Chromobodies to Quantify Changes of Endogenous Protein Concentration in Living Cells*

Understanding cellular processes requires the determination of dynamic changes in the concentration of endogenous proteins. We demonstrate the dependency of the intracellular level of chromobodies (CB, fluorescently labeled nanobodies) on the amount of their endogenous antigens and present a broadly applicable strategy how to employ turnover-accelerating CBs, to quantify dynamic changes of endogenous protein levels by quantitative live-cell imaging. This will enable unprecedented insights into the dynamic regulation of proteins, e.g. during cellular signaling, cell differentiation, or upon drug action. Graphical Abstract Highlights Chromobodies are stabilized by antigen binding in live cells. Monitoring changes of endogenous protein levels in living cells with chromobodies. Broadly applicable system to generate turnover-accelerated chromobodies. Quantification of time- and dose-dependent compound effects. Understanding cellular processes requires the determination of dynamic changes in the concentration of genetically nonmodified, endogenous proteins, which, to date, is commonly accomplished by end-point assays in vitro. Molecular probes such as fluorescently labeled nanobodies (chromobodies, CBs) are powerful tools to visualize the dynamic subcellular localization of endogenous proteins in living cells. Here, we employed the dependence of intracellular levels of chromobodies on the amount of their endogenous antigens, a phenomenon, which we termed antigen-mediated CB stabilization (AMCBS), for simultaneous monitoring of time-resolved changes in the concentration and localization of native proteins. To improve the dynamic range of AMCBS we generated turnover-accelerated CBs and demonstrated their application in visualization and quantification of fast reversible changes in antigen concentration upon compound treatment by quantitative live-cell imaging. We expect that this broadly applicable strategy will enable unprecedented insights into the dynamic regulation of proteins, e.g. during cellular signaling, cell differentiation, or upon drug action.


In Brief
Understanding cellular processes requires the determination of dynamic changes in the concentration of endogenous proteins. We demonstrate the dependency of the intracellular level of chromobodies (CB, fluorescently labeled nanobodies) on the amount of their endogenous antigens and present a broadly applicable strategy how to employ turnover-accelerating CBs, to quantify dynamic changes of endogenous protein levels by quantitative live-cell imaging. This will enable unprecedented insights into the dynamic regulation of proteins, e.g. during cellular signaling, cell differentiation, or upon drug action.

Graphical Abstract
Several methods are available to detect changes in the concentration of specific proteins in biological samples. The rise of mass spectrometry (MS)-based analysis has enabled relative and absolute quantification of proteins with unprecedented sensitivity and accuracy (1,2). Considering that MS-based quantification requires expensive equipment and trained personnel, this technology is mostly applied for largescale proteomic analyses. Because of the ever-growing availability of specific antibodies, antibody-based techniques such as enzyme-linked immunosorbent assay (ELISA) or immunoblotting are commonly used to analyze relative concentration changes of single proteins of interest (POIs) 1 . However, the informative value of these methods is limited because only average protein amounts are determined and no intercellular resolution is provided. In addition, tracing changes in protein concentration over time is very laborious and time-consuming. Alternatively, immunofluorescence (IF) can be used to assess the subcellular localization and the relative concentration of POIs on single-cell level. Like any immunodetection method, this can suffer from inaccuracies based on batch-tobatch antibody variability, epitope inaccessibility and cross reactivity (3). In addition, because of cell fixation and permeabilization procedures, IF allows no direct analysis of dynamic changes (4). Considering that most cellular processes are dynamic in nature and rely on the spatiotemporal orchestration under native conditions, the assessment of time-dependent changes of endogenous protein levels within the physiological environment of living cells is preferable.
With the rise of genome editing techniques, fluorescent protein (FP) tagging of endogenous proteins provides a straightforward approach to optically monitor the relative amount of a POI (5). However, as repeatedly described, FP tagging can interfere with crucial protein parameters such as turnover, subcellular localization, and participation in multiprotein complexes (6 -8). During the last decade, intrabodies have emerged as beneficial tools to study the dynamic behavior of endogenous proteins in various cellular models. Because of their compact structure, small size, high stability and solubility, single-domain antibody fragments from camelids (VHH, nanobodies) possess many advantageous properties to be employed within living cells (9 -11). Acknowledging the potential of these binding molecules, numerous protocols and synthetic nanobody libraries for targeted selection of intracellularly functional nanobodies have been developed (12)(13)(14). By fusing nanobodies (NBs) to fluorescent proteins, so-called chromobodies (CBs) are generated. Upon cellular expression, they allow optical detection of endogenous proteins in live cells. With regard to imaging purposes transiently binding CBs addressing functionally inert epitopes are preferable to avoid unwanted effects on antigen mobility or by displacing natural interaction partners. To date, multiple target-specific CBs have been applied to visualize cellular processes in cultured cells and entire organisms without functional interference (15)(16)(17)(18)(19)(20).
Recently, we have generated a CB (BC1-TagGFP2), which specifically targets the soluble, nonmembrane-associated fraction of endogenous ␤-catenin (CTNNB1) without affecting its transcriptional activity. By live-cell imaging of cells stably expressing BC1-TagGFP2 we observed an increased CB signal along with an elevation of intracellular CTNNB1 level upon compound-mediated induction of the WNT/␤-catenin pathway (21). Notably, this was not because of altered transcription of the CB but attributed to a yet unexplained mechanism of antigen-mediated stabilization on protein level, which is in accordance to previous findings describing higher levels of bacterially injected NBs within the cytoplasm of mammalian cells in the presence of their cognate antigen (22).
Here, we demonstrate that antigen-mediated CB stabilization (AMCBS) is applicable for numerous CBs by showing this phenomenon for four different CBs targeting unrelated endogenous and nonendogenous antigens. To adapt CBs for monitoring changes in antigen concentration more precisely, we screened for N-terminal amino acids, which induce accelerated CB turnover. Based on our findings, we generated highly antigen-responsive CBs. As exemplarily shown for CTNNB1-specific CBs, stable chromobody cell lines allow visualization of rapid and reversible changes in the concentration of endogenous proteins upon compound treatment by quantitative live-cell imaging.

EXPERIMENTAL PROCEDURES
Expression Constructs-All oligonucleotide sequences used in this study for DNA amplification are listed in supplemental Table S1. All expression constructs and cell lines used in this study are listed in supplemental Table S2.
Intracellular Immunoprecipitation (IC-IP)-3 ϫ 10 6 HEK293T cells were transiently transfected with equal amounts of expression vectors encoding for GFP-CA, BC1-CB, destabilized versions of BC1-CB or eGFP. In case of detection of endogenous CTNNB1 8 h after plasmid transfection cells were treated with 10 M CHIR99021 for 16 h. Transfection efficiency was controlled by fluorescence microscopy and cells were harvested 24 h after transfection. Cell pellets were lysed as described and GFP-CA, BC1-CBs or eGFP were precipitated using the GFP-Trap (ChromoTek) according to the manufacturer's protocol. Input and bound fractions were subjected to SDS-PAGE followed by Western blot analysis using anti-RFP, anti-CTNNB1 and anti-GFP antibodies.

Cell Culture, Transfection, Stable Cell Line Generation and
Compound Treatment-The HeLa Kyoto cells (Cellosaurus no. CVCL_1922) were obtained from S. Narumiya (Kyoto University, Japan), and the HEK293T, A549, U2OS cell lines were obtained from ATCC, LGC Standards GmbH, Wesel, Germany (CRL3216, CCL-185, HTB-96). All cell lines were tested negative for mycoplasma using the PCR mycoplasma kit Venor GeM Classic (Minerva Biolabs GmbH, Berlin, Germany) and the TaqDNA polymerase (Minerva Biolabs). Because this study does not include cell line-specific analysis, all cell lines were used without additional authentication. All cell lines were cultivated according to standard protocols. Briefly, growth media containing DMEM (high glucose, pyruvate, ThermoFisher Scientific) for HeLa Kyoto, HEK293T and U2OS cells, or DMEM/F-12 (high glucose, pyruvate, ThermoFisher Scientific) for A549 cells supplemented with 10% (v/v) fetal bovine serum (FCS, ThermoFisher Scientific), L-glutamine (ThermoFisher Scientific) and penicillin-streptomycin (ThermoFisher Scientific) were used for cultivation. Cells were passaged using 0.05% trypsin-EDTA (ThermoFisher Scientific) and were cultivated at 37°C in a humidified chamber with a 5% CO 2 atmosphere. Plasmid DNA was transfected with Lipofectamine 2000 (ThermoFisher Scientific) in U2OS cells, Lipofectamine LTX (Thermo-Fisher Scientific) for A549 cells, or polyethylenimine (PEI, Sigma Aldrich Chemie GmbH, Munich, Germany) for HeLa Kyoto and HEK293T cells. Prior to transfection, cells were allowed to grow to ϳ70% confluency. To generate DNA/PEI complexes for transfection in a 96-well plate format 100 -200 ng DNA were mixed with 0.5 g PEI in 20 l serum-free medium. Transfection mixture was incubated for 15 min at room temperature and subsequently was added to the adherent cells. Transfections using Lipofectamine 2000 or Lipofectamine LTX were carried out according to the manufacturer's protocol. RNA interference-mediated knockdown of PCNA and vimentin was accomplished using Lipofectamine RNAiMax (Thermo-Fisher Scientific) according to manufacturer's instructions. Analysis of siRNA-mediated knockdown occurred 72 h post transfection. Stable HeLa_BC1-TagGFP2 cell line (21) was maintained in medium supplemented with 3 g/ml Blasticidine (Sigma) and stable A549_VB6-eGFP (19) with 80 g/ml Hygromycin (Carl Roth GmbH, Karlsruhe, Germany). Stable HeLa_CCC-TagRFP (18), stable HeLa_␣CA-mCherry (16), stable HeLa-Ub-R-BC1-eGFP were maintained in media supplemented with 0.5 mg/ml G418 (Roth). For the generation of the stable cell lines HeLa_Ub-R-BC1-eGFP and U2OS_Ub-R-BC1-eGFP plasmid transfection of the indicated chromobody construct was performed using Lipofectamine 2000 (ThermoFisher Scientific) according to the manufacturer's protocol. 24 h post transfection cells were subjected to a three-week selection period using 0.5 mg/ml G418 (Roth) followed by single cell separation. Single clones were analyzed regarding the level of Ub-R-BC1-eGFP expression, respectively. Compound treatment with 10 M CHIR99021 (Tocris) was performed for 16 h, with 100 g/ml cycloheximide (Sigma) for 6 h, with 5 ng/ml TGF-␤ (PeproTech Germany, Hamburg, Germany) for up to 72 h, with 10 M MG132 (Calbiochem) and 10 mM NH 4 Cl (Roth) for up to 10 h. Immunofluorescence-For immunofluorescence, HeLa cells (parental and stable CB-expressing cells) were seeded at 2000 -4000 cells per well in Clear 96-well plates (Greiner Bio-One GmbH, Frickenhausen, Germany) and cultured for 24 -48 h. For fixation, cells were washed twice with PBS and were fixed with 4% formaldehyde (v/v, PFA, AppliChem GmbH, Darmstadt, Germany) in PBS. Subsequently, cells were blocked and permeabilized using 5% bovine serum albumin (w/v, BSA, Roth) and 0.3% Triton X-100 (v/v, Roth) in PBS. Primary and fluorochrome-conjugated secondary antibodies were diluted in 1% BSA and 0.3% Triton X-100 in PBS and antibody staining was carried out according to standard procedures. Fluorescence images were acquired using MetaXpress Micro XL system (Molecular Devices, San Jose, CA) and 20ϫ or 40ϫ magnification.
Image Segmentation and Analysis-Depending on experimental setting and cell line ϳ1.5 ϫ 10 3 -1 ϫ 10 4 cells were plated per well in a black Clear 96-well plate (Greiner). Images were acquired with an ImageXpress micro XL system (Molecular Devices) and analyzed by MetaXpress software (64 bit, 6.2.3.733, Molecular Devices). Fluorescence images comprising a statistically relevant number of cells were acquired for each condition. For quantitative fluorescence analysis the mean fluorescence in a defined (segmented) area of interest (e.g. whole cell or nucleus) was determined. Using the Custom Module Editor (version 2.5.13.3) of the MetaXpress software, we established an image segmentation algorithm that identifies areas of interest based the parameters of size, shape, and fluorescence intensity above local background. For nuclear segmentation, nuclei in fixed cells were stained with 0.02 g/ml 4Ј,6-diamidino-2-phenylindole (DAPI, Sigma) whereas live cells were continuously incubated with 2 g/ml Hoechst33258 (Sigma). To segment the whole cell including the nucleus as well as the cytosolic compartment, the ectopically expressed antigen or its respective control was used to generate the corresponding segmentation mask (supplemental Fig. S1). The average fluorescence intensities in whole cells or nuclei were determined for each image followed by subtraction of background fluorescence. From these values the mean fluorescence and standard errors were calculated for three independent replicates and student's t test was used for statistical analysis.

RESULTS
Quantitative Image Analysis of Antigen-mediated Stabilization of a CTNNB1-specific CB-Previously, we have shown that an increase of endogenous CTNNB1 upon treatment with the GSK3-␤ inhibitor CHIR99021 (CHIR) was accompanied by rising levels of a CTNNB1-specific CB stably integrated and constitutively expressed in HeLa cells (cell line hereafter referred to as HeLa_BC1-TagGFP2) using immunoblot analysis (see Figure 11 of (21)). Here, we focused on an imaging-based readout and analyzed the level of both proteins in situ by immunofluorescence staining with a CTNNB1-specific antibody in combination with the detection of BC1-TagGFP2 fluorescence. Treatment with CHIR resulted in a strong CT-NNB1 signal increase in HeLa_BC1-TagGFP2 and parental HeLa cells. A corresponding increase in CB fluorescence was observed in HeLa_BC1-TagGFP2 (Fig. 1A). For a populationwide quantitative analysis encompassing hundreds of cells, we established an automated image segmentation algorithm (supplemental Fig. S1). Using this algorithm, we quantified fluorescence signals derived from the CTNNB1 antibody staining and BC1-TagGFP2 in individual cells. Quantification of the CTNNB1 signal in nontreated or CHIR-treated parental HeLa and HeLa_BC1-TagGFP2 cells revealed highly similar levels indicating that the presence of the BC1-CB does not affect endogenous levels of ␤-catenin (Fig. 1B, 1C). In favor of a more comprehensible data representation, we chose to depict the entirety of individual cell fluorescence as population-wide mean fluorescence. By this we determined a 4.6fold increase of CB fluorescence compared with a 19-fold increase of CTNNB1 during 16 h of CHIR treatment in HeLa_BC1-TagGFP2 cells (Fig. 1C, 1D).
Next, we tested whether BC1-TagGFP2 stabilization is also evident in the case of ectopically expressed CTNNB1. We transiently co-expressed parental HeLa cells with plasmids coding for BC1-TagGFP2, mCherry-CTNNB1 or mCherry as control (Fig. 1E). Although the expression of CB and antigen differed substantially between individual cells, the overall effect of AMCBS was evident as shown by a 1.7-fold increase of mean BC1-TagGFP2 fluorescence detected in mCherry-CTNNB1-expressing cells (Fig. 1F).
First, we transfected HeLa_␣CA-mCherry cells with constructs encoding either GFP-labeled p24 capsid protein (processed from GFP-gag-pol (26), for simplicity referred to as GFP-CA), or GFP as control. Whereas GFP-CA-positive cells displayed a markedly increased ␣CA-mCherry signal compared with nontransfected cells, no differences for ␣CA-mCherry fluorescence was observed in cells co-expressing GFP ( Fig. 2A). Moreover, in the absence of antigen, ␣CA-mCherry was homogenously distributed between cytoplasm and nucleus, whereas in GFP-CA-expressing cells the CB strongly co-localized with its target in the cytoplasm ( Fig. 2A). That this colocalization is likely because of intracellular binding of the ␣CA-mCherry to GFP-CA is supported by intracellular co-immunoprecipitation (supplemental Fig. S2). Using software-assisted image segmentation, we quantified ␣CA-mCherry fluorescence in nuclear, cytoplasmic or whole-cell area of GFP-or GFP-CA-expressing cells and found a 2.5-, 6.0-, or 4.3-fold increase, respectively (Fig. 2B). Quantitative immunoblot analysis revealed a 2.2-fold enrichment of ␣CA-mCherry (supplemental Fig. S3A). Considering a transfection efficiency of ϳ50% this is comparable to the microscopically determined values.
To test whether the CB level also responds to antigen depletion, we transfected HeLa_CCC-TagRFP cells with siRNAs targeting PCNA or control siRNAs and analyzed the fluorescence intensity derived from the CCC-TagRFP 72 h post transfection. Compared with control siRNAs, PCNA knockdown led to a ϳ70% reduction of CCC-TagRFP fluorescence (Fig. 2C, 2D). Accordingly, immunoblot analysis showed a strong reduction of CCC-TagRFP and PCNA (sup-  Fig. S3B). Finally, to test AMCBS in vimentin chromobody-expressing A549 cell lines (A549_VB6-eGFP, A549_VB6-TagRFP) (19), we analyzed whether TGF-␤-induced increase or siRNA-mediated depletion of VIM is accompanied by corresponding changes in CB fluorescence. Previously, we have shown by Western blotting that both treatments are highly effective to raise or deplete the cellular levels of endogenous vimentin in these cell models (see Fig. 4 of (19)). Here, quantitative live-cell imaging of A549_VB6-eGFP cells revealed a continuous increase of the CB signal upon TGF-␤ treatment with a ϳ4-fold elevation after 72 h (Fig.  2E, 2F). Conversely, we observed a reduction of the CB signal to ϳ20% upon siRNA-mediated knockdown of VIM compared with control siRNAs in A549_VB6-TagRFP cells (Fig.  2G, 2H). These data strongly support the hypothesis that antigen-mediated stabilization is a general phenomenon of numerous CBs, as we showed an antigen dependence for CBs targeting different cellular antigens. Notably, we were not only able to detect an increase but also a depletion of the corresponding target structure by monitoring the CB signal over time.
Optimization of CB Turnover-In the above-described cell models, CB expression is constitutively driven from a strong promoter (CMV) resulting in medium-to-high cellular accumulation of CBs. Because excessive CB expression might prevent the detection of small changes in antigen concentration, a low basal antigen-independent CB amount is desired. In this context, a set of destabilizing nanobody framework mutations were recently described to reduce the cellular amount of nonbound chromobodies (27). Thus, we asked whether the effect of the most destabilizing mutations (S70R, C92Y and S113F, according to Kabat numbering, (28)) is transferable to BC1-TagGFP2 (supplemental Fig. S4A). We generated corresponding expression constructs (BC1 S70R -TagGFP2, BC1 C92Y -TagGFP2, BC1 S113F -TagGFP2) and compared them to BC1-TagGFP2 upon co-expression with mCherry-CTNNB1 or mCherry (supplemental Fig. S4B). Analysis of CB fluorescence revealed a reduced mean fluorescence of the BC1 S70R -TagGFP2 and BC1 S113F -TagGFP2 mutants and a strong reduction for the BC1 C92Y -TagGFP2 mutant close to background fluorescence (supplemental Fig. S4C). Further quantification of the CB signals in mCherry-CTNNB1-or mCherryexpressing cells showed that antigen-dependent stabilization measured for all mutations is moderately increased compared with the nonmutated version (supplemental Fig. S4C). However, when we tested antigen binding of these modified chromobodies by intracellular immunoprecipitation of CTNNB1 (21), we could not detect any CTNNB1 in the bound fraction of the low expressing BC1 C92Y -TagGFP2 construct indicating that this framework mutation abolishes binding (supplemental Fig. S4D).
Although the framework mutation approach can improve antigen-dependent responsiveness, it also bears the risk of nonfunctional binding molecules. Therefore, we conceived a strategy to reduce ground levels of CBs, which is not expected to impair antigen binding. First, we tested whether CBs are degraded via the ubiquitin proteasome system (UPS) or lysosomal protein degradation. To analyze this in the absence of an antigen we chose the HeLa_␣CA-mCherry cell line, which lacks an endogenous binding partner of the ␣CA-CB. Following treatment of the cells either with the proteasome inhibitor MG132, or NH 4 Cl, which inhibits lysosomal degradation (29) we monitored CB fluorescence by time-lapse imaging for 10 h. The data revealed a clear increase in ␣CA-mCherry fluorescence after incubation with MG132, whereas NH 4 Cl-treated cells showed no changes in the CB signal (supplemental Fig. S5A, S5B). Considering the high similarity of sequence and structure of CBs we conclude that most CBs are likely degraded via UPS.
According to the N-end rule, one of the key determinants of protein half-life in the UPS is the N-terminal amino acid (30,31). This residue is recognized by E3 ubiquitin ligases, which initiate ubiquitinylation of accessible nearby lysine residues priming the protein for proteasomal degradation (Fig. 3A). To screen for N-terminal amino acid residues that confer accelerated CB degradation we implemented the ubiquitin fusion technique, which employs co-translational cleavage of ubiquitin from a fusion protein composed of an N-terminal ubiqwhereas the CB was directly detected using the TagGFP2 signal. For nuclear segmentation, cells were stained with DAPI. Scale bar: 50 m.  (30,32). This allowed us to generate CBs displaying any desired amino acid exposed at their N termini. To produce Ub-CB fusions, we designed an expression construct as outlined in Fig. 3B. To test whether this construct is processed as intended, we expressed the original BC1-TagGFP2, an ubiquitin fusion thereof comprising methionine (M) as the CBЈs N-terminal residue (Ub-M-BC1-TagGFP2) and the noncleavable mutant Ub G76V -M-BC1-Tag-GFP2 in HEK293T cells, and compared the apparent size of the matured protein by immunoblot. In line with efficient cleavage of Ub, we detected no size difference between BC1-TagGFP2 and the Ub-M-BC1-TagGFP2 construct, whereas Ub G76V -M-BC1-TagGFP2 displayed a size shift toward a higher molecular weight, which corresponds to uncleaved Ub-CB fusion protein (supplemental Fig. S6). In the following, we generated a set of 20 expression constructs differing only in the N-terminal amino acid exposed after Ub cleavage. Upon transient expression in HeLa cells, we determined mean CB fluorescence by quantitative imaging. With phenylalanine (F), alanine (A), lysine (K), tryptophan (T), arginine (R), and tyrosine (Y), we identified residues, which led to a reduced CB fluorescence of 40 -60% compared with methionine (M), whereas the expression construct carrying a serine (S) at its N terminus showed the highest fluorescence (Fig. 3C).
For further analyses, we focused on the amino acids Phe, Ala, and Arg, which conferred the strongest reduction of CB fluorescence and Ser as the most stabilizing residue. As the expression of ubiquitin ligases may vary between different cell lines, we additionally tested the levels of these modified CBs (Ub-F-BC1-TagGFP2, Ub-A-BC1-TagGFP2, Ub-R-BC1-Tag-GFP2, Ub-S-BC1-TagGFP2) in U2OS and A549 cells. Quantitative analysis of the fluorescence revealed similar CB signals as detected in HeLa cells (supplemental Fig. S7).
To test whether the effect of the identified amino acids is transferable to CBs targeting different antigens, we replaced the BC1 nanobody by ␣CA or VB6 in our chromobody expression constructs. For a comparative analysis we transiently expressed all modified CBs in HeLa cells in combination with mCherry as a transfection control and performed fluorescence imaging followed by automated quantification of the CB signals (supplemental Fig. S8A, S8B). Our data revealed a reduced fluorescence for all CB constructs carrying an N-terminal Phe and Arg residue as well as a stabilizing effect of Ser. Notably, the strong effect of Ala detected for BC1-TagGFP2 was not observable for Ub-A-VB6-TagGFP2 or Ub-A-␣CA-TagGFP2 (supplemental Fig. S8A, S8B).
According to our experimental design, differences in CB expression and fluorescence should only arise from differences in the CB degradation velocity. Thus, we analyzed CB fluorescence in HeLa cells transiently expressing Ub-M-BC1-TagGFP2, Ub-F-BC1-TagGFP2, Ub-R-BC1-TagGFP2, Ub-S-BC1-TagGFP2 over time upon inhibition of translation by cycloheximide (CHX) (Fig. 3D). Compared with Ub-M-BC1-TagGFP2 and Ub-S-BC1-TagGFP2, which displayed 69 and 80% of the initial fluorescence after six hours, fluorescence of CBs with N-terminally exposed Phe or Arg decreased more rapidly to about 28% of their respective initial fluorescence (Fig. 3E). From this we conclude that Phe and Arg are turnover-accelerating amino acids, which confer rapid degradation of CBs.
Turnover-accelerated CBs Show Improved Antigen-mediated Stabilization-To analyze whether the N-terminally modified CBs show an improved antigen-dependent enrichment, we compared the stabilization effect of CTNNB1 on Ub-R-BC1-TagGFP2, Ub-F-BC1-TagGFP2, and Ub-S-BC1-Tag-GFP2 to Ub-M-BC1-TagGFP2. Upon co-expression of the modified constructs either with mCherry-CTNNB1 or mCherry as a control in HeLa cells (Fig. 4A), we determined the average CB fluorescence within co-transfected cells. With respect to the degree of CB stabilization, N-terminal Met and Ser behave similarly with stabilization factors of 1.8 and 1.9, respectively. For Phe and Arg we detected the greatest stabilization effects with factors of ϳ2.3 (Fig. 4B). Taken together, the Arg-modified BC1-CB construct showed the most rapid turnover (Fig.  3E), lowest fluorescence in the absence and greatest stabilization in the presence of the antigen (Fig. 4B). In addition we tested the N-terminal Arg modification in the context of the ␣CA-CB (␣CA-TagRFP). Upon transient expression in HeLa cells we observed an increased stabilization of the Ub-R-␣CA-TagRFP construct in the presence of antigen compared with the nonmodified version (supplemental Fig. S9A, S9B).
We additionally analyzed a potential impact of the linker length between NB and FP moiety and the FP moiety itself on AMCBS. Hence, we first substituted the (G 4 S) 3  with mCherry or mCherry-CTNNB1 in HeLa cells, we found no significant differences between the linker lengths irrespective of the overexpression of the antigen (supplemental Fig. S10). To evaluate a potential impact of the FP on AMCBS, we substituted TagGFP2 either by eGFP, mCherry, or TagRFP in our N-terminally modified BC1-CB constructs and monitored degradation velocities in the presence of CHX. Within six hours we observed the most rapid degradation to 26% for Ub-R-BC1-eGFP (supplemental Fig. S11A), whereas corresponding CB constructs comprising either mCherry or TagRFP are less degraded to 43% or 57%, respectively (supplemental Fig. S11B, S11C). Additionally, we analyzed antigen-dependent stabilization as described above. Here, we detected a 2.9-fold fluorescence increase of Ub-R-BC1-eGFP in antigen-expressing cells compared with Ub-M-BC1-eGFP (supplemental Fig. S12A), which exceeds the stabilization factor of 2.3 observed for Ub-R-BC1-TagGFP2 (Fig. 4B). Similar stabilization factors were observed for Ub-R-BC1-mCherry (supplemental Fig. S12B) and Ub-R-BC1-TagRFP (supplemental Fig. S12C). Finally, to verify antigen binding of the modified CB constructs, we performed intracellular immunoprecipitations from HEK293T cells expressing Ub-M-BC1-eGFP, Ub-R-BC1-eGFP, the original BC1-TagGFP2 (21) as positive control, or GFP as negative control. Immunoblot analysis revealed that all CB constructs precipitate endogenous CTNNB1 in comparable amounts whereas no binding to GFP was observed (supplemental Fig. S13). From that we conclude, that intracellular antigen binding is neither affected by ubiquitin fusion nor by the introduction of turnover-accelerating amino acids.
Monitoring Rapid Changes in CTNNB1 Level in Living Cells With Turnover-accelerated Chromobodies-A major benefit of the AMCBS approach compared with end-point assays is its applicability to continuously monitor time-dependent changes in POI concentration in living cells. Previously, we demonstrated that an increase of endogenous CTNNB1 upon compound treatment can be visualized by real-time imaging of the CB signal in HeLa_BC1-TagGFP2 (21). For decreasing antigen concentrations, we supposed that CBs with accelerated degradation would reflect these changes more precisely. Thus, we tested, whether turnover-accelerated CBs can trace reversible changes in POI concentration in live cells. Because transient expression of CBs displays a substantial heterogeneity of intracellular CB levels, we generated monoclonal HeLa and U2OS cell lines stably expressing the turnoveraccelerated CTNNB1 chromobody (HeLa_Ub-R-BC1-eGFP, U2OS_Ub-R-BC1-eGFP). Quantitative fluorescence imaging of HeLa_Ub-R-BC1-eGFP cells revealed that in the absence of experimentally modulated levels of CTNNB1, CB fluorescence is nearly indistinguishable from autofluorescence of parental HeLa cells, whereas elevating endogenous CTNNB1 levels by CHIR treatment strongly increased CB fluorescence. Under both conditions, in situ detection of CTNNB1 showed that the mean CTNNB1 concentration was highly similar in parental HeLa and HeLa_Ub-R-BC1-eGFP cells (supplemental Fig. S14A and S14B), indicating that expression of the turnover-accelerated BC1-CB has no effect on the amount of endogenous CTNNB1. To test whether the Arg-modified CB construct retains its binding properties, we performed intracellular immunoprecipitation in HeLa_Ub-R-BC1-eGFP cells compared with HeLa_BC1-TagGFP2 cells. Our immunoblot analysis revealed that the modified BC1-CB can precipitate endogenous CTNNB1 upon induction with CHIR (supplemental Fig. S14C).
Next, we compared CB performance in the original HeLa_ BC1-TagGFP2 (21) and the newly generated HeLa_Ub-R-BC1-eGFP cell line to monitor reversible changes of endogenous CTNNB1 levels. To that end, cells were cultivated in the presence of CHIR for 16 h. Subsequently, cells were washed and continuously cultivated. One, two, three, four, six, eight, and 24 h post removal of CHIR, cells were fixed and immuno-stained for CTNNB1 followed by fluorescence imaging (Fig.  5A). In all cell lines we found a ϳ12-fold increase of endogenous CTNNB1 upon CHIR treatment, which rapidly returned to base level within four hours after compound removal (Fig.  5B). Concurrently, we also observed elevated CB signals after 16 h of CHIR treatment in both CB cell lines. Notably, in cells expressing turnover-accelerated Ub-R-BC1-eGFP, relative mean CB fluorescence not only increased more strongly compared with the original nonmodified BC1-TagGFP2, but also decreased more rapidly after CHIR removal. In contrast, in cells expressing the nonmodified BC1-TagGFP2 construct the CB signal remained at higher levels after removal of CHIR (Fig.  5B). A similar rapid decrease of the CB signal was observed in U2OS_Ub-R-BC1-eGFP cells upon removal of CHIR (supplemental Fig. S15). For a better illustration of the reversible changes in BC1-CB fluorescence we performed continuous live-cell imaging of HeLa_BC1-TagGFP2, HeLa_Ub-R-BC1-eGFP, and U2OS_Ub-R-BC1-eGFP cells upon addition and subsequent removal of CHIR (supplemental videos SV1-3).
To assess the accuracy of CTNNB1 estimation by BC1-CB fluorescence, we plotted the ratio of normalized mean fluorescences (CTNNB1/BC1-CB) against time (supplemental Fig. S16). In an ideal model, CB fluorescence would be proportional to the amount of CTNNB1 at any given time. Consequently, the ratio of the normalized fluorescence values would always be 1. Comparing the overall deviation from the ideal model, the newly generated cell line comprising the turnover-accelerated BC1-CB clearly outperforms the original cell line. For additional validation, we generated whole-cell lysates from CHIR-treated parental HeLa, HeLa_BC1-Tag-GFP2 and HeLa_Ub-R-BC1-eGFP cells at indicated time points and monitored CTNNB1 and CB levels by immunoblot analysis (supplemental Fig. S17, Fig. 5C). These results are in line with our microscopically obtained data, confirming a more rapid turnover of the Ub-R-BC1-eGFP upon removal of CHIR. In summary, these data show that the turnover-accelerated CBs substantially improve the ability to monitor reversible changes in the amount of endogenous CTNNB1 by AMCBS.
Determination of Compound Effects Using Turnover-accelerated Chromobodies-In preclinical drug development the use of cellular in vitro models has strongly increased over the last years. These models serve as powerful tools to screen for novel candidates as well as to evaluate cellular compound efficacy. However, to our knowledge there are no methods available to continuously monitor dose-response and kinetics of drug action on the level of endogenous target proteins. Here, we applied our turnover-accelerated BC1-CB-expressing cell line U2OS_Ub-R-BC1-eGFP to monitor the effect of two well established GSK3-␤ kinase inhibitors CHIR and 6-bromoindirubin-3-oxime (BIO) on endogenous CTNNB1 in a quantitative live-cell imaging setup. We continuously imaged U2OS_Ub-R-BC1-eGFP upon incubation with different inhibitor concentrations for 42 h (Fig. 6A) and quantified nuclear fluorescence every 3 h (Fig. 6B). For both inhibitors an overall dose-dependence of the effect was obvious. Prolonged timelapse image analysis revealed that higher inhibitor concentrations (Ͼ2.5 M for CHIR and Ͼ5 M for BIO) induced onset of cell death after 30 -36 h. Notably, we also identified differences in the kinetics of drug action. Although for every concentration of CHIR the corresponding maximum effect is ob-served after 22 h and gradually declines thereafter, the effect of BIO reaches a plateau at about 24 h for nontoxic concentrations. Interestingly, the maximum tolerated concentration of both inhibitors (1 M CHIR and 2.5 M BIO) led to a similar increase in BC1-CB fluorescence of about 3.5-fold (Fig. 6B). Although the molecular basis for the inhibitorsЈ cytotoxicity might have multiple reasons including off-target effects, it is also possible that U2OS cells only tolerate a certain elevation of CTNNB1 for a defined period. Taken together these data demonstrate that turnover-accelerated chromobodies in combination with quantitative imaging allows a precise determination of dose-and time-dependent compound effects on the level of individual endogenous proteins in living cells.

DISCUSSION
Chromobodies (CBs) comprising nanobody (NB)-derived binding moieties genetically linked to fluorescent proteins have become valuable tools to visualize endogenous antigens within living cells (10,14). Recently, we and others observed that NBs are stabilized in the presence of their antigen (21,22,27). Originally observed for a CB targeting hypo-phosphorylated CTNNB1, here we show antigen-mediated stabilization (AMCBS) for four different CBs targeting soluble (cytoplasmic/ nuclear CTNNB1), structural (vimentin), nuclear (PCNA) and virus-derived (p24-CA) antigens. Considering similar observations reported by others, it is conceivable that AMCBS is a general phenomenon common to numerous CBs. Notably, by monitoring CB signals in different stable cell systems, we not only visualized elevation but also, for the first time, depletion of cellular antigens over time. Although substantial changes in protein levels over longer periods can be sufficiently visualized using CBs in their original format, it has to be considered that specific signals of bound CBs in response to smaller and/or more rapid changes in protein levels can be obscured by the diffuse signal of nonbound CBs. To cope with this issue, recently, several approaches have been described to modify such intracellular nanoprobes accordingly.
To repress the expression of nontarget-bound intrabodies a DNA-binding KRAB domain was fused to an intrabody, thereby establishing a negative transcriptional feed-back mechanism (33). However, like any regulatory circuit involving transcription, this approach reacts sluggishly to rapid changes of cellular POI levels. Moreover, because of DNA-binding, KRAB domain-containing intrabodies accumulate in the nucleus even in the absence of the target protein (33). When we added the KRAB domain to our VIM-and CTNNB1-specific CBs, we observed a strong enrichment of both CBs in the nucleus, which impedes target concentration analysis by AMCBS (data not shown). Consequently, this system is not suitable to monitor nuclear proteins and rapid changes in protein levels. Another approach to lower the concentration of unbound intrabodies was reported for constructs comprising a PEST domain, which promotes rapid ubiquitin-independent proteasomal degradation (34). Upon introduction of PESTmodified CB expression constructs in live cells, we observed a substantial decrease in CB fluorescence. However, this was accompanied by a rapid onset of cell death irrespective of the addressed antigen (data not shown).
Recently distinct point mutations within the framework regions were described to destabilize, and accordingly lower the amount of intracellular NBs. Such modified NBs were shown to be re-stabilized in the presence of overexpressed antigen and thus are functional to detect recombinant or viral antigens e.g. by flow cytometry (27). Although it was stated that only NBs comprising framework mutations are stabilized in the presence of the antigen, our analysis revealed substantial antigen responsiveness even for nonmodified CBs. This indicates that antigen-mediated stabilization is inherent to CBs per se and does not depend on mutational destabilization. Moreover, as shown for the BC1 C92Y -TagGFP2 version, the introduction of mutations within the framework regions bears the risk to lose functional binding molecules. Notably, this is in line with previous reports of multiple NBs that show a participation of the framework regions in antigen binding (35)(36)(37)(38).
Here, we conceived a strategy to reduce base levels of nonbound CBs, which do not affect antigen binding. We focused on the N terminus of CBs, which has never been reported to participate in antigen binding and employed the rather old concept of the N-end rule (30,39). With either Met or Ala at the N terminus, CBs are subjected to the Ac/N-end rule pathway that involves N-terminal acetylation, which presumably results in a long half-life of the protein (30). The other major degradation pathway of the ubiquitin proteasome system is the Arg/N-end rule pathway. To screen for CB turnover accelerating N-terminal amino acid residues we implemented the ubiquitin fusion technique (40) and identified Arg and Phe, which, when exposed at the N terminus of all tested constructs, mediates the fastest CB turnover. The identification of these representatives of basic or bulky hydrophobic residues is in accordance with short half-lives described for other proteins displaying those residues at their N termini (31). Additionally, we observed significant differences in the degradation velocities of CBs comprising different fluorescent proteins. We assume that variances in position and number of lysine residues accessible for ubiquitinylation within the fluorescent moiety have an impact on turnover rates of the corresponding CBs.
Although our findings provide strong evidence that CBs are degraded via the ubiquitin proteasomal system, the precise molecular and structural mechanisms, which are responsible for the stabilization of CBs upon antigen binding within living cells remain to be elucidated. E3 ubiquitin ligases recognize the N-terminal amino acid and initiate ubiquitinylation of accessible nearby lysine residues (31). Notably, such ubiquitinylated proteins are stable within living cells unless they also expose an unstructured region, which is needed to initiate degradation (41). CBs have two tightly folded domains: the binding moiety, which shows a typical immunoglobulin fold and the cylindrical FP structure. Consequently, unstructured regions are restricted to the complementarity determining regions (CDRs) of the NB, the interconnecting linker, and to the short alpha helices forming the caps on the ends of the FP beta-barrel. Forming the paratope, the CDRs are in close contact with the antigen and thus it can be speculated that antigen binding masks these potential initiation sites and thus prevents degradation of the CB. Additionally, it is conceivable that CB molecules become partially immobilized in the cell upon antigen binding and are therefore less likely to interact with proteasomes compared with freely diffusible CBs. The possible reduced mobility of the bound CBs is in agreement with findings showing that larger protein complexes have a limited and/or reduced diffusion coefficient (42). Finally, competition of antigen and ubiquitin ligases for CBs, which would also facilitate the escape of antigen-bound CBs from ubiquitinylation could also contribute to the observed phenomenon of AMCBS.
For monitoring protein levels optically, FP tagging of endogenous proteins is a straightforward approach. However, FP tagging can interfere with crucial protein parameters such as turnover, subcellular localization, and participation in multiprotein complexes (6 -8). To avoid a permanent FP fusion, recently a technique was described, which relies on the cotranslational separation of the POI and the fused FP reporter translated from a bicistronic expression constructs (43). Although monitoring FP fluorescence indicative for the expression of the POI provides a simple and efficient read-out, endogenous proteins can be only addressed upon genome editing bearing the risk, that only one allele is modified. Moreover, as the half-life of the POI might differ from that of the FP, this method is suited to measure relative amounts at steady state but is likely insufficient to measure rapid dynamic changes in POI concentration (43). For proteins such as CTNNB1, whose concentration is not primarily regulated by transcription but degradation, this approach is not applicable.
The herein described turnover-accelerated chromobodies substantially expand the possibilities of these multifunctional nanoprobes. AMCBS with highly antigen-responsive CBs combines for the first time visualization of subcellular localization and redistribution of endogenous proteins with monitoring and quantification of rapid changes of protein levels by quantitative live-cell imaging. Like for any molecular probe applied in quantitative live-cell imaging, a potential influence of CB binding on antigen levels has to be carefully evaluated. Here, we demonstrated that the level and the dynamics of endogenous CTNNB1 is not affected by the presence of CTNNB1-specific CBs. Notably, similar observations were made for the PCNA-CB (CCC-TagRFP) and VIM-CB (VB6-eGFP) as reported previously (18,19). Moreover, the generation of organisms stably expressing CBs also strongly indicates that CB binding in trans does not affect the levels of tightly regulated antigens (18,44). From a technical perspective, this AMCBS approach is readily applicable, as fluorescence microscopy instrumentation is widely available in cell biology laboratories. Because of continuous improvements of nano-/chromobody screening protocols (45,46), the number of available chromobodies is constantly growing (14) and will enable time-resolved quantification of many further proteins of interest in the near future. AMCBS in combination with target-specific chromobodies could be further adapted to detect post-translational modifications or the presence and abundance of specific splice variants, which cannot be detected with conventional FP fusions.