Setup of the assay procedure and optimization of assay conditions
The aim of this study was to develop a simple and direct assay for HIV-1 release based on our previously characterized fluorescently labelled HIV derivative [19, 20], which should allow the quantitation of particles in tissue culture supernatants from virus producing cells by direct measurement of fluorescence intensity. Towards this end, we first optimized the detection conditions to increase the sensitivity of the measurements in tissue culture supernatants. Due to intrinsic fluorescence of tissue culture media components at the wavelengths used for detection of fluorescent proteins, optimization of the assay initially required the selection of the fluorophore best suited for detection against medium background. Fluorescence measurements of serial dilutions of bacterially expressed, purified enhanced green, cyan and yellow fluorescent proteins (eGFP, eYFP, eCFP) as well as monomeric red fluorescent protein [22] in PBS and various tissue culture media indicated that enhanced yellow fluorescent protein eYFP was most suitable among these fluorophors for detection against complete medium background. EYFP concentrations above10 ng/ml were detectable over medium background (Additional file 1, Figure S1). Omission of phenol red or fetal calf serum from the medium did not significantly alter the sensitivity of detection.
Based on these findings, we constructed an HIV derivative carrying the eyfp gene within the gag open reading frame, analogous to the previously described construct pCHIVeGFP [20]. The fluorescent protein was inserted between the matrix (MA) and capsid (CA) domains of the Gag polyprotein, upstream of the proteolytic processing site between MA and CA (Figure 1A). Transfection of cells with pCHIV and derivatives results in expression of all viral proteins, except for Nef, under the control of the CMV promoter and to formation and release of entry competent virus like particles (VLPs). The genome encoded by pCHIV is replication incompetent due to deletion of the complete viral long terminal repeat regions essential for reverse transcription and integration.
As shown in Figure 1B, cells transfected with pCHIVeYFP displayed the expected yellow fluorescence. Furthermore, HeLaP4 cells which express the receptor and co-receptors of HIV formed large syncytia upon transfection with pCHIV derived constructs, indicative of HIV Env protein expression. Cell lysates as well as particle preparations from the supernatant of transfected cells were prepared and analyzed by immunoblot (Figure 1C). Analogous to our results obtained with the related construct pKHIVeGFP [19], the Gag.eYFP fusion protein expressed in the viral context was processed upon particle maturation by the viral protease to yield MA.eYFP, and released pelletable particles containing the MA.eYFP fusion protein in a late domain dependent manner (Figure 1C and data not shown). Accordingly, particles should be detectable in the tissue culture supernatant by measurement of eYFP fluorescence intensity.
Specific detection of released particles by fluorescence measurements
The exploitation of fluorescent MA for the quantitation of specific particle release from cells transfected with pCHIVeYFP by fluorescence measurements required appropriate controls. The fluorescence readout - analogous to p24 determination by ELISA - does not necessarily reflect the amount of complete particles released, but rather quantitates the amount of the main particle component Gag present in tissue culture supernatants. Consequently, unspecific release of free Gag, p24 CA or MA.eYFP from damaged or lysed cells could impair the accuracy of results obtained by either ELISA or fluorescence based measurements. To control for unspecific protein release due to cell damage, we used cells expressing unfused eYFP protein, which is not secreted from intact cells, but will be detected in the medium when the sample contains cell debris or cytoplasmic content from lysed cells. Secondly, a positive control for the successful inhibition of release was established. For this purpose, we generated a release deficient variant of the fluorescently labelled HIV derivative by replacing the 'late domain' motif PTAP within the p6 domain of Gag.eYFP with the amino acid sequence LIRL. The late domain is known to mediate interaction with the host cell ESCRT machinery [see [10] for review] and its mutation to LIRL has been previously shown to affect HIV release efficiency [23]. An inhibitory compound which directly blocks viral late domain - ESCRT interaction or an siRNA which downmodulates expression of a late domain interacting host factor can be expected to yield a comparable decrease in release efficiency as the mutation of the late domain, while compounds or siRNAs targeting the Gag assembly process may theoretically accomplish even stronger reduction.
Expression of the Gag.eYFP.late(-) fusion protein and the release deficiency of this construct was confirmed by immunoblot analysis of cell lysate and pelleted tissue culture supernatant (Figure 1C). While comparable amounts of Gag.eYFP and Gag.eYFP.late(-) were detected in cell lysates (compare lanes 3 and 4), the amount of Gag derived products in pelleted particles was significantly decreased in the case of the late-variant (compare lanes 7 and 8). Free eYFP was expressed in transfected cells (lane 1), but not detected in pelleted supernatant (lane 5). We then determined the relative release efficiency of Gag.eYFP.late(-) by fluorescence measurements.
Figure 1D shows the result of a comparison of release efficiencies between 293T cells transfected with plasmids pEYFP-C1, pCHIV and pCHIVeYFPlate(-). At 48 h post transfection tissue culture supernatants were harvested and cells were lysed by detergent treatment as described in methods. Supernatants from particle producing cells displayed typical spectroscopic characteristics of eYFP with excitation and emission maxima at 512 nm and 527 nm (Additional file 2, Figure S2). Relative release efficiencies were calculated by dividing fluorescence intensities measured in tissue culture supernatants by the corresponding total fluorescence intensities measured in supernatants and cell lysates. Cells transfected with plasmid peFYP-C1 displayed high intracellular fluorescence but the relative amount of eYFP released into the supernatant was low in comparison to cells transfected with pCHIVeYFP. Cells expressing the late domain defective variant of the labelled virus reproducibly showed an approximately five-fold reduction of relative release efficiency as compared to cells expressing the wild-type variant.
Sensitivity of measurements
Our previous characterization of eGFP labelled HIV derivatives had revealed that - while bulk release efficiency of labelled virus was comparable to wild-type - cells expressing the labelled HIV derivative displayed electron dense aggregates at the plasma membrane, suggesting altered release kinetics; furthermore released particles displayed reduced infectivity. Full wild-type particle infectivity and budding site morphology was restored by co-transfection of an equimolar amount of wild-type HIV proviral plasmid, leading to the formation of mixed particles carrying 50% of labelled Gag [19]; assuming an average number of approximately 2400 molecules of Gag per particle [24], particles contain thus approximately 1200 molecules of fluorescent protein. For this reason, equimolar mixtures of eYFP labelled proviral derivatives with the corresponding non-labelled provirus were used in all further experiments.
In order to determine the sensitivity of fluorescence based particle quantification, VLPs were prepared from the tissue culture supernatant of cells co-transfected with an equimolar mixture of pCHIV and pCHIVeYFP by ultracentrifugation through a sucrose cushion. The amount of CA p24 present in the particle preparation was determined by ELISA, which represents a currently used standard method for HIV particle quantification. Since both eYFP and p24 CA are expressed from pCHIVeYFP as parts of the Gag polyprotein, they are translated and packaged into VLPs at equimolar amounts; in the case of the mixed particles used here, one molecule of eYFP will thus correspond to two molecules of CA. Serial dilutions of fluorescent particles in complete medium were set up and their relative eYFP fluorescence intensity was determined (Figure 2). The lowest particle concentration detectable above background by fluorescence measurements in complete medium corresponded to approximately 15-30 ng CA/ml (equivalent to ~8-15 ng eYFP/ml), which was in good agreement with the results from measurements using purified fluorescent proteins. Omission of FCS or phenol red from the medium or replacement of the medium by PBS only resulted in a minor increase in sensitivity (data not shown). From these and similar data sets we concluded that robust fluorescence intensity quantitation can be performed for particle concentrations corresponding to 50 ng p24/ml or higher. This sensitivity is significantly lower than that of advanced ELISA based detection systems [25]. However, this does not affect the usefulness of the system for the intended applications, since the amounts of particles released under optimized assay conditions were in a range easily detectable by fluorescence measurements (see below).
Assay validation and time course of particle release
We then analysed the time course of fluorescence release from cells transfected with equimolar mixtures of either wild-type or late domain defective variants of pCHIV/pCHIVeYFP, respectively. Samples of tissue culture supernatant were removed at consecutive time points between 24 and 48 h post transfection and subjected to fluorescence measurements. Cells transfected with peYFP-C1 were used as a negative control. In order to investigate whether budding of HIV from the membrane could promote the release of free eYFP from cells, we included cells transfected with a mixture of pCHIV and peFYP-C1 as an additional control. At 48 h post transfection, cells were lysed by detergent treatment and the fluorescence intensity of the lysate determined in order to normalize for transfection efficiency and Gag or eYFP expression in the different samples. As shown in Figure 3A, fluorescence in the supernatant above background was detectable at 24 h post transfection with pCHIV/pCHIVeYFP (filled circles) and increased over the following hours. Again, release from the late(-) variant (open circles) was found to be diminished. While fluorescence in the supernatant of peYFP-C1 transfected cells (open triangles) remained at background levels, release of low levels of free eYFP was promoted in the presence of HIV expression (filled triangles). This might be explained by unspecific packaging of cytosolic eYFP into budding virus particles, since we observed that in this case release was also influenced by the presence of a functional HIV late domain (data not shown).
Results obtained from fluorescence measurements were validated by determination of p24 CA in the same samples by ELISA. Again, we calculated the amount of p24 in the supernatant relative to the amount measured in cell lysates at 48 h. Results are shown in Figure 3B. Similar concentrations of p24 were detected in supernatants from cells transfected with pCHIV/pCHIVeYFP (filled circles) and with pCHIV/peYFP-C1 (filled triangles), demonstrating that insertion of eYFP within Gag did not affect release efficiency; p24 concentrations measured corresponded to ~70 ng/ml at 24 h post transfection and ~0.5 μg/ml at the end of the experiment, respectively. Evaluation by ELISA yielded similar results for pCHIV/pCHIVeYFP (filled circles) and pCHIVlate(-)/pCHIVeYFPlate(-) (open circles) as the fluorescence based assay (compare Figure 3A and Figure 3B), confirming the validity of the fluorescence assay method. A slightly larger deviation between parallel samples was observed in the ELISA experiment, presumably because this procedure involved more handling steps. The observation that the relative ratio of intra- to extracellular fluorescence differs somewhat between the two types of measurement is explained by the fact that unprocessed Gag (which represents the predominant form in cell lysates) is recognized less well than processed CA (the predominant form in particles), by immunoprecipitation [26] or ELISA. Even prior heat and detergent treatment of samples, which has been reported to make unprocessed Gag accessible to ELISA detection [27], did not confer full reactivity of the precursor form in our hands. Thus, data derived from the ELISA measurements are less suitable for reliable normalization of CA levels relative to intracellular Gag expression levels. In contrast, in vitro cleavage of immature fluorescent particles using purified PR did not result in a change in fluorescence intensity, demonstrating that the fluorescence based quantitation of Gag was independent of its processing status (data not shown). We concluded that the assay described here is suitable to monitor HIV-1 particle release and can be used to test inhibition of HIV-1 particle formation in tissue culture.
In order to define conditions yielding efficient late domain dependent release accompanied by low levels of unspecific release we compared the ratio between release of the wild-type and late(-) virus variants at the different time points. The ratio of late(-) particles released compared to wild-type particles increased at later time points, presumably due to unspecific Gag release from damaged or dead cells upon prolonged protein over-expression (Figure 3C and data not shown), indicating that harvest times of 33-36 h post transfection were optimal under the conditions used.
Adaptation of the assay to medium throughput formats
Further adaptation of the assay to a semi-automated multi-well format was performed to render the assay suitable for random screening. Two different protocols were established suitable for screening of chemical compound libraries or for siRNA screening approaches, respectively. For the purpose of compound screening we established a protocol for the bulk transfection of 293T cells. Briefly, 293T cells were harvested and transfected in suspension using equimolar mixtures of pCHIV/pCHIVeYFP, pCHIVlate(-)/pCHIVeYFPlate(-), or plasmid peYFP-C1 alone, respectively. Cell suspensions were then distributed into wells of a 96-well plate or 384-well plate, respectively, using an automated reagent dispenser and plates were incubated at 37°C. The addition of chemical compounds dissolved in DMSO was mimicked by the addition of a final concentration of 0.5% DMSO to the medium at 14 h post transfection. At 36 h post transfection, plates were briefly spun to pellet potential cell debris. Aliquots of tissue culture supernatants were transferred to fresh plates using an automated liquid handler, the remaining supernatant was removed from the cell layer and cells were lysed on the plate as described in methods. Relative fluorescence intensities of supernatants and cell lysates were determined using a TECAN Safire multi-well reader. Supernatants and lysates from untransfected cells were used to determine background fluorescence. As shown in Figure 4 and Additional file 3, Figure S3, this procedure yielded robust results in 96-well plates. A limited data set of 60 positive (cells transfected with pCHIV/pCHIVeYFP, 0.5% DMSO) and 36 negative (untransfected cells) control wells from a pilot screen including 6 replicate plates was used to estimate statistical effect size, yielding a Z factor [28] of 0.71 (Additional file 3, Figure S3). Mean fluorescence intensities in the supernatant of 176 wells from 88 plates (Figure 4A, grey bars) showed low plate-to-plate variability. The amount of eYFP labelled protein released into the supernatant by the late domain defective variant was reduced to 16% of the wild-type control. Very similar results were obtained when fluorescence intensities of the supernatant normalized to the total fluorescence measured in the supernatant and cell lysate were compared (Figure 4A, black bars). Thus, normalization for intracellular protein expression was not necessary to improve the accuracy of measurements in a standardized 96-well setup. In the context of a random screening assay, however, the parallel determination of intracellular eYFP intensity in a screening assay can serve to identify compounds which do not specifically interfere with HIV particle formation but rather exert a general effect on protein expression. Comparison of relative release efficiencies determined for labelled wild-type particles across rows 2-11 of 96-well plates in the absence of inhibitory compounds did not reveal any systematic aberration across the plate (Figure 4B). We conclude that the assay allows for medium throughput screening of potential inhibitors of HIV-1 assembly or release.
Transfer of the protocol to a 384-well format yielded less robust results (Figure 4C). Due to difficulties in reproducible preparation and measurement of cell lysates in the small well format, normalization of supernatant fluorescence to total fluorescence led to larger variability between wells (data not shown). While the assay in its present form could be used to determine relative amounts of Gag released into the supernatant using a 384-well format, further optimization would be required for a reliable parallel quantification of intracellular Gag.
Finally, a modified protocol was established to be used in screening of siRNA libraries for factors involved in HIV release. For this, 293T cells were distributed onto 96-well plates which had been pre-coated with an siRNA containing transfection mix using a previously described reverse transfection protocol [29, 30]. Dried pre-coated plates could be either used directly for knockdown experiments or be stored for up to 15 months without any loss of efficacy, thus ensuring improved reproducibility and comparability between different plates of the same batch as compared to liquid transfection of siRNA. 293T cells were seeded onto these plates and incubated in order to achieve target gene knockdown, followed by transfection with pCHIV/pCHIVeYFP and further incubation. In order to validate this approach we performed siRNA mediated knockdown of the cellular ESCRT protein Tsg101, which has been previously shown to impair HIV release [31]. At the conditions used, prolonged knock-down of Tsg101 resulted in reduced cell viability and cell death; to avoid this, we chose a relatively short incubation period of 18 h for the initial transfection with siRNA, followed by transfection with pCHIV/pCHIVeYFP. Incubation was then continued for 36 h before read out of fluorescence intensities in supernatant and cell lysates was carried out. As shown in Figure 5A, siRNA mediated knock-down of Tsg101 resulted in an approximately 4-5 fold reduction in release efficiency under these conditions, which was only slightly less pronounced than the reduction observed using the late(-) virus variant.
We have performed a pilot screen, in which we tested an siRNA library (silencer siRNAs, Ambion, Applied Biosystems) targeting approximately 740 human kinases with one individual siRNA per well and three distinct siRNAs per target gene. Since individual siRNAs from this library have not been tested for their silencing properties and time course of gene silencing, we chose longer incubation times of 30 h after the initial transfection with siRNA and 42 h after the second transfection. The results of this screen and the validation of primary hits will be described elsewhere. Exemplary for the reproducibility of results obtained using the established protocol, Figure 5B shows the correlation of relative release values (RFIsup/RFItotal) obtained for two replicate sets of 480 wells from the siRNA screen. The values obtained for RFIsup/RFItotal were slightly higher than those observed for the compound screen setup due to the modified assay procedure. Pearson's correlation coefficients between replicates of identical siRNA sets were generally found to be in the range of 0.7 - 0.8, indicating that the setup described here allowed us to obtain reproducible results in a 96 well format.