A quantitative tri-fluorescent yeast two-hybrid system: from flow cytometry to in-cellula affinities

We present a technological advancement for the estimation of the affinities of Protein-Protein Interactions (PPIs) in living cells. A novel set of vectors is introduced that enables a quantitative yeast two-hybrid system based on fluorescent fusion proteins. The vectors allow simultaneous quantification of the reaction partners (Bait and Prey) and the reporter at the single-cell level by flow cytometry. We validate the applicability of this system on PPIs with different affinities. After only two hours of reaction, expression of the reporter can easily be detected even for the weakest PPI. Through a simple gating analysis, it is possible to select only cells with identical expression levels of the reaction partners. As a result of this standardization of expression levels, the mean reporter levels directly reflects the affinities of the studied PPIs. With a set of PPIs with known affinities, it is straightforward to construct an affinity ladder that permits rapid classification of PPIs with thus far unknown affinities. Conventional software can be used for this analysis. To permit automated high-throughput analysis, we provide a graphical user interface for the Python-based FlowCytometryTools package.


Plasmid creation 130
In order to generate the pSB_1Bait plasmid, the pLexA (22) vector was linearized using 131 EcoRI and SalI (Thermo Scientific) to remove all DNA between the LexA cDNA and the ADH 132 terminator. The Barstar WT coding sequence was ordered for synthesis to Eurofin Genomics as 133 part of a new expression cassette. At the 5' end we added the sequences of the HA-Tag and our 134 MCS-spacer (EcoRI, AscI, and XhoI). At the 3' end, after the stop codon of Barstar, we inserted 135 one XhoI site, created 3 stop codons (1 per ORF) and regenerated the SalI site. The upstream 136 (LexA) and downstream (Terminator) 30bp required for homologous recombination in yeasts (30, 137 31) were also added. This new optimized expression cassette was amplified by PCR (Phusion DNA 138 polymerase, Thermo Scientific), using the primers primSB_0001 and 2 (see Suppl. Table S1), and 139 inserted in the previously linearized pLexA vector. As a result, we obtained the pSB_1Bait_Barstar. 140 The coding sequence for Tag-RFP was subsequently introduced in the EcoRI site through PCR 141 from pTag_RFP-Actin (Evrogen), using the primers primSB_0003 and 0004, combined with 142 homologous recombination in yeasts to obtain the pSB_1Bait_RFP-Barstar plasmid. The 143 pSB_1Bait_RFP-Empty and pSB_1Bait-Empty vectors were generated by digesting the 144 pSB_1Bait_RFP-Barstar and pSB_1Bait_Barstar, respectively, with XhoI (Thermo Scientific), 145 followed by self-ligation. The coding sequences of the mutants of Barstar, and Ras_G12V_C186A 146 were ordered to Eurofin Genomics, amplified (primSB_0018 and 0019) and introduced in the 147 pSB_1Bait_RFP-Empty linearized with XhoI by homologous recombination. 148 To create the pSB_1Prey vector, the pB42AD plasmid (22) was linearized using EcoRI and 149 XhoI. The sequence coding for the non-toxic Barnase mutant H102A was ordered from Operon 150 MWG. At the 5' end we inserted the same MCS-spacer sequence as in the pSB_1Bait vector to 151 allow easy transfer from one plasmid to the other. At the 3' end, we inserted one XhoI site, created 9 3 stop codons and one NcoI restriction site. The upstream (HA-Tag) and downstream (Terminator) 153 30bp required for homologous recombination in yeasts were also introduced. This new expression 154 cassette was then amplified by PCR (primSB_0010 and 0011) and inserted in the pB42AD by 155 homologous recombination in yeast to obtain the pSB_1Prey_Barnase-H102A vector. The coding 156 sequence of the yEGFP was amplified from the pGY-LexA-GFP_KanMX (kindly provided by D r 157 Gaël Yvert) using the primers primSB_0012 and 0013, and then introduced in the EcoRI sit of our 158 MCS as previously to generate the pSB_1Prey_yEGFP-Barnase-H102A vector. The pSB_1Prey-159 Empty and pB_1Prey_yEGFP-Empty were created by removing the coding sequence of Barnase 160 H102A with XhoI and performing a self-ligation. The coding sequences of the other Preys (CRaf 161 RBD WT, and CRaf RBD A85K) were ordered to Eurofin Genomics, amplified by PCR 162 (primSB_0020 and 21) and introduced in the pSB_1Prey_yEGFP-Empty linearized with XhoI by 163 homologous recombination. 164 To create the reporter plasmid, the pSH18-34 (22) was digested using the unique SalI (In 165 the modified Gal1 promoter) and RsrII (downstream to the -Galactosidase coding sequence) 166 restriction sites. We subsequently reconstructed the expression cassette using four PCR products: 167 1) The Gal1 promoter delta Gal4 with 8 operator LexA and the Kozack sequence with 168 a new downstream MCS (AscI, NheI) (primSB_0076 and 0077). 169 2) The Gal1 Nterm sequence (I10-C20), originally expressed by the pSH18-34, is used 170 as spacer (primSB-0078 and 0079) between the two copies of the Tag BFP. 171 3) The coding sequence of the Tag-BFP (from pTag_BFP-Actin, Evrogen) borded with 172 2 XhoI sites, (primSB_0084 and 0085). 173 4) The terminator sequence (primSB_0080 and 0081).
These 4 amplicons were then used to perform directly a gap repair in yeasts. Thus, we 175 obtained the pSB_3RO plasmid. A second copy of the Tag-BFP (primSB_0120 and 121)  were selected on SD-W medium. Matrix mating assay were performed for one night with 50µl of 205 Bait and Prey strains (each) resuspended in YPAD medium at 0.1 OD at 30°C. The next morning 206 YPAD medium was removed and the yeast diploids were harvested and amplified in 1ml of SD-207 UHW for 3 days at 30°C. 208 The qY2H assay was performed in pre-heated (30°C) and oxygenated SGR-UHW 209 supplemented with Galactose 0.25% (Euromedex) and Raffinose 1% (Sigma-Aldrich) to induce 210 the expression of the Prey proteins. To ensure we obtained an excessive number of cells (about 211 10 7 ) for the analysis, a culture of 100 ml was inseminated with 600 µl of saturated diploids per 212 couple of interest. It turned out that for a typical analysis a number of 10 6 cells is adequate, so that 213 actually 10 times smaller cultures and insemination volumes can be used. The yeasts were 214 incubated for 2h at 30°C without shaking, and then harvested after a centrifugation step of 10min 215 at 1000g. The yeast were resuspended in 1ml PBS (Dominique Dutscher), centrifuged 1min at 216 13000 rpm, and washed again with 1ml of PBS. The yeasts were resuspended in 500µl of PBS 4% 217 PFA (Sigma-Aldrich, Catalog n°P6148) and incubated for 10min at 4°C. The fixation reaction was 218 blocked by 2 washing steps with 1ml PBS, and one incubation of 15min at 4°C in 500µl of PBS 219 0.1M Glycine (Euromedex). Finally, the yeasts were washed twice in PBS, and stored in 1ml of 220 PBS at 4°C for not longer than 24 hours.

Flow cytometry and data analyses 223
The expression levels of BD-Bait, AD-Prey and reporter were acquired in linear scale using 224 a MacsQuantVYB flow cytometer (the settings are presented in Suppl. Table S2) confirms results of previous studies that the global Y2H read-out correlates with in-vitro affinity 271 (13-17). In addition, our approach discloses the influence of the expression levels of the reaction 272 partners, i.e., the BD-Bait and AD-Prey fusions. Their levels may vary significantly between the 273 studied couples (Fig. 2 & 3) and to a smaller extent between different experiments. These variations 274 complicate the discrimination of Bait-Prey couples based on their affinities. Fig. 4A (Fig. 4B). Similarly, the couple BD-Barstar Y29F / AD-Barnase H102A has the lowest 288 AD-Prey expression level and displays a weaker reporter level than expected. Standardization of 289 the AD-Prey level corrects the reporter levels of the studied couples according to their reported in-290 vitro affinities (Fig. 4C). Changing the location of the gating intervals leads to the same conclusions 291 (see Suppl. Fig. S2). Nevertheless, suggestions can be made to choose the optimal the gating 292 intervals (see "Recommendations"). 293

Affinity ladder permits rapid classification of PPIs based on their strength 295
Often the goal is to rank PPIs based on their affinity or to obtain an upper and lower bound 296 for the dissociation constant. With the above gating approach, an affinity ladder can easily be 297 generated with a set of PPIs with known dissociation constants (Fig. 5). Standard software of flow 298 cytometers can be used to perform the required gatings and calculate mean fluorescence intensities. 299 We provide a graphical user interface to automate the generation of the affinity ladder (see 300

Experimental Procedures). 301
The generated affinity ladder can then be used for a rapid visual classification of PPIs with 302 thus far unknown affinities within a given range (here from nano-to picomolar). This is 303 demonstrated at the example of the mutation D35A of Barstar. Thus far, no in-vitro affinity data is 304 available for the interaction of this mutant with Barnase H102A. With the affinity ladder of Fig. 5  305 we can rank the affinity between those of HRas/CRaf 122 nM) and HRas/CRafA85K (11 nM). 306 Our experiment indicates that the Barstar mutant D35A exhibits a significantly higher 307 affinity for Barnase H102A than the Barstar mutant D39A (420 nM). To validate this observation, 308 we performed independent alchemical free-energy calculations (see Supp. Mat.). Through the use 309 of a thermodynamic cycle (Suppl. Fig. S3A) we calculated the difference in binding free energy 310 between the mutants Barstar D35A and Barstar D39A. We obtained a value of -1.

In-cellula is not in-vitro 333
The observed agreement between in-cellula reporter levels and in-vitro affinities (Fig. 4B) 334 cannot be presumed a priori. The in-vitro experiment measures the affinity between the interactors 335 alone (or with tags) whereas the qY2H system relies on fusions proteins (Fig. 1). If the fused 336 domains influence the interaction between the Bait and Prey, e.g., by blocking the binding 337 interface, the resulting in-cellula reporter level would be impaired and most likely not correlate 338 with the in-vitro affinity.

In-vitro experiments measure the affinity under well-defined buffer-controlled equilibrium 340
conditions. In contrast, our in-cellula experiments take place in non-equilibrium microvessels (37) 341 where the interaction partners can interact with the endogenous complex solution of biomolecules. 342 This may lead to effectively smaller concentrations of the reaction partners. Also, post-translational 343 modification(s) could impact the interactions. 344 With prior knowledge about the Bait or Prey the amino-acid sequence can be optimized to 345 take into account some effects, like specific sub-cellular localization. An illustrative example is the 346 protein HRas, which is usually found to be associated to the cytoplasmic membrane through its C- Beside potential sequence optimizations (as proposed above), we recommend the following 352 precautions to be taken for the measurement with the qY2H system: 353

1)
As in any Y2H screen, BD-Bait and AD-Prey constructs should be tested against negative 354 controls, i.e., an empty AD and BD construct, respectively, to identify potential false positive 355 interactions. Also, a mating of yeasts expressing only empty (but fluorescent) AD and BD 356 constructs is recommended; it serves to remove the background of the system. 357 2) The PPIs should be tested in both orientations, i.e., with the proteins switched between the 358 Bait and Prey vectors, to identify the orientation with the lower background and with the 359 higher reporter level (for standardized levels of reaction partners). 360

3)
We recommend to pre-transform BD-Bait-expressing haploids with the read-out plasmid; it 361 increases the read-out; two subsequent transfections are more efficient than a single double 362 transfection. Use only freshly transformed yeast cells for the qY2H experiment. Storing diploids yeast cells for a week in the refrigerator decreases the level of AD-Prey and read-364 out by a factor two to five. 365

4)
For the construction of the affinity ladder, the gating interval for the red fluorescence 366 intensity (BD-Bait) was positioned at the lowest possible location to avoid saturation effects, 367 i.e., it was set just above the 95-% threshold of the negative control. The gating intervals of 368 the green fluorescence intensity was set to a medium range value to reach the desired 369 sensitivity but to avoid saturation and protein burden effects (39). The width of each interval 370 gate should be about 20-30% of the value of its lower border. 371

5)
If the gating intervals are not directly applied at acquisition time on the flow cytometer, at 372 least 10 6 cells should be acquired for analysis. This number is sufficient to reach a converged 373 ladder after gating (see Suppl.   were extracted from 6 OD of diploid yeasts grown for 2h at 30°C in SGR -UHW (0.25% galactose). 574 The BD-Bait and AD-Prey fusion proteins were simultaneously detected by Western blotting using 575 HA tag (originally present only in the AD-Prey proteins, and newly added to the BD-Bait proteins). 576 The expected molecular weights of the fusion proteins are indicated by red (BD-Bait) or green 577 (BD-Prey) triangles. Except for Ras G12V C186A (empty red triangles), all the proteins are 578 detectable at their correct molecular weight. B. The differential expression level was then 579 quantified using ImageJ program. A four to nine fold overexpression of the AD-Prey compared to 580 the corresponding BD-Bait (when detectable) can be observed. 581 qY2H affinity ladder. The same dual gating approach as in Fig. 4 was applied to all studied couples of Table 1. Here one representative experiment is presented. The cumulative mean of the negative control (BD-Empty / AD-Empty) was removed in all cases. The resulting cumulative mean curves are ordered according to their reporter dissociation constants (Table   1). For the couple BD-Barstar D35A / AD-Barnase H102A no dissociation constant is reported in the literature. Our affinity ladder allows to rank the constant between 11 and 122 nM.