Lingyan Shi, PhD
Assistant Professor of BioengineeringCreating New Raman Scattering Microscopic Views into Cellular Dynamics
A new BioEng assistant professor interviews Lingyan Shi, PhD about her award-winning research program.
By Asst. Prof. Benjamin Smarr, Bioeengineering and the Halicioglu Data Science Institute with Asst. Prof. Lingyan Shi, Bioengineering.
Long ago, my first research experience was at UC San Diego in the lab of Prof. Mark Ellisman, working with what at the time was cutting edge microscopy techniques. I rejoined UC San Diego in 2020 as new faculty in the department of Bioengineering, and COVID-19 quickly robbed me of the chance to meet colleagues. I have moved from cellular mechanisms in the brain to the biological rhythms between brain and body. So it was with some nostalgia that I “sat down” (in Zoom) with Asst. Prof. Lingyan Shi, who is once again inventing the frontier of light microscopy at UC San Diego. She was recently inducted into the Society of Hellman Fellows, a prestigious award UC San Diego offers extraordinary young faculty. I talked with Lingyan about the work that earned her this award.
When we meet, Lingyan is warm and excited to engage. As soon as we’re through the normal pleasantries, she dives in.
Asst. Prof. Lingyan Shi Explains about Raman Scattering Microscopy (SRS)
L: Let’s talk about clustering! You cluster different physiological features in time. I cluster different dimensions of light. I bet together we could figure out some really good clustering.
B: That would be great. First, bring me up to speed about your new light microscopy.
L: Of course. Label-free hyperspectral imaging. I’m using a modified Raman scattering microscopy approach. We take ultra-high resolution images with a stimulated Raman scattering microscope (SRS), and this gives us the spectral profile of each pixel in the image. We then filter each pixel and compare the spectral profiles of the tissue to the profiles we generate on molecules of interest. This allows separation of excitation of different molecules at different energies, and we can then separate those molecular signatures into separate channels. For example, unsaturated fat-rich regions give off specific blue-shifted spectra, so we can isolate those, and create an image channel for the unsaturated lipids in the image. Deconvolution allows us to achieve super resolution for fine comparisons of intra-cellular compartments.
B: So you can make as many channels as you have molecular signatures. That’s very cool. Hence the need for good high-dimensional clustering. You want to classify each pixel by its molecular contents.
L: Yes, the pixels which contain a specific spectrum can color fields of view by the density of specific molecular peaks. We can have as many colors as molecules, so these images get very rich – beyond what we can really make sense of just as an image. So clustering lets us make images that more easily convey the information we seek.
For example, if you use an antibody to label a specific protein, we could compare antibody-labels to other information in each pixel of tissue, to find out to what structures it colocalizes.
B: Can you make movies with this technique to watch proteins getting trafficked? I remember how hard that used to be – you’d need special labels for all the compartments the protein might be going through, and we quickly ran out of channels.
L: Exactly. With a movie, we can even tell when the labeled molecules are getting into the golgi or the endoplasmic reticulum, or when it’s integrated into the cell membrane. We can also tell whether it is in multimers of protein, and whether those are homo- or heteromers. It’s very flexible.
Once the library is fully established, we don’t need antibodies really. For example, you have done some work looking at meal timing and how that affects circadian rhythms and health. To me that’s very exciting. A big part of lifestyle is dietary nutrition, and nutrition shows up as what gets into your cells. We could collaborate by labeling glucose in food, administer timed or unrestricted meals, and then image where that deuterium gets absorbed and how differently it is distributed based on the meal timing. We could compare that to other nutrients too and see how it affects what the cells metabolize.
B: How specific can these spectral profiles get? Is it just “lipids” vs “proteins”?
L: We can compare different subtypes of lipids. We have a project looking at the distribution of omega-3-22 (DHA) vs omega-3-24 fatty acids. It is very specific. And sometimes you get lucky: for example, Amyloid B has its own specific Raman profile and unique Raman peaks, so this approach can be used to identify presence of plaques in biopsies with targeted peak for in situ imaging.
Identifying specific spectrum is like extracting sound waves from noise to allow speech recognition. If we can profile a molecule of interest, like a viral capsid, you could train ML to recognize that profile from imaged tissue. It’s even quantitative, because the size of the Raman peaks is directly proportional to the number of molecules.
B: Wow. It sounds like this is great science, but also could have clinical impacts – faster, label-free biopsy analyses for instance. Knowing how much of the cellular environment changes with time of day, I’d love to finally meet in person and think up some collaborations.
L: I would love that. Let’s do it!