Profiling Using SH2 Domains
Kazuya Machida,1Christopher M. Thompson,1Kevin Dierck,3Karl Jablonowski,4Satu Ka ¨rkka ¨inen,5
Bernard Liu,4Haimin Zhang,6Piers D. Nash,4Debra K. Newman,7Peter Nollau,3Tony Pawson,8
G. Herma Renkema,5Kalle Saksela,5,9Martin R. Schiller,2Dong-Guk Shin,6and Bruce J. Mayer1,*
1Raymond and Beverly Sackler Laboratory of Genetics and Molecular Medicine, Department of Genetics and Developmental
2Department of Molecular Microbial and Structural Biology
University of Connecticut Health Center, Farmington, CT 06030, USA
3Department of Clinical Chemistry, Center of Clinical Pathology, University Medical Center Hamburg-Eppendorf, Hamburg 20246,
4Ben MayInstitute forCancer Researchandthe CommitteeonCancer Biology, TheUniversityof Chicago, Chicago,IL 60637,USA
5Institute of Medical Technology, University of Tampere and Tampere University Hospital, Tampere FIN-33014, Finland
6Department of Computer Science and Engineering, University of Connecticut, Storrs, CT 06269, USA
7Blood Research Institute, The BloodCenter of Southeastern Wisconsin, Milwaukee, WI 53201, USA
8Samuel Lunenfeld Research Institute, Mount Sinai Hospital, Toronto M5G 1X5, Canada
9Department of Virology, Haartman Institute, University of Helsinki and Helsinki University Central Hospital, Helsinki FIN-00014,
Protein tyrosinephosphorylation controls many
aspects of signaling in multicellular organisms.
One of the major consequences of tyrosine
phosphorylation is the creation of binding sites
for proteins containing Src homology 2 (SH2)
domains. To profile the global tyrosine phos-
phorylation state of the cell, we have developed
nearly the full complement of human SH2
domains. Here we provide a global view of
SH2 domain binding to cellular proteins based
on large-scale far-western analyses. We also
use reverse-phase protein arrays to generate
profiles for phosphopeptides, recombinant
proteins, and entire proteomes. As an example,
we profiled the adhesion-dependent SH2 bind-
ing interactions in fibroblasts and identified
whose tyrosine phosphorylation and binding
to SH2 domains are modulated by adhesion.
These results demonstrate that high-through-
put comprehensive SH2 profiling provides valu-
able mechanistic insights into tyrosine kinase
Protein tyrosine kinases (PTKs), proteintyrosine phospha-
tases (PTPs), and their substrates play a critical role in
regulating processes such as proliferation, differentiation,
motility, and immune responses, as well as pathological
conditions such as cancer (Hunter, 2000). Profiling the
global tyrosine phosphorylation state of cells under vari-
ous physiological conditions has therefore become an
important goal. Current efforts in this area fall into two
broad categories, identification based and detection
based (Machida et al., 2003). The first relies on sensitive
mass spectrometry (MS) for unambiguous identifications
of phosphopeptides and usually requires phosphoprotein
enrichment and large amounts of sample (Kim et al., 2005;
Kumar et al., 2007; Olsen et al., 2006). The latter approach
generally exploits the potential of antibodies that recog-
nize specific phosphorylated sites, or pan-anti-phospho-
tyrosine (anti-pTyr) antibodies. Although detection-based
methods may not result in phosphoprotein identification,
they have advantages in terms of sensitivity and through-
put. For example, reverse-phase protein microarrays cou-
pled with phosphospecific antibodies have been used to
profile immunological responses and cancer progression
(Chan et al., 2004; Paweletz et al., 2001). However, phos-
phospecific antibodies appropriate for quantitative analy-
sis are available for relatively few phosphorylated sites;
thus comprehensive profiling of tyrosine phosphorylation
is still a challenge.
Aprimary mechanism usedby thecell to interpretphos-
photyrosine-mediated signals relies on small, modular
protein domains that bind specifically to tyrosine-phos-
phorylated proteins. The Src homology 2 (SH2) domain
is the most prevalent of these modules and plays a central
role in tyrosine kinase signaling pathways (Machida and
Mayer, 2005; Pawson et al., 2001; Sadowski et al., 1986;
Yaffe, 2002). Recently, Nash and coworkers reported
that the human genome contains a total of 120 SH2
domains in 110 distinct proteins (Liu et al., 2006). The
Molecular Cell 26, 899–915, June 22, 2007 ª2007 Elsevier Inc. 899
Figure 1. Characterization of Purified SH2 Domains
(A) Solubility and pTyr dependency of GST-SH2 fusion proteins. GST-tagged SH2 domains were expressed in E. coli, and solubility was classified as
good (>50%), moderate (25%–50%), or poor (<25%) based on anti-GST immunoblotting of soluble fraction relative to whole bacteria lysate. pTyr
dependency is assessed by pull-down assay and far-western blotting (for details, see Table S1). Dendrogram is adapted from Liu et al. (2006).
900 Molecular Cell 26, 899–915, June 22, 2007 ª2007 Elsevier Inc.
High-Throughput SH2 Profiling
SH2domain isrelatively small(?100aminoacids) andcan
fold independently; hence the isolated domain can be ex-
pressed in bacteria, purified, and used for in vitro binding
assays (Mayer et al., 1991).
SH2 domains differ in their binding preferences for
specific phosphorylated ligands, resulting in specificity
in signal transduction (Pawson, 2004). The ligand selectiv-
ity of SH2 domains has been investigated in vitro using
directed phosphopeptide library screening (Songyang
et al., 1993). These studies revealed that, in addition to
phosphotyrosine, amino acids in positions from ?2 to +4
relative to phosphotyrosine contribute to high-affinity
binding in most cases; more extended contacts (?6 to +6)
have been observed in some structural studies (Hu
and Hubbard, 2005; Liu et al., 2006; Pascal et al., 1994).
Although SH2 domains have been studied extensively
both biochemically and structurally, numerous questions
remain to be addressed. For example, do all SH2 domains
bind phosphotyrosine? Do specific tyrosine-phosphory-
lated sites bind to one or many SH2 domains? Do SH2
domains in closely related protein families have similar
binding specificity? To what extent do SH2 domains
vary in the number of binding partners in the cell? To
answer such questions, it is clear that comprehensive
experimental data are needed to complement computa-
tional approaches (Joughin et al., 2005).
We first proposed using SH2 domains as the basis for
a robust, simple, and biologically relevant method to pro-
file the global tyrosine phosphorylation state (Machida
et al., 2003; Nollau and Mayer, 2001). Because SH2 do-
sites and play an important role in mediating tyrosine
kinase signaling in vivo, we reasoned that a battery of
SH2 domains could be used to probe differences in global
patterns of tyrosine phosphorylation relevant to signaling.
We therefore established a modified far-western method
(Nollau and Mayer, 2001). This approach, which we term
SH2 profiling, holds promise as a means of characterizing
changes in tyrosine phosphorylation in response to phys-
iological signals on a system-wide level, and also as a
molecular diagnostic tool for classifying clinical samples
such as tumors (Dierck et al., 2006; Machida et al., 2003).
In the present study, we employ a comprehensive,
quantitative, high-throughput phosphotyrosine profiling
strategy incorporating nearly the entire complement of
human SH2 domains. First, we have generated GST-SH2
fusions for currently known human SH2 domains. For
the 74 SH2 domain probes with sufficient solubility, we
compare their binding preferences for cellular proteins
by far-western blotting and evaluate the correlation be-
tween similarity in binding patterns and sequence similar-
ity (evolutionary distance). Next, we have developed
a high-throughput quantitative reverse-phase SH2 bind-
ing assay, dubbed the SH2 Rosette assay, which allows
a large number of binding reactions to be performed
simultaneously with high sensitivity and reproducibility.
In a two-step screening strategy, small amounts of protein
sample (whole-cell lysate) are used for initial Rosette
screening, followed by gel electrophoresis-based far-
western analysis with selected SH2 domains to confirm
screening results and set the stage for identification of
specific SH2 binding proteins of interest. This flexible
SH2 profiling approach provides an effective tool for ana-
lyzing the tyrosine phosphoproteome and complements
existing proteomic strategies.
Characterization of SH2 Domains
A global SH2 profiling method should incorporate a com-
prehensive set of SH2 domain probes. We collected
cDNAs for all presently known SH2 domains and con-
structed bacterial vectors expressing N-terminally GST-
tagged SH2 proteins. SH2 domains are highly conserved
in their overall structure; thus the domain boundaries
were readily discerned from primary sequence. To date
we have generated 110 SH2 domain constructs, repre-
senting 95% (114/120) of all human SH2 domains (104/
110 of SH2 domain-containing proteins) (Liu et al.,
2006). We categorized the solubility of GST-SH2 domains
into three classes (good, moderate, and poor) based on
small-scale protein purification experiments (Figure 1A
and see Table S1 and Table S2 in the Supplemental
Data available with this article online). Approximately half
of the GST-SH2 domains tested were highly soluble and
well behaved when expressed in E. coli, but the remainder
were less soluble. Poor solubility was particularly associ-
ated with several families of related SH2 domains, as illus-
trated in the dendrogram (Figure 1A). SH2 domains of the
uble, as also noted by others (Babon et al., 2006; Jones
et al., 2006). Jones et al. reported some success solubiliz-
ing such domains by denaturation and refolding, but we
did not try this approach.
Next we determined the phosphotyrosine dependency
of SH2 domain binding. This was evaluated by (1) GST-
SH2 pull-down assays of lysates from pervanadate-
treated or untreated cells, followed by anti-pTyr immuno-
phosphatases, thus strongly enhancing tyrosine phos-
phorylation in vivo); (2) GST-SH2 pull-down assays of
pervanadate-treated lysates in the presence/absence of
phenyl phosphate, followed by anti-pTyr immunoblotting
(phenyl phosphate specifically competes with phospho-
tyrosine for SH2 domain binding); or (3) far-western blot-
1A and 1B and Figure S1). Based on these assays, 76%
(B) Binding preferences of SH2 domains. Pervanadate-treated cell lysate (mixture of NIH 3T3, A431, HepG2-PDGFR, and MR20) was separated on
gradient gels and identical blots probed with GST-SH2 domain probes as indicated. Binding conditions and exposure times were optimized to
provide approximately equivalent total signal for each domain. Results from two independent experiments are shown.
Molecular Cell 26, 899–915, June 22, 2007 ª2007 Elsevier Inc. 901
High-Throughput SH2 Profiling
(85/112) of GST-SH2 probes tested showed clear pTyr-
dependent binding (Table S1), including the Rin2 SH2,
which contains histidine at the bB5 position instead of
the highly conserved arginine in FLVRES motif believed
to be critical for phosphotyrosine binding (Figure S1B;
Kuriyan and Cowburn, 1997; Liu et al., 2006; Machida
uble and did not display any detectable binding activity,
uated (indicated as ‘‘undefined’’ in Figure 1A). We note
that some of these poorly soluble SH2 domains, e.g.,
Stat and Socs, have been shown to bind specific tyro-
sine-phosphorylated sites in cell-based assays. The fact
that all evaluable GST-SH2s showed clear phosphotyro-
sine-dependent ligand binding confirms that the SH2 do-
main is a dedicated phosphotyrosine binding module, in
contrast to other protein binding modules such as the
tion-independent fashion (Uhlik et al., 2005).
The protein binding specificity of the 74 human SH2
probes with adequate solubility and low nonspecific bind-
ing was investigated by pull-down and far-western blot-
ting experiments (Figure 1B, Figure S1, and Table S1). In
both assays, SH2 domains displayed diverse binding pat-
terns as expected. We noted that a few SH2 probes were
positive in the pull-down assay (gave a strong signal when
probed with anti-pTyr antibody) but gave relatively weak
signals in the far-western assay. This may be due to the
detection of indirect binding partners (i.e., other members
of large pTyr-containing complexes) in the pull-down as-
say; by contrast, the far-western assay detects only direct
To assess the similarity of the far-western binding pat-
terns for different SH2 domain probes, the location and
signal intensity of each band in the blots were quantified,
and data were analyzed by a hierarchical clustering algo-
rithm to provide aquantitative measure of overall similarity
(Figure 2A). The reproducibility of binding patterns was
very high, as demonstrated by coclustering of indepen-
dentreplicate samplesforthesameSH2 domaininalmost
every case (replicate was closest match in 70/74 cases).
Coclustering of replicates also implies that, even for very
expected, the SH2 domains from closely related protein
families generally share similar, unique binding patterns;
for example, binding patterns for Nck1 and Nck2, and
for Grb2, Grap, and Grap2, are virtually identical (clusters
1 and 2, Figure 2A). Roughly a third of the SH2 domains,
including those of almost all nonreceptor tyrosine kinases,
clustered together in a group with relatively broad speci-
ficity (Figure 2A, cluster 3). However, more subtle distinc-
tions within this cluster are clearly apparent. For example,
the eight Src family kinases can be subclassified into two
distinct families (Blk,Lyn,Hck,andLck; andFyn,Yes,Fgr,
and Src) based on SH2 binding pattern (Figure 2A), which
correspond to their classification based on sequence
similarity (Figure 1A).
We also performed a global comparison of the related-
ness of SH2 domains based on primary sequence (evolu-
tionary distance) with relatedness based on binding pat-
terns (Figure 2B). Overall, the correlation between
sequence similarity and binding pattern was rather poor,
except for closely related family members. This analysis
also revealed several cases in which there was striking
discordance between binding pattern and sequence sim-
ilarity. For example, the SH2 binding patterns for the four
closely related tensin family members (tensin, tenc1,
Cten, and Tem6/Tens1) are quite distinct (see Figure 2B,
blue). There are also a few examples of distantly related
SH2 domains with very similar binding patterns, the
most notable example being the Nap4 SH2 (a member
to Src family SH2 domains (Figure 2B, green). Agreement
between clustering based on sequence similarity and
binding pattern can also be visualized by comparing the
two trees directly (Figure S2). In this analysis, SH2
domains for a number of small protein families such as
the Cbls, Fer/Fes, and Sap/Eat2 show good agreement
(Figure S2, red circles).
Development of High-Throughput SH2 Domain
We previously described the use of far-western blotting
to profile the global phosphorylation state of a sample
(Nollau and Mayer, 2001). This 1D gel electrophoresis-
based approach requires relatively large amounts of
protein, however, and quantitative analysis of binding
patterns is difficult and highly sensitive to experimental
variables such as electrophoresis conditions. In order to
increase the throughput of the assay and decrease
sample requirements without sacrificing data quality,
we envisioned a two-step screening strategy, using
quantitative reverse-phase dot blotting for initial profiling,
followed by 1D gel analysis as needed. In reverse-phase
assays, protein analytes are immobilized in the solid
phase and probes such as antibodies or modular protein
domains are applied in solution (Liotta et al., 2003; Mac-
Beath, 2002). This approach has advantages where the
concentration of analyte is low (e.g., tyrosine-phosphory-
lated proteins in whole-cell lysates), because efficient
binding can be driven by high concentrations of the
SH2 domain probe. By contrast, forward-phase assays
(where domains or antibodies are arrayed and analyte
is in solution) suffer from poor binding kinetics under
these conditions, leading to relatively low signal and
We have adopted a hybrid dot-blotting method, which
we dubbed the SH2 Rosette assay, in which multiple
samples (up to 12 or more) are manually spotted on a ni-
trocellulose filter in register with a well of a 96-well cham-
ber plate. Duplicate sample arrays are spotted in multiple
wells, allowing a maximum of 96 different SH2 domain
probes to be tested in one experiment (Figure 3A). SH2
binding and washing are performed in the chamber plate,
and binding is detected by chemiluminescence and
902 Molecular Cell 26, 899–915, June 22, 2007 ª2007 Elsevier Inc.
High-Throughput SH2 Profiling
quantified by densitometry. With this rapid and quantita-
tive approach, it is possible to profile the level of binding
to virtually all SH2 domains in the genome using minimal
amounts of protein sample. A single spot consumes only
0.1 ml (<500 ng) of sample; thus no more than 50 mg sam-
ple is required for a 96-probe binding reaction. Accuracy
of spotting and experimental variation was validated as
shown in Figure 3B. Interspot variation (accuracy of spot-
ting) and interassay variation (experimental variation)
were 12.9% and 8.5%–16.0%, respectively, for higher
protein loading (50–200 ng per spot). As expected,
slightly larger variation was observed closer to the limit
of detection (<50 ng). Detection sensitivity varies de-
pending on the specific SH2 domain and sample, but
typically is in the range of 5–50 ng of whole-cell lysate
Figure 2. Bioinformatic Analysis of SH2 Domain Binding Patterns
(A) Unsupervised hierarchical clustering analysis of SH2 domain binding preferences. Each gel lane (Figure 1B) was divided into 48 bins, signal
intensity of each bin was quantified by densitometry, and values were used as the basis for clustering.
(B) Pairwise distance values based on sequence alignment (Liu et al., 2006) (Y axis) plotted against pairwise (Euclidean) distance values based on
binding patterns (X axis) for each pair of SH2 domains. Data for Src family kinases (SFKs) versus Src family kinases (red), Src family kinases versus
Nap4 (green), and tensin family versus tensin family (blue) are highlighted.
Molecular Cell 26, 899–915, June 22, 2007 ª2007 Elsevier Inc. 903
High-Throughput SH2 Profiling
Figure 3. Strategy for High-Throughput SH2 Domain Binding Assay
a set of protein spots in separate wells of a 96-well chamber plate. After ECL-based detection and densitometric quantitation, resulting values are
used to compare different samples and SH2 probes.
904 Molecular Cell 26, 899–915, June 22, 2007 ª2007 Elsevier Inc.
High-Throughput SH2 Profiling
SH2 Profiling for Specific Phosphopeptides
To decipher cellular signaling pathways, it is important to
determine which SH2 domain(s) can bind with high affinity
to specific tyrosine phosphorylation sites. We first tested
theutility ofthe SH2 Rosette assayfor unbiased screening
both for phosphopeptides and for purified phospho-
Platelet-derived endothelial cell adhesion molecule
(PECAM-1) is an immunoglobulin superfamily member
expressed in endothelial cells, leukocytes, and platelets.
It participates in cell signaling pathways via two potential
phosphorylation sites in its cytoplasmic domain, tyrosines
663 and 686, both of which conform to the consensus for
immunoreceptor tyrosine-based inhibitory motifs (ITIMs)
(Newman and Newman, 2003) (Figure 4, top). N-terminally
biotinylated peptides corresponding to both sites were
synthesized in both unphosphorylated (Tyr663 and
Tyr686) and phosphorylated (pTyr663 and pTyr686) forms
and immobilized on membranes.
As shown in Figure 4, SH2 domain probes differed in
their ability to bind the two phosphorylated peptides, but
in no case did they bind the unphosphorylated peptides.
The binding results were consistent with previously
reported interactions. Ten SH2 proteins (Shp1, Shp2,
Src, Fyn, PLCg1, SHIP, PI3K (P85a), Grb2, STAT3, and
STAT5) have been reported to bind phosphorylated
PECAM-1 (Newman and Newman, 2003), and the SH2
Rosette assay detected all but Grb2, which is thought to
bind indirectly (Newman and Newman, 2003) (STAT
probes were not used in our experiment due to their insol-
ubility). We found that the N-terminal SH2 domain of Shp2
(Shp2N) preferentially interacted with pTyr663, while the
tandem Shp2 SH2 domain probe (Shp2NC) bound both
pTyr663 and pTyr686, consistent with a previous surface
plasmon resonance (SPR) study (Pumphrey et al., 1999).
such as the strong binding of the Csk and Sap SH2
domains to pTyr686 (Figure 4). Thus this comprehensive
approach can validate previously known interactions
and identify new ones that can be further explored by
We have previously reported that performing SH2 bind-
ing reactions under competitive conditions (in the pres-
ence of a single labeled SH2 probe and multiple unlabeled
competitor SH2 domains) can improve binding specificity
(Nollau and Mayer, 2001). To test the effects of competi-
says in the presence or absence of a cocktail of unlabeled
SH2 domains as competitors. As shown in Figure S3, the
overall binding patterns of SH2 domains to the PECAM
peptides were comparable with or without competitors.
However, competitive conditions did provide some addi-
tional information regarding those domains with the high-
est specificity (relative affinity) for a site. For example,
Shp2 SH2 probes bound well to pTyr663 even in the pres-
ence of excess unlabeled competing SH2 domains, sug-
gesting this interaction is highly favored. On the other
hand, binding of all SH2 domains tested to pTyr686 was
dramatically reduced by competition, suggesting many
SH2 domains bind this site with similar affinity.
SH2 Profiling of Recombinant Phosphoproteins
We also used the Rosette assay format to profile SH2 do-
main binding to purified recombinant proteins. Members
of the p21-activated kinase (PAK) family are important
regulators of cytoskeletal dynamics, cell motility, and
cell survival (Bokoch, 2003). N-terminal tyrosine phos-
phorylation ofPAK2has beenreportedto leadto superac-
tivation of the kinase, although the specific mechanism is
not yet known (Renkema et al., 2002). The N-terminal half
of PAK2 (amino acids 1–212, hereafter PAK2N; see
Figure 5A), which contains three tyrosines, was tyrosine
phosphorylated by expression in TKB1 bacteria, which
harbor an activated EphB1 tyrosine kinase. To identify
possible phosphorylation-dependent binding partners,
purified PAK2 proteins were immobilized and subjected
to SH2 Rosette analysis (Figure 5B).
Several SH2 domain probes bound strongly to PAK2N,
including the Sap SH2 (arrow in bar graph, Figure 5B), al-
though most bound poorly relative to the positive control
(v-Abl-transformed cell lysate). To extend these findings,
we carried out far-western analysis using wild-type
PAK2N or mutants lacking specific phosphorylation sites
(Y130F, Y139F, and/or Y194F), with SH2 probes from
Sap and its close relative, Eat2 (Figure 5C). Binding of
the Sap SH2 domain to PAK2N was completely abolished
ical for the interaction (Figure 5C, indicated by an arrow).
On the other hand, no single site was essential for binding
to the Eat2 SH2 domain. The differential dependence
of Sap and Eat2 SH2 domains on pTyr194 was further
confirmed by SPR analysis (Figure S4).
uct of the gene mutated in X-linked lymphoproliferative
syndrome (XLP) (Sayos et al., 1998); its interaction with
PAK kinases has not been reported. The binding consen-
sus of Sap and Eat2 SH2 domains has been determined
by random phosphopeptide library screening to be (T/
S)IpYxx(V/I) (Poy et al., 1999). While sequences surround-
ing pTyr130 and pTyr139 do not strictly conform to this
consensus (Figure 5A), pTyr194 matches reasonably
well. The dramatic difference between Sap and Eat2 in
their dependence on pTyr194 was not anticipated, how-
ever, and indicates that the Eat2 SH2 domain has
a much less stringent binding specificity than that of its
close relative Sap. Although the association of Sap and
(B) Validation of SH2 Rosette assay. Various amounts of v-Abl-transformed 3T3 lysate were spotted in 24 duplicates on a nitrocellulose membrane
and probed with anti-actin antibody to evaluate interspot variation (accuracy of spotting) (upper left), and three different GST-SH2 probes were used
to evaluate interassay variation (upper right). Bar graphs show values of densitometric quantitation. %CV, coefficient of variation. Error bars indicate
Molecular Cell 26, 899–915, June 22, 2007 ª2007 Elsevier Inc. 905
High-Throughput SH2 Profiling
PAKhadnotbefore beenexperimentallydocumented, itis
interesting to note that a bioinformatic approach using
binding free energy estimation and peptide sequence
analysis predicted the PAK2 Tyr194 site as a potential
Sap SH2 binding site (McLaughlin et al., 2006).
Application of Rosette Assay to Entire Proteomes
We next tested whether the SH2 Rosette assay was
sensitive enough to profile levels of SH2 binding sites in
complex protein mixtures such a whole-cell lysates. We
first analyzed lysates of cells transformed by the viral
oncogene products v-Abl, v-Src, and v-Fps. Although all
three of these activated tyrosine kinases can transform
murine fibroblasts (Figure 6A, top panel), little is known
regarding the extent of similarity in the range of sites
they phosphorylate in vivo (Kamps and Sefton, 1988).
Whole-cell protein lysates from parental NIH 3T3; NIH
3T3 stably expressing v-Abl, v-Src, or v-Fps; NIH 3T3
overexpressing the CrkI and CrkII SH2/SH3 adaptors;
and hepatocellular carcinoma HepG2 cells without or
with pervanadate treatment (HepG2-POV) were immobi-
lized, and the SH2 Rosette assay was performed
(Figure 6A). We found that cells transformed by v-Src,
v-Abl, and v-Fps were easily distinguishable on the basis
of their binding preferences for individual SH2 domains.
For example, as shown in the bar graph in Figure 6A (bot-
tom panel, asterisks), the Fer SH2 bound relatively weakly
to the v-Abl lysates, compared with v-Src and v-Fps ly-
sates; similarly, the Grb2 SH2 bound relatively weakly to
v-Src lysates, and the Emt and Nck SH2s bound relatively
Figure 4. Profiling SH2 Binding to PECAM-1 ITIM Motifs
Peptides corresponding to tyrosine 663 and 686 of human PECAM-1 were synthesized in phosphorylated (pY663 and pY686) or unmodified (Y663
and Y686) form. Indicated amount of peptides and internal positive control sample (pervanadate-treated v-Abl-expressing NIH 3T3 lysate) were
immobilized as described in the Experimental Procedures. Blocked membrane was probed with anti-phosphotyrosine antibody (anti-pTyr), avidin-
HRP (avidin), and GST-SH2 domains as indicated on the right, and signal levels for each binding reaction (summed values for multiple concentrations
of peptide) were quantified (bar graph, bottom, arbitrary units). The ‘‘relative ratio to control’’ values indicate relative binding of each probe to phos-
phopeptides compared to positive control. For example, Shp2N probe showed strong preference for pTyr663 (3.23 more binding than control).
906 Molecular Cell 26, 899–915, June 22, 2007 ª2007 Elsevier Inc.
High-Throughput SH2 Profiling
weakly to v-Fps lysates. Thus global SH2 profiling re-
vealed broad differences in the classes of tyrosine-phos-
phorylated sites present in cells transformed by different
tyrosine kinase oncogenes.
We used 1D far-western blotting to confirm and extend
the Rosette assay results. As expected, relative binding
values obtained from dot blotting accurately predicted re-
sults seen in analytical far-western blotting (Figure 6B).
Thus the SH2 Rosette assay provides a rapid, high-
throughput method to identify differences among com-
plex lysates that can then be subjected to more detailed
experimental analysis. For example, one could explore
whether the relatively low level of Grb2 SH2 binding sites
in Src-transformed cells versus Abl- and Fps-transformed
cells reflects differences in the levels or mechanisms of
activation of the Ras pathway in the different cell lines.
SH2 Rosette Assay for Profiling
Adhesion-Dependent Cell Responses
To test whether this approach could be used to profile
compared the SH2 binding profiles of adherent versus
Figure 5. Profiling SH2 Binding to Recombinant Phosphoproteins
(A) Schematic representation of PAK2 proteins used in the experiment. Amino acid sequences surrounding Tyr130, Tyr139, and Tyr194 are shown at
(B) SH2 Rosette screening results for PAK2N. Purified PAK2N WT protein and positive control cell lysate (v-Abl-expressing 3T3 cells) were immobi-
lized, and SH2 Rosette assay was performed. Quantified values (arbitrary units) for each probe are shown. Strong binding of Sap SH2 domain to
PAK2N WT is indicated by arrow.
(C) Binding of Sap SH2to PAK2N depends onpTyr194. Equal amounts ofWT and mutant PAK2N proteins wereseparated bySDS-PAGE,transferred
to membranes, and probed with indicated GST fusion probes or anti-phosphotyrosine antibody (pTyr). (?) indicates unphosphorylated (BL21-
derived) protein; (+) indicates phosphorylated (TKB1-derived) protein.
Molecular Cell 26, 899–915, June 22, 2007 ª2007 Elsevier Inc. 907
High-Throughput SH2 Profiling
908 Molecular Cell 26, 899–915, June 22, 2007 ª2007 Elsevier Inc.
High-Throughput SH2 Profiling
suspended cells. For most adhesion-dependent cells,
e.g., fibroblasts and epithelial cells, attachment to a solid
substrate is essential for cell proliferation. When cultured
in suspension, such cells undergo growth arrest often
leading to apoptosis (termed anoikis); in contrast, tumor
cells can survive and continue to proliferate. Several non-
are activated upon cell adhesion and mediate down-
stream signaling, including proliferation and survival
signals (Giancotti and Tarone, 2003). To discern SFK-
dependent phosphorylation events stimulated by cell
adhesion, we utilized SYF cells, which are knockout-
derived mouse embryonic fibroblasts lacking the Src,
Yes, and Fyn SFKs, along with SYF cells rescued by re-
expression of c-Src (SYF-Src).
Cells were cultured in suspension or adhesion condi-
tions for 45 min or 24 hr, and the levels of tyrosine-phos-
phorylated binding sites for each SH2 domain were deter-
mined by Rosette assay (Figure 7A and Figure S5). The
overall tyrosine phosphorylation level of SYF cells (as-
sessed by pTyr antibody) was much lower compared to
SYF-Src cells under all conditions, indicating the predom-
inant role of SFKs in fibroblast cells. Adherent SYF-Src
cells had elevated phosphorylation levels compared to
phorylation occurs upon cell adhesion. Almost all SH2 do-
ing the global increase in phosphotyrosine. Particularly
dramatic adhesion-dependent binding increases of more
than 2-fold were seen for roughly one-fourth (11/45) of
the SH2s analyzed (Figure 7A, asterisks).
pendent on adhesion were likely to play an important role
in adhesion-mediated signaling, and we further investi-
gated these by far-western analysis (Figure 7B). Different
SH2 domains exhibited diverse patterns of binding to cel-
awiderangeof phosphorylated proteins, closelymirroring
the overall anti-pTyr pattern. On the other hand, the Cten
and Csk SH2 domains each bound in an adhesion-depen-
dent manner to a single predominant band of ?130 kDa
and ?70–80 kDa, respectively.
We next attempted to identify the adhesion-dependent
SH2 binding proteins using antibodies against proteins
previously implicated in adhesion. Crk-associated sub-
strate (p130Cas) and focal adhesion kinase (FAK) (both
?130 kDa) and paxillin (?70 kDa) were immunoprecipi-
tated from lysates of adherent or suspended cells, and
their overall tyrosine phosphorylation levels were deter-
mined with anti-phosphotyrosine antibody (Figure 7C). In
each case, phosphorylation was enhanced by adhesion.
The SH2 binding specificity for p130Cas, FAK,and paxillin
was then determined by probing identical membranes
with individual SH2 domains (Figure 7D and data not
shown). These results are summarized in Figure 7E. Inter-
estingly, the binding patterns of SH2s may be classified
into five groups according to their ability to bind to these
three proteins. Some SH2 domains, such as Fyn, Src,
and Arg, bind strongly to all three proteins upon adhesion,
while other SH2 domains are much more specific; for
example Nck and Cten bind only p130Cas, and Csk binds
only paxillin. Taken together, the SH2 Rosette assay and
subsequent analyses provide insight into the network of
interactions between SH2 domain-containing proteins
and focal adhesion proteins that are likely to play a critical
role in adhesion.
SH2 Profiling and Phosphoproteomics
Given the importance of signaling mediated by tyrosine
phosphorylation, there is great interest in strategies to de-
fine or profile the global state of tyrosine phosphorylation
in the cell. Many tyrosine phosphorylation sites exert their
biological activity through binding to other proteins
containing SH2 domains, so by characterizing the SH2
domain binding activity of a cell we can very efficiently
capture information relevant to the activation state of sig-
naling pathways in that cell. In this report, we describe
a flexible, high-throughput, quantitative SH2 profiling
ple (including unpurified total cell lysate) and can serve as
a first-step discovery tool to highlight tyrosine phosphory-
lation sites of potential interest that can then be further
investigated by hypothesis-driven experimentation.
It is important to note that quantification of binding sites
for the entire complement of SH2 domains, or SH2 profil-
ing, provides two quite distinct kinds of information: (1) in-
formation that can be used to assign samples into classes
that may be correlated with biological activities; and (2)
information regarding the SH2 protein-phosphoprotein
interactions that are likely to occur in vivo. Regarding the
first type of information, we and others have proposed,
for example, that SH2 binding profiles may be useful as
a molecular diagnostic tool for classifying tumor cells
and thereby predicting clinical outcomes (Dierck et al.,
2006; Nollau and Mayer, 2001; Yaoi et al., 2006). For this
purpose, the quantitative levels of binding sites for
Figure 6. SH2 Profiling of Oncogene-Transformed Cell Lines
(A) Analysis of PTK-expressing NIH 3T3 cellular lysates by SH2 Rosette assay. Left panels, photomicrographs showing morphology of NIH 3T3 cells
expressing v-Abl, v-Src, and v-Fps PTKs. Middle panel, 400 ng (spot numbers 1–12) or 40 ng (spot numbers 13–24) of lysates from the following cells
were probed with HRP-labeled GST-SH2 domains noted on right: NIH 3T3; NIH 3T3 stably expressing v-Abl, v-Src, v-Fps, CrkI, or CrkII; HepG2;
pervanadate (POV)-treated HepG2 (HepG2-POV); and positive control (v-Abl-3T3-POV); or lysates were probed with antibodies for phosphotyrosine
(anti-pTyr) or actin (anti-actin). Lower panel, average signal of duplicate spots was quantified (arbitrary units). Asterisks indicate probes that strongly
discriminate among different PTK-transformed cell lines.
(B) Confirmation of SH2 Rosette results by far-western blotting. 1D gel-based far-western blot, and dot blotting-based result with quantitation, are
shown for each SH2 probe along with GST control and anti-pTyr. Error bars indicate the SEM.
Molecular Cell 26, 899–915, June 22, 2007 ª2007 Elsevier Inc. 909
High-Throughput SH2 Profiling
910 Molecular Cell 26, 899–915, June 22, 2007 ª2007 Elsevier Inc.
High-Throughput SH2 Profiling
different SH2 domains serve only as a means of discrimi-
nating one class of samples from another (class discovery
and class assignment). A particular quantitative pattern of
SH2 binding may correlate, for example, with response to
a particular therapy or risk of recurrence in certain classes
of cancers. Although SH2 profile-based classification is
likely to reflect actual biological differences in signaling
state, SH2 binding data need not correspond to the actual
physical interactions that occur in vivo. For example,
quantifying binding sites for the SH2 domain of a protein
that is not expressed in a tumor cell under analysis might
still provide useful information for classifying that tumor.
Global SH2 binding patterns can also illuminate actual
in vivo binding interactions and signaling pathway activity,
thereby providing valuable information on signaling at the
systems level. The signaling output of a tyrosine kinase
pathway depends on a number of factors, including the
specific protein sites phosphorylated, the absolute num-
distribution, and the identity and local concentration of
potential SH2-containing binding partners for those sites.
Comprehensive SH2 profiling provides information about
the overall concentration of binding sites for different
SH2 domains, making it possible to predict what interac-
tions are likely to occur under a given set of conditions.
More importantly, changes in SH2 binding patterns pro-
vide information about which interactions are likely to
change most dramatically upon signaling. For example,
we have examined binding of SH2 domains to proteins
from SYF-Src cells in adherent versus suspended condi-
tions (Figure 7A). We find that for roughly 25% of the
SH2 domains tested, adherent cell lysates had much
higher levels of binding sites compared to suspended
cell lysates. By focusing on those domains for which bind-
ing activity changed most dramatically, we rapidly identi-
fied a number of potential protein interactions that are
dependent on, and therefore likely play a functional role
in, the adhesion response. For example, we show that
the Csk SH2 domain binds in a highly specific fashion to
paxillin only in adherent cells. Although this interaction
had been described previously (Sabe et al., 1994), only
through comprehensive SH2 profiling analysis was the
remarkable specificity and adhesion dependence of this
interaction, compared to all other potential SH2-pTyr
interactions, fully apparent.
Binding Properties of SH2 Domains
properties of SH2 domains. SH2 and PTB domains are the
major known binding modules for tyrosine-phosphory-
lated proteins. PTB domains are entirely distinct from
SH2 domains on the basis of primary sequence and struc-
ture, and only 25% of known PTB domains are thought to
bind in a phosphorylation-dependent manner (Uhlik et al.,
binding activity to assay, we did not encounter any exam-
ple in which binding to total cellular proteins was not
enhanced by tyrosine phosphorylation. These results
indicate that the SH2 domain, in contrast to the PTB
domain, represents a bona fide, dedicated phosphotyro-
sine recognition module. Of course, this analysis does
not exclude the possibility that some SH2 domains may
bind tightly to specific unphosphorylated proteins, as
reported in several cases (Li et al., 1999; Mahajan and
Earp, 2003; Pero et al., 2002).
We have determined binding profiles of 74 human SH2
domain probes using far-western analysis of pervana-
date-treated cell lysates and quantified the relatedness
of these binding patterns by hierarchical clustering. This
analysis provides a comprehensive comparison of the
binding patterns of SH2 domains for physiological mix-
tures of tyrosine-phosphorylated cell proteins. One ad-
vantage to the far-western approach used here is that it
accurately reflects both the strength of binding to, and
the relative abundance of, physiologically relevant binding
partners, i.e., those found in actual cell lysates. This is
quite distinct from (and complementary to) strategies
using oriented random peptide libraries or peptide arrays
to identify the preferred peptide binding sites for SH2
domains (Hwang et al., 2002; Songyang et al., 1993).
Such experiments reveal the optimal binding sequence
but do not provide information about actual protein bind-
ing sites present in cells. SH2 pull-down/MS approaches
can identify specific SH2 binding sites in a sample (Bla-
goev et al., 2003) but provide little information on which
sites are most abundant in that sample. By combining
data obtained through each of these approaches, along
with expression data for SH2-containing proteins and po-
tential partners, itwillbe possible to construct anaccurate
and comprehensive map of all SH2-mediated interactions
in the cell under different physiological conditions.
Figure 7. SH2 Profiling of SYF Cells in Adherent and Suspended States
SYF cells or c-Src reconstituted SYF cells (SYF-Src) were cultured in adhesion or suspension conditions for 45 min (see Figure S5 for details).
(A) Quantitative SH2 Rosette binding data (arbitrary units) for adherent (white bar) or suspended (black bar) SYF-Src cells. Asterisks indicate more
than 2-fold higher binding in adherent versus suspended cells. Error bars indicate the SEM.
(B) Far-western blotting with Src, Csk, Nck, Crk, p85a, and Cten SH2 domain probes, and immunoblotting with anti-pTyr and anti-tubulin. A and S
denote adherent and suspended cells, respectively.
antibodies. Lysates before or after IP (indicated as pre- and post-IP, respectively), and precipitates (IP), were immunoblotted with anti-phosphotyr-
osine antibody (upper panel) or antibodies for Cas, FAK, and paxillin (lower panels).
(D) Identical filters to those in (C) were probed with SH2 domains of Src, Csk, Nck, Crk, p85a, and Cten.
(E) SH2 domain probes were classified into five groups according to their ability to bind p130Cas, FAK, and paxillin in adherent SYF-Src cells ([D] and
data not shown).
Molecular Cell 26, 899–915, June 22, 2007 ª2007 Elsevier Inc. 911
High-Throughput SH2 Profiling
The comprehensive far-western binding data we have
obtained enables us to compare the overall binding pat-
terns for different families of SH2 domains. In general,
we find that SH2 domains of nonreceptor tyrosine kinases
recognize a broad spectrum of binding sites in the cell,
while other SH2 domains, particularly those from adaptor
proteins such as the Grb2 family, often recognize a more
limited set of phosphoproteins. This is consistent with
a general role for SH2 domains in tyrosine kinases in bind-
ing to a wide range of kinase substrates and thereby
facilitating their sustained, processive phosphorylation
(Duyster et al., 1995; Mayer et al., 1995; Songyang et al.,
1995). SH2 domains of adaptors, on the other hand, are
likely to have evolved to mediate more specific protein
complexes (Pawson, 2004).
We have also compared the similarity of SH2 domains
by sequence alignment with their similarity based on bind-
ing profile (Figure 2B). Except for closely related protein
families, the correlation between evolutionary distance
and binding pattern is weak. This is not surprising, be-
cause NMR and X-ray crystal structures have shown
that only a few residues of the SH2 domain directly con-
tact the ligand and determine specificity; thus for more
distantly related domains, most of the sequence differ-
ences are irrelevant to binding. We note, however, that it
should be possible to use our binding data to identify
those residues that correlate most strongly with binding
specificity, which are likely to directly participate in pep-
tide binding. We performed a pilot experiment in which
we aligned SH2 domains based only on residues pre-
dicted from structural studies to contact ligand. In this
analysis, NAP4 and other SOCS family members clus-
tered relatively close to the Src family (data not shown),
consistent with our binding studies showing that the
NAP4 and Src family SH2s have very similar binding pro-
files (Figure 2). Songyang et al. previously proposed to
classify SH2 domains into four groups based on the
‘‘specificity determining’’ bD5 residue (Songyang et al.,
1993). We found no obvious correlation, however,
between these four groups and our SH2 binding-based
clusters (data not shown).
Another important aspect of binding specificity is the
extent to which there is a one-to-one correspondence be-
tween a particular phosphorylated site and a particular
SH2 domain. This is an important question for under-
standing the ‘‘wiring’’ of signaling networks. Currently,
there is often an implicit assumption that a particular
phosphorylation site binds only to a single SH2 domain,
despite the relatively limited binding specificity of SH2
domains and the relatively large number of SH2 proteins
expressed inthecell.Ourdata indicatethatwhile specific-
ity varies with the site, in some cases many SH2 domains
can compete on roughly equal terms for binding to the
same sites.This was mostspecifically addressed bycom-
petitive binding assays in the case of specific PECAM
phosphorylation sites (Figure S3). Fifteen SH2 domains
bound with reasonably high affinity to the Tyr663 and/or
Tyr686 sites, and in virtually all cases binding of any single
SH2 domain was strongly inhibited by a cocktail of unla-
beled SH2 domains, suggesting multiple domains bind
with similar affinities. Only in one case, the binding of the
Shp2 SH2(N) to the Tyr663 site, was binding largely resis-
tant to inhibition by competition. Thus we can conclude
that binding of Shp2 SH2(N) to this site would be strongly
favored in vivo. In the case of Tyr686, which SH2 protein
binds in vivo will be strongly affected by the local concen-
trations and availability of the competing domains.
Features of the SH2 Rosette Assay
To generate comprehensive, quantitative, high-through-
put SH2 domain binding data using minimal amounts of
sample, we adopted a reverse-phase array (dot-blotting)
format. A variety of methodologies have been developed
recently to profile the binding of SH2 domains to tyro-
sine-phosphorylated samples, and a detailed discussion
of the comparative advantages and disadvantages of
the different SH2 profiling platforms is presented in the
Supplemental Data. An important practical difference be-
tween the Rosette method and forward-phase methods
(where SH2 domains are arrayed on a surface or coupled
to coded beads, and the analyte is in solution) is in their
binding kinetics. Forward-phase assays work best for pu-
rified analytes such as synthetic phosphopeptides, where
the high concentrations needed to drive efficient binding
to immobilized SH2 domains can be achieved. This ap-
proach can also provide the dissociation constants for
binding of the analyte to each SH2 domain in the array
(Jones et al., 2006). Reverse-phase assays, however,
are better for analyzing complex samples containing rela-
tively low amounts of phosphorylated targets, such as un-
fractionated cell lysates. In this case, high concentrations
of the SH2 domains in solution can be used to drive bind-
ing to the immobilized sample.
Two of the strengths of the SH2 Rosette approach are
its wide applicability and its sensitivity. We have shown
that many types of samples including whole-cell lysates
can be profiled, and the number of samples and of SH2
domains per assay is flexible and easily adjusted. The as-
say uses the same detection methodology and reagents
as 1D far-western-based SH2 profiling (Nollau and Mayer,
as anti-phosphotyrosine immunoblotting, depending on
the specific SH2 domain and sample (Nollau and Mayer,
2001). We show here that sensitivity of the SH2 Rosette
assay is more than sufficient to profile lysates of cells
with relatively normal levels of tyrosine phosphorylation,
suchasSYF-Srccells(Figure 7).The factthatSH2 domain
probes are fused to GST, which dimerizes in solution, and
are detected by binding to highly oligomeric glutathione-
HRP conjugate, likely increases assay sensitivity through
avidity effects as earlier noted for far-western blotting
(Nollau and Mayer, 2001).
We were initially concerned that the quantitative dot-
blotting format might result in loss of specificity compared
with 1D far-western blotting, but our results indicate that
this is not a significant obstacle. We found the overall
912 Molecular Cell 26, 899–915, June 22, 2007 ª2007 Elsevier Inc.
High-Throughput SH2 Profiling
signal-to-noise ratio to be similar for dot-blotting and 1D
ble (Figure 6B). And although some differences in binding
patterns will inevitably be obscured in dot-blotting com-
levels for two samples are similar), this disadvantage is
any potentially interesting results can easily be subjected
to far-western analysis in a two-step screening strategy.
Although this format is designed for highly sensitive
quantification of SH2 binding sites, as opposed to the
direct identification of SH2 ligands, if identification of spe-
cific binding partners is needed this can be accomplished
through additional analysis. We have shown that binding
partners can be identified in some cases by antibody-
based screening of candidate proteins (Figure 7D). Alter-
natively, unbiased identification of proteins is possible
using affinity purification and MS.
Taken together, the strategy of comprehensive screen-
ing by SH2 Rosette assay in combination with 1D analysis
is a flexible, sensitive, and specific approach to profile the
global tyrosine phosphorylation state. Because it is tech-
amounts of unpurified sample, it offers an unprecedented
opportunity to analyze the tyrosine phosphoproteome in
a wide variety of research and clinical applications.
Expression and Purification of Recombinant Proteins
cDNA clones for the human complement of SH2 domains (Liu et al.,
2006) were purchased from ATCC except for those cloned by RT-
PCR from a mouse cDNA library (Lnk, Sck, Syk, and Txk) or as noted
in the Acknowledgments. SH2 domains were cloned into pGEX fusion
vectors (Amersham Pharmacia) and verified by sequencing. GST-SH2
fusion proteins were purified as previously described and stored in
small aliquots at ?80?C (Mayer et al., 1991; Nollau and Mayer, 2001).
Detailed information for each construct is available in Table S2. For
PAK2N-H6 proteins, amino acids 1–212 of the PAK2 tyrosine mutants
(Renkema et al., 2002) were subcloned into the pET23d vector (Clon-
tech). Expression was induced in BL21(DE3)pLysS cells or TKB1 cells
(both Stratagene) and the proteins were purified using Ni-NTA agarose
beads (QIAGEN) according to the manufacturer’s instructions.
To generate a broad-spectrum tyrosine-phosphorylated cell lysate for
comparing SH2 binding patterns (Figure 1B), equal amounts of lysate
from pervanadate-treated NIH 3T3, HepG2, A431, and MR20 cells
were combined. As a control, lysates of each cell line were prepared
intheabsence of vanadate,combined,and treated with tyrosine phos-
phatase PTP-1B for 1 hr at room temperature (RT). Proteins were pre-
cipitated with 3 vol cold ethanol overnight, redissolved in SDS sample
buffer (180 mM Tris-HCl [pH 6.8], 30% glycerol, 6% sodium dodecyl-
sulfate [SDS], 15% b-mercaptoethanol, and 0.03% bromophenol
blue), and boiled for 3 min prior to gel electrophoresis or dot blotting.
More details of cell culture and lysis conditions are available in the
online Supplemental Data.
Characterization of SH2 Binding Profiles
GST-SH2 domains were labeled with horseradish peroxidase-conju-
gated glutathione (GSH-HRP) and far-western blotting was performed
as previously described (Nollau and Mayer, 2001). Conditions were in-
dividually optimized for each probe (final SH2 probe concentrations
SH2 binding patterns (Figure 1B), identical blots of pervanadate-
treated mixed lysate were prepared using precast polyacrylamide
gels (NuPAGE 4%–12%, Invitrogen) and a multigel apparatus (Xcell6
unit, Invitrogen). Replicate experiments for each domain were per-
formed on different days using different gels and membranes. Signals
kinElmer) and exposed to X-ray films. Developed films were scanned,
and signal intensities of whole lanes were quantitated by densitometry
roughly equivalent signal levels were selected, and when necessary
image sizes were adjusted to align precisely, using images of the
same filters reprobed with anti-pTyr as a reference. Aligned gel lanes
were sliced into 48 bins, and signal in each bin was quantified. Each
densitometry-quantified protein lane was projected as a vector into
a high-dimensional space, and the Euclidean distance between these
vectors was calculated as the pairwise binding pattern distance.
Hierarchical clustering analysis was performed using Cluster v3.0
(Eisen et al., 1998) with average linkage clustering and absolute corre-
lation values. The result was visualized using Java TreeView v1.1.0
Sequence Alignment Quantification
The SH2 domain alignment was generated as described (Liu et al.,
2006). Briefly, the 120 SH2 domain borders determined by SMART/
Pfam were extended in the N-terminal and C-terminal ends by 10–15
amino acids. The sequences were analyzed by ClustalW using Bioedit
or Mega with settings containing a full multiple sequence alignment
with parameters set at pairwise gap open, 0; extend, 0; multiple align-
ment gap open, 1; extend, 0. The ClustalW alignment was run in
addition with settings including neighbor joining tree with bootstrap
settings at 1000. The protein distance matrix (pairwise) was compiled
through the distance matrix software from the ClustalW alignment
integrated in the Bioedit/Mega software package.
Peptides were immobilized to gelatin-coated nitrocellulose mem-
branes by 4% paraformaldehyde treatment as described (Too et al.,
1994). Lysates were spotted onto nitrocellulose membranes (BA83,
Schleicher & Schuell BioScience) at 0.1 ml per spot using gel loading
tips guided by atransparent template on a light box. Dried membranes
and 0.05% Tween-20), and blocked for 1 hr at RT in TBST containing
10% nonfat milk, 1 mM Na3VO4, and 1 mM EDTA. Each GST-SH2
probe was labeled with GSH-HRP and applied to a separate well of
a 96-well chamber plate (MBA96, Neuro Probe) assembled with a
blocked membrane. After 1 hr of incubation at RT, chambers were
separately washed three times with TBST, and then the entire filter
was washed for 15 min.
Supplemental Data include supplemental text, Supplemental Experi-
mental Procedures, Supplemental References, five figures, and three
tables and can be found with this article online at http://www.
We thank the following colleagues for generously providing cDNAs: S.
Sugano (Bks and Chimerin2), H. Band (CblA), G. Mills (Emt), J. O’Shea
(Jak3), R. Goitsuka(Mist), B. Vanhaesebroeck (p85b),T. Katada and K.
Saito (Rin2 and Rin3), S. Shoelson (SH2B), M. Welsh (Shb), Y. Kawachi
(PTK70), H. Mano and K. Oshima (Tec), and Kazusa DNA Research
Institute (FLJ00138). We also thank M. Lalande and M. Landers for
providingamousecDNA libraryand M.Hansenfor helpfulsuggestions
Molecular Cell 26, 899–915, June 22, 2007 ª2007 Elsevier Inc. 913
High-Throughput SH2 Profiling
on cDNA database searches. This work was partially supported by
grants from the NIH (CA107785) and Department of Defense
(DAMD17-03-1-0540) (to B.J.M.). G.H.R was supported by grants
from the Academy of Finland (202423) and the Medical Research
Council of Tampere University Hospital (9F062). S.K. was supported
by the Tampere Graduate School in Biomedicine and Biotechnology.
K.S. was supported by grants from the Academy of Finland, Medical
Research Council of Tampere University Hospital, Medical Research
Council of Helsinki University Hospital, and the Sigrid Juselius
Foundation. M.R.S. was supported by the NIH (K01 MH65567).
Genome Canada, through the Ontario Genomics Institute, supported
work in T.P.’s lab.
Received: January 26, 2007
Revised: April 16, 2007
Accepted: May 17, 2007
Published: June 21, 2007
Babon, J.J., McManus, E.J., Yao, S., DeSouza, D.P., Mielke, L.A.,
Sprigg, N.S., Willson, T.A., Hilton, D.J., Nicola, N.A., Baca, M., et al.
(2006). The structure of SOCS3 reveals the basis of the extended
SH2 domain function and identifies an unstructured insertion that reg-
ulates stability. Mol. Cell 22, 205–216.
Blagoev, B., Kratchmarova, I., Ong, S.E., Nielsen, M., Foster, L.J., and
Mann,M.(2003).Aproteomicsstrategy toelucidatefunctional protein-
Bokoch, G.M. (2003). Biologyof the p21-activated kinases. Annu. Rev.
Biochem. 72, 743–781.
Chan, S.M., Ermann, J., Su, L., Fathman, C.G., and Utz, P.J. (2004).
Protein microarrays for multiplex analysis of signal transduction path-
ways. Nat. Med. 10, 1390–1396.
Dierck, K., Machida, K., Voigt, A., Thimm, J., Horstmann, M., Fiedler,
W., Mayer, B.J., and Nollau, P. (2006). Quantitative multiplexed profil-
ing of cellular signaling networks using phosphotyrosine-specific
DNA-tagged SH2 domains. Nat. Methods 3, 737–744.
Duyster, J., Baskaran, R., and Wang, J.Y. (1995). Src homology 2
domain as a specificity determinant in the c-Abl-mediated tyrosine
phosphorylation of the RNA polymerase II carboxyl-terminal repeated
domain. Proc. Natl. Acad. Sci. USA 92, 1555–1559.
Eisen, M.B., Spellman, P.T., Brown, P.O., and Botstein, D. (1998).
Cluster analysis and display of genome-wide expression patterns.
Proc. Natl. Acad. Sci. USA 95, 14863–14868.
Giancotti, F.G., and Tarone, G. (2003). Positional control of cell fate
through joint integrin/receptor protein kinase signaling. Annu. Rev.
Cell Dev. Biol. 19, 173–206.
Hu, J., and Hubbard, S.R. (2005). Structural characterization of a novel
Cbl phosphotyrosine recognition motif in the APS family of adapter
proteins. J. Biol. Chem. 280, 18943–18949.
Hunter, T. (2000). Signaling—2000 and beyond. Cell 100, 113–127.
Hwang, P.M., Li, C., Morra, M., Lillywhite, J., Muhandiram, D.R.,
Gertler, F., Terhorst, C., Kay, L.E., Pawson, T., Forman-Kay, J.D.,
and Li, S.C. (2002). A ‘‘three-pronged’’ binding mechanism for the
SAP/SH2D1A SH2 domain: structural basis and relevance to the XLP
syndrome. EMBO J. 21, 314–323.
Jones, R.B., Gordus, A., Krall, J.A., and MacBeath, G. (2006). A quan-
titative protein interaction networkfor the ErbB receptors using protein
microarrays. Nature 439, 168–174.
Joughin, B.A., Tidor, B., and Yaffe, M.B. (2005). A computational
method for the analysis and prediction of protein:phosphopeptide-
binding sites. Protein Sci. 14, 131–139.
Kamps, M.P., and Sefton, B.M. (1988). Most of the substrates of onco-
genic viral tyrosine protein kinases can be phosphorylated by cellular
tyrosine protein kinases in normal cells. Oncogene Res. 3, 105–115.
Kim, J.E.,Tannenbaum, S.R.,and White, F.M.(2005).Globalphospho-
proteome of HT-29 human colon adenocarcinoma cells. J. Proteome
Res. 4, 1339–1346.
Kumar, N., Wolf-Yadlin, A., White, F.M., and Lauffenburger, D.A.
(2007). Modeling HER2 effects on cell behavior from mass spectrom-
etry phosphotyrosine data. PLoS Comput. Biol. 3, e4.
Kuriyan, J., and Cowburn, D. (1997). Modular peptide recognition do-
mains in eukaryotic signaling. Annu. Rev. Biophys. Biomol. Struct. 26,
Li, S.C., Gish, G., Yang, D., Coffey, A.J., Forman-Kay, J.D., Ernberg, I.,
Kay, L.E., and Pawson, T. (1999). Novel mode of ligand binding by the
SH2 domain of the human XLP disease gene product SAP/SH2D1A.
Curr. Biol. 9, 1355–1362.
Liotta, L.A., Espina, V., Mehta, A.I., Calvert, V., Rosenblatt, K., Geho,
D., Munson, P.J., Young, L., Wulfkuhle, J., and Petricoin, E.F., III.
(2003). Protein microarrays: meeting analytical challenges for clinical
applications. Cancer Cell 3, 317–325.
Liu, B.A., Jablonowski, K., Raina, M., Arce, M., Pawson, T., and Nash,
P.D. (2006). The human and mouse complement of SH2 domain pro-
teins—establishing the boundaries of phosphotyrosine signaling.
Mol. Cell 22, 851–868.
MacBeath, G. (2002). Protein microarrays and proteomics. Nat.Genet.
Suppl. 32, 526–532.
Machida, K., and Mayer, B.J. (2005). The SH2 domain: versatile signal-
ing module and pharmaceutical target. Biochim. Biophys. Acta 1747,
Machida, K., Mayer, B.J., and Nollau, P. (2003). Profiling the global ty-
rosine phosphorylation state. Mol. Cell. Proteomics 2, 215–233.
Mahajan, N.P., and Earp, S.H. (2003). An SH2 domain-dependent,
phosphotyrosine-independent interaction between Vav1, and the
Mer receptor tyrosine kinase: a mechanism for localizing GEF action.
J. Biol. Chem. 278, 42596–42603.
Mayer, B.J., Jackson, P.K., and Baltimore, D. (1991). The noncatalytic
src homology region 2 segment of abl tyrosine kinase binds to tyro-
sine-phosphorylated cellular proteins with high affinity. Proc. Natl.
Acad. Sci. USA 88, 627–631.
Mayer,B.J.,Hirai,H.,and Sakai,R.(1995).Evidence thatSH2domains
promote processive phosphorylation by protein-tyrosine kinases.
Curr. Biol. 5, 296–305.
McLaughlin, W.A., Hou, T., and Wang, W. (2006). Prediction of binding
sites of peptide recognition domains: an application on Grb2 and SAP
SH2 domains. J. Mol. Biol. 357, 1322–1334.
Newman, P.J., and Newman, D.K. (2003). Signal transduction path-
ways mediated by PECAM-1: new roles for an old molecule in platelet
and vascular cell biology. Arterioscler. Thromb. Vasc. Biol. 23, 953–
Nollau, P., and Mayer, B.J. (2001). Profiling the global tyrosine phos-
phorylation state by Src homology 2 domain binding. Proc. Natl.
Acad. Sci. USA 98, 13531–13536.
Olsen, J.V., Blagoev, B., Gnad, F., Macek, B., Kumar, C., Mortensen,
P., and Mann, M. (2006). Global, in vivo, and site-specific phosphory-
lation dynamics in signaling networks. Cell 127, 635–648.
Pascal, S.M., Singer, A.U., Gish, G., Yamazaki, T., Shoelson, S.E.,
Pawson, T., Kay, L.E., and Forman-Kay, J.D. (1994). Nuclear magnetic
resonance structure of an SH2 domain of phospholipase C-gamma 1
complexed with a high affinity binding peptide. Cell 77, 461–472.
Paweletz, C.P., Charboneau, L., Bichsel, V.E., Simone, N.L., Chen, T.,
Gillespie, J.W., Emmert-Buck, M.R., Roth, M.J., Petricoin, I.E., and
Liotta, L.A. (2001). Reverse phase protein microarrays which capture
914 Molecular Cell 26, 899–915, June 22, 2007 ª2007 Elsevier Inc.
High-Throughput SH2 Profiling
disease progression show activation of pro-survival pathways at the Download full-text
cancer invasion front. Oncogene 20, 1981–1989.
Pawson, T. (2004). Specificity in signal transduction: from phosphotyr-
osine-SH2 domain interactions to complex cellular systems. Cell 116,
Pawson, T., Gish, G.D., and Nash, P. (2001). SH2 domains, interaction
modules and cellular wiring. Trends Cell Biol. 11, 504–511.
Pero, S.C., Oligino, L., Daly, R.J., Soden, A.L., Liu, C., Roller, P.P., Li,
P., and Krag, D.N. (2002). Identification of novel non-phosphorylated
ligands, which bind selectively to the SH2 domain of Grb7. J. Biol.
Chem. 277, 11918–11926.
Poy, F., Yaffe, M.B., Sayos, J., Saxena, K., Morra, M., Sumegi, J.,
Cantley, L.C., Terhorst, C., and Eck, M.J. (1999). Crystal structures
of the XLP protein SAP reveal a class of SH2 domains with extended,
Pumphrey, N.J., Taylor, V., Freeman, S., Douglas, M.R., Bradfield,
P.F., Young, S.P., Lord, J.M., Wakelam, M.J., Bird, I.N., Salmon, M.,
and Buckley, C.D. (1999). Differential association of cytoplasmic sig-
nalling molecules SHP-1, SHP-2,SHIP and phospholipase C-gamma1
with PECAM-1/CD31. FEBS Lett. 450, 77–83.
Renkema, G.H., Pulkkinen, K., and Saksela, K. (2002). Cdc42/Rac1-
phorylation. Mol. Cell. Biol. 22, 6719–6725.
Sabe,H.,Hata, A.,Okada,M.,Nakagawa,H.,and Hanafusa, H.(1994).
Analysis of the binding of the Src homology 2 domain of Csk to tyro-
sine-phosphorylatedproteinsinthesuppression andmitotic activation
of c-Src. Proc. Natl. Acad. Sci. USA 91, 3984–3988.
Sadowski, I., Stone, J.C., and Pawson, T. (1986). A noncatalytic do-
main conserved among cytoplasmic protein-tyrosine kinases modifies
the kinase function and transforming activity of Fujinami sarcoma virus
P130gag-fps. Mol. Cell. Biol. 6, 4396–4408.
Saldanha, A.J. (2004). Java Treeview—extensible visualization of mi-
croarray data. Bioinformatics 20, 3246–3248.
Sayos, J., Wu, C., Morra, M., Wang, N., Zhang, X., Allen, D.,
van Schaik, S., Notarangelo, L., Geha, R., Roncarolo, M.G., et al.
(1998). The X-linked lymphoproliferative-disease gene product SAP
regulates signals induced through the co-receptor SLAM. Nature
Songyang, Z., Shoelson, S.E., Chaudhuri, M., Gish, G., Pawson, T.,
Haser, W.G., King, F., Roberts, T., Ratnofsky, S., Lechleider, R.J.,
et al. (1993). SH2 domains recognize specific phosphopeptide se-
quences. Cell 72, 767–778.
Songyang, Z., Carraway, K.L., III, Eck, M.J., Harrison, S.C., Feldman,
R.A., Mohammadi, M., Schlessinger, J., Hubbard, S.R., Smith, D.P.,
Eng, C., et al. (1995). Catalytic specificity of protein-tyrosine kinases
is critical for selective signalling. Nature 373, 536–539.
Too, C.K., Murphy, P.R., and Croll, R.P. (1994). Western blotting of
formaldehyde-fixed neuropeptides as small as 400 daltons on gela-
tin-coated nitrocellulose paper. Anal. Biochem. 219, 341–348.
Uhlik, M.T., Temple, B., Bencharit, S., Kimple, A.J., Siderovski, D.P.,
and Johnson, G.L. (2005).Structural and evolutionarydivision of phos-
photyrosine binding (PTB) domains. J. Mol. Biol. 345, 1–20.
Yaffe, M.B. (2002). Phosphotyrosine-binding domains in signal trans-
duction. Nat. Rev. Mol. Cell Biol. 3, 177–186.
Yaoi, T., Chamnongpol, S., Jiang, X., and Li, X. (2006). Src homology 2
domain-based high throughput assays for profiling downstream mole-
cules in receptor tyrosine kinase pathways. Mol. Cell. Proteomics 5,
Molecular Cell 26, 899–915, June 22, 2007 ª2007 Elsevier Inc. 915
High-Throughput SH2 Profiling