|Home | About | Journals | Submit | Contact Us | Français|
We introduce a fluorescent reporter for monitoring protein–protein interactions in living cells. The method is based on the Split-Ubiquitin method and uses the ratio of two auto-fluorescent reporter proteins as signal for interaction (SPLIFF). The mating of two haploid yeast cells initiates the analysis and the interactions are followed online by two-channel time-lapse microscopy of the diploid cells during their first cell cycle. Using this approach we could with high spatio-temporal resolution visualize the differences between the interactions of the microtubule binding protein Stu2p with two of its binding partners, monitor the transient association of a Ran-GTPase with its receptors at the nuclear pore, and distinguish between protein interactions at the polar cortical domain at different phases of polar growth. These examples further demonstrate that protein–protein interactions identified from large-scale screens can be effectively followed up by high-resolution single-cell analysis.
A mechanistic understanding of cellular biology requires a comprehensive knowledge about the protein interactions of the cell (Uetz et al, 2000; Gavin et al, 2002; Krogan et al, 2006; Yu et al, 2008; Breitkreutz et al, 2010). Split-protein sensors comprise a family of related techniques that contributed in small- and large-scale experiments to this cumulative endeavour (Miller et al, 2005; Tarassov et al, 2008; Hruby et al, 2011; Lowder et al, 2011; Stynen et al, 2012). The Split-Ubiquitin method (Split-Ub) is the prototype of these techniques (Müller and Johnsson, 2008; Stynen et al, 2012). Here, the N- and C-terminal halves of Ubiquitin (Nub and Cub) are coupled separately to the two proteins under study. Upon interaction of the fusion proteins, Nub and Cub are forced into close proximity and reassemble into a native-like Ubiquitin (Ub). The native-like Ub is recognized by Ub-specific proteases that cleave off a reporter protein that was genetically attached to the C-terminus of Cub (Johnsson and Varshavsky, 1994).
The cleavage of the Split-Ub reporter orothic acid decarboxylase (Ura3p, CRU) from Cub leads to a qualitative difference in bulk cell growth (Wittke et al, 1999). This and other proteome-wide interaction techniques produce binary protein–protein interaction maps. The information encoded in these maps could fundamentally transform our understanding of cellular processes. However, to be effectively used in cell biology these networks would need to acquire a spatial as well as temporal dimension in order to place the interactions into a functional and cellular context (Alexander et al, 2009). The reason for the prevalence of binary interaction maps is mainly technical. Currently, robust and easy-to-use approaches for the characterization of cellular protein interactions in space and time are not available.
Here, we report on a new method based on Split-Ub and two spectrally different fluorescent proteins (SPLIFF) to monitor the interaction between two proteins during the cell cycle with high spatial and temporal resolution. We further show that SPLIFF can bridge large-scale protein interaction screens with high-resolution single-cell analysis.
To create a robust and sensitive fluorescent reporter for protein interactions, we designed a Split-Ub module where Cub is sandwiched by two spectrally different fluorescent proteins (Cherry-Cub-GFP, CCG) (Figure 1A). Coupling CCG to the C-terminus of protein Y (Y-CCG) reveals the cellular localization of the fusion protein. Interaction of Y-CCG with a Nub-coupled interaction partner X will result in the cleavage of GFP from Y-CCG to create Y-CC. The liberated GFP is subsequently degraded. As Cherry stays attached to Y, the ratio of red to green fluorescence serves as a ratiometric reporter of protein–protein interactions and constitutes the actual readout of SPLIFF. To improve the spatial and temporal resolution of the assay, we defined the start of each reaction by fusing two yeast cells of opposite mating type, each expressing one half of the Split-Ub sensor (Figure 1B). The conversion of a Cherry/GFP-labelled (Y-CCG) into a Cherry-labelled fusion protein (Y-CC) is then recorded online by two-channel time-lapse fluorescence microscopy. This strategy is borrowed from chemical stopped-flow experiments where the kinetics of chemical reactions is recorded after forcing the reagents into a single chamber.
In the following experiments, we demonstrate the applicability of SPLIFF by investigating protein interactions that occur at different cellular locations over widely different time periods and by providing examples of successful transitions from large-scale to single-cell analysis.
The interaction between the nucleolar protein Net1p and the NAD-dependent histone deacetylase Sir2p exemplifies the experimental approach (Straight et al, 1999; Figure 2). The CRU or CCG modules were attached in frame behind the ORF of NET1 by homologous recombination. The Nub module was fused 5′ to the SIR2 ORF in α cells. The expression of the Nub fusion was controlled by the PCUP1 promoter and could be adjusted by varying the levels of copper in the medium. The growth on 5-fluoroorotic acid (5-FOA) of cells co-expressing Net1CRU and Nub-Sir2p revealed the interaction in the chosen configuration of Nub and Cub attachment (Figure 2B). Subsequent time-lapse microscopy of the diploid cells originating from the mating of Net1CCG- and Nub-Sir2p expressing a- and α-cell visualized the course of interaction in the nucleus from the a-cells. The reaction was completed within 20min even before nuclear fusion has occurred (Figure 2C and D; Supplementary Movie 1).
To test the workflow for transferring interactions revealed by large-scale interaction experiments into single-cell analysis, we screened Net1CRU against an array of 382 different Nub fusions (Nub array) (Hruby et al, 2011). Among others Ubc9p, Cdc14p, and Fkh1p were identified as further Net1p-binding partners (Table I; Visintin et al, 1999). The interactions of Net1p with all three Nub fusions were subsequently analyzed by SPLIFF. The kinetic profiles of the Net1CCG conversions were very similar to the profile induced by Nub-Sir2p (Supplementary Figure 1). Nub-Cdc14p induced a slightly but significantly slower conversion of Net1CCG than Nub-Sir2p (Supplementary Figure 1; Supplementary Movie 2). Nub-Pea2p, a Nub-fusion that was not identified by the large-scale experiment, did also not interact in the SPLIFF analysis with Net1CCG (Supplementary Figure 1).
To identify the rate-limiting steps of these reactions, we compared the accumulation of GFP-Cdc14p in the nuclei of a-cells with the Nub-Cdc14p-induced Net1CCG conversion (Figure 3). In the first 10min after mating, the kinetic profiles of both reactions were similar. However, whereas the best fit of the Net1CCG conversion is described by a sigmoid curve, the nuclear accumulation of GFP-CDC14p proceeded linear and consequently slower (Figure 3B). We conclude that the conversion of Net1CCG to Net1CC is not limited by the association rate of the fusion proteins and the subsequent Ub assembly and degradation of the attached GFP. The sigmoid shape indicates that the interaction between Cdc14p and Net1p is dynamic. Nub-Cdc14p exchanges binding partners and thereby catalytically converts Net1CCG into Net1CC.
The nuclear pore complex of yeast consists of >30 different proteins that are organized into three different layers. Integral membrane proteins (POMs) anchor a ring of coat nucleoporins followed by adaptor nucleoporins. The adaptors position the channel nucleoporins to regulate the transfer of cargo between the nucleus and the cytoplasm (Figure 4A; Aitchison and Rout, 2012). Nup49p and Nic96p together with Nsp1p and Nup57p form the Nic96p sub-complex (Figure 4A). Nic96p is a member of the adaptor nucleoporins whereas Nup49p, Nsp1p, and Nup57p belong to the FG repeat-bearing channel nucleoporins. FG repeats interact transiently with importins and exportins that are bound to cargo and the Ran-GTPase Gsp1p during their shuttle across the pore (Figure 4A; Aitchison and Rout, 2012). After mating, the nuclear membranes of the two yeast nuclei fuse and the nuclear pore complexes from both cells mix (Bucci and Wente, 1997). According to the growth rate of the co-expressing diploids Nup49CRU interacted strongly with Nub-Nup57p, -Nic96p, -Nsp1p, and only very weakly with Nub-Gsp1p (Supplementary Figure 2). We measured the kinetic profiles of these interactions with Nup49p as CCG fusion. The SPLIFF analysis of Nup49CCG visualized for the first time the interaction between a Ran-GTPase and a FG-repeat protein in a living cell (Figure 4B, C and F; Supplementary Movie 3). Nub-Gsp1p converted Nup49CCG much faster than the Nub fusions of Nic96p or Nsp1p (Figure 4F; Supplementary Figure 2; Supplementary Movie 4). The nuclear resident Nub-Cdc14p did not interact with Nup49CCG (Figure 4D and E). Contrary to the transport factors, constitutive members like Nsp1p or Nic96p are known to be stably incorporated into the nuclear pore complex (Rabut et al, 2004). We therefore surmise that the slow exchange of the unlabelled proteins against the corresponding Nub fusions of Nsp1p or Nic96p is rate limiting for the interaction with Nup49CCG. Notably, the interaction between Nup49CCG and its Nub-labelled interaction partners started only after nuclear fusion was completed (Supplementary Movie 4).
The proteins of the polar cortical domain (PCD) form a protein network below the plasma membrane. The composition of the PCD in yeast is not fully defined and highly dynamic (Gao et al, 2011). The PCD is located first at the mating projection (PCDI), later at the site of bud growth (PCDII), and finally at the site of cytokinesis (PCDIII). During these transitions, the components of the PCDs are dissolved before they are recruited to their new site (Figure 5A). Measuring the interactions between members of the PCD thus requires tracking of the CCG fusion in their different locations during the cell cycle. Spa2p is member of the polarisome complex that organizes the actin cytoskeleton at the PCD (Sheu et al, 1998). We first identified the pool of Nub-labelled interaction partners by mating a Spa2CRU expressing strain against the Nub array and selected a subset of those for single-cell analysis (Table I). Among the found binding partners was a further member of the polarisome Nub-Pea2p (Sheu et al, 1998). Time-lapse microscopy of the mated Spa2CCG and Nub-Pea2p expressing cells revealed for the first time that both proteins interact throughout the cell cycle (Figure 5A, B; Supplementary Movie 5). The specificity of the measurements was confirmed by co-expressing Spa2CCG and Nub-Ptc1p, a Nub fusion that, according to the large-scale analysis, does not bind to Spa2CRU (Figure 5C–E).
Spa2p forms homo-oligomers (Table I). Our SPLIFF analysis revealed that this reaction occurred much faster than the Spa2p-Pea2p heteromerization (Figure 5E; Supplementary Figure 3; Supplementary Movie 6). Why is the Spa2p-Pea2p interaction slower? We presumed that the continuous presence of the unlabelled Pea2p in the polarisome might effectively hinder Nub-Pea2p from binding and converting Spa2CCG. Consequently, cells lacking unlabelled Pea2p should display a faster interaction. Consistent with our hypothesis, we observed a 5.3-fold increase in the rate of Nub-Pea2p induced Spa2CCG conversion upon deleting the chromosomal PEA2 in the Spa2CCG expressing a-cells (Figure 5E; Supplementary Figure 3; Supplementary Movie 7). The ratio of the initial rates of conversion RCiPEA2 to RCiΔpea2 (Δf=5.3) provides a quantitative measure (Supplementary Figure 3). By using simplifying assumptions, we calculated the fraction of bound Spa2p (Fb) by Fb=(Δf−1)/Δf. Based on this calculation, we estimate that only 19% of the Spa2CCG molecules in wild-type cells are free to react with Nub-Pea2p. The remaining 81% (Fb) of the binding sites are occupied by endogenous Pea2p.
We identified Hof1p, a member of the cytokinesis machinery, as a new ligand of Spa2p (Table I) (Lippincott and Li, 1998; Meitinger et al, 2011). The SPLIFF analysis revealed that Hof1p already interacted with Spa2p at the PCDI during cell fusion. Furthermore, the kinetic profile of the interaction between Spa2CCG and Nub-Hof1p differed strikingly from the profile of the Spa2p/Pea2p interaction in tracing interactions during fusion and cytokinesis but not during bud growth (Figure 5E; Supplementary Movie 8). An allele of HOF1 lacking the N-terminal FCH domain (Nub-Hof198–669) added an interesting mechanistic detail to the understanding of this protein interaction. In contrast to the full-length protein, Nub-Hof198–669 converted Spa2CCG only during cytokinesis but not during cell fusion (Figure 5E). The results thus allude to different modes and degrees of interaction between Hof1p and Spa2p during the three phases of polar growth.
Many proteins simultaneously occupy different locations in the cell. Each of these locations might reflect an altering set of binding partners. We examined the cellular distributions of four different protein interactions to demonstrate the spatial resolution of SPLIFF.
The Nub fusion of Kel1p was found as interaction partner of Spa2CRU (Table I). Kel1p is involved in cell fusion during mating (Philips and Herskowitz, 1998). Kel1CCG is spread across the entire bud during its growth (Figure 6A). To specify the regions of Kel1p–Spa2p interaction, we mated Kel1CCG- with Nub-Spa2p expressing cells. Time-lapse analysis of the diploids revealed that the interaction occurs not only during cell fusion but also during bud growth (Figure 6A and B). We analyzed the spatial distribution of the red and green fluorescence in medium-sized buds (Figure 6C–E). Both intensities matched closely at the base of the bud yet segregated at its tip (Figure 6C–E). The reconstituted differential interaction maps describe a gradient in the ratio of red to green fluorescence that peaks at the bud tip and trails off toward the mother cell (Figure 6D and E). This distribution very probably corresponds with a gradient of interactions caused by the unequal distribution of Spa2p and Kel1p across the bud (Figures 5A and and6A6A).
Nap1p is a multifunctional protein that is involved in the transport and assembly of histones and the formation of the septin ring at the bud neck of the cells (Mortensen et al, 2002; Ohkuni et al, 2003). Nap1p is localized in the cytosol, the nucleus, and the bud neck. We recently confirmed Nub-Kcc4p as binding partner of Nap1CRU (Hruby et al, 2011). Kcc4p is a septin-localized protein kinase (Barral et al, 1999; Okuzaki and Nojima, 2001). To find out exactly when and where the interaction between Nap1p and Kcc4p occurs, we monitored the Nub-Kcc4p-induced conversion of Nap1CCG in the cytosol as well as in the nucleus. The interaction started immediately after cell fusion (Figure 7A and B). The slower decrease in nuclear versus cytosolic green fluorescence indicated that Kcc4p interacts with Nap1p primarily in the cytosol (Figure 7B; Supplementary Figure 4). Continuous time-lapse analysis of Nap1CCG identified at a later time point during early bud formation a red fluorescent zone devoid of green fluorescence beneath the incipient bud site (Figure 7A and C; Supplementary Movie 9). This region of preferred complex formation contrasted with the nucleus as well as with the cytosol as two compartments of comparatively weak Nap1p-Kcc4p interaction activity (Figure 7C). Fifty minutes after mating a decrease in the nuclear Cherry signal might indicate a trapping of Nap1CCG in the cytosol through the continuous co-expression of Nub-Kcc4p (Figure 7B). Nub-Pea2p is also concentrated at the incipient bud site but does not interact with Nap1CRU. Consequently, Nub-Pea2p did not induce a similar dissociation of red from green fluorescence in Nap1CCG-expressing cells (Figure 7D and E).
Stu2p is a microtubule-binding protein that is associated with the cytosolic and nuclear portion of the spindle pole body (SPB), and the nuclear, and astral microtubules (Figure 8A) (Usui et al, 2003; Al-Bassam et al, 2006; Kitamura et al, 2010; Winey and Bloom, 2012). The large-scale Split-Ub assay of Stu2CRU identified among others the known binding partner Kar9p, and Spc72p (Table I) (Miller et al, 2000; Liakopoulos et al, 2003; Usui et al, 2003). Mating of cells expressing Stu2CCG and the corresponding Nub-fusion proteins allowed us to dissect the microtubule structures into zones of differential Stu2p interaction activities (Figure 8B and C). The interaction between Stu2CCG and its two Nub-labelled binding partners started immediately upon nuclear fusion and the accompanying alignment of the two SPBs. Later during the cell cycle the unequal distribution of the red and green fluorescence across the Stu2CCG-stained microtubules clearly identified the tip of the astral microtubules as preferred site of Stu2p–Kar9p interaction (Figure 8B; Supplementary Movie 10). In contrast, time-lapse analysis of the cells co-expressing Stu2CCG and Nub-Spc72p revealed a preferential conversion of Stu2CCG into Stu2CC at both tips of the spindle (Figure 8C). One hour later a small fraction of green fluorescent Stu2CCG still remained in the mother whereas the SPB of the daughter cell was stained completely red (Figure 8D). Mated cells that were cultivated for several generations displayed a complete conversion of Stu2CCG into Stu2CC (Figure 8E).
Due to their indifference in regard to order, location, and timing, large-scale protein interaction networks are still a largely untapped resource of information that waits to be fully exploited by cell biologist (Alexander et al, 2009; Kim et al, 2006; Przytycka et al, 2010; Vidal et al, 2011; Deeds et al, 2012). Up to now dynamic and spatial features of protein interactions could only be revealed by fluorescence resonance energy transfer between two fluorescently tagged proteins or fluorescence cross correlation spectroscopy of two freely diffusing cytosolic proteins (Bacia and Schwille, 2007; Slaughter et al, 2011). However, both methods require careful optimizations before a positive hit from one of the high-throughput studies can be visualized as protein interaction in the cell. A transfer of large-scale interaction data into the analysis of their dynamic and spatial aspects is thus hardly possible (Lowder et al, 2011). We routinely achieved these transitions by replacing a large-scale compatible version of the Split-Ub technique by SPLIFF to subsequently probe the detected interactions with high spatial and temporal resolution in living cells.
Among the further advantages of SPLIFF is the nearly instantaneous detection of binding once the Nub and Cub fusions are free to interact. The Cherry signal that remains at the protein identifies the place of interaction and serves as internal reference for the decrease in GFP signal. The GFP signal reflects the fraction of uncut Cub-fusion proteins and by extrapolation the unbound fraction of the protein with regard to its Nub-labelled binding partner. As the cleavage of GFP is the temporally delayed consequence of interaction, the localization of the Cherry-Cub fusion protein can only be an approximation of where the interaction has originally occurred. To set closer limits for the preferred sites of interaction or for visualizing gradients of interactions across the cell, it is thus obligatory to analyze the signals of both, the bound and unbound fraction of the Cub fusion. A realistic interpretation of the relative green and red fluorescence intensities (FIs) further necessitates a continuous online detection once Nub and Cub fusions encounter each other. We used cell fusion by mating as a controlled way of mixing the two fusion proteins. The procedure unambiguously defines the start of the reaction and allows tracking it throughout the cell cycle. This feature also enhances the reproducibility of the measurements as averaging different single-cell measurements with slightly shifted cell-cycle phases might obscure specific aspects of the interactions. For example, any static or unsynchronized analysis would have missed our discovery that the interaction between Hof1p and Spa2p is restricted to cell fusion and cell separation but does not occur during bud growth. Together with Sho1p, Hof1p is now the second member of the HICS complex that is involved in cell fusion and cytokinesis (Labedzka et al, 2012). The question whether further members of the cytokinesis network also participate in cell fusion as part of identical or different protein complexes can now be systematically investigated by SPLIFF.
The feature of simultaneously detecting bound and unbound fraction was instrumental in localizing the interaction between the microtubule-nucleator Stu2p and the component of the SPB Spc72p. The prevalence of unbound Stu2CCG at the nuclear microtubules, the high fraction of bound Stu2CC at both tips of the spindle, and the known localization of Spc72p strongly suggests the cytosolic part of the SPB as the preferred site of Stu2p–Spc72p interaction. As the converted Stu2CC dissociated from Nub-Spc72p its newly acquired localizations during later time points blurred the assignment of its interaction to a specific place in the cell. We therefore postulate that the subsequent increase in the Cherry signal at the other Stu2p-containing structures was most likely due to the dynamic exchange of Stu2CCG by Stu2CC that was originally generated by Nub-Spc72p at the SPB. However, without additional knowledge or experimentation our analysis could not strictly exclude that a certain fraction of these signals might have been caused by a minor population of Nub-Spc72p localized at these sites specifically during the later phases of the cell cycle. Ultimately, once conversion of Stu2CCG into Stu2CC was nearly completed the Stu2p-containing structures were uniformly stained by Cherry and a detection of interaction no longer feasible.
The moderate affinity between Nub and Cub might stabilize transient or weak protein complexes by reducing the rate of their dissociation (Müller and Johnsson, 2008). Applying the rate of CCC to CG conversion and not the CCG/CC ratio as the preferred measure to specify the cellular place and duration of a certain interaction should avoid that this temporal trapping might interfere with the spatial and temporal resolution of SPLIFF.
Our studies thus defined the rapid and continuous detection of interaction, the detection of the unbound fraction, and the ability to control the start of the reaction as three requirements for the accurate SPLIFF analysis of protein interactions. Split-protein sensors lacking any of these features call for an even more deliberate interpretation with regard to localizing and timing a certain protein interaction. For example, the fluorescence signal generated upon reconstitution of Split-GFP (BiFC) might simply reflect the localization of the binding partner with the strongest localization signal and not necessarily the cellular site of their interaction (Kerppola, 2006; Stynen et al, 2012).
As the conversion of CCG to CC is irreversible, the continuous interaction between the Nub- and Cub-coupled fusion proteins will deplete the pool of CCG substrate and reduces the rate of its conversion. As a consequence, quantitative differences in the rate of conversion do not automatically correspond with quantitative differences in the dynamics of the underlying protein interaction. Despite this and other limitations of SPLIFF in exploring quantitative aspects of protein interactions, we propose that the calculated occupancy (Fb) of a protein is a useful quantitative reference for classifying protein complexes. For example, the Fb of 81% indicates that the Spa2p/Pea2p complex is much more stable than the Net1p/Fkh1p complex as the absence of unlabelled Fkh1p revealed no significant increase in the respective RCi of Net1CCG (Supplementary Figure 1; Ghaemmaghami et al, 2003).
The application of SPLIFF in higher eukaryotes is a very realistic perspective, as all essential components of this technique are also functional in these cells (Pratt et al, 2007).
Time-lapse experiments were carried out using a DeltaVision fluorescence microscope (Applied Precision) provided with a steady-state heating chamber and equipped with a mercury arc lamp and a camera CoolSNAP HQ2-ICX285 (Photometrics). In all cases a 100x NA 1.4 UPlanSApo oil immersion objective (Olympus) was used. Unless otherwise stated, images at 0.8μm intervals for a 4-section z stack were collected in 5min intervals on three channels: Brightfield, GFP (excitation 470/40nm and emission 525/50), and Cherry (excitation 572/35nm and emission 632/60). The CCD capture time was adapted to the intensity of GFP and Cherry signal in every construct to reduce bleaching and phototoxicity.
Confocal microscopy was conducted using An Axio Observer Z.1 Spinning Disc microscope (Zeiss). Images were acquired under the control of the AxioVision 4.8.2 software (Zeiss) as Z series with 15–20 twenty slices. The distance between two Z slices was set to 260nm. For image acquisition, a Plan-APOCHROMAT 63 × /1.4 Oil DIC ∞/0.17 objective and 1 × 1 binning was used. The camera gain was set to 1. For excitation of GFP or Cherry, a 488-nm diode laser at 80% laser power or a 561-nm diode laser at 100% laser power was used.
All files generated by the microscope software (Resolve3D softWoRx-Acquire Version: 4.0.0 Release 16) were processed and analyzed with ImageJ 1.45s (US National Institutes of Health; http://imagej.nih.gov/ij/download.html). Z stacks were projected onto two-dimensional images. In order to quantify the fluorescence over time, we selected regions of interest (ROI) and a region within the cell that could serve as background for the quantification. In experiments where the CCG fusion was distributed throughout the cell a section from a haploid Nub-fusion-expressing cell occupying the same field was used as reference. To quantify localization-dependent protein interactions, we analyzed the spatial fluorescence profiles of the structures of interest. Z stacks were projected to two-dimensional images. The intensity variations of the GFP and Cherry Channels were followed on a chosen path or across a certain area of the cell using either ImageJ Plot Profile tool or Interactive 3D surface plot tool, respectively.
To obtain the relative fluorescence intensities (RFIs) of each channel, we first calculated the FI by:
to then calculate the RFIs by:
(loc) specifies the signal of the ROI areas and (back) the signal derived from the background areas. In time-dependent interaction experiments, (x) represents time. FI(start) represents the FIs shortly before the fusion of the cells. The RFIs of both channels were then used to calculate the extent of the interaction (FD) as measured by conversion of the CCG fusion to the CC Fusion by:
In localization-dependent interactions, (x) represents distance along the chosen path. In static localization-dependent interactions, FI(start) represents the maximal value of the analogous structure in a haploid CCG-expressing strain, which resides in the same field as the investigated diploid. In the time- and localization-dependent interactions of Stu2CCG, FI(start) represents the maximal intensities of both channels for each time frame shown.
Images were prepared for publication with Acrobat Illustrator (Adobe) using linear contrast and intensity adjustments. When needed, DeltaVision fluorescence microscope images were deconvoluted using the software Resolve3D (softWoRx-Acquire Version: 4.0.0 Release 16). 3D reconstructions were obtained from Z stacks acquired by confocal microscopy using ImageJ 3D projection tools.
Large scale Split-Ub assays were performed as described (Hruby et al, 2011; Dünkler et al, 2012). Measuring interactions between individual Nub- and Cub-fusion proteins by spotting yeast cells expressing both fusions onto 5-FOA and SD Ura-containing media was essentially as described (Eckert and Johnsson, 2003).
The CCG- and Nub-fusion-expressing a- and α-cells were separately incubated in SD medium at 30°C overnight. If not indicated otherwise cells were diluted into fresh medium without copper and grown to an OD600 of 0.6–1. After mixing equal amounts of the a- and α-cells (between 0.5 and 0.75ml each), the culture was immediately spun down, and resuspended in 50μl of SD medium. 3μl of this suspension was immobilized by fixing a coverslide with parafilm strips on a custom-designed glass slide containing solid, non-fluorescent agar-SD without copper if not indicated otherwise. The slide was immediately incubated at 30°C under the microscope. Pictures were taken after 45–75min of incubation when the first zygote formations became visible.
The switch from the CRU to the CCG module was achieved by cloning the same PCR product used for the construction of the respective CRU fusion in front the CCG cassettes on a pRS306 or pRS304 vector. Linearization of the obtained plasmids with restriction enzymes cutting only once in the amplified sequences of the ORFs of the respective genes and homologous recombination after transformations of the yeast were as described (Dünkler et al, 2012). Successful recombination was verified by colony-PCR. Descriptions of the used constructs and yeast strains can be found in Supplementary Tables 1 and 2.
Culture media and yeast genetic methods were performed following standard protocols (Dünkler et al, 2012). Media for the Split-Ub interaction assay contained 1mg/ml 5-FOA. Yeast strains JD47, JD53, and JD51 are as described (Dünkler et al, 2012).
By plotting the relative decrease in GFP fluorescence FD we calculated by linear regression the initial rates of conversion (RCi) for the first 15min of the reaction.
We assumed that the conversion of Y-CCG fusion into the Y-CC fusion after mating with the Nub-X expressing strain is described by a bimolecular reaction where the rate of conversion is given by
Y-CCCGfree is the concentration of unbound Y-CCG-fusion and CNub-Xfree the concentration of unbound Nub-X. k is the rate constant of the reaction and a complex composite of a mix of different reactions including among others the rate of Nub and Cub reassembly, the cleavage rate by the ubiquitin-specific proteases, the diffusion constants.
We assume that shortly after cell fusion k and CNub-Xfree are very similar if not identical in the two diploids that were derived by mating of the Nub-X-expressing strain with a Y-CCG-fusion-expressing strain where the gene X was either deleted (Δx) or not (X). It then follows that the quotient of the initial rates:
CT is the sum of the concentrations of free and bound Y-CCG fusion whereas CB is the concentration of Y-CCG fusion bound to X.
dCY-CCG/dt(Δx) and dCY-CCG/dt(X), the RCis of both reactions were derived by calculating a linear regression of the first 15min after the reaction has started. The averages from at least four different independent measurements were taken. An analysis of the two reactions was only performed when the curves of the kinetic profiles were significantly different.
Curve fitting was performed in R version 2.15.1 ( http://www.R-project.org) by minimizing the least squares deviation of a set of regression functions to the data. The best fitting function was determined by the Akaike information criterion. For comparison of two groups, the sum of the residual sum of squares of the individual regressions is then compared to that of the regression for the combined data using the F-test (Motulsky and Ransnas, 1987).
We thank members of the Johnsson laboratory for discussions and constructs, and Judith Müller and Kai Johnsson for comments on the manuscript. This work was supported by the BMBF Initiative SysTec (0315690B) and by the BMBF Initiative GerontoSys2 (SyStaR, 03158 94A).
Author contributions: DM with contributions of AD and JN performed and analyzed the experiments. HK and KJ performed the statistical evaluation and curve fitting of the data. NJ, DM, and AD designed the research. NJ wrote the manuscript with contributions of all other authors.
The authors declare that they have no conflict of interest.