Cryo Electron Tomography of Native HIV-1 Budding Sites

The structure of immature and mature HIV-1 particles has been analyzed in detail by cryo electron microscopy, while no such studies have been reported for cellular HIV-1 budding sites. Here, we established a system for studying HIV-1 virus-like particle assembly and release by cryo electron tomography of intact human cells. The lattice of the structural Gag protein in budding sites was indistinguishable from that of the released immature virion, suggesting that its organization is determined at the assembly site without major subsequent rearrangements. Besides the immature lattice, a previously not described Gag lattice was detected in some budding sites and released particles; this lattice was found at high frequencies in a subset of infected T-cells. It displays the same hexagonal symmetry and spacing in the MA-CA layer as the immature lattice, but lacks density corresponding to NC-RNA-p6. Buds and released particles carrying this lattice consistently lacked the viral ribonucleoprotein complex, suggesting that they correspond to aberrant products due to premature proteolytic activation. We hypothesize that cellular and/or viral factors normally control the onset of proteolytic maturation during assembly and release, and that this control has been lost in a subset of infected T-cells leading to formation of aberrant particles.


Classification of Gag lattice types in tomograms of resin-embedded sections
The position of the plasma membrane in the budding sites was manually traced using IMOD [1] on every third xy-slice in the central 30 slices of the budding sites (where the membrane was cut in an angle of 90±~20° by the tomographic slice). From this set of linearly connected points, a triangular mesh surface of the plasma membrane was interpolated and smoothed using the IMOD functions smoothsurf, imodmesh and imodfillin, combined with custom-made scripts for MATLAB (Mathworks, Natick, Massachusetts, United States). The points in this surface model have an approximate spacing of one voxel, and the surface normal at every point can be calculated using the connection to the neighboring points. Using this, the tomogram grey values along a line orthogonal to the membrane were calculated for every point, and the average of these one-dimensional plots over the membrane surface was calculated.
When necessary, the resulting line plots were scaled to achieve the same object pixel size. They were cropped to either [-27 nm, -6 nm] or [-27 nm, 4.5 nm] (where 0 nm is the center of the membrane and the negative direction is the cell/virion interior) to contain either the membranebound density or the membrane-bound density and the membrane. After normalization of the scaled and cropped line plots to average 0 and standard deviation 1, they were subjected to further analysis.
Classification of the line plots was performed in MATLAB using functions in the statistics toolbox. Principal component analysis [2] was performed with the MATLAB function princomp, hierarchical clustering with the MATLAB functions pdist, linkage and cluster. Briefly, the distance between the 15 element vectors (containing the membrane-bound density) was calculated as the Euclidean distance and was used for linkage analysis with linkage, using Ward's linkage criterion of increase in intragroup variance. Linkage analysis using the criterion of mean intercluster distance gave indistinguishable results.

Calculation of the theoretical frequency distribution in Figure 7E
The gray circles in Figure 7E represent the theoretical frequency distribution expected if all cells had 18% budding sites in class 1 (thin lattice). This distribution was calculated by taking into account the small number of budding sites sampled per cell, in the following way: (i) If the same number n of budding sites had been recorded on every cell, the probability of having k (k≤n) budding sites in class 1 (thin lattice) is given by the Binomial distribution P n (k) = (n!/(k!(n-k)!))×0.18 k ×(1-0.18) (n-k) .
The frequency (x-axis in Figure 7E) is given by k/n and its probability (y-axis in Figure 7E) is given by P n (k).
(ii) The actual experimental data contain a varying number of budding sites per cell. We denote the probability of n budding sites on a cell in the data set by W n . The theoretical frequency distribution displayed in Figure 7E is then given as the weighted average of the P n (k) by W n , where the frequencies k/n for each individual P n (k) was rounded to the closest 0.1 for the histogram representation. This combined frequency distribution has local maxima around 0.25 and 0.5 since these values are among the few that can be formed for certain values of n (2,4,8) that occur often in the data.    Figure S3 is seen to be slight variations in the position of the NC-RNA layer (B). There are also variations in the membrane region between the four classes, but since this part of the density profile is not used in the PCA or clustering (but only the shaded part as indicated in Fig. 6C), it will not contribute to the variation in class 2 found there.