Transcript

New tools for precision multimodal measurement of single cells and single molecules

From Berkeley Lab · Episode page · Watch on YouTube

· 1 h 11 min · 10,663 words · 70 timestamped sections

This is the full, timestamped transcript of "New tools for precision multimodal measurement of single cells and single molecules", an episode of Berkeley Lab published January 6, 2026 (1 h 11 min, about 10,663 words). It mentions 14 people, works and ideas, including Aaron Streets, RNA sequencing, Raman scattering, microfluidics, lipid droplets.

Who speaks in this episode?

What is discussed, and when

  1. – 10:00: Aaron Streets, microfluidics, Dr. Shamani Jan
  2. – 20:00: RNA sequencing, microfluidics, Chan Zuckerberg Biohub, Human Cell Atlas
  3. – 30:00: Raman scattering, microfluidics, RNA sequencing, spleen
  4. – 40:00: lipid droplets, adipocytes, Raman scattering, PDMS
  5. – 50:00: microfluidics, lipid droplets, Raman scattering
  6. – 1:00:00: spleen, RNA sequencing, Visium
  7. – 1:10:00: microfluidics, spleen

Notable quotes from "New tools for precision multimodal measurement of single cells and single molecules"

  • Jan: Welcome everyone. I'm happy to see quite a number of people, despite being so close to the holiday break. So it's my honor for today's instrumentation colloquium to have…
    about Dr. Shamani Jan · at
  • Imaging modalities and combine microfluidics and other imaging things to really study details of life and cells. And he will tell us a lot about all these different things, and I'm really excited to have him here today.
    about Aaron Streets · at
  • These other molecules into DNA, encode them in DNA sequences, then you can use a sequencer to read out the composition of the proteome, or the certainly the RNA and I think RNA…
    about RNA sequencing · at
  • Look at specific genes, or you can look genome wide and just ask, if I cluster cells by their gene expression can identify unique types, and this has been the dominant method…
    about Human Cell Atlas · at
  • Throughput. That's what I'm gonna talk about. So the way that we approach this problem is with microfluidics. And broadly, we use microfluidics to help us combine genomic…
    about microfluidics · at
  • There's a lot of technologies, and so we want to basically combine imaging with sequencing, and the imaging that we do a lot of in the lab is a technique called stimulated…
    about Raman scattering · at
  • Them warm before they develop the Shivering response. But there's a lot of morphological differences between different kinds of fat and we're really interested in studying how…
    about lipid droplets · at
  • Really interested in looking at lymphatic tissues and how immune cells develop in thymus and spleen and lymph nodes. So we can take an image of a spleen. This is a single field…
    about spleen · at

Who and what is mentioned in this episode?

People

  • Aaron Streets — Person · Main speaker (invited colloquium guest) · 2 mentions · first at
    Delivers the technical talk on combining microfluidics, Raman imaging, and sequencing for single-cell and spatial biology; central technical authority in the session.
  • Dr. Shamani Jan — Person · Host / colloquium organizer (introducer and moderator) · 1 mention · first at
    Introduces Aaron Streets, frames his background and research focus, and moderates Q&A with in-room and Zoom participants.

Companies & orgs

  • Chan Zuckerberg Biohub — Company · Affiliation · 1 mention · first at
    Research organization where Aaron Streets is a part of, contributing to biological studies.
    Also in the biophysics graduate group and the computational biology Center for Computational Biology, which is a graduate group. And I'm also part of the Chan Zuckerberg…
  • Peking University — Company · Affiliation · 1 mention · first at
    Institution where Aaron Streets' colleague did the first single-cell RNA sequencing.
    Pretty is pretty standard. The hard part about single cell RNA sequencing is doing it for one cell, because you have to know that you have one cell in your tube. And so my…

Products & tech

  • PDMS — Product · Research tool · 1 mention · first at
    Material used for fabricating microfluidic devices in the lab.
    Actuatable valves. So in short, this is a this chip is about a couple centimeters by a couple centimeters. These channels are embedded in the PDMS using a molding and casting process.
  • Visium — Product · Research tool · 1 mention · first at
    Equipment used for spatially resolved transcriptomics in tissue sections.
    Computationally, right? And so what we've been doing is we've been imaging our spleen section and then printing it on transferring it on this array using the visium site…
  • micro pipette — Product · Research tool · 1 mention · first at
    Tool used historically for isolating single cells for RNA sequencing.
    Measurement in 2009 where he isolated a single cell by diluting a population of cells in the petri dish and using the micro pipette. Actually, at that time, he used the mouth…

Concepts & topics

  • RNA sequencing — Concept · Core topic · 11 mentions · first at
    Primary method discussed for identifying cell types by their gene expression profiles.
  • Raman scattering — Concept · Core topic · 9 mentions · first at
    Key technique discussed for imaging cells without fluorescent labels, providing rich chemical information.
  • microfluidics — Concept · Core topic · 8 mentions · first at
    Central to the discussion on combining genomic measurements with optical measurements for single-cell analysis.
  • lipid droplets — Concept · Research focus · 8 mentions · first at
    Key subject of study in adipocytes, examining their morphology and metabolism.
  • adipocytes — Concept · Research focus · 2 mentions · first at
    Primary cell type studied in the context of lipid metabolism and gene expression.
    That we're really interested in measuring is lipid metabolism in fat cells, or adipocytes. Specifically, we're looking, we're interested in how different types of fat cells…
  • Human Cell Atlas — Concept · Research initiative · 1 mention · first at
    Project mentioned in the context of high-throughput single-cell RNA sequencing.

Places & events

  • spleen — Place · Research focus · 8 mentions · first at
    Key organ studied in the context of immune response and tissue imaging.

Full transcript with timestamps

Unknown: Yeah, it is working. Let me just flip through a couple One two I okay, how you doing? I Rebecca Carney: Hey, welcome everyone online. We're going to get started about five after. Jan: Welcome everyone. I'm happy to see quite a number of people, despite being so close to the holiday break. So it's my honor for today's instrumentation colloquium to have Aaron streets here from UC Berkeley. Aaron is did a lot of stuff in California. So he did his undergrad in LA and if I got it right, in arts and physics, so a double major kind of which is quite interesting. And then he did his PhD in Stanford in Applied Physics, basically. And then he did a postdoc in Beijing. And after that, he came back to California and actually got a position here at UC Berkeley as an assistant and now an associate professor, and he's in the bio engineering department, and he's really pushing methods to get new

imaging modalities and combine microfluidics and other imaging things to really study details of life and cells. And he will tell us a lot about all these different things, and I'm really excited to have him here today.

Aaron: Well, thank you so much Jan for the invitation. In addition to spending a lot of time in California since college, I also grew up right down the hill. I grew up in Berkeley, and so I love being up on top of the. This is my favorite view in all of the world, I think, from the Lawrence Hall of Science specifically. And yeah, so I've been in the I'm in the bioengineering department. I'm also in the biophysics graduate group and the computational biology Center for Computational Biology, which is a graduate group. And I'm also part of the Chan Zuckerberg Biohub. And the lab has students from from all three of these programs. And even we even have a student from MCB now, and so we, and sort of the overarching theme of our work, is to build tools to study biology at the single cell level. And I put single molecule here too, because we have some a new suite of tools for studying single molecules of DNA using sequencing, but I'm not going to

talk about that work. I'm going to focus on the single cell and the imaging and the microfluidics and Okay, so the sort of the driving theme is not just, how do we study biology single cell level, but how do we actually define what is the identity of a cell? How do we quantify the identity of the cell? And this is a very pervasive question in all of biology, how do we characterize cellular behavior or cellular phenotype or function? And we do this for sort of two broad classes of regions. One, we might want to see how cells change as we as the organism undergoes some sort of perturbation like aging or some disease or or maybe a response to a molecule or a drug or injury, and in addition, but in a lot of those measurements can be done in ensemble. So you can take, for example, blood and characterize the DNA or the RNA or the proteins or the biomarkers in the blood for somebody who is old or young or sick or healthy, and see how

those markers that might indicate the prevalence of certain type of type of cell, how they change? But more you know, it's also really important to be able to do this for individual cells, because cells can be very different, and sometimes the response that we're looking for is not the change of the cell, but the change of the cellular composition. So how does the frequency of a certain type of cell change and disease, we have more white blood cells or less, or more immune cells or less. And so being able to measure cellular identity at the single cell level and high throughput is important. So that begs the question, how do we measure, how do we quantify the identity of a cell? Well, there's different molecules we can use to to do that, we can, we can, for example, ask what RNA is in the cell. We can ask what proteins are in the cell. We can ask what other molecules, metabolites, lipids, other molecules that are maybe not encoded directly in

the genome. And of course, we can also ask, well, what's the genome of that cell? Most cells in our body have the same genome, two copies of our of our genome. Some cells have different copy numbers. Some cells have mutations. Cancer cells have mutations. But you know, we can measure any one of these molecules. Also. We can look at a cell and say, What, and ask, what does it look like? And in fact, you know, the oldest and still one of the most powerful single cell measurement tools is the microscope. And in addition to characterizing the morphology of the cell, the microscope is incredibly powerful for characterizing these molecules with with fluorescent labeling and really high powered optics and a visible spectrum. We can detect individual molecules, of these, of any of these molecules. We can detect individual RNA molecules and RNA molecules, individual DNA molecules, proteins, some of the metabolites are harder because

they're smaller and in the power of optical microscopy, is that we can not only count single molecules, but we can see where they are and where a molecule is in a cell. It might be really important for what that cell is doing or why it's doing it. The challenge with optical microscopy, however, is that it has limited bandwidth. And when I say bandwidth, what I mean here is there's a limited number of channels or or frequencies or wavelengths that you can label molecules with and detect with visible light, limited by the visible spectrum. So with a really good spectral detector, you might be able to detect 20, maybe 50, if you really are good unique molecules. But if you wanted to identify a cell by characterizing its RNA or its protein on a genome wide scale, there's hundreds of 1000s of RNA molecules. There's hundreds of protein molecules in a cell, you need a higher throughput measurement, and that's where

sequencing comes in. DNA. Sequencing is essentially a highly multiplexed visible light microscope that can read out A, T, C and G sequentially to decode the sequence of DNA. So if. You can, if you can turn the identity of these molecules so you can clearly sequence the DNA from a cell using DNA sequencing. But if you can somehow turn the identity of these other molecules into DNA, encode them in DNA sequences, then you can use a sequencer to read out the composition of the proteome, or the certainly the RNA and I think RNA Seq. RNA sequencing has become recently, in the past 10 years, one of the most powerful and quantitative ways of identifying cell type, because the transcriptome of a cell that's RNA composition tells you which genes from the genome are being used, which proteins are being transcribed and later translated. And while the transcriptome doesn't reflect exactly the proteome, it

gives us a very quantitative annotation of the number of the amount of expression of all of the 20,000 genes in the genome. And so if we can sequence the RNA from a cell, we can get this very rich data set that's quantitative that tells us sort of the gene expression profile, and we can use that data to sort cells by their gene expression profile. You can use you can look at specific genes, or you can look genome wide and just ask, if I cluster cells by their gene expression can identify unique types, and this has been the dominant method for endeavors like the human Cell Atlas, which takes on the challenge of sequencing or identifying all the cells in human body. There's trillions of cells, at least 1000s of distinct types of cells, and maybe more, depending on how you define a cell. And so high throughput single cell RNA sequencing has become an important tool for annotating all of the cell types in human body. The way that we do it is

we we take the RNA, we reverse transcribe it into DNA, so that we have DNA that represents all the RNA, and the n we sequence it on a DNA sequencer. And that technique, that method, that molecular biology method, has been around for over 20 years, and it's pretty easy to do now. I mean, it just costs a lot of money you sequence but, but you know that molecular biology is pretty is pretty standard. The hard part about single cell RNA sequencing is doing it for one cell, because you have to know that you have one cell in your tube. And so my colleague at Peking University did the first single cell. Fu Cho tongs, in his post doc Cambridge, did the first single cell RNA Seq measurement in 2009 where he isolated a single cell by diluting a population of cells in the petri dish and using the micro pipette. Actually, at that time, he used the mouth pipet micro pep to pull out one cell from under a microscope, put

that cell in the tube, and then perform the molecular biology protocol to do the RNA sequencing for that one cell. And you know, one of the some of his most interesting and impactful early papers was that experiment many, many times to get enough single cell measurements to have good statistics to say what's going on. Because this is a noisy measurement. Even though there's a lot of RNA in a cell, it's really not that much from the point of view measuring it. And so this is the single cell RNA sequencing is very noisy, but it's still very informative. And so not only are you want to get lots of cells for one cell type just to be able to know what the center of that distribution looks like, and then you have to do that for as many cell types as you have to really find any differences. And so that's a laborious process, and that's one of the reasons why it's hard to do high through to do things.

RNA sequencing. This is a figure from a review that I thought I had put the reference up. Maybe it's out of off of the screen. It's Valentine and the database in 2015 I think, where this is just a sort of a metadata analysis of papers, of manuscripts over the past 20 years. And the x axis is time, and the y axis is how many cells that paper sequenced or analyze. So the black dots are RNA sequencing, but then the red, the orange and the green dots are other methods, like microarray or qPCR. And you can see that for a long time, people were measuring 10s and hundreds of cells, one at a time, and then right around 2014 Unknown: 2015 there's sort of this inflection in the number of cells people could measure. And we wrote a review where we analyzed, you know, not we analyzed. We talked about why this happened and and right around 1014 is when people started using micro fabrication to develop high throughput

devices for single cell RNA sequencing. And so there's a number of technologies that came out, micro fluidic platforms, micro well arrays, you could rapidly sort single tails into liquid handling automation, microfluidic droplets, etc, and also some very sophisticated DNA barcoding methods so that it will allow you to attach a unique DNA barcode to the RNA from a cell, so that you can do highly multiplex single cell RNA sequencing without doing. Individual reactions to drastically decrease the cost and made these kinds of measurements accessible to people that didn't develop technology. And so then after that, people in biology labs that were interested in doing this kind of measurement could do high throughput single star RNA sequencing measuring up to 1000s, 10s of 1000s, and now hundreds of 1000s of cells in a single experiment. And so that's kind of what we've been working on for the past 10 years as well, is develop is developing

new tools to increase the throughput, how many cells you can measure per experiment, and the sensitivity, how many of those transcripts you can actually detect. Because conservatively, I think, you know, we're good at detecting about 20% of the transcriptome, which is still a lot of data, but it's very ends up being very sparse. The data that comes from these measurements is a matrix of cells by genes. So however many cells you measured by 20,000 genes, and each element in that matrix is the number of counts you measured for that gene, for that cell, most of those elements are zero. Net matrix is very sparse, but it's still enough information to categorize what a cell is very quite precisely. One of the things we've been doing more recently is not only trying to improve the sort of performance of these measurements, but also starting to think about how we can simultaneously measure other

modalities in the same experiment. Because one of the while sequencing is very powerful, one of the problems with sequencing is that it's destruct destructive. You have to lyse the cell and take out all the RNA and sequence it, and so you lose all of the interesting morphological information that you get with imaging. And also not everything is encoded in the transcriptome, the lipids, the metabolites, the molecules that are that are metabolized by the enzymes that are translated from transcripts those aren't directly encoded.

And so to get a more holistic view of what a cell is, it might be nice to be able to take a picture of it and then sequence it. And so we're working on ways of doing that in high throughput. That's what I'm gonna talk about. So the way that we we approach this problem is with microfluidics. And broadly, we use microfluidics to help us combine genomic measurements with optical measurements. And when I say genomic measurements, I basically mean a large class of experiments where you're turning different molecules into DNA and then sequencing it. So that might be like reverse transcription of RNA into DNA. But also, there's a way that you can tag proteins with antibodies that are bar coded with a unique DNA molecule, so that when you sequence, you can sequence this unique sequence that is representative of that particular protein. And so in a way, you turn the identity of that protein into a DNA molecule that you can sequence. So

there's a lot of technologies, and so we want to basically combine imaging with sequencing, and the imaging that we do a lot of in the lab is a technique called stimulated Raman scattering microscopy. That's what this image here. This is a Hela cell. And so a little intro to stimulated Raman scattering microscopy. Some of you may know this technique like the back of your hand, so forgive me if I'm starting, you know, before your expertise, but I'll do that for everything. One thing I'll note is that when we are if and when we are successful making these kinds of measurements, we're left with a lot of data, and it's not all easy to interpret. You know, we have high dimensional, large data sets with many cells and many cells and many different layers of information for each cell. And so we've also been partnering with some people in EECS to figure out how to develop new algorithms to learn sort of

lower dimensional representations of this high dimensional data so that we can analyze it, so that humans can understand or is it so that machines can understand it. And on one hand, this gives us the ability to cluster cells in a low dimensional space based off of multi dimensional data and get sort of more refined definitions of cells. But also analyzing these molecules at the single cell level, in low dimension, allows us to understand how these molecules interact, how information flows between the transcriptome, the proteome, the shape of the cell, through the lens of different cell types. So there's a lot of rich information that you can glean from this kind of data. And there's a whole field. And in fact, there's more papers about computational methods for analyzing this type of data than there are papers for new methods for getting this kind of data these days. It's really interesting. So that's a whole field. That's a whole field in

of itself, and we dip our toes in that a little bit as well. So I'm going to try to get through sort of two classes of projects. One is imaging and sequencing single cells. And then I'm going to zoom out and try to see if we can image whole tissues, because the problem with single cell analysis is that it's very valuable for categorizing the cellular anatomy of organisms, but you've ripped the cell out of its native environment, you lose a lot of information about what that cell is. Sometimes the identity of the cells is really where it is. Is not just what it is. So an immune cell in the spleen might be different from an immune cell in the brain or in the gut or something like that. So being able to image these cells in situ is also a really important challenge in the field. Okay, so the imaging technique is stimulated Raman scattering. So most light scattering that we see is elastic. Right? The energy of

the photon, the incident photon is the same as the scattered photon. And so, you know, this green laser is Rayleigh scattered in Raman scattering. Raman scattering is inelastic. And so the scattered photon is red shifted, is has a lower energy than the incident photon, and some of the energy is transferred into vibrational modes in the matter that that was the scattering center. And so if you can measure the difference between the incident energy and the scattered energy, you might be able to in first thing about the vibrational energies in the material that you're scattering light off of.

And specifically so these, this is, you know, energy states of a molecule, let's say a molecule like this. There's the electronic energy states which we excite for fluorescence, and then there's these ground states, which are based a lot on the energy on the vibrational spectrum of that molecule, so that the energy might be defined by the atomic mass, the nature of the molecular bond and and then also different modes of vibration can cause some finer grained splitting of these energy states. And so there's a lot of rich information into the chemical identity of molecules that you could glean from the this, this vibrational spectrum, which you can probe with infrared spectroscopy, or with stimulated Raman or with Raman scattering. And this is a an example of the Raman spectrum of a cell. And so you see that we have these large peaks in the CH region. And then there's this, we call this a fingerprint

region, where there's a lot of contributions from a lot of different molecules. And then there's a silent region. There's a silent region. The Raman spectrum has a lot of information, but it's not always easy to figure out what that information is corresponding to. A lot of proteins, for example, have very similar Raman spectra because they have very similar amino acids in the subtle shifts in their energies is based on the structure that protein are hard to measure without very high resolution spectroscopy. But, yeah, this is a very information rich space, and if you can profile it, you might be able to learn a lot about a cell. So spontaneous Raman scattering spectroscopy method, you just use a continuous wave layer. Usually we use this on the on a sample and a cuvette, and we just, you know, have a diffraction gradient, and we collect all the scattered energies, the diffracted energies from we block the sort of the primary incident

wavelength, and then look at the scattered energies. And this is a very low cross section interaction, but if we can measure across them that we will find peaks energies that correspond to transitions that were based on the molecules that were present in our sample. And so this takes a long time to integrate. So Raman spectroscopy is relatively slow. Could take seconds to minutes and some kind times longer. But it's not well suited for imaging because it takes a long you need to collect a lot of photons. So stimulated Raman scattering is kind of analogous this stimulated emission fluorescence. And the idea here is that we choose two photons that the difference in their energy resonates with the vibrational energy that we want to probe. So let's say we now we, instead of doing spectroscopy, we know we're looking for, let's say, the energy associated with ch two asymmetric stretching. So we

choose two photons such that the difference in their energy resonates with that vibrational mode. And collectively, those those photons sort of stimulate the molecule to this virtual state, and it relaxes down to the resonance state, and this releases a two photons, Raman loss and a Raman gain photon. And so you can measure this, this blue shifted anti Stokes photon. And this is a technique called coherent anti Stokes Raman scattering. This is a very popular imaging technique. But also this, the presence of this molecule creates a transition of our pump and our Stokes photons such that we energy from the pump wavelength or the pump band gets transferred to the Stokes band. So if we can measure the loss in this frequency or the gain in this frequency, then that is linearly proportional to the number of Raman scattering centers in our sample. And so this has incredibly This is because it's coherent excitation. We have a much

higher cross section for this interaction, and we can get enough signal fast enough where we can imagine scanning this laser across the sample and imaging. And so the challenge, however, is that we're looking at. Loss or a gain. We use femtosecond or picosecond lasers to do this excitation. We need high peak power at the at the scattering center, and we're trying to look at a loss or gain in signal on the background of this very high intensity laser. And so the strategy is to modulate one of the beams. In this case, we modulate the Stokes beam, and let's say, 10 megahertz, and then we collect the unmodulated pump, the transmitted pump beam. So we'd look at forward scattering, and if the molecule that resonates with the difference that we chose is present, we will see a loss of the pump signal at the 10 megahertz modulation frequency. So we put, we detect the forward, scattered and transmitted pump beam, we filter out the Stokes beam, and then we

put that signal into a locking amplifier that's that's synchronized with this 10 megahertz modulation. And we look, we in that locking amplifier amplifies this stimulated rum on loss, which we see this decrease at the 10 megahertz. And this signal is linearly proportional to the number of scatterers. And so we align these two lasers, we synchronize them in time, we modulate the Stokes, we put it through a scanning laser scanning microscope, like a two photon microscope, and we detect in forest scattering the pump signal. And because this is a non linear process, the signal is exponentially proportional to the intensity at the focal plane. So we only get signal from the focal plane that we're focusing on. So kind of like two photon microscopy or confocal microscopy, we can scan an X, Y and Z and do a three dimensional reconstruction of the abundance of our target molecule in the cell. And so, for example, if we wanted to say where all the

lipids in a cell, we could, we could lipids have a lot of CH two. So we could tune our our our optical parametric oscillator so that the difference in the pump and the Stokes resonate with the CH two, asymmetric stretching vibration. And then we can scan that across our sample and reconstruct the position of all the lipid molecules, and all of them, but some fraction of them, in in a cell. So this is a Hela cell, and these bright puncta are lipid droplets which have a lot of CH two. So it's, it's, there's no labels. There's no fluorescent labels. In this method, it's, it's quantitative, so the signal is linearly proportional to the concentration of the molecule of interest. You can get 3d sectioning. You can do this on live cells, so we can do time lapse imaging. These are HeLa cells in oleic acid. So they're up taking fatty acids and metabolizing them into lipids.

And we can see this over many, many hours, because there's no fluorescent labels. There's no photo bleaching. So we can do this for a long time. We have to be careful about photo damage, because these are femtosecond lasers, and so we've done a lot of characterization of at what photon flux can induce reactive oxygen species or differential expression, or can impact proliferation. But at low power, at low average powers, we can we can make sure to not perturb the cells too much. And the challenge, however, is that while there is some chemical specificity, there's a lot of overlap of signal on the Raman spectrum. So not all of this signal is coming purely from lipids. There's some of the signals coming from CH three, and the sensitivity is not near what it is in fluorescence. So we can't do single molecule imaging with Raman, at least not in this configuration, but it's very good for a lot of a lot of

types of measurements. For example, this, one of the things that we're really interested in measuring is lipid metabolism in fat cells, or adipocytes. Specifically, we're looking, we're interested in how different types of fat cells metabolize lipids differently. There's there's white adipose tissue, which is the kind that we don't like, which stores fat in these large unilicular lipid droplets. This is a single adipocyte, actually, and then there's brown adipose tissue, which stores lipids in small lipid droplets, and is metabolized those lipids to produce heat. And it wasn't actually known that adult humans had brown adipose tissue. Up until relatively recently, we knew that infants, human babies, had brown fat in the back of their neck because it helps keep them warm before they develop the Shivering response. But there's a lot of morphological differences between different kinds of fat and we're really interested in studying how the

gene expression for different types of fat cells correlates with the lipid droplet morphology in those fat cells. So combining stimulated Raman scattering imaging with single cell RNA Seq, we think is a powerful way of answering some of these questions. So the way that we do that is with microfluidic devices. So this is a PDMS chip. This is a multi layer integrated microfluidic device with pneumatically. Actuatable valves. So in short, this is a this chip is about a couple centimeters by a couple centimeters. These channels are embedded in the PDMS using a molding and casting process.

PDMS is a clear polymer. It's silicone, the same stuff we use to seal our bathtub, but a clear version. And so we can imprint channels into the PDMS using photolithography and microfabrication, and then specifically with these multi layer devices, by using pressure to actuate certain channels, these control channels, which here are illustrated in orange, those channels expand. And because PDMS is flexible, the membrane between multiple layers of channels deflects, and we can make microfluidic valves. So this is a microfluidic pump. And with this pump, we can sort single cells into trapping chambers where we can image those cells and then send them to other chambers in the chip for for Molecular Biology reactions such as lysis RNA, reverse transcription of RNA and an amplification. So this is an example of a channel. So this is the device that we use. And each one of these channels that we

sort the cells into, looks like this, and so the blue shapes are valves we can open and close that will sequester one chamber from the next, and the pink shapes are the chambers that we can put the cells in with reagents. So once we trap and image the cell, we can push it into this next channel where there's lysis buffer, which lyses the cell. This is my microphone. And then we can do revert form revert. We can add in the molecular components to do reverse transcription, to turn the RNA into single stranded cDNA, amplify that DNA, then take it out of the chip and sequence it. And we can do this in parallel. So it allows us to parallelize this reaction so that a single grad student doesn't have to do that measurement for lots of different cells. And so in this way, we can take a Raman image of an adipocyte and then sequence it and know and really that there's lots of ways of imaging cells, and there's lots of ways of sequencing cells, but

this is the way that we do both. Because the hard thing about single cells that they're so small it's hard to track. Okay? Image the cell. Now I want to sequence that same cell. Micro fluidics provides a really nice and efficient platform to keep track of the cells. It's almost simple in that sense. So these are mouse 3t 3l one cells. These are fibroblasts that can be differentiated into adipocytes, and these are all supposed to be functionally similar to white adipose tissue, but as you can see, all of these cells have very different lipid droplet morphology. Some of them have one big lipid droplet. Some of them have many little lipid droplets. And so we're really interested in trying to figure out what drives this variability. Is it physics like? Is it just randomly? Like, if you take oil and water and shake it up, you'll get a bunch of emulsions, you know? And the size distribution of those

droplets evolve over time, and it kind of somewhat stochastic, and they follow certain principles. And maybe this variability is just coming from taking snapshots of different cells at different times of the ripening of these lipids. Maybe this is uninteresting, or maybe interesting, but but passive, or maybe there are genes that are being expressed variably in these cells that are somehow differentially metabolizing lipids and producing different shapes of lipid droplets. What's maybe it's a little bit of both. In fact, it probably, most definitely is. And so to address this, we wanted to see if we could sequence RNA from each one of these cells, quantify the lipid droplet profile, quantify the gene expression, and then see if there's any similarities. And so this is just a snippet of what we can do that. We're still mining this data for interesting biology, but we can take the images and and really to do

single cell analysis at the image level, you still need to extract data. So we extracted lipid droplet, size, number, the size distribution, size of the cell, and we have like, a 20 dimensional space of morphological traits. And then we extracted 20,000 gene expression measurements from each cell. And what I'm showing here is just like a two dimensional principal component of the morphological space and the gene expression space. And I'm just, I just selected images of cells that lie in sort of two of the extreme groups that we found, sort of by hand. These are this. There's some cells that have few large lipid droplets, and some cells that have many little lipid droplets, and you can see that they sort of span both the morphological space and the gene expression space. And so we can use this analysis to ask what genes are correlating to this morphological difference. It's very noisy, because unlike a lot of single cell analysis, where

you're comparing. In different cell types, we're comparing differences in gene expression within a cell type, but we have been able to extract differential expression of genes that are known to mediate lipid droplet metabolism. So one of the genes that's that's down regulated in these cells is a gene is that encodes a protein that's found on the surface of lipid droplets that helps mediate coalescence of lipids. And it's been shown that when you knock out this gene, these kind of cells can't transition into these kind of cells. And so this is just, there's no knockout here, but this is just in a natural cell culture environment. The cells that have this morphology also have a lower expression of that particular gene, which is really interesting. So there's a lot of sort of hardware feats that we had to solve, one of which is that simulated Raman scattering is for we do forward scattering.

It requires a very high numerical aperture objective to deliver the incident light and then a pretty high numerical aperture condenser to collect as much light as possible for the high sensitivity measurement. So that means you get need to get really close to the sample on both sides, which is hard to do with microfluidics because they're these thick chips. And so a postdoc in the lab came up with a way of making these paper thin microfluidic devices that are multi have multiple layers. It's quite straightforward to make a very thin network of channels that's a single layer, but to have these pneumatically actuatable valve based channels, integrated channels that we have, it's just challenging, just from a fabrication standpoint, to make these and so we had to come up with a new fabrication protocol that would give us these very thin integrated fluidic circuits so we can do the realm on imaging.

I don't have a clock here. What time is it? I just want to make sure so I have 20 more minutes. Okay. Ah, perfect. Thank you. Now I'm staring at the clock. Another thing is that we did was that when we're able to process these cells in parallel, because all the cells in a chip go through the same reactions. And this first demonstration only had 10 was only had a throughput of 10. We only did 10 cells at the same time. So we've been increasing the density of these chips. We could do more cells at the same time. But then when we took the cells off, that when we took the DNA that we amplified from the cells off the chip, we still had to prepare that DNA for sequencing for each cell, and that's very laborious, because a student had, you know, granted, we cut the experiment in half, basically, but as soon so to take each tube and prepare DNA sequencing library, where you add adapters and sequencing barcodes so that the sequencer

can can sequence the DNA, and that takes a lot of time. So we also developed a fabrication protocol where we could print unique DNA sequences into the lysis chamber of the chip, so that when the cell was so that when reverse transcription happened, the each cell would get a unique DNA barcode appended onto the end of its cDNA, and we would know we could deterministically print those barcodes into the chip during fabrication, so that we knew that channel one had barcode one, et cetera, so that we could pool all of the cDNA from the cells together and make one sequencing library And then sequence that and then decode which read came from which cell based on that DNA barcode. So that was a helps us increase the throughput, so we can kind of pool, so we print using them, piezo acoustic dispenser that can dispense, like picoliter droplets. We dispense those into the channels before we finish

the fabrication that helps us increase the throughput. And these kind of barcoding strategies have been really important in high throughput genomic studies of single cells, because when you start to get to 1000 or 100,000 of cells, you have to do it in batch, you know, and so, but you need some way of getting these barcodes onto the single cell at some point in the reaction. And so microfluidics has been a very important way of combining cells with barcodes, even for non imaging experiments. If you have any questions, please feel free to interrupt. I know that we have questions at the end, but I'm going to shift gears a little bit, and so if there's anything that comes to mind, I don't mind fielding it in real time. Okay, so that is an example of how we image at just one Raman shift. If, and we've been, we've been really interested in looking at lipids, and so imaging here at the CH two shoulder of this spectrum is relatively powerful, but there's

a lot of other molecules here. And in fact, even if you just look at the CH region of the Raman spectrum you have, you have a lot of contribution from CH three symmetric stretching, which is abundant in proteins. So there can be a lot of crosstalk between the protein signal and the lipid signal. And also there's a lot of information in the relative abundance of these transitions in a cell, not. Just that there's crosstalk, but you might want all of this information. And so how can we, how can we make a spectral image that's not just a single color, but where each pixel has a spectrum, has a Raman spectrum? Because then we could have a sort of an image stack where x, y coordinate, and then the third coordinate could be spectra, spectral information that could be a very information rich so the way that we do that, this is a more granular description of our microscope. We have a optical parametric

oscillator. This is great because you can just dial in. It has two outputs, a fundamental and then attunable pump. And you can just, with the computer, tune in which which wavelength you want for the pump, and it's pretty fast. It takes a few seconds but then actually, sometimes it takes 10 seconds, and it has this feedback. And so what we've done is a lot of people will use a picosecond pulse laser because it has a very narrow bandwidth, a femtosecond pulse laser, which is also useful for Raman. The drawback is that it has a very broad spectral bandwidth because it has a shorter temporal pulse. But what we do is we take advantage of that broad spectral bandwidth in the femtosecond laser, and we put each the pump and the Stokes through glass, and we chirp those pulses so that we can disperse the wavelength in time. And so now, instead of having a broad spectral pulse that arrives at the sample, you know, a lot of

the the optics on the table is really just synchronizing, is combining the beams in space and then synchronizing them in time, so that the pulse have to arrive at the sample at the same time. But by chirping the pulses, we can then tune the time delay between the pulses to excite different Raman shifts within the within a single pulse or within multiple pulses. So for example, as we tune this retro reflector right here, we can change the temporal difference between this pump and the Stokes. That'll give us, sort of, you know, different Raman energies based on the shift between the pump and the Stokes.

So this allows us to more rapidly access spectral information while we're scanning the laser in space. So we can, we can scan sort of inspector in space, sort of simultaneously, or at least in sequence, really fast. So allows us to take images, sort of 60x magnification images on the order of a few seconds. If we wanted to take a big image, we could do that in hours. It still takes a long time to take a large image. And so what this allows us to do is that for every This is a collection of a few adipo sites, but we can make a spectral stack, where for every pixel, we have a Raman spectrum. And this is constrained to the CH region for these images. And what I'm showing here, this is a scan through the spectral stack. And I'm showing you here is that the different Raman spectra for different regions of these cells. So looking at these images, we know that these are the lipid droplets because we were working with fat cells. But

let's say you didn't know what the organelles of the that you were looking at were. You can now cluster pixels by their spectrum. And we've and sort of in an unbiased way, pull out chemically distinct regions of the cell. So this, this image here, this three color image is just saying, let's cluster pixels in the three clusters based on their spectrum. We don't know what they are, and then looking at you can tell, oh, this cluster is the lipid droplets. The green is the cytoplasm, and the blue is the nucleus, because it has different combinations of lipids and proteins. There's not much lipid in the nucleus, there's a lot of proteins in the cytoplasm. There's more lipid in the lipid droplets. So these little, unique spectral signatures give us very rich information about the cell, even if we don't know what we're looking for necessarily, what we can also do is, given this information, we can say, well, you know, because this here, I

think we took 40 images across this Raman shift, but most of the information in This part of the spectrum is contained in two wave numbers. And so we can reduce this to like 95% of the information with just two images, and we can do two color imaging that gives us really good contrast for lipid and protein generically. And so with that, I like to transition into how we've been using this imaging to look at whole tissue, so cells in situ. And I'm going to move away from the lab on chip part, because here we're just going to image tissues not on a chip. Any questions about the microfluidics part?

Aaron: Yeah, in the back. Audience member: Thank you. Yeah, thanks. I'm interested in the part where you mentioned printing DNA into the chip. And I'm just wondering, like, physically, what that means, like, is it embedded in the PDMS? Is it attached to the Unknown: Yeah, yeah. What we So the way we've done it for that is that we use a machine. Called the cyaneon selenin, which is a piezo, acoustic droplet dispenser that can it's like a robot, but it just has really small, high precision droplet dispensing so it can pick up some DNA and drop it. And when we fabricate these devices, we fabricate the PDMS part, and then we bond it to a glass substrate using, usually using oxygen plasma, but sometimes just using heat bonding, or even PDMS PDMS bonding with a PDMS layer in the glass. So before we close the chip, we print into the channel, and then we just let them dry. So these are these particular barcodes get

resuspended when the liquid goes in, and then they're in solution, so that we can, but in some cases, you might want to immobilize those barcodes to the glass. And there's what. There's chemistry you can get for that too. Thanks. Did you have one? Okay okay so Aaron: yeah, please, please, Another audience member: yeah, just nicely speaking. Is there a push to go to higher and higher throughput for whether you're studying single molecules or single cells. Yeah. Unknown: I mean, when I talk to so, I have a colleague in EECS who does a lot of convolutional neural network develop, architecture development for analyzing this kind of large, large, large dimensional data. And his answer to your question is yes, more please. Thank you. Like I think it's you know, most people say more is more. More cells is better. And I think when you know, I used to think that more data on more information about a cell could be more informative. And I think

there's certain questions where that's true, but a lot of the work in machine learning in this space lately has shown that less more shallow, sparse data on more cells actually paints a better picture of the variety in the system than fewer samples with higher depth. So that's kind of interesting. So you know, but I think it depends on the question. So in the in the context of adipose tissue, for example, the function of those adipose sites, especially in diabetes and insulin resistance, has a lot to do with the cells that are nearby, the immune cells, the macrophages and other cells that might be releasing cytokines or other molecules that influence the fate or the function of those cells. And so while single cell analysis is very powerful, you lose a lot of information when you take the cells out of their context, and a lot of times just a structural arrangement of an organ might tell you a lot about the disease

state, as opposed to just the composition of those cells on a list. And so the question, obviously, for a century, people have been using microscopes, so therefore image not just cells, but tissue. Cells, but tissue histology, cancer. It's one of the main ways to diagnose cancer, to take a slice of the tissue and look at it under a microscope and see what parts of the tissue have been disrupted. And so, you know, we're really interested in using Raman to look at tissue sections. And we've been asking, how can we do this same kind of analysis, where we get the RNA Seq information and the Raman information from the same tissue? And so this is the microscope again, and what I'm just gonna show you here is that we can, you know, one of the graduate students the lab has really automated this system very well so that we can take we can tile with a high resolution, 60x objective. We can tile across a whole Mount Tissue on a slide and get and stitch it

together. There's a lot of really beautiful algorithms these days for stitching together tiled images, and then while imaging, he can tune the retro reflector so that we can take as many images as we want, but really two based on those two peaks to give us the most information, the CH spectrum. So this is a I'm gonna show you images of spleen. We're gonna switch biological systems into the mouse the murine spleen, lymphatic tissue. This is another part of our lab which is really interested in looking at lymphatic tissues and how immune cells develop in thymus and spleen and lymph nodes. So we can take an image of a spleen. This is a single field of view with the 60x microscope at one wave number. And then we can automatically tune the retro reflector to hit a different Raman shift to take it another wave number. We knew this very quickly, or relatively quickly. Now, when we make when we

interpret these images, there's a little bit of processing because there is overlap in the spectrum between lipid and protein, but if you know what those spectra looks like, you can do some it's very easy linear deconvolution to get images that represent primarily lipid and primarily protein. So we do some processing, then we can make these really beautiful. Positive images that show us, you know, the protein localization and lipid localization in this tissue, and we can do that over the automatically, over the course of a whole spleen section. So this is like a 15 micron slice of a mouse spleen, and this is a two color Raman image of that spleen, and you can see all these shapes. And if you've never seen a spleen before, this mean anything other than being a nice microscope image. So the spleen is one of our important immune tissues, and I'll just do a brief, really quick it's part of our innate immune system and our adaptive immune system. So

it's sort of a blood filter, so blood runs through the spleen, the spleen catches a lot of old red blood cells and and filters out the non functional red blood cells in a part of the spleen, it's called the red pulp. And then there's these modules, or these lobes of what's called the white pulp, embedded in the spleen, which are full of immune cells. And that's where the adaptive immune system kicks in. And if there's antigens that are from, you know, virus or bacteria or some invasion that are presented, then those immune cells in the white pulp light up, because the blood is one of the places where we can survey all infection. And so blood that has those antigens from from the those foreign bodies is filtered through the white pulp, and the immune cells can respond there. So you have a lot of B cells and T cells in the white pulp, and they light up in the in the context of infection. And so we wanted to combine Raman imaging

with gene expression analysis of these tissue sections. And there's a whole field of technology development in the context of of RNA in situ, RNA sequencing or spatially resolved transcriptomics and quick crash course, the idea is, can we do RNA sequencing and retain the location of where the RNA came from in a tissue, instead of pulling the cells out and sequencing them in a tube. And there's two broad classes of approaches for this, sequencing based and imaging based. The way imaging based works is it's essentially a fluorescence microscope that utilizes fluorescent in situ hybridization and just does a lot of sequential rounds of labeling and quenching to literally read out as Ts, Cs and Gs, and sequence in the spatial context of a tissue. So it's basically like a sequencing machine, which is a microscope where you put a tissue section in to be sequenced, instead of just the DNA molecules. And people like Cush, you, who, I

think, gave a talk in this cloak and a couple months ago, that doesn't do this exactly, but he has the tools that could allow you to do sequential, single molecule detection of RNA. It's very hard. You need really good, automated microscope, but it's very powerful and has very high resolution because it's up diffraction limited imaging. Another approach is called sequencing based and this is what we've been using in this context. And the way that sequencing base works is that you take a tissue and you label the RNA in the tissue with a spatially encoded DNA barcode. So, like any RNA that's in d1 would have a unique DNA barcode ligated onto it, and you do that. And so you do that such a way that you know which barcode is in which place. Basically, I need to develop an array of DNA sequences, and then you basically, you know, put that tissue on that array, dissolve the tissue, and let the RNA

attach to those barcodes. And this requires a company to make this array for you, but then you can sequence all those barcodes and then reconstruct the position of those RNA molecules computationally, right? And so what we've been doing is we've been imaging our spleen section and then printing it on transferring it on this array using the visium site assist, and then sequencing. And really, all of this, all the sequencing is done by this piece of equipment that we buy. And so we so the technology that we've developed is just a way of transferring our section from the Raman slide to this array, but it produces this beautiful data. So now for every pixel we have a Raman measurement, we also have a measurement of 20,000 genes, and this is what happens when you cluster those pixels based on their gene expression. You see red pulp and white pulp. It's like almost seems like it's more information

here, but when you infect the mouse with toxo Gandhi, the spleen lights up because now it's responding metabolically to this infection. And if you do. RNA Seq on that spleen, you see all these new clusters that correspond to T cells and B cell activation and macrophages. And so now the question is, okay, what can we learn by having this sort of chemical map and this gene expression map, and I'll just kind of race through this, maybe the purpose of this is more that we can ask a lot of questions instead of and I don't have many of the answers for you, but one of the things that we're interested in is, well, can we use the Raman information to ask questions about how genes are being differentially regulated within this spleen? And so what I'm showing you here is just the colors correspond to different clusters of cell types based off a gene expression. So you have a lot of these immune response

signatures in this infected spleen. So you can zoom in and you can see that these modules that have been that we can identify with the Raman, see that it's a bit dark, these are oftentimes recognized as megakaryocytes and histology images, and when we overlay the gene expression measurement, we see genes that are associated with platelets and the genes that megakaryocytes Express. And so this sort of confirms that we have some concordance of information. So now what we've been doing is really extracting information from the Raman image to say, Okay, well, if we can identify a quantitative change in the ratio of lipid to protein based on the Raman image, that might give us some signature that we can key in on to ask what's happening differentially on the gene expression phase.

And what we found is that this lipid to protein ratio measurement gives us a very strong signal for where the red blood cells are. The red blood cells have a unique signature in this type of analysis. And so we can segregate the Raman image by the red pulp and white pulp, and then ask we're getting Okay, they're there. We can kind of then separate out the white pulp into regions based on the relative amount of red blood cells, and then perform different so now we have the so I'm just going to jump through this now. We can say, here's the white pulp, and we're separated into red blood cell enriched white pulp and red blood cell depleted white pulp. Then we can ask what genes are being differentially expressed in those two regions of the white pulp. And we find this is just all within the past couple of months. We find some really strong signatures for immune response. And so we're really

excited about taking this kind of data, and exploring what we can learn about how the mouse spleen responds to a generic and acute infection model using this really multi dimensional data, another thing we can do is we can use a Raman to segment out single cells and then map the gene expression measurements back to those single cells. And this gives us the power to combine the single cell measurements with this sort of spatially resolved measurements, all kind of that's all to be continued for my next seminar. So in summary, we do a lot of imaging and sequencing, and what I didn't talk about was some of the single molecule sequencing techniques that we use, or the high throughput sorting that we've started to do using droplet microfluidics, which is another exciting direction that the lab is going. So thank you very much, and I'm happy to take any questions here or back there, depending on what's

happening next in this room, I actually say Bryce and Leon took all of those beautiful pictures. And of course, I want to thank my funders and my collaborators. Jan: Yeah, thank you very much. So we have time for a few questions. Also, if you're on Zoom, please type your question or unmute yourself and just speak up. We have questions from zoom, but any in the room to start with, Aaron: somebody raised their hand, Jan: yeah, please go ahead. Aaron: Yes, I can hear you. Azriel Goldschmidt: You can that's good, impressive range of stuff. I could not follow a lot of it, but a lot of it sounded pretty exciting. In, in, in your slide, 21 you had a video in the middle.

Aaron: We'll see if I can get back to it. Unknown: Yeah. In, in where you were showing, kind of the lipids are identified there. I realize not all this. Oops. I can't get here. I'll do this quicker. All right side, 21 Yep, the video of the cell spinning around the one we had three beat three things and in the middle, that's right. And I was wondering. So it seems something seemed to happen in the in the one in the middle, yeah, where something seemed to have collapsed. And, yes, this is over six hours. This is sped up, and that cell detaches. It's either getting ready to divide or dying. That's. That you're going to different one of two very important Aaron: perturbations, Unknown: but it's not something you are doing is just passive watching in this sense, yeah.

This is just Yeah. This is just classic watching cells. That's right. So we caught an interesting event. These cells are being incubated with oleic acid, but that's not doing that's not perturbing their division or their death. Too much, Azriel Goldschmidt: right? Cool. Thanks. Yeah, no Aaron: problem. Thank you for the question.

Maurice: So this so these cells are, you know, I can see the size right there. They're pretty big. What is the smallest item that you could manipulate in these microfluidic channels if you wanted to deal with, you know, single nanoparticles, for example. Unknown: Great question with the way that we've been sorting these cells in this experiment, was that we look at them under a microscope and use this peristaltic pump, which has resolution down to nanometers, to bring a cell into this trapping region, and then we have some machine vision that sees that there's a cell there and traps it. But for smaller particles, it's hard that they're hard to see and then hard to manipulate. One of the good things about microfluidics is because we're at a very low Reynolds number, everything's laminar and actually stokes flow regime, and so there's no so you can have a lot of control. But for smaller things like

bacterial cells or nanoparticles. One way that people manipulate those is by doing person plus on loading into large numbers of containers, usually droplets. So we can generate aqueous droplets in oil at a rate of 1000s to up to hundreds of 1000s per second. And so if you have a limiting dilution of your particle of interest. And you have, you know, one for every 10 droplet. You can get single particles into these droplets pretty reliably. Then you have to just figure out what your assay is and how to measure them. But that's kind of usually the way people approach the smaller particles.

Jan: I had one other we do one question online, yeah, Unknown: what are your thoughts for looking at smaller things than mammalian cells with the simulated, simulated ramen, micro fluids, bacteria and microbes? Do you think it is attainable, and are there issues with cell immobilization? Yes, yeah, that's we start getting in so we are diffraction limited in terms of optical resolution, but we use an infrared and near infrared photons, and so we have about, you know, hundreds of nanometers, sort of conservatively, a resolution. We can still see bacterial cells.

There's some really cool measurements of lipid accumulation in microbes or algae or engineered yeast, and that's an interesting direction for screening genes for productions of biofuels. We worked a little bit with Adam mark and on that, but the throughput became much higher than what we had the capacity for for screening his libraries. But this is an active area of research. You bring up a good point, because the imaging takes seconds to minutes, and so if things are moving around, they're going to be very blurry.

So immobilization would be an important concern, depending on the question. Jan: Thank you for that. One question from my side was you mentioned so you have your chips and you do the in situ imaging with your Raman. So how is the problem with, like, the optical transparency or quality of your chip material, not degrading your signal? Because I can imagine that Unknown: when you're doing the Raman imaging, the chip itself has to Yeah. So this is, so there's, yeah, that's a great point. So with this chip that we made, actually, I don't have the data here, but reducing the amount of PDMS between the objective and the condenser is critical, because scattering centers can significantly deplete a Raman signal, and also they can add background, like cross phase modulation, anything that could thermally respond to the femtosecond laser and shift even a little bit the refraction and PDMS can do that can give a

noise signal that's at the same 10 megahertz as the modulation frequency. And so making this thin chip was really critical. It wasn't as important for lipids, which have a very strong signal, but for some of the less abundant molecular populations is quite important. And having really clear glass, for example, Jan: your chip is still robust enough to do all the operations with the because you have some pressures applied to push things through your channels. Unknown: Yeah, yeah, yeah. Exactly We do liquid exchange and stuff like that. We have done a whole experiment where we looked at neurotransmitter. Or transport in neurons by taking deuterated dopamine, which has a unique realm on signature in that silent region. And that requires very well tuned imaging setup, because the signal is very subtle, and so we haven't been able to do that in PDMS. We have to use glass chips for that Jan: good if there's no further questions online or in the room,

then let's thank Aaron again for Thank you very much. Aaron: I appreciate it. Bye.

Questions about this episode

Who speaks in "New tools for precision multimodal measurement of single cells and single molecules"?

Aaron Streets (Guest), Dr. Shamani Jan (Host).

What is discussed in "New tools for precision multimodal measurement of single cells and single molecules"?

In order: 0:00 Aaron Streets, microfluidics, Dr. Shamani Jan; 10:00 RNA sequencing, microfluidics, Chan Zuckerberg Biohub, Human Cell Atlas; 20:00 Raman scattering, microfluidics, RNA sequencing, spleen; 30:00 lipid droplets, adipocytes, Raman scattering, PDMS; 40:00 microfluidics, lipid droplets, Raman scattering; 50:00 spleen, RNA sequencing, Visium; 1:00:00 microfluidics, spleen.

What is mentioned in this Berkeley Lab episode?

14 people, works and ideas are mentioned, including Aaron Streets, RNA sequencing, Raman scattering, microfluidics, lipid droplets, adipocytes. Each one on this page links to its own page with every episode that mentions it.

How long is "New tools for precision multimodal measurement of single cells and single molecules"?

The episode runs 1 h 11 min; the transcript is about 10,663 words in 70 timestamped sections.

More from Berkeley Lab

About this transcript: the episode's spoken words as captioned on YouTube, split into timestamped sections. Reclip.Me locates every person, work and idea in the text, keeps the time it is first named, and quotes the words around it verbatim. Last updated January 24, 2026.