SENSE.nano 2020 Symposium Session 1: Sensing at the level of cell and subcell

Our first session is on the cell, or at the scale of the cell or subcell. A thought-provoking question that we'll, we'll answer after the panelists, after the speakers. When was the first microscope used functionally? It's a very broad and big question, but when was the first microscope? And then when was the first image of DNA? So have that in the back of your mind, and, and after our speakers, before we get going to the panel, we'll share the answers. Our panelists, our speakers for the cell are Dr. Ashwin Gopinath, Dr. Peter So, and Dr. Timothy Swager. We'll hear from Ashwin on democratizing single molecule nanoarrays. We'll hear from Dr. Peter Soh on interferometric imaging for studying sickle cell diseases. And we'll hear from Timothy Swager on using dynamic lens systems for biosensing. Again, we'll hear from these speakers one talk after the other.

Q—

questions will be asked after all 3 speakers have presented. With that, I'll turn it over to Dr. Gopinath.

Thanks, Brian. And thanks for inviting me to share some of the work that we've been doing. I'm fairly new to MIT. I just started my lab a couple of years back. So all the work is just stuff that has happened in the last couple of years. So, in between all the COVID craziness, that's unique. It's like a testament to my postdoc who's able to do all of this. So, that's it. So, I'm going to be talking about democratization of single molecule nanorays. And so, it's kind of a little vague to understand what I mean by that. So, I just want to kind of take a toy example of what it is like to do single molecule optical experiments, and then jump off from there to say what it is that we are doing. Typically, if you want to interrogate single molecules using any kind of optical microscopy, you take a surface, you put a whole bunch of molecules on the surface, and then you— these molecules will have either fluorophores on them, you're doing FRET, there are different kinds of analysis that you can do, and you typically end up getting a cartoon picture. This is the data that you end up getting. Now, that's not necessarily the most optimum use of the surface. you could increase the concentrations, and when you increase the concentrations, you would have an overlap of a whole bunch of these molecules very close to each other. And because of diffraction-limited problems, you, you'll have to use advanced super-resolution techniques to kind of eliminate a whole bunch of data points, or to use, uh, to interrogate and understand what is happening. On the flip side, if you actually reduce the concentration, the amount of data that you can end up getting on that surface. So within this setting, the question is, what would be the ideal organization of single molecules to maximize the amount of data collected? And the answer is kind of simple. It's just, you know, you opt— you put the molecules such that the distance between them is just a little bit over lambda by 2. There is the dielectric constant in there, but let's ignore that for the time being. So if you have a hexagonal close packing of single molecules guaranteed to be at lambda by 2 periodicity, that's about as much of data you can actually get from the surface without using sophisticated single molecule resolution or like super resolution techniques. So that's the question that we are trying to answer. But what is a technique that allows us to do this? Now, and the tool of choice within my lab is DNA origami. So DNA origami is this technique that was invented in the mid-2000s, about 15 years back, when Paul Rudman at California Tech took a long single strand of DNA and folded it into arbitrary shapes, demonstrated that you can actually fold it into arbitrary shapes using short DNA strands. And what makes origami uniquely special is because this technique is extremely generalizable. That's just by changing these staple strands, or the strands that folds this long, long single strand into a particular shape, you can get any shape of your choice. You can make 2-dimensional shapes, you can make 3-dimensional shapes, and curved shapes, solid state, solid structures, anything that you can think of. And just bear in mind, this was all made, designed in a computer, made, and the, it's all, the constituent elements are all the same. They are just DNA. And these are, these are viewing DNA as building like a polymer, a programmable polymer, rather than thinking about it as a biological entity. So this extreme amount of modularity that is— that DNA origami brings to the table.

Ah.

But beyond this, what makes DNA origami kind of special is— so let's think about an abstract map of a triangle, for instance. The constituent element, every single constituent element within this structure is unique. So it's a unique set of DNA strands that form this particular structure. You can label a particular point within the structure using a unique DNA. What that means is, if you may, if you design the structure and make it, and the next day you decide, I want to put 3 different molecules on the inner vertex of this triangle, bear in mind the inner vertex is 30 nanometers and the external edges are about 120 nanometers. All you need to do is change 3 molecules in the design and you can attach, now, now, now it's ready to attach 3 other molecules. that has DNA tag on them, and you are left with another triangle now with 3 molecules on them. And for the last 60 years, people have been trying to attach DNA onto anything. So you have nanoparticles, quantum dots, carbon nanotubes, you name a nanomaterial, you can attach a DNA— there is a chemistry available to attach DNA onto. And what is even more interesting about DNA origami is every single point, every single modification point is 5 nanometers from every other point. So you can think of this origami as a breadboard with 5 nanometer resolution. Now when I say 5 nanometer resolution is different from the minimum resolution in a CMOS or like FAB setting, it's basically saying this is the minimum length size on which you can organize other molecules or other entities that you can tag with DNA. And for the last 15 years, people have been trying to attach carbon nanotubes, gold nanoparticles, proteins, polymers, quantum dots, other proteins. This is a small subset of origami-enabled devices. These are all devices wherein you have taken the core DNA origami, used it as a breadboard, attached other molecules onto it to build a sophisticated structure out of it. But the problem with all of this is that these are all in solutions. You can't create connections into it, and you can't— you have not created a you know, a link to the solid-state world. And what does that give you? That gives you, if you have the ability to organize DNA or DNA origami at specific locations, since you can uniquely modify the DNA origami with individual DNA strands, you can now attach individual proteins or individual carbon nanotubes or any inorganic structures or any other molecule that can be uniquely tagged with DNA onto a solid, solid, solid surface. And the way in which the, the technique that I developed to do this involves a chemically patterned surface. You take a typical chemically planar surface, you do a self-assembled monolayer to create a passivating layer, and then you create patches in the shape of the origami. In this case, it's about 100 nanometers in size. very easily accessible using standard fab. And what— and because of the nature of the interaction between the origami and the DNA and the surface, you can have unique assembly capabilities. For instance, in this particular case, you have a binding site that's in red, and you have origami in blue with blue strands, and you have these magnesium bridges that actually link them together. Because it's a cooperative interaction, it acts as an energy minimization process. So, if you have simulations of the binding site and the origami, it wiggles around and slots itself into the particular binding site. And this is not just a theory, we can have experiments to kind of demonstrate this. And that allows us a lot of room to tweak different parameters. You can tweak depending on the concentration, the size, and things of that nature. You can design the entire pathway, and what you're left with are surfaces like this. This is a silicon dioxide surface. surface in which there is an origami every 400 nanometers, and on the origami you can modify just a single entity or a defined number of entities.

And we get—

we can do this on multiple substrates. We have alignment error less than 8 degrees. Not only can we put molecules on the surface, you can also carry reactive groups within the origami to do chemistries onto the surface. And so you can design on a surface. It's possible for you to give me a particular point And I can go in and put a particular molecule there and just have a single reaction that's doable. And this is all done on standard microfabrication tools. So the lithography is all standard. We just make 100-nanometer features and everything else is done by self-assembly. And this is also scalable. And this is— so this is the work that we did a few years back wherein we took 65,000 different photonic crystal cavities, and the intensity of light that comes out of any given photonic crystal cavity is dependent on the location and number of molecules that you put in there. So it's not an art piece, it is to basically say that if you have 65,000 sites on a surface, I can put a defined number of molecules exactly where you want. So if you zoom into any one of these pictures, each point is a pixel, which is a photonic crystal cavity. So a suspended silicon nitride membrane, and inside which you have a certain mode distribution for the electric field, and depending on that, I can decide where I need to put a particular molecule for a certain intensity of light to come out of that. And by that, I can pick and choose, given an XY coordinate, I can put a molecule and get it to interact uniquely with the photonic crystal cavity, resulting in a certain intensity of light that comes out. Not only can we put molecules exactly where we want, we can actually even control the orientation. What that means now, and that, that is all done, it's the same underlying physics wherein the assembly is the same. You can actually put You just change the shape of the origami. And when you change the shape of the origami, you can actually orient it in a particular direction that you want. And we have better alignment, better efficiency, and things of that nature. And it allows us to do interesting things wherein you're playing with the orientation of molecules here. Where you have— in this, this is a polarimeter where you have a 12-pronged wheel wherein every prong has got an origami in a particular shape, direction, and a dipole in a particular direction. So when you change the intensity, the polarization of light, Only the ones where the, the orientation of the molecules align with the excitation will that— will it light up. It's done in oil, so it's just— you see junk kind of floating around. So it's 65— it's now it's 35,000— 3,500 molecules. Not only can we do it on a surface, we can actually do it on photonic crystals as well. You can change the orientation, keep the x, y position same, just change the orientation and control the interaction between the cavity and the molecule. But all of this sounds good, but the problem here is that everything is expensive. Doing this in a cleanroom is very expensive. So you can actually do this, you can make these kind of hexagonal grids, but it's expensive. So you can't use it for like high-throughput, like you can't do it as a mainstay within an experimental biophysics lab. So the question was, can we actually reduce the cost? And can we do it for under a dollar without using it, without ever getting into a cleanroom? and get the same, same heat. So what it turns out, you can actually do this. What you have to do is you take a surface, you just put a poly— a polystyrene layer on top of it, and it self-assembles when you get the right— under the right conditions, and you end up getting a periodic grid of microspheres. And depending upon the size of the sphere, you have certain periodicity, and you can use that as a shadow mask to create a self-assembled layer, self-assembled monolayer, and regions where the origami can uniquely impact. So, it can act as a unique form of lithography. And then what you're left with is a protocol to basically make these kind of single molecule arrays, and the cost of everything is less than a dollar. So, this is reproducible, extremely high throughput, and the yield— slightly, you take a hit on the yield because the quality of the Lithography is not as good, but you can get— it's good enough for many of these experiments. And so, so what you end up getting is traditionally single molecule optical assay looks sort of like this. You have each of these spots is a particular molecule which has a fluorescent fluorophore on them, and depending upon its— you need to do a lot of analysis to basically figure out whether there is one or two and make sure that two molecules are not interacting with each other and things of that nature. So that limits the amount of data that you can get. But now what you can do is you can make grids like this. So this is post-processed grid. So these are each single pixel, each single point is a single molecule on which a certain assay is being done. So you have a 50x improvement in the number of usable data points, and this all costs less than a dollar to actually put a particular substrate. And these are all the same substrates on which traditional biophysical experiments are done. So you can start doing like single molecule counting, wherein you're counting the number of, let's say, Zika virus in this particular case, because Zika virus was hot when we were doing these experiments. And we can improve the super resolution imaging on these settings. We can also do like traditional temporal dynamics. All of these, the data, the amount of data that you end up getting is way higher. And all that is because the way in which you can actually combine different forms of self-assembly to position single molecules at the specific location. So that said, I want to thank the folks involved. Rishabh is my postdoc, fantastic postdoc, and, and, and Rizal and Paul are my collaborators from Arizona and Caltech. And these are the folks who've been generous, I mean, during my time at Caltech as well as my time here at MIT. So with that said, I welcome questions.

Good afternoon, everybody. I am Peter So. So I would like to tell you a little bit some of our work on developing novel microscopy for a number— for the study of a number of diseases. Since the time is short, originally my title includes sickle cell disease and also cancer metastasis. Just to make the story a little bit simpler for the time that we have, I will tell you mostly about sickle cell disease. Thank you, Peter. For many of you already know that sickle cell disease is a fairly common disease in the world. About 200,000 children are born worldwide with it, and there are about 100,000 patients in the US. One of the, one of the complications of sickle cell disease is sickle cell patients, even in advanced countries like US, still have about 10 to 15 years shortened life shortened life expectancies, as well as they have so-called vaso-occlusion crisis, as is seen here. For normal people that do not have sickle cell disease, red blood cells flow through the capillary bed in a normal way, but during sickle cell crisis, some of the cells, due to deoxygenation, they sickle and they become more rigid and more sticky. And they effectively clog up the capillary bed of the patients, and because of that, patients feel significant pain. Today, there are very few drugs that are available to treat this disease. Hydroxyurea is one of the drugs that have been used for many decades. Right now, we have 2 other drugs, but none of them completely remove vascular occlusion crisis. So today we have interest of developing method that would allow us to find biomarkers that would— that can be used on the blood of sickle cell patients such that we can develop more new drugs for this disease. So in order to do that, I want to tell you about an optical imaging method called interferometric microscopy that allows us to look for biophysical markers of sickle cell disease. Specifically, the, uh, the part of the methodology that we developed is a method to measure the thickening of the red blood cell membrane that typically have amplitude fluctuation scale on the nanometer range and also very fast fluctuation on the millisecond time scale. By measuring this type of fluctuation, we can extract mechanical parameters that go back into a red blood cell model that allow us to extract biological parameters of the membrane of the sickle red blood cell. So we, so we believe that some of those parameters would allow us to identify new biophysical markers. But before I tell you about how we applied interferometric imaging in some of our study of sickle cell disease, I want to tell you what is interferometric imaging. So the idea is very simple. If we have an object that have different refractive index than water, for example, the red blood cell, when you have light wave going through the cell, the wavefront might be distorted, and the distortion of the wavefront is proportional to the— if I know the refractive index, it proportional to the height or the thickness of the red blood cell. There are many ways to measure the wavefront distortion. You can do normal interferometry as shown here. But what the way that we have developed is a much more stable method. It is called diffraction phase microscopy. The idea is very simple. We have the wrapper cell sitting on the sample space, and then we have light coming in, illuminate the sample, generate a plane wave. And if we put a camera at what we say the grating plane, That would be just a normal white light wide-field microscope. Instead of doing that, we put in a grating at this plane so that we generate 2 diffraction orders, and one of them we do a spatial low-pass filter so we generate a reference plane wavefront that interferes with the signal from one of the orders that generates what we call interferogram, and the change in the fringe pattern of the interferogram allows us to deduce the thickness of the red blood cell with nanometer— actually sub-nanometer scale sensitivity. And so what we have done is apply this method to study blood from sickle patients. As it turns out, if you look at the blood from sickle patients, they have different cell types and they have different densities. You can first do a centrifugation and do a density fractionation. You'll find that typically we can separate the blood of the patient into 4 different fractions. So the red bar corresponding to sickle patients that have 4 different fractions, they are not on hydroxyurea, one of the drugs that I mentioned earlier. The green bar are the patients that are on hydroxyurea. You can actually— you can't really tell a difference in the distribution.

There's no statistical significance there.

However, if you look at a normal person that have no sickle cell disease, typically for normal people, they only have the lower density fraction of 1 and 2, whereas there is no higher density fractions. So one of the things that's very interesting, if we— so one of the reasons that we density fractionate the Cell first is because denser cells have more material, you have higher refractive index. As I mentioned earlier, in order to calculate the thickness of the cell, I need to know its refractive index. So using the density fractionation, I can identify a population of the cells that have roughly the same average refractive index, and then I can calculate for each fraction What are the fluctuations that I showed you a little bit earlier? And we found that the more lower density fraction of the cell fluctuates more, whereas the higher density cells fluctuate less. That means that actually, if we go back to the biomechanics model, we found that the lowest density cells have lower shear modulus, whereas The higher density cell has a higher shear modulus, whereas the bending modulus are constant. And the good thing about interferometric microscopy is we not only allow us to extract biomechanical parameter, we also at the same time measure the shape of the cell. So we have neurological parameters. For example, we can measure the eccentricity, how elliptical the cell is. surface-to-volume ratio, the volume of the cells, and you can see statistically, except for the bending modulus, all 4 fractions are very different. So if we take a look at patients that are on drug versus off drug, can we use this biophysical marker to look at the difference of them? So as you can see here, All of them are statistically significant except for some of the, uh, measurement in the eccentricity. So both the biophysical parameter like fluctuation and shear modulus and some of the morphological parameter like cell volume and surface-to-volume ratio, we can for each fraction statistically distinguish whether the patient, the group of patients, is on drug versus off drug. One of the questions that one might ask is how does this correlate with some of the mechanisms of how hydroxyurea works? What people postulate before is hydroxyurea mostly works by increasing the fetal hemoglobin concentration of the patient. As it turns out, for fetal hemoglobin, the hemoglobin molecule do not have the molecular defect of the sickle cells. It is only when we become an adult, the adult hemoglobin have a subunit that have the defects. By increasing the fetal hemoglobin concentration, One of the hypotheses is that it allowed, uh, it allowed the dilution of sickle hemoglobin, but therefore it has less chance of having vascular occlusion crisis. However, if we take a look at the low fetal to sickle hemoglobin ratio versus the high, which is the light brown versus the dark brown color, you will find that for all 4 factions, there is no statistical difference between the fluctuation or the moderates. On the other hand, another parameter that people have also looked at for a long time is the corpuscular volume, which is effectively how big the cell is. And in this case, we found that there is, in all 4 cases, concentrations, there are statistical differences. For cells that have higher MCV, they tend to have a difference between whether patient is on drug or off drug. So this is an interesting measurement. Seems like biophysical marker provide better correlation with whether patient is on drug versus off drug. Compared to some biochemical marker such as fetal, fetal to sickle hemoglobin ratio. So what are the limitations of the method that I showed you earlier? One of the limitations is I need to fractionate the cell into 4 fractions. One of the reasons is because I need to know what is the refractive index of the cell itself in order to calculate the height and get the dimension. So the thing that we would like to do is, is there a method to improve our technique such that we can, for individual single red blood cell on a single cell basis, measure not only the, um, the shape of the cell, the biomechanics of the cell, but also measure the refractive index of the cell, and from there We can relate it to the concentration of hemoglobins in the cell. So one of the methods that we have developed— I won't go into the detail of the math— is in addition to mass, we can measure the wavefront distortion, which is related to the refractive index change. I can also measure how much the light is transmitted through the cell, which is related to absorption. Of the light by the hemoglobin. So because the refractive index and extension coefficient both is a concentration of hemoglobin, so if I measure both the transmission of light, the amplitude, and the, and the phase image, I can simultaneously solve this set of equations and calculate on a single cell basis The concentration of the hemoglobin as well as the height of the cell. And so this is the result, the data that comes in, the phase map and the amplitude map that we can get both from the diffraction phase microscope at a single cell basis. From there, we can on a single cell basis calculate the shape, which is the height, and also the concentration of the hemoglobin on the cell. in the cell. So right now what we could do is able to, in an image of many cells, for each of the cells now identify both the concentration and the volume. And we can now also compare this type of measurement to some clinical measurements. So right now, so for typically when the sickle patient give blood so that the clinician can monitor their disease state, they typically use a blood analyzer that gives an average measurement of what is hemoglobin concentration on all the red cells in the sample, as well as what I mentioned, MCV. They give you the average size of the population, which is the, the mean value that has been measured with the current clinical measurement. But the good thing is with our new method, we not only have the mean, but we can measure the distribution, and that potentially can provide a lot more new information. So I would like to wrap up my talk, um, and I would like to say that such as quantitative phase microscope is a very promising method to look at individual patients' red blood cell population, and that allows us to extract biomechanical and morphological parameters. And some of these biophysical markers appear to have better correlation with whether patient is on drug or off drug compared to some biochemical markers. And so some of them might have potential use in screening for new drugs. I would like to thank everyone that did the work. Most of the work are done by a graduate, a former graduate student, Pouya Hosseini, and also a research scientist in the lab, Saeed Yaqoob, and have funding from NIH and other institutions. Thank you very much. Thank you.

Hello, I'm Tim Swager, Department of Chemistry, and today I'm going to tell you about our work on liquid lenses for biosensing. And what I'm showing on this first slide is something that looks like a bunch of flying saucers. It's actually showing a couple things. One is that we can orient these droplets. They are actually spheres, but only half of them are showing up because they're indexed. The other side are index-matched to the continuous phase, which is water. And the reason we're interested in these types of systems really is drawn upon inspiration from nature. So if you think about the fidelity which biology has to conduct molecular recognition, know one cell from another, really has to do with the fact that on the surface of a cell, there are all these functionality here, and they're dynamic. They move around. The lipid bilayer here that's just showing the kind of compartment, the inside versus outside of the cell, is liquid crystalline. It's very dynamic. It's technically a fluid from a physics standpoint. So we've always been interested in how can we create sensors that really emulate these dynamics, this ability to have things to cluster, to come together in unique ways to create the same fidelity of interaction that is in a cell, and do it in something that we can do sensor measurements with. And so what we work with is creating things that don't really look like cells normally, but they have similar dynamics and surface characteristics of cells. And we're working with fluid droplets in a water phase. So what I'm showing here on the right are 2— are polydisperse, they're different sizes of droplets. The inner side that is in red is dyed red. It's a hydrocarbon. And the outer one is a fluorocarbon. And we can make this with a variety of different liquids. But in general, these have surfactants on their surface that look a lot like these lipid bilayers. And they have the same fluidity and hence mechanics of a cell. And so the way we can make these in high precision is to use microfluidic devices. You can, you can buy these, you can make many droplets a day, you're making them kind of continuous, it's a continuous manufacturing method. And we can make them monodispersed, as I'm showing here, over a range of different sizes. Now, the way you control the structure of these droplets is you have to think about their surface tension. In particular, we're here, we're emphasizing the surface tension with regard to water. We have to sometimes worry about the internal surface tensions, and I'll mention in just a moment. But if you want the fluorocarbon to be on the outside, what you do is you take a surfactant that looks like it has a little piece of Teflon on the end, which is the fluoropolymer, and a polyethylene glycol, which is a water-soluble polymer often used in biology, and you put that into your water phase and it assembles on the surface and the fluorocarbon goes on the outside. And you can see we get If they're monodispersed, we can basically make these nice little crystalline monolayer arrays just by letting the droplets settle in a petri dish, or we can use a hydrocarbon surfactant and put the hydrocarbon on the outside. Now, we can take mixtures of these 2, and we can create a continuum of different droplets, and one of the things that's important is they're all round. They are round because the dominant interfacial tension is between the water and the outside, either the fluorocarbon shown here on the bottom or the hydrocarbon on the top. And it turns out the fluorocarbon-hydrocarbon mixture we're using in this case has such a low interfacial tension that it doesn't really mind having a bigger surface area. Here on the left, you see it has a very large surface area on the inside. It doesn't mind that. Okay, so we can keep everything perfectly round and just let the minimization of surface tension, which ends up being a perfect sphere drive these shapes. Now another thing I'm showing here, I'm showing that these droplets, the fluorocarbon is always on the downside here, and that's because fluorocarbons are denser than hydrocarbons. So one of the things we're going to use in this is we use gravity alignment of all these droplets. We can also do it with magnetic alignment, and that's in fact is how that first slide I showed on my title slide was made. But we're going to use this gravity alignment here of our droplets. And so, if you take these droplets and you put them on a surface and you project light through them, you take collimated light and you direct it through them, you get a series of ray diagram patterns as shown here. And this is work that was done in collaboration with my colleague Matthias Kohler, who's in mechanical engineering, an expert in optics, and his student Sarah Nagelberg, who's the lead author on this paper. But anyway, we have these perfectly round droplets. They all behave as a lens. And so you see the first 3 at the right here, they all have focusing conditions at different focal lengths. And in fact, as I showed you, you can see that it's very powerful because I show you those 3 droplets. These are actually photographs of them. If I were to show you those on flashcards, I don't think you could probably tell the difference if I were to show you them repeatedly. But they all have very different focal lengths. So very small differences can get very large focal lengths. Now, one I'm going to tell you about today that's actually relevant for just what I want to describe today is the one here, number 5, and this is what we call the perfect Janus. And that's composed of droplets that have exactly equal interfacial tension between the hydrocarbon phase and water and the fluorocarbon phase and water. That is, we balanced it and we had 1 to 1 by volume of both of those fluids, and so we get something that's just a perfect hemisphere that's half fluorocarbon and half hydrocarbon. And when collimated light goes into it, collimated light comes out, and it will transmit images through it. And so we've made use of all these different optical properties here. We can have, again, the perfect Janus, which will transmit an image as I've shown down here in the bottom, that little happy face looking out from underneath. But if we basically change its morphology, we can do that by having enzymatic reactions, for example, at the surface of these droplets, we can get to, in this case, that diversion of the light. So we have defocusing optics, and you can see that happy face is now not readable. We can also take droplets, and I'm going to tell you— I'm going to focus today on this topic where we take droplets, if we have them in this perfect Janus state, which again will transmit an image, but if we can somehow bind them together and cause them to tilt, this is called agglutination, we can get a highly scattering state. And up here at the top is just showing some of the refractive indices that are necessary to get these things to work. We have to match things properly with water, but it turns out to be pretty easy to do just with fluorocarbons and hydrocarbons. So how this can work now to create a sensor, so if we take droplets and they tend to perfectly align as shown here, here in this Janus, perfect Janus droplet, and we put them in, we look down from a microscope up above, we'll see that we won't see the actual structure of this because we're looking from down above. And remember, they're perfectly aligned because of gravity. And if we put them over the top of a QR code, we can image the QR code. We can image it with a smartphone. However, if we now throw in E. coli, Now, E. coli, we put a sugar, a mannose, on the surface of these droplets, and E. coli binds through what we call a polyvalent approach. That is, there are multiple connections between the E. coli and the sugar molecules. What you can do is you can bind the droplets together. So ordinarily, if I didn't have E. coli around, I could basically tap this this array of droplets, they would tilt on their side, but they would come right back again into their normal position. But we have E. coli around, they get stuck together, and they tend to scatter light. And so that can cause the, the QR code to be unreadable. So this is a very simple way to detect bacteria. It can also work to detect other polyvalent things such as antibodies or viruses that I'll talk about at the very end. Now we can We can detect these with a smartphone. You can actually even count the number of these agglutination events to get quantitative data. So you can either do threshold or quantitative data. Now, sugars are great, but they're not as specific. And so when we're going on to make other types of sensors, we link antibodies to our surfaces. We use the conventional kind of what was called a click reaction to attach these antibodies. And we confirm the antibodies are there by just using a fluorescent protein that binds generically to the stem of this antibody, and you can see here that we get fluorescence indicating that indeed we can bioconjugate, we can put things on the surface. So we can do this. We've used this to detect Listeria. We're using Listeria and anti-Listeria antibodies, and you can see now that without the Listeria around, they're all nice and aligned, but when we have Listeria around, we see these dark areas, and these little round circles here are something that we've done through an automated image processing program that basically will count the number of these events, and we can get a calibration curve and get out limits of detection. You can see we get down under 100 colony-forming units per milliliter, or viable cells per milliliter. Now, this all works great, but you look at these dark areas and say, how do you know that's not just a piece of debris? debris. And so what we've been using now is to go and put in different optical methods, particularly dyes, that can help us discriminate between when a droplet is really tilted on its side or agglutinated and when it is not. And so what we have here is I'm just showing you kind of schematic. I'll show you the chemicals in a minute for those of you that are interested. We put in 2 dyes, one that is kind of like what we call blocker dye, kind of a dark dye that doesn't absorb it. I mean, it doesn't emit light, but it just absorbs light. It absorbs light going through it, absorbs light that might end up in the lower layer and try and re-emit light. And so what happens is that just blocks everything. And if you look at these with a microscope, they don't show much color. You might see a little orange bleed through in this one droplet here at the bottom, but you don't see much of color. So if you look at it with a bifurcated fiber optic doing fluorescent imaging, you won't see much data or much signal. However, if we agglutinate them now by exposing these antibody-labeled droplets to Listeria, we're going to tilt them on their side. Now you see that red dye is exposed. And so we do this by basically creating dyes that have orthogonal solubility. In this case, we have one that— a fluorescent dye that's only soluble in fluorocarbon, another dye that's only soluble in hydrocarbon, and we choose them so that they're— the absorption of the blocker dye overlaps the emission of the emissive dye, and we decide what wavelengths to measure these at by looking at ratios of their extinction coefficients. And here on the lower right, you can see that indeed, when we put that blocker dye in, we can lower the emission coming out of these droplets. And so the way that these— this works now, just to show you kind of an animation so that we're all clear about this, is we have antibodies on the surface of our droplets. They're aligned with gravity, but when Listeria comes in and it wants to bind these, it agglutinates them, and you can see that they get turned on their side. Initially, you can't see the red dye when looking at it from the top or measuring the fluorescence from the top, but now when they're agglutinated, you see that red dye exposing itself. And so when we look at this with our fiber optic assembly, we see an increase in the fluorescence with the E. coli— with the, excuse me, the Listeria concentration in this case. And we can get calibration curves that now are getting us down to, again, this similar ballpark of 100 colony-forming units per milliliter, and very rapid tests for Listeria. So this is actually very useful for food processing where you might want to determine rapidly if you have Listeria in your food before you send it off to grocery stores. And this is actually a real problem. There aren't technologies that are efficient enough and fast enough to make this happen. So we can do this from the top as I've shown you before. Here, but we could also switch things around and do it from the bottom where we switch these dyes up and put, you know, the blocker dye on the bottom and the other dye on the top. And this has the advantage if you're going to go through turbid solutions. So remember, the dye— the droplets are going to sink to the bottom. So this way we won't have to go through a bunch of solution to get our fluorescence signals out.

And it—

this is just a plot showing you it works just as well as the other way. And so we can do this now in very complicated, highly scattering media. So just in summary, we have invented these liquid lens systems that have high fidelity to create ultrasensitive biosensors. So the detection limits we have here in terms of the rate and limits of detection are basically best in class. They have really high fidelity. They basically emulate cells, and so you're not putting things on noncompliant surfaces. They're also, through the agglutination method, is capable of detecting viruses and antibodies, both of which have the same sort of multipoint or multivalent types of associations. And we have now transferred this to basically make COVID-19 antibody tests with limits of detection of 0.2 micrograms per milliliter, which is considerably below what you would express if you basically have been exposed to this disease and developed antibodies. antibodies. So this has order of magnitude of sensitivity beyond what you need, and we think this could have clinical relevance in the future. So one of the special things about MIT is the people. Perhaps the most special thing is the people, and I've been fortunate to be able to have a really talented and large group of coworkers. This is the research group proper here shown on the left in front of the State House, and then also my most valued colleague and collaborator Matthias Kolla and his student Sarah are shown at the right. And so with that, I'd just like to thank you for your attention and look forward to the panel discussion.

Ashwin and Peter and Tim, thank you very much. So we're giving our speakers a chance to walk out of their virtual stage and join us on the, on the, on stage together. While we're doing that, let me just share with you the answers, or my answers, to the questions of when did the first microscope come into existence. The first microscope happened way back in 1590. Some folks that were developing lenses for optical lenses for sight combined them together to be able to have the very first microscope. But the first microscope was not used to see anything at the really small scale. Until maybe 1675, where the first bacteria were actually observed. And you can see the mechanical translating, the little knobs, the mechanical buttons that are on this schematic of the, of that very early microscope. Now, much, much, much later, we started using electrons instead of just light to image. That gave us the ability to see things at smaller scale. And so the first tunneling electron microscope, uh, came about in about 1931. And then about a decade later, later, the scanning electron microscope that gave us now submicron resolution, 50 nanometer resolution. And in a more modern version of that, you can see a, a hair mite holding on to a strand of hair on a person's body. So the first microscopes, a deep and rich, rich History. Now, the first image of DNA happened in 1952. This was called Photo 51. It's basically— it took 60 hours to generate the image. And used a technique of an electron beam passing through a DNA sample and scattering that beam to create an interpretable— an image that was interpretable to reveal the structure of the DNA. So with that, now I think our panelists are waiting, and I'll welcome them back. Ashwin and Peter and Tim, thank you very much for your, your very captivating talks. And so, participants, I know we have some questions that have been coming through. I'm trying to synthesize a couple of them. I will direct questions variously at the individual speakers and then maybe some for the group together to get some conversation started. But there's a first question that I'll ask of Ashwin. In this area of self-assembly, what are the opportunities to actually have multiple molecules being assembled together? Or are we only constrained to use a single molecule, or can we start to stack things together like Legos and combine them in interesting ways?

So, so the short answer is yes, you can do it. And there are a lot of different ways in which one could actually do that. But as you increase the number of molecules that you're going to try to put together, the error rates kind of go up in certain variation of error rates go up. So currently there are papers that are trying to take, let's say, 3 molecules, 3 of these origami structures, for instance, putting it together and making slightly larger structures. And the larger limit is something in the vicinity of like, I think, 64 or maybe even larger specific structures. But the error is kind of, you know, it needs to be worked out. So the short answer is yes, you can do it. But the longer answer is there are, there are still, you know, kinks to be worked out, technical kinks to be worked out. But there is no reason why we can't actually go at like solve it and just make it much larger structures that are uniquely addressable.

Very good, thank you. So actually, I'll take the next question to Peter. There's a question that came in on, just out of curiosity, the impact of transparency on cells. And actually, the more narrowed question was, are all cells transparent? How important is it to have transparency in being able to tease out all these various properties of the cells?

Yeah, so the answer is the cell is not fully transparent because of the fact that I have been talking about sickle cells, so I'm looking at red blood cells. So hemoglobin, actually, the property that I use, the refractive index, that's assuming— that's not looking at how fast the light moving through the cell, which is related to the density of the material, But I have been also looking at the amplitude of the information to allow me to extract hemoglobin concentration. In that case, I am assuming some absorption. So the extinction coefficient measures the absorption of the hemoglobin going through the cell. So, so in typical biological system, most single cells, except something that highly dense that have a pigment like hemoglobin, They are very transparent, but on the other hand, they always have a refractive index change.

Very good. Thank you, Peter. The question that was coming in on— for Tim that I'll first ask, just maybe a clarifying point. So with the Janus lenses and this approach of imaging, You're not changing the shape of the lens or the microfluidic lens. It's more the orientation of the lens. Is that correct?

Yeah, in the things I showed today, that is correct. But we have a number of other assays we developed that actually make use of that. I alluded to an enzymatic reaction that uses it, but we can also do other molecular recognition and we get confinement effects. And, you know, total internal reflection inside the different phases and other things like that. But we do both. But today I focused mainly on just orientation of them without changing their internal structure.

Very good. Thank you. So let me turn this maybe a broader question to, to the 3 of you. In terms of looking at the perspective, maybe in situ, of single molecule sensing in a flow setting, whether it's in the body or outside of the body, how do these techniques, various imaging or characterization techniques, translate to a scenario where the cell is not stationary and moving in some way? Is there opportunity there?

Maybe I'll take the first step. I mean, single cell, if you can do a labeling in the body, you can detect it in the blood vessel. And a number of techniques have been developed by the Massachusetts General Hospital group on doing some of this sort of work. But looking at single molecule is probably a little bit more challenging. Most single molecule work external to the body in many ways. And in that case, if you can label it, it's not a big issue. I mean, even doing it in flow is not a major issue as long as you have bright enough.

Ashwin or Tim, any concepts on how your sort of approaches and discussions today could be used in a flowing scenario?

Well, in vivo, I think it's going to be, as Peter was saying, laboratory stuff, yeah, but in vivo is really difficult to imagine how you might do that. But certainly in very controlled environments, I think you can do it, but in the body, that's a big dream.

Yeah, so, so, so, um, in the body, I mean, again, I, I mean, I, I second everything that Timothy said. Uh, uh, it's sort of, uh, you know, if you want to put origami in the body or even make a sensor with, with the origami kind of, uh, adding specificity, doing it in a controlled setting, totally doable. Uh, but putting it in the body, it's going to be challenging. There are ways that one could think of it. Not easy. We are far from it.

So yeah. So small steps as we figure out before we start to do it in real time and live.

Yeah, I might also add that I think that one of the recent calls for proposals from the Zuckerberg Foundation is actually not as ambitious as what you want to look at single molecule in the body, but they want to go as deep as they can beyond whatever we can do and just to follow single cell. If one can actually do that, I think that would already be a pretty big breakthrough if one can actually study single cells Very deep in the body.

So not necessarily flowing, but just being able to see inside the body and characterize—

They might flow and they might migrate, right? I think that's why it is interesting to look at individual cells. So if one can imagine looking at white blood cells that are in some cancer solid tumor deep inside, Would be interesting. Right now, most of the models kind of expose the tumor and put it in animal, put it in the surface. So you can get to some depth, but not in a really in situ setting. Very good.

Thank you. Another broad question here was, I think it may have been directed to an individual speaker, but I'll ask it a little more broadly. You know, the, the, how could these various approaches be used to characterize the, the proteins on the surface of cells? So the G proteins that sort of control how we're interacting with things outside or inside of the cell, how can we sort of characterize that cell surface phenomenology or chemistry with these, these various approaches?

Well, so maybe I can, I can do that. So we've done things not exactly like that, but we have aspirational dreams to look at those. But we certainly can— if you change surface tensions by having things either cluster or expand, you change the morphology of a cell. And basically one of the things we can do is we have the ability to kind of open up little apertures that let light in or out of are droplets. And so you can do things like that with these droplets through a change of morphology.

So one of the advantages of origami is the fact that it's made out of DNA, and one can think of basically combinatorially putting ligands on the origami at specific orientations at specific locations and doing kind of like a binding assay and then reading it out by sequencing to kind of figure out what are the organization of ligands and like clustering of surface receptors that would be harder to do in a traditional setting, at least especially in a really high-throughput fashion. So because of the fact that everything is built out of DNA and you can actually tag specific locations of the origami, one can imagine a way in which you can do these microscopy problems, like changing these kind of microscopy, traditional microscopy problems into a sequencing problem. So efforts in this direction is coming out of Peng Yin's lab over at Harvard, but that's just the beginning. There's a whole bunch of other things that one can think of there. And the question of like GPCR itself, you can do tagging of the origami with, let's say, lipid nanodiscs that carry GPCR or a particular kind of receptors, and then you can actually, you know, figure out what kind of binding is occurring and things of that nature. There are inklings of this work kind of starting up, but there is— it's open. There's wide— there's a large slew of questions that one can ask there using this as a tool.

Very good. Thank you. Another broad question. It was originally directed at, I think, Tim Swager, but in terms of the sensitivity of the microlens in particular, But broadly, as we talk about the sensitivity of these various approaches, you know, what is their spatial-temporal limit? And are they reversible? And I'll put that question first to Tim in terms of the sensitivity of the microlens in terms of the minimum concentration that can— of bacteria cells that can be detected. And then once the tilting happens, does it untilt if the bacteria goes away? Is it reversible?

Yeah, yeah. So first of all, we can— I think we can get down to single bacteria detection. It's more about time and how— I think when you're getting the single one, you got to find the droplet. It's just that they just have to find each other. And that's what takes a lot of time sometimes. But I would say in terms of when you get to really high sensitivity, usually what you're needing is a really strong binding event, like an antibody-antigen binding event. Which, you know, by most cases, unless you're applying some sort of mechanical or some sort of energy to cause it to dissociate, is going to be largely irreversible. I mean, it's like a bond, right? So they're bound tight enough, and they may even be bound through multiple locations once they get on there. So when we get to really high sensitivity type of sensors, they're dosimetric. They're essentially irreversible unless we, you know, you could maybe use high shear or things like that to cause them. And we have done things, have seen that. But by and large, we treat them as irreversible kind of dosimetric signals. But the good thing is these droplets, the things we're using, basically cost nothing. I mean, the containers you're using cost more than the fluids you're using. Just, you know, sub-microliters of fluids and water and surfactant, just, you know, sub-nanogram quantities of surfactants and recognition elements. And so the cost is nothing. So you really kind of use them as kind of disposable, one-use types of sensors. You don't really think about them as being reusable.

Any comment on sensitivity or the repeatability of the sensing measurement technique, either Ashwin or Peter?

Yeah, so I think all the different techniques— all 3 of us have talked about very different techniques, and certainly sensitivity are also very different. So the technique that I talked about at sensitive on the almost 10 picometer scale. So if I want to look at deformation of things, I can see that very accurately and very fast. So, so one of the applications is in biomechanics. And just to link back to some of what Tim has been talking about, there are some ideas that we have not pursued of looking at protein binding to surfaces, and from there cause surface tension change. And we might be able to detect it very sensitively based on the high displacement sensitivity that we have.

So everything that we are doing, we are just basically taking existing highly sensitive detectors, be it like surface plasmon or, you know, micro-ring resonator. There's a lot of single molecule sensitivity sensors that are nanofabricated and it's available, the designs are available. The problem is basically making it specific to actually have a— putting a capture site exactly in the hot zone wherein the sensing actually occurs, and that's the problem that we are trying to solve, or it's solved. So now the big question is to kind of pull, like, to increase the amount of multiplexity so that you can actually simultaneously be looking at, let's say, 100 different kinds of molecules, all of them at single molecule sensitivity, but with specificity as well. That's one arc of problems. The other one is trying to take ideas from DNA nanotechnology, where you're constructing exquisite little constructs wherein you act as a molecular transducer, wherein you're, you're looking at molecular interactions, and those, those interactions leading to a measurable CMOS signal, be it electrical, optical, whatever else. So you have on the surface exactly where you want some kind of molecular event occurring, be that just 2 proteins interacting with each other. When they're close to each other, it has a certain signal. Further away, it has another signal. So something like that sitting on top of a transistor. Now the transistor's turn-off, on-off characteristics is going to change depending upon whether the 2 proteins are close to each other or not. So now it becomes a real-time measurement of trying to do protein-protein interactions. That's useful for things like drug discovery and stuff like that.

But there are—

it's basically, you know, once you solve the problem of how to connect the molecular world specifically to a detector, then you're left to the creativity of how you actually— what are the molecules that you put on top of it and what are the questions that you can ask from there. So that's what we are focusing on.

Very good. Thank you. So I'll ask one question and sort of pose this to each of you. It's an interesting question the way that it's posed. Are the technologies, your various approaches, translatable to real life? But specifically, can we just use dirty examples? Can we use dirty samples, or how purified of a bacteria or cell or material do we need to be looking at? Or can we take a real-world sample and directly characterize, or how do we sort of make— what's that translation step to make it applicable to those dirty samples?

I'll start. So we do have a startup on the droplets. And one of the things that's actually— when we first did this, before we got enough confidence to go forward, we went to the USDA and did a test of salmonella in chicken wash. So chicken wash, you take a chicken with salmonella and you put it in a bag and you mix it all up, and if you were putting your hand in that, it's greasy, gross stuff, you know. And so the concern I had was, gee, you know, this— there are proteins that behave like surfactants here, couldn't this mess everything up? Well, you know, we have a lot of fluorocarbon in there, and the fluorocarbon phase has nothing to do with the biology. They really just do not interact. And so that tends to give us really the ability to do things in pretty, you know, I would say confounding or, you know, you might call dirty is what the question was, environments. And so we, you know, we found that our methods well work in dirty samples and we are working to commercialize it.

For me, it's an easy answer because most of the work that we have done that I described in this talk is doing sickle cell disease. So I wouldn't call it dirty, but they are patient bloods. So they are blood that we've drawn, our clinician collaborators drawn by patients. So they are real-world sample or real patient sample. And And clean or dirty, I think, but they are real.

They are as clean or as dirty as they should be. So in our case, we've got the organization of the origami basically down to— pushed it about as far as we can get within an academic setting. And we are trying to basically get that out into the foundries to basically try to see if we can automate this. So, to be able to actually put single molecule binding sites on top of transistors or other kinds of detectors or just on an array, and then it's just a question of what are the kind of biochemistries that you put on that spot to make it useful. And we are exploring that. So, I would say we are close, but I wouldn't say it depends upon what you're going to put on those sites. Like for instance, if you put antibodies on those sites, it becomes useful for protein detection. single molecule counting there. You could put, you know, you could make other variations and themes and variations of that flavor of sensors. The question is, what is the— what is it going to give us? And that's a question that we are exploring right now at the same time that we are scaling up and just basically trying to get these kinds of single molecule arrays done in a non-academic setting.

And very good. Thank you. The real-world limitations and opportunities of characterizing and measuring at the cell and subcell level. Ashwin, Peter, Tim, thank you very much for getting us through the, the very first session. And participants, thank you for your questions.

Thank you.