SCIEX
Duration: 26:03 Min
Comprehensive mRNA-LNP Characterization with Capillary Electrophoresis & Mass Spectrometry
Transcript
0:01 Hey. Hello, everyone. My name is Fang. I'm going to show multiple different workflows regarding
0:34 the mRNA lipid nanoparticle characterizations by both capillary electrophoresis and mass spectrometry.
0:42 Thanks, right, we have heard. We have got a very detailed CQA characterization for the
0:48 mRNA part, but as a comprehensive – I don't know why it's running automatically – but
0:54 compared – Sayegh, as a comprehensive analytical solution provider, we essentially provide,
1:01 at least here, right, demonstration of the different concepts to the entire workflow
1:06 for the mRNA lipid nanoparticles from plasmid DNA characterization, right, that's the raw materials,
1:12 and to the IVT-expressed mRNA integrity and encapsulation efficiency. And different
1:18 mRNA CQAs, I will touch a little bit upon, again, the poly(A) and 5' capping efficiency using an
1:26 orthogonal method. And then the other part of the lipid nanoparticles, the ionizable lipids,
1:32 which is another core component to make a successful and – sorry, why is it running – and
1:41 to make a successful and stable mRNA lipid nanoparticles. And last, right, what happens
1:47 after those mRNA lipid nanoparticles are injected into the patient, right, the MAD-ID,
1:55 right, post-administration tracking. And we also have a concept to – a proof concept workflow to
2:03 track what happens after the ingestion. So, without further ado, let's dive onto the first part,
2:10 right, the raw material, plasmid DNA. And plasmid DNA is a critical starting point
2:16 for protein productions, mRNA, and also some viral vector products. And here we essentially
2:24 are going to showcase a brand-new kit, our chemistry workflow, DNA 20 kb plasmid and linear
2:32 kit that's actually only launched last month. It is using intercalation dye for the detection
2:39 and is compatible with our single-platform – single-capillary platform legacy P100 plus
2:44 system and also the multi-capillary BioPhase 8800 system. So, we have a miniature version of the
2:52 BioPhase 8800 on the booth, so feel free to check out that. And because we're using LIF detection,
2:59 the sensitivity can be very good. It goes down to on the nanogram per microliter – or picogram
3:06 per microliter scale. And when it comes to plasmid analysis, one of the key points is not
3:12 actually the sizing, but rather the topology distribution. How much supercoil, open-circular,
3:18 or linear, or the aggregates of those supercoils exist in our sample. So, with this new kit,
3:25 we're able to demonstrate with a single-platform method that we can analyze plasmids ranging from
3:32 2 to 20 kilobases with the same method for a high-throughput environment. With this method,
3:38 we can separate the supercoiled, open-circular, and linear species, or sometimes the multiples
3:44 of the supercoiled if they exist in the sample in this one-platform method. So, this is more for
3:50 high-throughput. I have one method to run all kinds of plasmids. If we do need, as if needed,
3:57 a calculation is needed, a higher resolution is needed, we can always optimize the injection
4:03 volume, the separation voltage, and sample concentration to achieve a more suitable method for the
4:08 sample type. So, there's a lot of area for the optimization, too. So, it depends
4:15 on the environment you have. It depends on the sample requirement you have. The method can be
4:20 very flexible. But it's one ready-to-use kit. Another thing, right, we have the supercoiled
4:26 plasmids. And before it actually can be used either for mRNA IVT or protein production,
4:32 it needs to be linearized. And then, the next step for the plasmid DNA acquisition will be,
4:38 after linearizing, is my reaction complete? Do I have other fragments that I missed during
4:44 the initial assessment? So, the same kit can also be used for linear DNA quantification.
4:51 And if we have that running the same method, it also helps us to identify some of the peaks
4:57 in the original plasmid samples. So, by aligning the migration time, and, of course, if we use
5:03 a ladder together with it, we can determine or estimate the size of our linearized mRNA.
5:10 This is not a full mass spectrometry or sequencing, but it gives us a rough estimate, oh, am I having
5:16 the right product, roughly? And the linear DNA resolution is also very tunable. It depends on
5:24 the capillary length, right? In this particular case, it's showing a 30-centimeter capillary.
5:30 If we use the same material, the same ladder, but running a slightly longer capillary,
5:34 the 50-centimeter-long capillary, we can achieve basically baseline resolution even between 7,000
5:41 and 8,000 base pairs. And in this particular case, the plasmid we used and linearized is 7.8
5:50 kilobases. It sits right in between the 7,000 and 8,000 base markers. So, that's linear.
5:57 And another thing, right, we have known from both regulatory and previous studies that
6:04 host genome DNA can be critical in the sample. If we have too high or too big or too small,
6:12 then it causes problems later on. So, because we have a high-resolution size separation
6:19 and the LIF detection provides a very high-sensitivity quantitation method,
6:25 with these two coupled, we're able to actually do a host genome size analysis of a sample,
6:32 like how much is left and what size are those host genome DNAs. And because the cutoff is 200,
6:39 right, by spiking a 200-base pair marker in the sample, we can clearly see how much below
6:45 200-base pair and how much above 200-base pair. Another thing is this separation is
6:52 sequence-independent. We don't need to know the sequence. We don't need to design a product or
6:57 cell-specific primer or targeting to understand and determine the quantitation of a host genome's
7:04 DNA. So, that's about plasmid DNA. And moving on to mRNA, do we have the right product being produced
7:13 during the IVT process, right? Then the BioPhase ADN-100 system or the single capillary P100
7:21 system, when it's coupled with RNA 9000 purity and integrity kit, it allows us to use a CGT-LIF,
7:29 so against the LIF-based detection, to estimate the mRNA size when using it with a marker,
7:35 to quantify the intact mRNA amount or percentage purity. Or another thing is if we run a standard
7:44 curve, we can actually also determine or estimate the absolute quantification of the mRNA, and with
7:50 that data, by running a degraded sample and intact sample, we can even estimate the percentage
7:56 encapsulation efficiency. So, for encapsulation efficiency, right, because that's the ultimate
8:03 final drug product in one of the CQAs, and the common method will be the ribonuclease assay.
8:09 Following the same sort of philosophy and workflow as the ribonuclease assay, we do a
8:14 serious dilution of our intended mRNA to generate a concentration versus signal standard curve,
8:23 and then we run two samples, right? The total mRNA from a free, essentially degraded,
8:30 mRNA sample and the free mRNA from formulated mRNA lipid nanoparticles. And because now we have the
8:39 area counts or peak intensity from those two samples and calculate against our series diluted mRNA,
8:45 we can determine the encapsulation efficiency as well as the integrity of the formulated mRNA
8:49 versus the formulated lipid nanoparticles to get some insights on that as well.
8:56 Thank you.
8:45 and standard curve, we can have a microgram per mL or milligram per mL concentration for the mRNA
8:52 in both situations, and with a mathematical conversion, right, the total mRNA minus the free
8:58 mRNA divided by the total, the exact same calculation that we follow in the RiboGrain assay,
9:05 we can also calculate the percentage encapsulation efficiency. So, here's an example,
9:10 right, for this particular case, and it demonstrates how we can use the ladder and
9:16 this kit to estimate the size. The sample we are using here is a Firefly Luciferase mRNA,
9:23 so Fluc mRNA, and the theoretical size for this mRNA is 1.9 kilobase. And with running the ladder
9:31 and the sample in the same sequence, we can use the built-in algorithm in the BioPhase software
9:38 to estimate the size of those peaks. And the main peak is our Fluc, intact Fluc mRNA,
9:44 that's 1.9 kb as theoretical size, and we can average about 1.94 kb. And with a minor species
9:52 that's about 100 base pairs less, and with the hypothesis identity for peak two is actually
9:58 the tailless version of that mRNA, and because for this particular one we know the tail length
10:05 for that mRNA. And number three is actually what we just observed is accurate for the mRNA,
10:14 so that gives about 2.4 kb. And so, that demonstrates how we can use it for size
10:20 discrimination. And then when we deform, right, the mRNA lipid nanoparticle with Triton X solution,
10:27 incubation, and then our SLS, so sample loading solution, those two will break up the lipid
10:34 nanoparticle and also denature the mRNA to make them more straight. And then now we observe the
10:42 single intact mRNA in the sample as the aggregate peak, right, is minimized. And with that,
10:51 we are actually – so, this is, again, the purity assay for size discrimination. So,
10:57 because it's a purity assay, another key component we typically look for in purity analysis assays is
11:04 this method stability indicating. If I do have a bad sample or a stressed sample, can the assay
11:10 tell me that this is a bad sample? So, what we did was actually incubate the mRNA lipid nanoparticle
11:17 at 37 degrees for two days and five days, respectively, and then doing the same denatured
11:25 breakup, right, deformed mRNA lipid nanoparticle on the same analysis, and we can clearly see two
11:32 distinctive peaks, the 0.9 kb and the 1.1 kb show up in the fragment site. And as we get longer and
11:41 longer in the heat stress compared to two days and five days, we also saw these two peaks increase.
11:48 And with the total quantification, right, for intact lipid nanoparticles, the control sample,
11:54 which is always stored at minus 80 degrees, will have 92 percent intact purity. As we heat
12:01 stress the sample, we can see the intact percentage decrease significantly, and those two are the
12:08 fragments we were able to identify from the control to the stressed sample. So, it qualified
12:14 it as a good purity indicating method because it tells different stability. Now, compare those
12:21 two samples again, right, coming back, we have that standard curve, and we're able to add in
12:27 all these peaks together because we do believe those are still intact mRNA, free mRNA in there.
12:33 So, the free mRNA in the sample is 21 micrograms per mL. The total mRNA after the denaturing
12:40 comes back 436 micrograms per mL. Using that calculation, it gave us an encapsulation efficiency
12:48 of 95 percent. To demonstrate, right, if this is another orthogonal method that we can rely on,
12:56 while we're doing mRNA integrity assay, just adding a standard curve, we can get additional
13:02 quality attributes. We actually ran the RiboGrain assay, which is considered to be the gold standard,
13:08 the go-to method, right, for encapsulation efficiency. Based on that assay, this sample
13:14 gives about 92 percent encapsulation efficiency, and with the CE method, it's 95 percent. So,
13:20 it actually correlates pretty well, and we actually ran some other stressed samples with
13:25 30 and 70 different encapsulation efficiency with RiboGrain, and it came back around 40 and
13:33 70 percent, too. So, it correlates well, even though it's not an exact number match between the
13:38 two assays. So, moving on, for the poly(A)-tail and 5'-perm capping, I think we just heard fantastic
13:47 workflows on that, so I'm not going to dive too much. And what CE provides is an orthogonal method,
13:54 and what we're able to show is, for the poly(A)-tail, after the same digestion we just heard,
14:01 is we can do a single nucleotide resolution from 9 nucleotides all the way up to actually
14:08 160 nucleotides. That's what we were able to demonstrate. If we have a sample to go even further,
14:15 maybe we can also do that, right? But for this particular sample, it has a 120-nucleotide tail,
14:21 theoretically, and this is the poly(A)-tail distribution we were able to observe.
14:27 The center is located at 120-121 nucleotides, as expected, and then we saw almost nominal
14:35 distribution from 97 all the way out to 156. All of them are single base resolution, so we can group
14:43 them for quantification, or we can report the individual nucleotide distribution in that
14:48 poly(A)-tail. And then the capping and uncapping, right? Because this CE workflow provides single
14:56 nucleotide resolution, and between the capped and uncapped nucleotide or digestive product,
15:03 there's also one nucleotide difference. So, we are also able to separate the uncapping
15:09 versus the capping structures. But if we do want to know exactly what is the structure
15:15 for the uncapped species, right? Is it 2-phosphate, 3-phosphate, or all that modification,
15:20 we still need to rely on mass spectrometry. So, this is just another way to monitor the sample.
15:27 It's not intended to replace any deep characterization method. If we need to know that information,
15:32 we still need to go back to LC-MS using the phenyl column and also the non-TOF 7600
15:42 mass spectrometry. So, this is just another orthogonal method. So, enough of the mRNA core, and then
15:49 moving on to the lipids. There's generally four types of lipids involved in making a lipid
15:55 nanoparticle, right? The structure, the cholesterol in the ionized lipid, and the packed lipids.
16:02 In here, we focused on, for the deep characterization, we focused on the ionizable
16:07 lipids and using the ALC-0315 as an example because that's one of the popular ones that's
16:16 being used, I think, in the COVID vaccine and a couple of other products and pipelines we worked
16:23 with our collaborator. So, focusing on the ALC-0315, what I'm going to show is how can we
16:32 use LC-MS workflow to do a very clear, right, unambiguous identification of the impurity in
16:40 that sample, and also how can we monitor the quantification of that lipid. So, here is just
16:47 one example to begin with. In here, this shows the ALC-0315 structure. So, this particular part,
16:56 is the head group, and that's the tertiary amine that generally can be oxidized and cause big
17:04 problems, right? Another thing to keep in mind is the ionizable lipids are 100 percent synthesized.
17:14 So, with that in mind, we generally, with a synthetic molecule, right, a lot of times it
17:19 comes with a wealth of impurities that with small differences cannot be removed in a typical
17:27 purification process. So, the lipid raw material control can be very critical, and
17:34 quoted by our collaborators, right, that sometimes the quality of that ionizable lipid can actually
17:41 determine the destiny of the mRNA lipid nanoparticle project. And with that in mind,
17:47 when we look at the LC-MS, right, so what happens in this particular study is we formulated the
17:54 ALC-0315 ionizable lipid and incubated it at 60 degrees Celsius as a heat stress,
17:59 and for three days and five days. Then we run the control
18:01 sample, the heat stress sample, in one chromatography sequence. And by comparing
18:08 the difference and relying on the mass spectrometry for identification, we identified the 12.1 minutes,
18:17 that's the normal, right, the pure ALC-0315. And as we can see, compare the control versus
18:33 the stressed study, three or four, right, there's more down here that's much lower abundant,
18:39 but these four peaks change relatively higher abundance. And based on MS1 identification,
18:48 we're able to identify, okay, one is loss of SO group, and second is loss of the head group,
18:54 third one is oxidation, and the fourth one is loss of alkyl chain. But then the question is,
19:01 which one of them are the problematic one, right, PQA versus CQA? Which one is critical that we
19:08 need to monitor and control tightly? Based on literature, we know oxidation can be problematic,
19:14 but where is this oxidation happening? Is it on the N-terminal, which is the big one,
19:20 where the corporates can be the huge problem for the mRNA lipid nanoparticle,
19:28 because it's intact with the mRNA, and for the modified mRNA, causing loss of efficacy,
19:35 or is it some other position on the chain that's less disruptive? And then for that, MS1 is not enough anymore,
19:43 because no matter where the oxidation happens, it generates the same aspects. With the EAD,
19:50 right, electron association dissociation… electron association dissociation fragmentation technology
19:58 that's unique to the 7600 Xenoscope system, we're able to find unique fingerprints on the lipid product,
20:10 the ALC-0315, and by piecing those fragments together, the software we use is molecular profiling,
20:17 and that will spit out the ion formula based on M over Z ions, and then we're able to clearly
20:26 identify which pieces, right, each one of the fragments correspond to, and confidently identify
20:33 that the oxidation is happening actually on the N, the head group, which is probably a problematic
20:40 one, and then we can continue to monitor and characterize that particular modification during
20:46 the process of production or when we procure the material. And then another one to demonstrate
20:54 the power of the workflow is we obtained, right, our collaborator actually got the ALC-0315
21:05 and 9th bold lipids from three different vendors. We ran the same workflow and did a deep
21:11 characterization of those raw materials, and we're clearly able to provide valuable information for them
21:19 to make a decision, right, because as we can see, vendor two has the lowest purity percent,
21:26 while vendor one provides 98.2 percent of the intact or intended molecules, and also when we
21:34 look at the different modifications and impurities, they are actually ranging from 0.01 percent to
21:41 0.3 percent, adding all of them together, reaching about 1.7 percent of the impurity.
21:48 And I'm not showing the data here, but even with the 0.01 percent impurity, we're clearly seeing
21:55 all the fingerprint ionization fragments to give us unambiguous identification of that particular
22:01 modification. So, that's the lipid raw material, right? That gave us confidence to pick the right
22:07 material and to make the right lipid nanoparticles. But what happens afterward?
22:13 Following the administration, lipid nanoparticles can travel to different parts of the body
22:18 and undergo different metabolite changes, and frequently, right, ionizable lipids,
22:25 because it's a non-endogenous cationic lipid, can actually be used as a surrogate for
22:33 quantitative analysis of lipid nanoparticles in invaluable samples. Here, we are demonstrating
22:39 by spiking the lipid nanoparticles into plasma samples and incubate them. It's not actually a
22:46 clinical sample, but it's just a spiking matrix. And, again, we're using LC-MS, and the software
22:54 is a molecule profiler for those metabolite identifications. And, right, again, as I mentioned
23:03 previously before, there are four types of lipids to make a typical lipid nanoparticle. Here,
23:10 it demonstrates the fragment, right, metabolite product that can be identified in this workflow,
23:16 right, the cationic lipids. In this case, it's shown as ALC-0315. The sterol lipids in here is
23:23 using cholesterol, standard cholesterol, and the helper lipids, and we're using phosphatidylcholine
23:30 in here with a 36-carbon chain. And then, the PAK lipids, that's the PAK2000. So,
23:36 with the same workflow, we can identify all four kinds of lipids that we put into the lipid
23:42 nanoparticles. And then, using the ALC-0315 as the surrogate, because cholesterol exists in the
23:52 sample, too, and it's not easy to tease out which one is from the lipid nanoparticle,
23:59 while the cationic lipids, because it's a non-endogenous one, we can follow the percentage.
24:05 And in here, we essentially mix the lipid nanoparticles at different levels into the
24:10 plasma. And then, after incubation, right, monitoring, mimicking that PKPD process,
24:17 we're then using solid phase extraction to extract the lipids back out, essentially clean
24:26 up the sample from the plasma, and using the LC-MS workflow to analyze it. And all the data
24:32 points at each level of concentration were done in triplicate. And using the extracted ion that
24:40 can be essentially identified or input in the processing method in the molecular profiler,
24:46 we were able to quantify the different samples, and then we can see a very good linearity across
24:54 four logs of concentration. When we have four logs of linearity, it basically indicates that
25:01 we can quantify down to 0.1 percent with high confidence of these particular ionizable lipids.
25:09 So, with that, essentially, we're able to showcase workflows, maybe not exactly the
25:16 sample you are working on, but generally, right, plasmid DNA calculations, mRNA integrity,
25:23 encapsulation efficiency, structuralization of the mRNA, including poly(A)-tail and the capping
25:31 structures, and lipid impurity identification, and post-administration lipid nanoparticle
25:36 monitoring. So, all this gathered, right, that's essentially provided on the three instruments,
25:42 the single-capillary legacy P800 plus system, the multi-capillary BioPhase AD 800 system,
25:48 and our Xenoscope 7600 mass spectrometry. So, those two, we actually have a miniature on our booth,
25:56 and if you find any of the information interesting and helpful, feel free to stop by and find out
26:03 more details. With that, I'll take any questions.