Scientist Stories: Michelle Digman, In vivo Optical Imaging

First of all, I would like to thank Martin and all the organizers for inviting me to share the work that we've been doing in the lab. And I would like to just emphasize that I'm going to show you two different applications in imaging and fluorescence in terms of human disease. So the goal here is to use non-invasive tissue fluorescence imaging techniques due to the fact that cells in themselves have intrinsic fluorescence. Fluorescent signals. And if we use different approaches, let's say if we're using an excitation of blue light, we're able to actually visualize the structure within the tissue stratum. So within the tissue, there are natural fluorescence properties or fluorophores, and they consist of specific types of fluorophores. So here you can see collagen, elastin, keratin, and ADH. FAD, porphyrins, and fibrin. We have two approaches to actually visualize or illuminate these fluorophores. We can use either one-photon excitation imaging or two-photon excitation. In all of our applications I'm going to show you today, we're using two-photon excitation because it offers the capability to actually look, probe deeper inside tissues. Not just that, but it's also the type of laser that is used is a pulse laser, and so we are able to generate a very small focal volume, as you could see here, without actually illuminating above or below the focal point, which actually allows us to go deeper in tissues. And so for two-photon excitation, we generate a high density of photons at the focal spot, and if we have two coincident photons, they can excite the electrons for us to be able to excite the intrinsic fluorophores inside the tissue. Our approach is to use fluorescence lifetime imaging microscopy, and the advantage of this as opposed to, let's say, a spectral emission imaging microscopy is the capability to identify unique signatures within the tissue. So the fluorescence lifetime is the average time that a molecule spends in the excited state before emitting the photon. And how we do this is that we— our approach is to use serial detectors in a photon counting mode. And we programmed a programmable gate array, or FPGA chip. And as you can see here, we actually have 4 sampling windows. And we collect the time arrival of the photons within 20-nanosecond time scales. That's allowing us to count each photon at a time as we're imaging through, let's say, the tissue. And we have a total of 56 slides or steps within the 4 windows, and that allows us for high-quality and statistical analysis of the lifetime. So our actual contribution to this area is not FLIM, but it's the contribution of actually graphically representing this data. Typically what researchers do is they obtain this lifetime decay curve, and this has to be fitted on using single exponential or multi-exponential decays. However, with the polar coordinate system, which is what you see here, it allows us to just select all the pixels in our image and plot them in this polar coordinate plot, making the data much simpler and easier to analyze. So in our examples, you're going to see how we've calculated a fraction of contribution of the highest concentrated fluorescent fluorophores in tissues, maybe you've gone to image and done some imaging on tissue and you've noticed this very high background signal. And that's the contribution of NADH. And so we can calculate the fractional contribution of free and bound NADH, and with using this polar coordinate, you could see the actual signature of each different type of fluorophore within cells. We've also developed a metabolic trajectory that allows us to calculate whether cells or tissues are undergoing, let's say, shifting glycolysis or oxidative phosphorylation. And this is due to the fact that when NADH in the cells is free, the conformation of this fluorophore is actually in a bound state. It's in a closed state. So the fluorescence lifetime of this particular molecule is actually pretty short. As you can see here, the lifetime is about 0.4 nanoseconds., and if you add LDH or a binding protein, this conformation opens up and the lifetime becomes longer, and here you could see the lifetime extends to 3.4 nanoseconds. And the rule of this polar coordinate system is just the sum of vectors, and so if you draw a line between these two possible combinations that could be in your cells, you'll be calculating the fraction of free or bound NADH inside cells. So here I'm just using a pseudo-color to isolate spatially the signature of each one of these components. So what's important in metabolism is the fact that as cells undergo a normal metabolism, they're actually using mitochondrial respiration, and this is an NADH-consuming step. And of course, the cells can also undergo glycolysis, and depending on what's fate— or step they're in, their fraction of free and bound NADH will shift. So in normal tissues, cells break down glucose into pyruvate, and this process actually lowers the NADH, free to bound NADH states. And when cells are differentiating or undergoing high glycolysis, there's a higher population of free NADH compared to the bound NADH state. So how do we know this, and what was our signature trademark to look at the shifts in metabolism and to spatially map them inside cells was to use these particular inhibitors. We've— we used potassium cyanide as just a simple explanation for this, where it inhibits the electron transport chain. So before adding the potassium cyanide, you could see oxidative phosphorylation or high free— high bound NADH within the cells. And upon adding potassium cyanide, you see a shift in the trajectory of this polar coordinate system, which shows a higher fraction of free NADH. You can revert the process of, let's say, a high free NADH to a more bound state using H2O2. And you could see the differences in cells that are cultured in high and low glucose media. And here you could see the out— uptake of glucose is much higher, and higher free NADH component is found. And finally, you can take tumor cells and add a metabolic targeting inhibitor. This is dichloroacetate, and you could see that we've shifted now from a glycolytic state to a more oxfos state, which was the purpose of this particular drug targeting. For the first The first application I'm going to show you actually an application using a model of Huntington's disease. And as you all know, Huntington's disease is an autoneurodegenerative disease. It's characterized by emotional, cognitive, and psychiatric disability, as well as we're using— in our example, we're using the exon 1 polyQ expanded form of the protein. So, what's important about this particular application was to look at mitochondrial respiration within cells. So, mitochondrial dysfunction has been linked to a range of different diseases, including neurodegeneration, and early detection, of course, in mitochondria can— anomalies can be an effective type of diagnosis or an effective way to diagnose early detection and get treatment. And there, of course, there's a variety of different approaches to look at metabolism within Huntington's disease. You can use, it's very profound energetic impairment. You can use fMRI studies or PET, which reveals a marked reduction in glucose utilization in the striatum. But energy failure at the end leads to a significant role in dysfunction of the respiration system. So there is a current limitation in this area of research because glucose utilization is acquired for brain function, and how glucose is processed is not really clear. So we used actually the phasor fluorescence lifetime imaging, which I just explained to you how that works, to look at the disease model in Huntington's disease. And we used two different approaches. One is an in vitro model for human embryonic kidney cells, and the other And there's a transgenic Drosophila expanded and unexpanded polyQ HTT1X1. So the first thing to do was to express the Huntington protein. Here we have just the GFP alone. Here is the Huntington, just the normal wild-type Huntington protein, which has a length of 25 repeats. And here we have the 97 polyQ repeat. And just to show you that we can multiplex, in other words, we can actually label the cells with the green fluorescent protein will not hinder our NADH signal. We're using two-photon excitation at a wavelength of 740 nanometers, and we're looking at the emission of blue light. And what you can see here is the intensity of the NADH signal. In principle, we don't need to label these cells with GFPs, just to note where the actual Huntington protein is being expressed, and it has formed an inclusion. And with our trajectory, what we see is that— based from the wild-type or normal cells, we actually see a shift in the amount of free NADH as this expression is increased. We can also do image segmentation within the cells, and that allows us to just segment just the nuclear compartment or the cytoplasmic compartment or to segment different regions, let's say mitochondria, if you want to see specifically if the mitochondria has shifted in free and bound NADH. In this quantitative analysis, and here's the phaser analysis showing the normal unexpanded form, wild type, has a higher bound NADH fraction. It's undergoing oxidative phosphorylation, and once the cells are expressing the 97 expanded form in the cytoplasmic pool, we see a shift to more free NADH. Also, the nucleus shifts., and in fact, the nucleus is already higher free NADH in the wild type, and it shifts to a higher fraction of free NADH in the 97q. And what is happening here is that it's already showing that the nuclear compartment is showing a reduced nuclear activity characterized by this shorter lifetime, and I don't have time to go into how we know that, I'm very happy to describe that to you. So finally, we used an animal model for the Huntington expression. In this case, this is the transgenic Drosophila fly, which is expressing the poly-Huntington in the eye disc. And so the eye disc was isolated from the larva at the late 3 instar. And of course, what drives this expression level is that ELAV and the GMR, which is activated at the front of these morphogenic— what you see, the morphogenic furrow, and it's a temporal expression. Thereby, when we express the expanded form of the protein, you could see as a function of days that the toxicity of the polyQ repeat. So this is a summary of the wild-type versus the expanded form, and you could clearly see that the normal Drosophila iDISC is more bound NADH, and in the expanded form, which is the poly120Q, is the 120Q, and we also use a 96-expanded CAG repeat model, shifts the metabolism to a more free NADH, indicative of glycolysis. So just to summarize the first part of this example is that the phasor FLIM approach actually allowed us to quantitatively monitor the change in glycolysis to oxidative phosphorylation to glycolysis in cell cultures as well as in the eye disc Drosophila model. And within the nuclear compartment, we noticed that XOM1 is expressing— in expressing cells increased the levels of free NADH that influences both CTB BP and sirtuin activity in the nucleus. And that's been shown already by other researchers, but it was important to show the nuclear impact of this form. So finally, I would like to show you an example, some work that we're doing in the lab in regards to also human-associated diseases. And then this is looking at the heterogeneity as a driver of cancer drug resistance. And so one of the things that we're interested in is cancer stem cell hypothesis, and what you see here is that recent data has suggested the progression of these brain tumors is driven by a small subpopulation of tumor cells, which are called the cancer stem cells because of their capacity to renew. And glioblastoma in particularly have a therapeutic resistance due to tumor heterogeneity. And they're very difficult to target. So these cells require staining in order for— or surface marker staining. And this is a limitation because the subpopulations of these glioblastoma cells are not easily detected unless you do staining or you break them up. So one of the approaches here is to culture— to take the primary cells from tumor sample and then culture them in media. And we use two different types of culture, the adherent media condition and the neurosphere condition for the tumor cells. And we have two different subclone populations.. And here we are showing the adherent tumor cell population. These are the tumor mast cells on this side. And here we have the neuronal spheroid population that's cultured in neuronal spheroids, and these are the stem cell initiating tumor cells. And what we see is that they have a very unique lifetime signature. Essentially, the— stem cell-like cells actually have a higher bound NADH, and the cancer mass cells have a lower fraction of bound NADH. And similar higher fractions of bound NADH was observed with enriched 51A relative to 51B. So it's not our cell culturing conditions that is changing the metabolism within these cells. So these cell populations were mixed in a 1-to-9 ratio, and they were inserted into the cranial xenograft model in a zebrafish— I'm sorry, inside a mouse. And what we see here is the population expressing the green fluorescent protein that have the U251 cells versus the neuronal spheroids, which are in red. And we have a spatial heterogeneity. Essentially, we map exactly where these two— these types of tumor cells are within the tumor tissue. And again, we see that the neuronal spheroid has a higher bound NADH versus the tumor mast cells have a lower bound NADH. So the theory here is that transitioning of tumor populations happens within subpopulations. Cultures, and the majority of the tumor— so what happens is that these can actually give rise to tumor mass-producing cells. And of course they can divide into similar populations, and then they could give rise either to tumor-producing cells, and these can actually revert back to the stem-like tumor-promoting initiating cells.. And so what we were interested is, as a function of passage, as a function of time, we'd like to monitor the change of stem-like tumor-initiating cells to tumor mass cells. So as you can see, as a function of passage number, these stem-cell-like populations will transition to the tumor mass populations, and we wanted to monitor the changes of metabolic respiration within these cells.. And what you see here is a function of passage number, the fraction of bound NADH actually had lowered. So you could see that the free NADH actually has increased of these populations. We can actually monitor to see if the metabolic changes inferred with our FLIM imaging is in correlation with other types of techniques, including the Agilent Seahorse metabolic analyzer. And here, where a able to quantify the glycolysis through the extracellular acidification rate, as well as the oxygen phosphorylation— oxidation phosphorylation is quantified through the oxphos consumption, or the oxygen consumption rate, or OCR. So how this assay works is essentially that we're measuring the spare respiratory capacity using metabolic inhibitors. And what we see with the seahorse assay is that we're measuring higher rates of this basal oxygen consumption in the tumor mass population. And this higher rate of oxygen consumption is— suggests oxidative phosphorylation in the tumor mass population, in contrast with FLIM, which reveals higher levels of bound NADH in the stem cell-like population. So that was kind of contradictory to what we are looking at with our FLIM data. So what we're concluding here is that measuring higher rates of basal oxidation consumption in tumor mass populations suggests that higher levels of these oxidative phosphorylation in tumor mass populations is not really explaining why we see these changes in our FLIM analysis. And here you can see the transition of the seahorse is exactly happening at passage number 4. And if we look at the OCAR, or the oxygen consumption rate, you also— the change in metabolic respiration happens transitioning at passage 4 conditions. So the changes in cellular respiration here increase respiration in the basal OCAR and the max ECAR transitioning here occurs between passage number 0 and 4, and where the free NADH change, what we see with our lifetime information, starts at around passage 8, so a little bit further shifted in time. So the question is, could this change be due to other types of interactions? And previous studies have shown that that there are differences in chromosome 7 copy number between subpopulations. And chromosome missegregation controls tumor heterogeneity, and this chromosome 7 copy number can be identified by FISH. So if you look at now as the passage number— as we increase the passage number, you could see that the chromosome 7 population in terms of segregation decreases. And the transition— so we looked at transition and growth rates, how dynamic the changes in population and equilibrium happens due to different types of factors, especially when they're transitioning to tumor mast cells. You can see that the increase in normalized growth rate has occurred after passage number 10. In also relationship to the expression of EGF receptor, we understood that the tumor mass populations have a higher overexpression of EGFR. And if we look now at these two different cell types, we could see that there's a change in EGFR expression which precedes changes of FLIM and NADH. So when we put all the data together, we could see that changes of EGFR expression increased between passage 2 and passage 7. The respiration rates have also increased between passage 0 and passage 4. And the population of free NADH increased at passage 8, passage number 8, which correlates with cell growth and correlates with the percent of tumor mass cell populations increase. So finally, we have developed a method to noninvasively look at function of subpopulations of cells. We can use this as a predictor to actually calculate the type of cell fate that these cells will become. And we've not only done this with glioblastoma cells, but we've also done this in the embryo to look at populations of cells and actually determine what cell fate these cells will have based on their autofluorescent signature. So imaging enables the analysis of the cells at the single-cell level. We're able to visualize the changes within these cell populations, either in a tumor mass or within cells, and in the future this method can be applied to monitor the enrichment of cancer cell populations for treatment. So finally, I'd like to acknowledge the folks that have worked on this. Dr. Larry Marsh, who's our collaborator in the Huntington field, and Dr. Adila Syed, who's at the Optical Biology Core and is doing Huntington-associated work, Dr. Yihong Zhu from the School of Medicine, and of course the Laboratory for Fluorescence Dynamics, who I'm actually— that's part of my lab that develops these imaging techniques to be able to quantify spatial information from images, and all the funding sources. And I'd like to thank you for your attention.