Real-Time Single-Molecule Imaging of the Infection Pathway of an Adeno-Associated Virus

See allHide authors and affiliations

Science  30 Nov 2001:
Vol. 294, Issue 5548, pp. 1929-1932
DOI: 10.1126/science.1064103


We describe a method, based on single-molecule imaging, that allows the real-time visualization of the infection pathway of single viruses in living cells, each labeled with only one fluorescent dye molecule. The tracking of single viruses removes ensemble averaging. Diffusion trajectories with high spatial and time resolution show various modes of motion of adeno-associated viruses (AAV) during their infection pathway into living HeLa cells: (i) consecutive virus touching at the cell surface and fast endocytosis; (ii) free and anomalous diffusion of the endosome and the virus in the cytoplasm and the nucleus; and (iii) directed motion by motor proteins in the cytoplasm and in nuclear tubular structures. The real-time visualization of the infection pathway of single AAVs shows a much faster infection than was generally observed so far.

Single-molecule detection techniques have been developed for imaging and for spectroscopic characterization of individual fluorescent molecules (1–3). Within the last years, these techniques have been increasingly applied to biological topics (4, 5). By overcoming the problem of ensemble averaging, these techniques enabled questions in molecular biology to be solved, which hitherto could not be answered by conventional ensemble measurements (6–8). Single-molecule imaging has previously been applied to study the diffusional behavior of single molecules in lipid bilayers (9, 10), in fluids (11–13), or recently, in living cells (14) and their membranes (15,16). Here, we show for the first time that this method can be used for the visualization and kinetic characterization of the infection pathway of single viruses in living cells.

The viral infection process is a very intriguing interaction in nature. It starts with the contact between the virus and the cell membrane and finally results in transport of the virus into the nucleus and gene expression. For antiviral drug design, as well as for the development of efficient gene therapy vectors, it is essential to understand these processes. Electron microscopy is one tool for obtaining knowledge on the different stages of the infection pathway, but it cannot be carried out in living cells with high time resolution (17). A first step in this direction was the establishment of capsid labeling methods and their detection in fixed cells by conventional fluorescent microscopy (18). However, using conventional fluorescent microscopy either a high concentration of viruses or a high degree of labeling or both had to be used. This masks the detection of individual processes, can create unphysiological conditions, altering the kinetics of the infection process, and due to steric hindrance can interfere with the virus-cell interaction. In this regard, real-time imaging of single viruses labeled with only one dye molecule in living cells is a breakthrough.

Our model system AAV is a promising vector for gene therapy (19), although its infection process is still not completely understood (20, 21). For this study, AAV was covalently labeled with Cy5 dye, with a ratio of single to double labeling of 82:18. We analyzed 1009 trajectories of single AAV-Cy5 particles in 74 cells at different stages of the infection, an example of which is shown in Fig. 1.

Figure 1

Trajectories of single AAV-Cy5 particles indicating infectious entry pathways of AAVs into a living HeLa cell. To demonstrate how cell constituents were determined, trajectories are projected onto a phase contrast image of the investigated cell cross-section, taken with a commercial CCD (Coolpix, Nikon) attached to the binocular tubus of the microscope (36). The traces showing single diffusing virus particles were recorded at different times. They describe various stages of AAV infection, e.g. diffusion in solution (1 and 2), touching at the cell membrane (2), penetration of the cell membrane (3), diffusion in the cytoplasm (3 and 4), penetration of the nuclear envelope (4), and diffusion in the nucleoplasm. Under excitation with red light (633 nm) the autofluorescence of cells was low enough to allow the observation of single fluorophores as bright spots (Fig. 3, A and B). A two-dimensional gaussian fit of the intensity profile (9,10) provided a localization accuracy of 40 nm for these spots. We took series of images at 40 ms intervals to construct two-dimensional projections of single AAV-Cy5 trajectories. The brightness of the fluorescence spots in these series shows the time dependence of the fluorescence intensity (see fluorescence time traces in Fig. 3, C through E).

The motion of AAV outside the cell could be characterized as normal diffusion with a diffusion coefficient of D = 7.5 μm2/s (22).

When AAV approached the cell membrane, a general deceleration was observed (23). Typical trajectories displayed (Figs. 1 and2A) movements toward the cell surface, followed by a mean of 4.4 repetitive touching events (Fig. 2B) with a mean touching time of t t = 62 ms (Fig. 2C), interrupted by short diffusion paths in the vicinity of the cell surface. It is not clear yet, whether these touching events represent binding/release processes at a receptor or simply unspecific adsorption of the virus to the cell surface. The touching resulted in a mean contact time of t c = 3.2 s, measured between the first and the last cell surface contact. For virus particles that entered a cell, the mean contact time decreased tot c = 1.2 s, because the series of touching events was stopped by penetration of the membrane. A penetration efficiency of 13% was measured (24), implying that the membrane itself is a major barrier for virus uptake. The precise reason for this phenomenon is unknown, but probably involves a failure of interaction of receptor and different cofactors essential for the virus internalization (25–27).

Figure 2

Uptake of AAV-Cy5 by a living HeLa cell. (A) Membrane “touching events” of AAV at the surface of a living HeLa cell. An enlarged sketch of trajectory 2 taken from Fig. 1. The path shows five touches to the cell surface. These hits are highlighted by circles and represent short periods of immobility. (B) The mean number of consecutive cell touches 〈n Touch〉 derived for viruses with negative docking (diffusing back into solution) calculated from 269 trajectories (frequency plot with gray bars). The accumulated frequency ofn Touch was fitted with a sigmoid function. Its normalized derivative, the probability density functionpdn Touch (dot curve) was fitted with a gaussian curve. From this fit a mean value 〈n Touch〉 = 4.4 ± 3.1 was determined. A model for the statistical calculation of cell surface touches is inserted in (B) (37). In an idealized model the cell is seen as a spherical shell of radiusr = a. Particles released at r = b move inward at a rate of k in or outward withk out. The probability p for touching the sphere is p =k in/(k in +k out) = a/b. The mean number of consecutive touches a particle undergoes oscillating between r = a and r = b before wandering away from the shell for good is 〈n Touch〉 = Σn=0 npn(1 − p) = p/(1 − p) = a/(b − a). This mean number can be estimated to be 〈n Touch〉 = 5 for a typical value of a = 5 μm and a typical value of b = 6 μm between the touches. Distribution of adsorption times for 137 negative docking (C) and 42 membrane penetration (D) trajectories. The bars indicate the accumulated frequency of viruses observed with specific docking times before leaving the cell surface or penetrating the cellular membrane. These accumulated frequencies are fitted with exponential functions. Based on these data, the 1/e decays of the docking times were determined to be 〈t〉 = 62 ± 30 ms (C) and 〈t〉 = 64 ± 30 ms (D), respectively, indicating no significant time differences between cell leaving and penetrating events.

The postulated uptake mechanism for AAV into living cells requires endocytosis (20, 21). To our surprise, adsorption times at the cell membrane for particles which penetrated cells were not increased compared to those followed by detachment from the cell surface. These times are in the range of 64 ms, which is very short, especially for adsorption followed by penetration (Fig. 2D). First measurements on CHO K1 cells gave similar values. Nevertheless, there is strong evidence that the AAV entry pathway starts with endosome formation, followed by virus release. We imaged 113 trajectories within the cytoplasm and analyzed their diffusion as discussed before (Fig. 3, A through G). Fifty-three trajectories showed a linear 〈r 2〉 – tmovement indicating normal diffusion. The distribution function of the calculated diffusion constants (Fig. 3H) showed two maxima atD = 1.3 μm2/s and D = 0.6 μm2/s, indicating two different components. The faster species, which is approximately five times slower than the diffusion coefficient of AAV in aqueous solution, can be assigned to the free virus. The reduction of the diffusion coefficient matches the results known from other macromolecules diffusing in the cytoplasm (28,29). Using the Stokes-Einstein equation, the slower component can be assigned to a particle of about 50 nm in diameter, which was identified as AAV inside an endosome. Control experiments done in cells at pH = 9 support this interpretation (Fig. 3I). In all experiments no free virus but only endosome diffusion has been observed directly after membrane penetration and only one virus per endosome was found.

Figure 3

Endosomal processing and diffusion in cytoplasm. (A and B) Two series of fluorescence images showing one and two fluorescence spots each indicating a single AAV particle. Time distance for each panel is 40 ms and the size is 5 μm by 5 μm. (C through E) Time traces of the fluorescence intensity (I) of the spots shown in (A) and (B) (correlated by color). The plots feature the characteristic on/off dynamics (blinking) and one-step photobleaching behavior typical of single molecules. Time trace noise is the result of diffusive movement perpendicular to the focal plane (38). (F) Visualization of the three trajectories projected onto the transmitted light image taken with the microscopic setup used for fluorescence detection. Cell membrane and nucleus, both outlined yellow, were determined by phase contrast imaging as demonstrated in Fig. 1. (G) The mean square displacement plotted with time. The magenta curve yielding a diffusion coefficient of D = 1.4 μm2/s is ascribed to a free AAV particle undergoing normal diffusion in the cell plasma. The green and yellow curves are attributed to endosomal movements, one with normal diffusion (D = 0.55 μm2/s green) and one with anomalous diffusion (D = 0.2 μm2/s, α = 0.6 yellow). (H and I) Probability density function pdD for diffusion coefficients derived for particle movement in the cytoplasm (obtained in analogy to Fig. 2B) from 53 trajectories at pH = 7 (H) and 10 trajectories at pH = 9 (I). We found two maxima of the probability density function for pH = 7 attributed to free AAV (D = 1.3 μm2/s) and to AAV inside the endosome (D = 0.57 μm2/s). The latter maximum is similar to the maximum (D = 0.64 μm2/s) found for pH = 9, at which endosomal release is not possible. The low value ofD in (H) was therefore assigned to AAV inside an endosome. The histograms show the distribution of diffusional constants.

Fifty-one trajectories showed deviations from the linear dependency of mean square displacement with time (Fig. 3, F and G), indicating anomalous diffusion processes (〈r 2〉 = 4Dt α) with a D value in the range of 0.3 to 1.5 μm2/s and 0.5 < α < 0.9. This kind of diffusion occurred in localized areas of the cell and can be ascribed to obstacles or adsorption sites hindering the free Brownian movement of the particles (30).

Furthermore, directed motion was observed for nine of the trajectories in the cytoplasm. For all these cases the mean square displacement could only be fitted with a parabola [〈r 2〉 = 4Dt + (vt)2] indicating diffusion with drift (30). The calculated diffusion coefficients were within the range of D = 0.4 to 0.9 μm2/s and velocities in the range of v = 1.8 to 3.7 μm/s. Treatment with nocodazole started 1 hour before infection removed the directed motion, indicating microtubule-dependent transport of viruses by motor proteins like kinesin or dynein. This kind of transport mechanism is known to occur for several viruses which use it as an efficient way to move toward the nucleus (17, 21, 31).

At 15 min after the start of the experiment (32), at least one AAV-Cy5 was detected in the nucleus of 50% of the cells. This indicates a much faster infection compared to the 2-hour infection time measured by Bartlett et al. (20) for the same system with conventional methods. Conventional methods require the accumulation of a higher number of viruses for detection; our method, which is sensitive to single viruses, gives much shorter and more accurate values.

More than a hundred trajectories of viral particles were measured in the nuclear area of cells with different sized nuclei. Fifty-seven of the trajectories showed normal diffusion, with diffusion coefficients ranging from D = 0.4 to 1.3 μm2/s with a mean value of 〈D〉 = 0.85 μm2/s. In addition, for 23 of the viral particles anomalous diffusion was observed with D in the range of 0.1 to 0.5 μm2/s and 0.6 < α < 0.9. These values indicate a somewhat slower, but in principal, similar movement of viruses in the nuclear area as compared to the cytoplasm.

To our great surprise, in 34 of the trajectories within the nuclear area the viral particles underwent directed motion along well-defined pathways (Fig. 4). Some pathways were used several times consecutively by different viral particles. In addition, all trajectories in a given nuclear area were “unidirectional,” i.e., oriented in roughly the same direction through the nuclear area (Fig. 4A). The typical number of pathways taken seems to be restricted to about one to five within one nuclear area. For all these trajectories, the mean square displacement was quadratically dependent on t (30), resulting in diffusion coefficients in the range of D = 0.25 to 0.9 μm2/s and velocities in the range of v = 0.2 to 2.8 μm/s. Again, treatment with nocodazole prevented directed motion. This would suggest active transport of the virus within the nuclear area. However, motor proteins such as kinesin or dynein as well as microtubules have not been observed in the nucleoplasm so far. On the other hand, it is well known that tubular structures, formed by invagination of the nuclear envelope exist and sometimes even transect the nucleus (33, 34). The core of these channels is continuous with the cytoplasm. Microtubules could polymerize into these channels and allow active transport of the AAV by motor proteins. This model would explain the observation of the typical number of pathways and the “unidirectional” movement described above, because (i) the typical number of large nuclear channels is one to five (33) and (ii) the growth of the microtubules have to start from the centrosome on one side of the nucleus, more or less “unidirectional” into the nuclear channels. If either kinesin or dynein are used as motor proteins, the “unidirectional” transport is comprehensible. It is remarkable to observe viruses diffusing freely in the cytoplasm and then suddenly starting to move with constant velocity into the nuclear area along well-defined pathways [real-time movie available at (35)].

Figure 4

Transport of AAV-Cy5 within the nuclear area. (A) The visualization of five trajectories projected onto the white light image of the cell nucleus. The position of the nucleus outlined in yellow was determined with phase contrast imaging. The lower two trajectories run along the same pathway consecutively. In this case, all trajectories showed directed motion from the left to the right side, i.e. were “unidirectional.” (B) Mean square displacement plotted with time. The parabolic shape of the curves indicates diffusion with drift as described by the equation 〈r 2〉 = 4Dt + (vt)2 with diffusion coefficients ofD = 0.25 to 0.35 μm2/s and velocities in the range of v = 0.2 to 1.4 μm/s.

In conclusion our single virus tracing measurements have allowed, for the first time, a detailed observation and quantitative description of the infectious entry pathway of single virus particles labeled with only one dye molecule into living cells. The low degree of labeling and the extremely low concentration of viruses per cell guaranteed natural and physiological conditions with a minimum of distortion of the virus-cell interactions in contrast to all measurements carried out so far. The presented experiments give a real-time movie script of a virus infection in a living human cell with a hitherto unachievable clarity and detail. Refinement of the method to include labeling of capsid and viral DNA with different dyes will further allow characterization of the disassembly and nuclear processing of viruses. Knowledge obtained by single virus tracing experiments as described here, will give new insight into the broad scale of virus-cell interactions and will be essential for the development of antiviral drugs and specific second generation gene therapy vectors.

  • * To whom correspondence should be addressed. E-mail: christoph.braeuchle{at}


View Abstract

Navigate This Article