Kelly Lundsten: So, my name is Kelly Lundsten. I'm the Business Segment Manager for Advanced Cytometry for BioLegend
It was about exactly a year ago this month that we first released Brilliant Violet 421. Brilliant Violet 421 was a bit of a revolution in fluorescence chemistry, and it has significantly helped us grow and expand our ability and the cleanliness with which we can do multicolor flow cytometry
And, as of next month, we'll have released seven different Brilliant Violet conjugates that will all be excited off the violet laser, that will replace Qdot nanocrystal or nanocrystal technology almost entirely. So, that's what we're going to talk about today, are all the implications of new fluorescent technologies on multicolor flow cytometry and how we're going to incorporate those into our panels
All right. So, first what I'm going to discuss--my expertise is actually in fluorescence chemistry, so I think it's extremely important to acknowledge not only the strengths, but also the weaknesses of all technology platforms that you use. And, all fluorophores are not created the same, and it doesn't mean that because something has a weakness, that we don't use it. It means that we understand how things work in order that we can balance them into a panel
And, that's actually why I call this talk Multicolor Qi, because there are no right or wrong answers. There are no ubiquitous statements that we can make to be true for all things. It's about understanding our biology, our tools, our instruments, our reagents, and being able to balance them together in order to get the results that we want in a high content, high throughput manner
So, first what we're going to talk about are the different fluorophore families and the limitations of those chemistries. We're going to also talk about how that relates, then, to a balance in panel composition, so things like--a lot of these might seem quite trite--balancing dye expression--or, I'm sorry, dye brightness with antigen expression. That seems extremely easy, right
And, when you're doing five-color flow cytometry or six-color fly cytometry, these things are actually quite easy. It's when you start getting above 8 to 10 to 12 to 15 to 17 colors that this becomes actually a very significant challenge, for lots of different reasons that we'll discuss as we go along. So, there are different implications for each of these different factors
Second would be identifying different biological concerns or assay concerns. These would be things like PMA stimulation, where your CD4, which otherwise you might want to put on a relatively dim fluor in order to free up some of your better, brighter fluors, no longer can go on a dim fluor. It has to go on something that's actually quite a bright fluorophore
That would be other things like using chemotherapeutic samples, things where you don't know the time from which that patient receives their last chemotherapeutic event and the time at which you're actually going to analyze
There is a ton of endogenous autofluorescents in the world. You are entirely autofluorescent for a lot of reasons, right? And, that's the same with a lot of drugs that we use and a lot of chemotherapeutic reagents that we use. They're often extremely highly autofluorescent
And, that's actually the same with your different cell types. You know, my skin is doing a fantastic job at protecting my DNA from the assault of UV radiation from the sun because it's filled with a bunch of autofluorescent molecules that are dissipating that energy for me in the form of antioxidants. So, whether or not I'm using skin or I'm using spleen or I'm using brain, all of these are going to have varying levels of background
There is going to be, then, limitations for our instrument. You know, we're really quite lucky that we have instruments that are 20 parameters or 18 colors. You know, a lot of the reagents that we have are not quite there yet to give us really easy to construct 18-color panels. I'm pretty much maxed out at 16 at this point. But, you know, we do have instruments that are going well beyond that, like the Astrios from Beckman Coulter as an example
So, you know, electronics is definitely leading the way, or engineering is leading the way. We need to start responding with better reagents and better fluorophores in order to have a faster, higher throughput system
And, then, finally, what I always say is when you have all these different variables, you kind of understand them. You have your best first guess at what a good panel construction would be. It's test and repeat, test and repeat. When you're going to 15 colors, I can guarantee you it's not going to work the first time you do it. It's not reasonable to think that you're going to have a lot of variables put together and, without proper testing, that it's just going to miraculously work out of the gate. It doesn't work that way
But, it doesn't mean that it's not worth doing, especially for samples that are precious or things that you want to do in a high throughput manner
So, we're also going to talk not just about how to balance these factors in panel construction, but also then all the controls that go into having confidence in your data
And, you know, there is no single perfect control for all things. Isotype controls are not the perfect ubiquitous control for the background of all things. That's not what happens in multicolor flow cytometry anymore, because all of a sudden we have a ton of protein because all of our PE and APC conjugates are protein, and they're very big protein--you know, pieces of protein
And, there can be a lot of non-specific binding due to the fluorophore, due to antigen specific interaction. There's a lot of background. It's not just non-specific binding, but it's the background of all these things spilling over into each other that requires us to start using things called fluorescence minus 1 controls, and those are gating controls
They let me know that not only do I have an accurate gate every time I run that assay, right, so that I have internally consistent information, but then it also lets me know when my reagent is starting to degrade on me and I have to get a fresh reagent, because the percents are changing based on the FMO. So, it allows me to have confidence and reproducibility inn the data that I derive from a 15-color assay
There is a time and a place for isotype controls. I used to say that they were completely worthless, but that's because I'm not an immunologist. I have since understood that there are things like macrophage or dendritic cells or monocytes that do have a significant amount of non-specific binding that is due to things like cy5 or, just generally, they're scavengers. They're going to pick up lots of things. And, so we're going to show an example of where that actually is an important control
But, generally, the entire talk--and really what concerns most people on a day-to-day basis is not compensation, although they put that word in their brain as if it's the thing to fear. What they're really worried about is a loss of resolution or a loss of sensitivity, because what we're going to talk about extensively is just how intensely meaningless that term "compensation" actually is
I can change that value just by turning up my voltage, but it does not change the resolvability of my population. Those two things are not synonymous. They're involved with one another, but they are not synonymous. So, we're going to go over that at every stage throughout this entire talk
So, let's just start with the fluors. This is going to be about the first third of the talk, honestly. I think, from my perspective, that it's the most important thing. You've got 15 of these things going on in your assay, and they're just as important as your antibody. You're not seeing your antibody. You're not seeing your marker. You're seeing your fluorophore. If your fluorophore isn't linear, right, if it's non-specifically interacting with things, you cannot trust your results, right? You need to understand how it's working in order to understand how to troubleshoot it
So, historically, we were always kind of stuck with or relying heavily upon organic--simple organic structures, like the Alexa Fluors, the Cye Dyes, now the Dylight Dyes, Megastokes Dyes. There are all sorts of fun dyes, right? One of the reasons that we love them is that they are derived in nature, right? The FITC structure, for example, is related to antioxidant molecule, a B vitamin that we find all over our skin. They can be tuned to whatever particular wavelength we are interested simply by adding side chains or adding different--you know, a phenyl ring or, you know, adding to the structure
For organic fluorophores, when they are really small, like this, things in the coumarin family, where they're naphthalene derived, they tend to be excited by extremely high energy, short wavelengths
And, this is an inherent limitation to their structure. It's advantageous that they are capable of being excited by short energy. However, one of the biggest limitations is that because--you can kind of imagine that each double bond is sharing an electron pair, and those electrons are what is actually absorbing the energy
This molecule, because it's so small, only has a limited capacity to ever absorb energy. And, basic physics tells us that I can't emit more energy than I absorb, right? So, inherently, blue will always be dim as a fluor, and because the structure is very simple, it's also going to be very sensitive to photobleaching. But, the brightness of this molecule, or its capacity to absorb energy, is only one factor. It's also background, and as we all know, in biology, because we're always protecting our DNA from the sun, most of these molecules are also autofluorescent. We have a ton of autofluorescents in the blue emitting range for exactly that reason
So, as I start to tune this molecule to become bigger and bigger and bigger and bigger, I encounter a family called the rhodamine-based fluorophores, and these are many of the Alexa Fluors that are in the visible range, things like Alexa 568 or Alexa 594, a Texas Red
It's extremely easy to be green, right? This molecule is super small, super soluble, has really nice tight emission and excitation. It gets significant harder as we try to tune this molecule to be further and further near infrared, for example. The molecule becomes bigger and bigger, and I only have so many places where I can add side chains where this molecule stays soluble
One of our biggest challenges in making PE Texas Red or any of the Texas Red derivatives is that these molecules start to become relatively hydrophobic as these rings become bigger and more bulbous, and so we end up having to use a ton of--the molar ratio of Texas Red to the phycoerythrin just becomes prohibitive, and it tends to precipitate quite a bit, and that's one of the limitations that we have in making a really good PE Texas Red tandem
So, in the near infrared, we come to rely really heavily on cyanine dyes. You can see that this structure is significantly different than the rhodamine based dyes. They have their own strengths and weaknesses, right
These guys--their strength is that they do get us into the near infrared, right? That's a pretty big strength. One of their weaknesses is that because they're a dimeric molecule, they're a reflection of each other. If I'm a reactive action species and I'm really hungry and I need to neutralize myself on a double bond someplace, I'm going to look at this linker and say, "Yum, yum, I know exactly where to neutralize myself.
So, these molecules a lot of--you'll hear a lot of people say, "I think Cy5 has been cleaved from my tandem." Cleaved? Not exactly. More like photobleached. There is the potential that it's been cleaved once one of these bonds has been broken, leaving behind half a molecule that's no longer capable of the same excitation and emission properties, but more often than not, these molecules can actually be quite photo-unstable as well
And, that's just a nature of being an organic dye, right? They will not have absolutely fantastic photo-stability. They can't. It isn't their job. Their job is to absorb energy, and being in that activated state is a sensitive place for them to be, which causes them to photobleach. It's a natural process of life for them
If, however, dyes are released that come under different names, things like H7 or different--V450, if the spectra is identical to two--between two different fluors of different names, then the structure is the same. They have not literally made an advancement outside of sulfonating a molecule or emanating a molecule. There aren't many changes that I can make to a molecule to actually make these guys all that much better. I'm still--I'm limited in how far I can tune them and still meet the requirements for physics
So, that's actually, you know, why we rely so heavily on proteins. So, proteins are the second family that we've historically used for the last 15, you know, plus--you know, 15, 20, years
And, why we love proteins like phycoerythrin is that this molecule is 240 kilodaltons. It's twice the size of my antibody, and it's a protein, right? I can't freeze it, because it does not like to be frozen and thawed. You know, that particular process is extremely hard on any protein, including my skin, right
But, why we love it is that--I like to put this structure up because I think most people think of PE as being its own fluorophore, and it's not. PE can have anywhere between 10 to 20 chromophores that are embedded in the protein structure. The majority of the 240 kilodaltons is protein. It's protein that's embedding the fluors so that I can have the benefit of a one-to-one conjugation between PE and my antibody, but be benefitting from the fluorescence of 10 chromophores
All right, that's why we love it so much, and that's why it's almost impossible to make a synthetic phycoerythrin. It's actually extremely difficult even to make it in a bioreactor because of the tertiary structure. The folding process is quite intricate. Some people try baculovirus for that reason, but it never produces typically as good of a PE as if you actually just purified it out of the algae
So, that purification process, that production process--you can imagine that sometimes we might have a limitation on how much PE is available when things like tsunamis happen, because there goes all of our algae that we were trying to use to purify for our biological assays
So, you know, PE, APC, PerCP, GFP--these guys will also be extremely sensitive to solvents. I don't know how many times on [unintelligible] you see over and over again, "When I do microscopy and I've got GFP expressing in my sample, I don't see it after I fixed it." Well, of course you don't see it. Would you dip your hand in methanol? Don't dip your phycoerythrin or GFP or any other biological specimen that contains a protein that needs to be fluorescent and rigid into a denaturing solvent. It's going to collapse--it's going to no longer--it's going to release it from its rigidity, which is an inherent factor that's required for it to be fluorescent. It's going to relax it so that it's no longer capable of being fluorescent anymore
So, in order to get us more into the near infrared, organic fluors have a strict stoke shift. They have a strict excitation and a strict emission, and there's not a very big distance between those two things spectrally. If you see a fluorophore that has a very short excitation but a very long emission, it's a tandem, because it's physically extremely difficult for the fluorophore itself to have a big stoke shift
So, that's why we rely on these guys--excellent donors, right? These guys have over 1 million extinction coefficient. Excellent absorbers, excellent donors, which makes them great partners for a fret, right? So we use things like Cy5, Cy3, Cy7, Cy-whatever to couple with these guys in order to get a further near infrared emission, thus giving us more colors that we can analyze at a single time, right? But, limitations are definitely stability and size, for sure
So, along come nanocrystals. You know, nanocrystals have strengths and weaknesses as well. Back in the day, you know, five, six, seven years ago, they were the only thing that got you above 10 colors. That's their strength, is that they gave us the ability to do multicolor
Their weaknesses are that in space, they are about the same size as a phycoerythrin, but they're like a bowling ball, right? So if PE is a big piece of protein snot, I'm an antibody, it's shackled to my ankle and I want to get through that door, I can probably pull a piece of snot that's twice as big as I am through that door. What I can't pull through that door is a bowling ball that's twice my size that's shackled to my ankle. The door is not big enough
So, an inherent limitation of Qdots is going to be their size and, thus, their limitations for permeability to a cell, because if we want to get them inside, the holes that we would have to blow in the surface of our cells are so big that in multicolor we could potentially end up perturbing the quantifiability of our cell surface antigens
So, that's why I like to say that the Brilliant Violet fluors as a family are almost like synthetic organic dyes and proteins got together, had a baby, and the baby just happened to have the best of both of them. It has the stability of an organic fluor. I can put it in methanol, I can put it--it's raised in acetone, so there's very little that I can do in terms of solvents to hurt this molecule. I can microwave it, if I want to
The only thing that I can't do is put it next to the sun, because like any other fluorescent molecule, it's going to still be light sensitive, right? It's about 70, 72 kilodaltons. It's in a big, long string. It's kind of shaped like a horseshoe, and that's because of the kinks of the double bonds that link each of the monomers together to form the polymeric structure
And, you can kind of imagine that that's exactly what it is. It's as if I took 70 kilodaltons worth of Pacific Blues, caused them to have enough of a linker that they didn't fret together. They're far enough away that they're not interfering with each other, but they are transitioning energy along a very long chain of double bonds, right? It makes it extremely efficient
And, so thereby, because I have so many of them on a single molecule and we conjugate more than one polymer per antibody versus PE is always a one-to-one with the antibody, it means I'm benefiting from a lot of monomers in terms of the fluorescence of that individual event
So, that kind of takes us to brightness. There's a lot of--you know, when we think of brightness, we can think of it on many different levels: the brightness of a single element, the brightness of our antibody. But, really what we--what we're mot concerned with is in situation, what's our signal to noise? Because brightness is only one element of that equation
So, there is a condition called collisional quenching that easily--when I used to work at Molecular Probes dealing with a lot of different chemistries, it was probably one out of five questions in technical assistance all the time. Our mentality is that more is better. More is never better in biology. We all know that. We are always maintaining homeostasis. We have this perfect place in the middle where everything is perfect
And, that's the same with the number of fluorophores that are conjugated to the antibody to result in the brightest antibody that we can have for quantitative detection, right
So, for example, if I had an antibody, a 145-kilodalton antibody, normally the molar ratio would be somewhere between about six Alexa Fluor 488s per antibody, about four fluoresceins per antibody, about three Alexa 647s per antibody. PE always a one-to-one because of the size, right? There's always these different ratios based on the fluor that's going to be ideal, where they'll be able to resist collision quenching but give us the brightest signal
So, say that I thought, "More is better. I'm going to add a molar ratio of 10-to-1 to this antibody." Well, great for me. All I did was waste my money, because what will happen--conjugation is a random event. They're conjugating to any available lysine residue, which is the most singly abundant amino acid in the construction of an antibody, so it's a good target to go after to make sure we're getting enough on
So, in this case, because it is random, there's always the likelihood that two fluorophores will accidentally conjugate too close to one another. Conjugation is irreversible. That's why we just don't randomly lose fluors from our antibodies. It's irreversible
So, because they accidentally conjugated too close to one another, when they jump up to their excited state, you can imagine that their pie cloud expands. Now you have an electron that's existing at any point in that orbit at any point in time, right? And, that's an expansion. If the fluorophores are conjugated too close to one another, their excited states interfere with one another. Literally, an electron from the excited state of one fluorophore randomly ejects itself and lands in the excited state of the other fluorophore that's its neighbor that's too close. It's a proximity event, because it's a statistical likelihood based on distance
So, when that electron hits the other one, both of them drop down to ground state because their excited states have been interfered with. That doesn't mean they're photobleached. They're not broken. They'll keep getting excited, they'll keep interfering, and they'll keep dropping down. You see this a lot with DAPI. If you put DAPI in your mounting media in microscopy and when you initially image it, it looks fine. Three months later, you pull out that slide again--every single molecule of DAPI has packed itself into the nucleus, so that when they're excited, they're smack next to each other, bound very tightly to that piece of DNA, and what you see are little black spots all over your nucleus. Those black spots are a quenching event that tells you that the density of the DAPI is too much in particular places
So, that's what limits the brightness of an antibody, but this also relates to tandems. We rely heavily on tandems. They are not a bad word. We need them. They constitute half of the fluorophores that we're going to use in any particular assay. But, there are definitely strengths and weaknesses to them as well. Because that conjugation event of the acceptor to the donor is a random event, there will always be the likelihood that I conjugate an acceptor that is not close enough to a donor to transition energy. Then there's always the likelihood that I'll have an acceptor photobleach, leaving the donor just to emit where it's supposed to emit
So, say that this was PE Cy7. There is always going to be the potential that the Cy7 acceptor has cross-beam excitation with my red laser, causing me spillover into its channel off the red laser, and there's always this possibility that over time, because Cy7 is particular sensitive to light, that I'll see them photobleaching, causing an increase in intensity or background into my PE channel, a change in the count value
And, that's why I really like it when people use tandems on things that they know that they'll replenish relatively quickly, so that they don't see that degradation effect over time because they're shutting down the amount of time they're using the reagent, basically. The faster you can go through a tandem, the better
It's why APC Cy7, for example, is great on things like Cy4 or CD4 or CD45 or CD19, things that you use really abundantly and can totally handle the dim dye
Okay, size is definitely an issue. You know, we kind of already talked about how big Qdots are. This is actually the smallest one. The ones that are about 655 or 605 are actually as big as this antibody. And, as they start to become 705 and 800, they become pod shaped, and that actually is a direct reflection of the emission spectrum of the Qdot as well. They start becoming extremely cumbersome
Another good one is GFP being 30 kilodaltons or 32 kilodaltons, in that range, where the chromophore itself is probably not even 400 daltons. It's tiny. All right. And, the majority of the bulk of this molecule is the protein that's simply allowing that fluorophore to exist, to work
Okay. So, there's--you know, the direct comparison for utility of the Brilliant Violet dyes are Qdots or eFluor nanocrystals from eBio, and they're like comparing apples to oranges in a lot of ways. Although we are using them off the Violet laser, they don't have the same strengths and weaknesses
For example, Qdots or any nanocrystal technology that's based on cadmium selenide or tellurium--or cadmium telluride is going to have excitation spectra that are extremely promiscuous between all the different lasers. You know, ideally, because they're semiconductors, right, they're going to be extremely happy to be receiving extremely high intensity, high energy wavelengths. That's what semiconductors do. They're excellent solar panel material for that reason, right
However, in systems where we have four lasers or three lasers on our Lsr2s or [unintelligible], that means that any particular Qdot is going to spill over into that identical filter set off of every single laser indiscriminately, okay? But, again, not to poo-poo them so much, because they were all we had for a really long time to get us to where we are now, okay
But, the Brilliant Violet fluors are significantly different. They're organic, which means they have discreet excitation and emission, so there is no generality that I can make about what that spillover value is going to be because everyone's configuration is very different. For example, our BV605 has extremely limitation excitation at the 488 or off the blue laser. It's fantastic if all you have is a three-laser system of violet, blue and red. It's great, because it isn't going to interfere with PE Texas Red at all. But, things like BV650, it is somewhat excited by the red, not nearly as bad as Qdot 655, but it's still going to have some cross-beam compensation that you'll have to deal with
The question isn't whether or not it has compensation. It's about whether or not that spillover is causing you to lose sensitivity. If you're maintaining population resolution, it doesn't matter what that value is, period, all right? It's all just mathematical balancing by the software
So, these are the current scope of every Brilliant Violet fluor that we have. They are now Brilliant Violet 421, Brilliant Violet 510, Brilliant Violet 570, Brilliant Violet 605, Brilliant Violet 650, Brilliant Violet 711, and Brilliant Violet 785--seven. We had to add a new PMT to our violet just to handle our own dyes
So, it's--and they can all be used simultaneously for phenotyping. There is no exclusion that needs to be applied, but there's a lot of logic that has to happen for that panel to be constructed cleanly
So, the only--of the seven polymers that we currently have, the seven products that we currently have, only two of them are polymers alone. Brilliant Violet 421 and 510 are both polymers alone. They are not tandems. But, again, any time you have a short wavelength excitation and a long wavelength emission, you assume you're working with a tandem. So, all of the rest, from 570 up, are tandems with acceptor fluorophores
So, I'm going to kind of break it down. We haven't completed a 100 percent perfect staining index at this point. We will very soon that we can send you. However, this is our pretty much best educated guess on how these guys are in terms of brightness
So, the brightest Brilliant Violet fluorophores that we have are Brilliant Violet 421 and Brilliant Violet 605. The brightest fluors that you can use at all in an assay are BV421 PE, PE Cy5, APC, and Brilliant Violet 605. Those are the five brightest fluorophores that exist right now
But, that does not mean that they have equal utility. PE Cy5 is a pain in the butt to include in an assay. It's got so much cross-beam comp that I'm often completely obscuring my sensitivity in APC. So, just because something is bright doesn't make it the right answer, right? They all have their own personality
So, in this instance--you know, initially when we were first making material, we always make some of the basic phenotypic markers, like CD3, CD4, CD8, and we realized really quickly that to take such a bright fluorophore like BV421, where there's very little background in that channel outside of autofluorescence--applying it to something like CD3 really doesn't make a lot of sense. If that's the marker that fits best in that channel--Pacific Blue is fine. You don't need to have the brightest thing on every marker. Sometimes that actually hurts you
But, it is not--you know, I save this channel for what I need to be most sensitive. So, you know, a really good example is it's extremely difficult to resolve FOXP3, even once you dump CD127. To get that positive population to pop out is really, really, really hard, right? Anybody's ever done that, it's--it can be a little frustrating and very variable between donors and between situations. So, you know, having CD25 on something that is quite bright in order to get that population to pop out is really important
Normally, I would not personally use a quadrant gate here. I would use my FMOs to show me where my negatives are for both of these markers, and I would gate on the space in between. But, that's a really good example of getting that population here to pop out and give me a very clearly positive signal--very positive on population
Brilliant Violet 605--I am in love with this molecule for cytokines. I think that I would like to see us conjugate it to every random cytokine, cytokine receptor, chemokine receptor, anything like that. It's going to be really fantastic to diversify a panel for the real special, interesting things that you want to look at
The next brightest--so this would be kind of the 4 out of 5 level brightness--fluors would be the Brilliant Violet 650, also really nice and bright. I'd like to see it on lots of cytokines as well. Brilliant Violet 711--in this case, we had it on CD8s. It really, really pops up there. And Brilliant Violet 510, which was quite a surprise. Again, this guy is just the polymer. He's not in a tandem. He's got no excitation off the blue laser because he's not a tandem, which is really helpful
So, if you're not using LIVE/DEAD Aqua and you want to use this channel to phenotype, we have figured out that this guy is at least 10 times brighter, minimum. And, we haven't done an extensive test of all the conjugates that we have, but we know that we're 10 times brighter or more. And, we'll have more testing on that soon, as well
All right. So, the third brightest--so this is kind of a 3 out of 5--and FITC would fall into this range. FITC actually is not a particularly bright fluor, but on a four-laser system, the FITC channel has extremely little spillover into it, so it's a really great channel to maintain resolution even though it itself is not particularly bright
In this instance, the Brilliant Violet 570 and the Brilliant Violet 785--they both fall into this category. The 785 we do--we determine brightness based on staining index, and the background--or the standard distribution of the negative on the BV785 is a little bit broader, which kind of takes into account making it a 3 out of 5, but great for phenotyping. I would like to see 785 on lots of maturity markers--HLA, DR, CD45RA, KI67, things that you would love to be able to add into a panel
So, just to start touching on compensation a little, there are a lot of things that we learned that most people have never seen in flow cytometry working with this family in testing. First of all, compensation is voltage. Those two things are hand in hand and absolutely inextricable
For example, I honestly believe that you should never, ever, ever, ever, ever look at compensation percents without taking into account the voltages. They should always be presented together, because one is meaningless without the other. What ends up happening--and we'll go over this later--is you can have an over-amplification of the spillover when the differential of the voltages between those two PMTs is too high. Any time you have 200 percent spillover of a spectra that looks like it's only bleeding over 10 percent, logically you know that that's not true. That's an artifact of something you created by having an imbalance in your voltages
So, just for as a really quick example of, on average, if things were well balanced, what you should expect, this would be about reasonable. So, this is the channel, and it's the removal of V500 from all of these channels--for example, of BV570 from all of these channels. BV57 out of PE, people are concerned with the acceptor being Cy3, being directly excited, again, by the 561 laser and emitting into PE. It's really not that bad. As long as the voltages, again, are balanced well, this should not be really hurting resolution all that badly
We specifically chose acceptors for these molecules in order that they weren't excited 100 percent by the other lasers, so that we weren't making it more difficult for ourselves, right
What I thought was most funny is everyone was concerned with BV570 bleeding into PE, when really what they should be concerned about is PE bleeding into BV570. If you think about how fluorophores work, all of those side chains--they're excited by short wavelength excitation. We just barely ever see it because we're not as concerned. We're so concerned with all the cross-beam comp that we don't really think about the fact that PE is also affecting the background contribution to the BV570 channel
But, again, this is because of this, you know, slight differential. You're seeing a little bit of an over-amplification here. But, we'll go over that a little bit more
And, again, it's that value--meaningless. What we're concerned with is background and resolution. We've come to equate that with compensation, and that's not a fair equation--equating
So, lots of things can cause background. The one I had to learn the most about was metabolic activation of the cells, and it makes absolute sense. Do you know how many of your cofactors are autofluorescent in the blue range? It's logical, right? They're the same sort of antioxidants that are protecting me there, you know, so they're going to be quite abundant. As the cell becomes metabolically more active--like neurons, for example, are highly autofluorescent because they're so extremely metabolically active. Obviously, that's going to make a really big difference in what your background is
And, in this instance, if I am looking at cytokine expression, I need to have a stimulated--I have to have stimulated and unstimulated controls to qualify correctly for what the actual background or negative population is
There could be all sorts of exogenous treatments that we talked about before, things like chemotherapeutics, but, you know, there are lots of other things that people are adding or samples that they're getting from unknown sources
There could be lots of things that contribute to your background. The most obvious is fixation. I know a lot of people that are in the habit of--you know, it takes a long time sometimes to stain up a 15-color panel, especially when you're doing stimulation and you're doing cytokine detection or FOXP3. It can take most of your day sometimes to do that staining. A lot of people will put their cells into PFA and just leave it
When you stick your hand into PFA, I can guarantee you it's fixed entirely at the surface within five minutes. When something is dead, it's dead. It's not coming back to life again. You do not have to continue the process of cross-linking. That is what the PFA is doing for fixation. It's cross-linking primary means. So, if I have two small, little hydrocarbons that both have a little baby primary mean on them and I cross-link them, shazam, all of a sudden I created a blue molecule, because that's biology. That's the diversity of biology
So, don't create a situation that isn't necessary. You're only hurting yourself or shooting yourself in the foot by increasing the background because you fixed way too long
Non-specific binding--obviously, a big issue, especially as I learned with Cy5 on some monocytes. But, there are lots of different reasons for non-specific binding. The best is using things like anti-mouse antibodies on mouse tissue or things that, you know, obviously just species-wise don't make a lot of sense
But, also, any time you use streptavidin, biotin is a cofactor. You have a ton of biotin in your mitochondria. Actually, in microscopy, we use biotin--or we use streptavidin fluorescently labeled by itself to do mitochondrial localization. That's non-specific binding, and an isotype control is not going to help me with that, right? Applying the secondary, the streptavidin, with no primary that's biotinilated, that's going to tell me what my background is. You use an endogenous biotin blocking kit in that case prior to adding your antibody
What most people don't think about until they start progressing beyond six to eight colors--I would say probably more than eight--is the additive bleed through of all the fluors into all the channels. That's why I talk a lot about the assaulted channels. You have certain channels, like PEB--PE, APC and BV421, where at least the fluor you're trying to detect in that channel is quite bright. It can handle a little bit of background from all your PE and APC tandems
Alexa 700 cannot. It's an extremely dim fluora that has a whole lot of fluoraphores that are bleeding over into it quite a bit, and when that additive effect becomes too much, I lose my ability to resolve anything that's positive for Alexa 700. Okay? And, that was obviously done on a fully stained sample
So, that's background, but obviously it's signaled to noise or signaled to background. Signal intensity is going to make a humongous difference, so we have to ensure not only that we have a good match of abundance at the target to brightness of the fluor, but also that we make sure it's got good access. I had to learn the hard way that CD11C can't have EDTAN solution [sp] because it blocks its binding. It's a calcium-depending binding. That's good to know, right
Or, another good example are people that are doing intracellular staining for a FOXP3, but they're trying to make it go real fast. Well, you've got to get through the nuclear membrane too. You've got to make sure that it's--you let it sit there for the full amount of time to have good permeatizations, that you've got good access so it can actually work, right? Or, use temperature as an accelerant to that process. And, then, obviously, just good balance in all things. You have to make sure that everything is well matched
All right. So, when we're actually setting up the panel, there are a lot of things to take into consideration, and there--this also determines how we gate, because gating is not an art, it's a science. Now, we're going to go over many different ways in which you can accurately gate through the rest of the presentation as we start looking at the data
So, you're going to have very obvious yes-or-no expression. Those are easy. They pop way out there. They're in a nice little tight population, super easy to gate, can use quadrant gates to get those guys. They are basic phenotypers. They do not deserve my best dyes, right? CD4, outside of a really bad PMA stimulation, would never deserve BV421 or PE. Right? That would be a misuse of a balance
You're obviously going to have a lot of different low expressers or maturity markers, things with a wide dynamic range. These need FMOs to tell you clearly what's positive and negative. Anything where there's a high and low population--it's quite difficult to discriminate those things, and it needs to be very internally consistent. Those often get my things like PE Cy7 or PE CF594 or Alexa 647 or something like that
And, then, finally, I have my rare or random events. Actually, that's usually where I save Alexa 647. Alexa 647 is quite bright of a fluorophore that I have a ton of flexibility with, because if I get some randomly new expresser or made, you know, antibody from a collaborator from Nebraska and it just shows up on my doorstep, without having to spend $1,000 to get it custom conjugated, I can do this conjugation myself and I know that I'm getting it conjugated to something that's bright enough that I can get a first pass assessment at the quality of what I received. It's a flexible place
But, generally I'm going to save all of my best fluors for the things that matter the most. This is usually why I'm doing the experiment, is this question. I'm going to save it for my best
There are obviously lots of instrument considerations. I'll show you what our current configuration is, but this can change wildly. We can make our cytometers into whatever we want. There are 10 laser cytometers out there
And, then, finally, I'm just going to go over once I have a good idea of how I'm setting up the panel, titrate your antibodies, how to control--how to apply your controls, how to gate for cytokines, and then compensation always, and how to balance a loss of sensitivity
So, primary tier is going to be yes-or-no answers. I like that we use this as a test, because we--I know people do actually sell NK1.1V450, for example. This is completely unacceptable, right? I am either an NK cell or I am not an NK cell. I'm not thinking about becoming an NK cell, right? This should be a clearly resolved population
It does not mean that I should put it on PE, because it is only a basic phenotypic marker. It shouldn't take one of my best fluors. What I would use it on often is, when I want to get to 15, 16 colors, I have to use PE Cy5, whether I want to or not. Most people are dumping their NK cells to clean up their T cell population. If I dump PE Cy5, this guy is not co-expressing with anything else on the rest of the cells that I want to analyze. Then I don't care what the comp is for this guy into anybody else because he's not co-expressed. I can dump him out and get rid of him. So, it's a good place to put that sort of fluor
My secondary tier is going to be anything where it's a dynamic range, a smear, a God knows what, right? I am absolutely thinking about becoming mature. I am thinking about becoming activated. You know, that sort of phenotype. I often use these on things like FITC. BV570 works fine for these particular two. PE Cy7 often is necessary, depending on how strong of an activation you have instigated
And, then, finally, my tertiary tier is going to be all my best fluors. If I can keep them for this, I will. And, it's usually when we're building a panel, we work backwards because I am initially limited by only what's commercially available in PE or only what I can resolve with BV421, and I have to go backwards to find out the constituents of my entire panel
So, first, before you do anything else when you're starting to consider that maybe, theoretically, this is the panel you want to look at, you have to titrate first. Titration is not a big deal in terms of staining index or resolvability when you're only working with a limited number of colors in an assay
But, I can tell you it matters tremendously when you get above 10, largely because you don't want just a mess. I always kind of say it's like your cell is a bun and you're just slathering butter all over that bun. Eventually, it's just going to get too sticky and too slippery for you to hold onto. You're confounding the ability for your antigens to find each other. Let alone, you're creating more and more and more background in each of your channels. That's going to further hurt your staining index
So, the very first thing you do is find your most conservative signal to noise. In this case, we use staining index because we want to take into account not just the MFI or the medium fluorescence intensity of the negative, but we want to also take into account its distribution, which equally affects my ability to resolve that population. So, this is a better equation to use. It's more accurate. But, again, I do this first
So, the rest of the data that I'm going to show you from here on, kind of elucidating gates, originated from a paper from Holden Maecker and a few others, and their--you know, there are a lot of efforts in the world right now to standardize multicolor for lots of reasons. We have lots of consortiums going on, lots of people who want to be able to compare patient samples between groups that are in Mexico versus the UK versus China, right? We want to be able to make valid comparisons between those results
We can't do that at 15 color right now, but lots of efforts like this are trying to do it at 8 color. So, when I look at this, though, I thought, well, this is a fun opportunity to look at an example to consolidate a panel. If I'm only doing clinical research, I can absolutely consolidate these five panels into probably two. So, I picked on B-cell panel, the dendritic monocyte and NK cell panel and the T-cell panel
I did not include a LIVE/DEAD. When you're working with PDMCs or you're working with whole blood, often the scatter helps you get rid of a lot of your LIVE/DEAD. And, also, you're phenotpying so extensively at 15 colors, usually at some point they're lost from what you're gating on. So, I did not personally include a LIVE/DEAD. Always a good idea for many reasons, but I dropped it here
And, I also wasn't able to include IGD or CD4 in my consolidation, so it wasn't entirely a consolidation of all three
But, this is what my panel looked like. This was my seventh panel, and I'm an expert at developing panels. It was a little bit of a frustrating learning process, but it was a learning process. We learned quite a bit about this
So, I'm going to show you some of the mistakes that were made, where we lost resolution and couldn't keep that particular configuration, but generally this is what I ended up with
So, online we have a brand new tool that really helps everyone do this themselves. It's called the Online Panel Selection tool, and what happens is that you choose the species, you choose the lasers you have, you choose the fluors that you have the capability to detect, and then you go through and list out all the markers that you want to have, and it will populate what's commercial available from BioLegend for those things. However, we don't carry PE Texas Red. There will be certain things that you don't see on this list. BV711 and 785 haven't been released yet, so I couldn't show them in this particular scenario because they're not commercially available yet
So, there are always some things that you don't see, something someone gave you that you have to include in, but still it's a great way to organize your thoughts and know what's commercially available. And, there's a little button on each side that you can click to see what the QC plot looks like to give you an idea of expression level and suitability with that particular fluor. So, just putting in a plug for that because this has made my life so much easier
This is our particular cytometer configuration. As of last week, this is no longer accurate because we finally added a seventh PMT, which shifts all of these, so that BV785 is here and it shifts all of them down. So, we now have an 18-parameter capable instrument. We could definitely do 16 colors quite cleanly now
Filters can change, though. You know, absolutely your percent comp, your resolution, lots of things will be affected by the band pass width or the [unintelligible] filter that's in front of that particular place. So, we're not saying that this is the word of God; it's just what works out best for us in our situation
Okay. So, in order to get my 15 colors, my 17-parameter immunophenotyping--I really liked how they did it in the paper, and I decided to mimic a lot of their gating style. Because I know that a lot of my monocyte populations, especially my dendritic cells--I wanted to go after PDCs and MDCs--I knew that I needed to collect as many cells as I possibly could just for fortitude. So, I made my scatter gate actually quite liberal, because I have 15 phenotypic markers that are going to work down that scatter gate, right? So, I have plenty to help me with parsimony
So, most of the populations are working off of CD19 versus CD3. The CD3 BV510 rocked. I was extremely happy with this. The reason I used these two together is that they won't be co-expressed tremendously. I'm really looking at my double negatives, my CD19 positives and my CD3 positives. That's where all my phenotyping is occurring. So, you know, these guys are going to have a decent amount of spillover into each other, and what I wanted to avoid was a lot of the distribution error that happens between channels like this
So, we're working off of this quadrant in here. However, quadrants are not always the right gate. We have been doing a lot of work with Brilliant Violet 711--great, bright fluorophore. I think you're going to be quite happy with it
However, if I could make a recommendation, I personally would not choose to put this fluorophore on any sort of bivariate or in a situation where I really needed to have confidence in my double positives, because there is a phenomena that's based on kind of a mathematical artifact of compensation that most of us refer to as spreading error It has nothing to do with the marker--well, it has to do with the brightness, thereby the abundance of the marker, but it's really a spectral artifact, hence why it happens when compensation is applied
So, what you should notice here is with bi-exponential gating, I had to transform this particular population by negative 500, which is in FLOJO [sp]. This is not DEVA [sp]
So, in that--you can see that the distribution of these positives are actually quite broad. That has nothing to do with the reagent performance. The dye itself is extremely internally consistent. This is very much a mathematical artifact, because as you get brighter and brighter, this number and the standard deviation digitally of this particular value is much larger than when something is dimmer. Okay
You see this all the time. This happens with all fluors that spill over into one another. It happens with APC Cy7 and Alexa 700. It happens with Qdot 605 and PE Texas Red. But, usually those transformation--that transformation effect is much smaller. This is really just kind of the perfect storm of two reagents that meet all those criteria to actually create a significantly greater spreading error. So, that's one thing to keep in mind when trying to actually fill out a full multicolor panel
It does not hurt my ability to resolve four and eight, because I am not interested in four positive, eight positive. So, I have no trouble here to go to my CD4s and my CD8 and find my memory populations based on that distribution. And, so that's where we go from here
So, what I do want to show is one of my first mistakes. I almost always like to use CD4 on APC Cy7. However, when Cy7 photobleaches, it leaves a PC behind. A PC is very bright and very happy to be excited in the red, so when APC Cy7 starts to degrade or treated harshly or left sitting around too long on the bench in the light, you're more likely to see a loss of resolution in your APC channel, where often you have something that you put there because it was hard to detect. Right? So, that was an example of this
In this instance, when I was using the CD4 APC Cy7 against a CCR7 APC, which is a brand new clone from us, that is fantastic for macaque and human. We find that we're actually getting significantly better resolution with this clone than previous clones--highly recommend it. My little plug for that
When we're using this, it's not a fantastic resolution. It's not dim, right? It's not lowly abundant, but it's not fantastically abundant. When I use the APC Cy7 and I start to see a population of background in the APC channel, I lose resolution here. When I apply--and this is actually--all this gating is based on FMOs. I can prove to you that these gates are true, okay, which was why FMOs are so extremely important, for exactly this reason, for internal consistency
When I use CD4 Alexa 700 where the spillover--this is not a very bright fluor. It's either photobleached or not photobleached. It's not spreading like APC Cy7 does. This is significant better resolution because my contributing background is a lot lower. So, this was a much better scenario
You do not see the effect when you're not gating on a population that's CD4 positive. If I don't have--if I'm looking at CD8 positive cells and looking at the same distribution and they do not have these fluors on them because they are not co-expressed, it doesn't matter. The background isn't there. An event is one cell. If that cell is CD8 positive, it's not CD4 positive, so you're not populating that background
It's one of the best strategies for maintaining resolution, is to think about those things. If two things assault each other a lot, make sure that they're not co-expressed. You can do that with immunophenotyping. It's a lot harder when you're looking at a lot of cytokine expression from one sub-population. That becomes a lot more difficult
All right, so the reason why I really need everyone to start using FMOs more regularly is because gating is not an art. It is a science. You cannot randomly say, "Hey, I start to see a slight differentiation between these two populations." That isn’t valid. So, when I'm looking at an activation marker like HLEDR [sp] versus CD38, try randomly drawing a gate there. If I apply the FMO, I know exactly where my negatives are and I have no trouble drawing that gate consistently between application, especially as my PE Texas Red starts to change over time. If I keep applying that FMO, I will always be internally consistent
All right. So, B-cells and monocytes--these are just some additional derivations of this 15-color immunophenotyping, where I'm looking at 27 versus 3 to look at my [unintelligible] versus memory B-cells. I have my classical versus non-classical monocytes, which absolutely required an FMO
I've got so many oddly distinct populations in there that until I overlay it, do I see that I have clearly negative here. These are low. These are high. Right? That's how my FMO helps me, because if I had just looked at this as a heat map, it would have been really hard for me to make that call on where the negatives actually were. I can't stress that enough. I believe that all data should be shown with their FMOs overlayed for things that are difficult to resolve. That's my own feeling on the matter
All right. So, then we're going after our dendritic cells and our NK cells. This is pretty easy, 16 versus 56. Again, it's an oddly distributed population. You know, we have three different things going on with this particular bivariate. We have cytotoxic NK cells, cytokine producing NK cells, and hyper responsive NK cells all contained within this plot
We go to plasmacytoid dendritic cells and myeloid dendritic cells based on CD11C expression, and that's where we need an isotype. And, so because--I'm doing 15 colors; I've got limited options. I really needed PE Cy5 to be on my CD11C, because it's bright. Right? I needed to benefit from that
So, when it came to actually using this, I made sure that it wasn't just an FMO that I used. I included the isotype control for PE Cy5 into the FMO for the panel in order to show what the background binding could potentially be. I hope that makes sense
So, you know, I am getting really decent resolution, but I need to prove to myself that this isn't non-specific binding. So, when I add the isotype control to my FMO, there is not active CD11C PE in this panel, in the red here. The FMO here is in blue, and I can see maybe a little bit--maybe. I don't really know. But, I can clearly have evidence to say that I am not seeing PE Cy5 non-specifically binding to this particular sub-population. That's all that's important, is that you can prove it and that it's internally consistent and representative. It's only going to make your data better
But, there will be spillover and there will be sensitivity between all of these probes. So, for example, again, I make sure that I list all of my PMT voltages, because they are not all equal, so that as I start looking at the different percentages--for example, all the Brilliant Violet fluors will spill over into each other. Okay
However, when I look at things like BV510, this is the--the channel is on top. The fluor is here. When I look at 510 into 570, 510 is 480; 570 is 500. It's probably not exactly 50 percent, really. It's probably lower than that, but that 20-volt differential accidentally over-amplified some of the photons that came from BV510 into the 570 channel. It's just an artifact of the amplification that is photo multiplication
Okay. So, there are a few other things that we have here. Cross-beam compensation--as we were talking about before, the BV570 has an acceptor that's excited by the 561 laser, so when we're looking at the 570 spilling into PE, it's about 20 percent in this instance. You can kind of look at the differential of voltage. It's not very high. And, then, again, vice versa. The spillover of PE into 570 is about 23 percent, and that's because of that short wavelength excitation
There are going to be other combinations where you're going to see the spillover. You're going to see APC into BV650, BV650 into APC, and you're going to see 605 into Texas Red, PE Texas Red. This is next to nothing, and that's fantastic, even off of our instrument that has a 561 laser
So, we're extremely happy with those sorts of results. It means we're not causing a massive additive problem the same way that Qdots do because of their absolute promiscuity of excitation
All right. But, then what I should also be looking at are the channels that are in here of diminishing sensitivity, so things like BV570, BV421, and especially Alexa 700. It's extremely difficult to maintain resolution in this channel, to the point where often I end up dropping it because I can't get what I want and still be able to detect something in that channel
Okay. So, there are a lot of things that affect compensation, but fundamentally, it is spectral spillover. Right? So, if you see two spectra that are overlapping and it only looks like 10 percent is spilling over but you're getting 100 percent comp, that's based on voltage. Okay? When--because everything else should be internally consistent
Filters do affect compensation, but they should be internally conserved. The instrument shouldn't be changing filters between application, right? The wavelength of the excitation laser or the age of the laser, how much output that laser is giving you--those things can affect the voltage that you would apply, thereby affecting the compensation number that will result
So, literally what it is, it's just so important for people to understand this in multicolor, how meaningless that number is. When I have something like FITC off of the same laser, FITC does spill over into PE. Say they're both excited by the blue laser. When I look at that tiny little spectral spillover here, it--I'm sorry, here into FL2, into this PE channel, it really doesn't look like it's very much, right? It's about 15 percent. If these voltages were identical, they were both 500 and they performed identically, the PMT hardware itself performed identically, then it should actually accurately represent only that spillover. But, that's never the case because we're always balancing these voltages. We're changing them
So, what ends up happening is the instrument is not smart. It only sees photons. It says, "I am detecting 100 percent number of photons into FL1, my FITC channel, from FITC. I am ratioing that to the number of photons that my FL2 is telling me is spilling over into the PE channel." Right? That's the percent comp theoretically, but what ends up happening is that when this voltage is amplified, it's no longer this. It becomes this. It becomes higher, because I have photo-multiplied the photons that are undetected into my neighboring channel and they are now out of proportion with the actual percent because the number of photons into my primary channel has not changed. It's an artifact. I have over-amplified my spillover
Have I changed my compensation? No. I've changed my mathematics. That's about it. Right? But, I have not changed anything about the resolution of those populations
So, when people tell you, "I use CD4 on all of my fluorophores and put them on comp beads and run them through the system or put them on cells and I run them through and those are my comp controls," for something like FITC, that would be fine. FITC is either on or off, but the spectra does not change. Right? So, yes, I could have CD4 Alexa 647, CD4 FITC, CD4 BV421, and I would never see a change in compensation because of something to do with the fluorophore
Where I would see it and where I cannot follow that strategy is with tandems, because this is not simply the spillover of the emission into another channel. It's the ratio of the donor to the acceptor, and this ratio is always changing, whether I bought that antibody today, yesterday, a month ago, a year ago, God knows when I bought it because nobody bothered to mark the date on it and it's been sitting in the fridge since I showed up at the lab. Those are all going to have entirely different ratios of acceptor to donor that absolutely will change that percent comp. I cannot just use CD4 for PE Texas Red or PE Cy7 or PE Cy5. It won't be accurate. I'll see an under- or over-compensated condition because of that
Really, in all reality, even if you're using comp beads, you should be treating them identically to your cells, because if you are staining for hours at a time, all of those fluors are being exposed to light for that entire time or they're being exposed to solvents, they're being exposed to fixatives that your beads should also be exposed to, because all we care is that our comp controls are accurate. We don't care that they're perfect; we care that they're accurate to what's on the cells. Right
Okay. So, the reason that this is absolutely meaningless--there was a good example while we were beta testing for these guys, where I was sitting at Upenn in Mike Betts' lab with a graduate student of his named Morgan Ruder [sp]. We were running through the full sample, which is here, with these voltages applied to these channels
We finished with the entire run, and I took the data off and quickly put it into FLOJO just to assess the comp matrix, which is what I was most concerned with at the time, and I said, "You know, this doesn't look right to me." And, she went back and realized that she had accidentally inverted the voltages between these two channels based on what she had determined to be optimal prior to running the assay
So, we had a tiny little bit of our full sample left, which was why there are significantly less cells in these plots. And, we said, "Well, I wonder what happens when we just switch these two voltages." First of all, BV650 does not bleed into BV605 75 percent. And, also, BV605 does not bleed into BV650 9 percent. There's someplace in the middle where this is actually accurate
But, by changing the voltages, by just flipping the voltages for each of these, running the same samples, all we did was flip the percent comps. These are still resolvable. I have not changed resolution. The only time I would change resolution is if my voltage mistake took me out of the linear range of the PMT. That's it. That's why these numbers are meaningless and why we should never refer to them as advantageous or disadvantageous for the fluor. They're only one thing that affects sensitivity
Okay. So, that's the end of this talk. There are a lot of people that have contributed to this entire project, lots of people who have beta tested for us over the last year to help us get to where we are now, applying all the different reagents to their many different biologies. And, there's a very tireless team here at BioLegend that has worked night and day to make these dyes possible. So, we definitely want to extend a thank you to all of them
So, thank you very much for your attention