High-speed atomic force microscopy is a powerful tool for studying structure and dynamics of proteins. So far, however, high-speed atomic force microscopy was restricted to well-controlled molecular systems of purified proteins. Here we integrate an optical microscopy path into high-speed atomic force microscopy, allowing bright field and fluorescence microscopy, without loss of high-speed atomic force microscopy performance. This hybrid high-speed atomic force microscopy/optical microscopy setup allows positioning of the high-speed atomic force microscopy tip with high spatial precision on an optically identified zone of interest on cells. We present movies at 960 ms per frame displaying aquaporin-0 array and single molecule dynamics in the plasma membrane of intact eye lens cells. This hybrid setup allows high-speed atomic force microscopy imaging on cells about 1,000 times faster than conventional atomic force microscopy/optical microscopy setups, and allows first time visualization of unlabelled membrane proteins on a eukaryotic cell under physiological conditions. This development advances high-speed atomic force microscopy from molecular to cell biology to analyse cellular processes at the membrane such as signalling, infection, transport and diffusion. H igh-speed atomic force microscopy (HS-AFM) 1 has been proven to be a unique and powerful tool for the concomitant analysis of the structure and dynamics of single biomolecules 2. HS-AFM was used to visualize myosin-V walking 3 , cellulase cellulose degradation 4 , F 1-ATPase rotary catalysis 5 , bacteriorhodopsin photocycle 6 , and OmpF diffusion and interaction 7. However, HS-AFM has been restricted so far to well-controlled molecular systems of purified proteins under well-controlled conditions. These systems were characterized by a limited number of pure molecular species with small corrugation, mainly because the fast z-piezo that follows the topography profile has a small extension range (typically 400 nm; ref. 8). Recently, the development of a wide-range HS-AFM scanner allowed the first nanoscale observation of molecular movements at two frames per second on living bacteria 9. Historically, AFM debuted as early as 1990 for cell imaging applications 10 , prompted by the need to perform cell structural analysis at a resolution superior to light microscopy. However, AFM images of cells revealed rather the cell interior, like actin filaments, than the membrane structure of the cells. Later, AFM and optical microscopy (OM) were combined in order to take advantage of OM's large-scale overview imaging capacities and its power to analyse fluorescence signal targets of proteins of interest 11. For this purpose, AFMs were built in a table-top configuration and mounted on inverted optical microscopes 12. In order to avoid modifications of the conventional inverted OM, two types of table-top AFM configurations were built. In one of them, the AFM tip and the scanner is combined into a single moving component, and in the other a large optical microscope sample stage on which the AFM sample is mounted needs to be moved. In both cases, there is a loss of AFM performance. In the tip-scanner configuration, the laser detection must be coupled to the moving cantilever, while for the second case, the sample stage that needs to be moved is complex and heavy. Both types of structures are prone to capture environmental noise and feature innate resonance frequencies. Furthermore, massive objects do not allow sub-millisecond mechanical response, thus precluding individual protein imaging at high resolution and high speed. In this work, we integrate an OM path into our HS-AFM, maintaining the structure of the HS-AFM setup 1 , and hence not compromising HS-AFM performance in terms of speed and resolution. To achieve this, we choose a completely different approach compared with most (if not all) AFM/OM integration developments: instead of building a table-top AFM mountable on an inverted optical microscope, accepting loss of AFM performance, we build an optical path into our HS-AFM setup, accepting minimal trade-offs in OM performance. We show that our setup can acquire bright field and fluorescence OM images of biological samples, and that it allows HS-AFM tip positioning with high precision guided by OM. Finally, we show that it achieves HS-AFM imaging of individual membrane proteins on eukaryotic cells and record their dynamics at an imaging rate of 960 ms per frame. This accomplishment opens the door to a wide range of real-time studies of molecular dynamics in membrane processes on living cells. Results Development of hybrid HS-AFM and fluorescence microscope. In order to integrate an OM path into the HS-AFM (Fig. 1 and Fig. 2a,b), we added and exchanged several elements to and from the HS-AFM setup, previously designed by Ando et al 1. The original laser diode of the HS-AFM system that reads the cantilever position was changed to a far-red superluminescent diode (SLD) with 750 nm wavelength and low coherence (Fig. 1, label 1). The use of the SLD 750 nm allowed the optical separation