Scientist Stories: Alexandra Pacureanu, From 3D Light to 3D Electron Microscopy

our first speaker, who is Alexandra Pacuriano. Alexandra, do you mind— yeah, switching your video on. Hello, nice to see you. So Alexandra is our first keynote speaker. Many of you probably know her very well. So she's working at the ESF in Grenoble, where she's leading the neuroimaging at nanoscale unit. And if you have followed Alexandra's work, you probably know that she's pioneering the ultrastructural imaging using X-ray beams at synchrotrons. And what is very interesting with her work, and what I found fascinating as well, is that she's pushing the boundaries in resolution, and also she's combining many different imaging modalities together, which are synergizing tremendously what one can get with different microscopes. So Alexandra, the floor is yours. Thanks a lot for accepting to join this symposium and to give a keynote lecture. I'm really looking forward to see your latest results.

Thank you very much, Yannick, and thank everyone for being here on Sunday afternoon or morning, depending where you are. So I'll try to share my screen to begin with. Can you see it all right?

Perfectly, yes.

And you can hear me? Yes, so I'm very grateful—

Alexandra, just sorry to interrupt you, you will see normally in the attendees on your right, you will see Vincenzo is showing you some lights, perfect lights, so that will tell you when you are reaching the end of your time.

Okay, thank you. Hopefully I can, I can follow and stay on time. Yeah, so I'm really grateful to, to be invited to, to talk at this really exciting conference. I followed it last year online, and it's really something that's touching to my interest on, on all the aspects. So I will start with something that's a little bit maybe old. So we are in the in the conference on from 3D light to 3D electron microscopy. I'll start with using X-ray light as illumination. And in a way, I see this a bit in the middle between the light and EM because sometimes we start with in vivo visible light imaging, then we go through X-ray, and then we go to electron microscopy, which is for now the ultimate resolution. So I'll start with the kind of a few points that I would like to highlight as to what is the motivation to use X-rays, and in particular nanoscale 3D X-ray microscopy to explore cells, tissues, or entire organisms. So as you probably know, X-rays, especially the hard X-rays, they are very short wavelength. It's usually sub-angstrom. And when you combine this with nanofocused beams, we can reach resolutions in the range of tens of nanometres for now. And something that maybe you're not so familiar with, a contrast mechanism which we call phase contrast with hard X-rays, it's greatly improved. So we can image soft tissues or cells with much more sensitivity. In the sense that we can also image tissues or cells unstained or frozen hydrated, so to keep them as close as possible to their native state. And then, thanks to their penetration power, with X-rays we can image samples without sectioning. So for the nanoscale size, we usually start at in the millimeter, so hundreds of microns to millimeter size samples. And the way we achieve 3D imaging is by tomography. I hope you're familiar with this. We rotate the sample over 180 degrees and we acquire multiple projections, and from this we can reconstruct the 3D structure inside the sample. And kind of the, one of the advantages of this type of imaging is that the 3D resolution is isotropic, so we don't have higher resolution in plane and low resolution in the vertical direction. And it's compatible with cryogenic imaging, so we can image samples which are either, sorry about that, which are either frozen hydrated or in the case of fixed and embedded samples, actually cryogenic imaging helps to protect from radiation damage, which is one of the issues with X-ray imaging. And so, Janik said that I usually work, so most of the time work with synchrotron sources, and these offer high brilliance. That means a lot of photons per second of the order of 10^14. And what this offers is the possibility to go fast, so to acquire large volumes. And also this high brilliance, it also includes high coherence, and this helps actually to improve the spatial resolution and the contrast. So overall, we believe that this kind of imaging, it's it's interesting in order to image several samples, which we need in biology, so not just one. So we try to follow evolution of organisms as they develop, but also evolution with diseases, and also how treatments can help. And also it's compatible both downstream and upstream with other techniques. So I'll try to show you a bit of that. So X-ray-wise, I'll focus on two types of techniques. So one is phase contrast imaging, and in particular, we use holography. I won't go too much into the technical details. So basically, there are two types of techniques that we use. One is holography and one is near-field dichrography, but most of the data I'll show is acquired with holography. And this, so this is the schematic of how we acquire data for holography. So what this gives you is the structure. So what you reconstruct is electron density in 3D. And then the other technique, which is complementary and then gives you a different type of information, is X-ray fluorescence. So this is unlike visible light fluorescence. We, in this case, we place samples in the focus beam and we raster scan, and we use a detector which counts photons and also measures their energy. So for each point we record the spectra and we can fit the spectra in order to separate different chemical elements. And depending on the excitation energy that you use, so in our case it's 17 keV typically, which allows us to to detect elements with atomic number larger than 13 with very high sensitivity, sub-ppm level. So in some cases you can even try to start to count the number of atoms. And this is label-free. So you can keep, for example, if you take a cell, it can be frozen hydrated without doing any other kind of sample preparation. And this is like kind of a pedagogical example to give you a rough idea. So if you take a cell and you image on one hand with phase contrast and on the other hand with X-ray fluorescence, both are in 3D, it's possible to obtain, for example, from the fluorescence images, it's possible to segment some of the organelles, not all of them, and then quantify inside the distribution of different elements. So you can see here the kind of dominant elements like phosphorus, sulphur, chlorine, and potassium in different labelled organelles. And also the elements that are present in lower concentrations nevertheless are important for the function of the cell. And also to give you an example where we use these two techniques together to actually address a scientific question, which is actually even a question which is quite with direct impact on our daily life. It's the toxicity of silver nanowires. So these wires, which are only, in this case, is only 33 nanometers diameter. They're very slender. They're used increasingly in electronic devices, such as flexible touchscreens. So we know that we are more and more exposed to this from the devices we interact with. And the question was like, how can this affect us? Can they penetrate through the skin? Can they affect the cells? So in this case, we took skin cell types in a culture and we sprinkled these nanowires on top. And then on one hand with— so these are frozen, then they were left a few a few days so that the wires had time to interact with the cells. And one question was like, do they penetrate inside? So here what you see, it's images from the cells with phase contrast. The contrast is not great, but consider this is hard X-rays, it's actually quite good. So there's absolutely no attenuation contrast, it's all phase contrast.. And this allowed us to see, if you see this kind of small and slender dark structures, these are the nanowires. So we could see that they indeed penetrate inside, and also that the cells actually have sufficient mechanical force to bend them and to kind of hold them together and encapsulate them in what I think are endosomes in order to try to to defend themselves. And on the other hand, when you look at these different maps of the elements present in the cell, we could see, of course, the silver nanowires, but we could also see that the cell is depositing sulfur. So this initially was not present on the wires. And in some of them, we also saw copper deposition. So these are means of the cells to defend themselves. And actually, we also did the studies for different diameters of the wires, and we could see that the ones that are larger diameters, like 90 nanometers, are actually more aggressive to the cells. So they puncture and kill the cells, where these are, they seem to be less aggressive. So this could give an idea to the tech developers how to go about making these devices. Right, so how does the microscope look like? So this is what we end station. So there are X-rays coming from here. Inside this vacuum chamber is the sample stage, so everything happens in here, and even the X-ray optics that make a focused beam are in here. And then the detector for imaging is actually placed outside the chamber, so it's quite far away. And this is a detector for the X-ray fluorescence, and actually there's also an inline visible light microscope which we use to identify, for example, cells that we want to scan precisely, which is aligned with the X-ray beam. So these are maybe not very interesting parameters for you, but we work at two different energies. One of them, it's more the 17 keV, it's well suited to excite the elements that are typically present in living tissues and cells.. And this other energy, which is higher, it's better suited for very thick and absorbing sample to have enough penetration. And there is, so this is a cryo system. So actually the sample stage, it's cool. So the samples can be imaged at -160 Celsius. And this is a picture under the hood, just very briefly. So we actually work both in room temperature and cryo. So this is like a sample that's mounted in room temperature conditions, and when it's in cryo, we use a Leica system to transfer from the lab to the cryo stage. Right, so now I'll focus more on results with the phase contrast technique, in particular holography. So this has a lot of applications, I think, in in life sciences, and it has a few advantages compared to other X-ray microscopy techniques, such as that it's a full-field technique, so it's actually quite quick to acquire data, and it's easy to go to record data at different scales. So by simply changing the position of the sample between the focus of the beam and the detector, you change change your voxel size, and interestingly, you also change your field of view. And typically, we actually record— so to obtain one 3D image, we record 4 tomographic scans at different positions between the focus and the detector. So you rotate around this vertical rotation axis. And this is just to give you an idea. So what we record, the raw data, is holograms. It actually looks quite different from the final sample. So it requires quite a lot of preprocessing to get to the final data. And yeah, what you see here, this is a gold lithography on, on silicium. So this is basically what the structure looks like, and this is what you see what the raw data looks like when you place the sample in different positions between the focus and the detector. Right, so I'll show you a few examples of what this can be useful for. And I'll start with this. This is a little bit as a thank you note to EMBL because they're organising this great conference this year. So this is phytoplankton, it's a diatom, and I made it green to be like the EMBL colours. So this is studied for different reasons. Reasons. One of them is that it actually— it's important for the planet because it removes carbon dioxide from the atmosphere, and it's actually also a good kind of probe of the health of aquatic systems. But it's also like quite amazing structures, like photonic structures. So these are very, very fine— one of these diatoms, it measures a bit more than 100 microns. And staying a little bit in this marine domain, one type of materials that we can image with X-rays and we can't really image with anything else in 3D because it's very difficult to section it's mineralized materials, and in this case, nacre or mother-of-pearl, which is studied by a lot of people because it has a lot of outstanding properties as a material, but also for addressing very fundamental questions such as how nature generates highly ordered structures. So nacre, as you probably know, is made of basically tiles or bricks of aragonite, so they are quite dense, and in between there is protein, organic matter. And these interfaces actually, they're very slender, they're only about 40 nanometers thick, so you need high resolution if you want to look at the 3D structure of maker. You need imaging in 3D and a very high resolution. And this was actually one challenging project for us, so we struggled to to achieve the resolution. Also because when you illuminate it with the, with X-ray, with a quite intense X-ray beam, the aragonite can absorb and become warm, and then the organic part becomes slippery, so the, the tiles start to move and then you get a blurred image. So we had to do this in cryo, and luckily in cryo it was stable, so we could resolve these interfaces. And This is work by the group of Igor Slonikov in Dresden. So they managed to then segment this interface and they discovered a structure that was not known. There are dislocations which are like screw-like and they have opposite topological signs. So this acts as dissipated topological defects. So at the beginning the structure is actually deposited in a sequential manner and is not ordered. And then this order is achieved through this screw-like dislocations. And then kind of switching to another type of organism from nature is pollen grains. So this, for many of us, it's a source of allergy, which is a bit unpleasant. However, pollen grains are actually quite amazing because on one hand, they ensure preservation of genetic material, they protect genetic material from mass extinction. So for example, through the Jurassic period, conifers, they survived thanks to the protection offered by the grain. And also they actually use quite a lot as a template for biomimetic nanomaterials or even for drug delivery in the pharma industry. So we took this pollen grains and we tried to study them with many, many different techniques. So we started with atomic force microscopy, cryogenic 3D X-ray specks contrast or holography. To the X-ray fluorescence, focused ion beam, SEM, and high-resolution TEM in order to try to understand a bit the secrets of how the pollen can protect genetic material so well, but also to kind of see what is this fine structure which wasn't really known. So what we discovered is that the exine, so the kind of the part of the wall of the, of the grain, it's a nanofoam. So this is like the first known biological nanofoam. And also we saw, for example, that it contains quite a lot of calcium in the outer part, which can maybe explain in part the fact that it it's quite sturdy, although it's very light, so it can fly and propagate the genetic material. Right, and now going more towards maybe biomedical approaches. So with X-rays, for a long time now, we've been looking into bones, right? And when you go to the nanoscale, actually you start to resolve the bone cell network, which is an interconnected network. It's a bit like neurons of the bone, if you want. And although for a long time this kind of cells, they were considered passive. So they are initially, they are osteoblasts, they are cells that generate bone, and then they kind of get stuck in a matrix that becomes mineral and they can't move anymore. But in the recent years, people started to be able to look more carefully into them, and it turns out that they actually have essential roles in maintaining the bone quality. So they act as mechanical sensors, and also, for example, they can detect when there are small fractures, small, very small, like nano cracks. For example, if you go for a run your bones may develop small defects. And the fact that this— so you see here, it's like about 300 microns of bone tissue where you see there are many, many, many cells in organizing kind of cylindrical multilayered structures. And here are blood vessels and nerves that come in this hole. So then actually what is around here, it's hard mineralised tissue, which you can see in the image on the right. So here you see in orange the cells, and then the structured tissue around is made of hydroxyapatite crystals combined with collagen fibres. So yeah, so this kind of cells are actually quite interesting to— and the network in particular, it's interesting to look into for many reasons. And when we talk about diseases, for example, we could see that people with osteoporosis show that this network, it's disrupted or disconnected, which is maybe one explanation why people get fractures like without any severe mechanical impact. Or without doing anything that normally would break your bones. And also we could see that the connections between them are actually quite varied. So on one hand, we could see that there are kind of hubs of the connection. So you can see one here in 3D. So it's not that only cell-to-cell connections, but there are also kind of meeting points and also that the— so you have branching and sometimes you have self-redundant connections where they go back. So yeah, people wonder, okay, what, how does this really work? And what is the effect of different types of connections on the quality of the tissue? And one thing that one can do is to use fluid flow simulation. So what happens is that through this kind of canals, interconnected canals, there's fluid flow which is helping as for mechanosensing. And what one can do is to kind of severe some of these connections and see how the fluid flow changes. So there's a community of people that are trying to advance in this direction, and this is becoming more and more— it's enabled more and more by X-ray microscopy at the nanoscale. Right, so now going away from mineralized tissue and going towards soft tissue. So one One advantage of this kind of technology I was saying is that the contrast, it's good even in unstained tissue. So these are biopsies of tissues from lung and heart from patients which have died with COVID And this type of tissue, it's fixed, it's embedded in paraffin, but it's not stained. So you can see that at subcellular scale, we can we have quite good contrast and we can look, for example, into the subcellular structure and what happens in the lungs of people affected by COVID. So what you can see here in the rendering is that normally the tissue is quite— it's like a lace structure. So it allows exchange of air, while in case of COVID it's really clogged. There's a lot of kind of debris of dead cells, but we also find quite a lot of kind of calcifications and even kind of iron agglomerations, which we are— for now we don't know exactly why they are forming, what they mean. But I would like to underline here that the fact that you can image unstained tissue, it also gives you the possibility to combine it with this other technology, like fluorescence, so to see in the same time structure and elemental composition. All right, so now I'll switch to neuroimaging, which is more my main interest at the moment. So I'll go quickly through a few points of why This kind of technology, it's particularly interesting for neuroimaging. So there's like a global challenge and a global kind of venture at the moment to develop scalable neuroimaging at nanoscale and to try to reconstruct comprehensively neuronal circuits in brains, in full brains if possible, in more brains, several brains ideally. And yeah, this is necessary, as you probably know, on one hand, just to understand the basic principles of the fundamental mechanism of brain function, which we still don't understand completely, but also for most of the neurological diseases, we have no treatments. We don't really understand them. So there's clearly a need to deep more deeply and more precisely into the brain. And yeah, and there's a question of a tyranny of scale here. So I took this from a Cell short paper which is written by a large group of world-leading neuroscientists which are basically pleading for resources and development of technology to to map for the first time a mammalian brain entirely at a scale that allows you to see all the neurons, all the other cells, but also the connections between neurons, so synapses. And as you can see here, if you start with the full brain of a worm, which only has about 300 neurons, so this map now exists, but it took about 20 years to create. People started this a long time, a very long time ago by electromicroscopy, so by slicing carefully and imaging and then stacking back together, recomposing a 3D image of the full brain. And then more recently, a first fruit fly brain has been imaged at Janelia. There are much more neurons in there, but also the cost is quite quite impressive, and there are many— takes years to just to acquire the data, not to talk about the analysis. So yeah, getting to a mouse brain, it's a bit like— in scale-wise, it's a bit like going from a seat on an aircraft to going from Boston to Lisbon across the ocean. So it's pretty daunting task.. And yeah, this is a nice picture of a fly. So this was an article in the New York Times a couple of months ago, which was celebrating basically the first map of the fruit fly. And I found it quite funny that, okay, the title was "How to Map a Fly Brain in 20 Million Easy Steps." So what we try to do with X-ray imaging is to reduce these 20 million easy steps to just a few steps.. And what we were happy about is that, okay, for a fruit fly brain, you can basically fit it in one field of view at 100 nanometers. Of course, this is not at all the resolution that you need to trace all neurons and synapses, but it gives you an encouraging starting point. And also, so in the Li lab at Harvard Medical School, we looked after imaging with X-ray holography in 3D, Jasper and Aaron, they sliced tissue and they looked at it with transmission electron microscopy. So you can see the contrast is actually fairly similar, but of course the resolution is not that good, but there is room for improvement. And you can also use actually the two techniques combined. So you can combine X-ray microscopy in 3D with volume electron microscopy in order to kind of more quickly cover more volume and trace the dendrites. For now, we don't really see the synapses, or not enough of them, but then you can target and study synapses in specific regions. And if you want to know more about this, I invite you to follow the talk of Aaron tomorrow evening. And then in some cases, you can actually use only the X-ray microscopy to obtain sufficient information to address scientific questions. So for the, in this case, Jasper was trying to kind of learn more about how walking is controlled in the fruit fly. So for this, we imaged almost the entire leg of the fly and a part of the ventral nerve cord, which is kind of their spinal cord. It was possible to trace motor neurons from the ventral nerve cord all the way into the muscle. You can also see not just the neuronal tissue, but also you can see muscles and all kinds of other organs which help in coordinating walking. Then this is just a small parenthesis that of course, this is very large data, very complex data. In order to reach the knowledge you need to segment usually. So the data analysis is really important. And luckily, there are now a lot of teams that work on automatic segmentation based on machine learning. And things are starting to be more promising that one day we're gonna be able to obtain actually the segmented data without years of manual work or without having the infrastructure of Google for computing. Right. So just quickly to show examples where you can address more directly questions related to the biomedical field. So for example, with people in Lund from the hospital, we looked at biopsies from the wrists of patients with diabetes. This was led by a neurosurgeon, Lars Dahlin, and he wanted to know more about how the nerves— what you see here is a cross-section to a fascicle of nerves, and these are myelinated axons. He wanted to see what is changing in the structure when people get pain and numbness in their limbs due to diabetes. Is like diabetes-induced neuropathy. So what we could see is that besides the nice and well-aligned and kind of constant diameter axons, there are also some that are smaller with less myelin and also kind of entangled. So this kind of gives you some insight into what is the underlying morphology for for neuropathy. And also, of course, there are many, many mouse models for diseases. And this is very useful to, for example, people at Sapienza in Rome, the group of Alessia Cedola, they were looking into a possible treatment of multiple sclerosis in a mouse model. And so it's a treatment with mesenchymal stem cells. So we looked at the spinal cord. So what you see here is the middle of the spinal cord and you see blood vessels and in blue is just cell bodies. And it was quite nice to see that following the treatments, there's both regeneration of motor neurons in the spinal cord and regeneration of the blood vessel capillaries, which were really disrupted in the diseased model. And actually, sometimes X-ray microscopy can serve to generate ground truth for other techniques. So for example, for in vivo MRI. So this is used as a diagnosis technique and there are biophysical models that basically translate the measurements. So you get probabilistic measurements of the flow to the brain. And this is relying on assumption on the morphology at the axonal level. So this is more on white matter. Typically, the axons are supposed to be all parallel with constant diameter, which is actually, when you look at the structure in detail, you see that this is not really accurate because on one hand, the diameter is varying quite a bit. And there is interaction between the axons and the cells around. And also some, in some areas, they're actually quite entangled. So one needs to develop a bit more accurate models, biophysical models, in order to use more reliably in vivo MRI for patients. And very quickly, how you can use also in the context of neuroscience both techniques together. So another model, mouse model in this case, Alzheimer's. What you see here, it's an image from hippocampus. It's also unstained tissue, so you see healthy neurons which appear brighter and nice and round, and then some which are dead or dying, which appear darker and seem to be dried out. We took then slices from here and looked with X-ray fluorescence to see what happens at the elemental level. And we could see that the neurons which are affected by the disease, they— well, on one hand you have like a double redistribution of the elements, but particularly we saw that there is a bit of a higher concentration of iron in the neurons which are affected. Right, so now I'll just finish by kind of mentioning the challenges that we face and that we try to address. So one of them, it's imaging maybe whole brains in multiple samples. So this is work with people in Australia. You still hear me? I hope so. Yes. Thanks, because I saw something weird on my computer. People at the University of Queensland in Australia, they are looking into autism spectrum disorder using zebrafish model. First, in vivo neurophysiology in whole brains. What we're trying to do is to kind of relate the morphology to the phenotypes with the— on the one hand, with the— so it's a genetic mutation that makes the model, but also with what you see in vivo, in the measurements. In this case, we tried to do whole-brain imaging on several on several animals. And for this, we need to kind of speed up, and we need to sort out stitching as well, because if you want a certain resolution, you cannot fit the whole brain in the field of view. So you need to kind of reliably be able to stitch it together. And then another aspect is improving the resolution. So one difficulty with X-rays is the radiation dose that you want. If you have higher resolution, you want to put more dose on your sample, and it can damage the sample. So for this, we're actually looking into image processing ways which allow us to work with low dose, which on one hand allows us to protect the sample, but also to go faster. And then increase the resolution by post-processing. And we have quite promising results with machine learning-based denoising methods. And then, of course, we want to go multiscale. And also sometimes not just anyway multiscale, but we also want to target specific regions.. So for this, one needs ways to kind of guide the targeting. So what you see here, this is work that we do with Andreas Schaper's lab at the CRIC. So you can see tissue from the olfactory bulb image at different voxel sizes. And you also can see how much the field of view is changing. So if you go from 650 nanometer pixel size all the way to 25 nanometers. So of course, if you want to cover the whole tissue at 25 nanometers, you have to be able to stitch and image many, many subvolumes. And yes, Alexandra, just to tell you that you're already like a few minutes over time.

Okay, sorry, yeah, I've got questions.

You have to, uh, I'm done. Yeah, it's the last slide. So yeah, so for this part, you, you can follow Carlos's talk, which is a bit later today. Um, and just briefly to mention, so this, all these technologies are actually available. If people are interested, you can apply for access or do collaborations. And we're also working from going from cryolite to cryo X-ray microscopy. So yeah, multimodal and multiscale for the win. So we've been trying a few combinations of technologies. Yeah, and thank you. So I have to say that this is a work of a lot of people, as you can imagine, and it's hard to cover everyone in one slide. And I'll just finish on this image that we need to stand together for peace and freedom, especially in these times. So I'll stop here.

Okay, my Zoom, it's going a bit— unfortunately, we don't see your video anymore, so we don't see you anymore.

Maybe your camera crashed. It's— yeah, my Zoom is not happy.