FRET-Seq: a High-Throughput FRET-Based Screening Platform to Improve FRET Biosensors in Mammalian Cells

Genetically-encoded biosensors based on FRET have been widely used to dynamically monitor the activity of protein tyrosine kinases (PTKs) in living cell with high spatiotemporal resolution. However, the limitation in sensitivity, specicity, and dynamic range of FRET biosensors have hindered their broader applications. Here, we introduced a systematic platform, FRET-Seq, which integrates high-throughput FRET sorting and next-generation sequencing, to identify FRET biosensors with better performance from large-scale libraries directly in mammalian cells.


Introduction
Developing protein-based imaging tools for monitoring the dynamics of protein tyrosine kinases (PTKs) in living cell is necessary to understand kinase activities. Genetically-encoded biosensors based on FRET (FRET biosensors) have been widely used to dynamically monitor the activity of PTKs in living cells with high spatiotemporal resolution 1,2 . The typical design of these FRET biosensors consists of two parts 3 : a pair of uorescent proteins for FRET action, and a sensing region (e.g. a ligand region (SH2 domain) and a substrate peptide containing a tyrosine phosphorylation site). A exible linker peptide is also used to connect each domain. However, the limitation in sensitivity, speci city, and dynamic range of biosensors have hindered their broader applications 4,5 . Several research groups have developed FRET biosensors 2,6,7 , but the general procedure for constructing FRET biosensors still involves trial and error. Thus, a more e cient method which allows for selection of biosensors with better performance in response to the PTK activity from a large number of biosensor variants is required.
Here, we introduced a systematic method to optimize the FRET biosensors for Fyn kinase directly in mammalian cell through function-based screening, which combines the advantages of several technologies, such as site-saturation mutagenesis, mammalian cell library, and high-throughput DNA sequencing. We also developed mammalian cell-based and function-based high-throughput screening by FACS to directly optimize the Fyn FRET biosensors for studying the activity of Fyn kinase. Because the FRET e ciency of intramolecular biosensors is primarily in uenced by the distance and the relative orientation of the two uorophores 2 , we chose to optimize the substrate sequence in the sensing region of FRET biosensors (ECFP/YPet pair) to increase sensitivity, speci city, and FRET dynamic range in reporting kinase activities. The step-by-step protocol for the Fyn FRET biosensor development is described.

Reagents
Bacteria and mammalian cells: 2) Ligate each gene fragment into the pSin lentiviral transfer vector between SpeI and EcoRI with T4 ligase. Several restriction sites are introduced as shown in Figure 1a. The nal product is called pSin-ELYK, where E, L, Y, and K stand for ECFP, LacZ, YPet, and kinase domain, respectively ( Figure 1a).
3) Transform the ligation product to DH5α with the mixture of IPTG and X-Gal, and seed cells onto the LB agar plate with ampicillin at 37℃ overnight. 4) Select only blue colonies for plasmid puri cation. 5) Verify the sequence of pSin-ELYK by Sanger sequencing (GENEWIZ) before use.

Plasmid preparation for the Fyn self-activating FRET (saFRET) biosensor
1) Replace the "K" domain in pSin-ELYK with the gene fragment of either active or dead Fyn kinase domain using T4 ligase.
3) Transform the DNA assembly product to DH5α with the mixture of IPTG and X-Gal, and seed cells onto the LB agar plate with ampicillin at 37℃ overnight. 4) Select only white colonies for plasmid puri cation. 5) Verify the sequence of the Fyn saFRET biosensor by Sanger sequencing before use.

Activity test of Fyn saFRET biosensors by a western blot
1) Transfect puri ed plasmids of Fyn saFRET biosensors to HEK293T cells using Lipofectamin3000.
2) After 36-48 h of transfection, harvest and lyse cells with lysis buffer.

3) Centrifuge cells and collect supernatants.
4) Mix supernatants with SDS-PAGE protein loading buffer, boil, and load the solution into SDS-PAGE gel. 5) Following the general protocol for the western blot, the Anti-GFP antibody is used to target the biosensor, and the Anti-Phosphotyrosine antibody is used to target the phosphotyrosine. The gel image is taken by the gel documentation.
1.4 Activity test of Fyn saFRET biosensors by a uorescent microscope 1) Transfect puri ed plasmids of Fyn saFRET biosensors to HEK293T cells using Lipofectamin3000.
2) After 36-48 h of transfection, seed transfected cells on a bronectin-coated glass-bottom dish and culture in 0.5% FBS DMEM medium for 12 h before being subjected to the inhibitor stimulation (such as 10 µg/mL of PP1 for Fyn kinase).
3) The detailed experiment for imaging the activity of the biosensor in response to the PP1 is described in "Microscopy, Image Acquisition, and quanti cation". The procedure for constructing these FRET biosensor libraries is described below: 1) The template for generating the biosensor library containing substrate variants is the sensing domain ( Figure 1b). 3) For the PCR condition, vary the annealing temperature from 55-70˚C and set the thermocycling condition at 20 cycles. 4) Load PCR products of four substrate libraries in 1.5% agarose gel. 5) Excise gels with the correct size and purify them using Zymoclean gel DNA recovery kit. Each substrate library is prepared in parallel for steps #6-12. 6) Replace the LacZ domain of the pSin-ELYK template (Figure 1a) between Esp3I restriction sites with puri ed DNA fragments (from #5) of the sensing domain containing substrate libraries sites using the Golden Gate assembly. 7) Purify and concentrate assembly products using the DNA Clean and Concentrator Kits. 8) Add puri ed products (the volume < 2.5 μL) from #7 to 25 μL DH10B cells, and then added the mixture into a 0.1 cm cuvette for electroporation by the electroporation system. 2) On the next day, replace the medium with the fresh medium 3-6 h before transfection.
3) Use the ProFection® mammalian transfection system for transfection. The mixture of viral packaging plasmids pCMV-△8.9 and pCMV-VSVG, and the pSin plasmid containing biosensor variants are added in the calcium-phosphate reagent from the Profection system. Following the provided protocol, the solution is added to Lenti-X 293T cells. 4) After 6 h of transfection, replace the medium with the fresh medium. 5) After 48 h of transfection, collect the medium containing virus particles, lter them through 0.45µm lter (Sigma-Millipore), and concentrate using PEG-it virus precipitation solution for overnight at 4℃.

6) Centrifuge virus precipitation solution to collect the virus pallet. Dissolve the virus pallet with PBS and stored at -20℃.
7) Measure the viral titer by ow cytometry before use.

Generation of mammalian cells containing biosensor libraries by virus transduction
1) Seed 2⨯10 6 HEK293T cells in 100-mm cell culture dish with 10% FBS DMEM medium.
2) On the next day, replace the medium with the fresh medium 6 h before transduction. 2) On the day of sorting, collect seven groups of cells in #1 and resuspend them in PBS with 5% BSA. They are used to create gating pro les for cell sorting in the BD FACS Aria II Cell Sorter. The information about excitation and emission of FACS is shown below.

3) Add the concentrated virus with the MOI (multiplicity of infection
-For ECFP, the excitation wavelength is 405 nm, and the emission lter is 450/50 nm.
-For YPet, the excitation wavelength is 488 nm, and the emission lter is 545/35 nm.
-For FRET signal, the excitation wavelength is 405 nm, and the emission lter is 545/35 nm.
-For FRET ratio, it is the ratio of emission of ECFP signal to FRET signal. 2) Following the provided protocol with some additional step, such as incubating the column with RQ1 RNase-Free DNase to remove the genomic DNA. RNA from each group (from #1) is puri ed, quanti ed, and checked its quality by gel electrophoresis.
3) Use the puri ed total RNA (~500 ng) as a template for cDNA synthesis using the SuperScript IV reverse transcriptase with a gene-speci c primer.
4) Add adaptor sequences with different indexes for Illumina sequencing into cDNA by PCR using Q5 DNA polymerase with low PCR cycles (< 16 cycles) as shown in Figure 4.

5)
Load the individual amplicon library with speci c index to 2% agarose gel.
6) Excise the gels with correct size of the library and purify them using Zymoclean gel DNA recovery kit.

7)
Verify the DNA sequence of puri ed amplicon libraries by Sanger sequencing and quantify them by Qubit prior to being sequenced by Illumina HiSeq4000 with 50-bp single-end sequencing.
6. Analysis of sequencing results and selection of biosensors with high sensitivity 1) Separate FASTQ les from Illumina HiSeq4000 containing the sequence data of the substrate from each library by indexes.
2) Check their sequencing qualities using FastQC. Twelve sequencing data are used for identifying the substrate of the biosensor with better performance.
3) Extract only sequence data of the substrate domain with the following criteria.
-The sequences must have Phred score > 20 at all positions covering the constant region of the substrate sequence.
-The sequences must contain the "TAC" sequence encoding tyrosine. 6) The criteria for selecting a biosensor with better performance is that the variants with E v > 1 in KAH and KDL groups and < 1 in KAL and KDH groups.
7) The data for each substrate sequence can be also visualized in the 4D plot using Matlab software. To achieve better illustration of the 4D-plot, E v of each group can be further normalized to E vn by E vn = log (E v ) when E v 1, and E vn = E v-1 when E v ≤ 1. These selected sequences with 4D analysis will be further evaluated by their product of E v (KAH) and E v (KDL) and calibrated with the product of WT biosensor to lter and identify biosensors with the best performance.
7. Microscopy, Image Acquisition, and quanti cation 1) Construct the biosensor with the predicted substrate sequence as described in "Plasmid construction". Brie y, the substrate of the wild-type sensing domain is replaced by the predicted substrate sequence by PCR with a pair of primers, including: -The forward primer anneals to the 5'-end of the sensing domain (SH2 domain).
-The reverse primer containing the predicted substrate sequence anneals to the 3'-end of the sensing domain (linker and substrate).
The sensing domain with new substrate sequence is used to replace the LacZ of pSin-ELYK plasmid using Golden Gate assembly ( Figure 1). The assembly product is later transformed, puri ed, and sequenced by Sanger sequencing before use.

Anticipated Results
Please see Associated Publication.
The sensing region containing a SH2 domain, a exible linker, and substrate variants. Restriction enzyme sites are indicated on the top.   The design of the amplicon library for high-throughput DNA sequencing Following the design of library preparation from Illumina, Inc., the sequence of interest is a part of sequences of the biosensor variants shown in the box.