A genetically encoded toolkit of functionalized nanobodies against fluorescent proteins for visualizing and manipulating intracellular signalling

Background Intrabodies enable targeting of proteins in live cells, but generating specific intrabodies against the thousands of proteins in a proteome poses a challenge. We leverage the widespread availability of fluorescently labelled proteins to visualize and manipulate intracellular signalling pathways in live cells by using nanobodies targeting fluorescent protein tags. Results We generated a toolkit of plasmids encoding nanobodies against red and green fluorescent proteins (RFP and GFP variants), fused to functional modules. These include fluorescent sensors for visualization of Ca2+, H+ and ATP/ADP dynamics; oligomerising or heterodimerising modules that allow recruitment or sequestration of proteins and identification of membrane contact sites between organelles; SNAP tags that allow labelling with fluorescent dyes and targeted chromophore-assisted light inactivation; and nanobodies targeted to lumenal sub-compartments of the secretory pathway. We also developed two methods for crosslinking tagged proteins: a dimeric nanobody, and RFP-targeting and GFP-targeting nanobodies fused to complementary hetero-dimerizing domains. We show various applications of the toolkit and demonstrate, for example, that IP3 receptors deliver Ca2+ to the outer membrane of only a subset of mitochondria and that only one or two sites on a mitochondrion form membrane contacts with the plasma membrane. Conclusions This toolkit greatly expands the utility of intrabodies and will enable a range of approaches for studying and manipulating cell signalling in live cells. Electronic supplementary material The online version of this article (10.1186/s12915-019-0662-4) contains supplementary material, which is available to authorized users.


Background
Visualizing the location of specific proteins within cells and manipulating their function is crucial for understanding cell biology. Antibodies can define protein locations in fixed and permeabilized cells, but antibodies are large protein complexes that are difficult to introduce into live cells [1]. This limits their ability to interrogate the dynamics or affect the function of proteins in live cells. Small protein-based binders, including nanobodies derived from the variable region of the heavy chains (V HH ) of camelid antibodies, offer a promising alternative [2]. Nanobodies can be encoded by plasmids and expressed in live cells. However, generating nanobodies against thousands of protein variants is daunting, and even for single targets, it can be time-consuming, costly and not always successful. A solution to this bottleneck is provided by fluorescently tagged proteins, which are widely used in cell biology [3,4] after heterologous expression of proteins or gene editing of endogenous proteins [5][6][7]. The most common application of fluorescent protein (FP) tags is to visualize protein locations, but they have additional potential as generic affinity tags to manipulate and visualize protein functions in live cells. These opportunities are under-developed.
Proteins tagged with red fluorescent proteins (RFPs) such as DsRed, mRFP and mCherry (mCh) are also popular. Extensive optimization has made them attractive tags [3,18], and libraries of RFP-tagged proteins have been developed in mouse stem cells [19] and yeast [14].
Nanobodies that bind to RFP [20,21] or GFP [21,22] are most commonly used in their purified forms for immunoprecipitation and immunocytochemistry. However, they also offer a generic means of targeting in live cells the huge variety of available tagged proteins and the many emerging examples of endogenous proteins tagged with FPs by gene editing. GFP-targeting nanobodies have been used for applications such as targeted proteasomal degradation [23,24] and relocation of proteins in cells [25], but these and other applications are less developed for RFP-targeting nanobodies.
Here we develop a plasmid-encoded toolkit of nanobodies that bind common FP tags, including RFPs, CFP, GFP and YFP, fused to functional modules for visualization and manipulation of cell signalling (Fig. 1). We fused the nanobodies to a variety of functional modules: fluorescent sensors for Ca 2+ , H + and ATP/ ADP; optimized SNAP tags for labelling with bright and photostable dyes [26]; and hetero-dimerizing partners that allow inducible recruitment or sequestration of proteins and visualization of membrane contact sites (MCS) between organelles. We developed two methods to allow crosslinking of RFP-tagged and GFP-tagged proteins: a dimeric nanobody, and co-expression of RFP-targeting and GFP-targeting nanobodies fused to complementary hetero-dimerizing domains. We also describe functionalized nanobodies directed to lumenal sub-compartments of the secretory pathway. We demonstrate the utility of nanobody fusions by visualizing local Ca 2+ dynamics at the surface of mitochondria, by manipulating the locations of proteins and organelles within cells, by characterizing MCS between mitochondria and the plasma membrane (PM) and by targeting lumenal Ca 2+ sensors to a sub-compartment of the endoplasmic reticulum (ER).
This versatile toolkit of genetically encoded, functionalized nanobodies greatly expands the utility of RFP-and GFP-targeting nanobodies. It will provide a valuable resource for studying protein function and cell signalling in live cells. We illustrate some applications and demonstrate, for example, that IP 3 receptors deliver Ca 2+ to the outer membrane of only some mitochondria and that MCS between mitochondria and the plasma membrane occur at only one or two sites on each mitochondrion.

Results
Targeting RFP and GFP variants with genetically encoded nanobody fusions in live cells The RFP nanobody (RNb) and GFP nanobody (GNb) used are the previously described llama variants LaM4 and LaG16, respectively [21]. They were chosen for their favourable combinations of high affinity (K d values of 0.18 nM and 0.69 nM, respectively) and the ability to bind a variety of RFP or GFP variants [21]. The latter attribute maximizes their potential for targeting a wide variety of FPs [3,4]. LaM4 binds both mCh and DsRed variants, but not GFPs [21]. In addition to binding GFP, LaG16 binds cyan, blue and yellow FPs (CFP, BFP and YFP), but not RFPs [21]. In contrast, the widely used VhhGFP4 nanobody binds GFP, but not CFP [22].
In HeLa cells with organelles (ER, mitochondria, nucleus and lysosomes) labelled with mCh or mRFP markers, expression of RNb-GFP (Fig. 2a) specifically identified the labelled organelle (Fig. 2b). Similar results were obtained with GNb-mCh (Fig. 2c) and organelles (ER, mitochondria and nucleus) labelled with GFP or mTurquoise (Fig. 2d). These results demonstrate that plasmid-encoded RNb and GNb allow specific labelling of a variety of RFP and GFP variants in live cells.

Targeting sensors to RFP and GFP
The effects of intracellular messengers such as Ca 2+ [27], H + [28] and ATP/ADP [29] can be highly localized within cells. To enable visualization of these intracellular messengers in microdomains around RFP-tagged and GFP-tagged proteins, we fused RNb and GNb to fluorescent sensors for Ca 2+ [30], H + [31,32] or ATP/ADP [33].
In HeLa cells expressing TOM20-mCh or TOM20-GFP to identify the outer mitochondrial membrane (OMM), the RNb-Ca 2+ sensors ( Fig. 3 and Additional file 1: Figure S1) and GNb-Ca 2+ sensors (Fig. 4) were targeted to the OMM. Both families of targeted sensor reported an increase in [Ca 2+ ] after treatment with the Ca 2+ ionophore, ionomycin (Figs. 3 and 4 and Additional file 1: Figure S1). This confirms the ability of the sensors to report [Ca 2+ ] changes when targeted to the OMM microdomain.
In some cells, the targeted Nb-Ca 2+ sensors revealed local changes in [Ca 2+ ] c after receptor stimulation with histamine, which stimulates IP 3 formation and Ca 2+ Fig. 1 Nanobody fusions for visualizing and manipulating intracellular signalling. Plasmids were generated that encode nanobodies specific for GFP variants (GNb) or RFP variants (RNb), fused to functional modules. Nanobody fusions with an N-terminal signal sequence to target them to the secretory pathway are also shown (ssGNb and ssRNb) release from the ER in HeLa cells [34]. Imperfect targeting of the RNb-GGECO1.2 to the OMM allowed Ca 2+ signals at the surface of individual mitochondria to be distinguished from those in nearby cytosol in some cells ( Fig. 3d-f and Additional file 2: Video 1). In the example shown, RNb-GGECO1.2 at both the OMM and nearby cytosol responded to the large, global increases in [Ca 2+ ] evoked by ionomycin. However, cytosolic RNb-GGECO1.2 did not respond to histamine, while the sensor at the OMM responded with repetitive spiking (Fig. 3d-f and Additional file 2: Video 1). The GNb-LARGECO1.2 sensor, which has the lowest affinity for Ca 2+ of the sensors used, revealed changes in [Ca 2+ ] c at the surface of some mitochondria, but not others in the same cell ( Fig. 4d-f, Fig. 4h and Additional file 3: Video 2). In the example shown, GNb-LARGECO1.2 at the OMM in all mitochondria within the cell responded to the large, global increases in [Ca 2+ ] evoked by ionomycin. However, in response to histamine, mitochondria in the perinuclear region responded, but not those in peripheral regions ( Fig. 4d-f, Fig. 4h and Additional file 3: Video 2). Ca 2+ uptake by mitochondria affects many cellular responses, including mitochondrial metabolism, ATP production and apoptosis [35]; and Ca 2+ at the cytosolic face of the OMM regulates mitochondrial motility [36]. The subcellular heterogeneity of mitochondrial exposure to increased [Ca 2+ ] suggests that these responses may be very localized in cells.
These observations align with previous reports showing that Ca 2+ -mobilizing receptors evoke both oscillatory [Ca 2+ ] changes within the mitochondrial matrix [37], and large local increases in [Ca 2+ ] at the cytosolic face of the OMM [38]. Our results establish that nanobody-Ca 2+ -sensor fusions are functional and appropriately targeted and can be used to detect physiological changes in [Ca 2+ ] within cellular microdomains such as the OMM.
For targeted measurements of intracellular pH, RNb was fused to the green fluorescent pH sensor super-ecliptic pHluorin (SEpHluorin) [31], and GNb was fused to the red fluorescent pH sensor pHuji [32]. Both Nb-pH sensors were targeted to the OMM by the appropriate fluorescent tags, where they responded to changes in intracellular pH imposed by altering extracellular pH in the presence of the H + /K + ionophore nigericin (Fig. 5).
For targeted measurements of ATP/ADP, RNb was fused to the excitation-ratiometric ATP/ADP sensor Perceval-HR [33]. RNb-Perceval-HR was targeted to the surface of mitochondria and responded to inhibition of glycolysis and oxidative phosphorylation (Fig. 6).
Targeting SNAPf tags to RFP and GFP in live cells SNAP, and related tags, are versatile because a range of SNAP substrates, including some that are membranepermeant, can be used to attach different fluorophores or cargoes to the tag [39]. Purified GFP-targeting nanobodies fused to a SNAP-tag have been used to label fixed cells for optically demanding applications [40]. We extended this strategy to live cells using RNb and GNb fused to the optimized SNAPf tag [41] (Fig. 7a, b). In cells expressing the mitochondrial marker TOM20-mCh, RNb-SNAPf enabled labelling of mitochondria with the cell-permeable substrate SNAP-Cell 647-SiR and imaging at far-red wavelengths (Fig. 7c). In cells expressing lysosomal LAMP1-mCh and RNb-SNAPf, SNAP-Cell 647-SiR instead labelled lysosomes (Fig. 7d), demonstrating that SNAP-Cell 647-SiR specifically labelled the organelles targeted by RNb-SNAPf. Similar targeting of SNAP-Cell 647-SiR to mitochondria (Fig. 7e) and lysosomes ( Fig. 7f ) was achieved by GNb-SNAPf co-expressed with the appropriate GFP-tagged organelle markers.
Chromophore-assisted light inactivation (CALI) can inactivate proteins or organelles by exciting fluorophores attached to them that locally generate damaging reactive superoxide. Historically, antibodies were used to direct a photosensitizer to its target, but the fusion of fluorescent proteins or SNAP-tags to proteins of interest is now widely used [42]. RNb-SNAPf and GNb-SNAPf make the SNAP strategy more broadly applicable to CALI applications. We demonstrate this by targeting CALI to the outer surface of lysosomes. We anticipated that CALI in this microdomain might, amongst other effects, disrupt the motility of lysosomes, which depends on their association with molecular motors [43]. RNb-SNAPf enabled labelling of lysosomes with the CALI probe fluorescein, using the cell-permeable substrate, SNAP-Cellfluorescein ( Fig. 8a, b). Exposure to blue light then immobilized the lysosomes ( Targeting H + sensors to RFP-tagged and GFP-tagged proteins. a Schematic of RNb fused to the pH sensor super-ecliptic pHluorin (RNb-SEpH) and bound to RFP. b Schematic of GNb-pHuji binding to RFP. c, d HeLa cells co-expressing RNb-SEpH and TOM20-mCh were imaged in modified HBS (MHBS) using epifluorescence microscopy and exposed to extracellular pH 6.5 (c) or pH 8 (d) in the presence of nigericin (10 μM). Scale bars 10 μm. e, f HeLa cells co-expressing GNb-pHuji and TOM20-GFP were exposed to extracellular pH 6.5 (e) or pH 8 (f) in the presence of nigericin. Scale bars 10 μm. g, h Timecourse from single cells of the fluorescence changes (F/F 0 ) of mitochondrially targeted RNb-SEpH or GNb-pHuji evoked by the indicated manipulations of extracellular pH. Results shown are representative of 3 independent experiments indicating a loss of motor-driven motility. Control experiments demonstrated that labelling cytosolic SNAPf with SNAP-Cell-fluorescein (Additional file 1: Figure S2A and S2B) had significantly less effect on lysosomal motility after exposure to blue light ( Fig. 8f and Additional file 1: Figure  S2C-E). These results demonstrate that nanobody-SNAPf fusions allow targeting of fluorescent dyes in live cells, which can be used for re-colouring of tagged proteins or targeted CALI.

Sequestration of proteins tagged with RFP or GFP
The fusion of GFP nanobodies to degrons allows proteasomal degradation of GFP-tagged proteins [24], but the method is slow and cumbersome to reverse. An alternative strategy is to sequester tagged proteins so they cannot fulfill their normal functions. We used two strategies to achieve this: artificial clustering and recruitment to mitochondria. We induced artificial clustering by fusing RNb or GNb to a multimerizing protein (MP) comprising a dodecameric fragment of Ca 2+ -calmodulin-dependent protein kinase II (CaMKII) [44], with an intervening fluorescent protein (mRFP or mCerulean) for visualization of the Nb fusion (Fig. 9a, b). RNb-mCerulean-MP caused clustering of the ER transmembrane protein mCh-Sec61β (Fig. 9c, d) and caused lysosomes tagged with LAMP1-mCh to aggregate into abnormally large structures ( Fig. 9e, f ). GNb-mRFP-MP had the same clustering effect on lysosomes labeled with LAMP1-GFP ( Fig. 9g, h) and caused clustering of GFP-tagged proteins in the cytosol (calmodulin, Fig. 9i, j), nucleus and cytosol (p53, Fig. 9k, l) or ER membranes (IP 3 R3, Fig. 9m, n).
For inducible sequestration, sometimes known as 'knocksideways' [45], we used two approaches based on heterodimerizing modules, one chemical and one optical. First, we adapted the original knocksideways method, where proteins tagged with FKBP (FK506-binding protein) are recruited by rapamycin to proteins tagged with FRB (FKBP-rapamycin-binding domain) on the OMM, and thereby sequestered. The method has hitherto relied on individual proteins of interest being tagged with FKBP [45]. RNb-FKBP and GNb-FKBP (Fig. 10a, b) extend the method to any protein tagged with RFP or GFP. For our analyses, we expressed TOM70 (an OMM protein) linked to FRB through an intermediary fluorescent protein (GFP or mCh, to allow optical identification of the fusion protein). RNb-FKBP sequestered the ER transmembrane protein mCh-Sec61β at the OMM (TOM70-GFP-FRB) within seconds of adding rapamycin (Additional file 5: Video 4) and rapidly depleted mCh-Sec61β from the rest of the ER (Fig. 10c-e). After addition of rapamycin, GNb-FKBP rapidly sequestered endogenous IP 3 R1 tagged with GFP (GFP-IP 3 R1) [7] (Fig. 10f, g, and Additional file 6: Video 5) and cytosolic GFP-tagged calmodulin ( Fig. 10h and Additional file 7: Video 6) at mitochondria expressing TOM70-mCh-FRB. Rapamycin caused no sequestration in the absence of the nanobody fusions (Additional file 1: Figure S3).
To make sequestration reversible and optically activated, we adapted the light-oxygen-voltage-sensing domain (LOV2)/Zdark (zdk1) system in which light induces dissociation of LOV2-zdk1 hetero-dimers [46]. Because this system is operated by blue light at intensities lower than required for imaging GFP [46], it is most suitable for use with red fluorescent tags. RNb-zdk1 (Fig. 11a) sequestered cytosolic mCh on the OMM in cells expressing TOM20-LOV2, and blue laser light rapidly and reversibly redistributed mCh to the cytosol (Fig. 11b, c). , and imaged using TIRFM. c, d Cells were then exposed to 488-nm light for 3 s to induce CALI. TIRFM images show a representative cell at different times before (c) and after (d) CALI, with the image at t = 0 s shown in magenta and the image at t = 60 s in green. White in the merged images from the two different times indicates immobile lysosomes, while green and magenta indicate lysosomes that moved in the interval between images. Yellow boxes show regions enlarged in subsequent images. Scale bars 10 μm (main images) and 2.5 μm (enlargements). For clarity, images were auto-adjusted for brightness and contrast (ImageJ) to compensate for bleaching of mCh during tracking and CALI. e Effect of CALI on the displacements of individual lysosomes, determining by particle-tracking (TrackMate), during a 60-s recording from a representative cell (images taken every 1 s; mean values shown by bars). f Summary data (mean ± SEM, n = 6 cells from 6 independent experiments) show the mean fractional decrease in displacement (Δ displacement) due to CALI in cells expressing RNb-SNAPf or cytosolic SNAPf (see Additional file 1: Figure S2). The fractional decrease in displacement for each cell was defined as (MD pre -MD post )/MD pre , where MD pre and MD post are the mean displacement of all tracked particles in 60 s before and after CALI. * P < 0.05, unpaired Student's t test Inducible recruitment of tagged proteins to membrane contact sites The ability of Nb-FKBP fusions to recruit membrane proteins to FRB-tagged targets suggested an additional application: revealing contact sites between membrane-bound organelles. ER-mitochondrial membrane contact sites (MCS) have been much studied [47], but contacts between the PM and mitochondria, which are less extensive [48], have received less attention. In HeLa cells co-expressing the PM β 2 -adrenoceptor tagged with mCh (β 2 AR-mCh), TOM20-GFP-FRB and RNb-FKBP, rapamycin caused rapid recruitment of β 2 AR-mCh within the PM to mitochondria at discrete puncta that grew larger with time ( Fig. 12a-e and Additional file 8: Video 7). Recruitment was not seen in the absence of co-expressed RNb-FKBP (Fig. 12f). Rapamycin also triggered similar punctate were imaged using epifluorescence microscopy. g-n HeLa cells expressing GFP-tagged proteins in the absence (g, i, k, m) or presence (h, j, l, n) of co-expressed GNb-mRFP-MP were imaged using epifluorescence microscopy. Results are representative of at least 5 cells, from at least 3 independent experiments. Scale bars 10 μm accumulation of β 2 AR at mitochondria in COS-7 cells expressing β 2 AR-GFP, TOM20-mCh-FRB and GNb-FKBP (Additional file 1: Figure S4). In similar analyses of ER-mitochondria and PM-mitochondria MCS, the initial punctate co-localization of proteins was shown to report native MCS, which grew larger with time as rapamycin zipped the proteins together [48]. Our results are consistent with that interpretation. In most cases, β 2 AR were recruited to only one or two discrete sites on each mitochondrion, which expanded during prolonged incubation with rapamycin, but without the appearance of new sites ( Fig. 12d, e, and Additional file 1: Figure S4). Rapamycin had no evident effect on recruiting new mitochondria to the PM, but it did cause accumulation of tagged TOM70 at MCS and depletion of TOM70 from the rest of each mitochondrion, indicating mobility of TOM70 within the OMM (Additional file 1: Figure S4). Our results suggest that inducible crosslinking using RNb-FKBP or GNb-FKBP identifies native MCS between mitochondria and PM, with each mitochondrion forming only one or two MCS with the PM. We have not explored the functional consequences of these restricted MCS, but we speculate is shown before (c) and after (d) treatment with rapamycin (100 nM, 10 min). The boxed region is enlarged in subsequent images. Scale bars 10 μm (main images) and 2.5 μm (enlargements). e Timecourse of mCh-Sec61β fluorescence changes (F/F 0 ) evoked by rapamycin recorded at a representative mitochondrion and in nearby reticular ER. Results show~80% loss of fluorescence from the ER devoid of mitochondrial contacts. f, g HeLa cells co-expressing endogenously tagged GFP-IP 3 R1, GNb-FKBP and mitochondrial TOM70-mCh-FRB were imaged using TIRFM. A representative cell (n = 6) is shown before (f) and after (g) treatment with rapamycin (100 nM, 10 min). The boxed region is enlarged in subsequent images. Scale bars 10 μm (main images) and 2.5 μm (enlargements). h HeLa cells co-expressing GFP-calmodulin (GFP-CaM), GNb-FKBP and TOM20-mCh-FRB were imaged using epifluorescence microscopy. A representative cell (n = 3) is shown before and after treatment with rapamycin (100 nM, 10 min). The image for TOM-mCh-FRB is shown in the presence of rapamycin. Scale bar 10 μm that they may identify sites where proteins involved in communication between the PM and mitochondria are concentrated, facilitating, for example, phospholipid transfer [49], the generation of ATP microdomains [50], or Ca 2+ exchanges between mitochondria and store-operated Ca 2+ entry (SOCE) [51] or PM Ca 2+ -ATPases [52]. We next tested whether PM proteins could be recruited to the MCS between ER-PM that are important for SOCE and lipid transfer [53]. In response to rapamycin, mCh-Orai1, the PM Ca 2+ channel that mediates SOCE [54], was recruited by RNb-FKBP to ER-PM MCS labelled with the marker GFP-MAPPER-FRB [55] (Fig. 13a, b). Recruitment was not observed in the absence of RNb-FKBP (Fig. 13c). We conclude that the method identifies native ER-PM MCS during the initial phase of Nb recruitment, and the Nb subsequently exaggerates these MCS.
One of the least explored MCS is that between lysosomes and mitochondria [56]. Recent evidence shows that these MCS control the morphology of both organelles [57] and probably mediate the exchange of cholesterol and other metabolites between them [58]. We assessed whether the nanobody fusions could be used to inducibly recruit lysosomes to mitochondria. GNb-FKBP enabled recruitment of lysosomes labelled with LAMP1-GFP to mitochondria labelled with TOM20-mCh-FRB, in response to rapamycin (Fig. 14a-c). Lysosomes were not recruited to mitochondria in the absence of GNb-FKBP (Fig. 14d).

Crosslinking RFP-tagged and GFP-tagged proteins
We generated a dimeric nanobody (GNb-RNb) that binds simultaneously to GFP and RFP (Fig. 15a) and demonstrated its utility by crosslinking a variety of GFP-tagged and RFP-tagged proteins. Cytosolic GFP, normally diffusely distributed in the cytosol (data not shown), was recruited to nuclei by H2B-mCh (Fig. 15b) or to mitochondria by TOM20-mCh (Fig. 15c). In the presence of GNb-RNb, mCh-Orai1 and endogenously tagged GFP-IP 3 R1 formed large co-clusters (Fig. 15d) that differed markedly from the distributions of GFP-IP 3 R1 (Fig. 10f) and mCh-Orai1 (Fig. 13) in the absence of crosslinking. Consistent with earlier results (Fig. 12 and Additional file 1: Figure S4), β 2 AR-mCh, which is normally diffusely distributed in the PM, formed mitochondria-associated puncta when crosslinked to mitochondria expressing TOM20-GFP (Fig. 15e). Whole organelles could also be crosslinked. Co-expression of LAMP1-GFP and LAMP1-mCh labelled small, mobile lysosomes in control cells (Fig. 15f), while additional co-expression of GNb-RNb caused accumulation of lysosomes into large clusters (Fig. 15g).
This crosslinking of GFP and RFP was made rapidly inducible with an RNb-FRB fusion that hetero-dimerizes with GNb-FKBP in the presence of rapamycin (Fig. 16a). Co-expression of GNb-FKBP with RNb-FRB in cells A B C Fig. 11 Reversible optogenetic recruitment of RFP-tagged proteins using RNb-zdk1. a Schematic of RNb-zdk1 fusion bound to RFP, showing the reversible light-evoked dissociation of zdk1 from LOV2. b HeLa cells co-expressing RNb-zdk1, mitochondrial TOM20-LOV2 and cytosolic mCh were imaged using TIRFM. A representative cell is shown before and after one or five 1-s exposures to blue light (488-nm laser at 2-s intervals) and after a 3-min recovery period in the dark. Scale bar 10 μm. c Timecourse of the mCherry fluorescence changes (F/F 0 ) recorded at a representative mitochondrion and in nearby cytosol after each of the indicated light flashes. There is a reversible decrease (~60%) in mitochondrial mCh fluorescence and a corresponding reversible increase (~70%) in cytosolic fluorescence. A single measurement of mCh fluorescence was made at the end of a 3-min recovery period in the dark (REC) before further light flashes. Results are representative of 5 cells from 3 independent experiments co-expressing TOM20-GFP and mCh-Sec61β led to rapid colocalization of GFP and mCh after addition of rapamycin (Fig. 16b, c, and Additional file 9: Video 8). Similar results were obtained with RNb-FKBP and GNb-FRB (Additional file 1: Figure S5). We conclude that GNb-FKBP and RNb-FRB provide a rapidly inducible system for crosslinking any GFP-tagged protein to any RFP-tagged protein.

Targeting secretory compartments with lumenal nanobodies
GNb and RNb were directed to the lumen of the secretory pathway by addition of an N-terminal signal sequence, giving ssGNb and ssRNb. Targeting of ssGNb-mCh to the Golgi, ER network or ER-PM MCS was achieved by co-expression of organelle markers with lumenal FP tags (Fig. 17a, b). In each case, there was significant colocalization of green and red proteins. Similar targeting of ssRNb-GFP to the ER network or ER-PM MCS was achieved by co-expression with mCh-tagged lumenal markers of these organelles (Fig. 17c, d). These results demonstrate that ssGNb and ssRNb fusions can be directed to the lumen of specific compartments of the secretory pathway.
Fluorescent Ca 2+ sensors targeted to the lumen of the entire ER [59,60] are widely used and have considerably advanced our understanding of Ca 2+ signalling [61,62]. Fluorescent Ca 2+ sensors targeted to ER sub-compartments A B C D E F Fig. 12 Recruitment of proteins to native PM-mitochondria MCS using RNb-FKBP. a Schematic of RNb-FKBP fusion bound to RFP. b, c HeLa cells co-expressing RNb-FKBP, mitochondrial TOM70-GFP-FRB and β 2 AR-mCh were imaged using TIRFM before (b) and after (c) treatment with rapamycin (100 nM, 10 min). Scale bar 10 μm. d, e Enlarged images from C of the yellow box (d) and cyan box (e) show punctate recruitment of β 2 AR-mCh to individual mitochondria at the indicated times after addition of rapamycin. Scale bars 1.25 μm. f TIRFM images of HeLa cells co-expressing mitochondrial TOM70-GFP-FRB and β 2 AR-mCh in the presence of rapamycin (100 nM, 10 min) show no recruitment in the absence of co-expressed RNb-FKBP. The yellow box shows a region enlarged in the subsequent image. Scale bars 10 μm (main images) and 2.5 μm (enlargement). Results (b-f) are representative of 5 independent experiments and the secretory pathway have received less attention but have, for example, been described for the Golgi [63,64]. Our nanobody methods suggest a generic approach for selective targeting of lumenal Ca 2+ indicators. Fusion of ssRNb to GCEPIA1 or GEMCEPIA [60] provided ssRNb-GCEPIA1 and ssRNb-GEMCEPIA (Fig. 18a). These fusions were targeted to the lumenal aspect of ER-PM junctions by co-expression with mCh-MAPPER [7] (Fig. 18c, d). Fusion of ssGNb to the low-affinity Ca 2+ sensors LAR-GECO1 [59] or RCEPIA1 [60] provided ssGNb-LARGECO1 and ssGNb-RCEPIA1 (Fig. 18b). These fusions allowed targeting to ER-PM junctions labelled with GFP-MAPPER (Fig. 18e, f ). The targeted Ca 2+ sensors responded appropriately to emptying of intracellular Ca 2+ stores by addition of ionomycin in Ca 2+ -free medium (Fig. 18g-k). These results confirm that Ca 2+ sensors targeted to a physiologically important ER sub-compartment, the ER-PM junctions where SOCE occurs, report changes in lumenal [Ca 2+ ]. Our results demonstrate that nanobody fusions can be targeted to lumenal sub-compartments of the secretory pathway and they can report [Ca 2+ ] within physiologically important components of the ER.

Discussion
The spatial organization of the cell interior influences all cellular activities, and it is a recurrent theme in intracellular signalling [65,66]. Hence, tools that can visualize and manipulate the spatial organization of intracellular components are likely to find widespread application. We introduce a toolkit of plasmids encoding functionalized nanobodies against common FP tags, including CFP, GFP, YFP and RFPs (Fig. 1). Use of this toolkit is supported by genome-wide collections of plasmids, cells and organisms A B C Fig. 13 Recruitment of PM proteins to ER-PM MCS using RNb-FKBP. a Schematic of RNb-FKBP fusion bound to RFP. b HeLa cells co-expressing RNb-FKBP, mCh-Orai1 and the ER-PM junction marker GFP-MAPPER-FRB were imaged using TIRFM. A representative cell (n = 5) is shown before (top row) and after (bottom row) treatment with rapamycin (100 nM, 10 min). The boxed region is shown enlarged in subsequent images. c HeLa cells co-expressing mCh-Orai1 and GFP-MAPPER-FRB alone were imaged using TIRFM. A representative cell (n = 3) is shown before (top row) and after (bottom row) treatment with rapamycin (100 nM, 10 min). The boxed region is shown enlarged in subsequent images. The results show no recruitment in the absence of co-expressed RNb-FKBP. Scale bars (b, c) 10 μm (main images) and 2.5 μm (enlargements) expressing proteins tagged with GFP and RFP [10][11][12][13][14][15][16][17]19] and by facile methods for heterologous expression of tagged proteins or editing of endogenous genes to encode FP tags [5,6]. The functionalized nanobodies provide new approaches to studying intracellular signalling in live cells.
Our toolkit expands the repertoire of functionalized RFP-binding nanobodies, which are less developed than their GFP-binding counterparts [67]. The RNb fusions provide new opportunities to use RFP, which often has advantages over GFP. For example, RFP is spectrally independent from blue-green sensors, which are usually superior to their red counterparts [30,32]; from the CALI probe, fluorescein; and from optogenetic modules, which are often operated by blue-green light [68].
Relative to direct fusions of sensors to proteins of interest, nanobody-sensor fusions have several advantages. Firstly, the generic nanobody toolkit (Fig. 1) can be combined with collections of FP-tagged proteins to provide many combinations; each would otherwise require the expression of a unique construct that may or may not function normally. Secondly, each sensor is attached to the same entity (nanobody), which binds to the same partner (FP). Since the biophysical and biochemical properties of sensors may be influenced by their fusion partners, this A B C D Fig. 14 Inducible recruitment of lysosomes to mitochondria using GNb-FKBP. a Schematic of GNb-FKBP fusion bound to GFP. b HeLa cells coexpressing mitochondrial TOM70-mCh-FRB (magenta), lysosomal LAMP1-GFP (green) and GNb-FKBP were imaged using TIRFM. Merged images of a representative cell (n = 5) are shown before and at times after treatment with rapamycin (rapa, 100 nM). Scale bar 10 μm. c Enlargements of the boxed region in b. Scale bar 2.5 μm. d HeLa cells co-expressing TOM70-mCh-FRB (magenta) and lysosomal LAMP1-GFP (green) were imaged using TIRFM. A representative cell (n = 3) is shown before and after treatment with rapamycin (100 μm, 10 min); there is no recruitment in the absence of co-expressed GNb-FKBP. The yellow box shows a region enlarged in the subsequent image. Scale bars 10 μm (main images) and 2.5 μm (enlargement) provides greater confidence that sensors despatched to different locations will respond similarly to their analyte.
Nanobodies allow re-colouring of FPs with alternative fluorophores that may have advantageous properties. For example, re-colouring of RFP-tagged proteins with RNb-GFP (Fig. 2b) enables visualization of organelles with GFP, which has enhanced photophysical properties relative to RFPs. Nanobody-SNAPf fusions can be used to attach fluorescent dyes, including CALI probes and far-red fluorophores, to FP tags (Figs. 7 and 8). Longer excitation wavelengths cause less phototoxicity and allow greater penetration through tissue, which may be useful in studies of transgenic organisms and tissues. We also envisage live-cell applications in pulse-chase analyses and using super-resolution microscopy, Förster resonance energy transfer (FRET) and fluorescence lifetime imaging.
Crosslinking methods have many applications in cell biology, including stabilizing protein interactions (e.g. for pull-downs), identifying and manipulating MCS, enforcing protein interactions (e.g. receptor dimerization), redirecting proteins to different subcellular locations (e.g. knocksideways) and many more. Functionalized nanobodies provide many additional opportunities to regulate protein associations. The nanobody-FKBP/FRB fusions, for example, allow rapid rapamycin-mediated crosslinking of any pair of proteins tagged with GFP/RFP or tagged with either FP and any of the many proteins already tagged with FKBP or FRB [76] (Figs. 10, 12, 13, 14, 15, and  16). Nanobody-FKBP fusions may allow crosslinking to SNAP-tagged proteins [75], and the nanobody-SNAPf fusions to HaloTag-tagged proteins [74] and FKBP-tagged proteins [75]. RNb-zdk1 fusions allow photo-inducible crosslinking of RFP-tagged proteins to LOV-tagged proteins [46] (Fig. 11). Nanobodies that crosslink GFP-tagged proteins to RFP-tagged proteins (GNb-RNb; and the GNb-FKBP/RNb-FRB and GNb-FRB/RNb-FKBP pairings) may have the most applications, as they can take the fullest advantage of the numerous combinations of existing RFP and GFP-tagged proteins (Figs. 15 and 16).
Functionalized nanobodies directed to lumenal compartments of the secretory pathway would provide useful tools, but they are under-developed. Their potential is shown by nanobodies retained within the ER, which restrict onward trafficking of target proteins and inhibit their function [77]. We show that functionalized nanobodies, including nanobody-Ca 2+ sensors, can be directed to sub-compartments of the secretory pathway (Figs. 17  and 18). Lumenal Ca 2+ provides a reservoir within the ER, Golgi and lysosomes that can be released by physiological stimuli to generate cytosolic Ca 2+ signals [78,79]. Compartmentalization of Ca 2+ stores within the ER [63] and Golgi [79] adds to the complexity of lumenal Ca 2+ distribution in cells. Furthermore, lumenal Ca 2+ itself regulates diverse aspects of cell biology, including SOCE [54], sorting of cargo in the Golgi [80], binding of ERGIC-53 to cargoes within the ER-Golgi intermediate compartment (ERGIC) [81], and exocytosis of neurotransmitters by A B C D E F G Fig. 15 Crosslinking GFP-tagged and RFP-tagged proteins and organelles using GNb-RNb. a Schematic of GNb-RNb bound to GFP and RFP. b-e HeLa cells co-expressing the tagged proteins indicated with GNb-RNb were imaged using epifluorescence microscopy (b) or TIRFM (c-e). Representative cells (n = 5-7) are shown. Control images for GFP-IP 3 R1 are shown in Fig. 10 and Additional file 1: Figure S3. f, g HeLa cells co-expressing LAMP1-GFP and LAMP1-mCh in the absence (f) or presence (g) of co-expressed GNb-RNb were imaged using TIRFM. Representative cells (n = 5) are shown. Scale bars (b-g) 10 μm (main images) and 2.5 μm (enlargements of boxed areas) secretory vesicles [82,83]. Hence, there is a need for tools that can effectively report lumenal [Ca 2+ ] within this complex lumenal environment. The lumenal nanobody-Ca 2+ sensors detected changes in lumenal [Ca 2+ ] at the ER-PM MCS where SOCE occurs (Fig. 18).

Conclusions
We present a toolkit of plasmids encoding functionalized nanobodies directed against common fluorescent protein tags, which will allow a wide range of applications and new approaches to studying intracellular signalling in live cells. We illustrate some applications and demonstrate, for example, that IP 3 receptors deliver Ca 2+ to the OMM of only some mitochondria and that MCS between mitochondria and the plasma membrane occur at only one or two sites on each mitochondrion.

Materials
Human fibronectin was from Merck Millipore. Ionomycin was from Apollo Scientific (Stockport, UK). Rapamycin was from Cambridge Bioscience (Cambridge, UK). SNAP substrates were from New England Biolabs (Hitchin, UK). Other reagents, including histamine and nigericin, were from Sigma-Aldrich.
For CALI and LOV2/zdk1 experiments, the 488-nm laser in the upright position delivered an output at the objective of 2.45 mW (PM100A power meter, Thor Labs, Newton, NJ, USA). For CALI, a single flash of 488-nm laser illumination (3-s duration) was applied, with 10-ms exposures to 488-nm laser immediately before and after the CALI flash to allow imaging of SNAP-Cell-fluorescein (i.e. 3.02 s total CALI flash). For LOV2/zdk1 experiments, repeated flashes of 488-nm light (1-s duration each) were used at 2-s intervals to allow imaging with 561-nm laser illumination during the intervening periods.
Before analysis, all fluorescence images were corrected for background by subtraction of fluorescence detected from a region outside the cell. Image capture and processing used MetaMorph Microscopy Automation and Image Analysis Software (Molecular Devices) and Fiji [95]. Particle tracking used the TrackMate ImageJ plugin [96], with an estimated blob diameter of 17 pixels and a threshold of 5 pixels. Co-localization analysis used the JACoP ImageJ plugin [97]. Pearson's correlation coefficient (r) was used to quantify colocalization. We report r values only when the Costes' randomization-based colocalization value (P value = 100 after 100 iterations) confirmed the significance of the original colocalization. Where example images are shown, they are representative of at least three independent experiments (individual plates of cells from different transfections and days).

Statistics
Results are presented as mean ± SEM for particle-tracking analyses and mean ± SD for colocalization analyses, from n independent analyses (individual plates of cells from different transfections). Statistical comparisons used paired or unpaired Student's t tests, or analysis of variance with the Bonferroni correction used for multiple comparisons. *P < 0.05 was considered significant.

Additional files
Additional file 1: Figure S1. Targeting RNb-GEMGECO Ca 2+ sensor to RFP-tagged proteins. Figure S2. Targeting CALI to lysosomes with SNAP-Cell-fluorescein: cytosolic controls. Figure S3. Rapamycin alone does not recruit RFP-tagged or GFP-tagged proteins to mitochondria. Figure S4. Recruitment of proteins to native PM-mitochondria MCS using GNb-FKBP. Figure S5. Inducible crosslinking of RFP-tagged and GFP-tagged proteins with RNb-FKBP and GNb-FRB.  Fig. 4d), but not in peripheral regions (e.g. yellow box in Fig. 4d). Ionomycin (5 μM) was added at 3 min. The video was acquired at 1 Hz and is shown at 33 fps. The clock is in min:s. Relates to Fig. 4d-f. (MP4 2807 kb) Additional file 4: Video 3. Effect of targeted CALI on lysosomal motility. HeLa cells expressing LAMP1-mCh and RNb-SNAPf were imaged using TIRFM and 561-nm laser illumination before (top) and after (bottom) CALI (3.02 s exposure to 488-nm epifluorescence laser illumination). The video was acquired at 0.5 Hz and is shown at 3 fps. The clock is in min:s. Relates to Fig. 8. (MP4 776 kb) Additional file 5: Video 4 RNb-FKBP rapidly sequesters an ER integral membrane protein at the OMM. TIRFM images of HeLa cells expressing TOM70-GFP-FRB, RNb-FKBP and mCh-Sec61β were treated with rapamycin (100 nM, added at 60 s). The ER membrane protein, mCh-Sec61β, is then rapidly sequestered at the OMM. The video was acquired at 0.5 Hz and shown at 33 fps. The clock is in min:s. Relates to Fig. 10c, d. (MP4 551 kb) Additional file 6: Video 5. GNb-FKBP rapidly sequesters endogenously tagged GFP-IP 3 R1 at the OMM. TIRFM images show HeLa cells with endogenously GFP-tagged IP 3 R1 and transiently expressing TOM70-mCh-FRB and GNb-FKBP, and then treated with rapamycin (100 nM, added at 60 s). GFP-IP 3 R1 is rapidly sequestered at the OMM. The video was acquired at 0.5 Hz and is shown at 33 fps. The clock is in min:s. Relates to Fig. 10f and g. (MP4 2601 kb) Additional file 7: Video 6. GNb-FKBP rapidly sequesters GFP-CaM at the OMM. Epifluorescence microscopy images show HeLa cells transiently expressing GFP-CaM, GNb-FKBP and TOM20-mCh-FRB, and then treated with rapamycin (100 nM, added at 30 s). GFP-CaM is rapidly sequestered at the OMM. The video was acquired at 0.5 Hz and is shown at 9 fps. The clock is in min:s. Relates to Fig. 10h. (MP4 362 kb) Additional file 8: Video 7. RNb-FKBP recruits a PM protein to the OMM in response to rapamycin. TIRFM images of HeLa cells expressing TOM70-GFP-FRB, RNb-FKBP and the PM protein, β 2 AR-mCh, and then exposed to rapamycin (100 nM, added at 60 s). There is a rapid translocation of β 2 AR-mCh to the OMM. The video was acquired at 0.5 Hz and is shown at 33 fps. The clock is in min:s. Relates to Fig. 12b-e. (MP4 2141 kb) Additional file 9: Video 8. Crosslinking GNb-FKBP and RNb-FRB with rapamycin recruits mCh-Sec61β to TOM20-GFP in the OMM. HeLa cells expressing GNb-FKBP, RNb-FRB, mCh-Sec61β and TOM20-GFP were stimulated with rapamycin (100 nM, added at 100 s). The TIRFM images show rapid recruitment of mCh-Sec61β to the OMM. The video was acquired at 0.2 Hz and is shown at 8 fps. The clock is in min:s. Relates to