Bench philosophy: Raman-Microspectroscopy
by Steven Buckingham, Labtimes 03/2015
Until recently, Raman-spectroscopy was mainly used by chemists and physicists to identify molecules in gasses, liquids and solids by the scattering of laser lights. Now, progress in lasers, detectors and software algorithms, brought this technique to the attention of life science researchers.
My regular readers will know how keen I am on my Next Big Thing in science: the rise of single-cell experiments. So, you can imagine my excitement when I heard of a technique that not only lets you analyse the chemical composition of individual living cells at sub-cell resolution but does so without any chemical or genetic markers.
Yes folks, it’s true − and it is called Raman microspectroscopy. It gives you a molecular fingerprint of a cell or sub-cellular compartment, based on what bonds are in there. It can even tell you what proteins (and other chemicals) are present in a cell, without damaging the cell and without having to label any proteins or stuff the cell full of stains or dyes.
So, how does it work? It’s all about photons − or, rather, what happens to photons when they collide with a molecule. Imagine a beam of light passing through a solution. When the photons in the beam collide with a molecule in the solution, there are three things that might happen to them: they can be reflected, absorbed or scattered. We won’t worry about reflection and absorption for now, it is the scattering that interests us here. In fact, nearly all of the scattering takes place in a specific way called “elastic scattering”, or “Rayleigh scattering”. This is a form of scattering, in which all the scattered photons have the same wavelength as the illuminating ones.
Raman microspectroscopy is moving into the spotlight for life scientists. But don’t panic! Not every Raman microspectroscope is as complex as the modulation-multiplexed stimulated Raman scattering microscope shown above, which has been developed by researchers at the Purdue University, USA, to monitor living cells. Photo: Vincent Walter/Purdue University
But that is not the only kind of scattering that takes place. About one in a million photons are scattered in a slightly different manner, called “inelastic scattering”. This was discovered by Chandrasekhara Raman in 1922, hence, it is also called Raman scattering. The important point is this: the scattered photons change their wavelength slightly, going slightly redder or bluer. Why does this happen? Because when they hit the cloud of electrons that shroud the bonds in a functional molecular group, they excite the electron and so lose some energy.
So, why is this important? The answer −and this is what makes Raman spectroscopy so exciting − is that the amount of energy lost depends on which atoms are in the functional group, the exact structure of the molecule and the functional group that the photon happened to bump into. The result: a “Raman shift” spectrum, in which each peak corresponds to a specific molecular vibration. Once you have that, all you need is a bracing dose of maths and physics, and you can use those peaks to work out what molecule the photons hit.
Oh, okay, I’ll own up. The real situation is a good deal more complicated than that. For a start, in real cells, of course, there is a bewilderingly complex soup of several thousands of chemical species. All the same, conceptually, the principle is straightforward: get a Raman spectrum and use some very hard sums to work out the mix of chemicals that would have produced it.
So much for Raman spectroscopy. Where does Raman microspectroscopy come in? This is the really clever part: attach a Raman spectrometer to an imaging microscope and voilà − you have a device that can get a Raman spectrum from pinpointed spots on any cell.
The technique of getting the spectrum is surprisingly easy. This is due to the steady, year-by-year improvements in several components: the lasers, good quality filters and high performance CCD cameras, for instance. But getting a signal you can interpret biologically is another matter. A good deal of care has to be taken to get the recording conditions just right. First, you have to reduce (or preferably eliminate) auto-fluorescence, so you may find yourself forking out on some expensive, specialised, glass cover-slips. You also have to take pains to reduce damage to the specimen resulting from the irradiating illumination. But there are tips and tricks to deal with these problems, too. For instance, you can use a confocal-type set-up with a narrow beam that reduces the effects of off-target illumination, while careful selection of the illuminating wavelength will take care of things like cell damage.
So, once you have got your spectrum, what can you do with it? You have to remember that, strictly speaking, Raman spectra don’t tell you what molecules are there, it tells you what molecular vibrations are there. In some cases, for example, where the chemical soup you are looking at doesn’t have too many ingredients, you can use that information to figure out, which major types of molecule are there. But in practice, the chemical soup inside cells is less like a plain onion broth and more like a full-blown chowder, which means you will have countless different molecular vibrations, all overlapping each other. A simple list of all the chemical species present in the sample? That may be a bridge too far.
But all is not lost. In fact, we can turn that very complexity to our advantage. That bustling, jostling crowd of overlapping peaks effectively gives us a high-grained fingerprint of the chemical composition of the cell at the sub-cellular level. Even without knowing the ingredients of the recipe, we can taste its flavour.
Statistical methods are the key to getting useful biological insights from the spectra. But what kind of statistics do you do on something as complicated as a Raman spectrum? I suspect we are talking more than t-tests here. One main technique that is a main stay of multivariate analysis of Raman data is Principal Components Analysis (PCA). This takes advantage of the fact that although spectra are complicated, there is actually quite a lot of redundancy in it. What PCA does, is to work out the main patterns of variation in a complex signal and use that to produce a reduced-dimensionality version. So a spectrum with, say 1,000 points can be re-organised and reduced to, say, 50 points (called Principal Components). The first two points of this reorganised data can be plotted as a scatter plot, which can be easily visualised.
Raman spectra deliver a molecular fingerprint of the cells. Photo: UC Davis Health System
How is Raman microspectrometry being put to biological use? Some labs have proved its power as an advanced, non-invasive, cell-profiling tool. For instance, Katja Schenke-Layland’s group at the Eberhard Karls University, Tübingen, Germany used Raman microspectroscopy to tackle the problem of identifying apoptosis and necrosis in cultured cells (Brauchle et al., Scientific Reports 2014 doi: 10.1038/srep04698). We know a lot about these two different cell-death pathways, but it can be tricky to observe them because standard biochemical tests are invasive, use up a lot of cells and can’t monitor events over time. Caspase assays, for example, can’t discriminate between late apoptotic and primary necrotic cells.
So Schenke-Layland et al. took Saos-2 cells and SW-1353 cells and induced apoptosis (by exposing them to room temperature) or necrosis (by heating them up). They then looked at the Raman spectra at different time points for the two conditions and found that they changed predictably over time, and in a way specific to the type of cell death. Could this be used to reliably identify, which cells were apoptosing and which were necrosing? Yes − when the spectra were used to teach a machine-learning algorithm (Support Vector Machines), the system was able to spot whether cells were apoptotic or necrotic with over 90% accuracy.
Similarly, Hideaki Fujita's group at the Quantitative Biology Center, Riken, Osaka, Japan used Raman microspectroscopy to track differentiation in neuronal and adipocyte cell lines (Ichimura et al. PLOS ONE 2014 doi:10.1371/journal.pone.0084478). The PCA plots of the spectra showed obvious marked differences before and after differentiation. Because the spectra were taken from the nuclei, the authors hinted that the changes may reflect epigenetic events.
So what are the drawbacks with the technique? Although the physical set-up for performing Raman microspectrometry is not complicated and getting spectra is fairly straightforward, the method does suffer from poor signal-to-noise ratios. To solve this, there are ways of souping up plain old vanilla Raman. One method uses a second, synchronised laser beam applied along with the illuminating beam but at a different frequency. What does that do? If the difference in frequency between the two beams matches the molecular vibration, a resonance effect amplifies the vibration and, therefore, amplifies the Raman effect. This is called “Coherent Raman Imaging”.
Another hindrance to the adoption of the technique is the reluctance by biologists to take up a methodology based on physical principles they find hard to understand. If you are to be confident interpreting your data, you need to have a fairly in-depth understanding of the physical principles underlying the technique, and Raman-Stokes shifts don’t come easy to your average biologist. But the software that makes interpreting spectra simple and practical enough to confidently interpret the data just isn’t there.
Until the day comes when you can get software off-the-shelf that puts the power of Raman into the hands of non-technical biologists, the technique is unlikely to escape its niche.
Last Changed: 25.05.2015