A Genomic and Proteomic Analysis of Activation of the Human Neutrophil by Lipopolysaccharide and its Mediation by p38 Mitogen-Activated Protein Kinase

Bacterial lipopolysaccharide evokes several functional responses in the neutrophil that contribute to innate immunity. While certain responses, such as adhesion and synthesis of tumor necrosis factor- α , are inhibited by pretreatment with an inhibitor of p38 mitogen-activated protein kinase, others, such as actin assembly, are unaffected. The aim of the present study was to investigate the changes in neutrophil gene transcription and protein expression following lipopolysaccharide exposure and to establish their dependence on p38 signaling. Microarray analysis indicated expression of 13% of the 7070 Affymetrix gene set in nonstimulated neutrophils, and LPS-upregulation of 100 distinct genes, including cytokines and chemokines, signaling molecules, and regulators of transcription. Proteomic analysis yielded a separate list of upregulated modulators of inflammation, signaling molecules, and cytoskeletal proteins. Poor concordance between mRNA transcript and protein expression changes was noted. Pretreatment with the p38 inhibitor SB203580 attenuated 23% of LPS-regulated genes and 18% of LPS-regulated proteins by ≥ 40%. This study indicates that p38 plays a selective role in regulation of neutrophil transcripts and proteins following lipopolysaccharide exposure, clarifies that several of the effects of lipopolysaccharide are post-transcriptional and post-translational, and identifies several proteins not previously reported to be involved in the innate immune response.


INTRODUCTION
Lipopolysaccharide (LPS), 1 a component of the outer cell wall of Gram-negative bacteria, evokes a variety of functional responses in the human neutrophil (PMN) after binding to a plasma membrane receptor complex that involves the Toll-like receptors (TLRs) (1)(2)(3)(4)(5). These "immediate" functional responses, including actin assembly, adhesion, activation of nuclear factor-kappa B (NF-κB), and priming for an enhanced secretory response and for release of reactive oxygen intermediates, appear to be central both to the innate immune response and to the pathogenesis of several inflammatory human diseases, including sepsis and the acute respiratory distress syndrome (6). p38 mitogen-activated protein kinase (p38 MAPk) has been shown to mediate LPS-induced PMN adhesion, NF-κB activation, and TNF-α and IL-8 translation and release (7), and its blockade attenuates LPS-induced PMN accumulation in the airspace (8). However, other cascades almost certainly lead to downstream effectors of the LPS signal; for example, actin assembly appears to be p38 MAPk-independent (9). An improved understanding of the transcriptional and translational responses of the neutrophil to LPS and the modulation of these responses by p38 MAPk might carry pathogenetic and therapeutic implications.
Historically, it has been believed that the downstream PMN transcriptional response to LPS is static and that PMN functional responses to LPS that depend on de novo protein sythesis are primarily limited to the release of cytokines (10). However, recent studies indicate a robust transcriptional response (11).
To date, most studies have relied upon and reported a short list of functional assays of the LPS-exposed PMN; therefore, no exhaustive investigation of either the transcriptional response or protein synthetic repertoire of the PMN has been reported. While several techniques have been used to evaluate transcripts, the screening of global changes in mRNA by microarray analysis has only recently become possible. In this way, thousands of genes can be screened in an unbiased fashion for transcript LPS incubation  PMNs were isolated by the plasma Percoll method (22), a technique which yields less than 5% monocytic contamination, and resuspended at a concentration of 15.4 x 10 6  Affymetrix oligonucleotide array  Five µg total RNA was isolated with Trizol (GIBCO) and RNeasy columns (Qiagen), and subsequently labeled with biotin as described by Affymetrix. Briefly, first strand synthesis was accomplished with Superscript II reverse transcriptase (GIBCO) using a T7oligo(dT) 24 primer for 1 hour at 42°C, followed by second strand synthesis using E. coli DNA polymerase I and RNase H (GIBCO) at 16°C for 2 hours. dsDNA was used as a template for in vitro transcription with T7 RNA polymerase in the presence of biotin-labeled UTP and CTP using the BioArray High Yield RNA Transcript Labeling Kit (Enzo). Fifteen µg cRNA was fragmented and used for hybridization to Affymetrix HuGene 6800FL Genechips. Each sample was hybridized initially 7 BioImage (GSI; Ann Arbor, MI) 2D-Analyzer software was used to locate, quantitate, and match protein spots on the control and LPS gel images. Analysis was performed by assigning 50 common anchor spots between paired images; the remaining spots were compared by a constellation matching algorithm. All data were then carefully reviewed by the operator to account for any discrepancies.
Protein loading between control and experimental gels may have varied because of inconsistencies in rehydration of the different IEF gel strips; therefore, gel images were normalized so that the sum of the integrated intensities of all matched spots on paired gels was made equal. Control and LPS-stimulated gel images from individual donor experiments were matched to generate composite images; composite images were then matched into a master composite image in order to track the LPS response of protein spots among different donors (26). Only those spots that were common (image-matched) to all original twelve (pH 3.0-10.0) gels were considered for further analysis. For these spots, the LPS-induced change in integrated intensity in the six experiments was subjected to statistical analysis with a twotailed student's t-test, and those spots with p<0.05 were identified by peptide mass fingerprinting (described below). For the narrow range (pH 5.0-6.0; 5.5-6.7; 6-11) 2-D PAGE experiments using pooled donors, only those spots with concordant regulation exceeding 1.5-fold, or which appeared de novo in the LPS gel, in two repeat experiments were further analyzed.
In-gel tryptic digestion  In-gel digestion of protein spots was performed with sequencing grade 8 MA) operated in delayed extraction mode. Samples (0.5 µl) were spotted onto a sample plate to which matrix (0.5 µl of 10 mg/ml CHCA) was added. The sample-matrix mixture was dried at room temperature and then analyzed in reflector mode. CHCA was also spotted alone as a negative control.
Spectra were the sum of 100 laser shots, and those peaks with a signal to noise ratio of greater than 3:1 were selected for database searching. Spectra were internally calibrated using autolytic trypsin peptides (m/z 842.51, 2211.10).
Database searching algorithm  The monoisotopic masses for each protonated peptide were: a) entered into the program MS-Fit (http://prospector.ucsf.edu) for searches against the Swiss-Prot, NCBI, and GenPept databases, and b) entered into Mascot (http://matrixscience.com), an algorithm testing statistical significance of peptide mass fingerprinting identifications. For MS-Fit searches, masses derived from trypsin, CHCA, keratin, and Coomassie Brilliant Blue G-250 were excluded. Search parameters included a maximum allowed peptide mass error of 0.1 Da (0.8 Da in the few instances in which linear mode was used), consideration of one incomplete cleavage per peptide, pI range of 3.0 to 10.0, and MW range of 1 kD to 200 kD. Accepted modifications included carbamidomethylation of cysteine residues (from iodoacetamide exposure following IEF) (28), and methionine oxidation, a common modification occurring during SDS-PAGE (29). Protein identifications were assigned when three criteria were met: 1) statistical significance (p<0.05) of the match when tested by Mascot (http://matrixscience.com); 2) >20% sequence coverage by the tryptic peptides; and 3) concordance (+/-15%) with the molecular weight (MW) and pI of the parent 2-D PAGE protein spot. The following special exceptions were considered: a) protein identifications not fulfilling criterion 2 were still assigned if criteria 1 and 3 were fulfilled and no other Homo sapiens proteins with peptide mass match p values <0.05 were identified by Mascot; b) if criterion 3 was not fulfilled (lower than expected MW), a cleavage product of the identified protein was inferred and the cumulative MW of the tryptic peptides was compared with that of the 2D-PAGE spot to ensure that it was not exceeded; c) if criterion 3 was not fulfilled (isolated discordance between theoretical and observed pI), post-translational modification of an unrecovered peptide was inferred; and d) if two or more Homo sapiens protein assignments with >4 mutually exclusive matching peptides were identified, a protein mixture in the 2-D PAGE spot was inferred and further analysis halted (quantitative conclusions regarding the individual protein constituents could not be drawn).

RESULTS
Genes differentially expressed in LPS-stimulated neutrophils  Human PMNs were a) left untreated and b) incubated in the presence of 100 ng/ml LPS for 4 hours. As a control to confirm that the PMNs were quiescent at baseline and that LPS resulted in normal stimulation, mRNA was isolated, cDNA was prepared, and PCR for TNF-α was performed. Little TNF-α expression was seen in non-stimulated cells, whereas LPS treatment led to an increase in expression in each of the donors subsequently used for microarray analysis (data not shown). No macrophage-colony stimulating factor (M-CSF) receptor was detected by oligonucleotide microarray analysis, confirming no significant monocytic contamination.
Human PMNs express a limited repertoire of mRNA transcripts at baseline, but repond to LPS with differential expression of genes in many families. Considering only those genes present by microarray analysis in all three donors, unstimulated PMNs expressed 13.0% (923 of 7070 genes) of the Affymetrix gene set. Gene classes represented at baseline include metabolic enzymes, structural proteins, receptors, signaling proteins, and transcription factors. By comparison, human monocytes expressed ~40% and human fibroblasts ~35% of the represented genes (data not shown). By the criterion of a >3-fold increase in expression in all three donors on Affymetrix oligonucleotide array analysis, exposure of PMNs to LPS for 4 hours resulted in the upregulation of 100 genes (Table I).
Genes from several different functional classes were induced in PMNs following LPS exposure. Of interest, a number of transcriptional regulators were induced, including transcription factors of the NF-κB family. The transcriptional NF-κB complex has previously been implicated in the regulation of the genes induced by LPS (11). The genes for several cytokines and chemokines were also found to be upregulated. These include TNF-α , IL-1β, IL-6, MCP-1, MIP-3α, and MIP-1β (Table I). To confirm results from the microarray analysis, PCR was performed. PCR analysis on selected genes indicates that the time course for changes can be rapid or delayed, but parallel the changes found in the array at the 4 h time point (data not shown). Other up-regulated genes included those for metabolic enzymes, immune response molecules, kinases, phosphatases, signaling molecules, adhesion and cytoskeletal components, interferon-stimulated genes, and those with unknown or miscellaneous function (Table I).
LPS stimulation of PMN also resulted in the downregulation of 56 genes (Table II). Downregulated genes were identified as transcriptional regulators, protein and lipid kinases and phosphatases, structural molecules, and signaling molecules. Genes for metabolic proteins were also evident, as were several uncharacterized genes.

2-D PAGE and image analysis 
In contrast to the limited number of transcripts found at baseline, PMNs were found to express a large number and variety of proteins in the nonstimulated state ( Figure   1A and 1C, Tables III-V). Reproducible protein expression patterns were found on the pH 3.0-10.0 gels, and the majority of proteins fell in the pH 5.0-7.0 range ( Figure 1A). The basic region (pH>7.0) consistently exhibited poor resolution, precluding meaningful image analysis and further workup (data not shown). Depending on the spot-finding parameters (minimum spot intensity, filter width) selected on the image analysis software, spot-by-spot manual editing was found to be necessary in order to avoid over-and underdetected spots; moreover, further manual editing was performed to screen for unmatched and mismatched spots following matching of paired control and LPS-stimulated gels. After spot editing, ~1200 well-resolved spots were evident on each pH 3.0-10.0 gel. In an attempt to improve resolution of the pI range bearing the greatest number of well-resolved spots, overlapping narrow pH range gels (pH 5.0-6.0; 5.5-6.7; 6-11) were also run. Of interest, a similar number of well-resolved spots (~1200) were detected on the narrow pH range gels ( Figure 1C Human PMNs respond to LPS with the differential expression of a large number of proteins. In the six individual pH 3.0-10.0 experiments, the number of protein spots that increased in integrated intensity by at least 50% following LPS exposure was 185, 122, 104, 104, 96, and 131, respectively. The number of protein spots that decreased by at least 50% following LPS exposure was 72, 151, 102, 98, 128, and 97, respectively. While gel-to-gel regional variability in resolution was expected to account for individual spots not being well visualized on particular gels, only those spots that were matched to all twelve original gels were analyzed further. Overall, the number of spots matched to all twelve original gels was 125. The number of spots that were both matched to all twelve original gels and that increased by at least 50% in integrated intensity in the individual experiments following LPS exposure was 46,13,17,27,22, and 20, respectively. The number of spots that were matched to all twelve gels and that decreased by at least 50% was 6, 22, 17, 22, 34, and 28, respectively. The LPS-induced change in integrated intensity of the 125 spots that were matched to all twelve original gels was subjected to statistical analysis with a two-tailed student's t-test and those spots with statistically significant (p<0.05) regulation among the six experiments were identified by peptide mass fingerprinting (Table III).
Identification of LPS-regulated proteins  On the pH 3.0-10.0 gels, several proteins were consistently upregulated (Table III), including regulators of inflammation (annexin III) and signaling molecules (Rab GDP dissociation inhibitor β). Several actin fragments were seen to be consistently upregulated in the six experiments following LPS exposure (Table III). Of interest, the proteasome β chain was also consistently upregulated. Downregulated proteins included other signaling molecules, such as Rho GTPase activating protein 1.
On the pH 5.0-6.0 and 5.5-6.7 gels, several proteins were found to increase by greater than 1.5-fold following LPS exposure (Table IV,  and ERK ( Figure 2A) (33). Assuming that no other multiply phosphorylated stathmin species have escaped detection, analysis of the integrated intensities of the PO 4 -stathmin and stathmin spots indicates that the percentage of the PO 4 form of total cellular stathmin has increased from 11% to 38% with LPS stimulation ( Figure 2B). This is similar to a previous report of an increase from <10% to 35-40% of the Ser 25 phosphorylated form in Jurkat cells stimulated with anti-CD3 (34).

Effect of SB203580 on LPS-stimulated gene expression  Gene expression analysis of PMNs
stimulated with LPS indicated that the majority of genes induced by LPS were unaffected by prior treatment of PMN with SB203580. Of the 100 genes upregulated by LPS, the upregulation of 23 was inhibited by greater than 40% (Table VI). The majority of these genes affected by SB203580 were inhibited by less than 60%, whereas only 6 were inhibited by greater than 80%, all of which represent previously identified interferon-stimulated genes. Induction of cytokine genes by LPS, with the exception of IL-6, was generally unaffected by SB203580.
Effect of SB203580 on LPS-stimulated protein expression  Similar to the effect of SB203580 on LPSstimulated gene expression, little effect of SB203580 was seen on expression levels for the majority of LPS-regulated proteins (Table VII). Two exceptions are annexin III and α-enolase, for which LPSstimulated expression was attenuated in the presence of the p38 MAPk inhibitor.
Comparison of microarray and proteomics results     Of the LPS-regulated proteins identified by peptide mass fingerprinting for which probes were present on the oligonucleotide microarray, poor concordance was found at the mRNA level (Table VIII) (Table VIII), with no notable effect of SB203580 on expression at either level. Similarly, CAP1, RhoGAP 1, and ficolin 1 were downregulated at both the mRNA transcript and protein level (Table VIII), with no notable effect of SB203580. Annexin III was downregulated at the transcript level and upregulated at the protein level, with an inhibitory effect of SB203580 seen only at the protein level (Table VIII). Gram-negative-derived LPS (i.e., from E. coli), are known to signal through TLRs (36,37). Importantly, many of the expression changes found in LPS-stimulated PMNs in the present study were also described in the bacteria-exposed monocytic cells, indicating that many of the gene expression changes seen in bacterial infection are likely mediated by TLRs (38,39), and that the LPS model system accurately reflects exposure of immune cells to infection. Nevertheless, the reliance upon DNA microarrays alone affords insight only upon the transcriptional response without corroboration at the protein level. In the present study, application of both DNA microarray and proteomics technology to our model system provides unique insight upon both the cellular biology of the activated PMN and the responsiveness and regulation of its transcriptional and translational machinery. As will be discussed below, our study identifies, in particular, novel aspects of the LPS-stimulated PMNs transcriptional regulation, activity in the innate immune response, signaling, cytoskeletal reorganization, and priming for granule release.

DISCUSSION
In the present study, the increase in NF-κB transcript abundance (Table I)  As expected from the literature, the genes for several cytokines and chemokines, including IL-1β, IL-6, and MIP-1β were found to be upregulated (Table I). On the other hand, the notable absence of upregulated cytokines in the proteomics experiments reflects their removal in the post-LPS-incubation wash performed prior to lysis for 2D-PAGE. Upregulation of these inflammatory mediators is well documented in PMNs exposed to LPS, and in animal models of LPS-induced sepsis syndrome and acute respiratory distress syndrome, a PMN-mediated illness (41,42). Several genes in this family were upregulated that have not, to our knowledge, been described in LPS-stimulated cells, including MCP-1, GRO3, IL-10RA, and HM-74, an orphan G protein-coupled receptor with homology to chemokine receptors. The downregulation of TRAIL, the lymphotoxin b receptor, and IL8RB were also observed.
The modulation of genes involved in cytokine signaling, including the adapter molecules TRAF1 (LPS and TNF receptor signaling) and A20 (TNF receptor signaling), as well as several kinases and phosphatases may indicate a change in cytokine responsiveness after LPS treatment. Relevant in this regard from the proteomics data are the upregulation of protein phosphatase 1, which has been shown to regulate PMN NADPH oxidase activation and translocation (43,44) as well as to regulate LPS-induced NF-κB activation (45), the downregulation of Rho-GAP1, which has been shown to regulate NADPH oxidase activity in the PMN (46), as well as upregulation of PO 4 -stathmin (Table IV), a phosphoprotein postulated to function as a relayer and integrator of multiple signal transduction pathways (34). Several non-cytokine, non-chemokine genes involved in the immune response were also upregulated, including the complement pathway members C3, C3AR1, and PFC; the protease inhibitors ELANH2 (elastase inhibitor), SLP1, PI-3 and PI-9; and the acute phase protein orosomucoid. LPS-regulation of C3AR1 and orosomucoid expression has not previously been reported. In the proteomics experiments, the downregulation of ficolin-1 (Table III) (Table II), has been implicated in cell motility and metastasis (49). Decreased motility may be beneficial in sustaining the inflammatory response at sites of infection.
In addition, LPS treatment results in an inhibition of apoptosis (50). Therefore, the longer residence time of the PMN at sites of infection is consistent with the long term genetically coded changes seen in these gene profiling experiments, and indicates that the changes in gene expression are functionally relevant to host-defense and immunity.
By providing information on post-translational modification, the proteomics data may provide further insights into the cytoskeletal remodeling effects of LPS upon the PMN. We contend that the actin fragments identified (Table III) are unlikely to represent technical artifacts. Rather, their specificity (identical MW/pI among different experiments), statistically significant upregulation by LPS, as well as the use of a lysis buffer containing chaotropes and multiple protease inhibitors, argue instead that these fragments are physiologic consequences of LPS exposure in the human PMN. More specifically, the upregulation of these fragments following LPS exposure (Table III) suggests that LPS may activate an actin-cleaving enzyme, which, in turn, remodels the cytoskeleton. Intriguing in this vein, calpain has recently been reported to play an important role in cell migration and cytoskeletal organization of fibroblasts (51). The possibility that LPS may induce calpain activation, and that calpain activation may regulate cytoskeletal reorganization and motility, is currently under investigation. An alternative possibility is that actin cleavage is a marker of neutrophil apoptosis (52).
Other LPS-regulated proteins may play important roles in cytoskeletal reorganization. The upregulation of protein tyrosine kinase 9-like (A6-related protein) may modulate LPS-induced actin polymerization, as it bears a high degree of homology to twinfilin (A6), an actin monomer-binding protein that localizes to sites of rapid filament assembly in cells and is believed to regulate actin filament turnover (53). In turn, LPS-induced downregulation of Rho-GTPase activating protein 1 (Table III) may regulate twinfilin (and protein tyrosine kinase 9-like) activity, as twinfilin has been shown to colocalize with Rac1 and Cdc42, and to be regulated by active Rac1 in NIH 3T3 cells (53). Activation of Rho proteins may be facilitated by LPS-upregulation of moesin (Table V) family proteins (55), and, interestingly, is also postulated to regulate the dynamics of both the actin and microtubule cytoskeletons via phosphorylation of stathmin (Table IV) (56). Calponin H2 is an actinbinding protein not previously reported in PMNs that is postulated to play a role in cytoskeletal organization (57). Its downregulation by LPS (Table V) likely modulates LPS-induced cytoskeletal reorganization. The upregulation of nonmuscle myosin heavy chain and a putative phosphorylated form of myosin heavy chain (putative protein kinase C substrate by prediction rules) in the LPS-exposed PMN (Table IV) is of uncertain significance; myosin has been implicated in multiple functions in the PMN, including locomotion, fluid pinocytosis, and phagocytosis (58). Of interest, however, S100A4 (downregulated , Table II) has been reported to regulate cytoskeletal dynamics by inhibiting protein kinase C-mediated phosphorylation of nonmuscle myosin heavy chain (59). (Table IV, Figure 2) may represent another mechanism by which the cytoskeleton is remodeled. Stathmin is a phosphoprotein reportedly involved in both signal transduction and in regulation of the microtubulin filament network; furthermore, phosphorylation of stathmin has been reported to modulate its tubulin-binding avidity (60). Inferences can be made about both the phosphorylation site on PO 4 -stathmin and the responsible kinase induced by LPS. Four phosphorylation sites in stathmin have been well described: Ser 16 , Ser 25 , Ser 38 , and Ser 63 (32,33). Ser 16 has been reported as a substrate for Ca 2+ /calmodulin (CaM)-dependent kinases (32), and Ser 25 as primarily a substrate for p38 and ERK (33), with p34 cdc2 also active but bearing a 5-fold preference for Ser 38 (34). As stated above, the phosphopeptide identified in PO 4 -stathmin, extending from residue 15 to 27 (1468.7 Da), is consistent with phosphorylation of either Ser 16 or Ser 25 ( Figure 2). While both p38δ and p38α MAPk isoforms are expressed in the human PMN, LPS has been shown selectively to activate the p38α isoform in human PMNs (9). The p38α isoform, however, has been shown to be relatively inactive at Ser 25 ; in fact, p38δ is approximately 100-fold more active at Ser 25 and selective p38α inhibitors do not inhibit the stress-activated phosphorylation of stathmin in 293 cells (33). Further support for the lack of involvement of p38 signaling in phosphorylation of stathmin in our system is the apparent lack of effect of SB203580 (a selective p38α and p38β inhibitor) on LPS-induced expression of PO 4 -stathmin (Table IV). As p34 cdc2 is relatively inactive at Ser 25 (34), we conclude that the phosphorylation site is likely to be Ser 16 , a reported substrate of CaM-dependent kinase. While CaM kinases have previously been implicated in gene activation in LPS-exposed myelomonocytic HD11 cells (61), stathmin signaling has not, to our knowledge, been previously reported in either PMNs or lipopolysaccharide signal transduction.

LPS induction of stathmin phosporylation
Cytoskeletal reorganization, a well-described regulator of granule release (62), may underlie LPSinduced priming for PMN granule release, but several LPS-regulated proteins may provide more specific clues. LPS exposure led to increased levels of grancalcin, a calcium-binding protein previously detected in PMNs and shown to translocate to granules and plasma membrane in the presence of physiologic concentrations of calcium (63). Similarly, annexin III, a calcium-binding protein highly expressed in PMN granule membranes and implicated in calcium-mediated secretion (64), and in granule fusion (65), was also found to be upregulated. Exocytosis of granule contents may also be facilitated by LPS-upregulation of Rab GDP dissociation inhibitor (Table III), which has been proposed to recycle Rab after vesicle fusion by extracting it from the membrane and loading it onto newly formed transport intermediates (66).
Parallel use of DNA microarrays and proteomics affords a powerful strategy for comparison of corresponding mRNA transcripts and proteins, thereby affording new insight upon the mechanisms by which the cell regulates its signaling responses to the external environment. Of interest, a poor correlation was found between corresponding transcripts and proteins (Table VIII) (Table VIII), with no notable effect of SB203580, consistent with a non-p38-mediated pathway of primary transcriptional downregulation.
Interestingly, annexin III was downregulated at the transcript level and upregulated at the protein level, with an inhibitory effect of SB203580 seen only at the protein level (Table VII)   Assuming that no other multiply phosphorylated stathmin species have escaped detection, analysis of the integrated intensities of the PO 4 -stathmin and stathmin spots indicates that the percentage of the PO 4 form of total cellular stathmin has increased from 11% to 38% with LPS stimulation. The decrease in integrated intensity for stathmin was equal in amount to the increase in PO 4 -stathmin following LPS exposure.  Table IV. Analysis of pH 5.0-6.0 2-D PAGE gels. Results are from pooled samples for control (n=3) and LPS-exposed (n=3) PMNs from human donors. Expression of the reported proteins was altered >1.5-fold following LPS exposure in two repeat experiments. "New" designates proteins seen in the LPS gel in two repeat experiments but not detectable in the corresponding control gels.  Control LPS-exposed