Manipulating signaling at will: chemically-inducible dimerization (CID) techniques resolve problems in cell biology
- First Online:
- Cite this article as:
- DeRose, R., Miyamoto, T. & Inoue, T. Pflugers Arch - Eur J Physiol (2013) 465: 409. doi:10.1007/s00424-012-1208-6
- 894 Views
Chemically-inducible dimerization (CID) is a powerful tool that has proved useful in solving numerous problems in cell biology and related fields. In this review, we focus on case studies where CID was able to provide insight into otherwise refractory problems. Of particular interest are the cases of lipid second messengers and small GTPases, where the “signaling paradox” (how a small pool of signaling molecules can generate a large range of responses) can be at least partly explained through results gleaned from CID experiments. We also discuss several recent technical advances that provide improved specificity in CID action, novel CID substrates that allow simultaneous orthogonal manipulation of multiple systems in one cell, and several applications that move beyond the traditional CID technique of moving a protein of interest to a specific spatiotemporal location.
KeywordsChemically-inducible dimerizationRapamycinSignaling paradox
CID case studies resolving biological problems
Biological system studied
KCNQ ion channels
PI(4,5)P2 sufficient to drive current through KCNQ channels
Plasma membrane targeting
Suh et al. Science 314, 1454 (2006)
Ca2+ influx, receptor-mediated endocytosis
Loss of PI(4,5)P2 disrupts Ca2+ influx, endocytosis
Plasma membrane targeting
Varnai et al. J Cell Biol 175, 377 (2006)
Clathrin-coated endocytic pits
Loss of PI(4,5)P2 disrupts clathrin-coated pits, dissociates Arp2/3 from plasma membrane
Plasma membrane targeting
Zoncu et al. Proc Nat Acad Sci USA 104, 3793 (2007)
Loss of PI(4,5)P2 dissociates AP-2 from plasma membrane at endocytic pits
Plasma membrane targeting
Abe et al. J Cell Sci 121, 1488 (2008)
PTEN tumor suppressor
Binding of PTEN to plasma membrane requires PI(4,5)P2
Plasma membrane targeting
Rahdar et al. Proc Nat Acad Sci USA 106, 480 (2009)
Loss of PI3P in maturing endosomes delays maturation, disrupts receptor recycling
Endosomal membrane targeting
Fili et al. Proc Nat Acad Sci USA 103, 15473 (2006)
Cell-surface receptor CD25
Recruitment of Rac1 to CD25 causes actin reorganization, receptor phagocytosis
Plasma membrane targeting
Castellano et al. J Cell Sci 113, 2955 (2000)
Axonal growth cone
HRas and PI3K form positive feedback loop, required for symmetry breaking
Plasma membrane targeting
Fivaz et al. Curr Biol 18, 44 (2008)
PI3K-Rac-actin polymerization feedback loop for neutrophil polarization, migration; Rac activation alone insufficient
Plasma membrane targeting
Inoue and Meyer. PLoS ONE 3, e3068 (2008)
Organelle targeting motifs, Ras signaling
Expanded repertoire of target organelles for CID
Plasma membrane, organelle targeting, membrane cross-linking
Komatsu et al. Nat Methods 7, 206 (2010)
ZAP70 tyrosine kinase
Recruitment of ZAP70 to plasma membrane in correct orientation induces signaling through MAPK and calcineurin pathways
Plasma membrane targeting
Graef et al. EMBO J 16, 5618 (1997)
Akt kinase in apoptosis
Recruitment of Akt to plasma membrane activates GSK3, NF-κB, sufficient to rescue cells from apoptosis-inducing stimuli
Plasma membrane targeting
Li et al. Gene Therapy 9, 233 (2002)
β-arrestin2 and vasopressin receptors
Association of β-arrestin 2 to GPCR causes internalization; recruitment of β-arrestin alone to plasma membrane sufficient to induce Erk1/2 signaling
Terrillon and Bouvier EMBO J 23, 3950 (2004)
Ca2+ channels at plasma membrane-ER junctions
Oligomerization of STIM1 causes accumulation at plasma membrane-ER junctions, CRAC channel formation
Luik et al. Nature 454, 538 (2008)
Postsynaptic density (PSD) of neurons
CaMKIIa translocation to PSD acts to recruit proteasome to dendritic spines
Postsynaptic density (PSD) targeting
Bingol et al. Cell 140, 567 (2010)
Photocaged rapamycin can be activated intracellularly
Photocaged rapamycin (“direct” route)
Karginov et al. J Am Chem Soc 133, 420 (2011)
Localized photouncaging of rapamycin gives local results
Photocaged rapamycin (“indirect” route)
Umeda et al. J Am Chem Soc 133, 12 (2011)
Plant hormone-based CID
Abscisic acid (ABA)-based CID system to activate transcription
Abscisic acid (ABA)-based CID system
Liang et al. Sci Signaling 4, rs2(2011)
Plant hormone-based CID
Gibberellin A3 (GA3)-based CID system combined with rapamycin system for intracellular logic gates
Gibberellin A3 (GA3)-based CID system
Miyamoto et al. Nat Chem Bio 8, 465 (2012)
Inducible protein activity
CID-activatable kinase activity independent of localization
CID-activatable kinase activity
Karginov et al. Nat Biotechnology 28, 743 (2010)
ER-mitochondrial contact regions
ER-mitochondrial membrane contact at many discrete points
Csordas et al. Mol Cell 39, 121 (2010)
Plasma membrane-ER contact regions
STIM1-Orai1 complex requires other members for productive interaction
Varnai et al. J Biol Chem 282, 29678 (2007)
Nuclear proteins in yeast
Rapid protein inactivation can be used to establish conditional mutant phenotypes
“Anchor-Away” for rapid protein inactivation
Haruki et al. Mol Cell 31, 925 (2008)
Adaptor protein complexes of clathrin-coated vesicles
AP-1 involved in retrograde trafficking, not anterograde
Inactivation by rerouting to mitochondria
Robinson et al. Dev Cell 18, 324 (2010)
Initial sequestration of protein at Golgi minimizes background activity of target GTPases
Initial sequestration of protein at Golgi to minimize background
Phua et al. ACS Chem Biol (2012)
Phosphoinositide signals leading to actin reorganization
PI(4,5)P2 increase and PI(4)P depletion have separate effects on actin reorganization
“Lipid liberation” to increase lipid without depleting precursors
Ueno et al. Sci Signaling 4, ra87 (2011)
Lipids as targets of CID
The lipids of the plasma membrane have proved to be an attractive target for manipulation by CID. Given the importance of lipids as second messengers, and the involvement of PM in such fundamental processes as endocytosis, receptor trafficking, and cell migration, its potential interest as a target is obvious. Conventional techniques, such as the use of engineered kinase and phosphatase enzymes, including constitutively active and dominant negative versions, has yielded significant insight, but these techniques have limitations in the amount of time required for induction of an effect (timescale of hours) compared to the short period (seconds) that lipid signaling takes, and the difficulty of extrapolating from global effects of lipid manipulation to the highly localized signaling events often seen natively. Fortunately, the ready accessibility of plasma membrane to cytoplasmic proteins has allowed the development of a large number of CID probes targeted to this region.
Researchers have used CID to look at effects of the important second messenger phosphatidylinositol 4,5-bisphosphate (PI(4,5)P2). Suh et al.  targeted Inp54p and PI(4)P 5-kinase to the plasma membrane to deplete or enhance, respectively, plasma membrane concentration of PI(4,5)P2. The potassium channel KCNQ was rapidly inactivated by depletion of PI(4,5)P2, even in the absence of other second messengers, and could be rapidly activated by increasing PI(4,5)P2. Importantly, when a PI 3-kinase was used to increase PI(3,4,5)P3 levels, no effect was seen on current through KCNQ channels, demonstrating once again the specificity of phosphoinositide effects. Another key study was performed by Varnai et al. . PI(4,5)P2 was known to be a crucial molecule in the plasma membrane. PI(4,5)P2 was known to regulate many important proteins such as ion channels, phospholipases C and D, and also as the source of the second messengers diacylglycerol and inositol 1,4,5-triphosphate. A type IV 5-phosphatase domain was fused to FKBP, while plasma membrane-targeted FRB served as the anchor. Prior to addition of rapamycin, the 5-phosphatase fusion is in the cytoplasm, far from its target substrate, and thus effectively inactive. Addition of rapamycin induced rapid translocation to the PM and dephosphorylation of PI(4,5)P2. This had a number of effects, including loss of ATP-induced Ca2+ influx, decreased activity of TRPM8 channels, and loss of receptor-mediated endocytosis (assayed by transferrin receptor uptake). As the depletion of PI(4,5)P2 is rapid, and specific at the plasma membrane, these effects can be presumed to be directly due to PI(4,5)P2 depletion, in contrast to previous results relying on long-term disruption of PI(4,5)P2 levels which could cause secondary effects and were thus potentially difficult to interpret. This set the stage for further studies to characterize the effects of rapid gain or loss of specific phosphoinositides.
A different 5-phosphatase was used for several subsequent studies of PI(4,5)P2 depletion and its effect on endocytosis. Zoncu et al.  reported that rapid loss of PI(4,5)P2 caused disappearance of clathrin-coated pits. A follow-up study by Abe et al. , in contrast, found only a modest effect on clathrin assembly, although they did see serious disruption in receptor-mediated endocytosis. Specifically, transferrin receptor recycling was decreased, as was LDL receptor endocytosis, while plasma membrane localization of AP-2, epsin, and CALM adapters was disrupted.
Rahdar et al.  examined the plasma membrane binding of the important tumor suppressor gene PTEN. Depletion of PI(4,5)P2 in the plasma membrane by Inp54p resulted in rapid loss of PTEN binding to the inner leaflet of the plasma membrane. Finally, Inp54p was used in conjunction with a pharmacologic inhibitor of PI 3-kinase to test the plasma membrane phospholipid-binding specificity of proteins with polybasic clusters. Both PI(4,5)P2 and PI(3,4,5)P3 depletion was required for loss of membrane targeting of proteins with polybasic motifs, demonstrating that both phosphoinositides are used in targeting of these proteins to the plasma membrane.
The plasma membrane is not the only target for specific phosphoinositide depletion. Fili et al.  studied Rab5a-positive endosomes. With Rab5a-2xFKBP as the anchor, FRB-myotubularin phosphatase 1 (MTM1) could be recruited specifically to endosomes, where the MTM1 degraded PI3P specifically. This loss of PI3P resulted in delays in endosome maturation, giving a tubularized structure, and prevented recycling of transferrin receptor. EGF receptor cycling was delayed but not entirely lost. The crucial importance of specific phosphoinositides was thus shown.
Small GTPases as targets of CID
Small GTPase molecules have proved another major class of cellular molecules whose interactions can be teased apart by careful use of CID systems, in addition to such standard techniques as dominant negative and constitutively active mutants or pharmacological inhibition. One early such study, by Castellano et al. , examined the GTPase Rac1 in a model of phagocytosis. A cell-surface receptor, CD25, was fused to FKBP and introduced into cells along with FRB-tagged constitutively active Rac1. Latex beads were used to cluster the CD25 receptors, and thus lead to concentrated patches of Rac1 upon rapamycin stimulation. Recruitment of the active Rac1 caused phagocytosis of the CD25-bound latex beads in an actin-dependent fashion.
The involvement of the small GTPase HRas in neuronal development was also studied using CID by Fivaz et al. . During neuronal development, polarity develops via a process of symmetry breaking resulting in one neurite being selected to grow rapidly to form the axon. The small GTPase HRas was known to activate PI3K in the developing neuron, and the hypothesis was formed that the two form a positive feedback loop. CID was used to target a cytosolic PI3K to the plasma membrane, which resulted in increased HRas activation (as seen by increase in the GTP-bound form). The precise mechanism of this feedback was not determined, although it is possible that the enhanced production of PI(3,4,5)P3 at the plasma membrane as a result of PI3K activity may serve as a signal to HRas.
CID also proved useful in testing a positive feedback model of the polarization of migrating neutrophils . According to the model, PI3K (activated by a chemoattractant), acting through a downstream effector Rac (a small GTPase), induces actin remodeling; the actin remodeling then feeds back to the PI3K, creating a positive feedback loop. CID was used to bring PI3K to the plasma membrane in the absence of chemoattractant, and neutrophils showed rapid PI(3,4,5)P3 production, actin remodeling, and polarization. When Rac was directly activated, however (using the guanine nucleotide exchange factor Tiam1), there was no PI(3,4,5)P3 production or polarization observed. This, combined with pharmacologic data, suggested the existence of some other input besides chemoattractant signaling through G-protein-coupled receptor (GPCR) is necessary, in what the authors compared to an AND-gate. These CID results have supported the compartmentalization model of GTPase signaling, whereby the compartment- and time-specific activation of GTPases is key to their signaling specificity . For example, activation of Ras specifically at the plasma membrane can induce membrane ruffling, while activation of Ras at the Golgi has no effect on cell morphology . This compartmentalization is in agreement with earlier models of Ras and MAPK signaling pathways [17, 19].
An interesting early use of CID by Graef et al.  was to study requirements of membrane recruitment and orientation in activation of the non-receptor tyrosine kinase ZAP70. Recruitment of ZAP70 to the plasma membrane led to its phosphorylation and the activation of Ras/MAPK and Ca2+/calcineurin signaling. A series of different synthetic dimerizers (rapalogs) was also used in bringing ZAP70 to the membrane that would provide different conformations and different degrees of rotational freedom for the ZAP70 relative to the plasma membrane. It was found that merely bringing ZAP70 to the PM was not in itself sufficient to induce its activity. Rapamycin, which locks FKBP-FRB into a rigid structure, did not lead to activation of ZAP70 although it did bring it to the plasma membrane; only those rapalogs that allowed greater conformational freedom led to activation. Additionally, this study was the first report that the SH2 domain was involved only in membrane binding and not kinase activity, as the entire SH2 domain could be replaced with the CID system without affecting kinase activity.
Membrane recruitment of another kinase, Akt, was performed by Li et al.  to uncover the importance of membrane localization in the activation of this important anti-apoptotic factor. The PH domain of Akt was replaced with FKBP, which could then be recruited to a myristoylated FRB. Recruitment of the recombinant Akt to plasma membrane caused its phosphorylation and the activation of its downstream target NF-κB. The true functional ability of this reconstituted system was proved by exposing Jurkat cells with the induced Akt to apoptotic factors including staurosporine, anti-Fas antibody, and the DNA damage induced by etoposide.
Another early study was done by Terrillon and Bouvier , who studied the role of β-arrestin2 in GPCR signaling. It was known that β-arrestin could associate with activated GPCRs, and that these β-arrestin-bound receptors were then internalized, but it was not clear whether β-arrestin bonding was sufficient to induce receptor uptake, or whether some additional signal was sent by the active (ligand-bound) receptor. By using CID, Terrillon and Bouvier were able to force β-arrestin to associate with two different vasopressin receptors, V2R or V1aR. Association of β-arrestin to the receptors was sufficient to induce their endocytosis in the absence of ligand, demonstrating that β-arrestin was acting as a signaling molecule for receptor uptake. Interestingly, when FRB-β-arrestin2 was recruited directly to PM via a myristoylated FKBP, this was sufficient to induce signaling leading to ERK1/2 phosphorylation, even without the involvement of GPCR receptors.
Stromal interaction molecule 1 (STIM1) is a Ca2+ sensor in the ER that plays an important role in Ca2+ release at ER-plasma membrane junctions. Luik et al.  used CID to induce oligomerization of STIM1 in the absence of Ca2+ signals. Oligomerization of STIM1 led to STIM1 accumulation at ER-plasma membrane junctions, and this oligomerization was sufficient to drive Ca2+ entry through CRAC channels.
Another example of localization driving function was shown by Bingol et al. . The protein kinase CaMKIIα is abundant in the postsynaptic density of neurons, and was known to be important in synaptic plasticity. Its extremely high concentration in dendritic shafts led to speculation that it might be playing a scaffolding role. Using PSD95 as the anchor in a CID system, CaMKIIα was induced to translocate to postsynaptic sites. This localization of CaMKIIα was shown to drive recruitment of proteasomes to the postsynaptic density. This finding is consistent with the involvement of CaMKIIα in synaptic plasticity, as proteasomes are expected to be present at a site of extensive cellular remodeling, but this requirement would have been difficult to determine directly without the use of CID.
A different strategy was used by Umeda et al. . Modification of rapamycin at the C40 group was again performed, but this time, the caging group includes biotin, which has an extremely tight binding affinity for the protein streptavidin (Fig. 2b). The caged rapamycin–biotin–streptavidin conjugate (cRb-A) is excluded from entry into cells due to the bulky streptavidin protein, which does not cross the plasma membrane. UV irradiation releases the biotin–streptavidin conjugate, leaving free rapamycin which can then freely diffuse into the cell to induce CID. Importantly, highly localized spatial control was demonstrated: irradiation of a small portion of a cell was able to induce localization of a Tiam1 fusion protein (as assayed by induction of membrane ruffling) in only the irradiated quadrant of the cell, with very little spill-over into adjacent regions of the cell. This high specificity is a valuable asset, though unfortunately, due to the nature of the caging moiety, it is only useful if the protein of interest is active at or very near the plasma membrane.
It would be useful to have multiple, orthogonal dimerization systems available that could potentially be used in the same cell simultaneously to generate complex outputs. Two different groups have recently published novel CID systems that rely on plant hormones and their cognate binding proteins. Liang et al.  utilized the plant hormone S-(+)-abscisic acid (ABA) and its binding proteins PYL1 and ABI1 (Fig. 2c). When utilized in a manner equivalent to rapamycin and FKBP-FRB, the ABA system was able to induce protein localization in mammalian cells. The ABA system was shown to be orthogonal to the rapamycin system, as both could be used simultaneously in the same cells without interference. Miyamoto et al.  utilized a different phytohormone, gibberellic acid 3 (GA3) and its cognate binding proteins GAI and GID1 (Fig. 2d). The GA3 system was also fully orthogonal to rapamycin in mammalian cells; this was exploited to produce simple proof-of-concept AND and OR logic gates using GA3 and rapamycin as inputs. Combination of one of the plant hormone-derived CID systems with rapamycin CID should allow the ability to probe multiple portions of signal transduction pathways simultaneously.
One application of CID is to cross-link different membrane-bound organelles via FRB and FKBP that are tethered specifically to different organelles in the same cell (Fig. 3b, c) . As this cross-linking requires that the organelles in question must already come into very close proximity, this is useful for studying cases where membranes from different organelles are naturally tethered already. This was utilized by Csordas et al.  to study the ER–mitochondrial junction, which is important for recruitment of Ca2+ stores from ER into mitochondrial matrix. A similar strategy was also used  to examine the ER–plasma membrane junction during Ca2+ release. Variation of linker sizes revealed an unsuspected requirement for some space between the ER protein STIM1 and the plasma membrane Ca2+ channel Orai1, implying that a previously unknown protein complex must bridge these two components. This result would have been difficult to achieve without the use of a CID to stabilize the relatively transient interaction between the two membranes.
Many proteins are only active at a specific cellular location, and can thus be inactivated by moving them to a different location within the cell. Haruki et al.  developed a CID system they dubbed the “Anchor-Away” technique to rapidly inactivate nuclear proteins (Fig. 3d). First, they used FKBP-fused plasma membrane proteins as the “anchor” to verify that FRB-tagged proteins that shuttle between nucleus and cytoplasm could be sequestered by rapamycin addition. Next, taking advantage of the continuous flow of ribosomal proteins into and out of the nucleus, they fused FKBP to a ribosomal protein (the “anchor”) and FRB to various nuclear proteins of interest. Addition of rapamycin would be expected to deplete the targeted gene from the nucleus, which they confirmed for over 40 different target genes involved in transcription, nuclear transport, and chromosome structure. Work by Robinson et al.  utilized mitochondrial-targeted FRB as the anchor and adaptor protein (AP)-complex proteins from clathrin-coated vesicles as the target. The ability to rapidly (seconds to minutes timescale) deplete AP-1 was used to show that AP-1 is involved in retrograde transport only, a point that had been unclear in previous studies relying on much more slowly-acting knockdown effects.
Unfortunately, when the POI in a CID system is an enzyme that is active at a particular membrane, there is a problem of background activity by the enzyme prior to localization to the target site. This is particularly likely when the POI is floating free in the cytoplasm and can thus come transiently in contact with the target membrane prior to the induction of CID. One recent technique resolves this problem by initially sequestering the POI to one organelle (specifically, the Golgi apparatus) and then upon addition of dimerizer sending the POI to its target at the plasma membrane (Fig. 3e) . As the initial localization to the outer surface of the Golgi membrane is via binding of an FAPP(PH) domain to PI(4)P, which is relatively weak relative to the FRB-rapamycin-FKBP interaction, the POI is targeted to the plasma membrane without inducing cross-linking of plasma membrane and Golgi. Further developments should increase the repertoire of initial sequestration sites and final targets.
One final new technique, dubbed “liberation”, is designed to get around the limitation of simply targeting phosphatases or kinases to increase or decrease specific membrane lipid concentrations (Fig. 3f) . Rapid generation of PI(4,5)P2 at the plasma membrane can be achieved by using CID to target a PI(4)P 5-kinase to the plasma membrane; however, in addition to increasing PI(4,5)P2 levels, this inevitably also depletes the pool of PI(4)P, which may have unintended consequences. In the “liberation” technique, a PH domain from PLCδ is fused to FKBP. The PH domain specifically binds to PM PI(4,5)P2, essentially sequestering it from accessibility to other interacting proteins. An FRB is co-expressed that is mitochondria-anchored. Addition of rapamycin rapidly induces translocation of the PLCδ domain to mitochondria, “liberating” the PI(4,5)P2 at the plasma membrane without affecting the level of PI(4)P or other metabolites. Indeed, while increased PI(4,5)P2 synthesis (and concomitant PI(4)P loss) causes formation of actin comets, PI(4,5)P2 liberation (without affecting PI(4)P level) induces membrane ruffling but no actin comet formation. These two different signaling pathways would have remained entangled without the use of this new technology to separate the effects of two different lipid second messengers.
CID is a robust tool for studying proteins in a highly subcellular localization specific fashion and in a highly temporally-limited manner (on a timescale of seconds). This has allowed us to get a handle on the “signaling paradox” to understand how cells are able to generate a large variety of signals using a relatively small number of fundamental signaling molecules. While CID has been most spectacularly used to study signaling pathways, particularly those involving membrane lipids and small GTPases, it has also proved useful in many other systems. Recent advances allow even greater specificity of localization or activation and allow the control of orthogonal signals at the same time. CID is expected to remain a major tool in the cell biologist’s toolbox for many years to come. It is hoped that readers will be inspired to consider how CID may be useful in their own research.
The authors’ research was supported by NIH grants GM092930 to T.I. We regret that due to length considerations, we could not discuss many worthwhile papers that have appeared in the literature over the years.