A Chemical Magnet: Approaches to Guide Precise Protein Localization

Abstract

Small molecules that chemically induce proximity between two proteins have been widely used to precisely modulate protein levels, stability, and activity. Recently, several studies developed novel strategies that employ heterobifunctional molecules that co-opt shuttling proteins to control the spatial localization of a target protein, unlocking new potential within this domain. Together, these studies lay the groundwork for novel targeted protein relocalization modalities that can rewire the protein circuitry and interactome to influence biological outcomes.

Graphical Abstract

Small molecules that function as chemical inducers of proximity (CIP) to rewire the circuitry of the cell have garnered significant attention as tools to interrogate biological processes and as therapies for many diseases¹. A central aim of these approaches is to leverage the cell’s intrinsic machinery to alter protein levels and/or activity by controlling post-translational modifications such as ubiquitylation, acetylation, and phosphorylation. Most commonly, CIPs are classified as molecular glues², which are monovalent molecules that engage two proteins, such as FK506 and auxin, or heterobifunctional molecules³. The development of heterobifunctional CIPs has emerged as an attractive approach due to its modularity in design that enables applications to a broad range of target proteins-of-interest (POIs) and protein modifiers. Connected by a chemical linker, one end of the heterobifunctional molecule binds to the POI, while the other end recruits the cellular protein modifier into proximity. This forced proximity leads to post-translational modification of the target POI by the action of a cellular protein modifier.

One highly explored utility of CIPs is for targeted protein degradation, which is enabled by a class of heterobifunctional CIPs known as PROteolysis-TArgeting Chimeras (PROTACs). PROTACs recruit an E3 ubiquitin ligase into proximity with a POI, leading to its ubiquitylation and proteasomal degradation⁴. PROTACs have gained significant traction as therapeutics with several clinical trials in diseases including cancer and autoimmune disorders^4,5. However, investigating proteins lacking binding ligands continues to be a challenge. Chemical-genetic strategies that employ PROTACs to degrade tagged proteins including dTAG⁶, HaloPROTACs⁷, and BromoTag⁸ have emerged as powerful technologies to pharmacologically study and validate undrugged proteins. On the heels of advances in PROTAC development, other heterobifunctional strategies to induce proximity between proteins have emerged to promote non-degradative post-translational modifications. Exemplary approaches include those that redirect cellular processes by inducing proximity between a POI with kinases for phosphorylation⁹, phosphatases for dephosphorylation^10,11, acetylases for acetylation^12,13, and deubiquitinases for removal of ubiquitin¹⁴. Milestones in these areas have been covered in depth in several outstanding reviews^15-18. In the field, there has been a long-standing interest in the development of approaches for dynamic spatial control over proteins. Here, we highlight recent studies from Gibson, Sadagopan et al.¹⁹, Ng et al.²⁰, and Shao et al.²¹ on the exciting development of CIPs to control subcellular protein localization.

Dynamic changes in subcellular protein localization are critical for normal cellular processes and are often dysregulated in disease. Eukaryotic cells are compartmentalized with organelles including the mitochondria, Golgi, endoplasmic reticulum, and lysosomes. Newly synthesized proteins in the cytoplasm are often translocated to their site of action to ensure the proper execution of cellular proliferation, survival, and differentiation. About 35% of newly synthesized proteins in the cytosol translocate to the endoplasmic reticulum, 25% to the nucleus, 5% to the mitochondria, and less than 1% to the peroxisome and lipid droplets²². Protein localization is influenced by a signal sequence. The nuclear localization signal (NLS) is a short peptide motif enriched in basic amino acids, where its strength is determined by interactions with the importin family that facilitate protein transport to the nucleus. By contrast, the nuclear export signal (NES) is a short peptide motif defined by the percentage of hydrophobic amino acids and its strength is determined by transport to the cytoplasm through the nuclear export receptor, CRM1²³.

The movement of proteins within each cellular compartment is tightly regulated, which can be a determining factor for signal transduction. For example, ERK predominantly resides in the cytoplasm in resting conditions and translocates to the nucleus upon growth factor stimulation, which is required for transcriptional activation and cell cycle progression²⁴. The dysregulation and/or mislocalization of oncoproteins or tumor suppressors contributes to cancer initiation and progression. EGFR, which is normally localized at the plasma membrane, can be mislocalized in the nucleus to drive aberrant gene expression programs and cell cycle progression in cancer²⁵. Tumor suppressors such as BRCA1, p53, and RB are present in the nucleus of normal cells but show cytoplasmic localization in cancer cells, leading to deregulation of target gene expression that is required for tumor suppression²⁶. Therefore, the ability to modulate protein localization and restore proper protein function of such proteins has exciting therapeutic potential.

Optogenetic approaches and self-localizing ligands (SLLs) are exemplary approaches that do not require protein heterodimerization and have been previously developed to manipulate protein localization in the cell. Optogenetic strategies have been employed for uncaging a genetically encoded lysine to unveil a bipartite NLS (OptoNLS) and induce nuclear localization of an OptoNLS fused protein²⁷. These photoactivatable approaches have also been applied to RNA-editing strategies that induce the fusion of an NLS to a protein to control its nuclear localization²⁸. These elegant approaches are highly enabling but are nonreversible and require ectopic expression systems. SLLs are synthetic ligands with small-molecule localization motifs that relocalize a protein from the cytoplasm to diverse compartments including the plasma membrane, nucleus, and cytoskeleton, with applicability in cell culture models and C. elegans^29,30. To further enhance the speed and reversibility of controlling protein localization, photoactivatable SLLs have been developed to recruit proteins from the cytoplasm to the nucleus³¹. Continued identification of small-molecule localization motifs is necessary to exploit the full potential of SLLs to relocalize proteins to diverse organelles and subcellular compartments. Complementing these approaches, the development of CIPs to control protein localization has remained under active investigation.

Chemical-genetic approaches that harness CIPs to control proteins fused to a universal tag effectively model the pharmacology of CIPs. These approaches have broad utility for studying biological processes prior to the heavy investment required to develop direct-binding CIPs. Previous work from Stuart Schreiber’s laboratory demonstrated that FKBP12/Cyclosporin A heterodimerization systems can be leveraged to control protein localization and function³². In this approach, a bifunctional molecule induces proximity between an FKBP12-tagged POI and a cyclophilin-tagged POI. This strategy was initially used to successfully recruit a POI to the plasma membrane (Fas) or nucleus (GFP), and as a method to induce transcriptional activation of a reporter gene. While powerful for modeling protein relocalization, this synthetic system required fusing two proteins with separate tags as well as the addition of a localization signal sequence. Furthermore, it was not extended to untagged cellular cargos to modulate the localization of POI. Through the advancement of synthetic ligands that bind mutant FKBP12^6,33,34 recent technologies have harnessed the endogenous protein machinery to control the levels and activity of mutant FKBP12-tagged proteins. For example, we developed the dTAG system, which employs heterobifunctional dTAG molecules to induce proximity with an endogenous E3 ubiquitin ligase to degrade an FKBP12^F36V-tagged POI in cellular and mouse models^6,35,36. Expanding beyond degradation, Christopher Parker’s laboratory developed the AceTAG approach for targeted acetylation of an FKBP12^F36V-tagged POI through the recruitment of P300/CBP¹³. Leveraging the FKBP12^F36V-tag, Gibson, Sadagopan et al.¹⁹ now advance the Schreiber laboratory’s prior work by developing a strategy termed Nuclear Import and Control of Expression (NICE). In this approach, the authors control nuclear localization of an FKBP12^F36V-tagged POI using heterobifunctional CIPs that recruit exogenously expressed BRD4 as a cargo protein.

BRD4 is a member of the BET bromodomain family and transcriptional co-activator that is localized in the nucleus and has been extensively used in heterobifunctional degrader approaches^37-39. The heterobifunctional compound, NICE-01, is composed of JQ1, a small molecule pan-BET bromodomain inhibitor used to recruit BRD4, conjugated to AP1867, a ligand that binds FKBP12^F36V (Fig.1A). As initial proof-of concept, the authors demonstrate successful import of FKBP12^F36V-tagged GFP into the nucleus upon NICE-01 treatment (Fig. 1B). Use of the system identified that having higher levels of BRD4, which was achieved with exogenous expression, compared to FKBP12^F36V-tagged POI, and ligand affinity, are important factors in NICE compound activity. Like PROTACs, NICE-01 also displayed the hook effect⁴⁰ with increasing compound concentrations.

Figure 1. Overview of strategies for targeted protein relocalization.

(A) Chemical structure of NICE Compound 1 (NICE-01), which is composed of an FKBP12^F36V binder conjugated with a linker to a BRD4 binder.

(B) Schematic illustrating that NICE-01 induces redistribution of an FKBP12^F36V-tagged POI from the cytoplasm into the nucleus through shuttling by BRD4.

(C) Chemical structure of TRAM Compound 1, which is composed of an FKBP12^F36V binder conjugated with a linker to an ecDHFR binder.

(D) Schematic illustrating that TRAM Compound 1 induces nuclear import of an FKBP12^F36V-tagged POI or nuclear export of an ecDHFR-tagged POI. The shuttling signal sequence strength influences nuclear import or export.

(E) Chemical structure of QS57, which is composed of a 14-3-3σ binder conjugated with a linker to a BRD4 binder.

(F) Schematic illustrating that QS57 induces redistribution of BRD4 from the nucleus to the cytoplasm through shuttling by 14-3-3σ.

Figure created with BioRender (BioRender.com).

Abbreviations: NICE, Nuclear Import and Control of Expression; POI, protein-of-interest; TRAM, Targeted Relocalization Activating Molecules

The authors next extended their approach for targeted relocalization of other key oncogenic proteins that are often altered and/or mislocalized in cancer such as PIK3CA and NPM1. Interestingly, the nuclear localization of FKBP12^F36V-tagged PIK3CA^E545K, which cannot passively diffuse across the nuclear pore, was slower and less efficient than FKBP12^F36V-tagged GFP. This suggests that induction of nuclear import with this strategy is not equally efficient across all targets. Next, the evaluation of NPM1 highlighted the potential applications of NICE compounds for the relocalization of cancer drug targets. In acute myeloid leukemia, mutations in NPM1 known as NPM1c introduce a neo-NES leading to its cytosolic mislocalization. Prior work has shown that targeted degradation or indirect nuclear relocalization of NPM1c successfully reverses the oncogenic effects due to its mislocalization⁴¹. Using high-resolution microscopy, the authors observed the relocalization of FKBP12^F36V-tagged NPM1c into multimeric BRD4 phase-separated condensates upon NICE-01 treatment. Notably, improved nuclear localization of FKBP12^F36V-tagged NPM1c was observed in cells with higher BRD4 expression. Whether this was a similar localization pattern to wild-type NPM1 or led to the functional effects observed in prior studies was not evaluated. Finally, the authors assessed the potential of recruitment of a transcription factor, IRF1, from the cytosol to the nucleus to rewire gene expression. An NES was added to the FKBP12^F36V-tagged IRF1 to first localize IRF1 to the cytoplasm. Exogenous BRD4 expression and NICE-01 treatment led to FKBP12^F36V-tagged IRF1 nuclear import and downstream target gene activation, highlighting the functional consequences of successful relocalization of IRF1.

While Gibson, Sadagopan et al. harnessed the potential of BRD4 to relocalize target proteins, Ng et al.²⁰ developed an orthogonal approach termed Targeted Relocalization Activating Molecules (TRAMs) to control protein localization through trafficking by a shuttling protein. To model effective protein relocalization into the nucleus or cytoplasm, the authors developed a heterobifunctional molecule, TRAM Compound 1, that engaged two widely used tags, FKBP12^F36V and the Escherichia coli dihydrofolate reductase domain (ecDHFR) (Fig. 1C). The authors leveraged an mCherry-ecDHFR fusion with and without an NES to determine the importance of localization signal strength for nuclear export or import of an FKBP12^F36V-tagged protein with TRAM Compound 1 (Fig. 1D). TRAM Compound 1 treatment successfully co-opted an mCherry-ecDHFR fusion appended with an NES to redistribute an FKBP12^F36V-tagged NMNAT1 from the nucleus to cytoplasm, with evidence of the hook effect at higher concentrations. Reciprocally, FKBP12^F36V-tagged NMNAT1 was able to redistribute an mCherry-ecDHFR fusion without an NES from the cytoplasm to the nucleus. Next, the authors turned to developing approaches to co-opt two nuclear hormone receptors, ERα and GR, as shuttling proteins to redirect several POIs. The authors developed heterobifunctional molecules to recruit ERα by using AP1867 to engage FKBP12^F36V-tagged proteins conjugated to raloxifene that specifically binds to ERα. The authors achieved nuclear redistribution of FKBP12^F36V-tagged SMARCB1^Q318X, a mislocalized truncation mutant in atypical teratoid/rhabdoid tumors, as well as TDP43^ΔNLS and FUS^R495X, which are mislocalized cytoplasmic mutants in ALS. Like Gibson, Sadagopan et al., the authors observed that the strength of localization signals as well as the stoichiometry of the shuttling protein was important relative to the POI, where high expression of ERα was often required for effective nuclear relocalization.

While the authors had relied on overexpression approaches of FKBP12^F36V-tagged fusions, they sought to evaluate the relocalization of endogenously expressed targets. The authors focused on FOXO3a, a tumor suppressor that is sequestered in the cytoplasm, and FKBP12^WT, an abundantly expressed protein. High exogenous expression of ERα was required to relocalize FOXO3a with an endogenous knock-in of the FKBP12^F36V-tag. Notably, to evaluate FKBP12^WT relocalization, the authors developed bifunctional molecules consisting of SLF, a binder of FKBP12, conjugated to raloxifene for ERα recruitment, or dexamethasone for GR recruitment. While ERα was unable to relocalize FKBP12, high-expressing GR cells displayed relocalization of FKBP12 from the cytoplasm to the nucleus. These results further support the importance of the stoichiometry of the POI and shuttling proteins to induce protein relocalization. To demonstrate the potential pharmacological benefits of protein relocalization, the authors hypothesized that chemically induced relocalization of NMNAT1 from the nucleus down to the axon could serve as a protective mechanism akin to the Wlds mutation in response to axonal injury⁴². Importantly, redirection of FKBP12^F36V-tagged NMNAT1 by shuttling through ecDHFR-tagged GAP-43, which localizes in axons, prolonged axon health in response to an axotomy, showcasing a functional application of their approach.

The heterobifunctional CIPs developed by Gibson, Sadagopan et al. and Ng et al. employed known binders of the shuttling proteins. To expand the spectrum of recruitable shuttling proteins, Shao et al.²¹ aimed to identify covalent binding ligands of the 14-3-3 proteins, which can regulate their substrates by binding and sequestering them from their native cellular compartments⁴³. The authors performed a cysteine-reactive covalent ligand screen to identify EN171, a molecular glue stabilizer of 14-3-3σ with its substrate ERα. The authors tested whether they could sequester endogenous BRD4 from the nucleus to the cytosol by inducing proximity with 14-3-3 proteins. EN171 was used to develop a heterobifunctional molecule, QS57, through conjugation with JQ1 (Fig. 1E). The authors observed that QS57 treatment led to the relocalization of BRD4 from the nucleus to the cytoplasm and that the observed responses required 14-3-3σ (Fig. 1F). While this was not expanded beyond BRD4 and further optimization of EN171 is required to improve its selectivity profile, this study importantly demonstrated the feasibility of modulating the localization of endogenous proteins.

Together, these seminal studies by Gibson, Sadagopan et al., Ng et al., and Shao et al. unveil innovative strategies to redirect proteins within the cell to the growing CIP toolbox. These approaches represent a significant leap forward in the precise control of protein localization. While these studies present exciting proof-of-concept rationale for advancing therapeutic modalities predicated on protein relocalization, several key limitations and questions need to be addressed. First, can we overcome the incomplete relocalization and heterogeneity in responses that were observed in each study? Addressing this question will improve the applications of these tools for broader downstream biological study. The identification of optimal and ligandable shuttling proteins with sufficiently high expression levels, ligand binding affinity, and localization signal strength will be important. Second, how generalizable are these approaches to other organelles? It will be informative to identify shuttling proteins and cargoes that may prove advantageous in relocalization to divergent organelles such as the mitochondria or endoplasmic reticulum. Further work is also required to demonstrate that co-importing protein partners, which may be necessary for functional biological changes in some cases, is feasible. Finally, advancing these studies to in vivo models will aid in the evaluation of the consequences of redirecting disease-causing proteins away from their site of action or restoring functions of disease-preventing proteins in relevant models will be important. As we navigate these questions and continue the advancement of CIP modalities, it is increasingly evident that these strategies will play a pivotal role in reshaping our understanding of protein function within subcellular compartments. Ultimately, these insights will contribute to the development of innovative and targeted therapies that hold great promise for the future of medicine.