0:00
Hello, everyone, and welcome.
0:02
Thank you for joining us.
0:03
Yes, my name's Ben Wilkes and I'm a Principal Biologics Market Development Manager at Waters.
0:09
If you're not familiar with Waters, we are a global manufacturing provider of analytical systems, consumables and software.
0:17
We work across a wide range of application areas, including pharmaceutical and biopharmaceutical research, development and manufacturing as well.
0:27
Today, I'll be talking about using automated LC-MS analysis in biopharma development particularly, and how that can bring increased advantages to your process development timelines and efficiencies.
0:41
So the three areas that I'm going to be talking about today are process attribute monitoring, product attribute monitoring, and connected bioprocessing.
0:50
I'll go through each of those throughout this presentation.
0:55
So to start with a bit of a problem statement.
0:58
We know that the goal of biopharma manufacturing is to bring better and safer treatments to patients faster.
1:05
And the three key drivers which we can identify in this area that can aid that robust and reliable manufacturing processes.
1:13
So avoiding product batch failure, if we look at a 28 kilo product batch manufactured in a 20,000 litre bioreactor, if that was to fail, we could be looking at a $1.4 billion net loss.
1:28
So there's a loss of revenue and that brings cost to the product as well.
1:32
So we can make the manufacturing process more efficient and reliable.
1:36
We can bring down the cost to the patient, gaining knowledge to accelerate the process.
1:41
So by making the manufacturing and the process development and the product development more efficient and quicker, particularly analytics and release testing by using data-driven decisions, we can bring that lag of analytics and delays down from weeks to days and also optimising assets and resources.
2:02
The drive towards continuous by processing and processing intensification is aimed at reducing facility footprint and cost of manufacturing and cost of facilities.
2:13
So if we can address all three of these aspects, robustness, reliability, time and footprint, we can hopefully improve process robustness and bring down the cost to the patient.
2:26
So how can making process development and manufacturing more effective bring advances to this?
2:32
If we look at a traditional upstream development campaign, we'll be running a number of culture experiments that might be lasting in the region of two weeks per culture experiment.
2:43
Samples will be taken periodically, perhaps just at the harvest point, and they can be sent to an analytical facility for a variety of analysis.
2:53
That's complex analysis with experts.
2:56
That takes time.
2:57
Typically we hear that the turnaround time is four to six weeks, sometimes longer.
3:03
But to run your next experiment, you're not going to be waiting for that data to come back generally.
3:07
So you'll be running the next set of culture experiments with limited visibility on the results from the previous experiment. Running blind leads to a poorly effective development process and it's not a robust manufacturing process and development process.
3:27
So what we're proposing here is use of at-line analytics providing near real time analysis.
3:36
We're using our LC-MS system as an at-line analyser.
3:41
We're able to take samples and analyse those in near real time in a matter of minutes.
3:46
Not days, not weeks, but a matter of minutes per sample and that gets you almost real time feedback which you can use to tune your process and make process decisions on the fly.
3:59
We can have faster turnaround time of the analytics by having increased data richness.
4:05
We can have more robust statistics in forming your process decisions and also getting towards feedback control of the processes.
4:14
We can increase process understanding.
4:16
We can have substantial savings in terms of cost of goods, time and resources and ultimately earlier access for patients at a better price point.
4:29
So with the BioAccord, we can provide one solution platform for point of need analytics, which is unique in the industry for providing process attribute monitoring, product attribute monitoring and also connected bioprocessing.
4:43
By linking all three of these together, we can accelerate development and provide a more robust process.
4:47
And I'll go through each of these in turn now.
4:51
So firstly, process attribute monitoring, particularly looking at cell culture media analysis.
4:58
That comes down to two main areas that we see.
5:00
One is media, raw material testing and the one that we hear from our collaborators quite a lot is vendor QC and lot to lot comparability between raw materials incoming.
5:13
So comparing the quality of these materials.
5:17
We can also look at media development and media optimisation from these raw material sides.
5:22
The other side of the coin is spent media analysis.
5:25
So in process analysis of your nutrients, your metabolites leading into optimising your feed strategy and also refining your chemically derived media.
5:36
So we can do this with a very fast and complete analysis method.
5:41
It's really simple that we've developed for the BioAccord.
5:44
We can analyse more than 200 compounds in a single analysis that takes less than 10 minutes.
5:49
It's 9 minutes per analysis time.
5:51
It's really simple and rapid.
5:53
The only sample preparation that we need is cell removal and dilution in water.
5:57
So really simple. By using the technology that we have, we have really specific compound tracking with high accuracy from the MS and its really insightful as well.
6:08
We can quantify known compounds, but we can also identify unknown compounds.
6:12
So if something's happening in your cellular metabolism that you don't know, we can work this out with the system.
6:20
Sensitivity and robustness are two key aspects of the system as well.
6:23
So we're monitoring at the sub nanogram per millilitre level for nutrients and I mentioned more than 200 compounds.
6:31
We're looking at amino acids, amino acid derivatives, organic acids, particularly the TCA cycle, metabolites, vitamins, nucleic acids, base nucleotides, and a whole host of other organic compounds.
6:43
And if you'd like to discuss that a little bit more and see what's in that compound library, come visit us at booth four.
6:50
I will add that this compound library's been developed for mammalian cell processes, CHO cell processes, HEC 293.
6:59
It's got compound libraries that cover all biotherapeutic processes.
7:07
So what do these results look like and how can we analyse these results on the system?
7:12
So this is just a small example of monitoring five bioreactors with duplicated samples every two days and we can look at absolute and relative compound tracking within one and the same analysis run.
7:24
So we have tryptophan here at the top represented by the results in a bar chart.
7:29
We added standards to go to calibration curve here.
7:31
So we have absolute quantitation of tryptophan here.
7:34
Down below we have choline, which we didn't use standards for.
7:37
So we can see the relative quantitation based on those results.
7:43
Just take that a step further.
7:45
We can look at all the natural amino acids from three different bioreactor conditions and we can easily plot and see visually what's happening across these three different conditions and pick out differences in trends.
8:00
So we can see quite clearly here that glutamic acid, we've got a very big change in one of the process conditions also for cysteine as well.
8:09
Whereas something like tryptophan is tracking quite closely across this process.
8:18
Now, an example by one of our collaborators at Jansen in the USA.
8:25
They were comparing two different cell lines and what they saw was in cell line B that had a sharp drop off in viability as demonstrated by this orange line here.
8:35
And they're interested what the root cause behind that was.
8:39
So by using this analysis for media analysis, they identified different components which were changing in the spent media, particularly I think it was arginine, citrulline, ornithine, N-Acetylornithine and N-E- acetyllysine and nicotinic acid, if my vision is reasonable.
8:59
Using embedded MVDA tools, they're able to confirm correlation and identify these key components which were being up and down regulated across the process.
9:07
And then they fed that back into their cell biology knowledge and identified that the urea cycle and the nitric oxide cycle were being activated here.
9:17
So that was what the driver behind this cell viability drop was.
9:21
And they're able to inform the process decisions in the process development back to form a closed loop here and optimise their process.
9:34
I'll move on to talking about product attribute monitoring now.
9:38
So like the process attribute monitoring, we've got a really fast and complete workflow, 6 minutes analysis time, and we can analyse multiple product attributes within that one analysis.
9:49
It's really efficient, high throughput with the six minutes analysis time.
9:53
We bring automated analysis and data processing.
9:56
Importantly, there is no purification required, there's only cell removal and dilution.
10:02
And we work at really low sample volumes, so sub 100 microliters per sample, perhaps even 50 microliters or lower.
10:11
As I mentioned, we can look at multiple attributes within that one analysis time of 6 minutes.
10:15
We can look at tighter product identity, purity aspects, glycan profiling and other PTMs.
10:25
It's designed to be accessible from a hardware and a software perspective, so easy to use for non-analytical scientists.
10:31
The idea is that this is a process tool and the last but not least, connectivity.
10:37
So we're connecting with different bioreactors and automated sample preparation, which I'll go into in a little bit more detail later.
10:47
The idea behind bringing this automated analysis and data processing is we generate actionable data.
10:52
So if you have glycan data, an LC-MS analyst or an MS analyst would typically be looking at a trace like this and inferring data that means nothing to most process analysts.
11:03
The table on the right-hand side is the actionable data and that is what we generate from this analysis here.
11:12
So typically when we look back at the process development timelines in the work frame that I detailed at the start, you might be looking at VCD and titer as the measures which you make process decisions and cell line selections on from a range of clients.
11:29
But what can we do if we had more data to make more informed decisions earlier in the process development stage?
11:36
One thing we looked at was product purity.
11:39
So this work was done by some of our collaborators at Sartorius in the States.
11:46
A typical chart of titer plotted against product purity, mAb purity here was the presence of light chain against the intact purities as a relative ratio.
11:56
And we can see across the 11-day process that whilst titer increases across the process as we'd expect, the product purity actually starts to decrease at around day 6 down to just below 80% from 85%.
12:11
So that will have an impact on your DSP processes as well as primary recovery.
12:16
We can use this information around product purity against titer to inform harvest day decisions and also your DSP processes.
12:24
But it's important to note that high titer does not always equate to high product purity.
12:33
One of our collaborators, Ole Wohlenberg at the University of Hannover, wanted to use this product purity measurement to inform his DSP processes, particularly light chain modifications.
12:46
So really simply and really quickly, within 5 minutes, he was able to identify all these different light chain modifications, so unmodified light chain cysteinylated plus glutathione LC dimers and also half antibody.
13:00
And what he was able to identify actually was his unmodified LC was in really low percentage abundance compared to what he thought here.
13:07
And that was having a big impact on his DSP process.
13:14
Another product attribute that we can look at is glycan profiling.
13:18
So here we have the G0F/G0F glycan plotted in two different process variations.
13:24
Again, this was done by our collaborators at Sartorius. Using a controlled glucose feed they achieved around 20% G0F/G0F at day 11.
13:37
By having this data available really quickly and reliably as an at-line tool, they're able to generate more processed knowledge as we go.
13:45
So they thought, what if we could use this to have adaptive glucose feeding?
13:51
So the hypothesis was can we target and modulate feeds to generate 35% G0F/G0F.
13:59
They overshot it slightly, but as a proof of concept using adaptive feeding, I think it's a good example.
14:07
And what this real time output and optimization can feed into is harvest time optimization, feed modulation and cell line selection, particularly important for biosimilar development and for CDMOs. You've got a cell line and you're dropping in different products and different genes into that.
14:22
So it's important in that aspect.
14:29
Sartorius took this as a step further as well.
14:31
I wanted to look at high throughput screening of multiple parallel bioreactors.
14:35
So this was an Ambr 15 experiment that Charles at Sartorius performed.
14:40
He had five different bioreactor conditions, each in duplicate with a control and a temperature shift and he was analysing for these four different Glycan profiles.
14:52
You can imagine the number of samples and the amount of time it would take to generate this data using offline analytics.
14:59
All of this was generated by Charles who is very much an upstream scientist with no analytical prior knowledge to this.
15:06
All of this was generated within 24 hours taking the samples across all of the amber culture stations.
15:13
So you can really increase and improve your data throughput and density.
15:23
Going back to the cell line selection, the cell line development decisions, we can take this information a bit further.
15:30
So we also had two different clones.
15:35
Clone decisions may be made primarily I said on titer and VCD and some other and process parameters as well.
15:41
And between these two clones, clone B had a higher titer, so typically would have been selected for process development.
15:50
By looking at the intact mass analysis and taking that a step further for subunit analysis with one of our collaborators called Genovis, we can digest antibody into 3 subunits with reduction.
16:05
Comparing these two clones under these conditions, we identified a small unwanted modification in clone B.
16:15
Imagine if we'd taken that further and we're taking that through process development that would have led to a big waste in resource time and an inefficient process.
16:26
So we're able to identify a process decision to fail earlier on here.
16:35
I can now talk about connected bioprocessing quickly.
16:39
So bringing together these workflows with automation, we have one of our solutions called the Andrew Plus Pipetting robot.
16:46
This is actually on our booth at booth 4 if you'd like to come and see it and discuss that a bit further.
16:50
That will provide all of the automation that a human can do with pipetting.
16:54
So we can automate sample preparation.
16:56
We can even automate cell removal and dilutions with the system.
17:01
It can feed it directly into our analyser for automated process, automated process analysis and processing.
17:10
It's termed a walk-up solution.
17:11
With end-to-end workflows.
17:13
You can put a plate sample on that is completely unprocessed.
17:16
You can process those samples, analyse them and produce a report at the end, all within a single software platform that is able to be used by completely unexperienced analysts that just work on upstream development.
17:28
To bring this to a point of need, I've mentioned quite a few times that we've been working with Sartorius and particularly we have a two-way data sharing data interface between the Amber 15 and the Amber 250 HT.
17:40
That is a data interface to load the sample list and export the analytical results back to the Ambr software.
17:48
Using that data interface, we can adapt to system behaviour dynamically using the process variables in the Ambr software and also feed that into DoE studies to optimise cell culturing conditions and product quality.
18:02
We've had some great collaborations with GSK.
18:05
Vithiya Vimalraj demonstrated this using the BioAccord and the Ambr 15 data interface to look at glycan profiling, molecule characterization, and cell media analysis.
18:17
And I've included a QR code there if anyone would like to watch the webinar that Vithiya did on this.
18:28
Another collaborator that we worked with was Doctor Li Wang at WuXi Biologics.
18:32
He was interested in looking at biosimilar development, particularly around the homodimer formation.
18:40
Again, a QR code for the webinar that he generated this data on.
18:45
But what he did was a DoE study on the Ambr 15 to identify variations between temperature shift, feeding strategy and medium and concluded that the product titer and three key amino acids were related to his increase in homodimer formation.
19:06
And to take our connectivity a step further, we've developed a JMP software integration.
19:10
We know this is one of the biggest data analytics tools that people are using for multivariate data.
19:16
We've heard that 80% of time taken for data analytics is related to data clean up and formatting.
19:21
So we've automated that for you.
19:23
We've developed this JMP add in which you import the data into JMP.
19:27
It will sort that data and process time series line charts for your process and your product attributes.
19:35
It will recognise which is which and separate those out and also bar charts to compare between different bioreactors and a whole host of other data analytics tools that we've generated into this add in that's freely accessible.
19:51
I’d just like to conclude with how we have linked up our process attribute monitoring, the product attribute monitoring, bringing those together with connected applications, JMP, Sartorius data interface and product automation.
20:04
So I'd like to thank you for listening and please come visit us at Booth 4 if you have any further questions.
