A method for conditional regulation of protein stability in native or near-native form

Summary

Here, we report a method to regulate cellular protein levels by introducing a ubiquitin variant between a destabilizing domain (DD) and the regulated protein. When produced in the absence of a stabilizing ligand the DD dominates and the entire fusion protein is processively degraded by the proteasome. In the presence of the stabilizing ligand the fusion protein is metabolically stable and becomes a substrate for abundant ubiquitin-specific proteases, liberating a native, or a near-native protein-of-interest. This technique is thus particularly useful for the study of proteins whose free N-terminus is required for proper function. Additionally, removal of the DD in the presence of stabilizing ligand leads to higher expression levels of regulated protein when cells experience transient exposure to a stabilizing ligand, such as in a living animal receiving a single dose of a pharmacological agent as the stabilizing ligand.

Graphical Abstract

eTOC Blurb

The ability to control protein abundance in cells without perturbing the function of the regulated protein is a valuable tool for biomedical research. Miyamae et al. developed a method to conditionally regulate the protein stability in native or near-native form using a destabilizing domain and ubiquitin variant.

Introduction

Technology to conditionally control protein levels in cells can be a powerful tool for biomedical research. Targeting protein production at the level of DNA and RNA involving transcriptional switches, recombinases, RNA interference, or genome-editing tools are relatively well-established strategies (Ryding et al., 2001, Fire et al., 1998, Cheng and Alper, 2014). Though these have been widely used to modulate gene function at the transcriptional level, these methods are neither readily reversible nor tunable and can be accompanied by incomplete knockdown or off-target effects (Sigoillot and King, 2011). Furthermore, techniques that target the DNA or RNA precursor suffer from experimental delays, since existing protein molecules must be degraded.

To address these limitations, a variety of methods have been developed to directly regulate protein levels using cell permeable small molecules (Rakhit et al., 2014). The majority of these approaches involve the use of small compounds that recruit enzymes involved in the ubiquitin-proteasome system to the protein-of-interest (POI), thereby inducing its proteasomal degradation (Gosink and Vierstra, 1995, Matsuzawa et al., 2005, Zhou et al., 2000). One such method is adapted from the auxin-mediated degradation system in plants (Nishimura et al., 2009). A POI may be fused to auxin-inducible protein domain, which undergoes association with an F-box protein and SCF E3 ligase in the presence of indoleacetic acid (auxin), thereby promoting polyubiquitylation and degradation of the POI.

In a conceptually similar vein, Crews and Deshaies developed dimeric synthetic compounds comprised of a ligand for E3 ligase that is covalently linked to a small molecule possessing high affinity for a protein target. These bifunctional molecules, called proteolysis targeting chimeric molecules (PROTACs), enforce the colocalization of the POI with the E3 ligase resulting in ubiquitylation and degradation of the POI (Sakamoto et al., 2001). The generality of this approach may be limited by the requirement for a high-affinity ligand for the POI that can be modified to synthesize the PROTAC. Lin and coworkers have developed a method in which the POI is fused to the hepatitis C virus (HCV) protease and a peptide degron (Chung et al., 2015). Importantly, a peptide substrate for HCV protease is included between the POI and protease. Upon translation, the protease cleaves at the intramolecular substrate site, thus liberating the POI from the protease-degron fusion resulting in release of near-native POI. The addition of protease inhibitor prevents cleavage of the degron, ensuring degradation of the entire chimeric protein. These techniques, in which addition of ligand results in protein degradation, are useful for loss-of-function studies.

Our laboratory has previously developed a complementary strategy for regulating protein stability using cell-permeable small molecules to stabilize the POI. We engineered mutants of FKBP12 that are rapidly and constitutively degraded when expressed in mammalian cells (Banaszynski et al., 2006). These domains are called destabilizing domains (DDs), and this instability is conferred to any protein tagged with the DD resulting in the degradation of the entire fusion protein. The addition of a high-affinity ligand, Shield-1 in the case of the FKBP-derived DDs, protects the proteins from degradation allowing them to perform their cellular function. We used a similar discovery approach to create orthogonal DDs based on E. coli dihydrofolate reductase (Iwamoto et al., 2010), the human estrogen receptor ligand binding domain (Miyazaki et al., 2012), and the bilirubin-binding protein, UnaG (Navarro et al., 2016). The genetic fusion of any of these DDs to a POI ensures specificity of degradation, while the small molecule dependence provides rapid, reversible, and dose-dependent control over protein stability (Iwamoto et al., 2010, Banaszynski et al., 2008). However, this method requires a POI to be permanently tagged with the DD, usually at the N- or C-terminus. The terminal regions of proteins often encode functionally important sequences that are indispensable for proper function (e.g., sites for post-translational modification or interaction with membranes or partner proteins). Thus, fusion to a DD may compromise the protein function in some cases.

In this study we describe an approach to conditionally regulate protein stability without permanent fusion of a POI to the DD. The technique is called a liberation-prone degron (LIBRON), and it retains the basic features of the DDs in that DD-POI fusions are stabilized by cell permeable ligands. In the ligand-stabilized state the POI is liberated from the DD through the enzymatic action of abundant cellular proteases. Unlike other strategies (Lau et al., 2010), the LIBRON is a single ligand-single domain system. The POI need only be tagged with the LIBRON construct, and there is no need for additional genetic constructs to express additional genes encoding E3 ligases, proteases, or adaptor proteins. We demonstrate that LIBRON can be used to conditionally regulate the function of a protein whose free N-terminus is required for its biological activity. An additional benefit of the system is that the liberation of the DD allows the cleaved POI to be metabolically stable even in the absence of stabilizing ligand, suggesting a potential advantage for regulating proteins in living animals.

Results

Development of LIBRON

The rationale behind the LIBRON approach dictated the placement of a protease-sensitive cleavage domain between the DD and the POI. In the ligand-free state, cleavage should be relatively slow and the influence of the DD should dominate, so the entire polypeptide would be degraded by the ubiquitin-proteasome system (UPS). However, when the DD is stabilized by a ligand the protease(s) would have ample time to act, thus physically decoupling the POI from the DD. We chose the 76-residue ubiquitin (Ub) domain to test this strategy by making a construct encoding the FKBP-derived L106P DD followed by Ub then the POI (Figure 1A). Eukaryotic cells constitutively express many different ubiquitin-specific proteases (informally called deubiquitinating enzymes or DUBs) (Mevissen and Komander, 2017). We envisioned that proteins fused to the C-terminus of Ub would be liberated from the DD due to the activity of the DUBs.

Figure 1.

Development of LIBRON. (A) Schematic illustrations showing the destabilizing domain (DD) and liberation-prone degron (LIBRON) methods. Ub* represents a variant of Ub called Ub(K0) in which all seven lysines are mutated to arginine. See also Figure S1. (B) NIH3T3 cells were transduced with the constructs encoding Ub* or the indicated mutant and were treated with stabilizing ligand, Shield-1 (1 μM), for 24 h, after which lysates were prepared and immunoblotted using anti-GFP antibody. GAPDH is the loading control. Representative results of three independent trials are shown. See also Figure S2, S3, and S5. (C) NIH3T3 cells stably expressing the indicated constructs were treated with Shield-1 (1 μM) for 24 h, and fluorescence was quantified by flow cytometry. The experiment was performed in triplicate (n = 3), and error bars represent standard deviation. Significant differences between permissive and non-permissive states and among background levels of each constructs were assessed by two-tailed Student’s t-test and Tukey’s comparison test. a: significant difference (P < 0.0001) v.s. 3T3, b: v.s. non-treatment of DD-sfGFP, c: v.s. non-treatment of DD-Ub*G75S-sfGFP, d: v.s. non-treatment of DD-Ub-G75C-sfGFP See also Figure S3.

Ubiquitin is often used to signal protein degradation, typically when lysine residues on Ub are tagged with additional Ub molecules to create poly-Ub chains. The LIBRON approach requires that degradation be induced by the DD rather than polyubiquitylation, so we used a mutant of Ub in which all seven lysines are mutated to arginine. This ubiquitin variant is called Ub(K0) and is abbreviated as Ub* in this manuscript. For development of the LIBRON, we fused the L106P-Ub* to superfolder green fluorescent protein (sfGFP) to act as a POI. This three-part chimeric protein was stably expressed in NIH3T3 cells, and protein stability as well as liberation of sfGFP were analyzed by immunoblotting. Using Ub* with a wild-type C-terminus (LRGG are residues 73-76 of Ub), liberated sfGFP was detected even in the absence of the stabilizing ligand, Shield-1 (Figure 1A). We reasoned that cellular DUB activity is sufficiently high that cleavage occurs faster than the DD can be recognized and processed by the UPS (Figure S1A).

We focused on the penultimate Gly75 residue of Ub in order to tune the susceptibility of the Ub* domain to the cellular DUBs (Renatus et al., 2006). For example, substitution of Gly75 to Val is known to confer strong resistance to DUB-cleavage (Dantuma et al., 2000, Drag et al., 2008). Indeed, expression of the L106P-Ub*-sfGFP chimera encoding the G75V mutation revealed that uncleaved fusion protein is detected in the presence of Shield-1 but that no sfGFP is detected in the absence of Shield-1 (Figure 1B). This indicates that the DD dominates control of metabolic stability without liberation of the sfGFP (Figure S1B). We next sought to identify mutants of Gly75 that are moderately but not completely resistant to DUBs, leading to slow cleavage of the fusion protein when the DD is stabilized with Shield-1 (Figure S1C).

We prepared 18 additional constructs in which Gly75 was mutated to the other 18 possible amino acids. These constructs were stably expressed in NIH3T3 cells, and sfGFP expression was analyzed by immunoblotting (Figure S2). The G75A mutant behaves like wild-type; cleavage is too efficient for the DD to regulate stability. Conversely, fifteen mutants behaved like G75V. These fusion proteins were all sufficiently poor substrates for the DUBs that little to no liberated sfGFP was detected in Shield-1-treated cells. We observed that two mutants, G75S and G75C, liberate sfGFP when the cells are treated with Shield-1, whereas low levels of protein are observed in the absence of stabilizing ligand (Figure 1B and Figure S2). Stabilized GFP levels were also analyzed by analytical flow cytometry and fluorescence imaging (Figure 1C and Figure S3A), all indicating that these two mutants can regulate the stability of liberated protein in a ligand-dependent manner. We also confirmed that these two mutants display the same behavior when used with the orthogonal DHFR-derived DDs (Figure S3B).

Characterization of LIBRON

To further characterize the G75S and G75C constructs, we initially examined dose-dependent control of protein levels. NIH3T3 cells stably expressing each construct were treated with Shield-1 at various concentrations, and GFP levels were assessed by immunoblotting. Both constructs showed that cleaved and uncleaved fusion protein levels are regulated by Shield-1 in a dose-dependent manner (Figure 2A), demonstrating that both constructs allow tunable control of protein expression. The two constructs showed somewhat different ratios of cleaved to uncleaved substrate. Levels of liberated sfGFP were reliably higher in cells expressing the G75S construct relative to G75C. We reasoned that this difference in liberation of the sfGFP is due to the suitability of the G75S and G75C sequences as substrates for the DUBs.

Figure 2.

Characterization of LIBRON. (A) LIBRON constructs liberate a POI and regulate its stability in a dose-dependent manner using Shield-1. NIH3T3 cells stably expressing LIBRON constructs were treated with the indicated concentrations of Shield-1 for 24 h, and cell lysates were resolved using SDS-PAGE and immunoblotted with antibody against GFP. Representative results of two independent trials are shown. (B) NIH3T3 cells stably expressing LIBRON constructs were treated with Shield-1 (1 μM) and monitored over time. Immunoblotting was performed with antibody against GFP. GAPDH is the loading control in both panels. Representative results of two independent trials are shown.

To obtain more information about the relative cleavage rates of these two constructs we assessed time-dependent stability by immunoblotting. Cells stably expressing each construct were dosed with Shield-1 and then harvested for analysis at several time points. These constructs produced both cleaved and uncleaved fusion protein in a time-dependent manner upon addition of Shield-1 (Figure 2B), although the ratios of liberated sfGFP to full-length fusion protein differed. The G75S construct revealed detectable levels of liberated sfGFP at the earliest timepoints, and levels of liberated sfGFP appeared higher than uncleaved fusion protein by 8 hours. In contrast, the G75C construct produced only uncleaved fusion protein during the early timepoints with liberated sfGFP being detected around 8 hours after addition of Shield-1. This result suggests that the G75S sequence is more efficiently processed than the G75C mutant by the DUBs. This was supported by flowcytometry analysis which showed that the background level of G75C mutant was significantly lower than G75S mutant (Figure 1C). To further test this explanation we monitored the fate of the sfGFP after protein synthesis was blocked (Figure S4). Cells were treated with Shield-1 for 24 h then further incubated with cycloheximide (CHX) in the presence of stabilizing ligand for several timepoints. Translation of new protein was blocked whereas the synthesized protein is still stabilized by ligand and subjected to cleavage by DUBs. This allows us to estimate how efficient the cleavage happens for each construct. The sfGFP was nearly completely liberated from the DD-Ub* G75S mutant by 4 hours. The G75C mutant appears to be more resistant to DUB activity as the sfGFP appears to be approximately 50% liberated at the same time point. These data clearly indicate the difference in the relative cleavage rate between the two mutants.

To gain further insight into the behavior of these two constructs we monitored the levels of sfGFP upon withdrawal of Shield-1 using analytical flow cytometry (Figure 3A). Because it is not a substrate for DUBs, the G75V mutant behaves as if sfGFP is directly regulated by the DD (Banaszynski et al., 2006). In cells expressing the G75V mutant sfGFP levels decrease rapidly upon withdrawal of Shield-1, with sfGFP approaching background autofluorescence levels within 3 h. At the other extreme the wild-type C-terminal Ub sequence is an excellent DUB substrate, so sfGFP is efficiently liberated from the DD-Ub* sequence. The levels of sfGFP are generally unaffected by the withdrawal of Shield-1 in cells expressing this construct (Figure 3A).

Figure 3.

Monitoring the fate of liberated sfGFP. (A) NIH3T3 cells stably expressing fusion proteins encoding the indicated ubiquitin variants were treated with Shield-1 (1 μM) for 24 h, at which point purified FKBP F36V protein was added to the culture media to remove intracellular Shield-1, and sfGFP fluorescence was monitored by flow cytometry. (B) NIH3T3 cells stably expressing LIBRON constructs were treated with Shield-1 (1 μM) for 24 h, at which point FKBP F36V was added to the culture media to deplete intracellular Shield-1. Cells were further cultured for the indicated time, after which lysates were immunoblotted with antibody against GFP. GAPDH is the loading control. Representative results of three independent trials are shown. See also Figure S4.

Based on the results shown in Figure 2 we would expect the G75S construct to produce somewhat higher levels of liberated sfGFP relative to the G75C construct, assuming similar levels of transcription and translation of the respective mRNAs. Indeed, both mutant constructs display behavior in between the wild-type and G75V sequences. One hour following withdrawal of Shield-1, cells expressing the G75S construct possessed higher levels of GFP than cells expressing the G75C construct. This trend remained through the experiment, with GFP levels produced by the G75S construct consistently higher than G75C. By 24 hours post-withdrawal the expressed GFP levels were similar, although levels in the G75S cells were consistently higher than G75C, perhaps reflecting that rapid liberation of sfGFP led to higher background levels of liberated POI even in the absence of stabilizing ligand.

As flow cytometry cannot distinguish between cleaved and full-length fusion protein, we examined cell lysates using immunoblotting. In cells expressing G75S or G75C mutants, the uncleaved fusion proteins were rapidly degraded upon withdrawal of Shield-1 (Figure 3B). However, liberated sfGFP was detectable for both constructs between 8-24 hours following Shield-1 withdrawal. These data demonstrate that the LIBRON technique could be used to provide more durable levels of a desired POI when levels of stabilizing ligand are difficult to control, such as in living animals. Shield-1 levels can be controlled with high precision in cultured cells. Using the LIBRON, proteins of interest liberated from the DD may also be functional for longer periods of time, even if the animal has only received a single dose of stabilizing ligand.

We next investigated the ability of LIBRON to liberate the DD when fused to the C-terminus of the POI using constructs encoding sfGFP followed by Ub* variants then FKBP DD. These constructs should produce a POI-Ub* fusion protein which is not fully native, but we assumed that the placement of the LIBRON at the C-terminus of the POI would provide useful regulation of the POI stability in cases where placement of the LIBRON at the N-terminus interferes with proper function of the POI, especially, in case that the protein folding is disrupted by the placement of domain or tag at the N-terminus. The constructs were stably expressed in NIH3T3 cells, and the expression of sfGFP-Ub* fusion was analyzed by immunoblotting. We observed that the sequence of sfGFP-Ub*-DD can be recognized by DUBs (Figure 4A). In cells expressing Ub* encoding the wild-type Gly-Gly dipeptide at the C-terminus of Ub* the DD was efficiently liberated, and the sfGFP-Ub* fusion was detected in the presence and absence of Shield-1. This result is consistent with the previous finding that a fusion protein with C-terminal Ub domain is not destabilized (Qian et al., 2002). The DUB-resistant G75V construct revealed that the full-length fusion protein was observed in the presence of Shield-1 but that sfGFP-Ub* was not liberated. The constructs encoding G75S and G75C displayed the expected cleavage profile with the order of preference reversed from constructs with the LIBRON at the N-terminus. Both mutants produced the liberated fusion protein in the presence of Shield-1 but G75C constructs showed higher background levels of liberated sfGFP-Ub* fusion in the absence of stabilizing ligand (Figures 4A and B). The G75S mutant appears to be more resistant to DUBs relative to the G75C mutant as these background levels were significantly different, and decreased to the autofluorescence level in cells expressing G75S mutant in the absence of Shield-1. Both constructs display similar decay kinetics to the LIBRON fused to the N-terminus of the POI (Figure 4C). These results suggest that the LIBRON system is able to control protein stability with liberation when fused to either the N-terminus or the C-terminus of a fused partner protein.

Figure 4.

Characterization of LIBRON fused to the C-terminus of sfGFP (A) NIH3T3 cells were transduced with the constructs encoding sfGFP followed by Ub* mutants then FKBP DD and were treated with Shield-1 (1 μM) for 24 h, after which lysates were prepared and immunoblotted using anti-GFP antibody. Cyclophilin B is the loading control. Representative results of two independent trials are shown. (B) NIH3T3 cells stably expressing the indicated constructs were treated with Shield-1 (1 μM) for 24 h, and fluorescence was quantified by flow cytometry. The experiment was performed in triplicate (n = 3), and error bars represent standard deviation. Significant differences between permissive and non-permissive states and among background levels of each constructs were assessed by Student’s t-test and Tukey’s comparison test. a: significant difference (P < 0.0001) v.s. 3T3, b: v.s. non-treatment of sfGFP-Ub*G75G-DD, c: v.s. non-treatment of sfGFP-Ub*G75V-DD, d: v.s. non-treatment of sfGFP-Ub*G75S-DD, e: v.s. non-treatment of sfGFP-Ub*G75C-DD. n.s. means “not significant”. (C) NIH3T3 cells stably expressing constructs were treated with Shield-1 (1 μM) for 24 h, at which point FKBP F36V was added to the culture media to deplete intracellular Shield-1. Cells were further cultured for the indicated time, after which lysates were immunoblotted with antibody against GFP. Cyclophilin B is the loading control. Representative results of two independent trials are shown.

Application of LIBRON

To test the potential generality of the LIBRON technique we fused the G75S and G75C constructs to other protein targets, including apoptosis regulator, Bax, and subunit of replication protein A, hRPA2. Cells stably transduced with each of these fusion proteins were incubated with or without Shield-1 for 24 h, then cell lysates were resolved by SDS-PAGE and immunoblotted with the indicated antibodies. The LIBRON constructs are able to regulate the stability of both proteins with liberation (Figure 5). Additionally, the trend in DUB efficiency that we observed with sfGFP as the POI was similarly observed for Bax and hRPA2, with the G75S mutant being more efficiently processed by DUBs relative to the G75C mutant.

Figure 5.

Potential generality of LIBRON. The LIBRON constructs were fused to the N-termini of two different proteins and independently transduced into NIH3T3 cells. Cells were treated with vehicle or Shield-1 (1 μM) for 24 h then lysates were immunoblotted with antibody against hemagglutinin antigen (HA; the tag was fused to C-terminus of each fusion protein). GAPDH is the loading control. Representative results of three independent trials are shown. See also Figure S5.

Finally, we sought to test the LIBRON technique in a functional setting. First we tested the functionality of H-ras, which is a small GTPase that regulates cell division, proliferation, invasion, and metastasis. H-Ras exerts its function by activating several downstream effectors including the Raf/MEK/Erk pathway resulting in cellular responses. The constructs were stably transduced into NIH3T3 cells, and protein stability was assessed by immunoblotting. We observed that LIBRON regulated the liberation of H-ras upon stabilization by Shield-1 (Figure 6A). To test the H-ras function liberated by LIBRON construct, we examined the phosphorylation of Erk1/2 by immunoblotting. In cells expressing either the G75S or the G75C construct, phosphorylated Erk1/2 is increased by addition of Shield-1, with the G75C mutant showing a stronger difference between treatment with vehicle or Shield-1 (Figure 6B). With G75C being less suceptible to cellular DUBs, lower levels of liberated H-ras in the absence of Shield-1 reduce tonic phosphorylation of Erk1/2.

Figure 6.

Application of LIBRON. (A) LIBRON constructs were fused to the N-terminus of H-ras-HA and transduced into NIH3T3 cells. Cells were treated with vehicle or Shield-1 (1 μM) for 24 h then lysates were immunoblotted with antibody against hemagglutinin antigen (HA; the tag fused to C-terminus of fusion protein). GAPDH is the loading control. Representative results of three independent trials are shown. (B) LIBRON constructs were fused to the N-terminus of H-ras and transduced into NIH3T3 cells. Cells were treated with vehicle or Shield-1 (1 μM) for 24 h then lysates were immunoblotted with antibody against phosphorylated Erk1/2. Erk2 is the loading control. Representative results of three independent trials are shown. (C) LIBRON constructs were fused to the N-terminus of Nef and transduced into Jurkat cells. Cells were treated with vehicle or Shield-1 (1 μM) for 24 h then immunoblotted with antibody against SIV Nef. GAPDH is the loading control. Representative results of two independent trials are shown. Image was prepared by deletion of lanes of SIV-Nef expression sample on the right side of lane of “Jurkart”. (D) FACS histogram of Jurkat cells expressing LIBRON-conjugated Nef. Cells expressing LIBRON-conjugated Nef were treated with Shield-1 (1 μM) for 24 h. CD3 on the cell surface was detected with FITC-conjugated anti-CD3 antibody using analytical flow cytometry. Representative results of two independent trials are shown. See also Figure S5.

The N-terminus can be an important functional site for many proteins, and the placement of a fusion partner at the N-terminus can reduce or eliminate proper function. In cases where a free N-terminus is important for protein function we expect the LIBRON technique to provide robust expression control of functional proteins. As a test system, we focused on the simian immunodeficiency virus negative regulatory factor (SIV Nef), which plays an important role for lentivirus replication in infected host cells. When expressed in primate T cells, SIV Nef downregulates cell surface receptor markers such as the T-cell receptor (TCR)-CD3 complex via endocytosis (Bell et al., 1998). Given that the N-terminus of SIV Nef is myristoylated and that this modification is essential for its cellular function (Swigut et al., 2003), we assumed that LIBRON could be used to regulate the function of SIV Nef expressed in T-cells in a ligand-dependent manner.

The two LIBRON constructs were fused to the second amino acid (Gly2) of SIV Nef without a linker, so a substrate suitable for myristoylation would be produced upon liberation. The constructs were stably transduced into Jurkat T-cells, and populations of cells were treated with Shield-1 for 24 h, then protein stability was assessed by immunoblotting. This experiment confirmed that LIBRON allowed the liberation of Nef upon stabilization by Shield-1 (Figure 6C). Next, we used fluorescent anti-CD3 antibodies to quantify CD3 levels on the surface of Jurkat cells expressing both constructs. Addition of Shield-1 should produce liberated, functional Nef, which should lead to low levels of CD3 on the cell surface. Indeed, in cells expressing G75S or G75C constructs, the population of CD3-negative cells increased by 1.8-2.1 times upon addition of Shield-1 (Figure 6D). In contrast, the G75V construct did not show the down-regulation of CD3 in the presence of Shield-1 since it does not allow Nef to be cleaved. These results indicate that LIBRON is an effective tool for ligand-dependent control of protein stability for partners whose N-terminus is essential for cellular function.

Discussion

We developed a variant of the existing DD technique called LIBRON that enables the release of the POI from the DD degron following ligand stabilization. Importantly, DD-mediated degradation controls the metabolic fate of the POI in the absence of ligand. To achieve this goal, we made the C-terminus of a lysine-free ubiquitin variant a poorer substrate for ubiquitin-specific proteases. The two single amino acid mutations, G75S and G75C, of the Gly75 residue possess modest resistance to the enzymatic activity of the DUB enzymes. Insertion of either Ub* mutant between the C-terminus of an FKBP-derived DD and the N-terminus of a POI allows us to liberate the POI without losing pharmacologic control of the stability of the fusion protein.

Interestingly, the two mutants displayed different profiles in cleavage rate. When the DD-Ub* LIBRON element is fused to the N-terminus of a POI, the G75S mutant appears to be a better substrate for the DUBs as more POI is liberated relative to the cysteine mutant. The G75C mutant, which is processed less efficiently by the DUBs, delivers lower background levels of the POI in the absence of stabilizing ligand. We found that the order of efficiency in cleavage of two mutant sequences is reversed when the relative positions of POI and DD are reversed. This observation suggests that the efficiency of liberation is governed primarily by the structural characteristics of the substrate.

Both LIBRON constructs are able to regulate the cellular function of the H-ras protein which induces phosphorylation of Erk1/2. We also demonstrated that LIBRON can regulate the function of SIV Nef protein, which requires myristoylation of its N-terminus. Cell surface levels of CD3 were downregulated by Nef in a ligand-dependent manner, suggesting that the LIBRON system should be useful for studies of proteins whose unmodified N-terminus is required to be functional. DUBs primarily recognize the ubiquitin domain, so it should be possible to liberate partner proteins displaying any nascent N-terminal residue. We also noticed that the effect on the functional control of SIV-Nef was modest when compared with H-ras. We reasoned that this likely depends on the vector system and basal protease activity. While H-ras and other proteins exemplified in this study were expressed by retroviral vector, the lentiviral vector was used for Nef expression. Stable and robust expression may lead to the higher background activity observed. Another possible factor is the basal activity of ubiquitin specific protease in Jurkat cells. Abundant DUB activity in Jurkat cells may liberate a small amount of Nef from the LIBRON domain before proteasomal degradation.

This experiment also highlights another important feature of the LIBRON system, which is that Nef is liberated from the LIBRON element when Ub* is connected to Nef without the use of a linker (Figure S5B). This Nef construct showed the same profile of cleavage as other POIs used in this study, all of which were fused to the LIBRON element with two extra amino acids (Met-His) introduced by a restriction enzyme site to facilitate cloning (Figure S5A). We also tested connecting peptides comprised of either Gly-Ser residues or the first several residues of wild type Ub at the N-terminus of the POI (to mimic a poly-Ub substrate), but the cleavage profile was not changed (data not shown). Our observations are consistent with those of others supporting a model wherein the structural features recognized by DUBs depend more strongly on the 4-5 final amino acids of ubiquitin than the amino acids that follow the ubiquitin domain (Mevissen and Komander, 2017).

Previously, Lin’s group developed a system for a similar purpose, termed SMASh in which the POI is fused to the hepatitis C virus (HCV) protease and a peptide degron (Chung et al., 2015). The placement of a peptide substrate for HCV protease between the POI and protease allows POI to be liberated co-translationally by cleavage of HCV protease. The addition of protease inhibitor prevents cleavage of the degron, ensuring degradation of the entire chimeric protein. In this respect, SMASh can be thought of as drug-dependent protein degradation system. The LIBRON and underlying DD technologies are drug-dependent protein stabilization systems. These two approaches are thus complementary, allowing users to choose the most suitable system based on their experimental design. Furthermore, SMASh requires the inclusion of approximately 10 amino acids as a cleavage site, which leaves a few residues at the C- or N-terminus of the POI following processing. On the other hand, the LIBRON does not require a linker between Ub* and POI, so it can produce the native form of its partner protein upon liberation. Additionally the LIBRON system uses fully human proteins, avoiding the use of potentially immunogenic non-human sequences, which is likely to be important to achieve a durable response in a translational setting.

One important feature of this technique is that the protein liberated from either N- or C-terminus site of LIBRON element remains stable for several hours after ligand withdrawal, provided its intrinsic metabolic half-life is relatively long. Using the DD system in cultured cells the experimenter has complete control over the dosing of the stabilizing ligand. The situation is more challenging when the DDs are used in living animals, where the pharmacokinetic properties of the stabilizing ligand will exert a strong influence on the performance characteristics of the technology. Rapid elimination of the stabilizing ligand would lead to rapid degradation of DD-tagged protein partners, possibly preventing the regulated protein from reaching the desired experimental or therapeutic levels. With the LIBRON technique the regulated protein is liberated from the DD, so clearance of the stabilizing ligand will not necessarily lead to rapid degradation of the partner protein. Although separating the DD from the POI reduces the level of pharmacological control, it provides a benefit for studies in living animals. Given that it can be difficult to maintain the physiological concentration of a stabilizing drug at optimal concentrations, the LIBRON technique may be attractive for regulating clinically beneficial proteins. Further preclinical studies will open the doors for conditional protein control in human cell and gene therapy.

STAR Method

RESOURCE AVAILABILITY

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by Lead Contact, Thomas J. Wandless (wandless@stanford.edu).

Materials Availability

All unique reagents generated in this study are available from the Lead Contact without restriction. Plasmid generated in this study will be deposited to Addgene (https://www.addgene.org/).

Data and Code Availability

The published article includes all biological data generated during this study. No unique code was generated in this study.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Cell line

The male murine fibroblast cell line NIH/3T3 (Cat#CRL-1658) was cultured in DMEM supplemented with 10% heat-inactivated fetal bovine serum, 2 mM glutamine, 100 U/mL penicillin, and 100 μg/mL streptomycin. The phoenix ecotropic cell line (female, gifted from The Nolan Laboratory, Stanford University) was cultured in DMEM supplemented with 10% heat-inactivated fetal bovine serum, 2 mM glutamine, 100 U/mL penicillin, and 100 μg/mL streptomycin. The human female embryonic kidney cell line HEK293T (Cat#CRL-3216) was cultured in DMEM supplemented with 10% heat-inactivated fetal bovine serum, 2 mM glutamine, 100 U/mL penicillin, and 100 μg/mL streptomycin. The human male cell line Jurkat E6-1 (Cat#TIB-152) was cultured in RPMI 1640 supplemented with 10% heat-inactivated fetal bovine serum, 2 mM glutamine, 100 U/mL penicillin, 100 μg/mL streptomycin and 10 mM HEPES (pH 7.2-7.5). All cell lines were maintained at 37 °C and 5% CO₂ in humified atmosphere.

Bacterial Strain

The Escherichia coli BL21(DE3) cells were used for recombinant protein production. E.coli BL21(DE3) cells were grown in Luria Broth (LB) medium at 37°C until OD600 of 0.5 - 0.7, followed by induction with 0.5 mM isopropyl-β-thiogalactoside (IPTG) at 18°C for 18 h.

METHOD DETAILS

DNA constructs

cDNA sequences used as gene-of-interest are: sfGFP (superfolder green fluorescent protein, Genbank AB971579), Ub(K0) (Homo sapiens Ubiquitin-K0, Addgene plasmid #17603), Hras (Homo sapiens replication Ras family small GTP binding protein H-ras, Genbank AF493916), mBax (Mus musculus BCL2-associated X protein, Genbank NM_007527), hRPA2 (Homo sapiens replication protein A2, Genbank NM_002946), and SIV Nef (SIV mac251 Nef, Genbank AF128391.1). Genes encoding the proteins tested as fusions to the LIBRON domain were cloned by standard techniques (Figure S5A) in the retroviral pBMN vector encoding IRES-mCherry. Genes encoding SIV Nef lacking the first methionine residue was fused to the LIBRON domain (Figure S5B) and cloned into the third-generation lentiviral vector, pELPS encoding IRES-mCherry.

Transfections and transductions

Retroviruses were produced by transfecting phoenix ecotropic packaging cells with pBMN plasmids and lentiviruses were produced by transfecting HEK293T cells using standard TransIT-LT1 (Mirus) protocols. Supernatants containing virus were harvested 48 h post-transfection and filtered with 0.45 μm filter. Cells were incubated with the viral supernatants supplemented with 4 μg/mL polybrene for 4 h at 37 °C then cultured in growth media for 48 h to allow for viral integration.

Immunoblotting and Antibodies

Cells were washed with PBS twice and lysed with the buffer. Lysates were resolved by SDS-PAGE and transferred to PVDF membrane (Millipore). Detection was achieved using horseradish peroxidase-conjugated secondary antibodies and Immobilon (Millipore). Antibodies were anti-GFP (mouse, JL-8, Clontech), anti-HA (rat, 3F10, Roche), anti-SIV Nef (mouse, sc-65911, Santa Cruz Biotech), anti-GAPDH (mouse, 6C5, Abcam), anti-phosphorylated Erk1/2 (mouse, M9692, Sigma-Aldrich), anti-Erk2 (rabbit, M7556, Sigma-Aldrich), and anti-cyclophylin B (rabbit, PA1-027A, ThermoFisher Scientific). Chemiluminescence signal was recorded by Odyssey Fc Imaging System (LI-COR) or by an exposure of the membrane to a X-ray film.

Flow cytometry

Cells were trypsinized and resuspended in culture media before flow cytometry analysis using a BD FACS Aria or FACS LSRII with no less than 10,000 events represented. Collected data were analyzed using FlowJo software.

Recombinant protein

Recombinant FKBP (F36V) was prepared as previously described (Egeler et al., 2011) with slight modification. pET15b vector encoding FKBP F36V was transformed into BL21(DE3) cells. Cells were grown in Luria-Bertani medium (LB) at 37°C to an optical density at 600 nm of 0.5–0.7. Protein expression was induced with 0.5 mM isopropyl-β-D-thiogalactoside at 18°C for 18 h, after which cells were collected by centrifugation at 5,500 rpm at 4°C for 20 min. The pellet was lysed in lysis buffer (50 mM NaH₂PO₄, 300 mM NaCl, 10 mM imidazole at pH 8) supplemented with 1 mM phenylmethylsulfonyl fluoride (PMSF) and 1 mg/ml lysozyme, then left on ice for 30 min. After sonication, Triton X-100 was added to 1%, and after 30 min incubation on ice, the solutions were centrifuged at 19,000 rpm for 1 h at 4°C. The soluble fraction was bound to Ni-NTA resin at 4°C for 1 hr. The resin was washed three times with wash buffer-I (50 mM NaH₂PO₄, 500 mM NaCl, 10 mM imidazole, 0.2% Triton X-100 at pH 8), then further washed several times with wash buffer-II (50 mM NaH₂PO₄, 500 mM NaCl, 20 mM imidazole at pH 8) until eluent was free of protein by absorbance at 280 nm. The protein was eluted with elution buffer (50 mM NaH₂PO₄, 300 mM NaCl, 250 mM imidazole at pH 8). Eluted solutions were combined and dialyzed into dialysis buffer (PBS with 4% glycerol) and stored at −80°C.

Withdrawal of Shield-1

To remove S1 from culture media for DD-Ub-GFP cells, cells were washed with culture media then cultured in media containing 5 μM FKBP (F36V) protein to provide an extracellular thermodynamic sink for the Shield-1 ligand.

Detection of CD3 level on the cell surface

3 x 10⁵ Jurkat cells were pelleted (200 x g) and washed with PBS once. The cells were resuspended with PBS containing 2% fetal bovine serum (FBS) and centrifuged again at 4 °C, then incubated with 100 μL of PBS containing 2% FBS and 4 μg/mL of FITC-conjugated anti-CD3 antibody (OKT3, # 317306, Biolegend) for 30 min on ice. Labeled cells were washed three times with PBS containing 2% FBS at 4 °C. Cells were resuspended in 200 μL PBS containing 2% FBS and analyzed by BD FACS LSRII.

QUANTIFICATION AND STATISTICAL ANALYSIS

Quantification data is represented as mean with standard deviation (SD). n represents technical replicated for each measurement and is reported in the respected Figure legends. Data was fitted and analyzed for statistical significance using Prism8 software (GraphPad). Comparisons of two groups were performed with two-tailed Student’s t-test, and comparisons of more than two groups were performed with one-way analysis of variance and Tukey’s test.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse monoclonal anti-GFP	Clontech	Cat#JL-8
Rat monoclonal anti-HA	Roche	Cat#3F10
Mouse monoclonal anti-SIV Nef	Santa Cruz Biotech	Cat#sc-65911
Mouse monoclonal anti-GAPDH	Abcam	Cat#6C5
Mouse monoclonal anti-phosphorylated Erk1/2	Sigma-Aldrich	Cat#M9692
Rabbit polyclonal anti-Erk2	Sigma-Aldrich	Cat#M7556
Rabbit polyclonal anti-cyclophylin B	ThermoFisher Scientific	Cat#PA1-027A
Horse anti-mouse IgG, HRP-linked	CST	Cat#7076
Goat anti-rabbit IgG, HRP-linked	CST	Cat#7074
Goat anti-rat IgG, HRP-linked	CST	Cat#7077
Mouse monoclonal FITC anti-human CD3	Biolegend	Cat#317306
Bacterial and Virus Strains
Escherichia coli BL21(DE3)	New England BioLabs	Cat#C2527I
Chemicals, Peptides, and Recombinant Proteins
Shield-1	AOBIOUS	Cat#AOB1848
Polybrene (Hexadimethrine Bromide)	Sigma-Aldrich	Cat#H9268
Immobilon	Millipore	Cat#WBKLS0500
FKBP F36V protein	This paper	Egeler et al., 2011
Experimental Models: Cell Lines
Mouse: NIH/3T3 cells	ATCC	Cat#CRL-1658
Human: Phoenix ecotropic cells	The Nolan Laboratory, Stanford University	N/A
Human: HEK293T cells	ATCC	Cat#CRL-3216
Human: Jurkat cells, Clone E6-1	ATCC	Cat#TIB-152
Recombinant DNA
pBMN-IRES-mCherry	This paper	N/A
pELPS-IRES-mCherry	This paper	N/A
pBMN-FKBP DD(L106P)-sfGFP-imCherry	This paper	N/A
pBMN-FKBP DD(L106P)-Ub*G75G-sfGFP-imCherry	This paper	N/A
pBMN-FKBP DD(L106P)-Ub*G75V-sfGFP-imCherry	This paper	N/A
pBMN-FKBP DD(L106P)-Ub*G75S-sfGFP-imCherry	This paper	N/A
pBMN-FKBP DD(L106P)-Ub*G75C-sfGFP-imCherry	This paper	N/A
pBMN-FKBP DD(L106P)-Ub*G75S-H-ras-imCherry	This paper	N/A
pBMN-FKBP DD(L106P)-Ub*G75C-H-ras-imCherry	This paper	N/A
pBMN-FKBP DD(L106P)-Ub*G75S-H-ras-HA-imCherry	This paper	N/A
pBMN-FKBP DD(L106P)-Ub*G75C-H-ras-HA-imCherry	This paper	N/A
pBMN-FKBP DD(L106P)-Ub*G75S-hRPA2-HA-imCherry	This paper	N/A
pBMN-FKBP DD(L106P)-Ub*G75C-hRPA2-HA-imCherry	This paper	N/A
pBMN-FKBP DD(L106P)-Ub*G75S-mBax-HA-imCherry	This paper	N/A
pBMN-FKBP DD(L106P)-Ub*G75C-mBax-HA-imCherry	This paper	N/A
pELPS-SIV-Nef-imCherry	This paper	N/A
pELPS-FKBP DD(L106P)-Ub*G75V-SIV-Nef-imCherry	This paper	N/A
pELPS-FKBP DD(L106P)-Ub*G75S-SIV-Nef-imCherry	This paper	N/A
pELPS-FKBP DD(L106P)-Ub*G75C-SIV-Nef-imCherry	This paper	N/A
pMD2.G	The Trono lab, École polytechnique fédérale de Lausanne	Addgene Plasmid #12259
pMDLg/pRRRE	Dull et al., 1998	Addgene Plasmid #12251
pRsv/Rev	Dull et al., 1998	Addgene Plasmid #12253
pET15b-FKBP (F36V)	Egeler et al., 2011	N/A
Software and Algorithms
FlowJo	BD	https://www.bdj.co.jp/biosciences/flowjo/flowjo.html
Prism8	GraphPad Software	https://www.graphpad.com/scientific-software/prism/

Highlight.

Control of cellular protein stability using temporary fusion of destabilizing domain

Ubiquitin variant was inserted between destabilizing domain and protein of interest

Target protein was released by endogenous deubiquitinase from C-terminus ubiquitin

Tuning of cleavage speed by a point mutation on C-terminus of ubiquitin

Significance.

Techniques to conditionally control protein abundance in cells without disrupting the function of the regulated protein are desirable tools for biomedical research. In this study, we developed a method called the liberation-prone degron (LIBRON) to regulate cellular protein levels by introducing a ubiquitin variant between a destabilizing domain (DD) and the target protein. The LIBRON can be used to liberate the protein partner from the DD degron through the enzymatic action of abundant cellular proteases following ligand stabilization. In the absence of a stabilizing ligand the DD dominates and the entire fusion protein is processively degraded by the proteasome. This technique enables the predictable control of the desired partner proteins without a DD fused to either terminus of a protein-of-interest. Since protein termini can encode important functional sequences, the LIBRON system is a useful tool for the study of protein whose unmodified termini are required to be functional. We also demonstrated that the partner protein becomes long-lived once it is liberated from the DD, even if the stabilizing ligand is removed. Given that it can be difficult to maintain consistent levels of stabilizing ligands in living animals the LIBRON system should provide unique benefits for in vivo studies or possible clinical use.