Kinome Profiling for Studying Lipopolysaccharide Signal Transduction in Human Peripheral Blood Mononuclear Cells* □ S

The DNA array technique allows comprehensive analysis of the genome and transcriptome, but the high throughput array-based assessment of intracellular signal transduction remains troublesome. The goal of this study was to test a new peptide array technology for studying the activity of all kinases of whole cell lysates, the kinome. Cell lysates from human peripheral blood mononuclear cells before and after stimulation with lipopolysaccharide were used for in vitro phosphorylation with [ (cid:1) - 33 P]ATP arrays consisting of 192 peptides (substrates for kinases) spotted on glass. The usefulness of peptide arrays for studying signal transduction was demonstrated by the generation of the first comprehensive description of the temporal kinetics of phosphorylation events induced by lipopolysaccharide stimulation. Furthermore analysis of the signals obtained suggested activation of p21Ras by lipopolysaccharide, and this was confirmed by direct measurement of p21Ras GTP levels in lipopolysaccharide-stimulated human peripheral blood mononuclear cells, which represents the first direct demonstration of p21Ras activation by stimulation

The DNA array technique allows comprehensive analysis of the genome and transcriptome, but the high throughput array-based assessment of intracellular signal transduction remains troublesome. The goal of this study was to test a new peptide array technology for studying the activity of all kinases of whole cell lysates, the kinome. Cell lysates from human peripheral blood mononuclear cells before and after stimulation with lipopolysaccharide were used for in vitro phosphorylation with [␥-33 P]ATP arrays consisting of 192 peptides (substrates for kinases) spotted on glass. The usefulness of peptide arrays for studying signal transduction was demonstrated by the generation of the first comprehensive description of the temporal kinetics of phosphorylation events induced by lipopolysaccharide stimulation. Furthermore analysis of the signals obtained suggested activation of p21Ras by lipopolysaccharide, and this was confirmed by direct measurement of p21Ras GTP levels in lipopolysaccharide-stimulated human peripheral blood mononuclear cells, which represents the first direct demonstration of p21Ras activation by stimulation of a Toll receptor family member. Further confidence in the usefulness of peptide array technology for studying signal transduction came from Western blot analysis of lipopolysaccharide-stimulated cells, which corroborated the signals obtained using peptide arrays as well as from the demonstration that kinase inhibitors effected peptide array phosphorylation patterns consistent with the expected action of these inhibitors. We conclude that this first metabolic array is a useful method to determine the enzymatic activities of a large group of kinases, offering high throughput analysis of cellular metabolism and signal transduction.
Massive parallel analysis using array technology has become the mainstay for the analysis of genomes and transcriptomes (1)(2)(3)(4)(5). Since the determination of the transcriptome, the understanding of cellular functioning has improved dramatically. Novel insights have led to the notion that the majority of the transcriptome is necessary to keep a cell functioning and could be regarded as the minimal transcriptome. Only a small portion of the transcripts present in the cell determines the identity of the cell, and these critical transcripts are expressed at low levels. Therefore small changes in the expression profiles in the transcriptome can lead to large changes in enzymatic profile of the cell leading to significant differences in cell functioning (6). Thus, a comprehensive description of cellular metabolism may be more useful than such a description of the genome and transcriptome.
Array technology has not yet been adapted to measure enzymatic activity in whole cell lysates, but progress has been made with the preparation of protein chips for the assessment of protein substrate interactions (7-10) and the generation of peptide chips for the appraisal of ligand-receptor interactions and enzymatic activities (11)(12)(13). Recently Houseman and Mrksich (14) showed that employing peptide chips, prepared by the Diels-Alder-mediated immobilization of one kinase substrate (for the non-receptor tyrosine kinase c-Src) on a monolayer of alkanethiolates on gold, allows quantitative evaluation of kinase activity. Hence, in principle an array exhibiting specific consensus sequences for protein kinases across the entire kinome (the combined activity of all cellular kinases) would allow a more comprehensive detection of signal transduction events in whole cell lysates. Obviously, employing this kind of array technology for this purpose would allow faster and more inclusive analysis of cellular metabolism in comparison to currently available technology, which focuses on the static determination of the relative concentration of metabolites but does not address the actual activity of various cellular signaling pathways.
The above-mentioned considerations prompted us to test the usefulness of peptide arrays containing spatially addressed mammalian kinase substrates for studying the kinome in a cellular context. We show that such peptide arrays allow us to make a comprehensive description of the phosphorylation events induced by lipopolysaccharide (LPS) 1 stimulation of peripheral blood mononuclear cells. Furthermore, analysis of the results revealed a role of p21Ras in LPS signal transduction, and this finding was confirmed by a pull-down assay. Thus the peptide array technology enabled us to identify the first example of p21Ras activation by a member of the toll receptor family.

EXPERIMENTAL PROCEDURES
Chemicals-The catalytic subunit of protein kinase A was purchased from Promega (V5161).
Single Kinase Analysis on Peptide Array-50 l of the protein kinase * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Peripheral Blood Mononuclear Cells Isolation-Peripheral blood mononuclear cells (PBMCs) were isolated from peripheral blood of healthy volunteers using standard density gradient centrifugation over Ficoll-Paque Plus (Amersham Biosciences AB), followed by washing and resuspension in IMDM supplemented with 10% fetal bovine serum, penicillin, streptomycin, and amphotericin (hereafter referred to as complete IMDM).

Sequences of substrate peptides and their location on the peptide array
All sequences are derived from the Phosphobase resource (phospho.elm.eu.org). The mean amino acid residue length that achieves an optimal specificity/sensitivity ratio for substrate phosphorylation is nine amino acids. The array consists of 192 peptides, denominated A1-P12, of which 8 are controls (rhodamine-labeled irrelevant peptides for quality control of the spotting process). On each carrier these 192 peptides were spotted twice.  Table II the exact amino acid sequences of the peptide array are listed. The 192 pseudo-substrates are grouped in 8 clusters of 24. Middle, the top 12 most PKA phosphorylated peptides and a comparison to the known substrate consensus sequence for PKA are depicted. Also the location of these peptides on the array is given. Bottom, a quantification of the phosphorylation of the 12 most phosphorylated peptides.
Peptide Array Analysis-For kinome array samples, 10 7 PBMCs were suspended in 5 ml of complete IMDM. Stimulations were terminated by an ice-cold phosphate-buffered saline wash. PBMCs were lysed in 200 ml of cell lysis buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM Na 2 EDTA, 1 mM EGTA, 1% Triton X-100, 2.5 mM sodium pyrophosphate, 1 mM MgCl 2 , 1 mM ␤-glycerophosphate, 1 mM Na 3 VO 4 , 1 mM NaF, 1 g/ml leupeptin, 1 g/ml aprotinin, 1 mM PMSF) and the volume of the cell lysate was equalized with distilled H 2 O. The cell lysates were subsequently cleared on a 0.22-m filter. Peptide array incubation mix was produced by adding 10 l of filter-cleared activation mix (50% glycerol, 50 M [␥-33 P]ATP, 0.05% v/v Brij-35, 0.25 mg/ml bovine serum albumin, [␥-33 P]ATP (1000 kBq)). Next, the peptide array mix was added onto the chip, and the chip was kept at 37°C in a humidified stove for 90 min. Subsequently the peptide array was washed twice with Tris-buffered saline with Tween, twice in 2 M NaCl, and twice in demineralized H 2 O and then air-dried. The experiments were performed three times in duplicate.
Analysis of Peptide Array-The chips were exposed to a phosphorimager plate for 72 h, and the density of the spots was measured and analyzed with array software.
Ras Activation Assay-The Ras Activation Assay kit (Upstate Biotechnology, Lake Placid NY) was used following the supplier's instructions. Determination of protein concentration required by the Ras Ac-tivation Assay kit was made with the BCA Protein Assay kit (Pierce) according to the supplier's instructions.

RESULTS
Peptide Array Design and Construction-In silico analysis of the Phosphobase resource enabled identification of consensus amino acid phosphorylation sequences for most kinases present in mammalian genome (15,16). Further analysis of this set of kinase substrates revealed that the mean amino acid residue length that achieves an optimal specificity/sensitivity ratio for substrate phosphorylation was nine amino acids, and thus for building an array nonapeptides were employed. Arrays were constructed by chemically synthesizing soluble pseudo-peptides, which were covalently coupled to glass substrates (extensively described in the supplemental data). This array consisted of 192 peptides (denominated A1-P12), providing kinase substrate consensus sequences across the mammalian kinome Depicted are the effects the major substrate phosphorylating kinases in peripheral blood mononuclear cell-derived lysates on array phosphorylation of proteins from which substrate peptide sequences are derived. Results of lysates obtained with or without LPS stimulation are shown. The results represent the radioactivity incorporated in the substrates (in PhosphorImager/Aida quantification units) and are the average of four independent measurements. The average inter-spot variance of the data shown is 5%. The results presented are the 38 substrates in which phosphorylation was most up-regulated by LPS treatment. AUTO, auto-phosphorylation; JAK, janus kinase; PKC, protein kinase C; CK I, casein kinase I; MBP, myelin basic protein; PKG, protein kinase G; EIF4F, eukaryotic initiation factor 4F, GSK3, glycogen synthase kinase 3; STAT1a/b, signal transducer and activator of transcription 1a/b; PTP-2C, protein-tyrosine phosphatase 2C; TOP II, topoisomerase II; ACC, acetyl-CoA carboxylase; HSP27, heat shock protein 27; MAPKAP2, mitogen-activated protein kinase-activated protein-2; BICKS, bovine B-50-immunoreactive C-kinase substrate; AP-1, activator protein-1.

. Phosphorylation of peptide arrays by cellular lysates of unstimulated cells and cells stimulated with LPS.
PBMCs were either stimulated with LPS or left untreated, which was followed by lysis of the cells and incubation of the peptide array in the presence of [␥-33 P]ATP. Each of the pictures represents a graphical average for better visual quality and is obtained from four different arrays out of two different experiments. Naturally, not these pictures but the original scanned pictures were used for quantification and statistical analysis of the array. The lower picture is an overlay in which green dots represent peptides that showed enhanced phosphorylation by lysates obtained from LPS-stimulated cells; red dots are peptides that show less phosphorylation when the lysate is stimulated with LPS, and yellow dots are substrates in which the phosphorylation is not altered by LPS stimulation.
(see Table I). To allow assessment of possible intra-experimental variability in substrate phosphorylation, on each separate carrier the array was spotted two times using a biorobotics microgrid spotter equipped with 100 micron split pins. Peptides were spotted onto and subsequently covalently coupled to branched hydrogel polymer-coated glass slides (supplemental data). Spotted slides were stored at 4°C. The final physical dimensions of the array were 19.5 ϫ 19.5 mm with each peptide spot having a diameter of ϳ350 m and peptide spots being 750 m apart.
Single Kinase Analysis-If the design of our peptide array was appropriate, addition of a purified kinase in the presence of ATP should result in the phosphorylation of the appropriate consensus peptide sequences without concomitant phosphoryl- This treatment should result in the phosphorylation of target peptides with 33 P, allowing detection using phosphorimaging. As depicted in Fig. 1, this procedure resulted in strong phosphorylation of several peptides on the array. Analysis of the peptide sequence led to the extraction of the known most optimal PKA consensus sequence (17), whereas accompanying phosphorylation of peptides not containing PKA consensus phosphorylation sites was negligible (Fig. 1). Incubation of the array with PKA and [␣-33 P]ATP did not lead to a detectable signal on the array (data not shown), demonstrating that spot phosphorylation was a specific binding of the ␥-phosphate of ATP to the nonapeptides. These results identified the array as useful tool to determine substrate specificity of kinases.
Kinome Profiling of Unstimulated Human PBMCs-Next we sought to determine whether this array is also useful for kinome profiling of actual whole cell lysates. Of the oligopeptides that were phosphorylated by lysates of unstimulated human PBMCs, a significant fraction was derived from consensus sites for cytoskeletal component-derived peptides (e.g. vimentin lamin D and lamin B receptor) (Fig. 2, Table II), which may reflect the motile phenotype of these cells (the fraction of human PBMCs consists mainly of monocytes and T-cells). In addition, peptides derived from enzymes implicated in basal cell metabolism (e.g. eIF-4F, DNA topoisomerase II, and glycogen synthase) were substrates for cell lysates of these unstimulated cells (Table II). Incubation of the array with lysates and [␣-33 P]ATP did not lead to a detectable signal on the array (data not shown), demonstrating that spot phosphorylation was not dependent on specific binding of ATP to peptides. Thus the kinome profile obtained from human PBMCs using in vitro phosphorylation of a peptide array is not inconsistent with the profile expected from resting mononuclear cells.
Peptide Phosphorylation by Lysates of LPS-stimulated Human PBMCs-LPS is a component of the cell wall of Gramnegative bacteria and a major activator of the innate immune system, mediating for instance septic shock in human disease, and is well documented to elicit activation of a variety of kinases in human PBMCs (18 -20). We considered LPS stimulation of human PBMCs an attractive model to test array-based kinome profiling. To this end, lysates of human PBMCs were treated for 15 min with vehicle or 100 ng/ml LPS and were compared with respect to their capacity to phosphorylate peptides on arrays. The LPS stimulation resulted in the specific incorporation of 33 P in a variety of peptides, suggesting that the LPS treatment activated the kinases for which these peptides represent a consensus phosphorylation sequence. In this array an LPS-dependent up-regulation of phosphorylation of a variety of peptides was detected (Table II), including those with consensus phosphorylation sequences for p38 MAP kinase, p42/ p44 MAP kinase, Jun-N-terminal kinase, and substrates for components of the phosphatidylinositol 3-kinase protein kinase B pathway. The validity of these signals was verified by Western blot analysis of PBMCs and subsequent probing with phosphorylation-state specific antibodies to different kinases (Fig.  3). The up-regulation of the protein kinase G autophosphorylation site was unexpected, as protein kinase G has not been implicated in LPS signaling. However, blasting this substrate against the humane genome revealed that the same substrate is also present in mitogen-activated protein kinase kinase kinase 9, and may thus actually be in strict accordance with the results obtained concerning MAP kinase activation. Thus peptide arraying seems a valid tool for studying LPS signal transduction.
A Comprehensive Description of LPS-induced Phosphorylation Events-The possibility to study a wide range of kinases in parallel makes it possible to make a comprehensive description of the temporal characteristics of LPS-induced phosphorylation, revealing the sequential activation and deactivation of the various kinases. Hence we stimulated human PBMCs for 5 min, 15 min, 30 min, and 60 min and analyzed the effects on kinase activity employing peptide arraying, and the results show different kinetics for phosphorylation of a variety of substrates (Fig. 4). Confirming the results described above, especially enhanced phosphorylation was detected for substrates of various MAP kinases, which are known to be involved in LPSinduced p44/42 MAPK (18 -20) (e.g. Raf, myelin basic protein). A peptide containing the STAT-1␣/␤ phosphorylation site incorporated more radioactivity when incubated with lysates from LPS-stimulated cells (Fig. 4) because STAT-1␣/␤ phosphorylation is a known cellular effect of LPS (21)(22)(23)(24). Also, peptides derived from NFB proteins were also phosphorylated (Fig. 4). Interestingly, the phosphorylation of these peptides peaked in 5-15 min and came back to basal levels after 60 min of LPS stimulation, which is in agreement with the expected time course. Remarkably, Bruton's tyrosine kinase, a member of the Tec kinase family, was recently reported to be involved in LPS signaling (25), and we found corresponding phosphorylation of peptide corresponding to its activation site. In addition, phosphorylation of peptides derived from cytoskeletal proteins became even more pronounced as in unstimulated cells (vimentin, lamin D, and lamin B1), in agreement with the effects of LPS on cell morphology and endocytosis (Fig. 4) (19, 26). The induction of eIF-4F corresponds well to the induction of gene expression by LPS (Fig. 4). Other remarkable effects are the changes in phosphorylation of substrates derived from p450 (CYPII) and ␣-crystallin (Fig. 4), and results correspond well to data published earlier (27,28). Interestingly, several peptides derived from proteins that had not been linked to LPS signal transduction as yet also display marked changes in phosphorylation; e.g. Muscarinic Receptor M2, Rhodopsin, NEU (erbB2), and Phox47. However, phosphorylation of these proteins by MAP kinases has been reported, and thus these effects may well be indirect (Fig. 4) (29 -32). A picture emerges in which phosphorylation of various substrates is dynamically regulated as a consequence of the LPS stimulation.
Effects of Kinase Inhibitors on Lysates-induced Peptide Array Phosphorylation-To determine whether the glass slide-based peptide arrays gave functional and realistic results with respect to phosphorylation, we used the MAP kinase inhibitors PD 98059 and SB 203580. Indeed these inhibitors prevented the phosphorylation of MAP kinase regulated substrates (Fig.   5) (myelin basic protein, MAPKAPK2, c-Jun), and the substrates that were regulated by MAP kinases (muscarinic receptor M2, rhodopsin, NEU (erbB2), and Phox47) were also inhibited (Fig. 5). However other substrates that are not directly

FIG. 5. Effects of two MAP kinase inhibitors in the phosphorylation of different kinase substrates in cells stimulated with LPS.
Cells were pre-incubated for 1 h with the inhibitors PD (PD98059, 50 M) or SB (SB203580, 10 M) and subsequently stimulated with 100 ng/ml LPS; after 0 and 60 min the cells were lysed and analyzed using the peptide array analysis protocol. The arrays were quantified, the values of unstimulated cells were set at 1, and the other conditions were compared with the unstimulated values. The graphs show the phosphorylation level of 20 different substrates with their respective S.E. The first two bars depict normal peripheral blood mononuclear cells, the second two bars are from cells pretreated with PD98059, and the last two bars depict lysates that are pretreated with SB203580. The first bar represents unstimulated cell lysates, and the second is from lysates that are stimulated with LPS. The graphs A-P are substrates that have been reported to be involved in LPS/tumor necrosis factor signaling or in MAP kinase signaling, and the graphs Q-T are of substrates that are not commonly associated with LPS signal transduction. phosphorylated by MAP kinases are also influenced by the inhibitors, indicating that cross-talk and/or feedback loops between different classes of kinases is possible (STAT-1␣/␤, ␣-crystallin, Tec).
Analysis of the Peptide Array Results Reveals a Role for p21Ras Activation in LPS Signal Transduction-Among the most important questions in LPS signal transduction is the molecular mechanism leading to the activation of the Raf/ MEK/p42/p44 MAP kinase-signaling cascade. Classically, activation of this cascade is brought about either via the p21Ras route (which is archetypical for receptor tyrosine kinase-coupled receptors) in conjunction with protein kinase C or via the sequential activation of phospholipase C␤ and protein kinase C␣/␤ (which are archetypical for G protein-coupled receptors). Because LPS signals via the Toll-like receptor 4 (20), neither of both possibilities is immediately obvious. However, p21Ras has been reported to be linked to Toll/interleukin-1 receptor) domaindependent signaling indicating that Ras might be activated upon stimulation with LPS (33). Apart from activating the p42/p44 MAP kinase signaling cascade, Ras activation also leads to stimulation of the phosphatidylinositol 3-kinase/protein kinase B pathway as well as to extensive cytoskeletal remodeling. We noticed that in our results LPS induced phosphorylation of peptides that are in accordance with activation of p21Ras as judged from the increase in phosphorylation of peptides associated with the MAP kinase pathway, the phosphatidylinositol 3-kinase pathway, and cytoskeletal proteins. Also other substrates consistent with the activation of p21Ras (MARCKS, Na ϩ /K ϩ ATPase, Annexin-2; Fig. 6) were phosphorylated after LPS stimulation. This prompted us to look at whether p21Ras might be involved in LPS signaling. Indeed in a p21Ras activation assay we detected increased GTP-bound p21Ras. This marks the first direct identification of p21Ras activation via the stimulation of a member of the Toll-like receptor family. These data show that it is possible to use peptide array technology to characterize changes in cellular metabolism and signal transduction even in a temporal manner and find novel interactions between signaling cascades. DISCUSSION We interpret our kinome analysis as a useful and valuable method to determine the enzymatic activities of a large group of kinases. Therefore, kinome profiling could be a realistic possibility and especially interesting as the current metabolomics effort has been hampered by the lack of techniques that allow high-throughput analysis of the flow of cellular metabolism. Current mass spectrometry techniques concentrate on the static determination of metabolite levels rather as the enzymatic activity of the biochemical process leading to these levels. In the present study we focused on kinome profiling, but one can easily imagine arrays for assaying cellular activity with respect to dephosphorylation, acylation, acetylation, ubiquitination (ubiquitinome), etc. Thus arraying for enzymatic activities may provide metabolomics with the equivalent of the DNA array analysis for genomics with respect to the possibility to quickly obtain a comprehensive description of cellular metabolism and cellular transcriptome respectively.
At present kinome arraying still suffers from "teething problems." One of the aspects that may have to be determined is the fact that phosphatases present in the lysate could influence the amount of phosphorylation; to prevent this we employed a full spectrum of inhibitors. Obviously, however, the net amount of phosphorylation in the cell critically depends on the net activity of kinases and phosphatases, and thus the results obtained may not reflect the actual phosphorylation status of substrates in the cell but rather the amount of enzymatic phosphorylating activity. In essence kinome arraying measures flow and not absolute levels of substrate phosphorylation. In principle, using pre-phosphorylated arrays, it should be possible to measure phosphatase activity as well. It may also be possible to perform experiments without phosphatase inhibitors, but until it is shown that under the in vitro conditions of array phosphorylation kinase and phosphatase activity show similar temporal characteristics, results should be interpreted with caution.
Another possible concern is that less abundant kinases may be more difficult to visualize because more abundant kinase will produce stronger signals but may not be less important for the control of physiological processes. One possible way to alleviate this problem is by using relative activities for each spot. In this manner each spot has its own range, and the differences in intensity are bypassed. In this manner one looks at the fold induction instead of the amount of phosphorylation. The fact that there are more residues inside a substrate that can be phosphorylated is also solved by using relative activation levels. Also important is that this type of array needs to be updated regularly as new and more specific substrates are being discovered.
Further development of this technique is now critically dependent on the generation of peptides having improved specificity for further cellular kinases as well as expansion of the array to include the entire kinome. In particular, a pressing issue is the concern that some peptides can be phosphorylated on more than one spot and that this results in over-phosphorylation of the peptide and therefore overrating the kinase activity. This may be alleviated by developing specific pseudopeptides, harboring only one phosphorylation site, and by developing peptides with increased specificity for one kinase. The problem that MAP kinases or other kinases perform a double phosphorylation can be circumvented by employing relative values in the analysis or other post hoc corrections in the analysis software. It is important to realize that no data are currently available that suggest that peptides with multiple phosphorylation sites for kinases are also phosphorylated on those two sites simultaneously; however the extent of the problem is unknown, and more research may be needed in this direction.
Disregarding these limitations, however, our present study has shown that the kinome reacts dynamically to stimulation with LPS and has helped in identifying p21Ras as a novel signal transducer in LPS signaling. Thus we feel that peptide arraying for kinome-wide analysis of biologically relevant samples is a highly promising tool for studying the biochemical changes underlying cellular signal transduction.