Flipping nanoscopy on its head

About the smallest object we can see with the naked eye is our own hair. With a magnifying glass, we can see about 10 times better, and light microscopy, until relatively recently, could resolve features about 300 times thinner than human hair (~250 nm). Recent developments in fluorescence “nanoscopy” made it possible to routinely image cellular structures at 20- to 30-nm resolution (1), but a gap remained at the molecular scale: Most proteins are smaller than 5 nm across. On page 606 of this issue, Balzarotti et al. (2) report a new concept in nanoscopy, termed MINFLUX, that achieves the true molecular resolution (2 to 3 nm) and dramatically reduces the number of photons required by “flipping” a common wisdom in nanoscopy on its head.

In traditional optical imaging, even a very tiny object such as a single fluorophore (<1 nm) becomes blurred because light diffraction makes it appear much larger—about half the wavelength of the light used. Nonetheless, the center of the imaged fluorescence spot can be determined with extremely high precision (down to 1.5 nm if 10,000 photons are used) (3). The Abbe diffraction limit can be overcome by determining the position of, or localizing, one molecule at a time, with single fluorescent molecules that can be switched on or off stochastically (4, 5). This localization-based strategy, together with other nanoscopy approaches (6, 7), ushered in the “resolution revolution” that enabled breakthrough biological discoveries in the past decade (1).

A common wisdom in nanoscopy is to localize a molecule near where the signal is the strongest, which requires high-emission photon flux that is often limited by the emission rate of the fluorophore. In MINFLUX, Balzarotti et al. devised the opposite strategy, in which they localize a molecule near the signal minimum (see the figure, top left). Imagine an incident beam pattern that excites a molecule maximally at position x₀ along its profile (signal = N₀ photons). In order to confidently declare that the molecule has moved by Δx to a new nearby position x₁ (N₁ photons), the associated signal change (N₀ – N₁) should be larger than the Poissonian noise (N₀^1/2). For example, if N₀ is 100, Δx has to be large enough to reduce the signal by 10 photons. But if we flip the excitation profile so that we get zero signal at position x₀ (signal N₀ = 0), much smaller Δx can be detected as long as N₁ > 1 (see the figure, top right).

Seeing better in dimmer light.

The MINFLUX method uses minimal excitation to resolve changes in molecular position within an excitation pattern. In common nanoscopy, a molecule’s position is determined near the maximum of the excitation intensity profile whereas in MINFLUX, it is determined near the minimum, requiring much fewer photons.

The position of a molecule (red circle) is probed by watching it brighten and dim as the donut-shaped excitation profile is moved around it. Dark gray is brightest excitation.

As such, higher resolution is achieved with much reduced photon flux, but the signal may also be zero or minimal because no molecule is present. Thus, MINFLUX requires some a priori knowledge of where the molecule is. However, this level of imaging requires very few photons and can be done with conventional microscopy. In other words, emitted photons from the molecule do not pay the main cost of determining its position, as in localization-based approaches, but to merely confirm its presence and fine-tune its position estimation within the excitation pattern. In addition, because the localization accuracy is determined with excitation modulation, MINFLUX has the additional benefit of polarization-independent accuracy, an occasional problem in existing nanoscopy.

Once the position of a molecule in a view field is roughly located, the excitation pattern is serially moved to multiple positions around the molecule with a small displacement (50 to 150 nm) (see the figure, bottom). The observed fluorescence signal of the molecule at each position is then compared with the expected signal based on the known intensity profile and placement of the excitation pattern to estimate the position of the molecule. In the current work, a donut-shaped excitation pattern with an intensity zero at center was used, but in principle, any excitation pattern should work.

Using MINFLUX, Balzarotti et al. resolved fluorescent molecules spaced only 6 nm apart from each other on a DNA origami structure, with only 1000 photons per molecule in ~2 min imaging time. In comparison, these molecules could not be resolved with the same number of photons in existing localization-based nanoscopy even under the most ideal conditions of no background and a perfect-detection camera. Similar resolution was achieved of DNA origami previously by using a method called DNA PAINT, but with 50,000 photons and an image acquisition time of 2 hours (8).

The minimal photon flux feature of MINFLUX is particularly advantageous for single-molecule tracking experiments that are often limited by rapid photobleaching of fluorescent proteins. Individual 30S ribosomal molecules labeled with photoactivatable fluorescence protein diffusing in live bacterial cells were followed for orders-of-magnitude more time points. The average length reached ~750 time points per trajectory compared with ~5 to 10 in standard tracking experiments, making it possible to detect temporal changes in diffusion coefficient. The tracking duration was only ~150 ms, however, in part because of fluorescent intermittency at the millisecond time scale.

With all its stunning performance, MINFLUX still operates under the fundamental limit of all optical nanoscopy methods—that is, temporal resolution must be traded off in order to improve spatial resolution because of the sequential nature of estimating molecule positions (9, 10). In its present form, four probing positions of the donut beam must be serially sampled to localize a molecule. Furthermore, in order to achieve molecular resolution, these beams must be placed within 50 to 150 nm of the molecule or so, which limits the effective field of view. Although each localization cycle takes only a few microseconds aided by hardware-based modulation of excitation profile, the serial scanning format and small field of view would require ~100 hours to image an area with the size of a human cell. Although there is still much to improve, it should be noted that the original single-molecule localization nanoscopy images took as much as an overnight acquisition 10 years ago (4) but can now be done in less than 1 min or so.

What can we expect of MINFLUX, and more broadly, superresolution and single-molecule imaging? Because MINFLUX can now reach a resolution less than 5 nm, single-molecule fluorescence resonance-energy transfer, which can determine distances of up to ~7 nm at 0.3-nm resolution with only about 100 photons (11), may be combined to obtain dynamic structural information continuously covering from the length scale of single amino acids to the cellular scale or larger. A considerable challenge would be to extend the molecular resolution to three-dimensional imaging, which most certainly would require interferometric methods (12). Moving toward multicolor imaging is likely to be more straightforward because the precision in position determination is largely wavelength-independent in MINFLUX and because more fluorescent reporters become eligible because of the reduced photon budget.

Ultimately, the true spatial resolution of an image is going to be limited by how densely the sample can be labeled, However, the greater resolving power achieved at molecular distances that has been enabled by MINFLUX is likely to stimulate further developments in probe and labeling technologies. MINFLUX also requires more hardware engineering as compared with other localization-based nanoscopy. Nevertheless, rapid commercialization, pending further developments necessary for cellular imaging, may make it available to biologists in the not-too-distant future. ■