A Split-Ubiquitin Based Strategy Selecting for Protein Complex-Interfering Mutations

Understanding the topologies and functions of protein interaction networks requires the selective removal of single interactions. We introduce a selection strategy that enriches among a random library of alleles for mutations that impair the binding to a given partner protein. The selection makes use of a split-ubiquitin based protein interaction assay. This assay provides yeast cells that carry protein complex disturbing mutations with the advantage of being able to survive on uracil-lacking media. Applied to the exemplary interaction between the PB domains of the yeast proteins Bem1 and Cdc24, we performed two independent selections. The selections were either analyzed by Sanger sequencing of isolated clones or by next generation sequencing (NGS) of pools of clones. Both screens enriched for the same mutation in position 833 of Cdc24. Biochemical analysis confirmed that this mutation disturbs the interaction with Bem1 but not the fold of the protein. The larger dataset obtained by NGS achieved a more complete representation of the bipartite interaction interface of Cdc24.

at position two of this codon is more probable than all other non-silent mutations.
To determine the concentration of base analogues needed for the intended mutagenesis rate of 3-5 amino acid mutations per kilobase, different dilutions of the base analogue mixture were prepared: While the ratio of dPTP to 8-oxodGTP remained constant, a mixture of 10 mM dPTP and 40 mM 8-oxo-dGTP (base-mix) was diluted with ddH 2 O and subsequently used for error prone PCR. A 1:3 dilution of the base-mix and 3 cycles resulted in an acceptable mutational rate of about 3.6 amino acid substitutions per 1000 bp (data not shown).
Biotinylated forward (cctcccggccgATGTCGAGTGACGATAATAATACGAA) and reverse primers (gtatcgtcgacCCATACAGACGAATGTTCAAGAATTTC) were used for the error prone PCR. Eag1 and Sal1 restriction sites are underlined. For the preparation of library DNA, the PCR product was captured on streptavidin coated magnetic beads (New England Biolabs) to remove the template DNA. After washing, the beads were resuspended in 40 µl of PCR-buffer (20 mM Tris-HCl pH 8.8, 2 mM MgCl 2 , 10 mM KCl, 10 mM (NH 4 ) 2 SO 4 , 0.1 % (v/v) Triton X100, 0.1 mg/ml BSA). From this suspension, 5 µl was used as template for subsequent PCR-amplification in which the same (but nonbiotinylated) primers and unmodified dNTPs were used. To lower the mutational bias in the library, four independent error-prone PCRs were performed and subsequently combined for large-scale diversification. The PCR product was digested with Eag1 and Sal1, purified, and ligated into the pMet-CRU-313 plasmid. Ligation was performed for 18 h at 16°C using 15.6 µg plasmid, 6.25 µg insert and 150 U T4 DNA-ligase (Thermo Fischer Scientific). After heat inactivation of the Ligase at 65°C for 15 min, the DNA was precipitated with ethanol, resuspended in ddH 2 O and used for electroporation into competent XL1blue E. coli. The transformation of the complete library resulted in a total of 90 large (150 mm) dishes. Each dish was flooded with 5 ml 2YT medium; the cells were scraped off the plates, adjusted to 43% glycerol and stored in aliquots at -80°C.
We sequenced 12 randomly picked clones with a primer allowing us to cover the first 1000 base pairs and further 18 randomly picked clones with a primer allowing to cover the last 800 base pairs. Alignment of these sequences with the Cdc24 ORF sequence revealed an average of 5 mutations per kilobase -slightly more than estimated in the test experiments (File S2). Library DNA for transformation in yeast was prepared by diluting one aliquot of the library stock in 150 ml 2YT medium and incubating at 37°C until an OD 600 of 1.5. Plasmid DNA was prepared form this culture by using a Plasmid Maxi Kit (Thermo Fisher Scientific).

Library transformation and selection in yeast
High efficiency transformation of the N ub -Bem1 expressing yeast strain (Hruby et al. 2011) was performed as described elsewhere (Gietz and Woods 2002). The transformed cells were directly transferred in liquid selection medium (SD medium lacking histidine, uracil, methionine and containing 50 µM CuSO 4 and 200 µg/ml geneticin). After 24h, aliquots of 1.5 ml of the selection mixture were pelleted and stored for plasmid isolation. Further 5 ml were pelleted, resuspended in fresh selection medium and subjected to another round of selection.

Plasmid isolation and analysis by Sanger sequencing
Plasmid isolation from the pelleted yeast was performed with the Charge Switch Plasmid Yeast Miniprep Kit (Life Technologies) according to the manufacturer's instructions. The eluted DNA was precipitated with ethanol and subsequently electroporated into competent XL1blue E.coli. Positive clones were identified by colony PCR and subsequently grown in LB medium containing 100 µg/ml Ampicillin. Templates for Sanger sequencing were prepared from these clones through rolling circle amplification by an external service provider (Seqlab Laboratories). Sequence alignments and analysis were performed with the CLC Main Workbench version 7.6.2 (CLC Bio/Qiagen) (File S2).

Plasmid isolation and analysis by Next Generation Sequencing (NGS)
Template amplicons for NGS were PCR amplified from plasmids isolated by Phenol/Chloroform extraction from a separate selection experiment that was performed under identical conditions as described above. Application of the Charge Switch Kit including an EcoR1 restriction site (underlined) to generate the 50 bp extension that is necessary for NGS template preparation with the Nextera system (see below). PCR was performed for 3 cycles and further 20 cycles using annealing temperatures of 52°C and 59°C, respectively. Re-cloning in a pEGX2T plasmid via the BamH1 and EcoR1 sites and subsequent sequencing of a number of clones revealed that the library complexity was not altered by the PCR amplicon preparation (data not shown). The purified PCR product was captured on streptavidin coated magnetic beads to remove plasmid template DNA and eluted from the beads by restriction with EcoR1 after washing in CutSmart buffer (New England Biolabs). The product was purified by a PCR Purification Minelute Kit (Thermo Fisher Scientific) and quantified using the QuantiFluor dsDNA System (Promega). Preparation of index-and adapter sequence tagged amplicon fragments was subsequently performed with the Nextera XT Library Preparation Kit (Illumina) according to the manufacturer's recommendations. Size distribution of the NGS-ready amplicon fragments was monitored with a QIAxcel device (Qiagen).
Sequencing was performed with a Miseq nano v2 flow cell (Illumina) on a Miseq sequencing device (Illumina) according to the manufacturer's instructions.
Sequencing was performed as paired read runs of the input library and of each selection round. Each read of the paired end sequencing run generated two files that were retrieved from the MiSeq system.
Quality filtration of the raw reads and subsequent mapping to the reference amplicon (corresponding to the wildtype Cdc24 428-854 sequence) was performed with the mapmuts.makealignments algorithm of the Mapmuts software package (Bloom 2014).
Only read pairs with an overlap of at least 25 bases and an average Phred-score quality of 28 or higher were included in the further analysis. Subsequently, amino acid identities at each position were counted using the mapmuts.parsecounts script (Bloom 2014). This script determines also the coverage at each position. Subsequent data processing was performed by spreadsheet calculation.
The first and the last positions of each sequence read pair showed a much higher mismatch rate (up to 10-fold higher, data not shown) than the rest of the read. As a consequence we removed the distal 9 bases for all sequence runs.
Very low counts in the input library might result in artificially high enrichment scores and thus only variants with amino acid counts higher than 4 were considered for further processing. Otherwise, counts were set to zero.
In average, values for three mutated amino acids were obtained at each site of the input library. Then the frequency F of each amino acid identity i at each position p was calculated by dividing the counts of the respective amino acid by the coverage at that position.
The enrichment score ES of an particular amino acid was calculated as follows: Because the enrichment score of a non-mutated (wildtype) amino acid is defined as 1 (Melamed et al. 2013), enrichment scores were normalized to the corresponding wildtype ES at a particular position. Manual Split-Ubiquitin assay JD53 cells expressing either N ub -Bem1 or N ub -ha were transformed with the plasmids carrying the respective CRU fusions. Cells were grown in selective media to an OD 600 of 1. Of this culture and further 10-fold serial dilutions 4.5 µl were spotted on media lacking methionine, histidine and uracil and containing 50 µM CuSO 4 . Cells were grown for two days at 30°C. The same dilutions were also spotted on media containing uracil to control for the equal growth of the strains under non-selective conditions (Hruby et al. 2011, Dünkler et al. 2012.

Purification of recombinant proteins
Protein expression of 6His-tagged PB Bem1 -SNAP was carried out in SB medium at 18°C for 5 h. The cells were lysed by treatment with lysozyme and sonication. Purification was done by IMAC using an ÄKTA Purifier chromatography system (GE Healthcare). The protein was eluted from a 5ml HisTrap HP column (GE Healthcare) in a linear imidazole gradient. Fractions containing the purified protein were pooled, buffered in HBSEP (10 mM HEPES, 150 mM NaCl, 3 mM EDTA, 0.05% Tween 20, pH 7.4) containing 30% Glycerol using a PD10 desalting column (GE Healthcare) and stored until use at -20°C.
Cells expressing 6His-tagged PB Cdc24 or PB Cdc24 (D833G) were cultivated in LB medium at 37°C. Cell lysis and protein purification with IMAC was performed as described above.
IMAC fractions containing the respective protein were pooled and subjected to size exclusion chromatography in HBSEP buffer using a Superdex 200 16/60 column (GE Healthcare). Fractions containing the purified protein were pooled, concentrated and used directly for SPR measurements.

Determination of binding affinities
Binding affinities were measured by SPR using a Biacore X100 system (GE Healthcare)