The speed of super-resolution microscopy methods based on single-molecule localization, for example, PALM and STORM, is limited by the need to record many thousands of frames with a small number of observed molecules in each. Here, we present ANNA-PALM, a computational strategy that uses artificial neural networks to reconstruct super-resolution views from sparse, rapidly acquired localization images and/or widefield images. Simulations and experimental imaging of microtubules, nuclear pores, and mitochondria show that high-quality, super-resolution images can be reconstructed from up to two orders of magnitude fewer frames than usually needed, without compromising spatial resolution. Super-resolution reconstructions are even possible from widefield images alone, though adding localization data improves image quality. We demonstrate super-resolution imaging of >1,000 fields of view containing >1,000 cells in ∼3 h, yielding an image spanning spatial scales from ∼20 nm to ∼2 mm. The drastic reduction in acquisition time and sample irradiation afforded by ANNA-PALM enables faster and gentler high-throughput and live-cell super-resolution imaging.

The spatial resolution of standard optical microscopy techniques is limited to roughly half the wavelength of light. As a result of diffraction 1 , the image of an arbitrarily small source of light imaged using a lens-based microscope is not a point but a point spread function (PSF), usually an Airy pattern, with a central peak approximately ~200-300 nm in width (Fig. 1a), resulting in a blurring of structures below this spatial scale. This diffraction limit restricts the ability of optical microscopy techniques to resolve the subcellular organization of individual molecules or molecular complexes, which are smaller than this limit; for example, the structure of a nuclear pore complex, which is made up of hundreds of individual proteins, with a diameter of only ~120 nm, remains obscured by conventional microscopy (Fig. 1b).Microscopy methods have emerged over the past two decades that can overcome the diffraction limit and enable the imaging of biological structures such as nuclear pores, viruses, chromatin complexes and cytoskeletal filaments at resolutions close to the molecular scale 2 . The most well known of these super-resolution microscopy methods fall into three main categories: stimulated emission depletion 3 , structured illumination microscopy 4,5 and single-molecule localization microscopy (SMLM) [6][7][8][9] , which is the focus of this Primer. SMLM methods usually employ conventional wide-field excitation and achieve super-resolution by localizing individual molecules [6][7][8][9][10][11][12][13][14][15] .They have become broadly adopted in the life sciences owing to their high spatial resolution -typically ~20-50 nm or better -and relative ease of implementation, although each super-resolution method has its unique advantages and limitations and is optimally suited for different applications (discussed elsewhere 2 ).SMLM is fundamentally based on the fact that the spatial coordinates of single fluorescent molecules (also called fluorophores, or emitters) can be determined with high precision if their PSFs do not overlap. Subpixel shifts in the coordinates of a fluorophore lead to predictable changes in pixel intensities that can be used to compute its precise location (Fig. 1c). The localization precision reflects the scatter of localizations that would be obtained if a molecule was imaged and localized many times, and is fundamentally limited by the signal to noise ratio (SNR) and not by the wavelength of light or the pixel size (Fig. 1d). To avoid overlaps between the PSFs of individual molecules, fluorescent emissions of distinct molecules are separated in time; the most common approach to obtain this temporal separation exploits the phenomenon of photoswitching, where fluorescent molecules can switch between an active 'ON' or 'bright' state, where they emit fluorescent light when excited, and one or more inactive 'OFF' or 'dark' state in which they do not fluoresce (Fig. 1e).Photoswitching for a particular molecule is a stochastic event; however, switching probabilities can be DiffractionThe be...

Actin-based motility is used by various pathogens for dissemination within and between cells. Yet host factors restricting this process have not been identified. Septins are GTP-binding proteins that assemble as filaments and are essential for cell division. However, their role during interphase has remained elusive. Here, we report that septin assemblies are recruited to different bacteria that polymerize actin. We observed that intracytosolic Shigella either become compartmentalized in septin cage-like structures or form actin tails. Inactivation of septin caging increases the number of Shigella with actin tails and enhances cell-to-cell spread. TNF-α, a host cytokine produced upon Shigella infection, stimulates septin caging and restricts actin tail formation and cell-to-cell spread. Finally, we show that septin cages entrap bacteria targeted to autophagy. Together, these results reveal an unsuspected mechanism of host defense that restricts dissemination of invasive pathogens.