WEBINAR | Taming the Viral Beast: Applying LC and LC-MS Technologies to the Analysis of rAAV Viral V
Hello, and a very warm welcome to everyone joining us to today's science webinar entitled, "Taming the Viral Beast, Applying LC and LC-MS Technologies to the Analysis of Recombinant Adeno-Associated Viral Vectors." My name is Matt McArdle, and I'll be moderating today's presentation. I'm delighted to introduce today's expert speakers from Waters Corporation. Principal scientist Stephan M. Kozah, and senior scientist Ximo Zhang. They bring several years of experience in LC-MS method development to this webinar, and today will be discussing the challenges associated with analyzing AAV samples, and how LC and LC-MS workflows can help provide solutions.
In the webinar we'll also cover how to use size exclusion chromatography for AAV aggregation and tighter analysis, how to optimize ion exchange chromatography for AAV empty/full capsid analysis, and developing LC-MS based methods for AAV protein and peptide analysis. After the presentation, we'll move on to our question and answer session. Please feel free to ask questions for the Q&A session at any time during the webinar using the tab at the left of your screen. Simply hover over the speech bubble icon, and click to ask your question. Without further delay I'm delighted to hand over to Dr. Kozah.
I want to thank both our speakers for presenting to us today. Thank you Matt, and thank you all for joining us today. Today we're going to be talking about some of the application of our current LCA and LC-MS technologies that we've applied for the analysis of recombinant adeno-associated viral vectors.
This is a brief outline. Matt covered this a little bit, and I won't go into it in too much detail, except that we'll talk a little bit about gene therapy and analytics, and how it applies to AAV. And then we'll go through some of the methods. And then Ximo will follow with a summary at the end.
The title of this slide is, "Gene therapy comes of age." And really, it's going to focus on AAV. Gene therapy has been used for a number of years and investigated. But with recent successes it's certainly seen a resurgence in interest. And there are a number of different ways in which gene therapy can take place.
But for today, we're really going to focus on the delivery of therapeutic genes versus a modified virus. And in this case, we're going to focus only on adeno-associated viruses. In the course of this analysis there are really two parts of an AAV viral vector for gene therapy.
There's the vector, or the capsid shell of a protein, and then there's the therapeutic gene that's going to be delivered within the captive. We today are just going to focus on the protein portion of that. We're currently working on and evaluating different methods for analyzing the therapeutic gene. And we'll be looking at such things as methods for capsid ID, empty/full ratio the purity of the actual sample, to basically monitor the stability in the batch to batch consistency of those particular preparations. In terms of the analysis of the AAV vector, we're really going to look at, really what are the goals of these analytics.
Are they stage appropriate? Are these methods going to be streamlined enough to really support process and product development, as well as make sure that we analyze the appropriate characteristics of the molecule in terms of the CQAs of the molecule, and the protein structure, and eventually the DNA that's within the AAV capsids. In reality, these tests can be really time consuming. In general, there's really a lack of an established analytical platform for these types of analysis.
But even more importantly, compared to a lot of our more current experiences with things like monoclonal antibodies, we really have a limited sample amount, as well as low sample concentrations, which makes method development far more challenging. Some of the characteristics we're going to look at for AAV, since AAV is the most used vector for these, or really, can we identify what the different serotypes that we're looking at are. These molecules are very big. And also, we need to consider the biosafety level.
And in this case, we're treating AAV capsids as really a biosafety level one. In this slide, it's really a summary of some of the methods we're going to cover today. On the left in the blue squares is really the traditional-- or are really the traditional-- methods for the analysis of some of these AAV characteristics. For instance, for purity, using analytical ultracentrifugation, or AUC, is commonly used. And we're going to show an example where we can get somewhat similar separations using SEC with either fluorescence or MALS detection.
For the amount of capsid per mil of the content of the sample, ELISA's typically used. We're going to look forward and see if we can use SEC as well for this type of measurement. In terms of purity, and in this case, we're going to think about just the empty/full ratio, or the amount of DNA that is in a particular amount of capsid.
Traditionally that is performed by AUC, as well as some other measurements that we'll talk about later. And what we're going to do is try to apply ion exchange chromatography using fluorescence for UV detection to monitor that. In terms of purity, which is traditionally done by SDS-PAGE, Ximo will looking at reverse phase UV, as well as MS detection, to monitor things such as the hydrophobicity-- changes in hydrophobicity profile fragmentation modifications to the structure in terms of capsid identity, where a western blot of the VP proteins, or viral capsid proteins, will be looking at peptide mapping of those proteins to look at fine changes in the amino acid sequence and/or the post translational modification of that capsid.
So, I'm going to start off with some work I've done on the analysis of these AAV capsids by size exclusion chromatography, looking at aggregation and fragmentation analysis for these capsids. To start with, we'll think about some of the basic principles of size exclusion chromatography, or SEC, in terms of understanding the separation that we're trying to drive for these AAV particles. The thing to remember about SEC is it separates proteins not based on their molecular weight, but really based on their size and solution. And the way that we can envision this separation taking place can be viewed in the bottom of the slide in the figures. If we look at the figure in the left hand side, we have an image that represents the cross section. We've taken a particle, and we've cut it in half with a broad ion beam of an SEC particle.
And what really drives SEC separations is, as the size of the particle is larger, as seen in the middle figure, it really can't enter some of the smaller core volumes. And when it does enter some of the larger core volumes, the larger particle won't be able to explore much of that available core volume, and it will be excluded from extending further into the particle structure if there are constrictions. So, ultimately, the larger a particle is, the less time it will spend diffusing in and out of the particle, and the more time it'll spend in the flow stream. So, those larger particles will come out first, and the smaller a particle is, the more of the internal core volume of the particle it can explore. So it will come out later.
It's important to remember, that in an ideal SEC separation, we'll develop a method in which we have ideally no absorption to the surface of the particle. We want the separation to be based truly on size. And as you can see, with those considerations, really the core size and the core volume of the SEC column that we select are really the most critical factors. And then, in terms of driving resolution, it'll be important to consider the size of the particle as well. In this slide, is kind of a thought experiment, or thinking about how big these molecules actually are.
So, we're pretty used to looking at monoclonal antibodies on the lower left, and a molecular weight of about 0.15 MDs, or about 150,000 kilodaltons. And we can see that an AAV particle, as shown in the middle, is actually significantly larger in terms of molecular weight than a monoclonal antibody. It's about 32 times greater in molecular weight.
But if we actually look at the size of these particular , structures the AAV, even though it's much more massive given its protein structure and the single strand DNA that's within that structure, it actually isn't that much bigger in solution. So it's going to behave very similarly, although a little bit bigger, certainly bigger, than a monoclonal antibody. When we start thinking about some of the other viral vectors that are used, however, for gene therapy, such as lentivirus or adenovirus, those sizes can approach 90 to 110 nanometers. So there, now we're looking at particles that are four times larger than that of an AAV.
And as seen in this picture, since you're going to see from the following data that the SEC column that we're going to have to use for the AAV is just barely a large enough average for diameter, that this column will not work for such structures as lentivirus or adenovirus. So, we are going to look at our BEH 450 angstrom SEC column for this, it was an average core diameter of 450 angstroms. And we have a pretty good sense that this column was going to be effective for the separation of AAVs, based on some previous data, we had with IGMs. Now, IGMs are nowhere near the molecular weight of AAVs.
However, based on their structure, their hydrodynamic radius is actually substantially larger. And we were able to find on our VH 450 angstrom column, that we could separate out the pentameric and dipentameric structures of an IGM. Therefore, we had a pretty good sense that we can go in, and look at, and separate out AAVs that were either in a single state-- so, by themselves-- or basically in a dineric state, where two AAVs are connected together.
So, we looked at the separation. And in this case, we started by just looking at an AAV8 sample, that was considered a null sample. That is a sample that doesn't have any DNA in it. Because we didn't want to confound our UV and RI detection with the presence of that single stranded DNA.
So, we could basically show whether or not we could, on this column, and by monitoring the SEC-MALS data, separate out the higher moleculate or dineric form, or two AAVs stuck together from the single AAV. And that's shown in the graph on the right, where we can certainly separate out the AAV that's by itself, versus a dineric form of that AAV. One of the challenges here is going to be really concentration. We had to take the samples that we had and actually get them to a concentration of nearly E to the 13th, in order to get enough RI detection to make it reasonably good, although it's still quite noisy, molar mass assignment based on the MALS. But we can certainly see, that we can separate on this column these two particular structures based on size.
So how are we doing this on a 450 angstrom column, when our particle that we're separating, or our analyte, is really about half the average core diameter. Well, remember it is an average core diameter. And I always like to look at really what's going on physically when you're thinking about SEC separations. So, we can see, on the right is an electron micrograph image of one of our BEH 2.5 micron particles.
And we can see that for this particle size, and for this pore size, that there are a lot larger pore structures near the surface of this particle, that are certainly capable of allowing AAV molecules to diffuse in and out of them. So what we envision happening here is, as we get deeper into pore structure, we're probably going to eliminate a lot of the pore volume in terms of the AAV being able to access it. However, near the surface of the particle we should have enough pore volume to actually drive the separation. And as we've seen in the previous slide, that's exactly what we've done. So, given this, I just wanted to give a couple of thoughts about things that I found as I went through and tried to develop some of these SEC methods.
Because remember, one of the important things we want out of this method is that we minimize the interaction of the AAV, in this case with surface of the part, but we don't want any hydrophobic or ionic secondary interactions taking place and confounding our separation based on size. Some of the things I found as far as general method optimization is that, I found that potassium chloride, or KCL, has been more effective than sodium chloride for minimizing some of these secondary interactions. I also found that the addition of other salts, as well as additives to the mobile face, such as arginine, or isopropyl alcohol, and using citrate in place of the-- or mass in place of the phosphate buffer, really didn't have any significant benefits. I also found that as we talked about, the concentrations of these samples can be really low. And if you look on the right with a sample that's at about the middle 10th E to the 11th concentration, we can see that the UV signal can get quite low in terms of trying to find the low level abundances of the dineric forms of the AAV.
And so what we've determined is that by using fluorescence detection, we can get about a tenfold fold increase in signal-to-noise, so that we can get a much better, and more precise, accurate measurement of the fragments that are in here, versus the aggregates in the monomer as well. And so what we're focusing on here, is the fluorescence is going to be based on the intrinsic protein fluorescence of the tryptophans that are in the primary VP1, VP2, PB2 proteins that make up the capsid shell. One thing to remember when using fluorescence is that, when you measure the high molecular weight levels, they will come out a little bit different for UV versus fluorescent. So, given the method development on AAV8 using partial and full AAV samples, I took the AAV8 method development, and applied it to the separation of different serotypes of AAV, from AAV1 to 5, as well as looking at 6, 8, and 9, and those are shown here. One of the things I found, is that by altering the amount of potassium chloride in these samples, the different serotypes actually require different levels of potassium chloride in order to give a good SEC profile, in which we can see that the main peak, the single AAV monomer peak, is coming down and actually returning to baseline quite nicely in all of these figures.
Without the appropriate amount of potassium chloride, and its data isn't shown here, what we would see is significant time before we actually returned back to the original baseline. And that's always an important thing to consider whenever you develop an SEC method is, how quickly do you return to baseline. Next, we will talk about how we can apply SEC as a way of monitoring the capsid content of our AAV separations. It becomes challenging to monitor concentration of very large analytes, particularly as the analyte size approaches the size of our UV wavelength, if we're trying to do UV absorbance or fluorescence.
One of the ways that we can minimize those changes, is by making sure that our sample is always in the same matrix as we make these measurements. And what we are going to do in this case is use SEC as really a buffer exchange, to make sure that our AAV capsids are in the same buffer as they go through the fluorescence or UV detection. As shown on the right, one of the things that we did notice is that, if we look at A280, because of the single stranded DNA that's in the capsid, we do get a significantly greater response for the amount of UV absorbance for the full versus the anti-capsid. In this instance, we're actually looking at the response vector, which would be the inverse of that.
And we're plotting it versus the mole fraction of full In this case, we have the samples fully occupied with a single stranded DNA, then the mole fraction of full is one, and if it's completely empty, the mole fraction is zero. So for A280 UV absorbance, we get about a three-fold increase in signal for the full fraction-- of an x full or the mole fraction of the full is one, versus when the mole fraction of the full is zero. Whereas, for fluorescence, we don't really see much change. Only about a 5% change. But we can use either of these methods, as long as we can determine what the mole fraction of the full is for the sample, or the empty full ratio can be derived from that as well.
Ultimately, we can use either our anion exchange method, that I'll talk about later, or AUC, or some other method for monitoring the empty full ratio in order to apply this correction. We can also see, because of the difference that we have for the fluorescence, that we may not actually need to correct full It's only about a 5% difference. So, if we don't require highly precise concentration back values, and we can tolerate a 5% variation in those. Or, if the samples that we're measuring really don't change, they're fairly consistent in terms of the mole fraction of the full sample, then we can really use fluorescence uncorrected. However, another option in terms of monitoring the mole fraction of the full will be to approximate that using our online SEC, and really monitoring the A260, A280 ratio.
And we'll show that on the next slide. Here we see, on the left, SEC chromatograms. They're running about two minutes on a short guard column, a 4.6 by 30 millimeter, 125 angstrom guard column.
Which is basically a pore size that should exclude the AAV from most, or if not all, of the pore structure. We can see the fluorescent signal, the A280 chromatogram, and the A260. And one of the things we certainly notice here, is that the signal-to-noise will be better for the fluorescence. On the right, what we have done, is we've plotted the peak area ratio for the A260 UV absorbance peak, divided by the A280 UV absorbance peak, versus the mole fraction of the full on the y-axis. And what we can see, is that we can predict this curve, as shown by the dashed line, based on the mixtures that we've made.
And it matches up quite nicely. Now, we don't necessarily recommend this as a way for measuring an empty full ratio. Because it will be somewhat more variable than using a method such as ion exchange or AUC.
But it's a reasonable approximation in terms of applying this to correct for our responses for our capsid content determinations. And that is what has been applied in this slide, where we can see that we have our different capsid titers, ranging from about 2E to the 11th, to about 2E to the 12th for our SEC peak areas. And these have been corrected for their measured empty full ratios. And we can see that in the orange, we see the fluorescence response, and signal. We get a little bit better r-squared than we do in the blue, but certainly the UV would certainly work.
However, one advantage that we would have, in terms of using fluorescence here, is as I said, if we don't really need to make this correction, and we can tolerate the variance of the fluorescence, there would be an advantage there. But also we can see that the fluorescence might be able to, based on just the signal to noise ratio, extend to a lower amount. However we didn't test that in these experiments. And to wrap up my section of this presentation, I'd like to talk about some work that my colleague Hua performed, in terms of developing an ion exchange method for the separation of AAVs, with an intent of looking at the empty full ratio for those AAV preparations.
AAV separations can be performed on IEX. There are, however, other common ways of monitoring the empty full ratio, or the full empty ratio of these AAV capsids. Probably the most common sort of gold standard method that we see for this determination, is analytical ultracentrifugation, where we measure the differential sedimentation rates. It is a fairly time consuming, labor intensive, and it requires actually a good amount of expertise to generate reliable and reproducible data. Uses a large amount of sample.
However, it does have the advantage that we can also pull out by AUC that evidence that there will be some capsids that don't have a full complement of single stranded DNA. They'll have less than a full complement of single stranded DNA. And those can be separated by AUC. Another way that we can monitor empty full ratios is really by spectrophotometry. We kind of saw an example about that as a rough approximation with the SEC work from before. However, to truly get this number to be more reliable, we have to tenature our capsids, and then measure our A280, A260 for the proteins, as well as for the DNA in a mixture, and make our determination about what percentage of the sample was comprised of the protein, versus which percentage of the sample is comprised of the DNA.
One of the disadvantages with this method is that we do have to worry about interferences as far as UV absorbance. We will demonstrate that ion exchange can also be used for this type of separation. The advantage of ion exchange is that it will use significantly less sample than the previous two methods, and it's fairly quick, but not really fast.
And ultimately, one of the disadvantages is that given our AAV samples, as we change the capsid structure, the sample may need-- or the method may need redevelopment. We'll also touch a little bit upon the use of charge detection mass spectrometry for the separation. And I won't go into that in too much detail, except to show a little bit of data generated by megadalton solutions for us on one of our samples. But ultimately CDMS will also use significantly less sample, and it's certainly a more definite measurement of the empty/full ratio. But it's a fairly new technique and requires a lot of expertise as well.
And as I pointed out in this slide, we can also get information on the partially full sample with CDMS. Shown on this slide is a separation that Hua developed for the determination of AVA empty capsids. So, we had samples in which we had fully empty and fully full capsids. And we mixed those at different amounts.
And we can see that overlay of those different mixtures on this slide. And we can actually get a really reasonable linear fit of the empty percent empty and sample for these particular capsid of preparations. It's somewhat of a pseudo-linear fit. And as we can see, this fit will really be dependent on how accurately we know the actual content, in terms of capsids per mil of our empty sample and our full sample. So, that will actually play heavily into how well this fits.
On this next slide, we are showing the overlay of two of those chromatograms. The two where we have the fully empty sample, which theoretically does not have any single stranded DNA in it. And that's shown in the green chromatogram.
And then the fully full sample, and that's shown in the black chromatogram. On the right hand side of the slide, we have examples of the CDMS, the charge touching mass spectrometry, provided by Ben Draper at MegaDalton Solutions, for these two particular samples. One of the things I want to point out with the profile in green, is that for the empty capsid we see a lot of charge heterogeneity that extends well beyond the main peak at 12 minutes, all the way under the empty peak. We were concerned about this, we didn't really know whether this represented partially full capsid, or some full capsid. So, by looking at these data with CDMS, and looking at the empty, we determined by CDMS that is about 99% empty.
So, we believe that, that overlap of charge heterogeneity of the green chromatogram of the empty into the full, is really based on charge heterogeneity of the protein structure, the structure of the capsid itself. So, from this perspective, when we look at ion exchange as a way of monitoring empty/full for AAV8, and we weren't able to obtain any empty and full examples for the other serotypes, but certainly for AAV8, we won't be able to use ion exchange to monitor partially full content for these structures. And there's probably some error associated with even the absolute empty/full. However, the advantage of the ion exchange, in terms of being easy to run using minimal sample, in terms of helping to guide process development to improve productivity of these AAV structures, while making sure that the empty amount of empty capsid isn't actually changing that much, it still has some value as a tool for that, as well as a secondary method for monitoring and for ratios, maybe even further on in terms of product characterization. And as a final slide on ion exchange, here we show where the separation on this particular ion exchange column using a gradient, we can see that the different serotypes actually behave somewhat differently. Some of them behave more similarly, like AAV2, 5, and 8 behave more similarly.
Whereas AAV1 and 6 behave more similar to each other. And AAV9 is actually not well retained. And we actually had to reduce the amount of ionic strength in the mobile phase in order to get it to retain. However, what that shows you, is that as you go from serotype to serotype, or variant of serotype to various serotype, you may require some additional reoptimization to allow your ion exchange separation. And that is the end of my portion of the presentation today.
And now, I would like to hand you over to Dr. Ximo Zhang. So, in the next part, I'll be talking about the impact and peptide analysis of the rAAV capsids, by using reverse phase LC and LC/MS. So, LC addressed earlier in this presentation, we learned that the shell of the AAV capsids are comprised of viral proteins, including VP1, VP2, and VP3.
And these proteins, in addition to protect the genes as a packaging material, they can directly impact the viral infectivity. Therefore, these proteins are always considered as an important part of the analytics of the AAV capsid. And as you can see on this slide, one AAV capsid is comprised of roughly 60 copies of the proteins.
And then these proteins, the mass is around 60 to 80 kilodaltons, so not that bad. And the ratios of the three proteins are around five to five to 15. But this is just a rough number. And then the accurate ratio identity and purity need to be categorized and closely monitored, as they are always considered as a CQA.
On the right side of this presentation, of this slide, you can see that it's a screenshot of the presentation by Pfizer at WCVB last year. In addition to the traditional gel method, which separates the viral proteins pretty well, and they also developed the RT-LC-MS method, in order to categorize the impurity profiles of the samples. And as this is grid work.
And you can see the resolution across these proteins are not really high, so that the ratio can not be directly measured. So, in this part of the presentation, I'll be talking about how we developed a new method of by RP-LC, in order to directly measure the ratio, identity, and purity of these viral proteins. So, to develop the RP-LC method, there are a few challenges that we need to address beforehand. So, the first challenge is that the lower concentration that we saw in the AAV samples, which is 10 to 50 times lower, compared to our typical mass samples.
And then, another challenge is that there were always surfactants in the AAV samples, such as poloxamer or PS-80. And these surfactants can also mess up with the secretion by reverse phase. And on top of that, we didn't have a lot of prior knowledge in developing these kind of method for AAV samples, because it's relatively new, and not a lot of literature are talking about them. So, our goal would be to evaluate our technology, and going through the extensive method of development, in order to develop the high resolution RP-LC method. So, our first step is to remove the surfactants in the sample.
And as you can see on this slide, our product corner is the reversed phase analysis of our formulated rAAV samples. Against these, the surfactants, the big peaks of the surfactants at the beginning and at the end of the chromatogram, as you zoom in, the separation around 10 to 15 minutes, you can see the increase of the baseline, which are from the responses of the surfactant. And after developing a method to remove the surfactant, you can see at the bottom of the chromatograms, on the bottom of this slide, you can see the baseline is flatter compared to the top chromatogram.
And that's from the reduced surfactant signal. And in addition to the improvement of the chromatograms, you cans also see the reduced noise of the ESI chromatogram. So, as shown on this slide, the right hand side, you can see the before and after comparison of removing the surfactant. And with the surfactant removed from the sample, the noise level on the ESI chromatogram, on the ESI spectrum, is greatly reduced. And our next step is to develop the RP-LC-MS method, in order to achieve a high resolution separation for the captive proteins.
So here, we start the newest MS platform, which is the BioAccord system. And we started from the BEH C8 column, which is kind of like the golden standard of the industry for separating these proteins. And on the right side of this slide, you can see the separation of the capsid proteins of AAV8 by using formic acid. And you can see the separation is now really good, and we can capture the dilution of the proteins. We can also see the bumpy baseline from the surfactant And by using a new multiphase modifier, which is the LC-MS or DFA, we can see that the resolution across the proteins are increased.
And also the baseline is flatter, that means the surfactants are being separated from the protein. And also the peaks are a lot sharper, which means some improvement on the MS signals potentially. And other than the mobile phases, and the stationary phase can also have an impact on the separation. So, here it's to show the comparison of different column chemistry on the separation of the VP protein.
And as you can see it here, compared to using C8 and C18 column, using the C4 column can provide better selectivity on the critical pair of the protein. So, at the beginning of the chromatogram, you can see the separation between these two peaks are increased by using the C4 column. But that's not always the case. So, during the method development process, we did observe the sample dependence on different AAV serotypes. So, this sample, which is AAV5, you can see that using the C4 column, the resolution is still OK, but we're kind of missing one of the VP proteins, which is the VP1.
And you can see at the bottom of this slide, which is the separation of using the C18 column. And then we can see they include the recovery of the VP1, and which you lose later than the VP3 peaks. And this might be because of the increase, the hydrocapacity, and that we observed on this amino acid sequence.
And with the optimized separation, we were able to obtain the accurate mass of these VP proteins. So, on this slide, you can see that the separation of the AAV8 with the 0.5 micrograms of the sample on the column. And then we were able to obtain the masses of all VP proteins, including a fragment of the VP3, which labeled at the lower right hand corner.
In addition to the accurate mass, we are also able to see some selected modifications, such as the phosphorylation. So, on this slide, it shows the VP1 and VP2 with the phosphorylation observed the MS spectrum. Which has the plus 80,000 masses. And with UV protection, we were able to obtain the relative abundance of each VP protein. And as we can see, when the resolution is high enough, we can just directly integrate the peak, and then calculate the relative prism of this peak. And then we can get the ratio of the VP protein.
The table with MS on the right side, you can see we're able to calculate the relative abundance of each VP proteins, including the fragmentation. And similarly, as you presented at the beginning, so using the fluorescence as the detection method, we were able to increase the sensitivity of the detection. So, here shows the comparison of using UV and fluorescence as the detection method for separating the VP proteins, and with the same amount of sample, which is 0.05 micrograms of VP protein. And we obtained about roughly 20-fold higher signal to noise ratio level by using the fluorescence. So, with the optimized LC-MS and fluorescence method, we were able to analyze a panel of AAV serotypes. So, here this is the six different AAV serotypes, including AV1, 2, 5, 6, 8, and 9.
And we can see there's a retention time shift on the main peaks of the AV serotypes on different AV serotypes. And with the additional detection method from the MS, we are able to identify these separate peaks. And therefore, it can be used as a semi platform method for identification and also categorization of the different AV serotypes. So here, I just want to show you a case study of how we used the developed method to support the process development of AAV vector manufacturing.
And this project was done through a collaboration with the Dr. Xiaoying Jin at Sanofi Framingham. And the beginning of story, that Xiaoying observed some different potencies between the AV samples that she collected from different batches of the AV samples from the bioreactor. And then I was going to see if we can compare the analytics of these samples.
And then in order to understand why their potency is different. So, the measures we decided to compare, is that the first one is the RP-LC profile, and also the ratio of the VP proteins, as well as phosphorylation level of the viral protein through MS. And then at the start, we separated the three different samples, 1A and 1B, which are from the same batch, and also to 2A, which is from the second batch with different potency. So, the first thing we noticed is that the sample 1A and 1B are quite similar, while the sample 2A has a significant lower percent-- significant lower signal-- compared to sample 1A and 1B. And when we overlay the chromatograms together, we can see that the percentage, or the VP2 peak, is significantly lower compared to the other two samples. When we calculate the relative abundance and make the comparison, you can see the results on the left side of this slide.
And that with the bar plus is the relative abundance comparison across different VP proteins. So, we can see from the batch lot, the VP2, the percentage of VP2 of sample 2A is really lower than compared to the other two batches. And also the percentage of VP3 is higher compared to the other two batches. And I know it's difficult to see, but there are arrow bars on this on this bar plot. And this is too small to see probably. So, that arrow bar was collected from three different analysis, which demonstrated that the differences with the results here are real, and not from the variations during the analysis.
And in addition, the results we got here kind of agreed with the CSTS analysis from Sanofi. So, we determined that we can trust the analysis method by the RP-LC. And the level of phosphorylation is listed on the right side, which is quite similar, that you can see that the symbol 2A does have a higher-- doesn't have a higher phosphorylation compared to VP1A 1B be on the VP2. And also the phosphorylation on the VP1 prime, which is one of the isoforms of VP1. It is slightly lower compared to the first batch of the samples.
So, generally speaking with the developed RP-LC MS method, we were able to monitor the changes in the samples from different batches, and also provide information, and also to monitor the changes in the process development. The next part of this presentation is the peptide mapping of the capsid proteins. The purpose of doing a peptide mapping on AV capsids-- on AV vectors, are having multiple effects.
So, the first point is that we can't use it as a method to ID the capsid protein. Or ID the AV capsids. So, although there are only 13 common AV serotypes available now. But people are doing molecular engineering to the capsids.
They are to changes transduction, activity, and also other properties. So, there are more and more AV serotypes available. And some of them only differ by a few amino acids. So, through the impact analysis, it would be difficult to categorize the difference through the serotypes.
And peptide mapping would be a good method, in order for identification of these AV peptides. And in addition, the PTM, some PTMs, such as the deionization or oxidation are difficult to be captured through the intact protein analysis. So that using peptide mapping can also help with PTM characterization.
And the peptide mapping can also determine the cysteine oxidation state, which is kind of like the peptide mapping, and can also categorize the unknown and impurities, such as the HDP analysis. And of course, there are challenges in developing the peptide mapping approach. So, the first line is still the limited sample amount. So, if we have enough sample of the AV, there will be just like routine peptide matching method for MAb. But then typically, the sample amount we're getting are around micrograms per milliliter, which is really low, and making it really difficult to go through a routine buffer exchange step. And that poses challenges in developing this peptide mapping approach.
And also, the surfactants can also mess up with the separation of these peptides. And if we really are going to do a buffer exchange to remove the surfactants, it will further reduce the recovery. So, which making the development-- which making the method development even more challenging. So, at last we finally were able to develop a peptide mapping approach on the AV capsid proteins. So, we did go see the samples, and to capsid proteins, and then go through a buffer exchange through fraction collection. And then digest the samples, and use LC-MS to analyze this protein-- these peptides.
So, here shows the results of the digestion method. And this method was developed with only 1.25 micrograms of protein as a starting material.
And we can see that the sequence coverage is pretty good, as you can see from the coverage map at the bottom of this slide. And also from the bar plugged at the lower right hand corner. You can see for all VP proteins the sequence coverage is about 95%, which is pretty good. And then we are also able to locate the terminal peptides, as you can see from the upper right hand corner.
So here are the MS/MS spectra of the VP1 and VP2 terminal peptides. And the methods are about 2.1 kilodaltons for VP1, and also just only a few hundred of VP2.
So we were able to see the pattern of these peptides using a lot of BY ions, and which helps us to identify these terminal peptides. So, in addition to the identification of interminal peptides, we are also able to identify some of the modifications in the samples. Such as this assimilation, phosphorylation, deamination, and oxidation.
So here on this slide, you can see the chromatogram one of those upgraded peptides. And down here is a little peak, which is one of the deaminated peptides, on the VP1 protein, I think. And here on the right hand side, you can see that the fragmentation spectrum that we got from one of our newer MS instrument, which is Synapt XS.
And despite of it's low abundance, we were able to get a pretty good fragmentation on this deaminated peptide. As you can see, there's a lot of BY ions, which means the identification is meaningful. So, to summarize the presentation, we developed a SEC method to measure the aggregation and the titer of the AV capsids. And also using the developed ion exchange chromatography method, we were able to identify and analyze the empty/full ratio of the captid. And also DFA as the RPLC-MS modifier.
It can include both separation and MS performance for the capsid protein analysis. And as always, the fluorescence detection provides the higher sensitivity in all the analysis, that we developed. And the peptide mapping is kind of tricky, because of the low sample amount and the surfactant. So, it will need a complete redevelopment of the method.
And if you are interested in any of the analysis that we are talking about today, there are a few AV application notes available on waters.com. So, the current members are listed on this presentation. And at the end, I just want to thank our colleagues at Waters who contributed to this project, as well as our collaborators at MegaDalton Solutions for providing the CBMS results. And also our collaborators at Sanofi, and our reliance for providing the sample and serotype information.
And at the end I just want to thank you for your attention. And now that we are happy to answer any questions that you may have. Thank you Dr. Kozah and Dr. Zhang
for that interesting presentation. I hope that everyone listening enjoyed the webinar and found it informative. We'll now move on to the last part of today's webinar, the question and answer session. As a reminder, please feel free to continue asking questions using the tab at the left of your screen.
Simply hover over the speech bubble icon, and click to ask your question. And we've had a lot of questions come through. And our first one is, are there any examples for successful analytical chromatography of lentivirus-- lentivirus, sorry. Do you want to take that Ximo, or do you want me? It is for-- OK I can take this. So, for lentivirus, I can only speak to the RCMS part probably. So, just like the AAV, and we can probably use this access to deassociate the virus, and then just analyze the proteins for-- through LC and MS. And we probably
need to do some method development through that. Currently our lab is just a BSL level one lab. So, we will need some further development on that part. And in the future we're probably going to look into that. And I would add that for the SEC and ion exchange, we haven't done any work with lentivirus as an intact molecule for those same reasons.
Thank you both. Next question is, which protocols could be applied for moving surfactant from AAV samples? OK, I'll take this one. This protocol was-- we recently developed this protocol to remove the surfactant. So generally speaking, it's just to use the Amicon swing filter to remove the surfactant by buffer exchange.
So, they have a detailed protocol, step-by-step protocol listed. And if you are interested, I can forward that to you later on. Thank you.
Next question is from Michael, and he asks, is there any special sample preparation necessary for analyzing the AAV by RP chromatography? For a sample preparation, I think typically people use about 10% of citric acid to dissociate the AAV capsid into the viral protein. But I think the reverse phase chromatography is also in acidic condition with a pH two to three-ish. And so, without any sample preparation, if you are sure that your sample doesn't contain any nasty stuff, you can just directly inject the AV capsid. it will dissociate, but still if you treat the samples with acidic acid for a while, you probably will have more reproducible results. Thank you.
Our next question is, does the SEC method detect peeks of AAV clipping forms? I can take that one. Yes, there was only one example-- actually two-- but the most predominant one where we saw protein fragments-- and remember we're using fluorescence detection, so it's selected for proteins versus DNA-- was I think the AE9 sample that we looked at. And that's one of the advantages of the 450 angstrom column is, you do have the ability to get a little bit better resolution on those fragment side of the intact capsid. Thank you for that.
And the same attendee asked another question, which is, were you able to identify the phosphorylation site by LC-MS/MS? Yes. So LCM-- specifically we looked into the sites for PTM through the peptide mapping method. And through peptide mapping, we did find different PTMs.
And there was some phosphorylation signs that this was still under NDA with our collaborators. So we didn't put any of those information entered in the slide deck. But through the method we developed, we can see that there are some sites that are prone to phosphorylation.
Thank you. Hopefully you've still got the slides in front of you, because one question actually asks, how much material was used for injection on slide 39? I don't know if you're able to answer that. 39, which one is 39 again? Let me check. OK, so slide 39 is probably the separation of the six AV serotypes by the optimized LC-MS method. So, here are the samples from different vendors, so they have different concentrations.
But they are roughly on the range of the injections-- the method was roughly from the range of 0.02 to 0.05 micrograms. So around 20 to 15 milligrams. You can see the sensitivity-- which is color, good, we still got the different signal with that much of sample. Thank you.
And we've got a lot of questions coming in at the moment. We're going to have time to fit in just a couple more. But keep them coming, we'll make sure we get to them in one way, shape, or form. Our nest question is, under what circumstances would you recommend fluorescence detection of a UV-based detection for the SEC assay? In general for the SEC method, we would typically recommend the fluorescence method, simply because the sensitivity is just going to be that much better for the low abundance aggregate forms, or fragment forms, that are generally present in these samples.
And generally, these concentrations are low, and I think using the fluorescence is going to give you an advantage. And it's also going to be very selective for monitoring the protein component versus the DNA component that's in the samples. Thank you. I think we've got time just to squeeze one more in.
The short SEC UV method for tighter measurement seems to be more efficient. Is it possible to be used on E/F analysis instead of IEX? As I mentioned in the presentation, while it's, I think, a reasonable way of estimating the approximate empty to full ratio, one of the disadvantages that it is going to have, is particularly as you get to very high purities, that the uncertainty in that empty to full ratio is going to get larger and larger if you look at that response curve. So in general, what we position that method for, would be if you're actually-- if you're looking at samples are relatively low in purity, so there's a fairly high abundance of empty, it may be a decent way of making quick estimates for those types of samples. But it's certainly not going to be good for looking at high purity samples. And furthermore, it will be good for estimating empty/full ratio, to make response factor corrections for the SEC method, titer method.
Thank you. Well, unfortunately that's all we have time for today. I'd like to thank our expert speakers for today's informative discussion, presentation, and Q&A session. And thank you to everyone joining us online, and for interesting questions. I hope you found it a worthwhile session. I know you've had a lot of questions in today.
And if we didn't get to yours, or you have any other questions, please feel free to email me at editor@SelectScience.net, and I'll follow up with your questions for our speakers. Remember, you can download related resources in the tap at the left of your screen, including a certificate of attendance. And if you'd like to listen again to today's webinar, or invite a friend to listen, it'll be available to watch on demand in a few days' time. Goodbye, and thank you once again for joining us.