Article | Published:

Integrated analysis of receptor activation and downstream signaling with EXTassays

Nature Methods volume 7, pages 7480 (2010) | Download Citation

Subjects

Abstract

The ability to measure multiple cellular signaling events is essential to better understand the underlying complex biological processes that occur in living cells. Microarray-based technologies are now commonly used to study changes in transcription. This information, however, is not sufficient to understand the regulatory mechanisms that lead to gene expression changes. Here we present an approach to monitor signaling events upstream of gene expression. We coupled different reporter gene assays to unique expressed oligonucleotide tags (EXTs) that serve as identifiers and quantitative reporters. Multiple EXT reporters can be isolated as a pool and analyzed by hybridization to microarrays. To test the feasibility of our approach, we integrated complementation assays based on a protease from tobacco etch virus (TEV protease) and transcription factor activity profiling. Thereby, we simultaneously monitored Neuregulin-dependent mouse ErbB receptor tyrosine kinase dimerization, effector recruitment and downstream signaling.

Main

The use of modern genomic and proteomic technologies has dramatically changed the current understanding of molecular biology concepts, as exemplified by the paradigm shift in the understanding of cellular signaling. Linear cascade models evolved into complex signaling networks that are characterized by 'pathway' cross-talk at multiple levels1. The canonical signaling pathway initiated by epidermal growth factor (EGF) family ligands activating tyrosine kinase ErbB receptors is among the most extensively studied signaling networks. ErbB signaling has been implicated in developmental regulation, cancer progression and psychiatric diseases2,3. ErbB receptor activation has been extensively analyzed by biochemical means and, more recently, by protein arrays as well as pulldown experiments4,5. These approaches revealed a complex pattern of first-order signaling adaptor protein interactions for each ErbB family member. However, the interaction dynamics or the ligand-induced downstream signaling in a given cellular context can not be addressed using these techniques. Other methods allow multiplexed analyses of cellular signaling events using flow cytometry, high-throughput microscopy or transfected–cell arrays6,7,8,9. However, these assays are not optimally suited to detect modification-dependent interactions, in particular those of structurally complex membrane-bound receptors. We recently described the split TEV system (complementation assays based on a tobacco etch virus (TEV) protease) that is well suited to monitor, for example, phoshorylation-dependent interactions of membrane receptors with transcription-coupled reporters10,11. In this system, TEV protease activity is reconstituted from inactive TEV fragments expressed as fusion constructs of interacting proteins.

More distal aspects of cellular signaling can be studied using DNA microarrays and RNA sequencing to assess transcriptional changes and transcription-factor binding profiles globally12,13. More recently, a modified cis-regulatory reporter assay has been developed that allows the parallel assessment of transcription factor activities14. Thus far, however, available techniques are restricted to a particular biochemical mechanism or class of molecules. They can therefore only provide a partial view of the molecularly very different signaling events that are associated with a given stimulus.

Here we present an integrated and highly scalable reporter system, termed expressed oligonucleotide tag (EXT) assay, or EXTassay, to simultaneously monitor cellular signaling at multiple levels within living cells. We combined split TEV system and cis-regulatory reporter gene assays to monitor ErbB receptor activation and downstream signaling. Using different split TEV assays, we quantitatively assessed Neuregulin-1 (NRG1)-induced ErbB receptor dimerization and phosphorylation-dependent interactions of ErbB receptors with first-order signaling mediators11. In parallel, we used different cis-regulatory reporter assays to monitor downstream signaling by different ErbB receptor pairs at the level of activated transcription factors.

Note: Supplementary information is available on the Nature Methods website.

Results

Design of EXTassays

In EXTassays, multiple reporter gene assay formats, such as split TEV and cis-regulatory reporter assays, are used to monitor different molecular events in parallel (Fig. 1). EXT reporters replace classical reporter proteins. To monitor receptor activation, we fused ErbB receptors to inactive N- and C- terminal fragments of the TEV protease (N-TEV and C-TEV) and a synthetic transcription activator Gal4-VP16 (GV) via a specific TEV protease cleavage site. Upon ligand binding, the ErbB receptors dimerize, leading to the reconstitution of protease activity and release of GV, which binds a cis element (G5) to induce the expression of defined EXT reporters. Similarly, we used fusion constructs of different adaptor proteins to measure their interaction with activated ErbB receptor complexes through proteolytic GV release and activation of different subsets of EXT reporters (Fig. 1). We unambiguously assigned EXTs to particular assays by expressing defined EXT reporter constructs with corresponding assay components. We monitored the activation of endogenous transcription factors with EXT reporter constructs carrying various cis-regulatory elements upstream of a minimal promoter. Such assays can be done simultaneously in one cell without interference. Transfected cells containing reporter constructs for one or multiple assays are grown under identical conditions. For the analysis, total RNA is extracted, EXTs are amplified by reverse transcriptase (RT)-PCR, labeled and analyzed with decoding microarrays. The relative signal intensities of multiple EXTs serve as quantitative measures for the corresponding cellular assays (Fig. 1).

Figure 1: Barcoding of multiple cellular assays with EXT reporter assays.
Figure 1

Schematic representations of the three different types of assays are shown at the top. To monitor receptor dimerization (left), ErbB–TEV protease N- and C-terminal fragment fusion proteins are expressed, which upon dimerization cause the proteolytic release of transcription factor GV that binds the corresponding cis element (G5) to activate EXT reporter expression. To monitor adapter recruitment (middle), a phosphotyrosine adaptor protein (pAd) is fused to C-TEV, which upon recruitment to an ErbB receptor fused to N-TEV leads to activation of different EXT reporters. To monitor endogenous transcription-factor activation (right), cis-element reporter constructs with corresponding unique EXTs are expressed in a cell as exemplified by serum response element (SRE) and cAMP response element (CRE) reporters. In an assay (bottom), the components are expressed in different pools of cells and plated together in the same dish. RNA is isolated, and EXTs are amplified by RT-PCR, labeled and hybridized to a 'decoding' microarray.

EXT structure and performance

The performance of our approach critically depends on the ability of the EXT reporters to quantitatively monitor and fully discriminate between multiple cellular events. Thus, we designed an EXT oligonucleotide synthesis strategy that yielded a highly complex library of EXTs with balanced melting temperatures (Tm) and virtually no intramolecular complementary regions (Fig. 2a). Each EXT comprised a variable stretch of 49 bases and defined invariable 5′ and 3′ sequences. The variable region consisted of ten 'words' flanking a core region. We used eight different 4-nucleotide (nt) 'words', each composed of three adenosines or thymidines and one cytosine15. The core region comprises nine bases of alternating adenosine and thymidine or guanine and cytosine residues with three central guanine and cytosine residues. The invariable regions serve as common templates for amplification and subcloning (Fig. 2 and Supplementary Fig. 1a). To synthesize oligonucleotides, we used a combinatorial mix-and-divide strategy using standard chemistries. The theoretical complexity of the EXT library was 5.5 × 1011 distinct sequences. We introduced the symmetrical core element to assess the performance of each EXT with a few central mismatches and to enhance and stabilize Tm profiles for hybridization. The core element also increased the complexity compared to a purely word-based structure by eightfold. The balanced base-pair composition was analyzed by computationally generating a large virtual EXT library of 105 sequences, which we compared to cloned and sequenced EXTs as well as random 49-mers. The virtual and cloned libraries had a narrow Tm ranges of 61.2 ± 0.7 °C and 61.2 ± 0.8 °C, respectively, in contrast to the broadened Tm distribution of random 49-mers (70.1 ± 3.3 °C) (Fig. 2b). The EXT design also resulted in a low inter-EXT similarity in the cloned EXT library as determined by basic local alignment search tool (BLAST) comparisons of all sequences versus each other (data not shown).

Figure 2: Structure and synthesis of EXTs.
Figure 2

(a) Each oligonucleotide in the library contains 5′ (13 nt) and 3′ (15 nt) invariable regions, a 49-nt variable region consisting of ten 4-nt 'words' (W) and a nonamer core sequence. The complexity is calculated as 810 (10 positions with 8 words) times 29 (2 bases at 9 core positions) yielding more than 1011 combinations. (b) Tm for 229 cloned and sequenced EXTs (histogram) and calculated Tm distribution for 105 randomly computed EXTs (random EXTs) and for 105 random 49-nt sequences (random 49-mers). Mean Tm values are given ± s.d. (c) Hybridization dynamics of selected individually labeled EXTs; amount of EXTs is indicated by the amount of incorporated Cy3 label. Data are given as mean values of different EXTs (± s.e.m.; n = 4) except for one EXT used at 85 fmol Cy3 dye with a signal intensity of 11,179 arbitrary units. Inset, magnification of data in the low-femtomole range.

EXT performance in microarray hybridizations

We randomly selected several EXTs to thoroughly evaluate their performance in microarray experiments using custom arrays obtained from two different vendors. First, we determined the hybridization dynamics with mixes of individually labeled control EXTs at defined concentrations. The analysis revealed a dynamic range of detection over more than two orders of magnitude. The correlation coefficient, R2 = 0.99, was determined with the linear curve fitting in the GraphPad software (Fig. 2c). Next, we designed partially complementary sequences for four 'control' EXTs (c1–c4), which included single- or multiple-base and single- or multiple-word mismatches and combinations of both (Fig. 3 and Supplementary Figs. 1,2,3,4). Using c1–c4 as targets, we optimized the hybridization conditions with respect to hybridization dynamics and to precisely determine cross-hybridization profiles (Fig. 3 and Supplementary Figs. 1,2,3,4). An extensive set of rigorous control experiments revealed that (i) the hybridization dynamics of all control EXTs were highly similar and (ii) sequence-related EXTs can be discriminated with high specificity. Even under extremely stringent cross-similarity exclusion criteria (maximal 70% sequence identity), the number of EXTs that theoretically can be fully discriminated with microarrays exceeded 107 (Fig. 3 and Supplementary Note 1). Furthermore, abundance differences could be detected reliably even at very low input amounts (Figs. 2c and 3). The amplification of EXT template mixtures by PCR (even when diluted up to 1010-fold) did not change the relative abundance of a given EXT (Supplementary Fig. 5). Based on these analyses, we conclude that EXTs are optimally suited microarray targets for multiplexed EXTassays.

Figure 3: EXT mismatch-discrimination profiles.
Figure 3

Average signal intensity profiles obtained from microarray analyses with four control target EXTs (c1–c4) hybridized to complementary perfect and selected mismatch EXT probes, with perfect match signals set to 100%. Representative, perfect (upper-case letters) and mismatch (lower-case letters) sequences and identity scores for c1 are shown. Mismatches were sorted by decreasing identity scores: (i) single base mismatches, (ii) single word mismatches, (iii) combined single word and single base mismatches, (iv) multiple word and (v) multiple word and base mismatches. Irrespective of positional effects, full discrimination was reached at an identity score of 69%. The background (red line) was determined as an average of EXTs with similarity scores below 50% (four representatives are depicted in v). Data are mean values for c1–4 (error bars, s.d.; n = 4).

Design and analysis of EXT reporter constructs

To generate large numbers of EXT reporter constructs, we used a multicomponent modular recombination cloning strategy (Supplementary Fig. 6). To compare the performance of the corresponding constructs with standard luciferase reporter plasmids, we cloned EXTs upstream of the luciferase gene and placed these constructs under transcriptional control of either a TATA-minimal or the viral thymidine kinase (TK) promoter and a Gal4-responsive cis element (G5), which can be activated in mammalian cells by GV. Such modular EXT-luciferase constructs performed as efficiently as conventional reporter constructs (Supplementary Fig. 6).

Multiplexing EXT analysis with microarrays

Next, we assessed the performance of multiple EXT reporters in cellular assays. To model a dynamic response of EXT reporters, we transfected a pheochromocytoma cell line (PC12-OFF) with ten different G5-TK-EXT reporter constructs (assay pool) and either 0, 0.1, 1 or 10 ng of a GV-encoding plasmid, respectively, to measure the GV-dependent EXT expression (Supplementary Table 1). In parallel, we transfected a mixture of 60 additional G5-TK-EXT constructs (reference pool) expressing EXTs (11–70) along with a constant amount of a GV expression plasmid (5 ng) (Fig. 4a and Supplementary Table 1). We added equal amounts of cells from the 'reference pool' to each of the 'assay pools' and analyzed both simultaneously. We used a large excess of the 'reference pool' to avoid an amplification bias between the samples. We extracted plasmid DNA and RNA and reverse-transcribed the RNA. We amplified and labeled input DNA and cDNA with Cy3 and Cy5, respectively (Fig. 4a). In parallel, we amplified DNase-treated RNA and found no contaminating plasmid DNA (Fig. 4b). We simultaneously hybridized plasmid DNA- and cDNA-derived samples as well as control EXTs in two-color fluorescence experiments to diagnostic microarrays. We normalized for variability of the EXT plasmid input and varying transfection efficiencies by dividing Cy5-labeled cDNA fluorescence values by Cy3-labeled plasmid DNA fluorescence signal intensities and depicted the array data as reference signal intensities (Supplementary Fig. 7a). The corresponding reference signal intensities of the EXTs 1–10 increased in a dose-dependent fashion with similar rates (Fig. 4c). In contrast, the signal intensities of EXTs 11–70 were constant in all samples (Fig. 4c and Supplementary Fig. 7b). We therefore concluded that cellular EXTassays with multiple EXT reporter constructs are a technically feasible approach and allow parallel and quantitative analyses of different assay pools without interference.

Figure 4: EXTassay performance under control conditions.
Figure 4

(a) Experimental design of a multiplexed EXTassay under controlled conditions. Indicated amounts of GV were transfected along with G5-TK-EXT-luciferase reporter plasmids (assay EXT 1–10 pool) to simulate a dynamic signal. Equal numbers of cells were transfected with a 'reference EXT pool' (G5-TK-EXT(11–70)-luciferase constructs) and 5 ng GV plasmid and were plated with each of the 'assay cell populations'. The 'reference' pool served as an internal assay control. EXTs were amplified by PCR and labeled by in vitro transcription (PCR-IVT) and then hybridized to EXT arrays. (b) EXT PCR products obtained with RNA and cDNA input after 30 cycles. The absence of plasmid DNA contaminations in the RNA samples was verified by PCRs on input RNA samples that were not subjected to reverse transcription and yielded no detectable products at the same cycle number. (c) GV G5-EXT dose-response curves measured by EXTarray hybridizations. Average reference signal intensities (RSI) ± s.d. are given for all EXTs of the assay pool (1–10; red line, n = 10) and the reference pool (11–70; blue dashed line, n = 60) as indicated.

ErbB receptor activation and downstream signaling

We next integrated split TEV assays10,11 with EXTassays. We used PC12-OFF cells that do not express NRG1-responsive receptors but become signaling-competent to NRG1 through heterologously expressed ErbB receptors16. First, we tested the experimental setup and all components used for ErbB receptor dimerization and adaptor recruitment with standard luciferase assays (Supplementary Fig. 8). Only ligand-binding–competent Erbb2 and Erbb3 complexes or Erbb2 and Erbb4 complexes responded to NRG1 stimulation (Supplementary Fig. 8) in accordance with previous findings and split TEV experiments2,10,11. In contrast, Erbb2 dimers did not mediate NRG1 effects but rather had elevated constitutive levels of dimerization (Supplementary Fig. 8), which is also in agreement with previous results2,10,11,17. To assess transcription factor activities downstream of ErbB signaling in PC12-OFF cells, we cloned EXT reporter constructs using 24 defined cis elements with two to three EXTs each (Supplementary Table 2 and Supplementary Fig. 9). To analyze different ErbB receptor assays in parallel, we used G5-regulated EXT-luciferase reporter constructs containing either a TATA-minimal or TK promoter (Supplementary Fig. 10 and Supplementary Table 2). To quantitatively measure ErbB receptor signaling at different levels with EXTassays, we transfected upstream and downstream assay components in pools. We collected the cells 2, 4, 12 and 36 h after NRG1 stimulation (Fig. 5a), and isolated plasmid DNA and RNA for subsequent EXTarray analysis as described. We normalized the corresponding microarray data using standard algorithms (Supplementary Fig. 11 and Supplementary Table 3). Correlation analysis as well as scatter plots revealed a high reproducibility between replicates of DNA input samples, and we detected substantial differences between control and ErbB receptor construct–transfected samples only at the RNA input level (Supplementary Fig. 11). The mismatch performance of spike-in control EXTs was very similar to that obtained under control conditions (Fig. 3 and Supplementary Note 1) and served as an additional array quality assessment (data not shown).

Figure 5: Integrated NRG-ErbB signaling with EXTassays.
Figure 5

(a) PC12-OFF cells were treated as indicated. (b) Heatmap of the reference signal intensities of all EXTassays for ErbB receptor pairs (none, no ErbB receptor; 2/2, Erbb2 and Erbb2; 2/3, Erbb2 and Erbb3, 2/4, Erbb2 and Erbb4). For each receptor condition at each time point, 19 different split TEV, 54 different cis-regulatory-EXT and 68 reference EXT reporters (141 in total) were monitored. Split TEV dimerization (ErbB-ErbB n = 6) and adaptor recruitment assays (ErbB-PI3Ka, n = 5; ErbB-SHC1, n = 4; and ErbB-GRB2, n = 4) are plotted at the top and cis-regulatory reporter assays below (see Supplementary Table 4 for all raw data and n values). (ce) Split TEV ErbB dimerization (c), split TEV ErbB-PI3Ka interaction (d) and AP1 cis-regulatory reporter assays (e) plotted as line graphs (left). Schematic drawing of the assays for the Erbb2-Erbb3 receptor pair (right). Data are given as the mean of all corresponding EXTs ± s.d. (n = 3) for two replicate microarrays (R1 and R2). (f) Principal component analysis of all response profiles (left, principal components 1 and 2 in plane; right, same plot turned to show component 1 and 3 in plane).

Differential signaling of ErbB receptor complexes

Next, we analyzed all EXT reporter–derived signals of the no ErbB receptor control with Erbb2 and Erbb2 (2/2), Erbb2 and Erbb3 (2/3), and Erbb2 and Erbb4 (2/4) samples. We plotted the results of all individual split TEV and cis-regulatory reporter assays in a single heatmap (Fig. 5b and Supplementary Table 4), which revealed time- and receptor-dependent differences. Different EXT reporters corresponding to 'identical' assays (Supplementary Table 2) performed similarly in replicate experiments (Supplementary Fig. 12).

Next, we plotted selected EXTassay profiles as individual line graphs, with corresponding profiles of the control samples to indicate the baseline of split TEV and cis-regulatory reporter assays in PC12-OFF cells (Fig. 5c–e and Supplementary Fig. 12). As PC12-OFF cells do not express NRG1-responsive receptors16, the overall profile did not change significantly over time (ANOVA for time F3,31 = 2.39, P = 0.07) (Fig. 5b). The pattern of cis reporters indicated differential baseline activities of the corresponding transcription factors. Most prominently, P53 and AP1 reporter activities were elevated several fold compared to activities of control constructs lacking defined cis elements (TATA, TAL) (Fig. 5b).

The EXTassay profiles obtained with the NRG1-binding–incompetent Erbb2 receptors (2/2) revealed marked differences of split TEV and cis-assay results compared to the control (Fig. 5b,e). In agreement with luciferase assays (Supplementary Fig. 8) and published data10,11, increased split TEV reporter activities likely reflect ligand-independent Erbb2 receptor dimerization and activation (Fig. 5b–d and Supplementary Fig. 13). In cis-assays, Erbb2 signaling induced a distinct pattern of transcription factor activities (Fig. 5b). Most prominently, AP1 and P53 cis-reporter activities were consistently reduced compared to the no-ErbB-receptor control throughout all time points, whereas SRE remained largely constant (Fig. 5b). Overall, the 2/2 EXTassay profile changed only slightly, yet significantly, over time (ANOVA for time F3,31 = 3.03, P < 0.04) (Fig. 5b).

The EXTassay profiles of the NRG1-responsive ErbB receptor pairs 2/3 and 2/4 revealed dramatic and significant time-dependent changes (ANOVA for time F3,31 = 11.50, P < 2 × 10−6 for 2/3 and F3,31 = 18.54, P < 2 × 10−9 for 2/4) (Fig. 5b–e). The 2/3 receptor dimerization and phospho-adaptor recruitment was more efficient than those for the 2/4 complex (Fig. 5b). Moreover, split TEV assays revealed kinetic differences between 2/3 and 2/4 receptor complexes. The 2/3 complex activation was longer lasting and more efficient compared to that of the 2/4 complex (Fig. 5c,d). In cis-reporter activation, the 2/3 complex elicited also stronger responses than 2/4 (for example, AP1 and SRE activation) (Fig. 5b,e), but the patterns were different (Fig. 5b,e and Supplementary Figs. 12,13). The receptor- and time-dependent effects were corroborated by a principal-component analysis, which closely grouped all control and 2/2 profiles. In contrast, the response patterns of the 2/3 complex (particularly at 2, 12 and 36 h) and of the 2/4 complex at 12 h were different (Fig. 5f).

EXTassay versus luciferase measurements

Next, we assessed ErbB dimerization and phospho-adaptor recruitment with split TEV and standard luciferase assays for the same time points as in Figure 5a. Luciferase assays basically recapitulated the main EXTassay results, but differences in kinetics and dynamic range of the different reporter classes became apparent. TATA-EXT and TK-EXT reporters robustly detected 2/3 and 2/4 complex activation 12 h after NRG1 stimulation, whereas for luciferase assays, we observed NRG1 effects only 36 h after stimulation (Fig. 6). Moreover, transient activation of the 2/4 complex was masked in luciferase assays as was the time-dependent decline in reporter activation by the 2/2 complex (Fig. 6). We reasoned that the improved kinetics of EXT reporters over luciferase assays may be due to different half-lives of mRNAs versus those of the luciferase protein. In PC12-OFF cells half-lives were indeed different: 5.1 h for the luciferase protein and 1.2 h for its mRNA (Supplementary Fig. 14).

Figure 6: Improved kinetic performance and sensitivity of EXT reporters compared to standard luciferase assays.
Figure 6

(a,b) EXT reporters and standard luciferase assays were compared to monitor NRG1-dependent ErbB receptor activation 2, 4, 12 and 36 h after stimulation (none, no ErbB receptor expressed; 2/2, Erbb2 receptor expressed; 2/3, Erbb2 and Erbb3, 2/4, Erbb2 and Erbb4 receptors expressed as N-TEV-GV and C-TEV fusion constructs, respectively). Split TEV dimerization assays (a) and split TEV PI3Ka recruitment assays (b) (G5-TK promoter constructs and G5-TATA minimal promoter constructs) with standard luciferase assays to the control situation lowest value set to 1. Data are given as the mean of EXT reporter assays (± s.d.; n = 3).

EXTassay analysis with high-throughput sequencing

To expand the EXTassay readout options to next-generation sequencing applications, we analyzed cis-reporter activity profiles (Supplementary Table 2) in HEK293, PC12 and PC12-OFF cells. We amplified corresponding EXT pools to attach Illumina sequencing adapters (Supplementary Fig. 15a). The analysis of the reads corresponding to EXT pools obtained from the plasmid input revealed a homogenous distribution between cell lines and different cis-EXT constructs, whereas cDNA-derived cis-EXT profiles were distinctly altered for different cis EXTs (Supplementary Fig. 15b). The overall profiles of PC12 and PC12-OFF cells were highly similar (R2 = 0.99) but substantially different when compared to HEK293 cells (PC12 versus HEK293, R2 = 0.25; and PC12-OFF versus HEK293, R2 = 0.23; Supplementary Fig. 15b). We conclude that high-throughput sequencing of EXT libraries is technically feasible with current technologies and provides an additional readout option for EXTassays.

Discussion

For most microarray approaches, probe design is hampered by the limited choice of sequences that are dictated by a given target and by difficulties in computationally predicting hybridization performance of target-probe pairs18. Owing to the homogeneous properties of the EXTs and corroborated by control experiments, critical issues regarding hybridization performance can be largely excluded from our approach. Optimizing barcodes with our or other strategies18 may improve similar screening procedures19.

EXTassays rely on short mRNA-based reporters in contrast to a recently developed cis-regulatory reporter assay with homogenous mRNA reporters which are rather large (700 bases)14. As the latter depends on digestion and fragment size analysis, its multiplexing is limited.

Kinetic resolution and sensitivity of EXTassays was far better than that of luciferase assays. A delay of enzymatic reporter gene assays compared to RNA readouts has been reported previously14,20. To improve the kinetic resolution of EXTassays, the RNA stability of EXT reporter constructs could be adjusted by introducing RNA-destabilizing elements21. The limitations of EXTassays are generally those of all reporter systems. Cell lines that are difficult to transfect may not be suited for the current setup of our assays using transient transfections. Moreover, functional interference of reporter constructs with endogenous proteins could occur. Cis-regulatory reporter assays might alter gene expression by sequestration of endogenous transcription factors. However, our data showed that the transfection of cis-reporters in pools reduced this risk.

So far, we analyzed 141 different EXT reporters in parallel, but our data suggest that this number could be expanded by several orders of magnitude (Supplementary Discussion). In general, EXT reporters can be linked to any assay format that uses a transcriptional readout. The list of assays includes standard reporter gene assay formats such as cis-regulatory or promoter-activity measurements, transcription factor activation and protein-DNA binding assays covered by one-hybrid approaches22. Applicable assays include all types of transcription-coupled protein complementation systems such as two-hybrid23, split TEV10, split-intein24, split-ubiquitin25 or proximity assays such as the TEV-Tango or mammalian protein-protein interaction trap techniques26,27. The large number of different cellular events that have already been implicated in cellular signaling1 calls for an integrated analysis of these events, which at the same time requires highly flexible and scalable measurement tools. The EXTassay platform may provide an opportunity to simultaneously monitor complex cellular signaling networks at multiple levels.

Methods

Synthesis of the EXT oligonucleotide library.

The synthesis of the EXT library was carried out using standard DNA oligonucleotide synthesis chemistry. First, we generated the 3′ invariable region and split the sample into 8 portions. Then, we continued the synthesis with eight different 4-nt combinations or 'words': CTTT, CAAA, ACAT, TCTA, TACT, ATCA, TTAC and AATC15. We again mixed the samples and we subdivided them to continue with different words at the next position. We repeated these cycles five times. Subsequently, we generated a symmetrical 9-mer core sequence by splitting the samples into two equal portion after each next nucleotide (WSWSSSWSW, where W = A or T and S = C or G) (International Union of Pure and Applied Chemistry (IUPAC) code28). We completed the synthesis by five more rounds of word synthesis followed by the 5′ invariable region. We purified the resulting mixture of the 77-mer oligonucleotides by HPLC and PAGE.

Cloning of the EXT library.

We amplified the library with the primers EXTlib-F and EXTlib-R, which are complementary to the 5′ and 3′ invariable regions, respectively (Supplementary Table 5). We re-amplified the PCR product with the primers EXT_B3 and EXT_B2 to introduce attB3 and attB2 recombination sites compatible with the Multisite Gateway Pro system (Invitrogen) (Supplementary Table 5). We generated a library of EXT shuttle clones (pENTR_EXTlib) by recombination cloning of the PCR products into the pDONR-P3P2 backbone of the Multisite Gateway Pro system (Invitrogen). We transformed the library into either electrocompetent Escherichia coli DH-10b or chemically competent Mach1 cells (Invitrogen). The bacteria were allowed to grow at 37 °C for 4 h and were used for plasmid DNA preparation (Qiagen). We plated a portion of the bacteria on solid medium and screened for full-length EXT constructs by colony PCR and subsequent sequencing. About a quarter of the clones had perfect full-length inserts, and all sequenced EXTs were unique. The complexity of the cloned EXT library was estimated to contain roughly 105 unique full-length clones.

Expression plasmids and EXT reporter constructs.

We amplified full-length open reading frames of mouse ErbB receptors and adaptor proteins by PCR and subcloned these by recombination to generate expression plasmids for the corresponding N-TEV-GV and C-TEV fusion constructs essentially as described10,11. To generate modular reporter constructs, we used a destination vector pDEST-GL3-basic derived from pGL3-basic (Promega) by cloning the recombination cassette rfC.1 (Invitrogen) into the SmaI restriction site upstream of the firefly luciferase reporter gene. We amplified functional units from plasmids or oligonucleotide templates attaching recombination sites of the Multisite Gateway Pro recombination system (Invitrogen). We categorized cis-regulatory elements using the Transfac database29. We generated the final EXT reporter constructs consisting of a cis-regulatory element, a minimal promoter and an EXT by three-fragment recombination using corresponding shuttle clones according to the specifications of the Multisite Gateway Pro recombination system (Invitrogen). All constructs were sequence verified.

Cell culture.

We grew PC12-tet-OFF (PC12-OFF) cells (Clontech) on poly(L-lysine)–coated dishes at 37 °C in a humidified air incubator supplemented with 5% CO2 in DMEM low glucose medium, supplemented with 10% FBS, 5% horse serum (HS), 1% penicillin-streptomycin and 1% GlutaMAX (Invitrogen).

NRG1 stimulation.

We used the EGF-like domain from mouse NRG1 tagged with GST (GST-NRG1-EGFβ) that was expressed in bacteria and purified for stimulation experiments essentially as described10. After transfection, we grew PC12-OFF cells for 24 h in DMEM supplemented with 5% FBS (Gibco) and added GST-NRG1-EGFβ to the final concentration of 125 ng ml−1. We used purified GST as negative control at the same concentration. We collected cells at the indicated times after stimulation.

Transfections for NRG1-ErbB EXTassays.

For microarray experiments, we performed transfections in suspension. We trypsinized the cells, pelleted them by centrifugation and resuspended in DMEM containing 1% HS to achieve the density of 1,000 cells μl−1. We prepared Lipofectamine-DNA complexes using 100 ng DNA and 0.4 μl Lipofectamine 2000 (Invitrogen) in 10 μl OptiMEM medium per 100,000 cells. After 20 min incubation at room temperature (21–23 °C), we combined the mixture with the cell suspension and incubated the samples for 4 h at 37 °C without agitation. To remove transfection reagents, we washed and resuspended the cells from different transfections (for example, split TEV and cis-regulatory reporter assays) in DMEM medium supplemented with 5% FBS. We cultured the cells under identical conditions before analysis. We performed luciferase assays to control for the absence of cross-transfections following this protocol.

For EXTassays, we used 5 ng plasmid DNA per each unique EXT reporter construct and 100 ng for the plasmids encoding the ErbB receptors. We transfected cis-EXT reporter constructs in groups of 20 plasmids using 105 cells per transfection. For each cis-regulatory element we cloned 2–3 reporter constructs carrying different EXTs, whereby each cis-assay was represented by n = 2–3 EXT measurements. We transfected the components of the protein-protein interaction assays with TK promoter and TATA box–driven EXT reporters for each assays using 105 cells. For the reference pool, we transfected 68 G5_TK_EXT reporter constructs with 100 ng GV in two groups of 34 constructs (5 ng each).

Luciferase reporter gene assays.

We always performed luciferase assays in 96-well format with six replicate wells per condition essentially as described previously10. We transfected the cells using Lipofectamine 2000 according to the manufacturer's recommendations. We lysed the cells 24–48 h after transfection with 30 μl passive lysis buffer per well (Promega) and measured the activity of firefly and Renilla luciferase using a Mitras LB940 microplate reader (Berthold Technologies). We normalized the firefly luciferase readings to the corresponding Renilla luciferase readings and displayed the data as relative luminescence units.

Half-lives of luciferase mRNA and enzyme activity.

We determined the time-dependent decay of luciferase mRNAs and enzyme activities using the TET-regulatory system30. We transfected PC12-OFF cells with the doxycycline-responsive tetO_CMV_Luciferase reporter carrying a minimal CMV promoter and the doxycycline-responsive cis element (tetO). PC12-OFF cells express the synthetic transactivator tTA that is not capable of DNA binding in the presence of doxycycline. We added 100 ng ml−1 doxycycline 12 h after transfection and collected samples at various time points up to 36 h. For each time point, we performed luciferase assays and quantitative reverse transcriptase PCRs with primers to firefly luciferase mRNA on replicate samples and used rat β-actin as internal standard. We determined the corresponding half-lives by nonlinear curve-fitting using GraphPad (GraphPad software) and the 'one phase exponential decay' mode.

RNA isolation and cDNA synthesis.

We prepared RNA by adding 1 ml Trizol reagent (Invitrogen) per 106 cells. We added one-fifth volume of chloroform and vortexed the samples thoroughly before centrifugation for 15 min at 4 °C, 12,000g. We supplemented the aqueous phase by an equal volume of 70% ethanol and purified the samples over an RNeasy column (Qiagen) including on-column DNase treatment according to the manufacturers' protocol. We additionally treated the purified RNA with 3 units of RQ1 DNase (Promega) for 30 min at 37 °C and repurified it over an RNeasy column. We precipitated the RNA with half a volume of 7.5 M ammonium acetate and 3 volumes of 100% ethyl alcohol. We generated first strand cDNA from 500 ng total RNA using Superscript III reverse transcriptase (Invitrogen) and 120 pmol of random nonamer primer. We assessed RNA purity with respect to possible plasmid DNA contamination with a negative control for each sample where all reaction components were included except reverse transcriptase ('−RT control'). We diluted the cDNA 1:8 with water and used it as template for PCR amplification.

Plasmid DNA input isolation.

Owing to its small size, plasmid DNA was not denatured by Trizol and was localized in the aqueous phase. Before RNA purification, we removed one-fifth of the aqueous phase and precipitated the DNA by adding one-tenth volume of 3 M sodium acetate and 2.5 volumes of ethanol. We labeled EXT PCR products from these samples by in vitro transcription using the T7-Megascript kit (Ambion) and hybridized products along with those obtained from cDNA. The plasmid-DNA derived EXT mixtures served to calibrate for different transfection efficiencies and relative abundance differences within plasmid mixtures.

PCR amplification of EXT reporter mixtures.

We amplified EXT mixtures from cDNA and plasmid DNA input samples using 'Decoder' primers Dec1 and Dec2 (Supplementary Table 5). We monitored the amplification of the EXT mixtures by real-time PCR to asses the sample quality with respect to possible plasmid DNA contamination. The PCR was carried out using a 7500 Fast Real-time PCR System and standard SYBR green reagents (Applied Biosystems) as described previously31.

In vitro transcription and RNA labeling.

We used unpurified PCR products as templates to generate RNA by in vitro transcription using the T7 MEGAscript Kit (Ambion). We replaced 50% of UTP by 5-(3-aminoallyl)-modified UTP (Ambion) to produce aminoallyl-labeled RNA (aaRNA). We allowed the reaction to proceed at 37 °C overnight. We treated the samples with DNase1 for 15 min and purified the RNA over RNeasy columns (Qiagen). We adjusted the sample volume to 100 μl with RNase-free water and added 100 μl buffer RLT. To ensure the efficient recovery of the short in vitro transcription products, we added 400 μl of isopropanol. We transferred the samples to the RNeasy columns and centrifuged these for 15 s at 10,000g. We washed the membranes twice with 500 μl buffer RPE and dried it by centrifugation for 2 min and eluted the RNA. For the labeling reaction, we lyophilized 5–10 μg of the aaRNA and resuspended it in 4.5 μl coupling buffer (Ambion). We dissolved one tube of the Cy3- or Cy5-monoreactive dyes (Amerscham), respectively, in 55 μl DMSO and added 5.5 μl to each sample. After a 30-min incubation at room temperature in the dark, we added 2.3 μl 4 M hydroxylamine (Ambion) and incubated the samples for 15 min. We purified the RNA using the modified RNeasy protocol described above.

Microarray hybridization.

We used custom-designed microarrays from Agilent (8×15K) and Roche NimbleGen (4×72K). We adjusted the amount of the target to the amount of the fluorescent dye. We loaded 10–30 fmol of Cy dye per EXT on each array. We hybridized Agilent microarrays at 65 °C using HI-RPM hybridization buffer (Agilent) supplemented with 15% formamide. We hybridized Roche/NimbleGen arrays at 48 °C or 50 °C using standard hybridization components supplied by the manufacturer. Washing of Agilent and NimbleGen microarrays was performed following standard procedures recommended by the manufacturers. We dried the arrays by dipping for 30 s into acetonitrile. We performed initial tests with self spotted microarrays on Codelink slides (GE Healthcare) that were produced and hybridized by the microarray core facility, University of Göttingen (http://www.microarrays.med.uni-goettingen.de/).

Data extraction and analysis.

We scanned all arrays with 5 μm resolution using the GenePix 4200A microarray scanner and the associated GenePix Pro software (Axon Instruments) or the G2565AA microarray scanner enabled by the Scan Control software (Agilent). We extracted the data using Agilent Feature Extraction (Agilent) or NimbleScan (Roche Nimblegen) software. We analyzed data using the R statistical computing environment and the Genomics Suite 6.4 (Partek). We normalized the overall intensities of the microarrays using the RMA algorithm implemented in the R package. We calculated the average probe intensities across four to eight replicate features depending on the array design. We normalized the average Cy5 signals representing the relative EXT expression levels to the Cy3 signals originating from the plasmid DNA input to calculate the reference signal intensities.

High-throughput sequencing.

We amplified EXTs in a two-step PCR to attach 5′ and 3′ Illumina sequencing adapters according to the manufacturer recommendations (Illumina) (Supplementary Table 5). PCR products were analyzed with Illumina's Genome Analyzer with 100 base reads encompassing Dec1 or Dec2 primer stretches and invariable EXT regions. Flanking sequences were stripped off and the resulting EXT sequences and aligned against the EXT database. Alignments showing any variations or indels were sorted out. We used plasmid input derived total read counts to correct for sample differences. We normalized RNA-cDNA derived sample-corrected reads to DNA input reads per EXT (given as reference read numbers).

References

  1. 1.

    , , & Reconstruction of cellular signalling networks and analysis of their properties. Nat. Rev. Mol. Cell Biol. 6, 99–111 (2005).

  2. 2.

    & EGF-ERBB signalling: towards the systems level. Nat. Rev. Mol. Cell Biol. 7, 505–516 (2006).

  3. 3.

    & Neuregulin 1 in neural development, synaptic plasticity and schizophrenia. Nat. Rev. Neurosci. 9, 437–452 (2008).

  4. 4.

    , , & A quantitative protein interaction network for the ErbB receptors using protein microarrays. Nature 439, 168–174 (2006).

  5. 5.

    , & Phosphotyrosine interactome of the ErbB-receptor kinase family. Mol Syst Biol 1, 0008 (2005).

  6. 6.

    & Fluorescent cell barcoding in flow cytometry allows high-throughput drug screening and signaling profiling. Nat. Methods 3, 361–368 (2006).

  7. 7.

    & Automated subcellular location determination and high-throughput microscopy. Dev. Cell 12, 7–16 (2007).

  8. 8.

    et al. Identifying off-target effects and hidden phenotypes of drugs in human cells. Nat. Chem. Biol. 2, 329–337 (2006).

  9. 9.

    , & Cell-biological applications of transfected-cell microarrays. Trends Cell Biol. 12, 485–488 (2002).

  10. 10.

    et al. Monitoring regulated protein-protein interactions using split TEV. Nat. Methods 3, 985–993 (2006).

  11. 11.

    , , & Analysis of transient phosphorylation-dependent protein-protein interactions in living mammalian cells using split-TEV. BMC Biotechnol. 8, 55 (2008).

  12. 12.

    DNA microarray technologies for measuring protein-DNA interactions. Curr. Opin. Biotechnol. 17, 422–430 (2006).

  13. 13.

    , & RNA-seq: a revolutionary tool for transcriptomics. Nat. Rev. Genet. 10, 57–63 (2009).

  14. 14.

    et al. Homogeneous reporter system enables quantitative functional assessment of multiple transcription factors. Nat. Methods 5, 253–260 (2008).

  15. 15.

    et al. In vitro cloning of complex mixtures of DNA on microbeads: physical separation of differentially expressed cDNAs. Proc. Natl. Acad. Sci. USA 97, 1665–1670 (2000).

  16. 16.

    , , & ErbB-4 activation promotes neurite outgrowth in PC12 cells. J. Neurochem. 74, 979–987 (2000).

  17. 17.

    et al. erbB-2 is a potent oncogene when overexpressed in NIH/3T3 cells. Science 237, 178–182 (1987).

  18. 18.

    , , & Design of 240,000 orthogonal 25mer DNA barcode probes. Proc. Natl. Acad. Sci. USA 106, 2289–2294 (2009).

  19. 19.

    , , & Signature-tagged mutagenesis: barcoding mutants for genome-wide screens. Nat. Rev. Genet. 7, 929–939 (2006).

  20. 20.

    & Discordant estimates of heterologous promoter activity as determined by reporter gene mRNA levels and enzyme activity. Biochem. Biophys. Res. Commun. 189, 912–918 (1992).

  21. 21.

    , & AU-rich elements and associated factors: are there unifying principles? Nucleic Acids Res. 33, 7138–7150 (2005).

  22. 22.

    , , & A gateway-compatible yeast one-hybrid system. Genome Res. 14, 2093–2101 (2004).

  23. 23.

    & A novel genetic system to detect protein-protein interactions. Nature 340, 245–246 (1989).

  24. 24.

    , & Intein-mediated reporter gene assay for detecting protein-protein interactions in living mammalian cells. Anal. Chem. 78, 556–560 (2006).

  25. 25.

    , , & A genetic system based on split-ubiquitin for the analysis of interactions between membrane proteins in vivo. Proc. Natl. Acad. Sci. USA 95, 5187–5192 (1998).

  26. 26.

    et al. Design and use of a mammalian protein-protein interaction trap (MAPPIT). Sci. STKE 2002, pl18 (2002).

  27. 27.

    et al. The genetic design of signaling cascades to record receptor activation. Proc. Natl. Acad. Sci. USA 105, 64–69 (2008).

  28. 28.

    Nomenclature committee of the International Union of Biochemistry (NC-IUB). Nomenclature for incompletely specified bases in nucleic acid sequences. Recommendations 1984. J. Biol. Chem. 261, 13–17 (1986).

  29. 29.

    et al. TRANSFAC and its module TRANSCompel: transcriptional gene regulation in eukaryotes. Nucleic Acids Res. 34, D108–D110 (2006).

  30. 30.

    & Tight control of gene expression in mammalian cells by tetracycline-responsive promoters. Proc. Natl. Acad. Sci. USA 89, 5547–5551 (1992).

  31. 31.

    et al. Global transcriptome analysis of genetically identified neurons in the adult cortex. J. Neurosci. 26, 9956–9966 (2006).

Download references

Acknowledgements

We acknowledge the contributions of F. Benseler and all members of the Sequencing Core Facility (Max Planck Institute of Experimental Medicine, Götttingen) for the EXT synthesis and excellent services, and the expert support by staff of the Microarray Core Lab of the University of Göttingen, namely R. Hitt, G. Salinas-Riester and L. Opitz. We thank R. Reinhardt, S. Klages and S. Scheer for high-throughput sequencing and primary data analysis, K.-A. Nave for support, F. Melchior, E. Wimmer, as well as all lab members for stimulating discussions, and M. Wehr for critically reading the manuscript and providing valuable feedback. This study was supported by grants of the Bundesministerium für Bildung und Forschung (FKZ0315180A) and partially by the European Union (LSHM-CT-2005-018637) to M.J.R.

Author information

Affiliations

  1. Max Planck Institute of Experimental Medicine, Göttingen, Germany.

    • Anna Botvinnik
    • , Sven P Wichert
    • , Tobias M Fischer
    •  & Moritz J Rossner

Authors

  1. Search for Anna Botvinnik in:

  2. Search for Sven P Wichert in:

  3. Search for Tobias M Fischer in:

  4. Search for Moritz J Rossner in:

Contributions

A.B. cloned EXT constructs, performed the assays, analyzed the data, contributed conceptually and wrote the manuscript. S.P.W. performed bioinformatic calculations and analyzed the data. T.M.F. cloned expression constructs. M.J.R. conceived the study, analyzed the data and wrote the manuscript.

Competing interests

M.J.R. filed a patent application covering the principles of EXTassays.

Corresponding author

Correspondence to Moritz J Rossner.

Supplementary information

PDF files

  1. 1.

    Supplementary Text and Figures

    Supplementary Figures 1–15, Supplementary Tables 1–5, Supplementary Note 1 and Supplementary Discussion

About this article

Publication history

Received

Accepted

Published

DOI

https://doi.org/10.1038/nmeth.1407

Further reading