Identification of Isoform-Specific Dynamics in Phosphorylation

Oct 21, 2014 - cytokine that mediates STAT5 activation is erythropoietin (Epo). Differential .... erythropoietin receptor (EpoR) as described previous...
0 downloads 0 Views 2MB Size
Article pubs.acs.org/jpr

Identification of Isoform-Specific Dynamics in PhosphorylationDependent STAT5 Dimerization by Quantitative Mass Spectrometry and Mathematical Modeling Martin E. Boehm,†,‡,∥ Lorenz Adlung,‡,∥ Marcel Schilling,‡ Susanne Roth,‡,§ Ursula Klingmüller,*,‡ and Wolf D. Lehmann*,† †

Molecular Structure Analysis, ‡Systems Biology of Signal Transduction, German Cancer Research Center (DKFZ), Im Neuenheimer Feld 280, 69120 Heidelberg, Germany § Systems Bioinformatics, Netherlands Institute for Systems Biology, VU University, De Boelelaan 1085, 1081 HV Amsterdam, Netherlands S Supporting Information *

ABSTRACT: STAT5A and STAT5B are important transcription factors that dimerize and transduce activation signals of cytokine receptors directly to the nucleus. A typical cytokine that mediates STAT5 activation is erythropoietin (Epo). Differential functions of STAT5A and STAT5B have been reported. However, the extent to which phosphorylated STAT5A and STAT5B (pSTAT5A, pSTAT5B) form homo- or heterodimers is not understood, nor is how this might influence the signal transmission to the nucleus. To study this, we designed a concept to investigate the isoform-specific dimerization behavior of pSTAT5A and pSTAT5B that comprises isoform-specific immunoprecipitation (IP), measurement of the degree of phosphorylation, and isoform ratio determination between STAT5A and STAT5B. For the main analytical method, we employed quantitative labelfree and -based mass spectrometry. For the cellular model system, we used Epo receptor (EpoR)-expressing BaF3 cells (BaF3-EpoR) stimulated with Epo. Three hypotheses of dimer formation between pSTAT5A and pSTAT5B were used to explain the analytical results by a static mathematical model: formation of (i) homodimers only, (ii) heterodimers only, and (iii) random formation of homo- and heterodimers. The best agreement between experimental data and model simulations was found for the last case. Dynamics of cytoplasmic STAT5 dimerization could be explained by distinct nuclear import rates and individual nuclear retention for homo- and heterodimers of phosphorylated STAT5. KEYWORDS: STAT5, relative isoform quantification, dimerization, phosphorylation, Epo stimulation, mathematical modeling



INTRODUCTION Reversible protein phosphorylation is a widespread principle in intracellular signal transduction. Signaling by cytokines is typically connected to different functions in hematopoiesis and growth regulation. Although cytokines interact with a variety of receptors with highly different structures and functions,1 their signaling commonly proceeds via a JAK/ STAT module.2,3 It is assumed that the diversity of responses of the JAK/STAT pathway is supported by the formation of heterodimers between different members of the STAT family.4 A typical cytokine-triggered signaling event including STAT5 is the binding of erythropoietin (Epo) to the erythropoietin receptor (EpoR), leading to activation of the preformed EpoR dimer by a conformational change.5 The cytoplasmatic tail of EpoR is associated with the janus kinase JAK2 that, following conformational change, is activated through trans-autophosphorylation and then phosphorylates a set of Tyr residues on EpoR. Phosphorylation of tyrosine at position 343 or 401 on EpoR is necessary for recruiting cytosolic STAT5 via its SH2 domain,6−8 whereby JAK2 phosphorylates recruited STAT5 at its activation motif at position Tyr-694 (STAT5A, human). © XXXX American Chemical Society

This pTyr site can then bind to the SH2 domain of a second pSTAT5 so that two pSTAT5 molecules form a dimer. Dimers of phosphorylated STAT5 translocate to the nucleus, bind to specific promoter regions, recruit cotranscription factors, and finally initiate the expression of target genes.9 In the human, mouse, and rat genomes, two closely positioned and highly similar STAT5 genes occur, which encode STAT5A and STAT5B, respectively. These STAT5 isoforms have a high sequence similarity of >90%. Studies on individual functions of the STAT5 isoforms have been carried out (for reviews, see refs 10 and 11). STAT5A knockout mice exhibit impaired mammary gland development and inhibited milk production, probably due to the important role of STAT5A in prolactindirected signaling.12 STAT5B deficiency leads to severe growth inhibition, immunodeficiency, and reduced sexual dimorphisms in human and mouse.13,14 The isoform composition of pSTAT5 dimers can influence signaling properties because individual DNA-binding specificities have been reported for the Received: July 3, 2014

A

dx.doi.org/10.1021/pr5006923 | J. Proteome Res. XXXX, XXX, XXX−XXX

Journal of Proteome Research

Article

manufacturer’s instructions. Sorted colony-forming unit-erythroid (CFU-E) cells were cultivated for 14 h in Panserin 401 (PAN-Biotech) with 50 μM 2-mercaptoethanol and 0.5 U/mL Epo (Janssen-Cilag). Prior to stimulation, CFU-E cells were washed three times and deprived of growth factors in Panserin 401 with 50 μM 2-mercapto ethanol and 1 mg/mL BSA.

homodimer of pSTAT5A as compared to the homodimer of pSTAT5B15 as well as differences in the transactivation domain (TAD).11 Due to the high sequence similarity between STAT5A and STAT5B, similar phosphorylation and random dimerization of STAT5 isoforms has been assumed.16 However, there are also reports of the preferential formation of homodimers.17 This may indicate selective dimerization between the two STAT5 isoforms, a different activation status of the isoforms, or a combination of both of these. These observations indicate that the isoform composition of pSTAT5 dimers is a possible feature to modulate the function of STAT5A and STAT5B. We studied the isoform composition of pSTAT5 dimers by a combination of quantitative mass spectrometry and mathematical modeling. As an experimental model system, we utilized BaF3-EpoR cells and activation by Epo, which is tightly coupled to JAK2/STAT5 activation and thus well suited to study pSTAT5 dimerization.



Cell Lysis, Immunoprecipitation, and 1D SDS-PAGE

Unless stated otherwise, lysis was performed with 2× 1% Nonidet P-40 (NP-40) lysis buffer to extract the cytoplasmic fraction of BaF3-EpoR cells. For Figure S3, different lysis protocols were compared for CFU-E cells. Besides NP-40 lysis, the following buffers were used: 2× RIPA buffer (100 mM Tris, pH 7.4, 300 mM NaCl, 2 mM EDTA, 2 mg/mL deoxycholate, 1 mM Na3VO4, 5 mM NaF) or 2× buffer A (10 mM Hepes, pH 7.9, 10 mM KCl, 0.1 mM EDTA, 0.1 mM EGTA, 2 mM NaF, 2 mM Na3VO4, 0.8% NP-40) and 1× buffer C (20 mM Hepes, pH 7.9, 25% glycerol, 400 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1 mM NaF, 1 mM Na3VO4) or 2× homogenization buffer 1 (20 mM Hepes, pH 7.9, 20 mM KCl, 0.2 mM EDTA, 0.2 mM EGTA, 2 mM DDT) and 1× homogenization buffer 2 (10 mM Hepes, pH 7.9, 400 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1 mM DDT). All lysis buffers were supplemented with 2 μg/mL aprotinin and 200 μg/mL AEBSF. For cytoplasmic lysis with NP-40 lysis buffer, samples were rotated at 4 °C for 20 min and then centrifuged for 10 min at 14 000 rpm and 4 °C as described.20 For whole-cell lysis with RIPA buffer (supplemented with 2% NP-40), samples were rotated at 4 °C for 20 min and sonicated prior to 10 min centrifugation at 14 000 rpm and 4 °C. For separation of cytoplasmic and nuclear fractions with buffers A and C, samples were rotated for 30 min at 4 °C and then centrifuged for 2 min at 14 000 rpm and 4 °C to extract the cytoplasmic fraction. Fifty microliters of buffer C was added to the remaining pellet followed by a 30 min incubation on ice with vortexing every 2 min. Then, 450 μL of buffer A was added, and the samples were sonicated and centrifuged for 5 min at 14 000 rpm and 4 °C. Supernatant was taken as the nuclear fraction. With homogenization buffers 1 and 2, samples were stored on ice for 20 min and to each was added 50 μL of 1× homogenization buffer 1 with 5% NP-40. Samples were centrifuged for 3 min at 14 000 rpm and 4 °C, and supernatant was taken as the cytoplasmic fraction. The pellet was washed three times with 1 mL 1× homogenization buffer 1 and then resuspended 30 μL of homogenization buffer 2. After shaking at 1400 rpm and 4 °C for 60 min, 520 μL of homogenization buffer 1 was added, and samples were sonicated. Centrifugation for 20 min at 14 000 rpm and 4 °C yielded nuclear fractions in the supernatant. The supernatants were either subjected to IP or up to 50 μg of lysates was used for immunoblotting. Protein concentrations of the lysates were determined via the BCA protein assay kit (Pierce) according to the manufacturer’s instructions. For IP, the STAT5 antibodies sc-1081 (L-20), sc-835 (C-17), and sc-836 (N-20) obtained from Santa Cruz Biotechnology were used. Upon addition of 10 μL of Protein A Sepharose CL4B (GE Healthcare), samples were rotated overnight at 4 °C. The immunoprecipitates were washed twice with 1× 1% NP-40 lysis buffer and once with 1× TNE buffer (10 mM Tris, pH 7.4, 100 mM NaCl, 1 mM EDTA pH 8.0, 100 μM Na3VO4) and then resuspended in 10 μL 2× SDS sample buffer (4% SDS, 50 mM Tris, pH 7.4, 10% glycerol, 5% 2-mercaptoethanol, 100

MATERIALS AND METHODS

Chemicals

The UPLC solvents and all buffers were prepared using UPLC−MS grade water, acetonitrile, TFA, and FA (Biosolve). For production of one-source peptide/phosphopeptide ratio standards, phosphopeptides were synthesized in-house by fluorenylmethoxy-carbonyl chemistry (13C,15N-labeled activated amino acids were obtained from Cambridge Isotope Laboratories, Tewskbury, MA, USA). To dephosphorylate these phosphopeptides quantitatively, antarctic phosphatase from New England Biolabs (M0289S) was used. Standard chemicals for cell lysis and gel electrophoresis were obtained from Sigma unless otherwise stated. Cell Cultivation and Stimulation

BaF3 cells18 were retrovirally transduced to express the erythropoietin receptor (EpoR) as described previously.19 BaF3-EpoR cells were cultivated in RPMI 1640 medium (Gibco) supplemented with 10% fetal calf serum (FCS) and 1% antibiotics (i.e., 10 000 U/mL penicillin and 10 mg/mL streptomycin sulfate) plus 10% WEHI as an IL-3 source and 1.5 μg/mL puromycin to maintain selection. Prior to stimulation, BaF3-EpoR cells were washed three times with RPMI 1640 and incubated at 37 °C for 3 h in RPMI 1640 supplemented with 1 mg/mL BSA (Sigma-Aldrich). For experiments, 1 × 107 cells were used in 250 μL per replicate and stimulated with up to 5 U/mL Epo at 37 °C and 900 rpm. Stimulation was terminated by adding 2× NP-40 lysis buffer. As a positive control for cell fractionation experiments, CFUE cells were employed. Briefly, at day 13.5 of gestation, livers of Balb/c mouse embryos were dissected from uteri of sacrificed mice and subsequently isolated. After resuspension in 500 μL of 0.3% BSA/PBS, fetal liver cells (FLCs) were passed through a 40 μm cell strainer (BD Biosciences). Cells were treated with 10 mL of red blood cell lysing buffer (Sigma-Aldrich). For negative depletion, FLCs of 40−50 livers were incubated for 30 min at 4 °C with 10 μL of rat antibodies against the following surface markers: GR1, CD41, CD11b, CD14, CD45, CD45R/ B220, CD4, CD8 (all purchased from BD Biosciences), TER119 (gift from A. Müller, Würzburg, Germany), and the rat monoclonal antibody YBM/42 (gift from S. M. Watt, Oxford, UK). Cells were washed three times in 0.3% BSA/PBS and incubated for 30 min at 4 °C with anti-rat antibody-coupled magnetic beads (BD Biosciences). MACS columns (Miltenyi Biotech) were applied for negative sorting according to the B

dx.doi.org/10.1021/pr5006923 | J. Proteome Res. XXXX, XXX, XXX−XXX

Journal of Proteome Research

Article

Figure 1. Overview of the isoforms of mouse STAT5: (A) domain map and length of STAT5A and STAT5B; (B) schematic display of the three hypotheses for possible dimer structures of pSTAT5 caused by the presence of both isoforms with differences in the C-terminal TAD domain (CC, coiled coil; SH2, src-homology 2 domain; TAD, transcriptional activation domain).

Figure 2. Concept for studying the dimerization of pSTAT5A and pSTAT5B. Starting from the same initial STAT5 system status for both isoforms, isoform-specific IPs lead to distinguishable isoform ratios and degrees of phosphorylation for hypothesizing (A) randomly formed homo- and heterodimers, (B) homodimers only, and (C) heterodimers only.

Immunoblotting was conducted in semidry chambers (GE Healthcare) on methanol-activated PVDF membranes with a pore size of 0.2 μm (Carl Roth). Blotting was performed in transfer buffer (192 mM glycine, 25 mM Tris, 0.075% SDS, 0.5 mM Na3VO4, 15% methanol) for 1 h at approximately 1.3 mA/ cm2. Proteins were reversibly stained with Ponceau Red. After blocking nonspecific antibody binding with 5% BSA diluted in TBS-T (10 mM Tris, pH 7.4, 150 mM NaCl, 0.2% Tween-20),

mM DTT, 0.01% bromophenol blue). Two IPs were pooled per lane. Proteins were separated per 1D SDS-PAGE.21 Samples were boiled at 95 °C for 2 min and subsequently loaded and separated by 10% SDS-PAGE with low bisacrylamide content (GE Healthcare) in 1× Laemmli buffer (192 mM glycine, 25 mM Tris, 0.1% SDS). Staining of gels with SimplyBlue SafeStain (Invitrogen) was performed according to the manufacturer’s instructions. C

dx.doi.org/10.1021/pr5006923 | J. Proteome Res. XXXX, XXX, XXX−XXX

Journal of Proteome Research

Article

In this study, we focused on possible isoform preferences in the formation of pSTAT5 dimers. For this purpose, we used a combination of isoform-specific IP, quantitative isoform abundance analysis, and quantification of the degree of phosphorylation by mass spectrometry (MS). The concept of the study is outlined in Figure 2. As a starting point, STAT5 is assumed to exist as a 1:1 mixture of STAT5A and STAT5B and with a degree of phosphorylation of 50% at the activation motif for both isoforms (Figure 2, left). In our experiments, we observed that dimerization is restricted to the phosphorylated forms of STAT5A and STAT5B (see Figure S1). Relative abundances of phosphorylated dimers result from distinct dimerization hypotheses. The three cases considered are randomly mixed homo- and heterodimers (Figure 2A), homodimers only (Figure 2B), and heterodimers only (Figure 2C). Depending on the dimerization hypothesis, STAT5A- or STAT5B-specific IP is shown to capture a different fraction of monomeric and dimeric forms of STAT5. Immunoprecipitated STAT5 fractions exhibit different isoform abundances and degrees of phosphorylation (Figure 2, right column). It is evident that all three dimerization models are distinguishable because they all give rise to individual isoform abundances and different degrees of phosphorylation. The analytical tools to distinguish among the dimerization hypotheses outlined in Figure 2 are introduced below.

membranes were incubated sequentially with the primary and secondary antibodies, and proteins were visualized with ECL or ECL advance western blotting detection reagents (GE Healthcare) and subsequently detected on an ImageQuant LAS 4000 (GE Healthcare). Primary antibodies were mouse anti-GAPDH (Abcam) and rabbit anti-laminA/C (Cell Signaling). Secondary antibodies were anti-mouse coupled HRP (dianova) and anti-protein A coupled HRP (BD Biosciences). In-Gel Digestion and Standard Addition

STAT5 gel bands at approximately 90 kDa were excised and cut into small pieces (feed size 1 mm). Gel pieces were destained, reduced with DTT (dithiothreitol), alkylated with IAA (iodoacetamide), and digested with 0.3 μg of trypsin in 100 mM NH4HCO3/5% acetonitrile buffer overnight.22 In-house produced one-source peptide/phosphopeptide ratio standards for STAT5A and STAT5B were added to the digests.23 Following a four-step peptide extraction performed sequentially with 100 mM NH4HCO3/5% acetonitrile, acetonitrile, 5% formic acid, and acetonitrile, the samples were concentrated in a SpeedVac (Eppendorf) and desalted with C18 Ziptips (Millipore) according to the manufacturer’s instructions. NanoUPLC−MS/MS Analysis and Data Evaluation

Samples were analyzed by nanoUPLC (nanoAcquity, Waters) coupled to an LTQ-Orbitrap XL mass spectrometer (Thermo). LC separations were performed on a 75 μm × 150 mm C18 column with a 1.3 μm particle size (Waters, Milford, MA, USA) by using a water/acetonitrile-based gradient up to 40% acetonitrile within 60 min. Survey full-scan MS spectra (m/z 350 to 2000) were acquired at resolution R = 60 000. Peptide ions were selected for LTQ-CID fragmentation (NCE = 35 V) by applying a top 6 DDA method. Quantitative evaluation of nanoUPLC−MS data was performed manually with Xcalibur 3.0.63 (Thermo). Full-scan MS spectra of corresponding peptide pairs were manually analyzed under consideration of all relevant charge states and up to 5 isotope peaks. Alternatively, quantification was performed using MaxQuant 1.4.0.3.

Relative Isoform Quantification of STAT5A and STAT5B

As experimental read-outs for this study, the degree of phosphorylation and the relative isoform abundance of STAT5A (= [STAT5A]/[STAT5A + STAT5B]) are required. Therefore, we first developed and validated a method for relative quantification of STAT5A and STAT5B by mass spectrometry. Sequence comparison of STAT5A and STAT5B (both mouse) revealed that about 60 of 800 amino acids are different (corresponding to 7.5%). Upon digestion with trypsin, about half of the fully tryptic peptides were isoform-specific. Of these specific peptides, again almost half have counterparts with high similarity in both sequence and length, so, theoretically, 10 peptide pairs are suited for relative isoform quantification by LC−ESI−MS. In our experimental data, 7 of the 10 expected fully tryptic peptide pairs were observed with good intensity. In addition, two highly similar peptide pairs with one missed tryptic cleavage site were observed with good intensity and included into the list of peptide pairs. In practice, STAT5A/B ratio quantification was performed label-free on the basis of 6 to 9 peptide pairs. Table 1 gives an overview, and in Table S1, the full sequences of the peptide pairs are listed. The relative isoform quantification was validated using recombinantly expressed standard proteins, which were independently quantified by element mass spectrometry via their selenomethionine tag.27 In addition, relative isoform

Mathematical Modeling

Linear regression analysis and χ2 statistics were computed with Microsoft Excel. We computed χ2 statistics that allow estimating the goodness of the fit. χ2 test statistics are defined as the sum of squared residuals between model and data weighted by 1/variance of the data. The test statistics are compared to a χ2 distribution with n − 1 degrees of freedom, where n denotes the number of data points under the assumption that the measurement error is normally distributed. Quantitative dynamic modeling was performed in MATLAB (Mathworks) using the D2D software package (https:// bitbucket.org/d2d-development/d2d-software).24



RESULTS AND DISCUSSION

Table 1. Amino Acid Differences and Unique Tryptic Peptides of Mouse STAT5A and STAT5Ba

Concept

The existence of the two closely related isoforms of STAT5A and STAT5B (Figure 1A) results in three possible STAT5 dimers with individual transcriptional activation domain (TAD) structures11 (Figure 1B). Differences in the TAD domain of pSTAT5 dimers are caused by the different lengths of STAT5A and STAT5B and varied sequences in the TAD domain. These TAD domain differences may be enhanced by phosphorylation of tyrosine and serine residues located there25,26 (for a summary of STAT5B, see http://atlasgeneticsoncology.org).

protein

different AAs (calcd)

unique tryptic peptides* (expt)

unique pairs of tryptic peptides* (calcd)

unique pairs of tryptic peptides (expt)

STAT5A_mouse STAT5B_mouse

65 of 793 58 of 786

20 of 45 17 of 42

10

7* + 2**

a

Both expected and observed numbers of unique peptides are listed. MW 700−4000; *, trypsin/P without missed cleavage site; **, with one missed cleavage site.

D

dx.doi.org/10.1021/pr5006923 | J. Proteome Res. XXXX, XXX, XXX−XXX

Journal of Proteome Research

Article

Table 2. Isoform Specificity Control of Three Commercially Available STAT5 Antibodiesa rel. abundance STAT5A (%)

rel. SD (%)

n

remarks

L-20

99.3

0.2

3

C-17

21.2

3.1

5

N-20

4.5

0.4

3

68.8 69.8

2.1 0.6

6 3

highly STAT5A-specific, company data sheet: recommended for STAT5A of mouse, rat and human origin preference for STAT5B, company data sheet: recommended for STAT5A and STAT5B of mouse, rat and human origin highly STAT5B-specific, company data sheet: recommended for STAT5A and STAT5B of mouse, rat and human origin no isoform preference, data reflect status in BaF3-EpoR cells no isoform preference, data reflect status in BaF3-EpoR cells

antibody

L-20 + C-17 L-20 + N-20 a

Three antibodies and two combinations of them were used for precipitation of STAT5 from lysates of BaF3-EpoR cells. Following IP, the ratios of STAT5A and STAT5B were analyzed by relative quantification of STAT5A and STAT5B by LC−ESI−MS as described in the text. Results are displayed as relative abundance of STAT5A normalized to the sum of STAT5A and STAT5B.

Table 3. Sequences of the Blocking Peptides for the Antibodies L-20, C-17, and N-20 and Their Alignment with the Sequences of Mouse STAT5A and STAT5Ba

a

bp, blocking peptide.

+ N-20 revealed identical results, with a relative STAT5A abundance of 69.3%, which represents the true STAT5 composition in BaF3-EpoR cells. On the basis of this data, the relative antibody specificities toward STAT5A and STAT5B were calculated as described in detail in the Supporting Information and are listed in Table 3. The observed isoform specificity for the N-20 antibody differed from the information given in the information sheet of the manufacturer. To obtain more detailed information on the epitopes recognized by the antibodies utilized in our study, we purchased the respective blocking peptides available from the manufacturer for the antibodies and used them for sequencing by LC−ESI−MS/MS. Alignment of the sequences obtained for the blocking peptides with the sequences of murine STAT5A and STAT5B is shown in Table 3. The very high STAT5A specificity of L-20 correlates with the full identity of the L-20 blocking peptide sequence with the STAT5A sequence, which is completely unique in this region. Also, the STAT5B preference of the N-20 and C-17 antibodies were confirmed, as both blocking peptides were identical with the STAT5B sequence but two or three amino acid differences were observed relative to the STAT5A sequence (R16H and V20A exchange in N-20; H762R, R768G, and L774Q exchange in C17).

quantification was performed manually and automatically by MaxQuant. The measured isoform abundance of each peptide pair is exemplarily indicated for six of the experiments used in our study (see Table S1 for automatic data analysis with MaxQuant and Table S2 for manual data analysis). Both data evaluation methods gave very similar results. The manual data evaluation typically resulted in relative standard deviation (SD) of 4%, compared to 8% obtained for automated evaluation. A comparison of manual and automated data analyses is displayed in Figure S7. For all experiments, both manual and automated analyses were performed and delivered comparable results. Overall, the isoform quantification data were considered to be sufficiently accurate and robust for the purpose of this study. Antibody Specificity for Murine STAT5A and STAT5B

The concept of this study requires the use of STAT5 antibodies with known isoform specificity. An isoform-specific antibody provides analytical access to subsets of STAT5 dimers and thus allows STAT5A/STAT5B dimerization behavior to be studied (Figure 2). Here, we tested three commercially available STAT5 antibodies for their isoform specificity. To this end, STAT5 was immunoprecipitated from lysates of BaF3-EpoR cells, which express both STAT5A and STAT5B. Antibodies, named L-20, C-17, N-20, and two combinations thereof (L-20 + C-17 and L-20 + N-20), were used for IP. Relative isoform quantification of STAT5A and STAT5B was performed by the LC−ESI−MS method described in the previous paragraph. The results are summarized in Table 2. The L-20 antibody preferentially precipitated STAT5A, whereas the C-17 and N20 antibodies showed a preference for STAT5B. Two combined IPs using the combination of L-20 + C-17 or L-20

Accurate Degree of Phosphorylation Measurement of STAT5

As outlined in Figure 2, quantitative determination of the degree of phosphorylation of STAT5 is a necessary parameter for analyzing the STAT5 dimer composition. For analysis of the degree of phosphorylation at the STAT5 activation motif, we E

dx.doi.org/10.1021/pr5006923 | J. Proteome Res. XXXX, XXX, XXX−XXX

Journal of Proteome Research

Article

chose a label-based method. In detail, we selected the onesource peptide/phosphopeptide standard method28,29 (for a detailed protocol, see ref 23). Because bottom-up analysis of the degree of phosphorylation is based on the molar ratio of a peptide/phosphopeptide pair, this ratio standard is a mixture of such a pair in stable isotope labeled form and with a known molar ratio. The special feature of this peptide/phosphopeptide ratio standard (one-source standard) is that it has a highly accurate molar ratio without the need for absolute quantification. Mostly, a molar peptide/phosphopeptide ratio of 1:1 is adjusted. The activation site of murine STAT5 is located in the well-detectable tryptic peptide AVDGYVKPQIK for STAT5A and AADGYVKPQIK for STAT5B (the phosphorylation site is underlined). The precision of the determination of the degree of phosphorylation using a ratio standard is very high. This is because the molar peptide/phosphopeptide ratio in the standard carries only a volumetric error, which can be kept as low as 1−2%. By this approach, we were able to separately determine the degree of phosphorylation for the two murine STAT5 isoforms.

which supports this conclusion. Isoform-specific IPs in lysates of nonstimulated BaF3-EpoR cells with the L-20 or N-20 antibody show that these antibodies are highly specific for STAT5A or STAT5B, respectively. Regression analysis shows that for all isoform-specific IPs the relative STAT5A abundance is a linear function of the phosphorylation status. Using the STAT5A-specific antibody L-20, the STAT5A abundance decreases with the increasing degree of STAT5 phosphorylation. Using the STAT5B-specific antibody N-20 or the C-17 antibody that has a preference for STAT5B, the inverse situation is observed. The common feature of both observations is that an increasing average degree of phosphorylation impairs the isoform purity of the immunoprecipitated STAT5 fraction. This trend hints toward the existence of pSTAT5A/pSTAT5B heterodimers. To investigate the composition of pSTAT5 dimers in more detail, we applied mathematical modeling. Static Mathematical Modeling of Three pSTAT5 Dimerization Hypotheses

The aim of mathematical modeling was to explain the experimental data of this study with respect to the three hypotheses of pSTAT5 dimerization as outlined in Figure 2. The best alignments between model simulations and experimental data could provide an indication of the cellular dimerization behavior of pSTAT5. First, the presence of unphosphorylated forms of STAT5 heterodimers could be excluded experimentally (see Figure S1C). The measured parameters comprise (i) the relative abundance of STAT5A in BaF3-EpoR cells (%A = 69.3%) (Table 2), (ii) the relative STAT5 isoform specificities of the employed antibodies (Table 3), and (iii) an identical phosphorylation status of STAT5A and STAT5B (see Figure S1). On the basis of this, the relative molar abundances of all monomeric and dimeric species of STAT5 were calculated for the degrees of STAT5 phosphorylation in accordance with the dimerization hypotheses considered (equations given in Table S3). In this way, graphs were obtained (three dimerization hypotheses, each combined with one of the three antibodies and a combined IP) as shown in Figure 4. On the basis of χ2 statistics (provided as Supporting Information), the trajectories of the homodimers only (χ2 = 497.5, n = 32, p = 1.64 × 10−85) and heterodimers only (χ2 = 48.9, n = 32, p = 2.17 × 10−2) models were significantly different from the experimentally obtained data. Only the homodimers and heterodimers model cannot be rejected on a statistical basis (χ2 = 26.4, n = 32, p = 0.702). These results demonstrate that the homodimers only model is not in accordance with the experimental data. The difference between the heterodimers only model and the homo- and heterodimers model is rather small. Nevertheless, χ2 statistics indicate that the scenario involving both homo- and heterodimers is the most likely explanation for our quantitative data.

Correlation between STAT5 Isoform Abundance and Degree of Phosphorylation

The concept for the analysis of pSTAT5 dimerization (Figure 2) demonstrates that the outcome of the experimental data is dependent on STAT5 isoform abundance, the degree of phosphorylation, pSTAT5 dimer formation, and the antibody specificity. To study pSTAT5 dimerization quantitatively, we performed a set of activation experiments with BaF3-EpoR cells and different doses of Epo (0−10 U/mL) for a 10 min stimulation time. Thereby, a set of data with widely differing degrees of phosphorylation was obtained, ranging from nearly zero to around 70%. For display, we selected the relative abundance of STAT5A (y axis) as a function of the average degree of phosphorylation of STAT5 (x axis) (Figure 3). We observed that the data points were grouped according to the use of the individual antibodies. The combined use of the L-20 and C-17 antibodies results in a generic IP of both STAT5A and STAT5B, so the observed STAT5A abundance in this case reflects the STAT5A abundance in BaF3-EpoR cells that is independent of the overall degree of phosphorylation,

Dynamic Mathematical Modeling of pSTAT5 Homo- and Heterodimer Formation

The comprehensive display in Figure 4 contains experimental data from a series of individual experiments performed at the 10 min time point for different doses of Epo or without Epo stimulation. To study STAT5 activation in a time-resolved manner, we set up an Epo stimulation experiment of BaF3EpoR cells for 4 h with 16 individual time points. The experiment was performed using the C-17 antibody. In the dynamic process of STAT5 activation, transport to the nucleus plays a crucial role. Although we could reliably detect

Figure 3. Stimulation experiments of BaF3-EpoR cells with different doses of Epo (0−10 U/mL) and a stimulation time of 10 min. The linear correlation between the average degree of phosphorylation and the isoform composition of STAT5 is shown for different antibodies as indicated. F

dx.doi.org/10.1021/pr5006923 | J. Proteome Res. XXXX, XXX, XXX−XXX

Journal of Proteome Research

Article

concepts about the kinetic processes by our quantitative dynamic models. The time-dependent behavior of the average degree of STAT5 phosphorylation and the STAT5A abundance in the cytoplasm are displayed in Figure 5, panels A and B, respectively. Figure 5C synthesizes both data, showing the change of the relative STAT5A abundance with the average degree of phosphorylation at the individual time points. As novel information, this graph shows that cytoplasmic STAT5 exhibits different isoform abundances before and after its maximum phosphorylation. Figure 5A−C also contains the corresponding model simulations, derived from our preferred hypothesis of random homo- and heterodimer formation of pSTAT5A and pSTAT5B using the static model described above. However, the static model fails to describe the experimental data. In particular, the static model cannot reproduce the difference in isoform abundance between the pre- and postmaximum phosphorylation (Figure 5C). Therefore, a dynamic model was developed that solves this problem by additionally considering the shuttling of pSTAT5 dimers between the cytoplasm and nucleus. The dynamic model simulations are displayed in the right part of Figure 5 in panels D−F, which otherwise contain the same experimental data as their direct counterparts on the left. It is evident that the dynamic model provides a better agreement to the experimental data (both model and experimental data are available as Supporting Information). In particular, as shown in Figure 5F, the model simulations can reproduce the observed difference in the STAT5A abundance before and after the maximum phosphorylation. The essential features of the dynamic model are as follows. STAT5 forms phosphorylated homo- and heterodimers as a function of Epo that decays exponentially.19 The dimer formation of pSTAT5 occurs very fast, so phosphorylation and dimerization are described as a one-step reaction (with rate kphos). The pSTAT5 dimers are then imported into the nucleus, dephosphorylated, and exported as unphosphorylated monomers. In this model, individual nuclear import rates for homodimers (kimp_homo) and heterodimers (kimp_hetero), and individual nuclear retention rates (knuc_homo, knuc_hetero) for these species were allowed (Figure S4A). Using this dynamic model, simulations were performed under the following scenarios. First, import rates were set equal, and individual nuclear retention rates were allowed. In this case, the difference between STAT5A abundance between pre- and postmaximum phosphorylation could be reproduced. However, the STAT5A abundance values at maximum phosphorylation did not fit (Figure S4B). In the second approach, the nuclear retention rates were set equal, and individual nuclear import rates were allowed. In this case, the STAT5A abundance values at maximum phosphorylation were in agreement. However, there was no difference in the STAT5A abundance between pre- and postmaximum phosphorylation (Figure S4C). By allowing variable import rates and individual nuclear retention rates for homo- and heterodimers, the best fit, as displayed in Figure 5F, was obtained. On this basis, a minimal model was developed (Figure S5A). Compared to the full model, the minimal model can describe the experimental data equally well (Figure S5B) and is fully identifiable (Figure S5C). It suggests that STAT5 heterodimers are retained in the nucleus during the entire time frame of the experiment (Figure S5D), whereas homodimers shuttle between the cytoplasm and nucleus. Thus, we identified distinct transport kinetics for STAT5 homo- and heterodimers as a probable reason for the measured dynamic change of relative STAT5A abundance over the course

Figure 4. Experimental data and model simulation data for the relation between the average degree of STAT5 phosphorylation and relative isoform abundance. Experimental data points are identical in all three graphs; solid lines represent model simulations for the distinct hypotheses. (A) Formation of homodimers only; (B) formation of pSTAT5A/pSTAT5B heterodimers only; (C) formation of pSTAT5 homo- and heterodimers (equations for the model simulations are summarized in Table S2).

the amount of STAT5 in the cytoplasm, the nuclear and DNAbound fractions of STAT5 in BaF3-EpoR cells were difficult to assess experimentally (Figure S3). We therefore conducted all measurements from cytoplasmic fractions and deduced G

dx.doi.org/10.1021/pr5006923 | J. Proteome Res. XXXX, XXX, XXX−XXX

Journal of Proteome Research

Article

Figure 5. Time-course experiments of STAT5 in BaF3-EpoR cells stimulated with 5 Epo U/mL. Experimental data was obtained upon STAT5 immunoprecipitation using the C-17 antibody. Experimental data (triangles connected by dotted lines) of a single experiment and model trajectories of the best fit (solid lines) are shown. Shaded areas indicate 95% confidence intervals of model simulations based on estimated errors. Data was obtained upon STAT5 immunoprecipitation using the C-17 antibody. (A) Average degree of STAT5 phosphorylation as a function of time compared to the static model; (B) STAT5A abundance as a function of time compared to the static model; (C) STAT5A abundance as a function of average degree of STAT5 phosphorylation compared to the static model; (D) average degree of STAT5 phosphorylation as a function of time compared to the dynamic model; (E) STAT5A abundance as a function of time compared to the dynamic model; (F) STAT5A abundance as a function of the average degree of STAT5 phosphorylation compared to the dynamic model. Parameters of the model were estimated according to the maximum likelihood estimation approach by minimizing −2 log(L)m, where L is the likelihood and m is the number of data points.20

molecules, as suggested by Iyer and co-workers.30 Different nuclear retention times of homo- and heterodimers are suggested by distinct sets of STAT5A and STAT5B target genes31 with presumably different DNA-binding kinetics. Also, structural differences in the TAD domain of the individual types of dimers (Figure 1) may lead to their individual nuclear retention times.

of the experiment. Two additional hypotheses tested under the assumption of equal shuttling rates for all molecules were (i) individual DNA-binding kinetics for the different dimers (Figure S6A) or (ii) the existence of specific phosphatases in the nucleus (Figure S6B). However, these models could not explain our experimental data as well as the model with distinct import and export rates. If we additionally consider distinct import rates of STAT5 (Figure S6), then models i and ii are in line with the data. This suggests that both distinct import rates and one of the following properties are necessary: distinct export rates (original model), distinct DNA-binding kinetics (model i), or different affinities to nuclear phosphatases (model ii) of STAT5 homo- and heterodimers. We hypothesize that distinct shuttling rates could be caused by isoform-specific sequence differences. A structural basis for individual import rates of STAT5 isoforms could be at position 147 (arginine in STAT5A vs threonine in STAT5B) in the coiled-coil domain that is supposed to interact with transporter



CONCLUSIONS The occurrence of the two closely related isoforms of STAT5 allowed to obtain insight into the dimerization pattern of pSTAT5 species. The experimental data were generated by label-based and -free quantitative mass spectrometry. It is shown that random dimerization of pSTAT5 species as homoand heterodimers gives the best agreement between experimental data and model simulations. Detailed analysis of a timeresolved activation experiment implies that the relative isoform abundance in pSTAT5 dimers is different in the pre- and H

dx.doi.org/10.1021/pr5006923 | J. Proteome Res. XXXX, XXX, XXX−XXX

Journal of Proteome Research



postmaximum activation phases. Mathematical modeling revealed that the change of STAT5A abundance with the average degree of phosphorylation could be reproduced only by assuming individual nuclear import rates and individual nuclear retention times for STAT5 homo- and heterodimers. The workflow introduced in this study may be directly expanded to the study of dimerization patterns of other signaling proteins present in two or more isoforms. Finally, this study highlights the ability of systematic testing of hypotheses by mathematical modeling to provide new insight into regulatory mechanisms of cell signaling.



ASSOCIATED CONTENT

Calculation of the isoform specificities of the antibodies used for immunoprecipitation toward STAT5A and STAT5B. Figure S1: Antibody specificity, isoform ratios, and phosphorylated dimers. Figure S2: Label-free quantification of the STAT5A to STAT5B isoform ratio by LC−MS using the pairs of isoformspecific peptides listed in Table S1. Figure S3: Subcellular fractionation immunoblots of BaF3-EpoR and CFU-E cells. Figure S4: Quantitative dynamic modeling of STAT5A/B activation and translocation. Figure S5: Minimal model of STAT5A/B activation and translocation. Figure S6: Hypotheses comparison by quantitative dynamic modeling. Figure S7: Comparison of STAT5A/STAT5B isoform abundances obtained by manual (Xcalibur) and automated (MaxQuant) data analysis of the same experiments shown in Tables S1 and S2. Table S1: Label-free STAT5A/STAT5B relative abundance analysis by LC−MS. Table S2: Manual quantitative evaluation of the same experiments as described in the legend of Table S1. Table S3: Predicted relation between average degree of phosphorylation and relative abundance of STAT5A for use of a single antibody and for three dimerization models. χ2 statistics, experimental data, and model file for the quantitative dynamic modeling approach are also provided. This material is available free of charge via the Internet at http://pubs.acs.org.

AUTHOR INFORMATION

Corresponding Authors

*(U.K.) Tel.: +49-6221-42 4481. E-mail: u.klingmueller@dkfz. de. *(W.D.L.) Tel.: +49-6221-42 4563. E-mail: wolf.lehmann@ dkfz.de. Author Contributions ∥

REFERENCES

(1) Paukku, K.; Silvennoinen, O. STATs as critical mediators of signal transduction and transcription: lessons learned from STAT5. Cytokine Growth Factor Rev. 2004, 15, 435−455. (2) Rawlings, J. S.; Roster, K. M.; Harrison, D. A. The JAK/STAT signaling pathway. J. Cell Sci. 2004, 117, 1281−1283. (3) Schindler, C.; Levy, D. E.; Decker, T. JAK-STAT signaling: from interferons to cytokines. J. Biol. Chem. 2007, 282, 20059−20063. (4) Delgoffe, G. M.; Vignali, D. A. A. STAT heterodimers in immunity. JAK-STAT 2013, 2, e-23060. (5) Lu, X. H.; Gross, A. W.; Lodish, H. F. Active conformation of the erythropoietin receptorrandom and cysteine-scanning mutagenesis of the extracellular juxtamembrane and transmembrane domains. J. Biol. Chem. 2006, 281, 7002−7011. (6) Damen, J. E.; Wakao, H.; Miyama, A.; Krosl, J.; Humphries, R. K.; Cutler, R. L.; Krystal, G. Tyrosine 343 in the erythropoietin receptor positively regulates erythropoietin-induced cell proliferation and Stat5 activation. EMBO J. 1995, 14, 5557−5568. (7) Gobert, S.; Chretien, S.; Gouilleux, F.; Muller, O.; Pallard, C.; Dusanter-Fourt, I.; Groner, B.; Lacombe, C.; Gisselbrecht, S.; Mayeux, P. Identification of tyrosine residues within the intracellular domain of the erythropoietin receptor crucial for STAT5 activation. EMBO J. 1996, 15, 2434−2441. (8) Klingmüller, U.; Bergelson, S.; Hsiao, J. G.; Lodish, H. F. Multiple tyrosine residues in the cytosolic domain of the erytropoietin receptor promote activation of STAT5. Proc. Natl. Acad. Sci. U.S.A. 1996, 93, 8324−8328. (9) Penta, K.; Sawyer, S. T. Erythropoietin induces the tyrosine phosphorylation, nuclear translocation, and DNA binding of STAT1 and STAT5 in erythroid cells. J. Biol. Chem. 1995, 270, 31282−31287. (10) Teglund, S.; McKay, C.; Schuetz, E.; vanDeursen, J. M.; Stravopodis, D.; Wang, D.; Brown, M.; Bodner, S.; Grosveld, G.; Ihle, J. N. Stat5a and Stat5b proteins have essential and nonessential, or redundant, roles in cytokine responses. Cell 1998, 93, 841−850. (11) Grimley, P. M.; Gong, F.; Rui, H. Stat5a and Stat5b: fraternal twins of signal transduction and transcription activation. Cytokine Growth Factor Rev. 1999, 10, 131−157. (12) Liu, X. W.; Robinson, G. W.; Wagner, K. U.; Garrett, L.; Wynshaw-Boris, A.; Henninghausen, L. STAT5A is mandatory for adult mammary gland development and lactogenesis. Genes Dev. 1997, 11, 179−186. (13) Chia, D. J.; Subbian, E.; Buck, T. M.; Hwa, V.; Rosenfeld, R. G.; Skach, W. R.; Shinde, U.; Rotwein, P. Aberrant folding of a mutant Stat5b causes growth hormone insensitivity and proteasomal dysfunction. J. Biol. Chem. 2006, 281, 6552−6558. (14) Nadeau, K.; Hwa, V.; Rosenfeld, R. G. J. STAT5b deficiency: an unsuspected cause of growth failure, immunodeficiency, and severe pulmonary disease. Pediatrics 2011, 158, 701−708. (15) Boucheron, C.; Dumon, S.; Constantino, S.; Santos, R.; Morigg, R.; Henninghausen, L.; Gisselbrecht, S.; Gouilleux, F. A single amino acid in the DNA binding region of STAT5A and STAT5B confers distinct DNA binding specificities. J. Biol. Chem. 1998, 273, 33936− 33941. (16) Liu, X.; Robinson, G. W.; Hennighausen, L. Activation of Stat5a and Stat5b by tyrosine phosphorylation is tightly linked to mammary gland differentiation. Mol. Endocrinol. 1996, 10, 1496−1506. (17) Meinke, A.; Barahamand-Pour, F.; Wohrl, S.; Stoiber, D.; Decker, T. Activation of different STAT isoforms contributing to celltype restricted signaling in response to interferons. Mol. Cell. Biol. 1996, 16, 6937−6944. (18) Palacios, R.; Steinmetz, M. IL-3-dependent mouse clones that express B-220 surface antigen, contain Ig genes in germ-line configuration, and generate B lymphocytes in vivo. Cell 1985, 41, 727−734. (19) Becker, V.; Schilling, M.; Bachmann, J.; Baumann, U.; Raue, A.; Maiwald, T.; Timmer, J.; Klingmüller, U. Covering a broad dynamic range: information processing at the erythropoietin receptor. Science 2010, 328, 1404−1408.

S Supporting Information *



Article

M.E.B. and L.A. contributed equally to this work.

Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS We are indebted to Dr. R. Pipkorn for synthesis of peptide standards and to Dr. A. Konopka for synthesis of protein standards. We acknowledge funding by the SBCancer network in the Helmholtz Alliance on Systems Biology as well as by the BMBF-funded (BMBF: German Federal Ministry of Education and Research) MedSys network LungSys, the CancerSys network LungSysII, the e:Bio network SBEpo, and the Virtual Liver network. I

dx.doi.org/10.1021/pr5006923 | J. Proteome Res. XXXX, XXX, XXX−XXX

Journal of Proteome Research

Article

(20) Bachmann, J.; Raue, A.; Schilling, M.; Boehm, M. E.; Kreutz, C.; Kaschek, D.; Busch, H.; Gretz, N.; Lehmann, W. D.; Timmer, J.; Klingmüller, U. Division of labor by dual feedback regulators controls JAK2/STAT5 signaling over a broad ligand range. Mol. Syst. Biol. 2011, 7, 516. (21) Laemmli, U. K. Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature 1970, 227, 680− 685. (22) Shevchenko, A.; Tomas, H.; Havlis, J.; Olsen, J. V.; Mann, M. In-gel digestion for mass spectrometric characterization of proteins and proteomes. Nat. Protoc. 2006, 1, 2856−2860. (23) Boehm, M. E.; Hahn, B.; Lehmann, W. D. One-source peptide/ phosphopeptide ratio standards for accurate and site-specific determination of the degree of phosphorylation. Methods Mol. Biol. 2014, 1156, 367−378. (24) Raue, A.; Schilling, M.; Bachmann, J.; Matteson, A.; Schelke, M.; Kaschek, D.; Hug, S.; Kreutz, C.; Harms, B. D.; Theis, F. J.; Klingmüller, U.; Timmer, J. Lessons learned from quantitative dynamical modeling in systems biology. PLoS One 2013, 8, e74335. (25) Kloth, M. T.; Catling, A. D.; Silva, C. M. Novel activation of STAT5b in response to epidermal growth factor. J. Biol. Chem. 2002, 277, 8693−701. (26) Weaver, A. M.; Silva, C. M. Modulation of signal transducer and activator of transcription 5b activity in breast cancer cells by mutation of tyrosines within the transactivation domain. Mol. Endocrinol. 2006, 20, 2392−2405. (27) Konopka, A.; Zinn, N.; Wild, C.; Lehmann, W. D. Preparation of heteroelement-incorporated and stable isotope labeled protein standards for quantitative proteomics. Methods Mol. Biol. 2014, 1156, 337−363. (28) Hahn, B.; Böhm, M.; Raia, V.; Zinn, N.; Möller, P.; Klingmüller, U.; Lehmann, W. D. One-source peptide/phosphopeptide standards for accurate phosphorylation degree determination. Proteomics 2011, 11, 490−494. (29) Hahn, B.; D’Alessandro, L. A.; Depner, S.; Waldow, K.; Boehm, M. E.; Bachmann, J.; Schilling, M.; Klingmüller, U.; Lehmann, W. D. Cellular ERK phospho-form profiles with conserved preference for a switch-like pattern. J. Proteome Res. 2013, 12, 637−646. (30) Iyer, J.; Reich, N. C. Constitutive nuclear import of latent and activated STAT5a by its coiled coil domain. FASEB J. 2008, 22, 391− 400. (31) Basham, B.; Sathe, M.; Grein, J.; McClanahan, T.; D’Andrea, A.; Lees, E.; Rascle, A. In vivo identification of novel STAT5 target genes. Nucleic Acids Res. 2008, 36, 3802−3808.

J

dx.doi.org/10.1021/pr5006923 | J. Proteome Res. XXXX, XXX, XXX−XXX