Phenotypes on demand via switchable target protein degradation in multicellular organisms

Phenotypes on-demand generated by controlling activation and accumulation of proteins of interest are invaluable tools to analyse and engineer biological processes. While temperature-sensitive alleles are frequently used as conditional mutants in microorganisms, they are usually difficult to identify in multicellular species. Here we present a versatile and transferable, genetically stable system based on a low-temperature-controlled N-terminal degradation signal (lt-degron) that allows reversible and switch-like tuning of protein levels under physiological conditions in vivo. Thereby, developmental effects can be triggered and phenotypes on demand generated. The lt-degron was established to produce conditional and cell-type-specific phenotypes and is generally applicable in a wide range of organisms, from eukaryotic microorganisms to plants and poikilothermic animals. We have successfully applied this system to control the abundance and function of transcription factors and different enzymes by tunable protein accumulation.


Comparison of current methods for conditional protein degradation directly acting on target protein levels
The lt-degron is a notable addition to the portfolio of existing tools for modulation of protein function. First, the obsolete rational trial-and-error generation and testing of supposed ts mutant variants of a POI can be avoided. Second, many reported tools focus on regulating the concentration of POIs through manipulating synthesis or the conformation rather than stability of present POI fusions or that they rely on transcriptional control and therefore is dependent on the intrinsic half-lives of inducers, mediators and targets. The lt-degron allows manipulation of degradation rates and therefore impinges directly on the level of the POI activity or function. Third, the lt-degron is a modular approach that gives control over a wide variety of target POIs in a number of host systems, Forth, and perhaps most importantly, the lt-degron works reliably and reversibly in multicellular organisms across the kingdoms due to a universal induction mode via temperature.
Then, the lt-degron method is primarily about conditional, reversible and direct control of protein levels. So far, the "classical" heat-induced N-degron system 1 involving the K1 cassette harboring DHFR P67L has only been used in yeasts and cell culture to conditionally switch POIs and requires restrictive temperatures as high as 37°C to 42°C. 5,7,8,9 The biggest caveat of this technology was to date that these temperatures are beyond the physiological ranges of many multicellular organisms, e.g. plants. 10,11,12,13,14,15 Alternative methods for conditional protein shut-off which directly act on the level of target protein degradation are still largely limited to cells in culture or yeast as unicellular eukaryotes. 16 State-of-the-art inducible approaches for protein destabilization are compared in the following and rely mainly on 1) portable or dormant degrons that can be activated post-translationally, 2) small molecules that act as molecular glue between POI and components of the degradation machinery, 3) degradation-mediating nanobodies targeted against the POI or a fusion containing the POI, or 4) reversible reconstitution of a degron or the POI

Dormant degrons
TIPI (TEV protease-mediated induction of protein instability) is a two-component pro-

Small molecule-mediators
The destabilizing domain (DD) system relies on a protein fusion technique involving protein stabilization dependent on the addition of the small molecule Shield1, a derivative of the immunosuppressant drug rapamycin. 21 The portable tag targeting the entire fusion to the proteasome is a constitutively unstable mutant variant of human FK506 binding protein 12 (FKBP12). The technology has been applied in a wide range of cell cultures. 16 Seed germination and plant growth seem not to be affected by Shield1 treatment suggesting that it is not toxic to plants 22 but further studies elaborating the potential use in multicellular organisms are required. A modification of the DD-FKBP system is LID (ligand-induced degradation)-FKBP that uses Shield1 for the opposite effect, i.e. causing POI instability after addition to the growth medium. 23 In a PROTACS (proteolysis/protein targeting chimeric molecules) approach, bivalent chimeric molecules mediate protein-protein interaction and target POIs and their fusions to a Skp1-Cullin-F-box (SCF) E3 Ub ligase complex for ubiquitination and degradation. The Fbox protein β-TRCP has been used in pioneering studies in cell culture together with Protac-1, a chimeric molecule consisting of a phosphopeptide which serves as E3 interaction interface and a small molecule ligand that enables interaction with the target POI. The ligand is used as a handle for then joining the two moieties. 24,25,26 Various POIs, selected ligands and E3 ligases may prefer different features of the interaction-mediating small molecules and a custom chemical synthesis is mandatory and likely to require a specific design for a particular small molecule ligand.
The auxin-inducible degron (AID) 27  Ub ligase complex. 30 Another small molecule-mediated approach is the tagging of a GOI by integration of an E. coli DHFR (eDHFR) degron within the genome via homologous recombination. This leads to sensitivity after deprivation of the structurally stabilizing antibiotic trimethoprim and effective depletion of eDHFR-tagged POIs. 31 It can only be applied in systems allowing homologous recombination and involves addition of a small molecule stabilizer.
Small molecule-mediated protein degradation is a strategy to create "chemical knockouts", however, the chemicals need to be introduced into the intracellular system which flags its preferential use in cultured cells. Application of these techniques can be difficult in multicellular organisms due to the need of application and penetration of the small molecules.

Nanobodies
The deGradFP (degrade Green Fluorescent Protein) protein depletion strategy directly targeting POI protein levels works in cell cultures and Drosophila. It depends on the presence of two stably transformed transgenes, i.e. a GFP fusion to a POI and an inducible anti-GFP nanobody. 32 This degradation mediator is comprised of the F-box domain of Drosophila Slmb and a single-domain camel antibody fragment and is under control of a chemically inducible promoters. Thus, the nanobody mediating target degradation is dependent on its own half-life and remains active until it gets degraded itself which happens in an unregulated manner. It will be challenging to introduce reversibility into this system. Two recent additions to the nanobody methods include transcriptionally controlled conditional systems. The one contains a modified recognition element, namely SPOP, replacing the generic ant-GFP nanobody and is directed against specific nuclear proteins. 33 Is was applied in cell culture and zebra fish embryos. The GFE3 system is based on a fusion between the E3 ligase RING domain of XIAP and the recombinant antibody-like protein GFP-GPHN.FingR (gephyrin.Fibronectin intrabodies generated with mRNA display). 34 FingR is derived from the fibronectin 10FNIII domain and binds to gephyrin with high affinity. The nanobody is induced by addition of an ecdysone analog and temporary expression of GFE3 was shown to inhibit synapse grow in zebra fish embryos.

Conformational inactivation
The light-dependent LOV2-mODC degron consists of the photosensitive LOV2 (LIGHT OXYGEN VOLTAGE 2) domain of Arabidopsis PHOTOTROPIN1 (PHOT1) which is activated upon irradiation with blue light, 35 undergoes a conformational change unmasking the previously cryptic degron of mODC (mouse ornithine decarboxylase). 36 The protein disruption technique using temperature-sensitive inteins 37, 38 involves conditionally splicing of chimaeric protein fusions. It depends on the challenging reconstitution of a functional protein from a synthetic POI consisting of two (or more) inactive POI fragments separated by intein sequences. Folding, stability, and solubility issues need to be taken into account and, most importantly, for each target, functional disruption sites have to be identified that guarantee reconstitution of functional POI. 39 In an example of a thermostable inteinmodified xylanase, recovery of enzyme activity occurred after activation by splicing at >59°C and was found to be significantly below the wild type levels. 40 This is a very good example for a biotechnological application of inteins in downstream processing after having obtained plant cell lysates or protein extracts.
Several of these methods, if modified, have certainly the potential to allow the generation and use of conditionally active proteins in multiple multicellular systems. More details on conditional genetic techniques can be found in a recent review of our lab. 16

Selection of metabolically unstable mutant DHFR variants in S. cerevisiae and generation of a low-temperature-controlled N-degron
In order to achieve lower restrictive temperatures, first, metabolically unstable DHFR variants were isolated after a random PCR mutagenesis of the wild-type DHFR sequence fused to the URA3 reporter gene encoding Orotidine 5'-phosphate decarboxylase (Ura3). A plasmid expressing DHFR-Ura3 from P CUP1 (pJH10) served as a template for an error prone PCR mutagenesis that amplified a fragment starting within P CUP1 and terminating within the 5' portion of URA3. The resulting PCR products were then used together with the large fragment of EcoRI + BamHI digested pJH10 (lacking the DHFR sequence) to transform S. cerevisiae. Incorporation of mutated versions of DHFR and recircularization of the plasmids occurred by in vivo recombination.
S. cerevisiae transformants expressing mutant DHFR-Ura3 proteins that were degraded by the proteasome were selected using strain JH5 (ura3-53 leu2-3,112 P GAL1 -UMP1). In this strain, the UMP1 gene, which encodes a proteasome maturation factor required for normal proteasome biogenesis, 6 is controlled by the galactose-inducible P GAL1 promoter. When P GAL1 is repressed in glucose-containing media, the strain behaves as a proteasome-deficient ump1∆ mutant with proteolysis defects. After mutagenesis, transformants were screened for functionality of the URA3 reporter by plating onto two different media: one medium without uracil and one with uracil in combination with the toxin precursor 5-fluoroorotic acid (FOA). 41 Yeast cells with an active URA3 gene can survive without external supplement of uracil but also convert FOA to fluorodeoxyuridine, which is toxic to cells. All yeast cells with sufficiently instable DHFR variants, will not be able to survive without uracil but recover on FOA. Thus, we used glucose media lacking uracil to select clones expressing DHFR-Ura3 fusion proteins with Ura3 activity. On galactose media with FOA, in contrast, variants were selected in which Ura3 activity was sufficiently low due to degradation by the UPS. This se-lection resulted in the isolation of the strain carrying plasmid pJH10 mutC2 containing the DHFR T39A,E173D variant carrying the two point mutations T39A and E173D which was used to generate construct K2 (Supplementary Figure 2c) (Supplementary Figure 5f,g). Testing the effect of the E173D mutation resulted in a higher flexibility and accessibility of Lys174, which was accompanied with conformational changes of the side chains of Arg29 and Lys33 (Supplementary Figure 5h,i).
None of the three here discussed single point mutations was predicted to lead to significant instability of the DHFR by various techniques 42,43,44,45,46 (Supplementary Table 2).
As modeling template, a mouse DHFR X-ray structure (PDB ID: 1U70) crystallized as a ternary complex with methotrexate and the cofactor NADPH was used. 47 Prior to modeling, all co-crystallized ligands were removed. Subsequently, the "Protonate 3D" tool of Molecular Operating Environment (MOE; version 2012.10; Chemical Computing Group) was applied to add all hydrogen atoms. Based on the prepared structure amino acid residue mutations under study were introduced using also MOE. For all these structures molecular dynamics simulations for 5 ns were performed with YASARA (http://www.yasara.org/index.html) 48 using the AMBER03 force field. Periodic boundary conditions including an appropriate water box were applied. The entire system was neutralized (pH 7.0) by adding sodium and chlorine ions. 49 Stability prediction of DHFR point mutations was performed with PoPMuSiC, CUPSAT, and TSpred which are computer-aided tools for the prediction of changes in protein stability upon point mutations an the rationale design of mutant proteins affected in their stability. PoPMu-SiC (http://dezyme.com/) evaluates the changes in stability of a given protein under single-site mutations on the basis of the structure of the protein. 45 The readout is the free energy change ΔΔG per sequence position where negative values for ΔΔG indicate lower stability. CUPSAT (Cologne University Protein Stability Analysis Tool; http://cupsat.tu-bs.de/) uses amino acidatom potentials and torsion angle distribution to assess the amino acid environment of the mutation site. In case of unfavorable torsion angles, the atom potentials may have higher impact on stability which results in a stabilizing mutation. 44 TSpred 46 based on PREDBUR predict the potential of point mutations based on their hydrophobicity and hydrophobic moment. 42,43