Targeted protein relocalization via protein transport coupling

Abstract

Subcellular protein localization regulates protein function and can be corrupted in cancers¹ and neurodegenerative diseases^2,3. The rewiring of localization to address disease-driving phenotypes would be an attractive targeted therapeutic approach. Molecules that harness the trafficking of a shuttle protein to control the subcellular localization of a target protein could enforce targeted protein relocalization and rewire the interactome. Here we identify a collection of shuttle proteins with potent ligands amenable to incorporation into targeted relocalization-activating molecules (TRAMs), and use these to relocalize endogenous proteins. Using a custom imaging analysis pipeline, we show that protein steady-state localization can be modulated through molecular coupling to shuttle proteins containing sufficiently strong localization sequences and expressed in the necessary abundance. We analyse the TRAM-induced relocalization of different proteins and then use nuclear hormone receptors as shuttles to redistribute disease-driving mutant proteins such as SMARCB1^Q318X, TDP43^ΔNLS and FUS^R495X. TRAM-mediated relocalization of FUS^R495X to the nucleus from the cytoplasm correlated with a reduction in the number of stress granules in a model of cellular stress. With methionyl aminopeptidase 2 and poly(ADP-ribose) polymerase 1 as endogenous cytoplasmic and nuclear shuttles, respectively, we demonstrate relocalization of endogenous PRMT9, SOS1 and FKBP12. Small-molecule-mediated redistribution of nicotinamide nucleotide adenylyltransferase 1 from nuclei to axons in primary neurons was able to slow axonal degeneration and pharmacologically mimic the genetic WldS gain-of-function phenotype in mice resistant to certain types of neurodegeneration⁴. The concept of targeted protein relocalization could therefore inspire approaches for treating disease through interactome rewiring.

Spatiotemporal control of subcellular protein localization creates a coordinated system to regulate cellular physiology⁵. Aberrant trafficking and localization of proteins underlies numerous diseases, including cancers^6,7, neurodegenerative diseases^3,8 and genetic disorders⁹. Mutations that reroute tumour suppressor proteins from the nucleus to the cytoplasm are common mechanisms for promoting oncogenesis. Translocation of RNA-binding proteins such as TDP43 and fused in sarcoma (FUS) to the cytoplasm is a hallmark of amyotrophic lateral sclerosis (ALS)² and has numerous deleterious phenotypic consequences, including increased cytoplasmic aggregates^10–12. Correcting these phenotypes through selective inhibition of a protein’s trafficking presents several challenges, given the conserved nature of nuclear export and import machinery. The global toxicity of nuclear import or export inhibition has restricted the application of molecules that target these pathways^13,14. Yet, the limited clinical success of global nuclear export inhibitors such as selinexor in certain contexts^15,16 suggests that targeted approaches to the control of protein trafficking could potentiate alternative therapeutic strategies.

The ability to exert targeted, pharmacological control of the subcellular location of an individual protein could provide avenues for addressing intractable diseases, as function-blocking molecules address only a small number of disease-relevant activities¹⁷. Gain-of-function molecules that directly drive the formation of neocomplexes between proteins lift several constraints that restrict the range of viable therapeutic targets^18,19. In the context of controlling subcellular localization, correcting diseased phenotypes that result directly from protein mislocalization or imparting beneficial function through protein relocation could expand the range of therapeutic options. A selection of tools for studying protein sequestration upon different stimuli with synthetic protein fusions to nuclear hormone receptor ligand-binding domains^20,21, LOV2 domains^22,23 or binding sites for localization sequence-containing nanobodies²⁴ have enabled fundamental studies of altered protein localization. Early work that used chemically induced proximity between FRB and FKBP12 fused to arrayed localization sequences driven by rapamycin demonstrated the potential of small-molecule control over localization programming^25–27. This approach was extended to the rapid sequestration of cytoplasmic and endosomal proteins onto mitochondrial surfaces to attenuate activity²⁸. Chemical control over localization has also been demonstrated with molecules that consist of protein-binding warheads linked to DNA intercalators²⁹ and membrane-targeting lipids³⁰.

Therapeutic modulation of target protein location by means of coupling to a cellular shuttle or anchor protein provides an alternative avenue for programmable molecular control. To advance the idea of localization control to address disease-driving proteins, an understanding of the cellular features that enable localization hijacking and expansion of the viable targets, molecules and mechanistic principles of transport control is needed. A quantitative analysis of localization reprogramming ability would greatly facilitate the extension of small-molecule-mediated protein relocalization to therapeutic contexts. Here we develop a quantitative approach to demonstrate targeted protein relocalization by means of targeted relocalization-activating molecules (TRAMs), which couple the trafficking of a target protein to the trafficking of a shuttle protein (Fig. 1a). We demonstrate the ability to use nuclear hormone receptors to mediate nuclear import of disease-relevant mutant proteins, and methionyl aminopeptidase 2 (METAP2) and poly(ADP-ribose) polymerase 1 (PARP1) as endogenous shuttles to redistribute endogenous proteins. Finally, we use targeted relocalization by means of a TRAM to generate a gain-of-function protective phenotype in a model of neurodegeneration. Through coupling shuttles and targets using drug-like molecules, we advance approaches for therapeutic modulation of cellular physiology on the basis of the concept of controlled protein trafficking.

Fig. 1 |. Development of a quantitative cell-by-cell analysis pipeline and application to targeted protein relocalization.

a, Subcellular nucleocytoplasmic target localization control by means of protein transport coupling by TRAMs. b, Model bifunctional molecule that engages EcDHFR and FKBP12^F36V domains. c, Viral constructs for generating stable cell lines to study localization control of NMNAT1. d, Co-immunoprecipitation of Flag-tagged NMNAT1 with mCherry shuttle protein in the presence of 1. e, Representative live-cell images for NMNAT1 relocalization driven by 1 in HeLa-derived EL-A after 3 h of treatment. f, Computational pipeline for cell-by-cell analysis of dual protein translocation. g, TRAM 1-coupled nuclear export dose–response curve derived using analysis pipeline after 3 h of treatment. h, Impact of relative shuttle/target protein ratio on nuclear export of NMNAT1. i, Representative nuclear import of a diffuse protein driven by NMNAT1 after 3 h of treatment with 1. j, Impact of relative stoichiometry on nuclear import of mCherry. Images in e and i are representative of three biological replicates. Data in g, h and j are mean ± s.d. of three independent experiments. Immunoblots in d are representative of three independent experiments. Scale bars, 20 μm. P values in g were determined by one-way analysis of variance (ANOVA) comparing each condition with DMSO. Uncropped blots can be found in Supplementary Fig. 7. Avg, average; cyto, cytoplasm; E, eluate; F, flowthrough; I, input; int, intensity; NES, HIV-1 REV NES (LPPLERLTL); NS, not significant; nuc, nuclear.

Demonstration of TRAM-induced relocalization

Nuclear localization sequences (NLSs) and nuclear export sequences (NESs) are recognized by members of the karyopherin family of proteins to regulate bidirectional transport of cargo between the nucleus and the cytoplasm³¹. We reasoned that overcoming the inherent localization of a protein by means of protein transport coupling would require (1) a sufficiently strong opposing localization sequence on the shuttling protein and (2) a requisite stoichiometry of shuttle protein. To examine these parameters, we chose nicotinamide nucleotide adenylyltransferase 1 (NMNAT1) as a model nuclear-localized target protein for relocalization. We hypothesized that as NMNAT1 possesses only a single predicted NLS and no known direct DNA-binding domain, it might be susceptible to small-molecule-mediated relocalization. To enable the study of proteins lacking known small-molecule binders, we synthesized the bifunctional molecule 1 (Fig. 1b), which engages the Escherichia coli dihydrofolate reductase domain (EcDHFR) and the FKBP12^F36V domain, both of which have found widespread use owing to their tight binding affinities and orthogonality to most mammalian proteins. We generated an mCherry fusion to EcDHFR and the NES from the HIV-1 REV protein³² to act as a shuttle protein. As transient transfection results in a range of confounding variables for examining the relative stoichiometries between two proteins and results in high levels of transgene expression³³, we stably incorporated mCherry-EcDHFR-NES and FKBP12^F36V-GFP-NMNAT1 into HeLa cells (Fig. 1c). Complex formation between both partners mediated by 1 was demonstrated by co-immunoprecipitation (Fig. 1d). We isolated clonal export line A (EL-A) with a relative mCherry/GFP median fluorescence intensity ratio of 1.0, which demonstrated dose-dependent translocation of NMNAT1 to the cytoplasm upon treatment with 1 (Fig. 1e), an effect that was abrogated at higher concentrations. Simultaneous treatment with the unlinked warheads 2 and 3 had no effect on protein localization (Extended Data Fig. 1a,b). Compound 1a, containing a longer linker between the two warheads, afforded a similar relocalization ability to 1 (Extended Data Fig. 1a,j,k). The quantification of nucleocytoplasmic transport when studying two proteins simultaneously has traditionally been done by a colocalization metric across a set of representative images. We reasoned that an open-source computational imaging analysis pipeline that provides single-cell-level analysis of two different proteins independently in both the nucleus and cytoplasm would be the most accurate approach for gauging protein transport and comparing different target proteins and shuttle proteins (Fig. 1f). We built this custom analysis pipeline by first segmenting cells using a nuclear stain reference, then masking cells to capture both a nuclear area and cytoplasmic area. We applied a gating filter on nuclear intensities to remove aberrantly high nuclear-stained objects such as debris and dead cells. Finally, we summed and averaged the intensities per cell for the nuclear masked area and cytoplasmic masked area to give a pixel intensity per area measurement akin to the relative concentration of fluorophore in the nucleus and cytoplasm for each cell. We verified that this approach would be applicable to different cell types and densities (Extended Data Fig. 2a–d).

Quantitative analysis was used to examine dose-, shuttle protein- and stoichiometry-dependent relocalization. We observed a concentration-dependent hook effect with 1 on the transport of NMNAT1 (Fig. 1g). Quantification of relocalization also revealed that although NMNAT1 was readily extracted to the cytoplasm with 1, NMNAT1 did not possess a sufficiently strong NLS to partially redistribute mCherry–EcDHFR–NES to the nucleus. To examine the impact of relative stoichiometry between NMNAT1 and EcDHFR–NES, we isolated two more clonal export cell lines (EL-B and EL-C) with varying mCherry/GFP ratios (Extended Data Fig. 1c). When treated with 1, cell lines with higher mCherry/GFP ratios exhibited a high degree of NMNAT1 redistribution to the cytoplasm (Fig. 1h). Cell lines with intermediate ratios resulted in diffuse NMNAT1 localization. EL-C, with the lowest relative expression levels of mCherry to GFP, exhibited 10% NMNAT1 relocalization upon treatment with 1 (Extended Data Fig. 1d). When cells were co-treated with leptomycin-B, an inhibitor of nuclear export by exportin 1, along with 1, no relocalization of NMNAT1 was observed and an enrichment in nuclear mCherry resulted (Extended Data Fig. 1d–g). Similar to EL-A, co-treatment with 2 and 3 did not cause protein relocalization in EL-B or EL-C (Extended Data Fig. 1d–g). No degradation was observed as a function of 1 in four clonal cell lines (Extended Data Fig. 1h,i). These results support active NES-driven redistribution of NMNAT1 through small-molecule-mediated coupling.

To examine whether diffuse proteins can be sequestered in the nucleus by non-DNA-binding proteins, we generated cell lines stably expressing mCherry–EcDHFR without an NES and FKBP12^F36V–GFP–NMNAT1. We isolated clonal import cell lines A–C (IL-A–C) with relative mCherry/GFP ratios spanning three orders of magnitude (Extended Data Fig. 3c) and analysed their ability to sequester mCherry upon treatment with 1. The initial percentage of nuclear mCherry was approximately 50% in all cell lines, consistent with diffuse localization. After treatment with 1, lines IL-A and IL-B with relative mCherry/GFP ratios of 0.001 and 0.01, respectively, were both capable of significant sequestration of mCherry (80% nuclear) (Fig. 1i,j and Extended Data Fig. 3a,d,e). Co-treatment with 2 and 3 had no effect on mCherry localization (Extended Data Fig. 3b). IL-C, with a relative mCherry/GFP ratio of 1.0, did not exhibit a large enrichment of nuclear mCherry upon treatment with 1 (Extended Data Fig. 3d,f). No diffusion of FKBP12^F36V–GFP–NMNAT1 was observed in cells due to coupling to mCherry–EcDHFR (Extended Data Fig. 3g). Treatment with 1 did not lead to degradation of either mCherry or NMNAT1 in all four clonal cell lines (Extended Data Fig. 3h,i). Together, these data demonstrate that sufficient levels of anchor or shuttle must be present to enforce TRAM-induced relocalization of a target protein.

In principle, the trafficking of any protein could relocalize another provided suitable localization strength, expression levels and effective molecular coupling. Nuclear hormone receptors such as the oestrogen receptor (ERα) and glucocorticoid receptor (GR) exhibit ligand-dependent translocation from the cytoplasm to the nucleus to exert downstream transcriptional activation. Several agonists or antagonists of nuclear hormone receptors could be repurposed to hijack the nuclear trafficking capabilities of these proteins (Extended Data Fig. 4d). These ligands have previously been incorporated into bifunctional molecules that act as degraders or transcriptional activators^34–37. To demonstrate the potential of these receptors as nuclear shuttles, we synthesized 4–7 (Extended Data Figs. 4a, 5a, 6a and 7a), which use dexamethasone- and raloxifene-based warheads known to engage but not downregulate GR and ERα. Using ectopically expressed model constructs, we validated that these molecules engage GR and ERα to nuclear enhance FKBP12 and FKBP12^F36V, and that ternary complex formation is necessary for this relocalization (Extended Data Figs. 4–7).

Shuttle–target pairs determine relocalizability

To effectively relocalize a target protein, a TRAM must sufficiently engage a shuttle protein capable of overcoming the target’s intrinsic localization. Quantification of the relative relocalization ability of proteins could enable rational choices of shuttle–target combinations. To develop a metric that details a protein’s relative relocalization ability, we generated polyclonal cell lines containing a set of genetic constructs pairing GR and ERα against well-characterized NESs (Fig. 2a). We treated our GR or ERα lines containing NES-tagged targets with varying concentrations of the relevant TRAM (4 or 5), then assessed the relocalization of the NES-tagged protein as well as the nuclear hormone receptor (Fig. 2b,c and Supplementary Figs. 1 and 2). We observed variable degrees of target relocalization that depended on both the shuttle identity (ERα or GR) and the NES sequence of the target. For GR, we observed ligand-activated relocalization. ERα remained nuclear regardless of the presence of TRAM. Despite different relative steady-state distributions of ERα and GR before TRAM treatment, the relative relocalizabilities of the NES-tagged target proteins remained consistent. ERα was able to partially relocalize the HIV-1 REV NES-containing protein, whereas the HIV-1 REV NES-containing protein caused partial nuclear export of GR (Fig. 2c). To analyse the relative strength of a small panel of NLSs, we paired them against HIV-1 REV NES (Fig. 2d) using 1 and observed a gradation in the relocalization ability of the different NLS constructs. Combined, these data highlight the importance of shuttle protein choice in effecting the desired degree of target protein relocalization.

Fig. 2 |. Nuclear hormone receptors are shuttles for cytoplasmic and diffuse proteins and provide a basis for quantitative assessment of target relocalization susceptibility.

a, Combinatorial pairing of GR or ERα with NESs, and HIV-1 REV NES with NLSs. b,c, Quantification of the mean percentage nuclear fluorophore in HeLa cells expressing ERα (b) or GR (c) and the NES sequences at different doses of 5 or 4, respectively, after 3 h. d, Quantification of the mean percentage nuclear fluorophore in HeLa cells expressing HIV-1 REV NES and the respective NLS sequences at different doses of 1 after 3 h. e, Sigmoidal best-fit curves derived from single-cell plots of the mean percentage nuclear fluorophore against relative GFP/mCherry ratio. Comparison of target NES-bearing protein relocalization into the nucleus by GR or ERα with 100 nM 4 or 5 after 3 h. DMSO treatment is plotted as a dotted line. f, Best-fit curves as in e comparing targets for each shuttle. g, Analysis of GR localization across MEK1 and HIV-1 REV NES. h, Comparison of NLS-containing target protein relocalization into the nucleus by HIV-1 REV NES at the optimal concentrations of 1 after 3 h (100 nM for SV40, NMNAT1; 10 nM for NPM1, P53). i, Relocalizability score of the NES-containing target proteins when partnered with GR or ERα. j, Localization score of the NES-containing target proteins when partnered with GR or ERα. k, Relocalizability score of the NLS-containing target proteins when partnered with HIV-1 REV NES. l, Localization score of the NLS-containing target proteins when partnered with HIV-1 REV NES. For f, g and h, DMSO treatment values are shown as a dashed line of the respective colour. Data in b, c and d are mean ± s.d. of three independent experiments. Scale bars, 20 μm. P values in b, c and d were determined by two-way ANOVA with Dunnett’s post hoc test comparing each condition with DMSO within each protein pairing. Max., maximum.

We previously observed a correlation of shuttle-to-target ratio with the degree of TRAM-induced relocalization. Using a doxycycline-inducible promoter system, we generated a heterogeneous population of cells containing each pair of proteins which enabled a comparison of the effects of shuttle-to-target ratios across sequences and receptors. We used our image analysis pipeline to estimate the ratios between fluorophores on a single-cell basis. This analysis was validated by comparison with ratios obtained from manual integration of fluorophore intensity in TRAM-induced nuclear import and export experiments. We then applied this analysis to the comparative NES and NLS relocalization datasets (Supplementary Fig. 3). For each pair, we derived best-fit dose–response curves from data examining protein localization as a function of the relative shuttle-to-target ratio on a per cell basis throughout the entire population (Supplementary Fig. 4). From this analysis we calculated the theoretical maximum relocalization of a target as well as the sensitivity of that target’s relocalization to the relative ratio of shuttle-to-target protein. To compare optimal dose–response curves, we selected the concentrations of TRAM that resulted in the highest theoretical maximum movement for each target. We then compared each of the four NES sequences paired with ERα or GR and observed that the effective ratio (ER₅₀) for relocalization was lower for each NES when paired with ERα and the theoretical maximum degree of relocalization was higher (Fig. 2e). Although we were not able to distinguish the relative relocalizability between HIV-1 REV NES and MEK1 NES using ERα as a shuttle, we saw clear differences when employing GR as a shuttle protein (Fig. 2f). Thus, we were able to distinguish relative relocalizability between targets by comparing them among different shuttles. Cells with lower ratios of GR to HIV-1 REV NES abolished TRAM-mediated nuclear GR enhancement, and resulted in GR nuclear export (Fig. 2g). Similarly, we compared the ability of different NLS sequences to be relocalized by the HIV-1 REV NES. We observed a clear trend in relocalization across different NLSs, demonstrating this analysis can be applied to both nuclear import and export (Fig. 2h). Together, these data enable quantitative comparisons of relocalization across target proteins and shuttle proteins as a function of protein ratios. We developed a metric in which numerical difference from zero reflects the degree of relocalizability. We normalized the observed change in target localization by the absolute difference between the steady-state localization and complete sequestration in that compartment (Fig. 2i). The relative ranking for relocalization scores among the four NES sequences examined was IκBα > P53 > MEK1 > HIV-1 REV. Likewise, the relative ranking for relocalizability of the NLS sequences was NMNAT1 > SV40 > P53 > NPM1. By combining relocalizability with a steady-state distribution descriptor, we generated a localization score that allows us to differentiate the localization strengths of proteins with similar steady-state distributions, such as HIV-1 REV and MEK1 (Fig. 2j). Across the NLS series, we observed that localization strength paralleled relocalizability (Fig. 2k,l). These metrics suggest intrinsic relocalization abilities for certain sequences and shuttles and imply a hierarchy that may be inferred from differential pairings of candidate shuttles and targets.

Mislocalized mutant protein relocalization

Mislocalized transcriptional regulators and RNA-binding proteins are hallmarks of several cancers and neurodegenerative diseases. Targeted approaches that return these proteins to the nucleus might help to ameliorate disease-driving phenotypes. We examined the ability to relocalize three mutant proteins known to exhibit aberrant localization in disease. SMARCB1^Q318X is a mutant C-terminal truncation of SMARCB1 found in atypical teratoid and rhabdoid tumours that has been reported to exhibit cytoplasmic localization³⁸. Mislocalized TDP43 mutants lacking NLS sequences (TDP43^ΔNLS) that seed aggregates in the cytoplasm are a conserved biomarker for ALS³⁹. Mutations in the FUS protein that alter its nuclear localization have been observed in patients with ALS⁴⁰. The FUS^R495X mutation is associated with an aggressive disease phenotype⁴¹, and forms cytoplasmic granules in neurons that have been linked to its pathogenic function⁴².

We engineered HeLa cells to constitutively express each mutant along with ERα on a doxycycline-inducible promoter (Fig. 3a). Upon addition of doxycycline and subsequent treatment with 5, we observed that the requisite relative ratio of ERα to each mutant required for maximum relocalization was achieved in different fractions of each population (Fig. 3b,c and Extended Data Fig. 8a–d). Mean population analysis showed significant relocalization across all concentrations of TRAM for each mutant (Extended Data Fig. 8e) and no ligand-induced protein stabilization (Extended Data Fig. 8f). Best-fit curves allowed us to extract the theoretical maximum relocalization of each mutant (Fig. 3d and Supplementary Fig. 5), from which we derived the relative relocalizability and localization strength of each mutant when paired with ERα (Fig. 3e). We observed that TDP43^ΔNLS and FUS^R495X were more readily relocalized by ERα than SMARCB1^Q318X, suggesting protein-intrinsic features that facilitate or hinder relocalization.

Fig. 3 |. Targeted relocalization can move mutant proteins into nuclei and out of stress granules.

a, Genetic constructs for stable incorporation. b, Representative images of HeLa cells stably expressing the target proteins SMARCB1^Q318X, TDP43^ΔNLS (TDP43(^{K82A/R83A/K84A})) or FUS^R495X and shuttle protein ERα under an inducible promoter after 3 h of treatment with 5. c, Representative violin plots of population shifts in percentage nuclear target protein upon treatment with 5 for 3 h. d, Best-fit curves derived from single-cell plots of percentage nuclear mCherry against GFP/mCherry. e, Relocalizability and localization strengths calculated for SMARCB1^Q318X, TDP43^ΔNLS or FUS^R495X using ERα as a shuttle protein. f, Schematic for ERα-induced relocalization of FUS^R495X from stress granules. g, Experimental workflow for cells pre-treated with NaAsO₂ for an hour to induce stress granule formation before relocalization through TRAM treatment. h, Timelapse live-cell imaging snapshots of FUS^R495X extraction from granules after treatment with 5. i, Representative fixed-cell immunofluorescent images of stress granule marker G3BP1 and mCherry–FUS^R495X after treatment with 5 or control compounds. j, Quantification of stress granules in cells after treatment with 5 or control compounds. k, Comparison of mCherry–FUS^R495X-positive puncta in cells treated with bifunctional molecule 5 or control molecules. l, Comparison of G3BP1-positive puncta in cells treated with bifunctional molecule 5 or control molecules. Images in b and h are representative of three biological replicates. Images in i are representative of three biological replicates. Three fields of view totalling a minimum of 40 cells per condition per repeat were used to generate mean values for j, k and l. Data in j, k and l are mean ± s.d. of three independent experiments. Violin plots in c are representative of three independent experiments. Scale bars, 20 μm. P values in j were determined by two-way ANOVA with Dunnett’s post hoc test comparing each dataset with its relative DMSO control. P values in k and l were determined by unpaired two-tailed t-tests. SG, stress granule; * denotes mutation as defined in a.

We hypothesized that FUS^R495X-positive cytoplasmic granules might be dissolved if TRAMs could induce redistribution of FUS^R495X to the nucleus. To assess protein extraction from granules, we examined FUS^R495X-expressing HeLa cells treated with sodium arsenite to instigate stress granule formation (Fig. 3f). After formation of granules, cells were treated with 5 and continuously imaged (Fig. 3g). We observed relocalization of FUS^R495X from both granules and the cytoplasm to the nucleus over the course of 2 h after exposure to 5 (Fig. 3h and Extended Data Fig. 9a,b). These observations are consistent with small-molecule-mediated extraction of FUS^R495X from cytoplasmic granules to the nucleus. To quantify the effect of FUS^R495X redistribution on the number of stress granules in cells, we performed fixed-cell immunofluorescence imaging to examine both mCherry–FKBP12^F36V–FUS^R495X and G3BP1 (Fig. 3i and Extended Data Fig. 9c). We observed a significant reduction in both FUS^R495X-positive and G3BP1-positive stress granules in cells treated with 5 (Fig. 3j). Treatment with 5 led to more substantial reductions in stress granules than simultaneous treatment with both unlinked warheads 3 and 10 (Fig. 3k,l and Extended Data Fig. 9d). This demonstrates that FUS^R495X can be relocalized from stress granules to the nucleus and that TRAM treatment can reduce FUS^R495X-positive and G3BP1-positive granules.

Endogenous protein relocalization

The ability to exploit the cereblon-damaged DNA-binding protein 1 (CRBN-DDB1) and Von Hippel-Lindau (VHL) E3 ligases as general effector proteins with well-characterized small-molecule warheads has contributed to the rapid advancement of targeted protein degradation. We sought similarly general shuttle proteins that are expressed at high levels in most cell types. We identified METAP2 as a suitable nuclear export shuttle, and PARP1 as a nuclear import shuttle with well-characterized warheads. To demonstrate that these proteins could be used as suitable shuttles, we synthesized 11–13 (Fig. 4a), which use fumagillol- and olaparib-based warheads to engage METAP2 and PARP1, respectively. Using ectopically expressed model constructs, we validated that TRAMs 11 and 12 are able to engage METAP2 and PARP1 to nuclear exclude or enhance FKBP12^F36V, respectively, and that ternary complex formation is necessary for this relocalization (Fig. 4b,c and Extended Data Fig. 10). We observed dose-dependent relocalization that reached a minimum of 12% nuclear localization by using 11 to engage METAP2 (Fig. 4d) and a maximum of 84% nuclear localization by using 12 to engage PARP1 (Fig. 4e).

Fig. 4 |. Endogenous targets are susceptible to relocalization with TRAMs that engage endogenous shuttles.

a, Bifunctional molecules that engage METAP2 or PARP1 and FKBP12^F36V or FKBP12. b, Representative live-cell images of FKBP12^F36V nuclear export when treated with 11 or unlinked warhead controls for 3 h. c, Representative live-cell images of FKBP12^F36V nuclear import when treated with 12 or unlinked warhead controls for 3 h. d, Dose–response curve for percentage of nuclear FKBP12^F36V after a 3 h of treatment with 11. e, Dose–response curve for percentage of nuclear FKBP12^F36V after a 3 h treatment with 12. f, Representative live-cell images of PRMT9 knock-in HEK293T cells treated for 24 h with 11. g, Dose–response curve for percentage of nuclear PRMT9 after a 24 h treatment with 11. h, Representative live-cell images of SOS1 knock-in HEK293T cells treated for 24 h with 12. i, Dose–response curve for percentage of nuclear SOS1 after a 24 h treatment with 12. j, Representative live-cell images of FKBP12 knock-in HEK293T cells treated for 24 h with 13. k, Dose–response curve for percentage of nuclear FKBP12 after a 24 h treatment with 13. Images in b, c, f, h and j are representative of three biological replicates. Data in d, e, g, i and k are mean ± s.d. of three independent experiments. P values in d, e, g, i and k were determined by one-way ANOVA with Dunnett’s post hoc test comparing each condition with the DMSO control. Scale bars, 20 μm.

The examination of targeted relocalization of endogenous proteins as a therapeutic approach is limited by a lack of small-molecule ligands for any desired target protein. To bypass this challenge, we used a modified Cas9-based protocol for microhomology-mediated end joining to install binding domains and fluorescent proteins on endogenous protein targets (Extended Data Fig. 11). We designed a guide RNA plasmid with two orthogonal U6 promoters and stem loops to facilitate rapid cloning for knock-in onto different targets. To illustrate endogenous protein relocalization using endogenously expressed shuttle proteins, we sought candidate targets that had undergone successful tagging in the OpenCell database⁴³. PRMT9 is a nuclear-enhanced protein that has recently been explored as a target in acute myeloid leukaemia⁴⁴. Using CRISPR–Cas9 tagging, we inserted a GFP–FKBP12^F36V cassette onto PRMT9, and demonstrated that we could nuclear exclude PRMT9 upon treatment with TRAM 11, which engages endogenous METAP2 as an export shuttle. Dose-dependent relocalization was observed for the export of endogenous PRMT9 by endogenous METAP2, with maximum export observed at 10 nM of 11 (Fig. 4f,g and Extended Data Fig. 12a,b). Nuclear export with TRAM 11 was not observed when cells were treated with unlinked warheads 3 and 14 (Extended Data Fig. 12c–e). Pre-treatment with 14 or competitive treatment with 3 mitigated the observed nuclear export (Extended Data Fig. 12f,g), illustrating the dependence of relocalization on effective ternary complex formation. To study endogenous protein nuclear import, we appended the GFP-FKBP12^F36V cassette onto endogenous SOS1, which is known to have strong cytoplasmic localization and membrane association. Treatment with TRAM 12 engaging PARP1 resulted in partial redistribution of SOS1 from the cytoplasm to the nucleus (Fig. 4h,i and Extended Data Fig. 13a,b), which was not observed when treated with the unlinked warheads and was abolished when competed with unlinked warheads 15 or 3 (Extended Data Fig. 13c–g).

Motivated by the finding that endogenous levels of shuttle proteins were sufficient to relocalize endogenous protein targets, we synthesized TRAM 13 from warheads directly engaging endogenous FKBP12 and PARP1 and inserted GFP onto endogenous FKBP12 for live-cell visualization. Upon treatment with TRAM 13, significant nuclear enhancement of FKBP12 was observed (Fig. 4j,k and Extended Data Fig. 14a,b), which was abolished upon co-treatment or competition with the unlinked warheads (Extended Data Fig. 14c–g). Here we showcase relocalization of a protein by a rationally designed bifunctional molecule that directly engages both endogenous shuttle and target proteins. Across all three endogenous protein relocalization examples, a homogenous population shift was observed upon TRAM treatment, unlike the ectopic expression models for which relocalization is often observed in a subset of the cell population. This suggests that the minimum relative shuttle protein-to-target protein ratio requirement can be achieved throughout the population in endogenous systems.

NMNAT1 relocalization slows axonal degeneration

Targeted protein relocalization by means of coupling to a shuttle protein could enable pharmacological approaches that mimic beneficial gain-of-function mutations. WldS, a mutant protein consisting of mouse NMNAT1 (mNMNAT1) fused to a truncated N-terminal region of UBE4B, has been shown to be protective against Wallerian neurodegeneration^45–47. Mice bearing the WldS mutation have shown increased resistance to neuropathies and ALS^48,49. The ability for small amounts of WldS to traffic to the axon is crucial for its protective function, where it serves to maintain the levels of NAD⁺ upon loss of axonal nicotinamide nucleotide adenylyltransferase 2 (NMNAT2) during early stages of axonal injury⁵⁰. We reasoned that small-molecule-mediated transport of NMNAT1 from the nucleus down the axon might serve a similar function to WldS and provide a proof-of-concept example for protein relocalization to drive a gain-of-function phenotype. We collected dorsal root ganglion (DRG) neurons from rat embryos and transduced them with adeno-associated viruses (AAVs) expressing mNMNAT1 linked to FKBP12^F36V and a truncated sequence of axon-localized GAP-43, fused to EcDHFR (Fig. 5a,b). When we treated these neurons with 1, clear redistribution of mNMNAT1 down axons was observed, which did not occur when neurons were co-treated with unlinked warheads 2 and 3 (Fig. 5c and Extended Data Fig. 15a). To assess the protective function of mNMNAT1 redistribution, after treatment with 1, we performed an axotomy and removed the cell bodies (Extended Data Fig. 15b). We observed persistent mNMNAT1 expression in axons post-axotomy only in neurons treated with TRAM 1. Distinct blebbing and fragmentation of neurites was observed in dimethylsulfoxide (DMSO) or unlinked warhead co-treated neurons 24 h after axotomy, with near complete neurite death observed after 48 h (Fig. 5d, Extended Data Fig. 15c and Supplementary Fig. 6). However, in neurons treated with 1, both axon health and axon termini health were maintained for at least 72 h after axotomy and attenuated axon death was observed only after 96 h (Fig. 5d–f and Extended Data Fig. 15d–g). These observations are consistent with a gain-of-function protective activity of NMNAT1 driven by targeted, TRAM-mediated relocalization from nuclei to axons.

Fig. 5 |. Targeted relocalization of NMNAT1 can protect against axon injury.

a, Schematic for DRG neuron collection, transduction of model proteins through AAV, small-molecule treatment and axotomy. b, Constructs delivered by AAV to study the gain-of-function potential for mNMNAT1 relocalization in neurons, and definition of labels for neuronal imaging. GAP43at is the 20-amino-acid dipalmitoylation domain (MLCCMRRTKQVEKNDEDQKI) found at the N terminus of growth-associated protein-43. c, Representative live-cell images of mNMNAT1 in primary DRG neurons when treated with 1 or control molecules. d, Representative live-cell images of axonal-tagged protein in axons and axon termini after exposure to 1 for 24 h followed by axotomy to induce degeneration. mRuby3 signal was used to monitor degeneration across treated neurons. Axontag, GAP43at–mRuby3–EcDHFR. e, Quantification of axonal degeneration over time in the axon. f, Quantification of degeneration at 72 h of axon termini. Images in c and d are representative of neurons collected from four embryos. For the plots in e, three replicates each from four embryos were combined for DMSO and 1 nM 1; for the 1 nM 2 + 1 nM 3 condition, three replicates each from two embryos were combined. For plots in f, a total of six replicates derived from four different embryos were combined for the DMSO and 1 nM 1 treatment data, and a total of four replicates derived from two different embryos were combined for the 1 nM 2 + 1 nM 3 condition. Data in e and f are mean ± s.e.m. P values in e and f were determined by two-way ANOVA with Tukey’s post hoc test comparing the 1 nM treatment of 1 with the DMSO (purple) or the unlinked warhead control (green) at each time point. Scale bars, 20 μm.

Discussion

Subcellular protein localization often serves as a control mechanism for protein activity, and alterations in a protein’s localization can form the basis of numerous diseases. Controlling protein location by hijacking the transport mechanisms of another protein could serve as a therapeutic approach for precision manipulation of protein function. In the present study, we have developed a quantitative single-cell analysis approach to examine the potential for TRAM-mediated coupling of a target protein with a shuttle protein. We have shown that the relative expression levels of the two protein partners is a predominant parameter in driving the ability to relocalize a target protein. We have also demonstrated that the localization of an endogenous target protein can be overturned by a coupling to an endogenous shuttle protein. Pioneering work using chemical methods to influence nucleocytoplasmic transport has relied on pitting two synthetic localization signals against each other²⁵. Extending these efforts to native systems, as we have done here, represents the next frontier in controlling mammalian protein localization pharmacologically. Quantification of relative localization strength between proteins remains an open challenge⁵¹, with context dependencies that are probably intrinsic to each individual protein. Quantitative analysis at the single-cell level is essential for assessing the ability for small molecules to coopt other mammalian machinery. Here we develop and deploy a relocalizability metric as well as a relative localization strength metric which might serve as a basis for further comparisons. These metrics could inform the rational choice of protein partners for programmable degrees of target protein relocalization.

Tethering two proteins together to alter the trafficking of one partner necessitates suitable shuttle proteins with ligands. Nuclear hormone receptors are common targets for small-molecule therapeutics and exhibit nuclear translocation or localization. We have developed molecules that employ binders of nuclear hormone receptors and can harness them to drive enrichment of a protein target in the nucleus. Using discrete shuttles that are differentially expressed in disease contexts such as cancer or are uniquely expressed in certain cell types enables the future prospect of targeted relocalization-based therapies that can generate beneficial phenotypic consequences. The mislocalization of transcriptional regulators or RNA-binding proteins has been observed in cancers and neurodegenerative diseases. Mutations that truncate or impair NLSs are particularly pathogenic, with no clear mechanism for therapeutic intervention beyond genetic manipulation. We demonstrate the ability of small-molecule-mediated targeted relocalization using nuclear hormone receptors to return these proteins to the nucleus from the cytoplasm or stress granules. Our data suggest that the redistribution of mislocalized proteins that form the core of stress granules back to nuclei might dissipate these granules.

Targeted redistribution of endogenous proteins driven by small molecules has been challenging to rigorously demonstrate in live cells, necessitating methods for analysing protein localization quantitatively with high sensitivity. Small-molecule ligands to engage shuttles of the requisite expression level and localization strengths must be paired with ligands for target proteins. Genetic tagging of endogenous proteins with ligand-binding domains has been transformative to study targets for protein degradation with universal small molecules. We show, by means of endogenous knock-in of binding domains, the ability to redistribute endogenous proteins using METAP2 or PARP1 as general nuclear export or import shuttles, respectively. We further demonstrate the utility of sensitive imaging analyses to report on the redistribution of endogenous, highly expressed FKBP12 using endogenous PARP1 with a TRAM directly engaging both proteins. The degrees of relocalization of SOS1 and FKBP12 were smaller (less than 10%) than that observed for PRMT9. These experiments demonstrate that strongly localized or highly abundant proteins can undergo TRAM-mediated relocalization, albeit to a small degree. The critical amount of relocalization needed to confer a biological phenotype is likely to be target dependent. Our work establishes that targeted relocalization of endogenous proteins is possible, and further work is needed to define the extents necessary for gain-of-function phenotypes in endogenous systems. The endogenous proteins examined exhibit population-wide redistribution as a function of TRAM treatment. The structure of linkers within bivalent molecules can often be a key parameter for efficacy in PTM-targeting induced proximity approaches. Our efforts with PARP1 and METAP2 were supported by previously published studies describing effective linking strategies^52,53. Further exploration of linking strategies of target and shuttle binding warheads for any protein pair might reveal cooperative binding events for highly effective TRAMs.

An attractive aspect of targeted protein relocalization is the ability to impart protein function in a new subcellular compartment. The WldS mutation has been extensively studied for its ability to confer protection against neurodegenerative conditions. A consensus protective function of WldS is maintaining the balance of NAD⁺ in axons during injury. Small-molecule approaches to target this axis have attempted to inhibit the widely expressed NADase enzyme SARM1 (refs. 54,55). Targeted relocalization of NMNAT1 might serve to compensate for the loss of NMNAT2 during early neurodegeneration and could present an alternative approach amenable to specific targeting using neuron-specific transport pathways. We demonstrate protection from axonal degeneration by partial relocalization of NMNAT1 from nuclei down axons induced by TRAMs.

In this demonstration, TRAMs bind concurrently to a target protein and shuttle protein and confer relocalization in a stoichiometric mechanism of action. This limitation is not observed with bifunctional degraders and other PTM modifier-recruiting molecules, which benefit from event-driven or sub-stoichiometric mechanisms. The stoichiometric mode of action of TRAMs also requires higher expression of a shuttle protein compared with a target protein, which might limit the scope of potential targets for relocalization. Recent reports combine occupancy-driven mechanisms that operate by means of stoichiometric target engagement with downstream amplified phenotypic consequences³⁴. In principle, movement of a target protein to a new subcellular location could enable numerous mechanisms for rewiring cellular activity. Neosubstrates of enzymes might be accessible entirely through TRAM-induced relocalization. Relocalization of mislocalized transcription factors or RNA-binding proteins might similarly restore substrate accessibility. To realize these ideas, further work on the extent of relocalization required to observe different phenotypes will be essential. As most warheads used for bifunctional molecule development are inhibitors, available ligands might not be suitable to fully realize the potential of TRAMs due to inhibition of effector functions. Whether a target protein can perform functions while tethered to a shuttle remains an important question for each new system. Medicinal chemistry campaigns can bridge these gaps by developing non-inhibitory binders of target proteins to facilitate gain-of-function phenotypes upon relocalization. To do so will require identification of appropriate targets and shuttles for a desired phenotype. Choices of targets and shuttle pairings can be informed through relative relocalizability metrics defined for a larger set of proteins. Insights into features governing relocalizability will further enhance therapeutic efforts. Combined with non-inhibitory molecule development, interactome rewiring by means of protein localization manipulation has the potential to contribute to advances in both fundamental biology and development of next-generation therapeutics.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-024-07950-8.

Methods

Cell lines

Adherent cells were cultured in 10 or 15 cm plates at 37 °C in a 5% CO₂ humidified atmosphere. HeLa and HEK293T cells were obtained from ATCC and were cultured in DMEM supplemented with 10% heat-inactivated fetal bovine serum and 1% penicillin/streptomycin.

Stable cell line generation with Sleeping Beauty transposase

Fusion protein constructs were cloned into a pSBtet vector flanked by Sleeping Beauty transposase recognition sites which also encoded constitutive expression of the puromycin resistance gene. The day before transfection, 300,000 cells were seeded into a six-well plate. The next day, 1.9 μg of transgene DNA was mixed with 0.1 μg of Sleeping Beauty transposase encoding plasmid in 200 μl of JetPrime buffer, then 4 μl of JetPrime reagent was added. The mixture was gently mixed and left to incubate for 10 min at room temperature, after which it was added dropwise to the well. The next morning, medium was replaced with fresh cell culture medium. Then, 3 d after transfection, antibiotic selection was initiated with 1 μg ml⁻¹ puromycin until all control cells had died.

Retrovirus generation

At 1 d before transfection, 1.5 million HEK293T cells were seeded into 6 cm dishes. The following day, 130 μl of complete DMEM was added into an Eppendorf tube. Next, 1 μg of transgene cloned into a pMXs vector was added along with 900 ng of the retrovirus pol/gag and 150 ng of VSVg DNA, and gently mixed. PEI (1 mg ml⁻¹) was brought to room temperature and gently mixed before adding 6 μl to the DNA mixture. The mixture was gently mixed and incubated for 20 min at room temperature. Medium in the 6 cm dishes was replaced with fresh complete DMEM. After 20 min of incubation, the transfection mixture was added dropwise to the 6 cm plate, then the plate was returned to the incubator. The next morning, medium was removed and replaced with 5 ml of DMEM containing 30% heat-inactivated fetal serum and 1% penicillin/streptomycin and the plate was returned to the incubator. After 36 h, the medium from the 6 cm plate was collected into a 15 ml Falcon tube and centrifuged at 1,000 rpm for 5 min. The supernatant was filtered through a 0.45 μM filter into 300 μl aliquots. Aliquots were stored at −80 °C.

Generating stable cell lines with retrovirus

On the day of infection, cells were seeded at a density of 2 million cells per well in a six-well plate, and medium was added to the well up to 3 ml total volume. Then, 3 μl of polybrene transfection reagent (EMD Millipore, catalogue no. TR-1003-G) was added to each well. Viral aliquots were thawed at 37 °C, then gently mixed and added to the well. The plate was then centrifuged at 2,200 rpm for 45 min at 37 °C, then returned to the incubator. The next morning, the cells were lifted and transferred into a 10 cm plate. When cells reached 70–80% confluency, antibiotic selection was started with 10 μg ml⁻¹ blasticidin until control cells died.

Lentivirus generation

In total, 1.2 million viral HEK293T cells were seeded per well in a six-well plate. The next day, 6 μl of Mirus-LT1 transfection reagent was mixed with 250 μl of Opti-MEM and left to incubate at room temperature for 15 min. Then, 1.5 μg of transgene cloned into the pU6 transfer vector was mixed with 0.1 μg of gag/pol, 0.1 μg of REV, 0.1 μg of TAT and 0.2 μg of VSVg packaging plasmids. The Opti-MEM and Mirus-LT1 mixture was then added to the DNA mixture, mixed and left to incubate at room temperature for 15 min, then added dropwise to wells. The plate was gently swirled to mix, then returned to the incubator. The next morning, medium in the well was replaced. After 36 h, the supernatant was retrieved into 15 ml Falcon tubes. The tubes were spun down at 300g for 5 min, then the supernatant was filtered through a 0.45 μm filter into 300 μl aliquots. Aliquots were stored at −80 °C.

Generating stable cell lines with lentivirus

In total, 200,000 cells were seeded per well in a 24-well plate. Polybrene was added (EMD Millipore, catalogue no. TR-1003-G) to a final concentration of 8 μg ml⁻¹ in a total volume of 1 ml in a well of a 24-well plate, then viral aliquots thawed at 37 °C were added dropwise into the well, and the plate was then returned to the incubator. The next day, the medium was replaced. At 48 h after infection, antibiotic selection was started with 1 μg ml⁻¹ puromycin until the control well died.

Live-cell confocal microscopy

Adherent cells were plated (35,000 cells per well in an eight-well chamber slide) 1 d before small-molecule treatment. Doxycycline-inducible lines were plated 2 d before small-molecule treatment at half the seeding density. At 1 d before treatment, cells were treated with doxycycline to a final working concentration of 1 μg ml⁻¹ in 250 μl of complete growth medium. On the day of treatment, wells were gently rinsed with 2 × 300 μl of PBS, then incubated with 250 μl of complete phenol red-free growth medium with small molecules added for the indicated time. After the indicated time, 1 μl of a fresh Hoechst 33342 solution (0.5 μg μl⁻¹ in PBS) was added directly to the existing medium in each well to achieve a final Hoechst 33342 concentration of 2 μg ml⁻¹, and incubated for 10 min. Cells were imaged with a Nikon A1R confocal microscope using a Plan Fluor ×60 oil immersion, 1.30-numerical aperture (NA) objective. The microscope was equipped with a 405 nm violet laser, a 488 nm blue laser, a 561 nm green laser and a 639 nm laser. Images were exported and analysed using the FIJI software package (v.1.53).

Fixed-cell confocal microscopy

Adherent cells were plated (50,000 cells per well in an eight-well chamber slide) 1 d before the experiment. Cells were then washed three times with DPBS and fixed with 4% paraformaldehyde in PBS for 15 min at room temperature, washed three times and permeabilized with 0.1% Triton for 5 min at room temperature. Cells were blocked in 10% goat serum in PBS for 1 h at room temperature and incubated with primary antibody overnight at 4 °C. Cells were washed with PBS three times, then incubated with secondary antibody and DAPI for 1 h at room temperature. Cells were washed with PBS and imaged with a Nikon A1R confocal microscope using a Plan Fluor ×60 oil immersion, 1.30-NA objective. The microscope was equipped with a 405 nm violet laser, a 488 nm blue laser, a 561 nm green laser and a 639 nm laser. Images were exported and analysed using the FIJI software package (v.1.53).

Custom imaging analysis pipeline construction

1. Segmentation: a binary image is generated from a nuclear-stained image (for example, DAPI) of a field of view using Otsu thresholding. This binary image is then used as a mask to label coordinates within an acquired field of view corresponding to nuclei. Watershed transform is carried out to segment touching nuclei, and individual nuclei within a field of view are assigned with different numerical labels starting from 1 (‘nlabels’).

2. Generating masks: ‘Celllabels’: ‘nlabels’ are then expanded by 6 pixels circumferentially to capture cell area. At the same time, ‘nlabels’ is also expanded by 3 pixels to generate ‘expandednlabels’. Cytoplasmic labels (‘clabels’) are obtained by subtracting the ‘expandednlabels’ from ‘celllabels’. ‘Nlabels’ is also shrunk by 3 pixels to avoid taking in nuclear boundary measurements for the cytoplasm or nuclear readings. These pixel numbers can also be user defined.

3. Gating: Average nuclear stain intensities per area (‘avg nuc int’) are then tabulated for these individual nuclei, and a histogram is generated. Kernel density estimation is used to generate a probability density function from this histogram. A normal distribution centred around the main peak of the probability density function on the basis of the full-width at half-maximum is then generated and overlayed. Assuming the population ‘avg nuc int’ follows a normal distribution, the ends of the normal distributions are set as the gate bounds. Any nuclei without ‘avg nuc int’ within the set gate bounds are gated out and labels for those nuclei and their respective cytoplasmic label are dropped to give ‘gatednlabels’ and ‘gatedclabels’. Alternatively, if the user wishes to set fixed values for the gate bounds, this can also be passed in as an argument in the script.

4. Summing intensities per cell: Fluorophore images for this field of view are overlayed with the ‘gatednlabels’ and ‘gatedclabels’ to extract fluorophore intensities tabulated for every single nuclear and cytoplasmic label. Average fluorophore intensities for each nuclei/cytoplasm label are also generated by dividing the intensity over the area of the label, to give a nuclear and cytoplasmic fluorophore concentration (pixel intensity per area) per cell.

   $$\text{Avg}\text{GFP}\text{nuc}\text{intensity}\text{per}\text{label}=\frac{\text{summed}\text{GFP}\text{intensity}\text{of}\text{the}\text{label}}{\text{area}\text{of}\text{label}}$$
5. Calculations and output: For each cell, the average fluorophore pixel intensity per area is then used in the following calculation to estimate percentage nuclear fluorophore:

   $$\text{Percentage}\text{nuc}\text{GFP}\text{per}\text{label}=\frac{\text{avg}\text{GFP}\text{nuc}\text{intensity}\text{per}\text{label}}{\text{avg}\text{GFP}\text{nuc}\text{intensity}\text{per}\text{label}\text{+}\text{avg}\text{GFP}\text{cyto}\text{intensity}\text{per}\text{label}}$$

These values, including the average fluorophore intensity per label, are tabulated in a dataframe and output to an Excel file with a separate sheet for each condition. A summary of the mean for each value for each condition is also tabulated and output as an Excel sheet. In addition, a folder is also created containing the gating histogram, as well as overlay masks onto the fluorophore images and the segmentation and nuclear masks from the original nuclear stain images for each field of view.

Modified imaging analysis pipeline for relative protein ratio estimation and validation

The pipeline built as described above was improved by plotting each individual cell label onto the overlay of the masks on the fluorophore images. This facilitates the user to be able to visually inspect the image and correlate the visual analysis of that cell with the pipeline-derived metrics for that cell. Ten cells for a field of view of a treatment condition were manually picked across a range of protein expression levels by visual analysis. For the ten cells in each image, the drawing tool in ImageJ was used to manually demarcate the cell boundary using the cytoplasmic protein fluorophore image. For each demarcated cell area, the raw integrated density of each fluorophore channel was obtained using the ImageJ measure function. For each cell, the relative ratio of the two fluorophores was obtained by dividing the raw integrated density of GFP by the raw integrated density of RFP to give GFP/mCherry for the whole cell.

$$\text{GFP/mCherry}\text{whole}\text{cell}=\frac{\text{cell}\text{raw}\text{integrated}\text{density}\text{(GFP)}}{\text{cell}\text{raw}\text{integrated}\text{density}\text{(RFP)}}$$

For each cell analysed by means of manual integration as detailed above, their average fluorophore nuclear intensity and cytoplasmic intensity obtained from the image analysis pipeline were summed to give an estimate of whole cell fluorophore intensity, and a ratio of this value between the two fluorophores was obtained to give the relative protein ratio estimate of GFP/mCherry masked.

$$\text{GFP/mCherry}\text{masked}=\frac{\text{avg}\text{nuc}\text{GFP}\text{intensity}+\text{avg}\text{cyto}\text{GFP}\text{intensity}}{\text{avg}\text{nuc}\text{RFP}\text{intensity}+\text{avg}\text{cyto}\text{RFP}\text{intensity}}$$

The estimate of the relative protein ratio calculated from pipeline metrics (GFP/mCherry masked) was plotted against the relative protein ratio obtained from manual visual integration using ImageJ (GFP/mCherry whole cell), for all ten cells in each condition, for three repeats. Simple linear regression was used to plot the trendline between the two values, and the Pearson correlation coefficient R value was calculated for each graph.

Single-cell plots and best-fit curves to extract theoretical maximal relocalizability

Percentage nuclear fluorophore metrics for each single cell were plotted against the GFP/mCherry masked values calculated from the image analysis pipeline. Data from three repeats were combined into a single plot for each condition and outliers that had GFP/mCherry masked values beyond 3 s.d. away from the mean were gated out. For nuclear import cases, agonist sigmoidal best-fit curves were plotted using the equation of $y=\text{bottom}+\left(x^{\text{HillSlope}}\right)\times (\text{top}-\text{bottom})/\left(x^{\text{HillSlope}}+\right.\text{EC}5\left.0^{\text{HillSlope}}\right)$, where $\text{EC}50$ is the half-maximum effective concentration. For nuclear export cases, antagonist sigmoidal best-fit curves were plotted using the equation of $y=\text{bottom}+(\text{top}-\text{bottom})/(1+(\text{IC}50/x{)}^{\text{HillSlope}})$. The $y$ values at which these best-fit curves plateaued were identified as the theoretical maximum relocalization for that condition.

Relative relocalizability and localization metric calculations

Relocalizabilty score is taken as the maximum relocalization demonstrated (extracted from the plateau of the best-fit curve) minus the DMSO-treated percentage nuclear fluorophore, over the maximal theoretical possible change in their respective direction of movement. Relocalizability score is on a scale of −1 to 1, where −1 is the most relocalizable nuclear protein moving from the nucleus to cytoplasm, and 1 is the most relocalizable cytoplasmic protein moving from the cytoplasm to the nucleus.

$$\text{Relocalizability}\text{of}\text{a}\text{cytoplasmic}\text{protein}=\frac{\text{theoretical}\text{maximal}\text{protein}\text{\%}\text{nuc}-\text{DMSO}\text{treated}\text{protein}\text{\%}\text{nuc}}{\mid 100-\text{DMSO}\text{treated}\text{protein}\text{\%}\text{nuc}\mid }$$

$$\text{Relocalizability}\text{of}\text{a}\text{nuclear}\text{protein}=\frac{\text{theoretical}\text{maximal}\text{protein}\text{\%}\text{nuc}-\text{DMSO}\text{treated}\text{protein}\text{\%}\text{nuc}}{\mid 0-\text{DMSO}\text{treated}\text{protein}\text{\%}\text{nuc}\mid }$$

Relative localization score is given by subtracting the absolute value of the relocalizability score from the initial percentage fluorophore in the steady-state primary compartment.

$$\text{Cytoplasmic}\text{localization}\text{score}=\text{DMSO}\text{treated}\text{protein}\text{\%}\text{cytoplasmic}-\left|\text{relocalizability}\text{score}\right|$$

$$\text{Nuclear}\text{localization}\text{score}=\text{DMSO}\text{treated}\text{protein}\text{\%}\text{nuclear}-\left|\text{relocalizability}\text{score}\right|$$

Flow cytometry

Adherent cells were lifted with trypsin and transferred into an Eppendorf or conical tube using complete medium. Cells were pelleted at 300g and washed three times with PBS. Cells were resuspended in PBS and analysed on either a BD Accuri C6+ or an LSR II flow cytometer using FACSDiva software (v.8.0). Data were processed using the FlowJo software package (v.10.0).

Endogenous tagging of proteins using modified PITCh

To facilitate ease of cloning different target single guide RNAs (sgRNAs) into the cassette, a modified all-in-one CRIPSR–Cas9 vector was made containing two orthogonal U6 promoter-stem loop pairs from the parental PITCh vector⁵⁶. The PITCh protocol was then followed to insert donor sequences into endogenous targets. Briefly, a day before transfection, adherent cells were seeded at a density of 25,000 cells in a 10 cm dish. On the day of transfection, medium was replaced with 7 ml of Opti-MEM. Both 1.2 μg of guide vector (containing target sgRNA, PITCh sgRNA and Cas9) and 0.6 μg of donor vector were added to 500 μl of Opti-MEM in a 1.5 ml tube. To another 1.5 ml tube containing 500 μl of Opti-MEM was added 30 μl of Lipofectamine 2000. Both tubes were gently mixed and left to incubate for 5 min at room temperature. The two solutions were then mixed and incubated for 30 min at room temperature, after which the mixture was added dropwise to the 10 cm plate containing cells. Cells were incubated for 24 h, after which the medium was replaced with 10 ml of complete growth medium. At 72 h after transfection, medium was replaced with complete growth medium containing puromycin (1 μg ml⁻¹), and selection continued by replacing medium with fresh complete growth medium supplemented with puromycin every 2 d. After complete death of control cells, puromycin selection was stopped and cells were expanded for immunoblotting.

Immunoblotting

Cells were lifted with trypsin then washed with PBS three times and lysed with RIPA buffer supplemented with protease inhibitor cocktail (Roche) and 0.1% Benzonase (Millipore-Sigma) on ice for 15 min in 1.5 ml tubes. The lysates were centrifuged at 21,000g for 15 min at 4 °C. The supernatant was collected, and the protein concentration was determined by BCA assay (Pierce). Equal amounts of lysates were loaded onto 4–12% Bis-Tris gel and separated by SDS–PAGE. The gel was then transferred onto a nitrocellulose membrane and blocked with 5% milk in TBS-T for 1 h at room temperature. The membrane was incubated with primary antibody overnight at 4 °C and washed three times with TBS-T. Subsequently, the membrane was incubated with secondary antibody for 1 h at room temperature and washed three times with TBS-T for visualization with an Odyssey CLx Imager (LI-COR).

Preparation of antibody-crosslinked beads for immunoprecipitation

Immunoprecipitation with crosslinking was adapted from the Pierce Crosslink Magnetic IP/Co-IP Kit (catalogue no. 88805, Thermo Scientific). Beads were prepared by conjugation with mCherry Polyclonal Antibody (catalogue no. PA5–34974, Invitrogen). For each immunoprecipitation reaction, 2 ml of 1× Modified Coupling Buffer was made by diluting 0.1 ml of 20× Coupling Buffer and 0.1 ml of IP Lysis/Wash Buffer with 1.8 ml of ultrapure water. The bottle of Pierce Protein A/G Magnetic Beads was vortexed to obtain a homogeneous suspension from which 25 μl of beads was aliquoted into a microcentrifuge tube. The tube was placed on a magnetic stand to collect the beads for 1 min. Storage solution was removed and discarded before rinsing the beads with 2 × 500 μl of 1× Modified Coupling Buffer. Then, 7.5 μl of anti-mCherry antibody (catalogue no. PA5–34974, Invitrogen) was diluted to a final concentration of 0.5 μg μl⁻¹ with ultrapure water, 20× Coupling Buffer (diluted 1:20) and IP Lysis/Wash Buffer (diluted 1:20). Next, 100 μl of the prepared antibody solution was added to the beads and incubated on a rotating platform for 15 min at room temperature. Beads were gently mixed every 5–10 min during incubation to ensure that the beads stayed in suspension. The beads were collected with a magnetic stand and the supernatant was removed and discarded. The beads were washed with 1 × 100 μl of 1× Modified Coupling Buffer and 2 × 300 μl of 1× Modified Coupling Buffer. Then, 217 μl of DMSO was added to a fresh tube of disuccinimidyl suberate (DSS) to prepare a 10× solution (25 mM). This stock solution was mixed, then diluted 1:100 in DMSO to give a 0.25 mM DSS solution. Next, 2.5 μl of 20× Coupling Buffer, 4 μl of 0.25 mM DSS and 43.5 μl of ultrapure water were added to the beads and the beads were incubated for 30 min at room temperature on a rotator. The beads were vortexed every 10–15 min during incubation to ensure that the beads stayed in suspension. Beads were collected with a magnetic stand and the supernatant removed. Then, 100 μl of Elution Buffer was added to the beads and gently mixed for 5 min at room temperature on a rotating platform. Beads were collected with the magnetic stand, and this process was repeated twice more. Finally, 200 μl of cold RIPA lysis buffer was added to the beads and tubes were inverted to mix. Beads were collected with a magnetic stand, and supernatant was removed to yield the antibody-crosslinked beads.

Co-immunoprecipitations with TRAMs

For mCherry-eDHFR-NES pulldown, antibody-crosslinked beads were incubated with lysate containing 50 μg of protein made up to a total final volume of 200 μl. To this lysate was added TRAM 1 to a final concentration of 100 nM. The solution was rotated overnight at 4 °C. Before elution, beads were washed three times with RIPA buffer containing 100 nM TRAM 1. Proteins were then eluted off the beads by boiling in 60 μl of elution buffer (5% SDS, 20 mM dithiothreitol (DTT)) at 95 °C for 10 min. Half the volume of the eluate and 2 μg of protein input and flowthrough were loaded onto a 4–12% Bis-Tris gel and separated by SDS–PAGE. The gel was then transferred onto a nitrocellulose membrane and blocked with 5% milk in TBS-T for 1 h at room temperature. The membrane was incubated with primary antibody overnight at 4 °C, and washed three times with TBS-T. Subsequently, the membrane was incubated with secondary antibody for 1 h at room temperature and washed three times with TBS-T for visualization with an Odyssey CLx Imager (LI-COR).

For pulldown of METAP2, 50 μg of protein extract was diluted to a total volume of 200 μl in RIPA buffer and incubated with 50 μl of magnetic agarose beads (Chromotek, GFP Trap (GTMA-20)). TRAM 11 was added to a final concentration of 100 nM and the solution was rotated overnight at 4 °C. Before elution, beads were washed three times with RIPA buffer. Proteins were then eluted off the beads by boiling in 60 μl of elution buffer (5% SDS, 20 mM DTT) at 95 °C for 10 min. Half the volume of the eluate and 2 μg of protein were loaded for the input and flowthrough lanes onto 4–12% Bis-Tris gel and separated by SDS–PAGE. The gel was then transferred onto a nitrocellulose membrane and blocked with 5% milk in TBS-T for 1 h at room temperature. The membrane was incubated with primary antibody overnight at 4 °C, and washed three times with TBS-T. Subsequently, the membrane was incubated with secondary antibody for 1 h at room temperature and washed three times with TBS-T for visualization with an Odyssey CLx Imager (LI-COR).

Immunoprecipitations for immunoblotting of endogenous tagged proteins

For immunoprecipitation of endogenous SOS1, 4 mg of protein extract was diluted to a total volume of 250 μl in RIPA buffer and incubated with 50 μl of magnetic agarose beads (Chromotek, GFP Trap (GTMA-20)), overnight at 4 °C. Before elution, beads were washed three times with RIPA buffer. Proteins were then eluted off the beads by boiling in 60 μl of elution buffer (5% SDS, 20 mM DTT) at 95 °C for 10 min. Half the volume of the eluate and 50 μg of protein were loaded for the input, flowthrough and wild-type HEK293T cell lanes onto 4–12% Bis-Tris gel and separated by SDS–PAGE. The gel was then transferred onto a nitrocellulose membrane and blocked with 5% milk in TBS-T for 1 h at room temperature. The membrane was incubated with primary antibody overnight at 4 °C, and washed three times with TBS-T. Subsequently, the membrane was incubated with secondary antibody for 1 h at room temperature and washed three times with TBS-T for visualization with an Odyssey CLx Imager (LI-COR).

Sodium arsenite treatment and imaging

Adherent cells were plated (25,000 cells per well in an eight-well chamber slide) 2 d before the experiment. At 1 d before treatment, cells were treated with doxycycline to a final working concentration of 1 μg ml⁻¹ in 250 μl of complete growth medium. On the day of the experiment, cells were treated with 30 μM sodium arsenite in 250 μl of complete phenol red-free medium for an hour to induce stress granule formation. After an hour, small-molecule solutions were added directly into the medium in the well, and cells were incubated for another 3 h. Cells were then washed three times with PBS and fixed with 4% paraformaldehyde in PBS for 15 min at room temperature, washed three times and permeabilized with 0.1% Triton for 5 min on ice. Cells were blocked in 10% goat serum in PBS for 1 h at room temperature then incubated with primary antibody overnight at 4 °C. Cells were washed with PBS three times, then incubated with secondary antibody and DAPI for 1 h at room temperature. Cells were washed with PBS and imaged with a Nikon A1R confocal microscope using a Plan Fluor ×60 oil immersion, 1.30-NA objective. The microscope was equipped with a 405 nm violet laser, a 488 nm blue laser, a 561 nm green laser and a 639 nm laser. Images were exported and analysed using the FIJI software package.

For continuous imaging, adherent cells were plated (25,000 cells per ml) in a 24-well glass-bottom plate (CELLTREAT, Scientific Products, catalogue no. 229125) 2 d before treatment. At 1 d before treatment, cells were treated with doxycycline to a final working concentration of 1 μg ml⁻¹ in 500 μl of complete growth medium.

On the day of imaging, medium in the well was replaced with complete phenol red-free medium containing 2 μg ml⁻¹ Hoechst 33342 and 10 μM sodium arsenite. After an hour, the medium in the wells was replaced with complete phenol red-free medium containing 2 μg ml⁻¹ Hoechst 33342, 10 μM sodium arsenite and the requisite concentration of small molecule. Immediately after treatment was administered, timelapse imaging was initiated. The plate was maintained at 37 °C with 5% CO₂ in a stage top incubator (Okolab) over the course of 2.5 h. Images were acquired by manually capturing images every 5 min at a fixed field of view. The Nikon A1R confocal microscope Plan Fluor ×60 oil immersion, 1.30-NA objective was used.

DRG neuron collection and culturing

All animals were housed and experiments performed in accordance with the US National Institutes of Health guidelines for the care and use of laboratory animals, and were approved by Stanford University’s Administrative Panel on Laboratory Animal Care (APLAC no. 20608). At 2 d before collection of primary DRG cells, 250 μl of poly-d lysine (PDL) was added to the bottom of an eight-well chamber slide to coat the bottom of the wells evenly. The slide was incubated at 37 °C in a sterile environment. The next day, PDL was removed and the wells were washed with tissue culture-grade sterile water. The chamber slide was left to dry with the lid open within a sterile biosafety cabinet for about 1 h. The slides were subsequently wrapped with parafilm then left in a 4 °C refrigerator.

Explant collection.

Primary DRG cells were isolated from embryonic (day 17.5–18.5) Sprague-Dawley rats (Charles River Laboratories). A timed pregnant rat was euthanized with carbon dioxide and the embryos retrieved and kept in ice-cold HBSS (Corning, catalogue no. 21–023-CM) supplemented with 10 mM HEPES (Gibco, catalogue no. 15630130). The embryo was decapitated, the ventral organs were removed and the body oriented with the dorsal side facing up. Under a dissection microscope, the skin and the spinal cord were removed and individual DRGs were extracted from either side of the cavity between spinal levels. DRGs from the same embryo were placed in the same 1.5 ml tube. The DRGs were kept in HBSS-HEPES on ice until dissection of all embryos was complete (2–3 h). The DRGs were then placed on a Petri dish on ice and any extra membranes around the DRGs were removed under the dissection microscope. These DRGs were then aseptically transferred into a laminar flow hood and washed with 10 × 1 ml of sterile HBSS-HEPES. After the final wash, the DRGs were resuspended in complete medium, which consists of neurobasal (Gibco, catalogue no. 12348017) medium supplemented with 1% GlutaMAX (Gibco, catalogue no. 35050061), 2% B-27 (Gibco, catalogue no. 17504044), 1% penicillin/streptomycin, 2.5 mg ml⁻¹ d-glucose (Sigma, catalogue no. G7021) and 50 ng ml⁻¹ NGF (2.5S beta subunit from Cedarlane Laboratories, catalogue no. CLMCNET-001.25) that was pre-equilibrated in a humidified 5% CO₂ incubator at 37 °C for 30 min. Single DRGs were seeded in the middle of a PDL-coated well in 20 μl of medium. The chamber slide was left to incubate at 37 °C for 30 min in a 5% CO₂ incubator to allow DRGs to adhere, then 250 μl of neurobasal medium supplemented with 40 μM 5-fluoro-2′-deoxyuridine (FUDR; Sigma, catalogue no. F0503) and 50 ng ml⁻¹ NGF was carefully added to the well. Neurobasal medium supplemented with FUDR and NGF was replaced every other day.

AAV production

On day 0, HEK293T cells were seeded in a six-well plate for 70% confluency on day 1. On day 1, HEK293T cells were co-transfected with plasmids encoding the transgene, RepCap2 and adenoviruses E4, E2A and VA using the PEI MAX transfection reagent according to the manufacturer’s protocol. The morning of day 2, medium was replaced with complete DMEM. On day 3, the viral supernatant was removed and filtered through a 0.45 μm filter.

DRG neuron infection with AAV

Two days after explant seeding, medium in the wells was replaced with 250 ml of 0.2-μm-filtered neurobasal medium supplemented with FUDR (40 mM) and NGF (50 ng ml⁻¹). Then, 8 μl of NMNAT1 transgene viral supernatant and 12 μl of EcDHFR transgene viral supernatant were added to each well. Next, 4 d after infection, neurons were examined for transgene expression.

Axotomy and neuron imaging

Before axotomy, fluorescence images of entire explants were acquired in an epi-fluorescence microscope (Leica DMI 6000B) using a ×10 (0.32-NA) objective. The microscope was equipped with an ORCA-Flash4.0 Digital CMOS camera, a Lumencor SOLA light source and filter sets: 370–39/409/448–63 nm (blue emission), 484–25/505/524–32 nm (green emission), 560–32/581/607–40 nm (red emission) and 640–19/655/680–30 nm (far-red emission).

A biopsy punch (1.0 mm, Royaltek) was sterilized with 70% EtOH, then dipped into neurobasal medium. The biopsy punch was used to sever the cell body from the neurites. The detached cell body was then extracted with sterilized tweezers. To ensure complete removal of the cell body, fluorescence images of entire explants were acquired again in an epi-fluorescence microscope (Leica DMI 6000B) using a ×10 (0.32-NA) objective after axotomy.

To observe degeneration of the neurites, axon termini and the proximal axon region were imaged with a Nikon A1R confocal microscope using a Plan Fluor ×60 oil immersion objective at 0 h, 3 h, 6 h, 12 h, 24 h, 48 h, 72 h and 96 h after injury. Sample size choice was made to be at least n > 3 for a given condition. No randomization of samples or blinding was performed.

Axonal degeneration quantification

Axon degeneration was quantified using a degeneration index calculated using ImageJ (v.1.53) as previously described⁵⁷. In brief, mRuby3 signal was used as a marker of the axon. The mRuby3 channel images were binarized, such that axons were converted to white and all other regions black. The total number of white pixels was defined as total axon area. Axonal fragmentation in degenerating axons leads to their particulate structure, unlike healthy axons which are continuous and intact. To detect degenerated axons, we used the particle analyser module of ImageJ and counted the area of particles. As we used a ×60 objective, particles were defined as having a size of 0–30,000 pixels and circularity above 0.2 (ref. 58). The degeneration index was calculated as the ratio of fragmented axon area over total axon area.

Synthetic chemistry

All synthetic chemistry procedures are available in the Supplementary Information.

DNA sequences

All DNA sequences used can be found in the Supplementary Information.

Extended Data

Extended Data Fig. 1 |. Examination of NMNAT1 export cell lines.

a, Unlinked warhead controls for 1. 2: trimethoprim (TMP), binds EcDHFR, 3: binds FKBP12^F36V. 1a, TRAM which engages EcDHFR and FKBP12^F36V containing a (PEG)₃ linker between the warheads. b, Localization of either protein upon treatment of EL-A with unlinked warheads 2 and 3. c, Comparison of the relative mCherry/GFP median fluorescence intensity ratio between the three isolated clonal export lines (EL-A-C). d, Ability of three NMNAT1 export lines exhibiting different relative export shuttle (mCherry) to NMNAT1 (GFP) ratios to translocate NMNAT1 from the nucleus promoted by 1 or inhibited by leptomycin B (LMB) after 3-hour treatment. e, Representative live-cell images for NMNAT1 relocalization treated with 1 or control molecules, or 1 and leptomycin-B (LMB) in Export Line A (EL-A) after 3 h. f, Representative live-cell images for NMNAT1 relocalization treated with 1 or 1 and leptomycin-B (LMB) in Export Line B (EL-B) after 3 h. g, Representative live-cell images for NMNAT1 relocalization treated with 1, or 1 and leptomycin-B (LMB) in Export Line C (EL-C) after 3 h. Levels of mCherry h, and GFP i, after treatment with 1 for 3 h across four isolated export cell lines with varying nuclear and cytoplasmic protein expression levels. j, Representative live-cell images for NMNAT1 relocalization with export line EL-A treated with varying concentrations of 1a. k, Dose response curve of NMNAT1 and NES localization for EL-A in response to treatment with 1a. MFI: Median fluorescence intensity. Images in e, f, g, j are representative of three biological replicates. Data in b, c, d, h, i, k are mean ± s.d. of three independent experiments. Scale bars are 20 μm. P values in b were determined by unpaired two-tailed t-tests comparing treatment to the DMSO control. NMNAT1 = FKBP12^F36V-GFP-NMNAT1, NES = mCherry-ecDHFR-NES(HIV-REV1) P values in d, k were determined by one-way ANOVA with Dunnett’s post hoc test comparing each condition to the DMSO control. P values in h, i were determined by two-way ANOVA with Šidák post hoc test comparing the treated condition to the DMSO control for each line. Hoechst stains the nucleus.

Extended Data Fig. 2 |. Illustration of gating and segmentation pictures automatically generated from analysis pipeline applied to three different cell types.

a, Imaging analysis pipeline applied on HeLa cells. Gating histograms generated with the set bounds indicated. Masks for nuclear and cytoplasm areas overlayed onto respective fluorophore images. Cytoplasm areas are circumferentially grown from nuclei masks retaining the relative size between the masks. b, Imaging analysis pipeline applied to HEK293T cells. Masks for nuclear and cytoplasm areas overlayed onto respective fluorophore images, with segmentation of highly confluent cells and gating out of dead cells (high nuclear stain over a small area) and debris. c, Representative live-cell images for NMNAT1 relocalization in EL-A treated with 1 or control molecules in cells seeded at high (0.04 × 10⁶ cells) and low density (0.01 × 10⁶ cells). d, Quantification of nuclear GFP and mCherry from c. Images in c are representative of three biological replicates. Data in d are mean ± s.d. of three independent experiments. Scale bars are 20 μm. P values in d were determined by one-way ANOVA with Dunnett’s post hoc test comparing each condition to the DMSO control. Hoechst stains the nucleus.

Extended Data Fig. 3 |. NMNAT1 can be utilized as a nuclear import shuttle.

a, Representative live-cell images of clonal import line A (IL-A) upon treatment with 1 or small molecule controls for 3 h. b, Quantification of localization upon treatment of the IL-A with unlinked warheads 2 and 3. c, Comparison of the relative mCherry/GFP median fluorescence intensity ratio between the three isolated clonal import lines (IL-A-C). d, Ability of three representative import cell lines (IL) with different relative target (mCherry) and shuttle (NMNAT1) ratios to redistribute a diffuse protein target after 3-hour treatment with 1. e, Representative live-cell images of clonal import line B (IL-B) upon treatment with 1 for 3 h. f, Representative live-cell images of clonal import line C (IL-C) upon treatment with 1 for 3 h. g, Quantification of NMNAT1 protein localization change upon treatment with 10 nM 1 in IL-A-C possessing different mCherry/GFP ratios. h, Levels of GFP after treatment with 1 for 3 h in four isolated import cell lines with varying nuclear protein expression levels. i, Levels of mCherry after treatment with 1 for 3 h across four isolated import cell lines with varying diffuse protein expression levels. Images in a, e, f, are representative of three biological replicates. Data in b, c, d, g, h, i are mean ± s.d. of three independent experiments. Scale bars are 20 μm. MFI = median fluorescence intensity. P values in h, i were determined by two-way ANOVA with Šidák post hoc test comparing the treated condition to the DMSO control for each line. P values in b, d were determined by unpaired two-tailed t-tests comparing treatment to the DMSO control.

Extended Data Fig. 4 |. Validation of GR as import shuttle for FKBP12.

a, TRAM engaging GR and FKBP12. b, c Unlinked warhead controls for GR-based TRAMs. 8 binds to GR and 9 binds to FKBP12. d, Nuclear hormone receptor (ERα, GR) hijacking for import of target proteins. e, Nuclear localization of mCherry-FKBP12 after treatment with 6 or unlinked warheads 8 and 9 for 3 h. f, Nuclear localization of GR after treatment with 6 or unlinked warheads for 3 h. g, Genetic construct for stable incorporation. h, Representative live-cell images of GR-dependent FKBP12 import upon treatment with 6 or unlinked warheads 8 and 9 for 3 h. i, FKBP12 nuclear import dose response derived from the analysis pipeline after a 3-hour treatment with 6 engaging GR. j, Nuclear mCherry localization upon treatment with 6 in competition with either 1x or 10x of 8 and 9 for 3 h. k, Nuclear localization of GR upon treatment with 6 in competition with either 1x or 10x of 8 or 9 for 3 h. l, Representative live-cell images of the unlinked warhead competition assay for GR-dependent FKBP12 import after a 3-hour treatment. Images in h, l, are representative of three biological replicates. Data in e, f, i, j, k are mean ± s.d. of three independent experiments. Scale bars are 20 μm. P values in e, f, i, j, k were determined by one-way ANOVA with Dunnett’s post hoc test comparing each condition to the DMSO control. Hoechst stains the nucleus.

Extended Data Fig. 5 |. Validation of GR as import shuttle for FKBP12^F36V.

a. TRAM engaging GR and FKBP12^F36V. b, Representative live-cell images of GR dependent FKBP12^F36V import upon treatment with 4 or unlinked warheads 8 and 3 for 3 h. c, FKBP12^F36V nuclear import dose response derived from the analysis pipeline after a 3-hour treatment with 4 engaging GR. d, Nuclear localization of mCherry-FKBP12^F36V after treatment with 4 or unlinked warheads 3 and 8 for 3 h. e, Nuclear localization of GR after treatment with 4 or unlinked warheads 3 and 8 for 3 h. f, Representative live-cell images of the unlinked warhead competition assay for GR-dependent FKBP12^F36V import after 3-hour treatment. g, mCherry localization upon treatment with 4 in competition with either 1x or 10x of 8 or 3 for 3 h. h, Nuclear localization of GR upon treatment with 4 in competition with either 1x or 10x of 8 or 3 for 3 h. Images in b, f, are representative of three biological replicates. Data in c, d, e, g, h are mean ± s.d. of three independent experiments. Scale bars are 20 μm. P values in c, d, e, g, h were determined by one-way ANOVA with Dunnett’s post hoc test comparing each condition to the DMSO control. Hoechst stains the nucleus.

Extended Data Fig. 6 |. Validation of ERα as import shuttle for FKBP12.

a. TRAM engaging ERα and FKBP12, raloxifene (10) unlinked control binds to ERα. b, FKBP12 nuclear import dose response derived from the analysis pipeline after a 3-hour treatment with 7 engaging ERα. c, Representative live-cell images of ERα-dependent FKBP12 import upon treatment with 7 or unlinked warheads 10 and 9 for 3 h. d, Nuclear localization of mCherry-FKBP12 after treatment with 7 or unlinked warheads 10 and 9 for 3 h. e, Nuclear localization of ERα after treatment with TRAM 7 or unlinked warheads 10 and 9 for 3 h. f, Nuclear mCherry localization upon treatment with 7 in competition with either 1x or 10x of 10 or 9 for 3 h. g, Nuclear localization of ERα upon treatment with 7 in competition with either 1x or 10x of 10 or 9 for 3 h. h, Representative live-cell images of the unlinked warhead competition assay for ERα-dependent FKBP12 import after a 3-hour treatment. Images in c, h are representative of three biological replicates. Data in b, d, e, f, g are mean ± s.d. of three independent experiments. Scale bars are 20 μm. P values in b, d, e, f, g, were determined by one-way ANOVA with Dunnett’s post hoc test comparing each condition to the DMSO control. Hoechst stains the nucleus.

Extended Data Fig. 7 |. Validation of ERα as import shuttle for FKBP12^F36V.

a. TRAM engaging ERα and FKBP12^F36V. b, FKBP12^F36V nuclear import dose response derived from the analysis pipeline after a 3-hour treatment with 5 engaging ERα. c, Representative live-cell images of ERα-dependent FKBP12^F36V import upon treatment with 5 or unlinked warheads 3 and 10 for 3 h. d, Nuclear localization of mCherry-FKBP12^F36V after treatment with 5 or unlinked warheads 3 and 10 for 3 h. e, Nuclear localization of ERα after treatment with 5 or unlinked warheads 3 and 10 for 3 h. f, Representative live-cell images of the unlinked warhead competition assay for ERα-dependent FKBP12^F36V import for 3 h. g, mCherry localization upon treatment with 5 in competition with either 1x or 10x of 3 or 10 for 3 h. h, Nuclear localization of ERα upon treatment with 5 in competition with either 1x or 10x of 3 or 10 for 3 h. Images in c, f, are representative of three biological replicates. Data in b, d, e, g, h are mean ± s.d. of three independent experiments. Scale bars are 20 μm. P values in b, d, e, g, h were determined by one-way ANOVA with Dunnett’s post hoc test comparing each condition to the DMSO control. Hoechst stains the nucleus.

Extended Data Fig. 8 |. Nuclear import of mutant proteins with ERα.

a, Representative live-cell images of HeLa cells stably expressing target mutant protein SMARCB1^Q318X (SMARCB1*) and ERα under an inducible promoter after a 3-hour treatment of 5. b, Representative live-cell images of HeLa cells stably expressing target mutant protein TDP43^ΔNLS (TDP* = TDP43^{K82A/R83A/K84A}) and ERα under an inducible promoter after a 3-hour treatment of 5. c, Representative live-cell images of HeLa cells stably expressing target mutant protein FUS^R495X (FUS*) and ERα under an inducible promoter after a 3-hour treatment of 5. d, Violin plots of the distribution of respective target mutant protein and ERα in the three cell lines under different treatment conditions, across 3 repeats. e, Mean percent nuclear target protein and ERα upon treatment with 5 for 3 h in HeLa cells stably expressing the target mutant protein and ERα under an inducible promoter. f, Representative immunoblot of mutant proteins after a 3-hour treatment with 5. Images in a, b, c are representative of three biological replicates. Data in e, is mean ± s.d. of three independent experiments. Data in f is representative of two independent experiments. P values in e were determined by one-way ANOVA comparing each condition to the DMSO control. Scale bars are 20 μm. Hoechst stains the nucleus. Uncropped blots can be found in Supplementary Fig. 7.

Extended Data Fig. 9 |. TRAM-mediated protein relocalization from stress granules.

a, Timelapse imaging snapshots of FUS^R495X when cells are treated with DMSO after stress granule formation. b, Timelapse imaging snapshots of FUS^R495X extraction from granules when cells are treated with 5. c, Representative fixed-cell immunofluorescent images of stress granule marker G3BP1 and mCherry-FUS^R495X after treatment with 5 or control compounds for 3 h. d, Quantification of stress granules in cells after treatment with varying concentrations of 5 or control compounds for 3 h. Images in c are representative of three biological replicates. Images in a, b are representative of two independent experiments. Data in d are mean ± s.d. of three independent experiments. Scale bars are 20 μm. P values in d were determined by two-way ANOVA with Dunnett’s post hoc test comparing each condition to their DMSO control.

Extended Data Fig. 10 |. Validation of PARP1 and METAP2 as protein shuttles in ectopic expression systems.

a, Unlinked warheads that engage METAP2 (14) and PARP1 (15). b, Co-immunoprecipitation of mCherry-FKBP12^F36V upon GFP-METAP2 pulldown in the presence of TRAM. I = input, F = flowthrough, E = eluate. c, Representative live-cell images of mCherry-FKBP12^F36V after treatment with 11 or with unlinked warheads, 14 or 3 for 3 h in the presence of METAP2. d, Quantification of FKBP12^F36V relocalization upon treatment with 11 for 3 h. e, METAP2 localization upon TRAM 11 treatment for 3 h. 11. f, Representative live-cell-images of FKBP12^F36V after treatment with 12 or the unlinked warheads, 15 and 3 or in the absence of PARP1 after 3 h. g, Quantification of mCherry-FKBP12^F36V relocalization upon treatment with 12 for 3 h. h, Quantification of GFP-PARP1 localization upon TRAM upon treatment with 12 for 3 h. i, mCherry-FKBP12^F36V localization upon treatment with 12 for 3 h in competition with either 1x or 10x of each individual unlinked warhead 15 and 3. j, Quantification of mCherry-FKBP12^F36V localization from i. k, Quantification of GFP-PARP1 localization from i. Images in c, f, i, are representative of three biological replicates. Immunoblots in b are representative of two independent experiments. Data in d, e, g, h, j, k are mean ± s.d. of three independent experiments. Scale bars are 20 μm. P values in d, e, g, h, j, k were determined by one-way ANOVA with Dunnett’s post hoc test comparing each condition to the DMSO control. Hoechst stains the nucleus. Uncropped blots can be found in Supplementary Fig. 7.

Extended Data Fig. 11 |. Validation of Cas9-CRISPR mediated endogenously tagged lines.

a, Schematic for modified PITCh knock-in into target proteins. b, Sequencing of the regions near the N terminus of the endogenous PRMT9 gene after donor cassette insertion into endogenous PRMT9. c, Sequencing of the region downstream the N terminus of the endogenous SOS1 gene after donor cassette insertion into endogenous SOS1. d, Due to high GC content of the genomic region upstream of the N terminus of SOS1, immunoblotting was used to validate successful incorporation of donor cassette onto endogenous SOS1 after GFP pulldown, I = input, F = flowthrough, E = eluate. e, Sequencing of the regions near the N terminus of the endogenous FKBP12 gene after donor cassette insertion into endogenous FKBP12. f, Demonstration of GFP-FKBP12^F36V tagging of endogenous PRMT9, SOS1 and GFP tagging of endogenous FKBP12 in HEK293T cells. Immunoblots in d,f are representative of two independent experiments. Uncropped blots can be found in Supplementary Fig. 7.

Extended Data Fig. 12 |. METAP2 as an endogenous nuclear export shuttle to relocalize endogenous PRMT9.

a, Representative live-cell images of endogenous PRMT9 localization upon treatment with different concentrations of TRAM 11 for 24 h. b, Violin plots illustrating the homogenous population shift in PRMT9 localization upon TRAM 11 treatment at different concentrations for 24 h. c, Representative live-cell images of endogenous PRMT9 localization upon treatment with unlinked warheads 14 and 3 for 24 h. d, Quantification of localization changes observed in c. e, Violin plots showcasing the lack of homogenous population shift in PRMT9 localization when treated with the unlinked warheads 14 and 3 versus TRAM 11 for 24 h. f, Quantification of the change in PRMT9 localization upon treatment with 11 after pre-treatment for 1 h with either 1x or 10x of 14 or competed with 1x, or 10x of 3 for 24 h. g, Representative images of the competition assay quantified in f. Images in a, c, g are representative of three biological replicates. Data in d, f are mean ± s.d. of three independent experiments. Scale bars are 20 μm. P values in b, d, e, f were determined by one-way ANOVA with Dunnett’s post hoc test comparing each condition to the DMSO control. Hoechst stains the nucleus.

Extended Data Fig. 13 |. PARP1 as an endogenous nuclear import shuttle to relocalize endogenous SOS1.

a, Representative live-cell images of endogenous SOS1 localization upon treatment with different concentrations of TRAM 12 for 24 h. b, Violin plots illustrating the homogenous population shift in SOS1 localization upon TRAM 12 treatment at different concentrations for 24 h. c, Representative live-cell images of endogenous SOS1 localization after treatment with unlinked warheads 15 and 3 for 24 h. d, Quantification of localization changes observed in c. e, Violin plots showcasing the lack of homogenous population shift in SOS1 localization when treated with the unlinked warheads 15 and 3 versus TRAM 12 for 24 h. f, Quantitation of change in SOS1 localization upon treatment with 12 in competition with either 1x or 10x of 15 or 3 for 24 h. g, Representative images of SOS1 localization in the competition assay in f. Images in a, c, g are representative of three biological replicates. Data in d, f are mean ± s.d. of three independent experiments. Scale bars are 20 μm. P values in b, d, e, f were determined by one-way ANOVA with Dunnett’s post hoc test comparing each condition to the DMSO control. Hoechst stains the nucleus.

Extended Data Fig. 14 |. PARP1 as an endogenous nuclear import shuttle to relocalize endogenous FKBP12.

a, Representative live-cell images of partial endogenous FKBP12 localization upon treatment with different concentrations of TRAM 13 for 24 h. b, Violin plots illustrating a homogenous population shift in FKBP12 localization upon TRAM 13 treatment at different concentrations for 24 h. c, Representative live-cell images of endogenous FKBP12 localization after treatment with unlinked warheads 15 and 9 for 24 h. d, Quantification of localization changes observed in c. e, Violin plots showcasing the lack of homogenous population shift in FKBP12 localization when treated with the unlinked warheads 15 and 9 versus TRAM 13 for 24 h. f, Quantification of the change in FKBP12 localization upon treatment with 13 in competition with either 1x or 10x of 15 or 9 for 24 h. g, Representative images of FKBP12 localization in the competition assay in f. Images in a, c, g are representative of three biological replicates. Data in d, f are mean ± s.d. of three independent experiments. Scale bars are 20 μm. P values in b, d, e, f were determined by one-way ANOVA with Dunnett’s post hoc test comparing each condition to the DMSO control. Hoechst stains the nucleus.

Extended Data Fig. 15 |. Nuclear export of NMNAT1 in neurons.

a, Representative live-cell images of NMNAT1 upon treatment with TRAM 1 or unlinked warheads 2 and 3 at 10 nM for 24 h. b, Representative images of before and after axotomy on explants. Axotomies were performed using a biopsy punch, the mRuby3-Axontag protein serves as an axonal marker, Axontag = GAP43at-mRuby3-ecDHFR. c, Representative images of axons and axon termini post-axotomy, that have been pre-treated with unlinked controls for 24 h prior. d, Representative images of mNMNAT1 in axons after exposure to 1 for 24 h followed by axotomy. mRuby3 signal was used to monitor degeneration across treated neurons. d, Quantification of axon degeneration over time. e, Quantification of axon degeneration over time in axon termini. f, Comparison of the degeneration index in axona at different time points with different treatment conditions. g, Comparison of the degeneration index (DI) in the axon termini at different time points with different treatment conditions. DI was calculated as the ratio of fragmented axon area over total axon area. For plots in d, f, a total of 12 replicates derived from 4 different embryos were combined in the plots for all conditions except 1 nM 2 + 1 nM 3. A total of 6 replicates derived from 2 different embryos were combined in the plots for the 1 nM 2 + 1 nM 3 condition. For plots in e, g, a total of 6 replicates derived from 4 different embryos were combined in the plots for all conditions except 1 nM 2 + 1 nM 3. A total of 4 replicates derived from 2 different embryos were combined in the plots for the 1 nM 2 + 1 nM 3 condition. Data in d, e is shown as the mean ± s.e.m. values. Data in f, g, is shown as the mean ± s.d. values. P values in d, e, were determined by two-way ANOVA with Tukey’s post hoc test comparing the 1 nM treatment of 1 with the DMSO (purple) or the unlinked warhead control (green) at each time point. P values in f, g, were determined by one-way ANOVA with Dunnett’s post hoc test comparing each treatment to the DMSO control at each time point. Images in a, b, c are representative of neurons harvested from 4 different embryos. Scale bars are 20 μm.

Data availability

All data supporting the findings of this study are available within the Article, Extended Data Figs. 1–15, Supplementary Information or associated data tables. All single-cell quantification data are available upon request.