"Hello and welcome to our webinar, the importance of bioreactor scaling for upstream process development. My name is Sebastian Ruhl and I'm working as a field application specialist at Sartorius.

My product focus is cell culture technology, encompassing all bioreactors and peripherals.

Today, I want to talk to you about scaling of bioprocesses and why this is important enough that it should not be treated as a trivial topic.

Starting with the agenda, we first take a look at the background of what we typically understand as bioreactor scaling and how it is performed today.

We will then take a look at scaling challenges, alternative and new approaches to scaling, followed by other scaling considerations.

At the end of the webinar, we will have a short Q and A session.

So let's first start with the question, what is bioreactor scaling?

In general, we can consider bioreactor scaling as the transfer of a process from one scale to another, whether it be a larger or smaller scale.

The aim of this procedure is to maintain a similar level of performance between the source and the target scale.

The main performance criteria are the cell growth, the specific productivity, and the overall product quality at the point of harvest.

An example of a typical scaling approach in Sartorius Biorectors would be process development in an Amber two fifty system and a two step scale up using a uni vessel five liter to an SDR two hundred liter vessel.

This typical example considers the scale up of a process from process development, PD, into commercial manufacturing scale.

On the other end, there is downscaling, which provides the opportunity to troubleshoot a process in smaller scale than the production scale in order to save time and money.

Process characterization can also be performed in smaller scales using verified scale down models to gain further insights into the parameters that you need for a successful scale up.

Pure and scaling approaches usually look at one main parameter that is typically kept constant over the entire bioreactor workflow or scaling train.

Sometimes additional secondary parameters are taken into account, and these may or may not be the same at each scale, but nevertheless should be within a desired range.

Typical parameters for scaling are a function or a combination of agitation and guessing.

Most popular are the specific power input and KLA.

That brings us directly to scaling challenges that may occur.

Now we discuss scaling challenges.

If you keep one parameter constant throughout your complete set of bioreactor scales, you will soon run into limitations.

First, there are the physical limitations of your bioreactors and these differ across different scales and manufacturers.

Then there are other challenges that are scale specific.

For instance, a power input of thirty watts per cubic meter will be sufficient in the two thousand ditter production scale, but can lead to inhomogeneities in smaller process development scales.

High agitation rates might induce high shear stress for your cells, while low agitation can create problems with homogenization.

High guessing, on the other hand, can lead to excessive foaming, while low guessing results in CO2 accumulation or oxygen limitation.

In these scales, a key challenge is to find the sweet spot that balances these parameters as much as possible without compromising on product quality and quantity at each scale.

Another challenge for many scientists is the need for sufficient information about how their bioreactors behave and perform in order to predict how well their processes scale.

Today, most characterization data is manually collected during engineering runs and stored on in house spreadsheets. These spreadsheets may not contain complete or comparable data sets.

The process of generating characterization data is resource intensive and time consuming and comes with its own challenges.

For instance, carrying out experiments to collect mixing data requires regular draining and refilling of the cultivation chamber at each scale.

The creation of torque data that is used to calculate the specific power input requires a correctly sized tool for each scale, which may not be available and significant alteration of the drive system, again at each scale.

On the topic of data, nowadays scientists can have the use of advanced data analytics software to assist with scaling, but this access is not available to all and not everyone will have the know how to use these tools for the purpose of scaling.

So what alternative approach is there?

What if the scaling approach can consider multiple parameters at the same time, regardless of whether these parameters are constant or differ at each scale?

And what if these considerations can take in the full bioreactor scaling train from the start?

The answer is likely that some of the challenges mentioned before would be reduced or mitigated, and this approach would help map the design space of bioreactors across each scale.

When taking several parameters like Reynolds number or tip speed into account, it is easy to get discrete agitation and guessing values from process development to production scale.

This simultaneous approach would also make it easier to spot issues early on the scaling train rather than finding these out near the end as you progress with your process transfer.

This would save time and money from potential scaling failures at production scale.

When scaling, it would be very useful to understand the risk involved with the chosen parameter settings at each scale.

By risk here, we mean risk relating to bioprocess parameters. For example, a long mixing time results in high risk due to the development and maintenance of heterogeneities within the bioreactor, whereas a high tip speed results in more risk due to the larger shear forces it creates within the bioreactor.

Taking into consideration the aggregation of these risks across the scales can enable one to predict the scale risk of the candidate process.

And now on to some more details on this new scaling approach.

There is a new software application called Biopart Process Insights that has been designed specially for scaling across a broad range of Sartorius bioreactors.

Contained within the software are fully characterized datasets for each of the supported bioreactors all in one place.

Parameters like specific power input, tip speed, and KLA are given for all supported impellers and sparger configurations, working volumes, agitation rates, and guessing rates.

There is a guided user interface workflow and under the hood an engine to model the data and perform the scaling calculations without the user needing to know how to do so.

This simplifies and standardizes scaling for any non expert scalers.

You can create your full scaling train from process development to production, and the software predicts and derisks the transfer of multiple scaling parameters across multiple scales at the same time.

Process insights can also be optionally used to complement design of experiments and multivariate data analysis to further support the scaling process.

My last section is on other scaling considerations based on my own experience in the lab.

Even though agitation and gassing rates are in my opinion the most important scaling parameters, there are many other things to consider when looking at a holistic scaling process.

A good starting point for your holistic scaling approach is always to think about the largest scale that you want to reach.

This is the scale that sets the physical limits for all smaller scales.

For typical suspension cell culture processes, guessing rates above zero point one BVM and stir speed rates of more than one point eight meters per second should be avoided.

It is also helpful to determine the number of necessary pre culture steps in advance. This number should be considered and made consistent for all smaller scales.

This will lead to a more consistent passage number and a better overall condition of the cells in every main culture.

The largest bioreactor most likely will have the longest preparation time.

If the media stays in the cultivation chamber overnight for conditioning, it is likely beneficial to do the same in every scale.

When there is a long heat up time in larger scale bioreactors, this should be mimicked in smaller scales to match any potential media degradation.

The same applies for the time taken to apply temperature and pH shifts.

When it comes to feeding, the feeding duration should be considered for all scales and should be defined by the required feeding time at the largest scale.

In some small scale systems, it might be necessary to adapt the feeding regime to the volume removed during the sampling process since the proportion of volume removed is more significant to the overall bioreactor volume.

This should prevent potential overfeeding or dilution of the culture.

Geometrical differences of the bioreactors should be taken into account too.

Not all bioreactors have similar geometric designs. Some systems might have different height to diameter ratio or other features such as a single impeller or different driven shafts.

If the process relies on certain analytical instruments or sensors, the availability of these for all scales should be checked in advanced for consistency.

Media preparation can differ with regards to the required volumes and small amounts are often supplied as ready to use liquid media, while larger volumes need to be prepared from powder in place.

Availability of mixing equipment and cooling capacity should be considered accordingly.

In some systems, different sparger types might be present which lead to specific forming characteristic and gas requirements.

Last but not least, most of the time traditional scale up strategies only look at certain time points of the process performance.

For instance, the point of harvest.

Further insights could be achieved by looking at time series data like integral viable cell concentration or the trend of the glucose consumption.

To sum up, these are just a few practical tips based on some of my observations and experiences in the lab and in my opinion are likely to help with scaling.

As a conclusion, I can say that in my opinion, the conventional one parameter scaling approach is no longer state of the art.

There are new advanced software approaches available that enable multi parameter scaling across multiple scales that even include risk prediction.

On the topic of other considerations, in a typical scale up process, many seemingly minor things can combine to have an effect that can easily improve the performance of the process.

These approaches should help simplify the scaling process and make it more consistent and accessible across upstream teams.

We now have reached the end of the webinar. I would like to thank you for taking the time today and hope that this webinar was beneficial for you and can help you with your future scaling projects.

We will now move on to the Q and A session.

Thank you, Sebastian, for that interesting presentation. We have already received some questions from the audiences. Our first question is about the CO2 in the modeling. What you can say about modeling CO2 as part of scale up?

Yes, the CO2 levels, they do not significantly affect all molecules in the processes, but can affect some.

So it is definitely an important parameter for us.

With special guessing strategies, you can then mimic the CO2 profiles that you see in larger scales, also in smaller scales.

Since the CO2 stripping usually performs a little bit better in smaller scales than larger scales, also, you can limit the CO2 stripping in small scales to mimic that in larger scales.

And when you take a look at the overall CO2 profiles and total base consumption, this can be used as a kind of quick check for your culture performance. You don't have to dig too deep into your data but take a look at this online data which is already available after your process directly.

And this is already a good indicator for the CO2 level for the CO2 levels in the in the process and the PCo2 levels and can already give you some some indicator if your PO2 PCO2 levels are consistent overall of a complete scale.

Scales.

Thank you, Sebastian.

Next question is, you mentioned about the scaling software, the process insight.

Can I add my own data or use non sartorius bioreactors in the scaling software?

This is a question we get every time we present this tool. And unfortunately, this is not possible.

We don't have access to competitor systems. So, no, this is not possible.

Thank you. And the next question is: What growth medium was used to characterize the data in the process insights? And what are the considerations when applying this data when scaling using my own my own media?

So the media that was used was a PBS buffer following the Decima guidelines that you can view online. They are open for everyone.

And the effects of Claronic and any foam and viscosity, they they are more or less should consider them definitely because they change they change the media because But since there are so many different anti foam solutions and also different pluronic pipes and other things that are so specific to your process, need to consider these special media additions for your own process.

It was pure PBS buffer.

Thank you.

Thank you.

Next question: Is exit gas velocity considered as part of scaling?

This is important to note.

So potential disruptive disruptive effects of these exit gas of this exit gas velocity, this is taken into account. Yes. So you can get this exit gas velocity out of the tool and you can take it into your scanning considerations.

It sometimes is a source for shear cells and also has an effect on the bubble breaking, which is also affecting the cells.

Yeah, great, thank you. What considerations are there when trying to scale perfusion cultures?

So, for perfusion cultures, sure, all the other considerations are taken into account or should be taken into account. But there's also this cross flow rate. So the perfusion rate more or less that no, not the profusion rate per se, but the cross flow rate, the movement of the medium alongside the hollow fiber. This can cause excessive shear on the cells.

When you reach very high cell densities, you can also have an effect on the viscosity. So this can increase, which also changes the behavior of the culture. Sparger clogging is definitely a topic when you have different sparger types into your bioreactor and you only use one at a time. A lot of high cell densities can clog the sparger.

The filter size and lumen and size and of the diameter of the filter is definitely should definitely be considered. There are several different filter core sizes available for different applications. So this always is a great effect on does the filter block or not.

Also for external filters, So filter fouling should be considered as well.

Thank you. We have a similar questions, this one that is for the adherence cell cultures.

It says what considerations are there when trying to scale adherent cell cultures?

For adherent cell cultures, the EDI size is definitely something that you should consider.

There are low steering speeds normally applied, so mixing time should be definitely considered if this is sufficient to always have a homogeneous culture since you don't want to shear off the cells from your micro carriers.

This is definitely something that needs to be well balanced.

And the micro carrier concentration into the bioreactor, which is also an important parameter here.

Great, thank you.

What impact does the bioreactor geometry have on scaling?

So when you scale between different scale of the bioreactor small size, large size what will be the impact of the geometry?

Yeah, so the standard is surely the classical round stirred tank reactor.

We also saw good performance of the culture in cubical stir tanks or also in two d rocked bioreactors in my practical experience.

But you need to take a look on other parameters. For instance, when you go for a two d rocked bioreactor, specific power input is not a parameter that you can so well measure in these systems. So things like mixing time and KLA might be of a bigger interest in the scaling approach.

Great, thank you.

Another question: What do you think causes the different lactate production patterns?

So, the difference in the lactate production patterns in different scales.

This is very special for every process. So, from my experience, excessive glucose concentration is one of the main one of the main sources for that.

Also, great fluctuation in PO2 can lead to X to increase lactate production. So if you have a very unstable PO2 or DO value due to maybe a non optimal mass flow controllers and many, many different, different impacts go in there. This can also increase lactate production. But normally, in a, normally normal run, you always see a certain lactate increase at the beginning in the first few days, for instance, in a fat batch process, but this should be consumed in a later stage with increased cell density. If this is not the case, there's definitely something not optimal in the culture.

Yeah, thank you. What you can say about the PCO2 patterns in different scales? So, how different the scales are capable of stripping the PCO2?

Yeah, we already answered this a little bit in the first question. So the large scale bioreactors most likely have reduced CO2 stripping. So PCO2 is by default a little higher in larger scales.

What we have done in our practical experience in the lab is that we follow for larger scales for the Biosart STR for instance. The system has a combisparter, so we used the microsparter part for oxygen, so small bubbles and high oxygen transfer to the culture.

And but with increased oxygen flow, kept a constant airflow through the ring sparger with larger bubbles. And this helped, this really helped us with CO2 stripping. So the larger bubbles were created through the ring sparger and CO2 could be stripped very well from the culture so that we saw similar PCU2 patterns when we compared small to larger scale. So the constant, to answer this quick, a constant airflow for larger scales through the largest through the largest spotter that is available is a good, yeah, is a good little little trick to get the PCO2 levels in a normal range.

Brilliant. Thank you.

Can some parameters including shear stress or Kolmogorov scale, etc, from CFD calculation also be utilized in the scaling?

CFD is is nice for is nice for an assumption on your specific power input, but since we're talking about most most of the time here large bioreactors which are film based and if you have ever installed a film based bioreactor you know that is installed every time a little bit different with regards to the to the film, which is which is flexible into into the holders. The CFD model for a single use film based bioreactor is not always so precise. So I most likely would always go for practical experiments then to fully trust the CFD model.

Great, thank you very much. And, yeah, that will also conclude this webinar.

We might proceed just with one more question.

If the manufacturing is expected to be performed with a low power pressure, for example, zero point one bar in the bioreactor, should we also apply it at the scale down model? Is that possible to do that with Sartorius lower scale vessels?

If you definitely want to apply this, zero point one bar is definitely very, very small. So the effect of increased oxygen saturation into the liquid will be very limited. But if this is planned to do so, I would try to implement this implement this as as good as possible. And for Sartorius glass bioreactors, the univessicles, there is an option available for that to go up to zero point four bars of overpressure.

So zero point one can well be established there. But please be sure that this is not the case for ember systems or the film film based bioreactors. We cannot apply an overpressure of zero point one bar there. Yeah, but for the glass bioreactors, it is in general possible.

But I would overall would see the effect as very limited. So it could even be possible that in the amber you will see a comparable performance without the overpressure.

Great, thank you. In the conscious of time, then we have to conclude this webinar. Thank you again, Sebastian, for that exciting presentation. On behalf of all of us from Sartorius, thank you for joining us today, and have a great rest of your day. Thank you."