Biological molecules are often imaged by attaching fluorescent labels — but only a few label types can be used at a time. A method that could smash the record for the number of labels that can be used together is now reported. See Letter p.465
If we could walk around a cell system, observing protein synthesis, witnessing the dynamic choreography of regulation processes and identifying the plethora of transported molecules, then our understanding of cell biology would expand explosively. Instead, we must settle for experiments in which just a few biological molecules of interest (biomarkers) can be labelled and imaged at once. But on page 465, Wei et al.1 report a method that dramatically enhances our ability to distinguish between labels, thus increasing the number of labels that can be used together — ironically, by adapting a microscopy technique that was originally developed as a label-free imaging method.
Biological samples can be readily prepared with multiple labels, but imaging one label among many is challenging. Fluorescent labels known as fluorophores (Fig. 1a) each emit a broad range of colours, and these ranges often overlap — which makes it difficult to discriminate between more than four or five labels. Furthermore, multiple lasers (sequentially applied) might be necessary to excite each fluorophore at its characteristic absorption wavelength to induce fluorescence. Labels known as quantum dots were developed to resolve these issues (and others)2 by having broader absorption and narrower emission profiles than fluorophores, but so far their simultaneous usage has been limited to eight colours3.
An alternative approach for bioimaging does not use labelling, but instead detects the vibrational signatures of molecules. These signatures are composed of one or more vibrational resonances, with each resonant frequency determined by the number and type of atom involved, and by the vibrational mode (such as stretching or twisting). The signatures can be obtained from a molecule's infrared absorption spectrum, or from its Raman spectrum — which is generated through 'inelastic' scattering of incident light, and forms the basis of an imaging technique called Raman microscopy.
Raman microscopy provides spatial resolution akin to current fluorescence microscopy techniques, but the signals produced are extremely weak4 (typically requiring tens of milliseconds to seconds per spectrum to acquire sufficiently strong signals for detection in biological specimens). A variant of the technique called coherent Raman imaging (CRI)5,6 uses pulsed laser sources to actively drive molecular vibrations. This substantially enhances signal intensities, enabling video-rate imaging for concentrated biomolecules. In general, however, techniques based on Raman scattering (coherent or not) require6 high molecular concentrations of about 10 millimoles per litre (approximately 6 million target molecules per femtolitre; 1 fl is 10−15 litres). It has therefore not been possible to use these techniques to map molecules that occur at low concentrations, such as those found on cell surfaces.
Wei and colleagues attack the detection-limit problem for Raman imaging in an unexpected manner: by using fluorescent labels. Their approach uses stimulated Raman scattering5,6 (SRS; a CRI method that involves two laser sources, Fig. 1b) to detect the enhanced vibrational signatures of labels that have been excited in an 'electronically pre-resonant' way. When laser light is tuned to excite an electronic transition in a molecule, inelastic scattering from vibrational modes coupled to that transition can be enhanced by a factor of 108 (ref. 7). However, fluorescence (and other processes that affect light emission) from the electronically excited molecule can still overwhelm the enhanced Raman scattering. By tuning a laser to the electronic pre-resonance (EPR) frequency — a wavelength slightly off the electronic resonance — a substantial enhancement is obtained in Raman scattering, but with minimal fluorescence (Fig. 1c).
The authors therefore used bioimaging labels that could be simultaneously probed for their vibrational signatures through SRS and excited at their EPR frequency using the same laser sources. This enabled a detection limit of 250 nanomoles of a target per litre — equivalent to about 150 molecules within 1 fl. They also demonstrated that small frequency shifts of the incident laser pulses away from the EPR frequency could abolish the SRS signal (see Fig. 1c, d of the paper1). This demonstrates the unprecedented selectivity of the EPR–SRS approach — such a small frequency change would have only a minuscule effect on conventional fluorescence.
Wei et al. show that their technique works for commercially available fluorophores — which will allow biologists to adopt the technique quickly. The researchers have gone further, however, and developed a palette of 22 labels tailored to have single, isolated vibrational resonances. This will allow labels to be easily distinguished in bioimaging experiments, and will facilitate the simultaneous use of many different ones.
The authors have validated their technique in several biological contexts, using a combination of new and commercially available labels. In one example, they looked at a mixture of 16 subpopulations of cells, each of which had been cultured separately with its own type of label. The selectivity of EPR–SRS enabled each subpopulation of cells to be distinguished — allowing the researchers to take a series of images, one for each label type, over 15 minutes, and then combine them into a single image (see Fig. 4a of the paper1). Bioimaging with more colour combinations has previously been reported using fluorescent labels8,9, but not a greater number of individual labels.
Wei and colleagues also demonstrated proof-of-principle eight-colour imaging of mixed cell cultures and live tissue slices, using a mixture of labels that identified cell type or a biochemical process. This provided them with a powerful method for simultaneously comparing the responses of different nervous-system cells to induced conditions. For example, the authors investigated the effect on different cell types of proteasomal stress — which occurs when the cellular mechanism for removing damaged proteins is inhibited. Their findings suggest that cells called astrocytes are more resistant to proteasomal stress than are the neurons that these cells support, which ties in with previous observations10.
The ability to simultaneously track an increasing number of molecular species will undoubtedly expand our understanding of biological complexity. The move from single-colour to polychromatic detection in a technique known as flow cytometry (systems with up to 18 colours are commercially available) has been key to enabling immunologists to identify numerous subsets of white blood cells11. This, in turn, has been crucial for analysing and diagnosing HIV infections and other haematological diseases11. By analogy, a polychromatic microscope could characterize the spatial arrangement and interactions of cell subpopulations in tissues and adherent cell cultures (those in which cells grow on a support, rather than as a suspension).
It remains to be seen how widely adopted Wei and colleagues' method will be — other labels, such as quantum dots, might meet the needs of current biological research. However, the complexity (from the user's perspective) and cost of an SRS microscope should not be so different from those of the two-photon microscopes used for fluorescence-based imaging. Moreover, new laser technologies will drive the cost down further, and the newly developed fluorophores should be easy to substitute for many current labels. The odds of widespread adoption therefore seem reasonable.Footnote 1
Wei, L. et al. Nature 544, 465–470 (2017).
Resch-Genger, U., Grabolle, M., Cavaliere-Jaricot, S., Nitschke, R. & Nann, T. Nature Methods 5, 763–775 (2008).
Cutler, P. J. et al. PLoS ONE 8, e64320 (2013).
Krafft, C., Dietzek, B. & Popp, J. Analyst 134, 1046–1057 (2009).
Camp, C. H. Jr & Cicerone, M. T. Nature Photon. 9, 295–305 (2015).
Cheng, J. X. & Xie, X. S. Science 350, aaa8870 (2015).
Asher, S. A. Anal. Chem. 65, 59A–66A (1993).
Speicher, M. R., Ballard, S. G. & Ward, D. C. Nature Genet. 12, 368–375 (1996).
Livet, J. et al. Nature 450, 56–62 (2007).
Jänen, S. B., Chaachouay, H. & Richter-Landsburg, C. Glia 58, 1766–1774 (2010).
De Rosa, S. C., Brenchley, J. M. & Roederer, M. Nature Med. 9, 112–117 (2003).