Fluorescence microscopy is the most popular way to image biomolecules, but it leaves many of them in the dark. Non-fluorescent, light-absorbing molecules can now be viewed by a method that turns them into mini-lasers.
What happens when a molecule absorbs a photon from a beam of light? It moves from the ground state to an excited, higher-energy state and then quickly relaxes, giving off the absorbed energy as heat. There are, however, notable exceptions — molecules called fluorophores, which, after some 'wiggling', relax by emitting a lower-energy photon. Employed as molecular tags, fluorophores are invaluable in biomedical microscopy and diagnostics because they render dark molecules visible with high specificity. But what if fluorescent tagging is not an option, as in applications such as endoscopy, and the molecules under investigation stay hidden in the dark? As Min et al.1 report on page 1105, it is still possible to squeeze photons out of such molecules to produce three-dimensional images of biological systems, such as living cells and tissues.
How is this possible? Well, another mechanism for molecular relaxation exists that can be induced by a beam of light. In this process, called stimulated emission, a photon encountering an excited molecule produces a copy of itself, thus adding another photon of the same colour and propagation direction to the beam. To be effective, the energy of the stimulating photon must match the gap between the excited and the ground state. In fact, the stimulating photons need to be slightly lower in energy than their excitation counterparts, because some of the excited molecule's energy is usually lost as vibrational motion (wiggling) before the photon arrives2. Stimulated emission is used to amplify light in lasers3, and to overcome the resolution barrier in fluorescence microscopy4. As a molecular process it is almost as effective as light absorption, because both processes depend on the molecule's photon-capture area of about 0.2 × 0.2 nanometres, which is roughly the area of the molecule itself.
The role of photon-capture area in light absorption and stimulated emission can be thought of as follows. Imagine two synchronized trains of laser pulses directly focused on a molecule: pulses containing excitation photons are followed by pulses of photons for stimulated emission (Fig. 1), with each pulse containing N photons. If one could produce focal light spots that are the size of the molecule, then every photon in the pulses would interact with the molecule — a single photon (N = 1) from the first pulse would be absorbed and excite the molecule, and another from the second pulse would duplicate itself immediately afterwards.
Unfortunately, because of diffraction, the pulses cannot be focused on a region smaller than about 200 × 200 nanometres. This area is more than a million times larger than that of a molecule, and so a single photon would most probably miss its target. If, however, more than 1 million photons are used per pulse, one of them will certainly do the job, albeit at the cost of millions of surplus photons that have to be discarded. Discarding surplus excitation photons from fluorescence photons is easily done using filters, and is routine in fluorescence imaging; but singling out the duplicate photons in a stimulating pulse is impossible. Fortunately, to detect the presence of a molecule using stimulated emission, it is sufficient to measure only the number of photons that are added to the pulse — but even that is not straightforward, because N fluctuates from pulse to pulse.
Min et al.1 overcome the challenge of gain measurement by rapidly modulating the excitation beam (and thus the production of duplicate photons), and synchronizing the detection of the slightly strengthened stimulating beam with the modulation5,6. Thus, they reduce the fluctuations in the signal to 'shot noise', which for a train of m recorded pulses amounts to √(N × m). Because a pulse containing about 1 million stimulating photons yields a new photon from each of n excited molecules in a sample, m such pulses deliver m × n new photons. The increase in signal should become detectable if m × n is greater than √(N × m), which implies that m > N/n2 pulses will make a group of n dark molecules visible. In other words, detecting 20 molecules requires a minimum of 10,000 pulses. The authors' laser system fires about 108 pulses per second, which delivers images of molecules for measurement times of approximately 0.2 milliseconds on a single object point — fast enough for imaging.
Much in this new microscopy contrast mode is determined by the lifetime of the excited state, which is less than 1 picosecond (10−12 seconds). The excitation and stimulated-emission photons are preferably squeezed into pulses of about 0.2 ps duration (each providing transient intensities of less than 10 GW cm−2 in the sample, which is tolerable for living samples), with the stimulated-emission pulses immediately following their excitation counterparts. This ensures that molecules do not relax before the stimulating photons arrive. Because each round is completed in less than 1 ps, the molecules are almost instantly ready for another round. Indeed, the pulse rate produced by Min and colleagues' laser system is thousands of times slower than the maximum rate that could in principle be used. The recording time required for each measurement could therefore be dramatically reduced by firing pulses at higher rates, which should be facilitated by future developments in laser technology. Increased pulse rates might, however, damage the sample being studied, so the irradiation dose would have to remain at a level compatible with (live) cell imaging.
Another method for imaging non-fluorescent molecules has previously been reported5, in which the loss of photons, rather than the gain, is measured when an excited state of the molecules absorbs photons to enter an even higher energy state. The problem with this is that molecules in excited states are reactive, which makes them prone to decomposition. A strong advantage of Min and colleagues' approach1 is that the excited molecules are always forced back to the non-reactive ground state. Nevertheless, both methods provide three-dimensional resolution7 because their signals stem mainly from molecules in the focal region, which can be raster-scanned through a sample to build up an image. Min and colleagues' method1 also yields more photons per molecule than stimulated Raman scattering, a phenomenon that has recently also been pioneered by the same group6 for imaging non-fluorescent molecules.
But how does the new method compare with fluorescence? Unlike stimulated emission, fluorescence is randomly emitted in space and is thus harder to collect. But for most practical purposes, fluorescence still wins out because the signal can be freed from background noise. However, if non-fluorescent molecules are to be studied and fluorescent labelling is impractical, as is the case in many applications, then the stimulated-emission technique will come into its own.
Stimulated emission has been prominently used in stimulated emission depletion (STED) microscopy4,8, to keep all the fluorophores in a sample dark except for those in a spot smaller than the diffraction resolution limit of the microscope. But in STED microscopy, apart from its ability to switch fluorophores off, there is no interest in stimulated emission per se. Indeed, in related nanoscopy methods8, stimulated emission has been replaced by other mechanisms for keeping fluorophores dark, such as flipping the spin of one of their electrons (triplet-state pumping) or relocating some of their atoms (cis–trans photoisomerization)8. By contrast, in Min and colleagues' technique1, stimulated emission is the actual goal. An intriguing possibility for the future would be to design a set of laser pulses that fulfil both roles of stimulated emission — switching off molecular signals and stimulating photon emission — to provide images of unlabelled, non-fluorescent molecules at sub-diffraction (nanoscale) resolution for the first time. So, for many reasons, Min and colleagues' method is a bold step towards unveiling details of live cells and tissues that would otherwise be left uncharted.