you are going to provide a continuous processing, usually we use that in series, because the break through from the previous Mustang capsule will become the loading for the next one. So, this is continuous, so 24 x7 for 365 days year, for other biological applications, Mustang capsules in continuous mode works. But, definitely, I don’t have experience with continuous Mustang capsules with EVs at the moment, so my suggestion is to test in parallel where you can definitely better control the critical process parameters. But, if you’ve got experience and expertise and a system to do so, you can definitely use those also in series. Again, it’s better to confirm through lab scale or using Mustang Q capsule, not XT device, so with a lower bed volume, what could identify the DBC, dynamic binding capacity, and from this number then scale up your application directly and choose the best way in parallel or in series.

MO: You mentioned the media washout and replacement. What culture media did you test and what is the difference between the growth and the harvest medium? 

DH: The washout media is depleted of serum and human platelet lysate, and other protein so that it is very low in protein. And it’s synthetic so it’s low in any contaminating vesicles, whereas the initial media that’s used to commence the culture under these slow conditions is a complex, complete media often with serum and/or human platelet lysate. And that’s designed to really ensure that the adherent cells grow very well early in the bioreactor. It’s important to have high growth, high viability and keeping the cells as biochemically active as possible, so the exosome biogenesis programme is maximized. So, the first media is designed for that and, once they’re producing large numbers of exosomes, we’ve discovered that you can take them out of the really complex, complete media, put them into the synthetic media for up to two or three days and sustain exosome production. That’s within a media which has none of the contaminants that cause problems in the downstream process. So, I think that’s the rationale behind the choice of media at different stages of the culture.

RC: Not that much. Just focus also on the sterile filtration of the media, because this has got a huge impact on the productivity, even if you use that in the bioreactor or in other process steps. But the fluid dynamics coming in the process should be addressed and controlled, like the process itself.

MO: How do you follow cell viability all along with the culture in bioreactor? By using glucose, lactic level, how do you exclude the presence of other EVs, besides exosome?

DH: So, as indicated in the presentation, the way we followed cell viability was using surrogates, using glucose and lactate measurements throughout the culture period. That’s not an ideal way, but it’s certainly informative. It tells you whether you’ve got active glucose metabolism and lactate production. For these bioreactor systems, the ideal thing would be to have a direct measurement of cell number and cell viability and I guess that’s the advantages of the new biomass probe that Pall have developed for this technology. They can give a direct read-out of cell viability and number, so I think that is the way to go in the future for any of the technologies. 

The other part of the question, how do you exclude the presence of other EVs beside exosome? That’s a very important question. Clearly, we can hopefully control that by careful use of selected media during the culture and when we harvest it. It really does depend on the entire downstream process and the combination of TFF and chromatography steps, and other filtration systems to select ideally a well-defined EV population of defined size and characteristics. So, generally speaking, that would be the approach, a careful selection of the tools that Roberto has described. He was speaking about the Mustang systems, so using those in the right order to carefully purify the right component is the answer to the question.

RC: That was a tough one. I believe the contaminants are really linked to product characterization, so if you’ve got a very nice and robust product characterization, you can definitely identify also contaminants and it’s true also the other way around, so if you want to focus on product characterization you also have to focus yourself on the contaminant’s characterization. That is not a waste of money there, especially for this complex solution. Then TFF in combination with Mustang capsules could help to drastically reduce others EVs sub-population if any.

MO: Looking at yield and also product purity is there any difference between adherent and suspension cells?

RC: The first choice is based on the Cell line requested to produce a specific Exosome population, some cell line are adherent only and others are suspension only, than there are exosomes that could be produced both ways.

Exosome production through suspension bioreactors requires maintenance of cells at high density in order to obtain sufficient accumulation of exosomes in the culture medium, prior to purification so the suspension cell culture will have more cell load and thereby, the TFF membrane will be challenged more in terms of capacity, fouling, product transmission and area requirement. Adopting an adherent cell line will drastically reduce the clarification surface area required due to a reduced solution turbidity and therefore turn down the TFF challenge for some specific Exosome production. Some Exosomes productivity data through adherent cell line using iCellis nano are yet to come.

MO: Did you have any challenges with contamination when using the system? Were there any challenges of cell lift off during the cell culture period?  

DH: Two very interesting questions. I’m very pleased to say we had no contamination in any of the experiments, so that was very pleasing. And I think that’s a direct consequence of the care taken by the people preparing the cells and preparing the bioreactor, setting it up, assembling it and autoclaving it and going through the protocol very, very carefully. So, very pleasing, we had no contamination whatsoever. The bioreactor was not run in a Class 2 biohazard cabinet, it was certainly assembled, the manifold was prepared and autoclaved in a Class 2 cabinet, but it was located on a bench during the entire operation for as long as 14 days. We did do the sampling of strips and the collection of samples in the biohazard cabinet but, despite all that manual handling, there was no contamination, which is very, very pleasing. 

The other question is a really interesting one. Were there any challenges of cell lift off during the cell culture period? That’s interesting, we never actually looked for that. I imagine there would be, cells would become detached. As they divide, there’s a brief period of detachment and, potentially, they could lift off the fiber material before re-adhering and bedding down on the vessel matrix fibers. But we didn’t electively look for that in the samples that we were collecting. It’s an interesting thing and something to be mindful of for extended culture periods, I think, particularly if you’re switching media and doing manipulations, so it’s a very good point.

MO: How do you quantify exosomes through a biochemical technique without NTA (nitrilotriacetic acid) here?

RC: We discussed briefly about characterization techniques during my presentation and there are some gold standard techniques like microscopy analysis like SEM and TEM or others physical like DLS or Zetaviewer rather than NTA or if you want to specifically target vesicular protein on the EVs membranes there are techniques called Labelling.  If you don’t have the system or you don’t have the access to these gold standard techniques, there are other secondary techniques able to be addressed. It depends on the robustness that you want from your process. Maybe, David, can you help me there?

DH: It’s a very interesting question. I’m not fully conversant with this. I believe there are some biochemical techniques which ask questions about some enzymes, in particular that are found with exosomes. And I think it’s a really good idea that we think about using biochemical approaches for measuring exosomes and their function. I’d like to follow up on that, it’s a good question.

RC: Maybe you can use labelling? Labelling of EVs does not affect their functionality and their localization can be tracked after its uptake by recipient cells without resorting to any conventional reporter-based strategies or lipophilic dyes. Another available and recent technique is immune-fluorescence, wherein the isolated EVs can be stained with fluorescently tagged antibodies to proteins present in EVs.  So, it’s a tricky question, because of interactions with different analytical techniques and that’s the bottleneck of the exosome production right now. So, the characterization is really what matters, also from the regulatory agencies.

MO: Have you tried to produce two consecutive batches with the same cellular population? Do you think a same cellular population could sustain harsh production conditions twice whilst engaging differentiation?

DH: We’ve certainly done that manipulation in flat bed cultures and in tissue culture flasks. And, for the cells that we’re working with, it seems not to impact adversely on the cell proliferation and the exosome production, so I think it’s certainly possible. It’s something we’d have to do as an experiment with a bioreactor system. I think this issue could be almost overcome if you’re using immortalized cells, rather than primary cells and I think they’re likely to be more robust. I think that gives you far greater flexibility with how you might use them in various bioreactors, and you could contemplate a series of periods where the cells are in a harvest media. And they’re re-fed, they go back into a more helpful media and you repeat the batch harvest again. So, I think it’s a very interesting approach that can be added into the mix.

MO: Regarding the sterility of TFF cassettes. Which method of sterilization is used and did you assess if the filtration behaviour changes upon sterilization with, for example, gamma ray sterilization or e-beam sterilization?

RC: There are a couple of questions mixed up, I believe. Regarding the sterility of TFF cassettes, we’ve got a Validation guide where we validate the adoption of gamma ray as a sterilization method USTR2901,  the objective of this paper was to verify that irradiated Omega membrane has no significant change in performance compared to non-gamma irradiated Omega membrane. Membrane performance tests consisted of water permeability and solute retention.  It’s preferable to use single-use cassettes if you need something that is biologically controlled to enhance process robustness. 

I personally tested different sterilizing filters, gamma radiated ones or autoclaved ones and we didn’t see any different performance on the EVS or other biological products. 

MO: You talked about linear media speed. Can you go on a bit about what that means in the control of the process? Is it a goal to improve media exchange/nutrient supply/removal of lactate etc, into a more continuous fashion? 

DH: Yes, I think it is. Roberto might have some comments about this, as well. It’s all about obviously maintaining adequate nutrient concentration where the cells are growing and removing inhibitors, lactate and other metabolic inhibitors, from the cells. So it’s all about providing the optimal nutritional conditions for the cells to grow well and we didn’t change the linear speed of the media flow. We used the recommended rate that Pall advised. Potentially, depending on the cell type, this could be changed, particularly if you get fast-growing cells or very high metabolic rates, so it is potentially something that could be changed to improve cell viability and exosome production. I think it has to be customized according to the cell, though. 

Can you even retain exosomes in perfusion? Yes, they are present in the perfusate. They are being secreted by cells all the time, they’re going through the system, so there’s a constant movement of fluid through the vessel, through the fibre bed, and exosomes and EVs are moving continuously through the culture. Roberto, do have you got any comments about this question, because it’s an interesting question?

RC:  

As David presented, we got a patented flow system designed to maximize productivity in our iCellis Bioreactor the media circulation is achieved by a built-in magnetic drive impeller, ensuring low shear stress and high cell viability. The cell culture medium flows through the fixed-bed from the bottom to the top. At the top, the medium falls as a thin film down the outer wall where it takes up O2 to maintain high kLa in the bioreactor. This unique waterfall oxygenation, together with a gentle agitation and biomass immobilization, enables the compact iCELLis system to achieve and maintain high cell densities – equaling the productivity of much larger stirred-tank units. 

MO: Can the sterilization method (for example,  gamma rays) for the TFF membranes affect their efficacy and pore sizes? 

RC: The answer is no,  definitely, You can use single-use TFF, coming from our production sites that works brilliantly over different biological functionality of fluids. For example, for AAV or EVS. So the answer is no, by the way supported also by our Validation Guide USTR2901 where we claim that Omega membranes exposed to >50 kGy gamma irradiation have no significant change in performance compared with non-gamma exposed Omega membrane .

MO: You told us during your presentation that immortalized cells are promising cell types for the production of EVs. Doesn’t that immortalization of cellular type affect the cargo, thus the therapeutic effects of the produced EVs? 

DH: Potentially, but it depends on the method one uses for immortalization. There are a number of different molecular strategies that can be employed for that purpose. We believe that over expression of the catalytic domain of hTERT is probably the preferred method. It does result in good rates of cell immortalization and stable cell lines, but the point is a very important one. One needs to certainly do good characterization of exosomes and their cargo prior to, and at the end of immortalization to define whether there has been any change. We would prefer to stay away from immortalization strategies, where there’s over expression of oncogenes or genes that are involved in tumorigeesis. So that’s the position that we have, but the point’s well made. This will be an issue that the regulators might be looking at very carefully, the impact of that process on the final product and something that the industry needs to address, I believe.

MO: What is the exosome binding capacity of Mustang Q membrane and recovery rate? 

RC: That’s an interesting one. It depends on the exosome that you are trying to collect, so the exosome that you are producing. We’ve got a huge operating  binding capacity for Exo-L and Exo-S, the binding capacity is not limited by the Exosomes dimension but it could be easier to handle Exo-S vs Exo-L products. Talking about numbers, we can definitely bind up to 1010 per mL or even more, definitely more if you perform a very accurate loading maintaining a low loading conductivity and choosing the right pH assay maybe developed by a DoE if necessary. Talking about numbers derived from field activities on Mustang Q we were able to bind 1010, even to 1013, it depends on the EVs purity. Other parameters that could decrease OBC are the amount of contaminants in the solution, that could operate as EVs competitors on the binding site. So, the best way to start the chromatography is to characterize as much as you can your product and the contaminants you still may get out from TFF, that’s very important.

MO: In the case of expT3, campaign two, was there any continuous medium exchange perfusion before day seven’s total media change?

DH: No, there wasn’t. The only media movement was the linear speed movement of the media through the system. It was just that we used the initial 900 mL volume that was just being re-circulated through the vessel through the entire period of culture.