Lipid nanoparticles (LNPs) are widely used vehicles for messenger RNA (mRNA)-based therapeutics and vaccines. However, ionizable lipids used in LNPs can be susceptible to N-oxide impurities that can cause functional loss of the mRNA cargo.

In a recent webinar, available on demand, Adam Crowe—Manager of Analytical Development at Precision NanoSystems ULC (part of Cytiva)—demonstrated how liquid chromatography coupled to mass spectrometry (LC-MS) using electron activated dissociation (EAD) was used to structurally elucidate lipid raw material, including localization of double bonds and saturated impurities and differentiation between oxidated species.

We are happy to be able to publish an expert session with Adam where he answers questions about the analytical landscape, the effects of N-oxides on mRNA, elucidating lipid structures with EAD and preventing lipidation of mRNA.

Analytical landscape

In your opinion, what are the best analytical techniques for assessing the purity of ionizable lipids?

Adam Crowe: At Precision NanoSystems, my team and I use at least three different methodologies to look at quality. We use charged aerosol detection (CAD) for the overall profile. We use the LC-MS/MS with EAD method described in my webinar, and we also use a fluorescence-based assay. The assessment should not be taken lightly, in my opinion. The detailed analysis of the ionizable lipid is paramount for the success of a project. In my experience, one of the most common ways that clinical programs based on LNPs fail is not enough careful assessment of the raw material.

Can you detect N-oxides with CAD?

Adam Crowe: Yes and no. You will run into two problems. N-oxides tend to elute very close to the main peak of the ionizable lipid. While you can chromatographically separate them, the gradients required are quite long and you will need prior expertise in what you are trying to separate. The other issue is the relative abundance. Because the N-oxides are an intermediate product that degrades further, you never form huge amounts of it. At ~0.1% relative abundance is when I start to get concerned about N-oxide formation. This makes it difficult for CAD to detect N-oxides because of the method’s limited dynamic range, adding to the challenge of having to know what to look for.

Effects of N-oxides on mRNA

At which levels do N-oxides impact mRNA efficacy?

Adam Crowe: This is an interesting question. It seems in very, very low abundance. We’ve had the luxury of looking at the adduct formation of 20 to 30 different ionizable lipids. Since the N-oxide itself is not reacting with the mRNA, but presumably an aldehyde—a degradation product of the N-oxide as described by Packer et al. in 2021—predictions are challenging. As a summary, I can say that when N-oxides are present in a significant quantity, meaning ≥1% abundance, we see very significant adduct formation.

How fast does mRNA lose its potency based on lipid-adduct formation?

Adam Crowe: mRNA potency loss is a multi-factorial problem. You have the effects from the adduct formation, which we talked about, but you also have degradation effects for RNA itself that can cause functional loss. You have to separate those effects. How quickly mRNA forms adducts is lipid-dependent. We have seen a broad range of several hours up to weeks, depending on the ionizable lipid used for formulation, while maintaining mRNA integrity. And I want to point out that there can also be events linked to other impurities derived from the synthesis of material used in LNPs. So, it’s a fairly complex question to answer.

Do you have any thoughts on acceptable levels of N-oxides or adducts?

Adam Crowe: As I mentioned before, N-oxide levels above a 0.1% threshold is where we start to consider adduct formation to be a problem. However, it’s the lipidation event itself that you will need to monitor and do rate calculations on to assess the severity. This is because the N-oxides are diagnostic, but not necessarily predictive of the rate of adduct formation. There are cases where you can see relatively low N-oxide amounts, but the rate of adduct formation on the RNA is quite fast. In such cases, it is likely that the N-oxides have already degraded to another reactive species.

Elucidating lipid structures with EAD

Do you use MS/MS with EAD only for raw materials, or do you also monitor N-oxides in formulated LNPs?

Adam Crowe: You can absolutely use the LC-MS/MS with EAD method I presented in my webinar for formulated LNPs. It’s obviously less complex to investigate a particular raw material compared to a formulated LNP because you have less species in a sample. However, it absolutely can be done. I recommend reaching out to your SCIEX representative as they might have further information on that topic.

Can you elaborate on how much MS method optimization is typically required and how much time you need to process the data? Can you efficiently transfer methods to new lipids?

Adam Crowe: It’s not a whole lot. Although EAD is very tunable, there are very discrete ranges for the type of fragmentation we are seeking for ionizable lipids. Generally, lipids require high-energy fragmentation for achieving relevant bond breakage—we used around 15 eV. If you want to determine the behavior of your specific lipids, you can set up a method with different energies within one injection. The data obtained by EAD are fragment-rich and manual analysis can take some time. However, SCIEX released a software that can process lipid EAD data, Molecule Profiler software, and it does a lot of the interpretation for you. Historically, at Precision NanoSystems, we would spend almost a week peering through the data and manually assigning what species are there. Now, this is done in a ~10-minute computational run through the software followed by a manual check, so it’s really convenient.

You showed a great comparison of EAD to collision-induced dissociation (CID) data. Can you explain how EAD compares to electron-capture dissociation (ECD) and electron-transfer dissociation (ETD)?

Adam Crowe: That’s a great question. There are a few unique differences to consider. Mainly, tunability and therefore efficiency of fragmentation, the mechanism of fragmentation and the reaction speed.

The kinetic energy of the electrons can be adjusted with EAD, which is crucial for fragmenting lipids. As Adam mentioned, around 15 eV is used to achieve diagnostic fragment ions to structurally elucidate ionizable lipids and their impurities. Adversely, ECD is a low kinetic energy fragmentation technique that usually operates around 0 to 1 eV. In addition, many lipids are singly charged, and ECD struggles with the fragmentation efficiency of low-charged ions. ETD suffers from a similar problem: It needs a multiply charged precursor to work efficiently.

While workarounds of creating sodium adducts of your ionizable lipid and impurities might get you to the multiply charged precursor, the energy for fragmentation is still limited with both ECD and ETD. The former uses only about 0 to 1 eV, as mentioned, while the latter uses a gas to transfer electrons, and there is only so much gas you can use. While EAD can also operate in ECD mode (low energy for fragmentation of proteins and peptides), the energy can be elevated to a mode referred to as “electron impact excitation of ions from organics” (EIEIO). The high-energy mode of EAD works very well for singly charged small molecules, including lipids.

The other aspect to consider is reaction speed. No transfer gas means fast reaction times that are compatible with fast data-dependent acquisition methods using analytical flow LC. The data presented were acquired using a reaction time of 30 ms for EAD. This means that even with very sharp peaks and multiples of these peaks coming off your LC, you can still acquire enough data points for reliable relative quantitation along with your qualitative data for impurity ID. Performing ID and relative quantitation in one injection can significantly save time and sample.

Could you please explain how to distinguish between the cis and trans double bond based on the EAD data?

Adam Crowe: The determination of the stereochemistry of double bonds (cis vs. trans or Z vs. E isomers) can be achieved with EAD as well. It’s based on differences in the relative abundancies of the radical fragment and the hydrogen-loss, non-radical fragment surrounding the double bound location when using fixed kinetic energies. It does require a bit of pre-work to know what a cis bond vs. a trans bond looks like. This differentiation was published by Baba et al. in 2017 prior to EAD becoming commercially available, and it can also be found in this SCIEX technical note.

Could you give some more detailed information about the MS method setup of the ZenoTOF 7600 system? Did you use targeted, data-dependent or data-independent analysis?

Adam Crowe: The method used was data-dependent acquisition (DDA)—or information-dependent acquisition (IDA), as some people call it in the industry—for fragmenting the top five candidates, combined with an inclusion list. The inclusion list contained the m/z of expected impurities of the ionizable lipid MC3, such as the addition of oxygen, demethylation, water loss, etc. More information on the method settings can be found in this SCIEX technical note. Depending on your needs, you can increase the candidate ions and adjust the inclusion list.

How could you determine that the oxygen incorporation on the side chain in one of the impurities was an epoxide and not a ketonic functional group?

Adam Crowe: This is a fantastic question. We were wondering the same when poking through the data: Is it an epoxide or a ketone? The precursor m/z would be the same in both cases. We therefore checked both options when investigating the MS/MS data. The fragments observed supported the theory of an epoxide, but we did not see fragments corresponding to the C=O linkage. I recommend reaching out to SCIEX if you want to learn more about that.

Preventing lipidation of mRNA

Do you have recommendations for how to mitigate adduct formation between the ionizable lipid N-oxide and the RNA?

Adam Crowe: It really comes down to the quality of your ionizable lipid. Ensuring that the amount of oxygen is minimized and that the ionizable lipid is not heated or exposed to oxidizing agents will help reduce the amount of N-oxides. Ensuring that the purification after synthesis is robust will help as well. Generally speaking, you want to carefully consider your manufacturing synthesis, mechanism, route and purification of your ionizable lipid to mitigate lipidation.

Have you explored options regarding the lipid synthesis to prevent N-oxide/aldehyde formation in the first place?

Adam Crowe: Confidential information about what type of synthesis and what ionizable lipid structures are amenable to adduct formation exists and can be taken into account. I cannot provide more detail on that, however, so I’ll direct you back to my earlier comment about ensuring high-quality synthesis. The quicker and dirtier your synthesis is, the more likely these reactive species are present.

What analytical services does Precision NanoSystems offer to clients? Do you test for such lipid impurities?

Adam Crowe: Absolutely. We are constantly looking for ways to improve LNP analysis and characterization. A huge element at Precision NanoSystems and in our biopharma services is to leverage analytics to reduce your time to get to the clinic. We therefore do very in-depth analyses on all incoming ionizable lipids, whether they’re from Precision NanoSystems or from clients themselves. We look for these adduct events and the N-oxide species, and overall we perform about 35 other tests. We’ve gained extensive knowledge over the years, and we’re almost at the point where we can predict whether or not your ionizable lipid is likely to be successful in a clinical program based on the results of our analytical tests.