A Magnetic Nanoprobe Technology for Detecting Molecular Interactions in Live Cells

See allHide authors and affiliations

Science  01 Jul 2005:
Vol. 309, Issue 5731, pp. 121-125
DOI: 10.1126/science.1112869

This article has been retracted. Please see:


Technologies to assess the molecular targets of biomolecules in living cells are lacking. We have developed a technology called magnetism-based interaction capture (MAGIC) that identifies molecular targets on the basis of induced movement of superparamagnetic nanoparticles inside living cells. Efficient intracellular uptake of superparamagnetic nanoparticles (coated with a small molecule of interest) was mediated by a transducible fusogenic peptide. These nanoprobes captured the small molecule's labeled target protein and were translocated in a direction specified by the magnetic field. Use of MAGIC in genome-wide expression screening identified multiple protein targets of a drug. MAGIC was also used to monitor signal-dependent modification and multiple interactions of proteins.

Modern medicine faces the challenge of developing safer and more effective therapies. However, many drugs currently in use were identified without knowledge of their molecular targets (1, 2). Bioactive natural products are an important source of drug leads, but their modes of action are usually unknown (2). Elucidation of their physiological targets is essential for understanding their therapeutic and adverse effects, thereby enabling the development of second-generation therapeutics. Moreover, the discovery of novel targets of clinically proven compounds may suggest new therapeutic applications (3). Target identification (ID) is also important in chemical biology, where high-throughput screening is used to identify small molecules with a desired phenotype (4, 5). Despite the great benefits of such a screen, this approach has been hampered by the daunting task of target ID (1, 4, 5).

Superparamagnetic nanoparticles (MNPs) are biocompatible and are in routine clinical use (6, 7). We built on these nanomagnetic probes to develop a target ID technology MAGIC, to directly probe molecular interactions inside living cells with high sensitivity and selectivity (fig. S1). Streptavidin-conjugated MNPs were used as a generic reagent to attach biotinylated molecules to the nanoprobe. After internalization of small molecule–coated MNPs into the cells, protein(s) bind to the small molecule on the MNP. Thus, when a magnetic field is applied, the MNP and associated target protein(s) can be concentrated. Fusion of a fluorescent probe to the target protein renders this translocation easily detectable by confocal microscopy.

To visualize and track the distribution of MNPs within cells, we labeled MNPs with fluorescein isothiocyanate (FITC) (8) (Fig. 1A). FITC-MNPs were not efficiently introduced into the cell cytosol. Most of the internalized FITC-MNPs appeared to be trapped within endocytic vesicles (Fig. 1A). Attachment of transducible fusogenic TAT-HA2 peptide on the MNPs markedly enhanced endocytic uptake and subsequent release from endosomes. It is currently understood that the high density of cationic residues in the protein transduction domain of human immunodeficiency virus TAT protein causes an electrostatic interaction with the negatively charged cell surface, thus enhancing the chance of endocytic internalization, whereas the N-terminal 20 amino acids of the influenza virus hemagglutinin protein HA2 destabilizes endosomal membranes, causing them to release their contents into the cytosol (9, 10). Cotreatment with chloroquine, which is known to enhance endosome disruption (11), further increased the concentration of MNPs outside endosomes (Fig. 1A).

Fig. 1.

Proof-of-principle experiments for MAGIC. (A) Enhanced intracellular uptake and endosomal escape of MNPs by TAT-HA2. HeLa cells were incubated with MNP coated with FITC (FITC-MNP) or with FITC and TAT-HA2 (FITC-MNP-TATHA2) in the absence or presence of 100 μM chloroquine for 12 hours. Live-cell confocal images were taken to show the intracellular distribution of FITC-labeled MNPs. Scale bars, 50 μm. (B and C) Translocation of MNPs inside cells by the magnetic field. HeLa cells were incubated with FITC-MNP-TATHA2 or QD565-TATHA2 in the presence of 100 μM chloroquine for 12 hours. Scale bars in (B), 50 μm. (D and E) DEVD-coated MNPs direct the translocation of caspase-3–mRFP by the magnetic field. HeLa cells were transfected with the expression plasmids for caspase-3–mRFP or EGFP, detached, mixed together, and replated. These cells were incubated with MNPs coated with DEVD and TAT-HA2 in the presence of 100 μM chloroquine for 12 hours. In (B) and (D), live-cell confocal images were taken before and after application of a magnetic field (M.F.), with the focal plane at the cellular basal surface. Cells expressing EGFP or caspase-3–mRFP are indicated by green or red arrowheads, respectively, in (D). Scale bar in (D), 50 μm. In (C) and (E), quantitative analysis of the signals was performed with more than 100 cells.

We next addressed whether internalized MNPs could be moved inside cells by an external magnetic field (Fig. 1, B and C). HeLa cells were transduced with MNPs coated with FITC and TAT-HA2. As a control, a luminescent nanocrystal quantum dot, QD565, which does not exhibit magnetism (12), was internalized into another set of HeLa cells with the use of TAT-HA2. Specific translocation of MNPs, but not QD565, was observed inside cells after brief application of a magnetic field (Fig. 1, B and C). Furthermore, this translocation was reversible; the MNPs rapidly diffused away upon removal of the magnetic field and were redirected when it was reapplied.

We determined whether the MAGIC principle could be used to detect the intracellular target for Asp-Glu-Val-Asp (DEVD), an apoptosis inhibitor known to bind caspase-3 (13) (Fig. 1, D and E). HeLa cells were transfected with the expression construct for enhanced green fluorescent protein (EGFP) or caspase-3 fused to monomeric red fluorescent protein (mRFP). Next, these two sets of transfectants were detached, mixed, and replated for incubation with MNPs coated with DEVD and TAT-HA2. A notable amount of “red” signal, but not “green” signal, was translocated in the direction of the magnetic field (Fig. 1, D and E), indicating a specific interaction between DEVD and caspase-3 inside cells. This translocation was reversible upon removal and reapplication of the magnetic field (Fig. 1, D and E) (movie S1).

To use MAGIC in systematic target ID for a bioactive small molecule, we exploited expression cloning based on a retroviral EGFP fusion protein expression library (Fig. 2). We sought to identify the receptor(s) for an immunosuppressant, FK506 (fig. S2A), in a genome-wide screen. A normalized EGFP-tagged cDNA library was generated (8); cDNAs from multiple human tissues were fused to the 5′ or 3′ end of the gene encoding EGFP, and mRFP was coexpressed as an internal control to address potential false positives (fig. S2B). This library was stably expressed in HeLa cells by retroviral transduction. After introduction of FK506-coated MNPs into these cells, a magnetic field was applied, and the subcellular localization of proteins expressed from cDNA-EGFP fusions was examined. Nineteen positive clones were identified that exhibited specific translocation of EGFP in the direction of the magnetic field while the subcellular localization of mRFP remained unchanged (Table 1 and Fig. 2). Reverse transcription polymerase chain reaction (RT-PCR) and BLAST analysis of mRNAs from these clones identified overlapping transcripts of several known FK506-binding proteins and proteins of unknown function (Table 1 and Fig. 2) (1417). Specific interaction of these proteins with FK506 was verified by competition analysis (fig. S3). These interactions were also readily detected with other cell types (fig. S4).

Fig. 2.

Molecular target ID based on MAGIC. (A and B) HeLa cells were infected with the retroviral EGFP–fusion protein expression library (fig. S2B). After incubation of these cells with MNPs coated with FK506 and TAT-HA2 in the presence of 100 μM chloroquine for 12 hours, the subcellular localization of EGFP was examined in the absence or presence of a magnetic field (M.F.). To address potential false positives, we simultaneously monitored mRFP bicistronically coexpressed with EGFP–fusion protein. Live-cell confocal images were taken with the focal plane at the cellular basal surface (bottom) or with the pinhole size increased to collect whole-cell images (whole). Arrowheads indicate positive clones. RT-PCR and sequence analysis of mRNAs from these clones identified several proteins (Table 1) including FKBP12 (A) and an unknown protein with NCBI accession number BAB15266 (B). Scale bars, 100 μm.

Table 1.

Protein targets of FK506 identified from expression cloning with the use of MAGIC. A total of 19 positive clones were identified from ∼107 cells. The number of independent cDNA clones for each protein isolated from the screen is shown.

NCBI accession number Protein name Number isolated Reported binding to FK506
NP_463460 FKBP12 4 Yes (View inlineView inline)
Q16645 FKBP12.6 2 Yes (View inline)
P26885 FKBP13 1 Yes (View inline)
AAA58475 FKBP25 2 Yes (View inline, View inline)
Q02790 FKBP52 3 Yes (View inline, View inline)
AAA86245 FKBP54 2 Yes (View inline)
NP_068758 FKBP65 1 Yes (View inline)
BAB15266 Unnamed 1 No
BAB15220 Unnamed 1 No
BAC03954 Unnamed 1 No
BAD18781 Unnamed 1 No
Total   19  

The specificity of these interactions could be distinguished readily from background and false positive signals on the basis of reversible translocation manipulated by magnetic field at the initial screening stage (Fig. 2). Overall, the screen identified diverse targets with high efficiency and no false positives (Table 1). The additional targets discovered for FK506 might give some clues about the molecular mechanisms underlying its unexpected therapeutic actions and debilitating side effects (17).

To evaluate the feasibility of MAGIC for probing intracellular signaling processes, we used the NF-κB/IκB pathway (Fig. 3). Members of the NF-κB family (e.g., RelA/p65) are sequestered in the cytoplasm by IκB family members (e.g., IκBα). Various stimuli, including proinflammatory cytokines such as tumor necrosis factor–α (TNF-α), induce the phosphorylation of IκBα on two serine residues, Ser32 and Ser36 (18). This phosphorylation results in recognition of IκBα by the F box protein βTrCP of the SKP1/cullin/F-box ubiquitin ligase complex, leading to ubiquitin-dependent proteolysis of IκBα. These regulated protein-protein interactions allow NF-κB proteins to translocate into the nucleus, resulting in the expression of target genes.

Fig. 3.

Monitoring biological signaling processes by MAGIC. (A) Schematic of IκBα-mRFP, IκBα-ECFP-FKBP12, EYFP-RelA, and mRFP-βTrCP. Symbols used in (B) and (E) are also explained. (B to D) Detection of signal-induced phosphorylation of IκBα. After transfection of HeLa cells with the expression plasmid for IκBα-mRFP, MNPs coated with antibody to phosphorylated Ser32 of IκBα, FITC, and TAT-HA2 were introduced into these cells. Magnetic field–induced translocation of FITC and mRFP signals was monitored before and after stimulation with TNF-α (10 ng/ml) for 5 min in the absence or presence of prior treatment with SC-514 (1 mM). (E to G) Detection of signal-dependent association of IκBα with βTrCP. After transfection of HeLa cells with the expression plasmids for IκBα-ECFP-FKBP12, EYFP-RelA, and mRFP-βTrCP, MNPs coated with FK506 and TAT-HA2 were introduced into these cells. Magnetic field–induced translocation of enhanced cyan fluorescent protein (ECFP), enhanced yellow fluorescent protein (EYFP), and mRFP signals was monitored before and after stimulation with TNF-α (10 ng/ml) for 5 min in the absence or presence of prior treatment of SC-514 (1 mM). Scale bars, 50 μm. Shown in (D) and (G) are quantitative analyses of induced signals after stimulation with TNF-α for 5 min with more than 100 cells.

We first tested whether signal-induced protein phosphorylation could be detected with MAGIC (Fig. 3, A to D). After expression of IκBα fused to mRFP in HeLa cells, these cells were loaded with MNPs triple-labeled with TAT-HA2, FITC, and antibody to phosphorylated Ser32 of IκBα (8). Before stimulation, translocation of IκBα-mRFP was barely detectable, despite active accumulation of FITC signal in the direction of the magnetic field (Fig. 3, B to D). In contrast, brief stimulation with TNF-α markedly induced the magnetic field–directed translocation of mRFP signal for IκBα, faithfully reflecting the phosphorylation of IκBα in response to TNF-α inside cells (fig. S5). Under these conditions, SC-514, an inhibitor of IκB kinase (IKK), prevented TNF-α–induced translocation of IκBα-mRFP (Fig. 3, C and D), further emphasizing the specificity of the method.

Next we examined whether signal-dependent protein-protein interactions could be detected with MAGIC (Fig. 3, A, E to G). To entrap IκBα with MNPs inside cells, we used MNPs coated with FK506 and TAT-HA2 (TATHA2-MNP-FK506) together with IκBα tagged with FK506-binding protein, FKBP12 (IκBα-ECFP-FKBP12; Fig. 3, A, E to G). After expression of IκBα-ECFP-FKBP12, EYFP-RelA, and mRFP-βTrCP in HeLa cells, these cells were transduced with TATHA2-MNP-FK506. Application of a magnetic field directed the translocation of ECFP signal (Fig. 3F), reflecting IκBα recruitment around MNPs by FK506-FKBP12 interaction inside cells. Under these conditions, EYFP-RelA, but not mRFP-βTrCP, was directed toward the magnetic field, indicating that interaction occurs between IκBα and RelA, but not between IκBα and βTrCP. Upon stimulation with TNF-α, a marked increase in mRFP-βTrCP translocation was observed (Fig. 3, F and G), demonstrating signal-induced interaction between IκBα and βTrCP. Notably, SC-514 blocked TNF-α–induced translocation of mRFP-βTrCP (Fig. 3, F and G). Thus, MAGIC can faithfully probe molecular interactions dynamically regulated by specific signals in live cells.

In certain disease conditions (e.g., systemic inflammation), several cellular functions, such as endocytosis, are known to be compromised (19, 20). To demonstrate MAGIC in cells impaired in endocytosis, we adopted microinjection that can introduce exogenous materials directly into cells without passing through endocytic vesicles (21) (fig. S6). The signal-induced phosphorylation and protein-protein interaction between IκBα and RelA were probed in HeLa cells pretreated with TNF-α, which inhibited the endocytic uptake of MNPs, as described (20). MNPs coated with antibody to phosphorylated Ser32 of IκBα or with FK506 were microinjected into TNF-α–pretreated HeLa cells expressing IκBα-mRFP or IκBα-ECFP-FKBP12/EYFP-RelA/mRFP-βTrCP, respectively (8). Signal-induced phosphorylation and interaction of the NF-κB/IκBpathwaywere readily detected by MAGIC in this experimental setting (fig. S6).

MAGIC offers several advantages over target ID methods currently in use. First, it directly translates a physical molecular interaction into a clear readout signal, unlike indirect readout methods that are dependent on intermediary interactions (22), overall expression profiles (23), or complex biological phenotypes (24). Thus, intrinsic false positives/negatives or error-prone deductions about molecular target(s) of a small molecule are obviated. Second, by probing such interactions in a physiologically relevant context, misleading outcomes produced by an artificial experimental setting (22, 2428) can be greatly diminished. Third, it is amenable to dynamic, single-cell analysis of interactions. Finally, MAGIC can be used to detect a variety of biological interactions and protein modifications within live cells in a broad range of tissues and disease states. With the great advantage of being able to detect dynamic interactions between the biomolecules within mammalian cells, this technology could be exploited in genome-wide interaction screens.

The benefits of MAGIC may be best achieved through efficient and nondisruptive introduction of MNPs into cells. Prolonged incubation of cells with TAT-HA2–conjugated MNPs may affect cellular physiology, and microinjection cannot be used for large populations of cells. Other technologies for delivering biologically active cargos into cells (10) will be helpful to complement MAGIC.

Supporting Online Material

Materials and Methods

Figs. S1 to S6

Movie S1

References and Notes

References and Notes

View Abstract

Navigate This Article