A rapid, reversible, and tunable method to regulate protein function in living cells using synthetic small molecules

Mutants of human FKBP12 F36V (107 residues) fused to a protein are degraded in mammalian cells unless the synthetic FKBP ligand Shield-1 (Shld1) is present. Fusing such a destabilizing domain (DD) to a protein of interest makes the whole fusion unstable, and binding of Shld1 to the DD shields the fusion from degradation and restores the protein's function.

A screen of an error-prone PCR library of about 3 × 10^4 FKBP12 F36V variants fused in frame to YFP, stably expressed in NIH3T3 cells by retroviral transduction, yields destabilizing domains. Three rounds of flow-cytometry sorting (SLF* at 5 µM, then ligand withdrawal for 60 hr to collect low-fluorescence cells, then SLF* again) enrich ligand-dependent mutants. Sequencing of 72 selected clones shows that all keep the F36V mutation, that mutations are spread fairly evenly along the sequence without clustering on the tertiary structure, and that most lie away from the ligand-binding site.

Shield-1 (Shld1) is an FKBP12 F36V ligand made from SLF* by replacing its carboxylic acid with a morpholine group, a position known not to interfere with FKBP binding, to enhance intracellular availability and potency. Shld1 stabilizes FKBP12 F36V destabilizing domains fused to YFP in cells.

Five FKBP12 F36V mutants (F15S, V24A, H25R, E60G and L106P) fused to the N terminus of YFP all lower YFP fluorescence in NIH3T3 cells in the absence of Shld1, to different degrees. L106P is the most destabilizing, leaving YFP fluorescence at only 1%–2% of the positive-control level, and 1 µM Shld1 fully restores L106P-YFP to its maximum level.

The Shld1 concentration needed to rescue N-terminal FKBP12 F36V destabilizing domain-YFP fusions in NIH3T3 cells differs by over an order of magnitude between mutants. The EC50 is 5 nM for V24A, the most efficiently rescued, whereas the more destabilizing L106P needs about 100 nM.

After Shld1 is added, YFP fluorescence of the F15S, V24A, H25R, E60G and L106P destabilizing domain fusions in NIH3T3 cells rises at approximately the same rate, reaches its maximum at 24 hr, and stays there for at least 48 more hr without further Shld1. In cells expressing L106P-YFP, YFP expression is detectable within 15 min of adding Shld1, and lower Shld1 concentrations (tested at 30 nM to 1 µM) reach steady-state expression more quickly than higher ones.

After removal of Shld1, the FKBP12 F36V destabilizing domain-YFP fusions decay at rates that correlate with how destabilizing each mutation is. H25R, the least destabilizing of F15S, V24A, H25R, E60G and L106P, decays most slowly, and L106P-YFP falls to negligible levels within 4 hr in NIH3T3 cells.

FKBP12 F36V destabilizing domain-YFP fusions (F15S, V24A, H25R, E60G, L106P) are undetectable by anti-FKBP12 immunoblot in NIH3T3 cells without Shld1 and strongly expressed after 24 hr of 1 µM Shld1, with no evidence of partially proteolyzed fragments. The proteasome inhibitors MG132 (10 µM, 4 hr) and lactacystin block degradation of F15S-YFP and L106P-YFP after Shld1 withdrawal, which indicates that degradation of the fusions is mediated at least in part by the proteasome.

Fusion of the L106P destabilizing domain to YFP lowers protein levels faster than siRNA lowers lamin A/C, in different cell lines. L106P-YFP in stable NIH3T3 cells is nearly completely degraded within 4 hr of Shld1 removal, whereas lamin A/C protein in HeLa cells transfected with an anti-lamin A/C siRNA begins to fall after 24 hr and is significantly reduced by 48 hr.

The level of L106P-YFP in stable NIH3T3 cells follows the Shld1 concentration over one week of treatment with various Shld1 concentrations. The measured fluorescence stays close to the values predicted from the Shld1 dose-response curve of L106P-YFP, which makes the destabilizing domain a dose-tunable regulator of protein level.

A YFP-FKBP12 F36V library screen identifies destabilizing domains that work at the C terminus of a protein. M66T, R71G, D100G, D100N, E102G and K105I, together with L106P placed at the C terminus of YFP, confer Shld1-dependent YFP fluorescence, with Shld1 EC50 values from 10 nM to 100 nM, similar to the N-terminal domains. After Shld1 removal, decay rates again track the degree of destabilization, and YFP fused to the most destabilizing domains, D100G and L106P, falls to negligible levels within 8 hr.

The L106P destabilizing domain is less destabilizing at the C terminus of YFP than at the N terminus. YFP-L106P retains about 10% of its Shld1-stabilized fluorescence without ligand, whereas L106P-YFP retains only 1%–2%. C-terminal L106P is nonetheless as destabilizing as any mutant identified in the screens.

FKBP12 F36V destabilizing domains give Shld1-dependent YFP fluorescence in four cell lines (NIH3T3, HEK 293T, HeLa and COS-1) after transient transfection. Without Shld1, residual YFP fluorescence relative to the maximum for each mutant is 5%–26% across the lines (for example 7% for F15S-YFP and 8% for L106P-YFP in NIH3T3, and 7% and 5% in HEK 293T), for N-terminal F15S and L106P and C-terminal D100G and L106P. Residual fluorescence is somewhat higher in transient transfection than in stable transduction.

FKBP12-derived destabilizing domains are also stabilized by FK506, a commercially available FKBP ligand. Unlike Shld1, FK506 perturbs the cellular environment by inhibiting calcineurin.

The FKBP12 F36V destabilizing domains F15S and L106P, fused at the N terminus, confer Shld1-dependent stability to the kinases GSK-3β and CDK1, the cell-cycle regulators securin and p21, and the small GTPases Rac1, RhoA and Cdc42. D100G and L106P, fused at the C terminus, do the same for CREB and the small GTPases Arf6 and Arl7. All 14 proteins tested in NIH3T3 cells show ligand-dependent stability, covering proteins that differ in size, fold, function and localization.

Fusion to an FKBP12 F36V destabilizing domain can both destabilize a normally stable protein and stabilize normally short-lived proteins. CDK1, an otherwise stable kinase, is degraded without Shld1, while the short-lived cell-cycle regulators p21 and securin, which the APC complex normally targets for degradation, are stabilized by Shld1 in NIH3T3 cells.

The transmembrane glycoprotein CD8α fused at its C terminus to the D100G or L106P destabilizing domain shows Shld1-dependent expression in NIH3T3 cells, measured by flow cytometry with a FITC-conjugated anti-CD8α antibody on live cells. Cell-surface CD8α decreases after Shld1 is removed, which suggests that the FKBP12 destabilizing domains can recruit cellular machinery for internalization and degradation of a membrane-bound protein.

Shld1 control of FKBP12 F36V destabilizing domain fusions of constitutively active small GTPases reversibly switches cell morphology in NIH3T3 cells. After 24 hr of 1 µM Shld1, L106P-RhoA Q63L induces stress fibers, L106P-Cdc42 Q61L induces filopodia, and Arl7 Q72L-L106P induces a shrunken-cell phenotype. After Shld1 is washed out (24 hr for RhoA, 48 hr for Cdc42 and Arl7), cells return to fibroblast-like morphologies indistinguishable from mock-treated cells, and more than 90% of cells show the predicted phenotype.