Simultaneous Visualization of Both Signaling Cascade Activity and End-Point Gene Expression in Single Cells

We have developed an approach for simultaneous detection of individual endogenous protein modifications and mRNA molecules in single cells in situ. For this purpose we combined two methods previously developed in our lab: in situ proximity ligation assay for the detection of individual protein interactions and -modifications and in situ detection of single mRNA molecules using padlock probes. As proof-of-principle, we demonstrated the utility of the method for simultaneous detection of phosphorylated PDGFRβ and DUSP6/MKP-3 mRNA molecules in individual human fibroblasts upon PDGF-BB stimulation. Further we applied drugs disrupting the PDGFRβ signaling pathway at various sites to show that this combined method can concurrently monitor the molecular effect of the drugs, i.e. inhibition of downstream signaling from the targeted node in the signaling pathway. Due to its ability to detect different types of molecules in single cells in situ the method presented here can contribute to a deeper understanding of cell-to-cell variations and can be applied to e.g. pinpoint effector sites of drugs in a signaling pathway.


Introduction
Studying molecules in their natural cellular context provides valuable means to learn more about the complex function and regulation of cells. Cellular activity can be determined at several levels of complexity. Protein or mRNA expression can be used to determine endpoint effects of an active signaling pathway, but in order to investigate propagation of the signals assays for protein interactions and post-translational modifications are needed. So far most of the studies on protein-protein interactions andmodifications as well as mRNA-expression levels have been performed in a bulk population, rather than in individual cells, due to restrictions in detection sensitivity and specificity of the methods available. These methods yield average information about the molecule of interest but will fail to detect intercellular variation [1]. Studies that are based on single cell data circumvent the risk that rare events are hidden within the bulk population or that the average for the whole population is not reflecting real variations within it. This is of particular importance when studying cellular signaling, since each cell in a population -even if they are clonal expansions from one precursor -will answer slightly different to a stimulus due to variations in e.g. cell cycle, accessibility to ligand and transcription factors, accumulated mutations and the cocktail of other signals this cell is receiving [2,3]. Visualization of multiple nodes in a signaling pathway will enhance the ability to monitor signal progression and provide a better tool to address heteroge-neity in the response of a cell population to stimulation, enabling studies on cellular communication.
To facilitate investigations of endogenous protein interactions and post-translational modifications in situ we recently developed the in situ proximity ligation assay (in situ PLA [4,5]). For detection of DNA and individual mRNA molecules in situ we developed the padlock probes that enable discrimination between single nucleotide polymorphisms (SNP:s) [6,7]. Both methods allow the detection of single molecules -either proteins or nucleic acid molecules -in fixed cells and tissues. Briefly, in situ PLA as applied here utilizes two antibodies from different species, one directed against the protein of interest, the other against its posttranslational modification. Upon binding to the same modified protein, species-specific PLA probes directed against the primary antibodies and carrying two different oligonucleotides are added ( Figure 1). Only if these bind in close proximity to the same target protein will they guide the hybridization and ligation of two connector oligonucleotides, generating a circular DNA molecule that is amplified by a strictly circle-dependent rolling circle amplification (RCA) using phi29 DNA polymerase. This reaction is primed from one of the oligonucleotides attached to the PLA probes and thus the single stranded RCA product (RCP) will stay attached to the place where the PLA probe was bound. The RCP collapses into a bundle of DNA, consisting of ,1000 repetitive elements, complementary to the DNA circle. This bundle can then be easily detected by hybridization with fluorescence labeled detection oligonucleotides, concentrating ,1000 fluorophores in a sub-mm sized bright spot, to be visualized by fluorescence microscopy.
Padlock probes for detection of individual mRNA molecules [7] also utilize RCA to visualize the detected molecules. The mRNA molecules are first reverse transcribed into cDNA, using a specific LNA-primer, to enable ligation of the padlock probes on cDNA target molecules, which is a more efficient target than RNA. To create a single-stranded cDNA target, the original mRNA is digested by RNase H, except for the part that is tightly hybridized to the LNA-modified nucleotides of the primer. This will keep the synthesized cDNA anchored to the mRNA in the cell. A padlock probe is then hybridized to the cDNA and upon perfect match the target complementary 59 and 39 ends will become joined by ligation, creating a circular DNA molecule ( Figure 1). Since the ligase is very sensitive to mismatches at the ligation junction, padlock probes can be used for discrimination of SNP:s in situ [6,7,8]. After circularization, the padlock probe is used as template for an RCA primed from the cDNA, ensuring that the RCP stays attached to the original target position in the cell. Visualization is subsequently done as described for in situ PLA. As both methods outlined here are able to investigate individual molecules of different types we wanted to combine in situ PLA and padlock probes for detection of individual protein modifications and mRNA molecules to provide an assay to measure activity at different positions in the signaling pathway, giving a more coherent view on the status of individual cells.
The platelet-derived growth factor receptor beta (PDGFRb) is a receptor tyrosine kinase associated with growth and motility. Upon stimulation of PDGFRb with PDGF-BB the receptor is dimerized and auto-phosphorylated at several sites leading to recruitment of GRB2 and SOS, which in turn activate the RAS-RAF-MEK-ERK pathway. Upon sustained stimulation with PDGF-BB the phosphorylated receptor becomes internalized while activated ERK upregulates expression of downstream mRNA-targets e.g. dual specificity phosphatase 6 (DUSP6), also called MKP-3 [9,10], a dual-specificity phosphatase that dephosphorylates ERK, thereby providing a negative feedback loop for its own expression [10]. We used this model system to visualize the kinetics of ligand stimulation at an initial stage -phosphorylation of the PDGFRband a late stage -expression of the downstream target gene DUSP6. Further, we applied drugs that target different nodes in the pathway to test if the assay will be applicable to pinpoint the molecular effect of a drug. Using the assay developed above, treatment with drugs caused inhibition of either both phosphorylation of the PDGFRb and DUSP6 expression or DUSP6 expression alone, depending on where on the pathway they act. We demonstrated detection of cell-to-cell variations with single molecule resolution in fixed cells. The method presented here could be especially valuable when screening for new drugs, where the effector molecules are still unknown or where the drugs should only target a certain part of a signaling pathway.

Cell culture
TERT immortalized human fibroblast (BJhTert) [11] cells were seeded on 8-well chamber slides (LabTek, Nunc), 20,000 cells per well. The cells were allowed to attach overnight and subsequently starved in Modified Eagle Medium (MEM, Gibco)+0.1% FCS (heat-inactivated, Sigma) for 48 h. For stimulation PDGF-BB was added to a final concentration of 100 ng ml 21 and the cells were incubated at 37uC for the indicated time points. If the cells were treated with drugs, the drugs Gleevec (Novartis) and 5-Iodotubercidin (Sigma), dissolved in DMSO (Sigma) to a 10 mM stock concentration, were applied at a final concentration of 10 mM in fresh MEM+0.1% FCS for 1 h at 37uC before PDGF-BB was added at the indicated time points. The drug concentration was held constant throughout the experiment.
For fixation the cells were put on ice and the medium was immediately exchanged for ice-cold DEPC (Applichem Lifescience)-treated PBS (DEPC-PBS, 900 ml per well). After a 1 min wash, the cells were fixed in 2% (w/v) paraformaldehyde (PFA, Sigma Aldrich) in DEPC-PBS for 30 min at room temperature (20-23uC). Prior to permeabilization in ice cold 70% ethanol for 30 min on ice, the cells were washed again in ice-cold DEPC-PBS. Now the silicon mask of the chamber slides was removed, the slides were dehydrated through a series of 70%, 85% and 99.5% ethanol for ,1 min each, hydrophobic barrier pen was applied at the borders of the slide wells and finally 8-chamber Secure Seals (9 mm in diameter, 0.8 mm deep; Grace Bio-Labs) were attached to the slides.

Preparation of secondary PLA probes
One mg donkey anti-rabbit IgG (711-005-152, Jackson Immunoresearch) and 1 mg donkey anti-mouse IgG (715-005-150, Jackson Immunoresearch) were concentrated over Amicon ultra 0.5 ml, 10 kDa (UFC501008, Millipore) and dialyzed in 7 kD dialysis cup (Slide-A-Lyzer MINI dialysis units, #69562, Pierce) against PBS over night at 4uC. Antibodies were activated in SMCC (Pierce) added in 25-fold excess over antibody for 2 h at room temperature. Primer oligonucleotide (59 thiol-AAA AAA AAA ATA TGA CAG AAC TAG ACA CTC TT (Eurogentec)) and blocked oligonucleotide (59 thiol-AAA AAA AAA AGA CGC TAA TAG TTA AGA CGC TTU UU (Biomers)) were degassed at 95uC for 3 min and reduced by incubation with 25 mM DTT for 1 h at 37uC. Both, antibodies and oligonucleotides were purified using an Illustra NAP-5 column (#17-0853-01, GE Healthcare), equilibrated and eluted with PBS+5 mM EDTA pH 7.5. Afterwards anti-rabbit IgG was immediately mixed with the primer oligonucleotide and anti-mouse IgG with the blocked oligonucleotide. PLA probes were then dialyzed against PBS at 4uC overnight. Before purification by HPLC the conjugates were concentrated to 25 ml by Amicon ultra 0.5 ml, 10 kDa columns and 10 mM DTT was added to the conjugate to reduce unbound oligonucleotide dimers. The conjugates were incubated for 10 min at room temperature and subsequently injected into HPLC (Hitachi D7000 series). Chromatography was done over a Superdex 75 PC 3.2/30 column with a flow rate of 0.06 ml min 21 to separate conjugates from free oligonucleotides and samples were collected at 16-20.5 min (conjugate peak).

Detection of phosphorylated PDGFRb with in situ PLA
The protocol is based on the protocol described previously in Jarvius et al. 2007 [5]. All reactions were performed in Secure Seals attached to the slides. Prior to each new step all liquids were removed from the Secure Seal chambers and for incubations exceeding 30 min the chambers were sealed with q-PCR film to avoid evaporation. All washes were performed by flushing the wells with ,500 ml of the appropriate washing buffer.
First the wells were rehydrated in 16 PBS-T (DEPC-PBS+0.05% Tween20 (Sigma)) for 10 min at room temperature. The cells were blocked in Protein-Block, serum free (Dako), 2.5 ng ml 21 sonicated salmon sperm DNA (Invitrogen) and 2.5 mM L-cysteine (Sigma) for 90 min at 37uC. After blocking primary antibodies (63 ng ml 21 rabbit-anti-PDGFRb (#3169 Cell Signaling) and 6 mg ml 21 mouse-anti-pY100 (#9411, Cell signaling)) were applied in blocking solution over night at 4uC. Primary antibodies were removed by flushing the chambers with ,500 ml 16 PBS-T. Secondary PLA probes (1.6 ng ml 21 donkey-antirabbit-primer and 5 ng ml 21 donkey-anti-mouse-blocked, final concentration, see above) were incubated separately in half the final volume of blocking solution for 30 min at room temperature before they were mixed and applied to the cells for 1 h at 37uC. Subsequently the cells were washed in 10 mM Tris HCl pH 7.5, 0.1% Tween20 for 5 min at 37uC and once with 16 PBS-T. The in situ detection-procedure was performed in a similar way as described by Larsson et al. [7]. Fixation, permeabilization and dehydration of cells were however done as described above. Subsequent to the application of the Secure Seals, the cells were rehydrated in PBS-T. Reverse transcription was performed with 1 mM of the appropriate LNA-primer in 20 U ml 21 RevertAid H minus M-MuLV reverse transcriptase (Fermentas), 0.5 mM dNTPs, 0.2 mg ml 21 BSA, 1 U ml 21 Ribolock RNase inhibitor (Fermentas) in M-MuLV reaction buffer. Slides were incubated for 3 h at 37uC and then washed twice in PBS-T. All subsequent washes were performed in the same way. Postfixation was done in 2% (w/v) paraformaldehyde for 30 min at room temperature before the slides were washed. To make the cDNA accessible for padlock probe hybridization and to simultaneously ligate the padlock probe to its target sequence, the cells were incubated with 0.1 mM padlock probe, 0.4 U ml 21 RNase H (Fermentas), 0.5 U ml 21 Ampligase (Epicentre), 1 U ml 21 Ribolock RNase inhibitor in Ampligase buffer (20 mM Tris-HCl, pH 8.3, 75 mM KCl, 10 mM MgCl 2 , 0.5 mM NAD and 0.01% Triton X-100), 20% formamide (Merck), first for 30 min at 37uC and then for 45 min at 45uC. After an additional wash RCA was performed with 1 U ml 21 phi29 DNA polymerase, 1 U ml 21 Ribolock RNase inhibitor in the supplied phi29 DNA polymerase reaction buffer, 0.25 mM dNTPs, 0.2 mg ml 21 BSA and 5% glycerol for 90 min at 37uC. RCPs were labeled with 0.25 mM detection oligonucleotide in 26 SSC and 20% formamide for 15 min at 37uC. After detection the slides were washed twice in PBS-T and Secure Seals were removed. The cytoplasm of the cells was thereafter counterstained and the slides mounted as described above.
Simultaneous in situ detection of individual phosphorylated PDGFRb and DUSP6 or ACTB mRNA The procedure for simultaneous detection of individual proteins and transcripts in single cells was based on merging the two protocols described above. The merged protocol started with in situ reverse transcription, converting mRNA to DUSP6 and ACTB cDNA. This was followed by post-fixation of the cells, digestion by RNase H and hybridization/ligation of the specific padlock probes to the target cDNA molecules. These steps were performed in the same fashion as described above for solely transcript detection. Thus, the padlock probes were locked onto their targets and the target transcripts were secured. The combined protein/mRNA protocol could subsequently proceed with the detection of phosphorylated PDGFRb as described above except that 1 U ml 21 Ribolock RNase inhibitor was added to the blocking, antibody and PLA probe mixtures and DEPC-ddH 2 O was used instead of ddH 2 O in hybridization and ligation. RCA was performed as described for in situ PLA except that 2.5 U ml 21 phi29 DNA polymerase were used and both, padlock probes and in situ PLA signals were amplified. For detection, the fluorescence labeled oligonucleotides (see above), with a final concentration of 100 nM, were added to the detection mix described for in situ PLA.

Image acquisition
Images were acquired using a Zeiss Axioplan 2 epifluorescence microscope, the AxioCam MRm CCD sensor and a 206/0.8 Plan Apochromat lens together with filters for DAPI, FITC, Cy3 and Cy5. For each condition z-stacks were acquired at three positions of the well and the positions were chosen using the FITC channel (cytoplasm staining).

Image analysis and illustrations
Image J 1.44 d was used to make maximum intensity projections (MIPs) of the image stack of each channel. In the FITC channel, ImageJ was applied to enhance the contrast, subtract background (rolling ball radius = 200 pixel) and run a median filter (radius 4 pixel). Background was also subtracted from DAPI-images (rolling ball radius = 50 pixel).
Subsequently, CellProfiler v.10415 was used to identify cell nuclei (thresholding method: two-class thresholding ''Otsu Per-Object'') in previously background subtracted DAPI images. The nuclei were then combined with background corrected FITC images to identify the cytoplasm of individual cells (method to identify objects: ''Watershed -Image'', thresholding method: twoclass thresholding ''Otsu Adaptive''). Signals in Cy3 (PDGFRb) and Cy5 (DUSP6 or ACTB) were enhanced (feature type: ''speckles'', feature size: 10), counted (diameter: 2-20 pixel, thresholding method: ''RobustBackground Global'' (for Cy3), diameter: 1-20 pixel, thresholding method ''RobustBackground Adaptive'' (for Cy5), threshold correction factor: 0.8) and related to the individual cells. Subsequently, if images did not contain signals, the amount of signals had to be manually corrected, since then the adaptive enhancement of Cy3 and Cy5 signals gave rise to false positive signals. Graphs were done in R [12]. In Figure 1B images were for better visualization treated in ImageJ as described here: One slice of the DAPI z-stack was selected, background was subtracted in ImageJ and thresholded for brightness and contrast. Cy3 and Cy5 z-stacks were projected to MIPs, the background was removed, brightness and contrast adjusted and a maximum filter (size: 1 pixel) was applied.

Kinetics of PDGFRb phosphorylation and DUSP6 expression
To determine the kinetics of PDGFRb phosphorylation and DUSP6 expression upon stimulation with PDGF-BB, BJhTert cells were starved for 48 h prior to treatment with 100 ng ml 21 PDGF-BB. The cells were stimulated for different length of time (0 min-240 min) and subsequently fixed. Thereafter the phosphorylation of PDGFRb was investigated using in situ PLA with one antibody directed against the receptor and one against phosphorylated tyrosine residues. In parallel, the expression on DUSP6 transcripts was monitored by padlock probes on separate slides ( Figure 2). As expected, the phosphorylation of PDGFRb first increased substantially during the initial time points, peaking at 5-10 min after stimulation, before it gradually decreased again towards background levels due to internalization of the receptor [9]. In accordance with what has been reported before [10], DUSP6 transcript levels remained constant for the first 30 min of stimulation, and then showed an increase until 3 h. Interestingly, the distribution in numbers of detected molecules per cell in the cell population is quite large, especially for PDGFRb phosphor-ylation in response to stimulation. Regarding DUSP6 expression, not all cells are responding to stimulation with PDGF-BB to the same extent. It rather appears that, for the time investigated, some cells burst with transcription while in others the signal is not propagated to the induction of DUSP6 expression.

Combined detection of protein modification and mRNA expression
Next, we tested the combined detection of protein modifications and single transcripts (Figure 1). For this purpose phosphorylation of PDGFRb and ACTB mRNA molecules -an abundantly  expressed housekeeping gene to enable a more detailed analysis of detection efficiency -were detected either individually or simultaneously. The cells were fixed and permeabilized before cDNA synthesis and padlock probe hybridization and ligation were performed. After the ACTB cDNA had been targeted, the cells were blocked and stained with in situ PLA reagents required to detect phosphorylated PDGFRb. Both, reacted padlock probe and in situ PLA probes were amplified in the same reaction by RCA and detected by hybridization of fluorescence labeled detection oligonucleotides. We investigated whether the combined detection shows similar results as when the two detection reactions were run separately, and whether there is a decrease in detection efficiency upon combining the two. BJhTert cells were stimulated with PDGF-BB for 0 min, 5 min or 180 min. The results of the combined detection were consistent with the results of individual staining and the levels of ACTB transcripts were unaffected by stimulation with PDGF-BB, as expected. The combined protocol leads to a decrease in detected signals for both in situ PLA and padlock probes of about 50% compared to when performed in separate reactions (Figure 3).

Drug induced inhibition of PDGFRb phosphorylation and DUSP6 expression
Having developed a method that enabled measurement of both protein activity and mRNA expression in single cells, we then tested if the assay would be suitable to detect perturbation of the PDGFRb-signaling pathway using drugs that affect the pathway at different levels. In that way we were able to investigate if the assay could help in zooming in on the action of a drug. As proof of principle, we examined the effect of Gleevec (a tyrosine kinase inhibitor [13] targeting the PDGFRb directly) and 5-Iodotubercidin (an inhibitor of ERK2 activation [14], located downstream of PDGFRb but upstream of the induction of DUSP6 transcription) on phosphorylation of PDGFRb and DUSP6 expression. As expected, treatment with Gleevec eliminated the increase in signals for both molecules, since it inhibits the activation of the PDGFRb and thus all downstream events. 5-Iodotubercidin, on the other hand, did only eliminate the upregulation of DUSP6 but not phosphorylation of the PDGFRb, since it affected the signaling downstream of the receptor but upstream of DUSP6 (Figure 4).

Discussion
We herein successfully combined in situ PLA and padlock probes for detection of individual phosphorylated PDGFRb and DUSP6 transcripts in a single cell assay. First, the kinetics of each molecule was investigated in separate reactions and both molecules were detected as expected. The level of DUSP6 expression previously reported, ,1% compared to ACTB [15], are in agreement with the levels determined herein, confirming the ,30% efficiency of mRNA detection using padlock probes in situ [7]. For protein detection the efficiency of in situ PLA is less, a few percent, depending on antibody affinity and quality of the PLA probes. However, detection of phosphorylated PDGFRb using in situ PLA correlates with levels determined in a population based assay using immunoblot [5]. Combining in situ PLA with padlock probes detecting mRNA in situ did not affect detection of phosphorylated PDGFRb upon stimulation with PDGF-BB. ACTB mRNA levels were largely unaffected by stimulation with PDGF-BB ( Figure 3). However, the combined detection scheme decreased the signals for both targets to ,50% compared to individual detection, most likely since the protocol is longer and involves more washing steps than individual detection.
We then demonstrated in a proof-of-principle experiment how this combined detection could be applied for screening of new drugs or signal transduction studies, treating BJhTert cells with either Gleevec or 5-Iodotubercidin. Subsequently PDGF-BB was added at different time points and phosphorylation of PDGFRb, as well as expression of DUSP6 transcripts, was measured in the same cells. As expected, treatment with Gleevec inhibited PDGFRb phosphorylation, and consequently also impaired the induction of DUSP6 expression. 5-Iodotubercidin inhibited ERK2 activation and therefore did not affect PDGFRb phosphorylation but did inhibit DUSP6 expression, which is dependent on ERK2 activation.
Both in situ PLA and padlock probes allow detection of individual target molecules at endogenous levels in individual cells. Even after synchronization of the cells by serum starvation for 48 h prior to stimulation, there was still a broad distribution in how many target molecules per cell that were detected. The difference in amount of RCPs per cell can partially be explained by differences in cell size but might also reflect differences in response time or intensity. Especially when looking at DUSP6 transcript expression, some cells seem to burst with expression, while others remain at background levels, as previously described for other mRNA molecules [3]. As the measurements are performed in fixed samples, transient interactions/activations and bursts of expression between the different time points will not be recorded. Nevertheless, the assay provides information on frequency of expressing cells and levels of expression at the time of analyses. This cell-to-cell heterogeneity would have gone undetected if instead an averaging method would have been used. The possibility to simultaneously measure multiple parameters provides an advantage when more complex cell populations are studied, e.g. tissue sections, as they consists of a mixture of cell types, at various differentiation stages, responding to various external stimuli. Moreover, a simultaneous detection is advantageous in cases with rare or/and small sample size.
The combined method for detection of individual protein modifications and individual mRNA molecules in single cells in situ presented here may be a valuable tool to enable broader studies of signaling pathways. A further advantage: as the method is based upon imaging, differences between individual cells with respect to target molecule location, amount and modifications can be investigated together with morphological characteristics of the cells. Importantly, data about protein modification and mRNA expression can be related to the same cell. This might be of particular interest when new drugs are tested, especially such that should act very specifically and locally in a signaling pathway. In this paper we show detection of one specific post-translational modification and one specific transcript, but further developments of each individual method could also be applied to this combined technique. The main advantage with the strategy described herein lies in simultaneous targeting of multiple nodes in a signaling cascade to determine how a drug effects the propagation of the signals. This will allow studies in heterogeneous populations and provides a tool to determine if a drug can rewire a signaling circuit i.e., changing the balance in nodes where there are alternative interaction partners. In that way more branch points and levels of regulation can be revealed simultaneously, providing additional information on the activity status in individual cells that will reduce cost and increase throughput in high content analyses.