Super-resolution imaging of ESCRT-proteins at HIV-1 assembly sites.

Plasmids

HIV encoding plasmid pCHIV and its labelled derivatives HIV eGFP , HIV mCherry and HIV mEos have been previously described [ 48 , 76 ]. The Vps4A-E228Q-mCherry plasmid was also described previously [ 45 ]. Plasmid synGag was obtained from Ralf Wagner (University of Regensburg, Germany)[ 77 ], pGag.eGFP was contributed by Marilyn Resh [ 59 ], and peGFP.Vpr was kindly provided by Tom Hope [ 78 ]. pGag.mCherry was derived from pGag.eGFP by replacing a BamHI/BsrGI fragment comprising the eGFP coding sequence with a corresponding restriction fragment comprising the mCherry ORF generated by PCR. The pCHMP4B-HA plasmid encoding CHMP4B fused to an HA-tag was kindly provided by H. Göttlinger [ 79 ]. The YFP-Tsg101 plasmid was a kind gift of Wesley Sundquist (University of Utah, Salt Lake City, USA).

Cells and transfection

HeLa cells (Japanese Collection of Research Bioresources Cell Bank, Osaka Japan) were grown in Dulbecco’s modified Eagle’s medium (DMEM), supplemented with 10% fetal calf serum (FCS). HeLa cells were seeded into LabTek II 8-well chamber slides at a cell density of 2*10 4 cells/well and transfected on the following day using X-tremeGENE transfection reagent (Roche) according to the manufacturer’s instructions. For the CHMP4B-HA/ HIV:HIV mCherry experiments, 100 ng of CHMP4B-HA, 50 ng of pCHIV (wildtype) and 50 ng of pCHIV mCherry were transfected per well. For the study of endogenous ESCRT proteins Tsg101, CHMP2A and ALIX, 50 ng pCHIV and 50 ng pCHIV mCherry were used or, for PALM imaging of the HIV-1 assembly sites, 50ng pCHIV and 50 ng pCHIV mEos were used. After transfection, cells were incubated for 14–15 hours at 37°C and 5.0% CO 2 . For validating immunostaining efficiency within the viral bud, 75 ng of pEGFP.vpr and 50 ng of synGag (wildtype) and 50 ng of Gag.mCherry or Gag.eGFP were used for transfection, respectively. For experiments with Tsg101-fusion proteins, 50 ng of YFP-Tsg101 and 50 ng of pCHIV mCherry and pCHIV were used for transfection, respectively. In the case of Gag.mCherry, the pCHIV plasmids were replaced by 50 ng synGag and 50 ng of Gag.mCherry, respectively. For Vps4A-depletion experiments, 70 ng pCHIVeGFP, 70 ng pCHIV and 35 ng Vps4A-E228Q-mCherry were used for imaging of CHMP2A and cells were transfected with 50 ng CHMP4B-HA, 35 ng pCHIVeGFP, 35 ng pCHIV and 35 ng Vps4A-E228Q-mCherry for imaging CHMP4B-HA. Super-resolution sample preparation For STORM imaging, immunostaining of Tsg101 was performed using primary mouse monoclonal anti-Tsg101 (clone 4A10, #GTX70255; Genetex) and secondary donkey anti-mouse IgG (#ABIN336468, purchased via antibodies-online.com ) antibodies. For labeling of secondary anti-mouse antibodies, the unconjugated antibody was mixed with Alexa Fluor 488 succinimidyl ester (Invitrogen) and Cy5 bis-NHS-ester (GE Healthcare) in a molar ratio of 1:4:1 (antibody: Alexa Fluor 488: Cy5) in 150 mM NaHCO 3 buffer (pH 8.2) and incubated overnight. Unreacted dye molecules were subsequently removed by gel permeabilization chromatography (Performa DTR Gel Filtration Cartridges from Edge BioSystems). For dSTORM experiments, secondary antibodies were only labeled with Cy5-bis-NHS-ester in a molar ratio dye:protein = 1:4. Monoclonal primary mouse anti-ALIX antibodies (clone 3A9, #634502, BioLegend) were directly labeled for immunostaining of ALIX according to the protocol above as was also done for monoclonal primary anti-HA antibodies (clone 3F10, #11867423001; Roche) that were used to stain CHMP4B-HA. For immunostaining of CHMP2A, we used unlabeled primary polyclonal rabbit anti-CHMP2A antibodies (#ab76335, abcam) combined with secondary donkey anti-rabbit IgG (#ABIN376979, purchased via antibodies-online.com ) that were labeled according to the protocol above. Anti-GFP primary polyclonal antibodies from rabbit (#ABIN121945, purchased via antibodies-online.com ) were used for labeling of peGFP.Vpr fusion protein together with the same secondary anti-rabbit IgG antibodies than were used for CHMP2A. STORM sample preparation followed a slightly modified version of the protocol for microtubule immunostaining for STORM imaging described in [ 80 ]. HeLa cells were fixed not earlier than 14–15 h post transfection using 4% (v/v) paraformaldehyde solution (PFA, Electron Microscopy Sciences) for 15 min. Cells were permeabilized with 1% (v/v) Triton X-100 for 2 min. The cells were again rinsed twice with PBS. Unspecific antibody binding was blocked by incubating the sample with a buffer containing 3% bovine serum albumin (BSA) and 0.2% Triton X-100 for 30 min followed by immunostaining with primary antibodies for 60 min. In case of additional labeling with secondary antibodies, the sample was then stained with labeled secondary antibodies for 45 min. Finally, post-fixation was done by treating the cells with 4% (v/v) paraformaldehyde solution in PBS for 10 min. STORM and dSTORM measurements were carried out using a glucose oxidase based oxygen scavenging buffer with mercaptoethylamine (Fluka) as reducing agent as described in [ 80 ]. PALM measurements were carried out using PBS buffer.

Data acquisition

TIRF imaging and all PALM, STORM and dSTORM measurements were carried out on a combined TIRF and wide-field (WF) microscope as depicted in S1 Fig. A 561-nm diode-pumped solid-state laser (CrystaLaser, Reno, NV, USA) was used to excite Gag.mCherry and the red state of Gag.mEos. For photoconversion of mEosFP, we used a 405-nm diode laser (LuxX 405–120, Omicron Laserage, Rodgau, Germany). Excitation of GFP- and YFP-fusion proteins and activation of Cy5 for STORM imaging was achieved by exciting Alexa488 with a 488-nm diode laser (LuxX 488–60, Omicron Laserage, Rodgau, Germany). Activated Cy5 was excited by a 642-nm diode laser (PhoxX 642, Omicron Laserage, Rodgau, Germany). We performed wavelength selection by coupling the different laser lines into an acousto-optic tunable filter (TF525-250-6-3-GH18A, Gooch & Housego, Ilminster, UK). A set of two lenses then expanded the excitation beam in order to increase the field-of-view illuminated by TIRF excitation, before the light was focused on the back focal plane of a 100x N.A.

Apo

TIRF oil immersion lens (Nikon, Tokyo, Japan). A rectangular glass prism introduced into the beam path after the focusing lens allowed switching between wide-field to TIRF imaging by changing the displacement of the beam path relative to the optical axis of the objective. A dichroic mirror (Di01-R405/488/561/635-25x36, Semrock, Rochester, NY, USA) then directed the excitation light into the microscope body, which consisted of a home-built microscope stage built out of Inwar to reduce thermal drift. After passing the dichroic mirror, the collected fluorescence signal passed a set of emission filters to select a specific fluorescence wavelength: a 670/30 bandpass filter was used for Cy5 (Laser 2000), a 535/22 bandpass filter for GFP and YFP (FF01-535/22-25, Semrock, Rochester, NY, USA) and a 593/40 bandpass filter (FF01-593/40-25, Semrock, Rochester, NY, USA) for the red state of Gag.mEos and Gag.mCherry. Finally, the fluorescence signal was projected onto an EMCCD camera (DU860D-CS0-BV, Andor, Belfast, UK). The resulting pixel size was 120 nm. ESCRT proteins were measured with either STORM or dSTORM, where a movie stack comprising 10,000 frames was acquired at a frame rate of 20 Hz, where each activation cycle consisted of one activation frame (λ exc = 488 nm), followed by 9 imaging frames (λ exc = 642 nm). Laser power at the output of the objective was 50 mW for imaging. For activation, the power of the blue laser was ~0.5 μW at the beginning of data acquisition for STORM and 5.0 μW for dSTORM and successively increased over the course of the experiment up to ~ 2 mW. This was necessary to counteract gradual bleaching of the activator dye. For PALM measurements of viral Gag protein, an analogue protocol was used with λ exc = 405 nm for activation (increasing power starting at 5.0 μW to 2 mW) and λ exc = 561 nm (1.1 mW) for imaging.

Data analysis

The data analysis protocol is based on the procedure described by Rust et al . [ 81 ] and by Bates et al . [ 82 ] and implemented into a self-programmed analysis software using MATLAB (MathWorks, Natick, MA, USA). The same algorithms were used for PALM, STORM and dSTORM experiments. First, structures in the fluorescent image representing local maxima were isolated in a square 11 x 11 pixel window. The point spread function of the fluorescent molecule was evaluated by fitting these regions to a continuous ellipsoidal 2D Gaussian using Levenberg-Marquard’s nonlinear least-squares algorithm. The intensity I of the Gaussian distribution at coordinates (x, y) is given by: I x , y = A + I 0 e 0.5 * - ( x - x 0 σ x 2 - ( y - y 0 σ y 2 where A is the background intensity, I 0 the maximum amplitude of the distribution, x 0 and y 0 the coordinates of the centroid and σ x and σ y the standard deviations in x and y-direction, respectively. Sample drift was corrected by pixel-wise cross-correlation of each frame n of the image stack with the first frame of this stack. The normalized cross-correlation function G n x , y for each frame is given by: G n x , y = ∑ i ∑ j I 1 i , j I n i + x , j + y ∑ i ∑ j I 1 i , j 2 ∑ i ∑ j I n i + x , j + y 2 0.5 The intensity I 1 from the first frame at coordinates ( i , j ) is correlated with the intensity I n from the n th frame at coordinates ( i + x , j + y ) . The respective drift in x - and y is given by the position in ( x , y ) of the maximum of G n x , y . In order to reduce fluctuations caused by fluorophore blinking, the obtained drift function was fitted to a polynomial and the fit function used for drift correction. We rejected all molecules where the Gaussian function fitted to the point spread function showed an ellipticity, E , higher than 15% ( S11A Fig. ). E depends on the Gaussian standard deviations in x and y -direction (σ x and σ y ) and is defined as: E = | σ x − σ y σ x + σ y | Points that appeared in two or more consecutive frames within a distance smaller than 1 px (120 nm) were considered as originating from the same fluorescent molecule. The positions determined from individual frames where the same molecule was observed were averaged for rendering of the final super-resolution image. Diffraction limited spots that appeared for only one frame or molecules with less than 300 detected photons were discarded. Localization displacements of single fluorophores molecules were used to estimate the actual resolution of our system by means of the FWHM value of the Gaussian function fit to the resulting displacement histograms in the x - and y -direction respectively ( S11B Fig. ). The average STORM image resolutions for the different analyzed proteins are summarized in S2 Table . For the final STORM image rendering with a pixel size of 12 nm in the super-resolution images, each localized molecule was represented by a 2D Gaussian function with a fixed amplitude I 0 = 1000 counts and a fixed standard deviation σ x = σ y = 1.2 px = 14.4 nm. Diffraction limited TIRF images of (d)STORM/PALM measurements, which were used to identify colocalizations of ESCRT and HIV-1 buds, were emulated by building the average time projection of the acquired image stack. An self-written image analysis algorithm was developed in ImageJ Macro language [ 83 ] and used to consistently assess the number of HIV-1 assembly sites and clusters and determine their colocalization. Images were individually analyzed as follows: First, a convolution filter (Gaussian blur) followed by background subtraction (“rolling ball” algorithm [ 84 ]) were applied. Next, point objects were selected based on their intensities (local maxima) and a multi-point selection was created. Objects corresponding to either assembly sites or clusters were then segmented by a watershed approach. As a last step, center of brightness, distribution of intensities and area of each object were measured. These readouts were collected, analyzed accordingly and generated the statistical information presented in Figs. 2 – 5 . Gag assemblies with an area > 0.860 μm² were excluded from further evaluation. ESCRT clusters were counted by rendering a STORM image with a pixel size equal to the widefield pixel size of 120 nm and all clusters with an intensity lower than 5,000 counts were discarded as well as all clusters where no distinct structure could be obtained in the final STORM image with a pixel size of 12 nm. Object sizes in the final STORM or PALM images were either estimated by means of the average full-width at half-maximum (FWHM) of 1-D Gaussians fitted to two orthogonal 1-D cross-sections through the middle of the respective cluster as demonstrated in S2B , S4B , S5B , S6D and S7B Figs., or alternatively using Ripley’s L-test [ 49 ]. In the case of estimation of the size of the cloud surrounding ALIX clusters, the points contributing to the central clusters were excluded from Ripley’s analysis to achieve more accurate results for the size of the cloud (see S5C Fig. ).

Supporting Information S1 File Supplementary results. Size of HIV-1 budding sites using Photoactivation Localization Microscopy (PALM) (DOCX) Click here for additional data file. S1 Fig Schematic of the custom microscopy setup used for super resolution imaging (green: excitation pathway; yellow: detection pathway). (EPS) Click here for additional data file. S2 Fig Super-resolution imaging of HIV-1 assembly sites. Cells were transfected with equimolar amounts of HIV mEos and untagged HIV. ( A ) Left panel . Time projected TIRF image of HIV mEos showing various HIV-1 assembly sites. Scale bar: 10 μm. Middle panel . A zoomed-in TIRF image of a single HIV-1 assembly site highlighted by the red box in panel A . Scale bar: 500 nm. Right panel . Drift-corrected super-resolution PALM image of the HIV-1 assembly site shown in panel B . Scale bar: 500 nm. ( B ) Gaussian fits of the cross-sections (red lines) through the PALM reconstructed point cluster shown in the right panel of A . ( C ) The size distribution of N = 159 HIV mEos clusters determined from the FWHM of the Gaussian fit of the super-resolution images is shown with an average size (FHWM) of 116 ± 36 nm. (EPS) Click here for additional data file. S3 Fig eGFP.vpr measurements. In all images, eGFP.vpr expressed in HeLa cells was detected by immunostaining 18 – 20 h post transfection or directly via the eGFP fluorescence by TIRFM. ( A ) TIRFM images of a HeLa cell expressing both pEGFP.vpr and Gag.mCherry, and immunostained using antibodies against GFP, visualized in the eGFP channel (left panel), mCherry channel (middle panel) and anti-GFP (right panel) channel. Overlays of images demonstrating the colocalizations of eGFP.Vpr and HIV mCherry (B) and colocalizations of eGFP.vpr and anti-GFP ( C ) (magenta: Gag.mCherry, yellow: eGFP.vpr, green: anti-GFP) are shown, scale bars: 10 μm. ( D ) Size distribution of eGFP.vpr clusters colocalizing with Gag.mCherry determined from the full-width at half-maximum (FWHM) obtained by fitting a Gaussian function to the cross-section through the respective eGFP.Vpr cluster. The average cluster size (FHWM) is 56 ± 12 nm. N represents the number of analyzed colocalizing clusters. (EPS) Click here for additional data file. S4 Fig Tsg101 supplementary information. ( A ) An untransfected HeLa cell immunostained with anti-Tsg101 primary plus labeled secondary antibody binding to Tsg101 primary antibodies, scale bar: 10 μm. ( B ) The size of Tsg101 clusters was determined from the average full-width at half-maximum (FWHM) obtained by fitting cross sections of the cluster to two orthogonal 1-D Gaussian functions (indicated by the red lines). Scale bar: 500 nm. ( C ) TIRF images of Tsg101 in HIV:HIV mCHerry late- expressing cells. Left panel . TIRF image of HIV mCherry late- assembly sites. Middle Panel . Average time projection of TIRF image series acquired for super-resolution imaging of Tsg101 in HIV:HIV mCherry late- expressing cells. Right panel : Overlay of the left (HIV mCherry late-, magenta) and middle (anti-Tsg101, green) panels. Scale bars: 10 μm. No colocalizations are observed. ( D ) Left panel . A merged time-projected TIRF image of immunostained Tsg101-FLAG (green) in a HeLa cell expressing Tsg101-FLAG-IRES-GFP together with HIV mCherry (magenta) and HIV prior to STORM analysis. Middle panel . A zoomed-in image of the selected Tsg101-FLAG cluster highlighted in grey in the left panel colocalizing with an HIV assembly site. Right panel . The corresponding drift-corrected STORM image of the Tsg101-FLAG cluster. Scale bars: 500 nm. ( E ) The size distribution of all Tsg101-FLAG structures colocalizing with HIV mCherry with an average cluster size (FWHM) of 50 ± 11 nm. ( F ) Size distribution of all non-colocalizing Tsg101-FLAG clusters in cells co-expressing HIV mCherry (wildtype) with an average cluster size (FWHM) of 51 ± 10 nm. N represents the number of analyzed colocalizing clusters. (EPS) Click here for additional data file. S5 Fig ALIX supplementary information. ( A ) An untransfected HeLa cell immunostained with anti-labeled ALIX primary antibodies. All zoomed-in insets show drift-corrected super-resolution STORM images of the respective clusters highlighted in red. Scale bars: 10 μm (large image), 200 nm (insets). ( B ) Size characterization of the central, condensed spot of ALIX colocalizing with HIV mCherry ; two orthogonal cross-sections through the center of the spot are fitted to a 1D Gaussian function whose mean FWHM represents the structure size. Scale bar: 500 nm. ( C ) Size characterization of the diffuse cloud-like ALIX structure from the example shown in panel B ; the central cluster is masked (inset) and the cloud analyzed by applying Ripley’s L-function, whose maximum at 187 nm indicates the spreading size of the cloud cluster. Scale bar: 500 nm. (EPS) Click here for additional data file. S6 Fig CHMP4B-HA control experiments. HeLa cells transfected with ( A ) only HIV:HIV mCherry or ( B ) only CHMP4B-HA and immunostained with anti-HA primary antibodies. TIRF images of the HIV mCherry channel and the CHMP4B-HA channel are shown in the left and right panels, respectively. All scale bars: 20 μm. ( C ) A HeLa cell transfected with HIV:HIV mCherry late- and CHMP4B-HA. Left panel . A TIRFM image of HIV mCherry late- assembly sites. Middle panel . A time-projected TIRFM image of the CHMP4B-HA channel before the STORM localization analysis. Right panel . An overlay of the two channels with HIV mCherry channel shown in magenta and the CHMP4B-HA channel in green. No colocalizing structures are observed in the merged image. Scale bars: 10 μm. ( D ) Left panel . A STORM image of a CHMP4B cluster colocalizing with wildtype HIV mCherry . Scale bar: 500 nm. The size of CHMP4B-HA clusters was determined using the FWHM obtained by fitting 1D Gaussian functions to the orthogonal cross-sections. Right panel . The cross-sections, indicated by the red lines in the left panel, and the Gaussian fits are shown for comparison. (EPS) Click here for additional data file. S7 Fig CHMP2A control experiments and analysis. ( A ) An untransfected HeLa cell immunostained with anti-CHMP2A primary and appropriate labeled secondary antibodies. All zoomed in images show drift-corrected super-resolution STORM images of the respective clusters highlighted in red. Scale bars: 10 μm (large image), 200 nm (insets). ( B ) A STORM image of a CHMP2A cluster. Scale bar: 500 nm. The size of CHMP2A clusters was determined from the mean of the full-width at half-maximum (FWHM) obtained by fitting a 1D Gaussian function to each of the cross-sections indicated by the red lines. ( C ) Left panel . A merged time-projected TIRF image of immunostained CHMP2A (green) in a HeLa expressing dominant-negative Vps4A mutant Vps4A-E228Q-mCherry together with HIV mCherry (magenta) and HIV prior to STORM analysis. Middle panel . A zoomed-in image of the selected CHMP2A cluster highlighted in grey in the left panel colocalizing with an HIV assembly site. Right panel . The corresponding drift-corrected STORM image of the CHMP2A cluster. Scale bars: 500 nm. ( D ) The size distribution of all CHMP2A structures in cells expressing the dominant-negative Vps4A-E228Q-mCherry mutant colocalizing with HIV mCherry . The average cluster size (FWHM) is 66 ± 23 nm. ( E ) Size distribution of all non-colocalizing CHMP2A clusters in cells co-expressing HIV mCherry (wildtype) and Vps4A-E228Q-mCherry with an average cluster size (FWHM) of 65 ± 22 nm. N represents the number of analyzed colocalizing clusters. (EPS) Click here for additional data file. S8 Fig Experiments with YFP-tagged Tsg101. ( A ) A HeLa cell expressing the ESCRT fusion protein YFP-Tsg101 and HIV:HIV mCherry . Signal from the YFP channel ( left panel ), HIV:HIV mCherry channel ( middle panel ) a time-projection of the immunostained Tsg101 signal labeled with primary anti-Tsg101 antibodies and fluorescently labeled secondary antibodies ( right panel ) are shown. Scale bars: 10 μm. ( B ) Left panel . A merged TIRFM image of HIV mCherry (magenta) and immunostained YFP-Tsg101 (green). Scale bar: 10 μm. A high number of colocalizations are observed in the field of view. Right panel . A STORM image of the condensed, circular Tsg101 structure highlighted in the middle panel in grey. Scale bar: 500 nm. ( C ) The size distribution of immunostained YFP-Tsg101 clusters colocalizing with HIV mCherry with an average size (FWHM) of 60 ± 19 nm. ( D ) A HeLa cell expressing the ESCRT fusion protein YFP-Tsg101 and Gag:Gag.mCherry. Signal from the YFP channel ( left panel ), Gag:Gag.mCherry channel ( middle panel ) a time-projection of the immunostained Tsg101 signal labeled with primary anti-Tsg101 antibodies and fluorescently labeled secondary antibodies ( right panel ) are shown. Scale bars: 10 μm. ( E ) Left panel . A merged TIRFM images of Gag.mCherry (magenta) and immunostained YFP-Tsg101 (green). A high number of colocalizations are observable in the field of view. Scale bar: 10 μm. Right panel . A STORM image of the condensed, circular Tsg101 structure highlighted in the middle panel in grey. Scale bar: 500 nm. ( F ) The size distribution of immunostained YFP-Tsg101 clusters colocalizing with Gag.mCherry with an average size (FWHM) of 60 ± 10 nm. N represents the number of analyzed colocalizing clusters. (EPS) Click here for additional data file. S9 Fig Dual-color super-resolution imaging of CHMP4B and HIV-Gag. ( A ) Overlay of the average time projections of independent TIRFM image series for HIV mEos and CHMP4B-HA (magenta: HIV mEos , green: CHMP4B-HA), scale bar: 2 μm. ( B ) A zoomed-in image of the single colocalizing structure ( upper panels ) and corresponding super-resolution PALM (mEOS) or STORM (CHMP4B-HA) images, respectively ( lower panels ), scale bars: 500 nm. ( C ) The Gaussian fit of cross-section profiles through the respective PALM/STORM images yielding a FWHM of 133 nm for HIV mEos ( left graph ) and 57 nm for CHMP4B-HA cluster ( right graph ). (EPS) Click here for additional data file. S10 Fig Position of ESCRT protein clusters relative to the HIV-1 bud. Overlay of super-resolution images of ALIX and ESCRT proteins Tsg101, CHMP4B-HA and CHMP2A (green) and the corresponding TIRF images of the respective evolving HIV-1 buds (magenta) reveal the relative position of the ESCRT protein cluster relative to the viral bud. (EPS) Click here for additional data file. S11 Fig Ellipticity threshold and determination of image resolution. ( A ) Ellipticity distribution of fitted point spread functions and applied ellipticity threshold of 0.15. ( B ) Determination of the resolution of the system by displacement analysis of all summarized single localizations in the x -direction. The resolution is given by the FWHM of the localization displacement distribution (here: 38.3 nm). See S2 Table for the localization precision of all experiments. (EPS) Click here for additional data file.

S1 Table

Summary of results for analyzed ESCRT-proteins Tsg101, ALIX, CHMP4B and CHMP2A regarding number of analyzed cells, cluster and colocalization events. (PDF) Click here for additional data file.

S2 Table

STORM image resolution for different analyzed proteins. (PDF) Click here for additional data file.