Scientist Stories: Ralf Jungmann, Spatial Proteomics and Molecular Imaging
In this talk, Dr. Ralf Jungmann (Max Planck Institute of Biochemistry) outlines his lab’s strategy to achieve molecular‑scale imaging using conventional fluorescence microscopes by moving the complexity from hardware to DNA‑based imaging probes. He introduces DNA‑PAINT, a method that uses transient DNA hybridization to switch fluorophores on and off, enabling super‑resolution without specialized photo‑switchable dyes. By employing DNA origami structures as dense fiducial markers, he demonstrates drift‑corrected localization precision down to ~1 nm, surpassing the diffraction limit. He identifies the size of conventional antibody‑based labeling as a bottleneck and presents SomaLogic’s somamer aptamers—10× smaller than antibodies—as superior probes, showcasing nanometer‑scale imaging of EGFR dimers and nuclear pore complexes. The talk then describes a DNA‑encoded sequential imaging (seqFISH‑style) scheme that creates up to 1,000 virtual colors through barcoded exchange rounds, and introduces “proximity‑paint,” a split‑docking approach that reports only when an epigenetic mark is spatially adjacent to a targeted genomic locus. Throughout, Jungmann emphasizes the long‑term vision of performing proteomics and transcriptomics on a standard fluorescence microscope, effectively creating a spatial omics platform. He acknowledges collaborators (Jan Ellenberg) and funders (Paul Allen Foundation).
Chapters
Our next presentation is going to be from the team comprised of Dr. Ralph Jungmann and Jan Ehlenberg. And today Ralph will be presenting their work. And he's a group leader at the Max Planck Institute of Biochemistry.
Thank you very much, Catherine. And thanks so much for the invitation. It's a pleasure to finally be here. So Jan is here on the picture, so he couldn't make it. I couldn't make it this year, I couldn't make it last year. So, we're going to continue a little bit on the idea of super-resolution imaging in the cellular— crowded cellular nucleus and mapping epigenetic marks on the 3D architecture of chromosomes. And so, this is more going to be along the line of Jason's talk, more tech talk in the end. And so, we're doing super-resolution microscopy using DNA molecules, and the idea is to not put more intelligence into more complex hardware. So, this is as complex as it's gonna get in terms of a microscope, but we're gonna put all the intelligence in the imaging probe. And so, sort of the long-term goal of the lab, or maybe beyond the time of the lab, is can you actually do proteomics and transcriptomics with a fluorescence microscope? And so, one of the postdocs that actually left the lab a few years ago termed this localisomics. And so, the idea is everything revolves around DNA-based imaging probes.. And we're going to have a little bit of instrumentation that we need for it. What you would also need is molecular resolution. So, what do I mean with that? So, standard wide-field microscope has about 200 nanometer spatial resolution. We would need biomolecular resolution in order to tell apart one protein from another protein in a small complex. And we're going to see if that's actually going to be possible. Then if it's really going to be an omics approach with a fluorescence microscope, the usual colors for fluorescence microscopes will not be able to cut it, so we don't need of 3 or 4 colors, but you need 1,000 virtual colors. So, we need some sort of clever barcoding, and nothing of that would be possible without clever labeling approaches for proteins and nucleic acids inside a cell. So, let's get started. What's the issue with light microscopes? So, hands up in the audience who knows how STORM and PALM microscopy works? Okay, that's at least not everybody, so I would say maybe 25%. So, that's the problem that the usual microscope has. If you have an objective and you have a sort of excitation light, no matter how good your objective is or no matter how high your magnification is, you're not going to be able to focus the light to an area that's smaller than 220 nanometers in XY or 500 nanometers in Z. So you have this three-dimensional point— what's called a point spread function where every molecule or every molecular distance that's smaller than this ellipsoid appears as a diffraction-limited blur. So let's translate this to this room here, to this audience. So let's go from 220 nanometers 2 nanometers, 2 centimeters, or meters, and all you guys are going to be fluorophores. And so we're going to make a little bit of an interactive experiment here. So what I'm going to do is you cannot emit light by the nature of your body so easily, but you can raise your voice and tell me your first name. So let's count down from 3 to 0, and at 0, I want everybody of you to tell me your first name. So let's do that. 3, 2, 1, now. Excellent. So because I'm not Lieutenant Commander Data from Star Trek Enterprise, I cannot demultiplex this information that comes here. That's exactly the problem that you have with a fluorescence microscope if you look at single proteins or single DNA molecules inside a cell. So the idea in super-resolution microscopy is to temporally separate the fluorescence emission in time. So at every given point in time, only one person in the audience would shout her or his name, and I would be able to localize that person exactly and recreate a super-resolution map at the end of the day. And so you might ask, well, how is super— what's super about DNA in that context?. And so the idea of STORM and PALM microscopy is that you have some photo-switchable molecules that you switch between bright and dark states, and you need some specific buffer solutions to make that work. That's rather complicated, especially if you want to do multiplex detections, so looking at multiple colors at the same time. And so, at the end of my PhD, we had this idea of decoupling the switching between bright and dark states from the photophysical properties of the dye, and we're going to drive this change from on to off states using DNA hybridization. So, more specifically, transient DNA hybridization. We call this method DNA paint. And the idea is that, say, this is an object that you want to image, and actually on this occasion it's a DNA nanostructure. So, the other half of my lab is concerned with DNA nanotechnology. So, every cylinder that you see here is a DNA double helix. Let's say you want to image the center point here. What you do is you extend it with a single-stranded extension and you have a complementary strand in solution. Every now and so often, this molecule binds, it gets locked in place for a certain amount of time, and then it autonomously unbinds again. So what happens is upon binding, you can localize that. So this is actually real sort of intensity versus time traces. You localize it while it's bound, and if it's unbound, you just have background fluorescence. So what's important is you can actually tune these bright and dark times depending on the stability of this duplex and the concentration of strands in solution. So what that allows you to do is The question you might ask, well, how far can you push the resolution? What do I mean with molecular resolution? Can you actually use this in a fluorescence microscope to achieve 5 nanometers or better than 5 nanometers spatial resolution? And the answer is yes, if you do the calculation, theoretically you should be able to do that because it all scales with the number of detectable photons in these images. So the brighter the dye, the more precisely you can localize its position. However, although you can do that theoretically, what— how the reality looks like, because we do that on standard fluorescence microscopy hardware, This is a standard raw trace, reconstructed super-resolution trace of a DNA nanostructure. So this is an origami that has its binding sites spaced 20 nanometers apart. And what you see is you don't see anything because it's masked by thermal and mechanical drift of the microscope. So what you can do is you can do some sort of coarse drift correction, can go from this image quality to here, and you see you have these points here which are spaced 20 nanometers apart. Now we go in and do a little bit more sophisticated drift correction. Using these origamis as fiducial markers. So, you go from this quality to this quality, and finally we select each single binding site. So, this is a single DNA strand that you see here as a drift marker, and so you end up with about 30,000 fiducials in a field of view. So, that's 120 by 120 microns. So, this would be like filling your field of view with 30,000 gold beads. So, with gold beads you wouldn't see anything because everything is full of gold beads, But here the advantage is that you have these binding sites so close together, you have enough space to actually still put a cell on your microscope. So what does that mean practically if you use these as drift markers? Your real sample might look like this, and then you go over this, over this, over this at the end, and you can already tell what it is. It's an MPI logo, and if you zoom in, the spacing of these points is 5 nanometers. So here's a scale bar. So if you calculate or measure the image localization precision in this image, It's a localization position of roughly a nanometer, okay? That's with an optical microscope. That's an off-the-shelf Nikon Ti, no modifications to it. 1 nanometer is the size of a dye molecule. So, we can actually do that with an LMU logo as well for obvious sort of political reasons in my case, and I have to say the localization position is actually slightly higher in that image. So you might say, well, problem solved, right? So, you're 5— better than 5 nanometer spatial resolution, you should be able to look at a single nucleus and tell apart epigenetic modifications and in addition to that actually visualize a single genome— a single gene region in the genome, right? Because 5 nanometers, you can make the math, that's about 15 to 20 base pairs. The answer is, is that true? The answer is no, obviously, otherwise I would sort of not have done that. So, what usually is the problem is the labeling probe. So, what you see here is a microtubule image in a fixed a HeLa cell labeled with primary and secondary antibodies. So, the secondary antibody doesn't carry a dye here, but it's extended with a DNA oligo. If you do that in 3D, you can actually color those maps. So, you have red, which is sort of further up in the cell, and you have blue and green, which is sort of microtubules right at the COVID slip of the surface. So, you zoom in here, and then because it's a 3D image, you can make an XZ slice in here and turn this sort of image 90 degrees. So, what you see here is actually you look down the axis of microtubules. So, the resolution is high enough that you actually can see it's a hollow cylinder. And now, if you remember back, Cell Biology 101, microtubule diameter is 25 nanometers, roundabout. So, if you do the measurement, you measure 55 nanometers. And the reason is not because the microscope is miscalibrated, but what we are actually imaging is the position of the labeling probe and not the position of the protein anymore. So, the antibody sandwich is, compared to our image resolution, so large that we are not representing the truthful position position of the protein anymore. So, we need better probes. And so, what's the ideal probe? Well, the ideal probe is small, and ideally you have probes against the whole proteome, so all proteins in a single cell. And luckily enough, I ran into a representative of a company called SomaLogic. They are making these molecules called somamers. They are aptamers, which are sort of DNA or RNA analogs of antibodies, which fold into a secondary structure and can specifically detect proteins of interest by shape complementarity. And the good thing is they're 10 times smaller than an antibody. The DNA molecules with modified bases, they're very easily synthesizable, and you can easily extend it for DNA paint imaging. So, and the cool thing is that company has 3,000 targets available already tested, and the whole— and the goal of the company is having probes against the whole proteome. So, basically you have a list with your UniProt ID of your protein, and you say, "I want to have this aptamer." So, how does that work? Let's say you want to image EGF receptors on the cell surface. It's worked by a very talented grad student, Sebastian, in the lab. So, the idea is depicted here. So, if the aptamer that binds the protein, we have a docking site where we can do DNA paint against. So, just to get an idea of scale here, so these are sort of 4A431 cells adherent to a glass surface. You see them in ones. You already don't see anything anymore. You zoom in twice, you even see less. So, SCABER here is 40 nanometers, so now that's the diffraction-limited image. If you do the super-resolution reconstruction, you actually start to see these spots. And if you measure diameters or distances here, this is 16 nanometers, this is 14 nanometers here. So, that's about the distance of two EGFR monomers in a dimer, right? So, we actually resolving monomers in hetero or homo dimers, for for the matter of fact here. Localization position is on the order of 3 nanometers, so the size of an aptamer. And if we look at this sort of quantitatively, we can count the number of binding events per spot. You can actually see it's a unimodal distribution. So, what that tells you is each of these spots is really a single EGF receptor bound by a single aptamer labeled with a single DNA docking site. So, you can now start and go in and ask systems biology questions on a single protein level. In single cells. You can also do that inside. So, here's labeling of the nuclear pore complex actually together with Jan, where we have a GFP-engineered cell line for NUP107. It's actually a CRISPR cell line actually for the matter of fact, and you actually see here, if you see a single nuclear pore complex, you see the nuclear side of the nuclear pore complex and the cytoplasmic side of the nuclear pore complex. So, the labeling efficiency and density is good enough to actually tell these things apart. All right, so now coming back from a tool development thing, what else would you need to paint the epigenome in single cells? So, one thing is multiplexing, and multiplexing is easy to do in our case because it's all DNA encodable. So, this is an older technique already I did in my postdoc lab in Peng Yin's lab at the Harvard Medical School. So, the idea is you label everything in place, but then you do the imaging sequentially. So, in your first round you take an image of your first protein, wash out the probes, and your second round, the image of a second protein sequence encoded, and you can use the same dye molecule over and over again. Because it's sequential imaging, you don't need to have spectrally distinct dyes. So, in origami, what that looks like is 10-color imaging with super-resolution with a single-color dye. So, it's all Cy3B for the matter of fact. But how would that allow you to actually look at 1,000 genes plus epigenetic modifications in a single cell, in a single image. And the idea is actually a combination of the technique developed by Long in the audience called seqFISH, and it looks like this. So, let's see, here's your genomic DNA. So, we design FISH probes, which are sort of DNA molecules that specifically bind to genomic regions of your DNA in the nucleus, and they have two flanking regions. They have a region here on the left and on the right, which is sort of one sequence an orthogonal sequence, a third orthogonal sequence, and a fourth orthogonal sequence. And the idea is it's sort of a barcoding scheme where you do 14 rounds of exchange and each time you have a 1-bit or a 0-bit in there which allows you to encode up to 1,000 genes using 14 rounds of exchange imaging. But then you would say, well, if you can implement that, you will have all 1,000 genes that you interrogate light up at the same time in your cell nucleus. So, you're not going to see anything anymore. And here comes the idea of proximity detection. Let's say here's your epigenetic modification and you have a binder against that. Basically, you only have half of the binding site available when there's no binder next to it. If there's a binder, you complement the site, and only then, if there's an epigenetic modification close by to your gene localization, you get actually a signal. So, does that work? Here's a little bit of a test of what we call proximity paint. Let's say if both If both sides are there, you have a split docking site, you get a pain signal. If only one side is there, the interaction is too weak so you don't see anything. On origami, we tested that right next to each other, 0.3 nanometers, and actually you see if the pain signal, the P-PEN signal actually works. So, this is a 20-nanometer grid where these two binding sites next to each other. If you leave one out, you don't get any signal anymore. And so, next time, next year when we come back, I hope that we can actually show you some progress of this working intracellular at better than 10-nanometer spatial resolution. And so that brings me at the end, not without thanking all the people in the group who actually did the work, Jan Ellenberg again, and obviously the Paul Allen Foundation for making all of this possible and letting a physicist actually be part of the cohort in epigenetics. Thank you very much.
More from Joshua
Scientist Stories: Hans Waterham, The value of phenotypic assays in peroxisomal disorders
Scientist Stories: Joel Zlotogora, Genetic carrier screening programs for reproductive purposes