## Abstract

A comprehensive study of changes in messenger RNA (mRNA) levels in human neutrophils following exposure to bacteria is described. Within 2 hours there are dramatic changes in the levels of several hundred mRNAs including those for a variety of cytokines, receptors, apoptosis-regulating products, and membrane trafficking regulators. In addition, there are a large number of up-regulated mRNAs that appear to represent a common core of activation response genes that have been identified as early-response products to a variety of stimuli in a number of other cell types. The activation response of neutrophils to nonpathogenic bacteria is greatly altered by exposure to Yersinia pestis, which may be a major factor contributing to the virulence and rapid progression of plague. Several gene clusters were created based on the patterns of gene induction caused by different bacteria. These clusters were consistent with those found by a principal components analysis. A number of the changes could be interpreted in terms of neutrophil physiology and the known functions of the genes. These findings indicate that active regulation of gene expression plays a major role in the neutrophil contribution to the cellular inflammatory response. Interruption of these changes by pathogens, such as Y pestis, could be responsible, at least in part, for the failure to contain infections by highly virulent organisms.

## Introduction

Neutrophils are the first cells to be recruited from the blood stream to sites of inflammation1,2 and are critically important for determining the outcome of some acute infections.3 They are postmitotic cells that synthesize lower levels of protein and RNA than most dividing cells, and they can interact and/or modulate inflammation. Nevertheless, on exposure to bacteria or other activating agents, neutrophils are known to synthesize and secrete a number of cytokines4,5 including interleukin-1 (IL-1),6 IL-8,7,8 oncostatin M,9 and small inducible cytokine A3/macrophage inflammatory protein 1α/(SCYA3/MIP1A).10-12

Neutrophils are readily isolated from human peripheral blood. The isolated cells are more than 99% pure, with the principal contaminant being eosinophils, which themselves have relatively low levels of macromolecular synthetic activity. The cells can be synchronously exposed to “natural” stimuli such as opsonized bacteria and offer an attractive system for the study of gene expression in terminally differentiated cells. Although the cell biology of neutrophil activation has been studied in some detail, studies of responses at the messenger RNA (mRNA) level have been circumscribed, focusing principally on one or a few cytokine mRNA species.

Approaches for simultaneously detecting changes in levels of many of the polyadenylated RNA in a cell population fall into 3 categories: hybridization to arrays of targets complementary to specific mRNAs, sequencing of many randomly chosen complementary DNA (cDNA) fragments, or display of specific cDNA fragments on gels. A method for display of 3′-end restriction fragments of each species of RNA13 has the advantages that the position of fragments corresponding to known genes is predictable and that no prior knowledge of the sequence is needed to detect previously “unknown” genes.

We have applied cDNA display to study changes in mRNA levels in neutrophils activated by exposure to various bacteria. Sufficient analyses were performed to detect, on a statistical basis, more than 90% of all changes in transcripts. We used time-course studies to get insight into the mechanisms underlying these changes. There is a dramatic and complex change in the gene expression profiles of activated neutrophils, indicating an important role for neutrophil gene regulation in the propagation and early evolution of the inflammatory response.

## Materials and methods

### Bacterial strains and culture

Yersinia pestis strains KIM5 and KIM614were derived from strain KIM (Kurdistan Iran man).15,16KIM6, a derivative of KIM5, lacks the 70-kb plasmid pCD1. This plasmid carries 60 genes, 47 of which have been implicated in a system that enables the bacteria to inject specific proteins directly into the cytoplasm of mammalian cells.17-20 The injection machinery, its substrate proteins, and its regulatory apparatus are encoded by this plasmid. Y pestis strains lacking pCD1 are completely avirulent. Escherichia coli K12 strain R594 (Flac-3350 galK2 galT22 λrpsL179IN(rrnD-rrnE)1) was chosen to serve as a generic avirulent enterobacterial isolate.

Overnight cultures of Y pestis grown in modified Tryptose Blood Agar Base without the agar comprising 10 g tryptose, 3 g beef extract, and 5 g sodium chloride (NaCl) per liter supplemented with 2.5 mM calcium dichloride (CaCl2) were diluted to a density of 3 × 107bacteria per mL and incubated for 3 hours at 26°C in a water bath. At this point the temperature was shifted to 37°C, and the incubation was continued for an additional 2 hours. The bacteria were collected by centrifugation, washed with Hanks balanced salt solution (HBSS; without Ca++ or Mg++), resuspended in HBSS to a final density of 1.75 × 109 bacteria per mL and opsonized by the addition of 1.5 volumes of normal human serum incubated at 37°C for 20 minutes, washed twice with RPMI 1640 supplemented with 10% heat-inactivated fetal calf serum (FCS), and resuspended to a final density of 7 × 108 bacteria per mL.

Overnight cultures of E coli K12 grown in LB were diluted 1:100 and incubated for 2 hours at 37°C. They were then washed and opsonized as described above, except that C7-deficient human serum (Sigma Chemical, St Louis, MO) was used. This precaution was not necessary with Y pestis, which is completely resistant to complement-mediated lysis.

### Cell separation and activation by bacteria

We isolated neutrophils by using dextran sedimentation, centrifugation through Ficoll-Hypaque Plus, (Amersham Pharmacia Biotech, Uppsala, Sweden) and very brief hypotonic lysis of erythrocytes.21 All reagents, serum, buffers, and media were free of LPS (less than 0.01 ng/mL by limulus amoebocyte lysate assay [Sigma]). Monocytes were enumerated in neutrophil preparations by flow microfluorometry. A neutrophil suspension was incubated with fluorescein isothiocyanate (FITC)–conjugated anti-CD45 and phycoerythrin (PE)–conjugated anti-CD14 (Becton Dickinson, Mountain View, CA). The cells were then fixed with fluorescence-activated cell sorter (FACS) lysis buffer (Becton Dickinson) and analyzed with a FACScan flow cytometer (Becton Dickinson). Monocytes were identified on the basis of their forward and side light scattering properties and expression of CD45 and CD14. At least 105 events were analyzed for each sample.

Freshly isolated neutrophils and opsonized bacteria, suspended in RPMI plus 10% heat-inactivated FCS, were mixed to final densities of 2 × 106 cells per mL and 4 × 107 bacteria per mL, respectively. These cultures or control neutrophils were then incubated for 2 hours, or other indicated times, at 37°C with gentle agitation.

Monocytes were isolated from the peripheral blood mononuclear cells by a spontaneous aggregation method at 4°C.22 To activate monocytes, they were exposed for 2 hours to opsonized E coliK12 at a ratio of 20 bacteria per cell, which is the same procedure that was used for activation of neutrophils. Time-course experiments were analyzed with neutrophils incubated for at least 3 time points including 0 minutes (negative control), 10-30 minutes (early), and 120 minutes (late) with E coli K12 .

### Northern blot analysis and in situ hybridization

Northern blot analysis of total cell RNA, extracted from neutrophils by the guanidine hydrochloride (HCl) method,21,23 was performed as described.24-26Levels of hybridization were measured quantitatively by the Phosphorimager System (Molecular Dynamics, Sunnyvale, CA) and normalized to the 18S ribosomal RNA signals. In situ hybridization was performed by a previously described method,27 using Cy3 and FluorX (Amersham Pharmacia Biotech, Piscataway, NJ) labeled oligonucleotide probes.

### Gel display of 3′-end restriction fragments

cDNA displays of cells activated by bacteria were prepared as previously described in detail.21,28 Bands were displayed on sequencing gels run to display products of at least 100 bases in length. Bands were excised, amplified by polymerase chain reaction (PCR), and sequenced. The enzymes used to digest cDNA for comparison of the effects of Y pestis with those of E coli were BamHI, BclI, BglII,BsrGI, ClaI, EagI, EcoRI,HindIII, NcoI, PstI, andXbaI. Enzymes used for time-course studies wereApaI, BglII, HindIII, KpnI,SacI, SpeI, SphI, andXbaI.

For most experiments, every band that differed in relative intensity between the control pattern and any of the experimental patterns was sequenced. In different experiments using the same restriction enzymes, many bands could be confidently recognized as corresponding to previously sequenced bands on the basis of both band pattern and sequence.

### Informatics

Most of the band intensities were quantified by eye, confirmed by a second investigator, and expressed as a single-digit numeric. A part of the bands was quantified by using the Phosphorimager. The integrated density of each band above background was calculated as a “volume” by the Phosphorimager, as described in Figure1. A least-square linear regression model of the logarithm of the “volume” (Ŷ) in terms of the visually quantified band intensity (X) gives = 0.56X + 8.22, with Pearson correlation r = 0.93, an acceptable reliability estimate. The standard error of a given 0, estimated for a new observation X0, is computed as:

$Δ=1+1n+(X0−X¯)2nςx2·∑i=1n (Yi−Y^i)2n−2$

where n is the number of observations,$X¯$ is X averaged, and Ŷirepresents the predicted values corresponding to the observationsXi and

$ςx2=∑i=1n Xi2n.$

Thus for any given X0, a 1 − α confidence interval (CI) for 0 is the set of values of Y such that

$Y^0−Δ·Tvalue(n−2,α)

where Tvalue (n − 2,α) is the 2-tailed t-value of a t-distribution with n − 2 degrees of freedom.

Fig. 1.

Correlation between band intensities and Phosphorimager quantification.

We confirmed the accuracy of our band quantification method by comparing our results with the Phosphorimager, whose sensitivity and reproducibility is comparable to scintillation counting.64About 50 bands from randomly chosen lanes were selected, and each of their intensities was quantified by eye, confirmed by a second investigator, and expressed as a single-digit numeric. The same gel was digitized by the Phosphorimager with 16-bit precision to form the image and analyzed by Imagequant software (Molecular Dynamics). We used the software program S-PLUS (Mathsoft, Cambridge, MA) for statistical analysis as described in “Materials and methods.” Dashed lines display the 95% CI.

Fig. 1.

Correlation between band intensities and Phosphorimager quantification.

We confirmed the accuracy of our band quantification method by comparing our results with the Phosphorimager, whose sensitivity and reproducibility is comparable to scintillation counting.64About 50 bands from randomly chosen lanes were selected, and each of their intensities was quantified by eye, confirmed by a second investigator, and expressed as a single-digit numeric. The same gel was digitized by the Phosphorimager with 16-bit precision to form the image and analyzed by Imagequant software (Molecular Dynamics). We used the software program S-PLUS (Mathsoft, Cambridge, MA) for statistical analysis as described in “Materials and methods.” Dashed lines display the 95% CI.

The accuracy of such confidence limits depends on the validity of the assumption of linearity and equally normal distributions ofY values across all values of X. However, measurements by the Phosphorimager at very low intensities are much less reliable. Therefore we fit the data to a linear regression model based on measurements at X > 1, yieldingŶ = 0.49X + 8.64 and a correlation ofr = 0.96. For all X0, Δ(X0) · Tvalue(38,0.05) ≈ 0.5.

Quantitative measurement of Northern blots of several mRNAs confirms that genes identified by gel display to be up- or down-regulated do indeed show increases or decreases of mRNA levels. These changes range from a 10-fold decrease to a 71-fold increase (Table1), and the logarithm of the values correlate to estimates from the gel display method (Pearson correlationr = 0.92).

Table 1.

Northern blot analysis of gene expression in neutrophils: quantitative changes in mRNA levels induced by incubation with opsonized E coli

Gene symbolTranscript encodingRelative Northern blot expressionDifferential display expression
Gene nameControl+ E coli
IL1B IL-1, beta 71 8
IL8 IL-8 16
SCYA3 Small inducible cytokine A3; MIP1A 41 6
ATP6L ATPase, H+transporting, lysosomal (vacuolar proton pump) 16 kd 2.9
ITGB2 Integrin, beta 2 (CD18, LFA-1) 0.60
IL8RB IL-8 receptor, beta (CXCR2) 0.21
IL8RA IL-8 receptor, alpha (CDw128a CXCR1) 0.16 1
NCF1 Neutrophil cytosolic factor 1 (p47-PHOX) 0.10
Gene symbolTranscript encodingRelative Northern blot expressionDifferential display expression
Gene nameControl+ E coli
IL1B IL-1, beta 71 8
IL8 IL-8 16
SCYA3 Small inducible cytokine A3; MIP1A 41 6
ATP6L ATPase, H+transporting, lysosomal (vacuolar proton pump) 16 kd 2.9
ITGB2 Integrin, beta 2 (CD18, LFA-1) 0.60
IL8RB IL-8 receptor, beta (CXCR2) 0.21
IL8RA IL-8 receptor, alpha (CDw128a CXCR1) 0.16 1
NCF1 Neutrophil cytosolic factor 1 (p47-PHOX) 0.10

The relative Northern blot expression is expressed as neutrophils + E coli/control neutrophils. The total RNA of identically treated neutrophils used for either Northern blots or differential display.

RNA from HeLa cells was used as negative control for each measurement. Hybridization probes were cDNA-labeled by random priming of PCR products amplified from neutrophil cDNA. Primer sequences are available on request. Procedures for filter hybridization were previously described.25 Equal loading of lanes was demonstrated by ethidium bromide staining and by rehybridization with a 5.8-kbHindIII restriction fragment of rat 18S ribosomal cDNA.26 Transcript levels in total RNA were quantified by the Phosphorimager System. Data are presented as the means of 2-4 determinations, within 20% of each other. The ratio of expression level of neutrophils incubated with E coli K12 for 2 hours to neutrophils incubated without bacteria for 2 hours is shown. See Figure 1 for the relation between Phosphorimager values and differential display estimates.

Each sequence was searched against nr and dbest databases of the National Center for Biotechnology Information (NCBI) by the Basic Local Alignment Search Tool (BLAST) program (http://www.ncbi.nlm.nih.gov/BLAST/blast_overview.html).29Matches to known genes were confirmed to come from the 3′-untranslated regions of mRNA except where otherwise noted. The length of sequence obtained was compared with the size of bands on the display gel as a quality check.

A database was created by Shigeru Yamaga, jointly with Wen Ming Xiao (Gene Logic, Gaithersburg, MD), using Microsoft Access (Microsoft, Redmond, WA) as a database engine. An individual record was created for every differentially expressed band, and related information was entered as hypertext links to sequence files, search results of GenBank and TIGR databases, bands of overlapping sequence, references to relevant literature, keys for various classifications of bands, presence of polyA signals, quality of sequence, and scanned gel images, etc.

We used LocusLink ID,65 when available, as a unique key to known genes, and we used the terms listed as gene symbol (the HGMW-approved symbol, where applicable) and gene name. For ESTs, we used UniGene cluster numbers as a unique key.66Subsequently all sequences were clustered by a modified PHRAP approach.30 Public gene database search was completed on November 9, 2000.

## Results

### Quality of cell and RNA preparation

Morphologically, our neutrophil preparations were more than 99% pure except for the presence of 1% to 3% eosinophils; band forms accounted for less than 3% of the cells. No cells with the typical morphology of monocytes could be identified by light microscopy, nor did flow cytometry reveal any monocytes. Occasional preparations with more than 0.5% monocytes were discarded. The yield of total RNA from the neutrophil preparations averaged 13 μg/108 cells (range, 7-17 g/108 cells). We examined the distribution of IL-8 transcripts by in situ hybridization using a combination of 2 CY-3 (red)–labeled oligonucleotides complementary to different regions of the mRNA. IL-8 transcripts were detectable in virtually all neutrophils after incubation for 2 hours with E coli, although the intensity of RNA staining was somewhat variable from cell to cell. Neutrophils incubated in the absence of bacteria showed considerably less intense staining (data not shown).

We prepared monocytes and neutrophils from the same blood sample. Both types of cells were exposed to E coli K12 for 2 hours and then harvested for cDNA display (Table2 and Figure2, left). In some cases RNA species that were among the most strongly induced in neutrophils were actually down-regulated in monocytes, excluding the possibility that monocytes activated by the bacteria were contributing to the observed pattern for these species. Northern blots also showed that RNA extracted from the neutrophils did not contain detectable transcripts forc-fms,31 the receptor for macrophage colony-stimulating factor (MCSF) (data not shown).

Table 2.

mRNA differently regulated in monocytes and neutrophils

Gene symbolGene nameNeutrophil (control)Neutrophil +E coliMonocyte (control)Monocyte + E coli
ANXA5 Annexin A5 2
SCYA7 Small inducible cytokine A7 (monocyte chemotactic protein 3)
IL10RA IL-10 receptor, alpha
HLA-DRA MHC class II, DR alpha
CDW52 CDW52 antigen (CAMPATH-1 antigen) 4
VIL2 Villin 2 (ezrin) 4
HM74 Putative chemokine receptor; GTP-binding protein
CD97 CD97 antigen
PPIF Peptidyl-prolyl isomerase F (cyclophilin F)
Gene symbolGene nameNeutrophil (control)Neutrophil +E coliMonocyte (control)Monocyte + E coli
ANXA5 Annexin A5 2
SCYA7 Small inducible cytokine A7 (monocyte chemotactic protein 3)
IL10RA IL-10 receptor, alpha
HLA-DRA MHC class II, DR alpha
CDW52 CDW52 antigen (CAMPATH-1 antigen) 4
VIL2 Villin 2 (ezrin) 4
HM74 Putative chemokine receptor; GTP-binding protein
CD97 CD97 antigen
PPIF Peptidyl-prolyl isomerase F (cyclophilin F)

Gene symbols are taken from the HGMW-approved symbol and/or LocusLink database of NCBI.65 Cells with E coli were incubated with opsonized E coli K12 for 2 hours as described in “Materials and methods.” The intensity of each band in the sequencing gels was quantified and expressed as a single-digit numeric as described in “Materials and methods.”

Fig. 2.

Representative segments of display gels of cDNA fragments: left, neutrophils and monocytes exposed to

E coli for 2 hours; right, neutrophils exposed to various bacteria for 2 hours. After incubating the leucocytes with or without bacteria at 37°C for 2 hours, total RNA was extracted, and double-stranded oligo-dT primed cDNA was synthesized by standard methods. The cDNA was digested with a restriction enzyme and ligated with a Y-shaped adapter. Ligated cDNA was amplified by PCR with a phosphorous 32 (32P)–labeled primer complementary to one arm of the Y-shaped adapter and a second primer complementary to the oligo-dT primer and containing one of the 12 possible dinucleotide extensions on its 3′-end, as previously described.1,2 Ec indicates E coli K12; Yp (KIM5), Y pestis KIM5 (pCD1+); Yp (KIM6), Y pestis KIM6 (pCD1); N, neutrophil; and M, monocyte. Gene symbols are the same as those described in Tables 4-6, except for the following:PAI2, plasminogen activator inhibitor, type II (arginine-serpin) (GenBank Accession no. Y00630); RPL3, ribosomal protein L3 (X73460); and RPS4X, ribosomal protein S4, X-linked (M58458).

Fig. 2.

Representative segments of display gels of cDNA fragments: left, neutrophils and monocytes exposed to

E coli for 2 hours; right, neutrophils exposed to various bacteria for 2 hours. After incubating the leucocytes with or without bacteria at 37°C for 2 hours, total RNA was extracted, and double-stranded oligo-dT primed cDNA was synthesized by standard methods. The cDNA was digested with a restriction enzyme and ligated with a Y-shaped adapter. Ligated cDNA was amplified by PCR with a phosphorous 32 (32P)–labeled primer complementary to one arm of the Y-shaped adapter and a second primer complementary to the oligo-dT primer and containing one of the 12 possible dinucleotide extensions on its 3′-end, as previously described.1,2 Ec indicates E coli K12; Yp (KIM5), Y pestis KIM5 (pCD1+); Yp (KIM6), Y pestis KIM6 (pCD1); N, neutrophil; and M, monocyte. Gene symbols are the same as those described in Tables 4-6, except for the following:PAI2, plasminogen activator inhibitor, type II (arginine-serpin) (GenBank Accession no. Y00630); RPL3, ribosomal protein L3 (X73460); and RPS4X, ribosomal protein S4, X-linked (M58458).

### Changes in gene expression profile in neutrophils exposed to bacteria

We undertook an extensive comparison of the cDNAs generated from control neutrophils and neutrophils treated for 2 hours with one of 3 bacteria: E coli K12, Y pestis strain KIM5, orY pestis strain KIM6. A total of 17 different restriction enzymes were used for these displays, and fragments from each enzyme digest were displayed with each of the 12 possible 3′-terminal dinucleotides on the oligo-dT primer. On average, approximately 100 bands per lane could be evaluated visually. In most cases we analyzed the sequences of all bands whose inserts were in the size range of 75-600 base pairs (bp) and whose intensities differed by more than 2-fold between control and treated samples. Bands up to greater than 1 kb in length were analyzed if they were prominent and showed clear differences between samples. Based on the number of bands observed and on the frequency of randomly distributed restriction sites, we should have achieved an average of 1.5 representations of mRNA of intermediate abundance, with a higher frequency for the abundant mRNAs.

Striking differences were evident in patterns of cDNA display between control neutrophils and neutrophils exposed to bacteria (Figure 2, right). A total of 1887 bands were sequenced (Table3). Of these, approximately 19% did not give good sequence. A portion of these sequences still gave high probability matches to known sequences so that the bands could be identified. Any single prominent band is unlikely to represent more than a few percent of total mRNAs. This implies that bands corresponding to one mRNA molecule per cell are visible, except where obscured by darker bands. Redundancy in analyses occasionally occurred, particularly for some of the most prominent RNAs. Multiple bands representing the same transcript could arise by buckling out of nucleotides during oligo-dT priming, but often resulted from alternate sites of polyA addition in mRNAs.

Table 3.

Distribution of gene categories

DNA fragment sequences of bands from the gel display of 3′-end restriction fragments were clustered by using Locuslink, Unigene, and similarity as described in “Materials and methods.” DNA fragments represent the number of bands on differential display gels from which DNA sequencing was performed.

F3-150

From 350 genes, we grouped 224 genes whose expression patterns were reproducibly and prominently modified by bacteria (E coli  K12 or Y pestis substrains KIM5 or KIM6) in 2 hours as described in Tables 4, 5, and 6.

In total, 350 known genes and 292 EST or anonymous sequences were found to change substantially in expression level by 2 hours after activation with bacteria. The anonymous cDNA could be derived from unrecognized alternate polyadenylation sites in genes represented elsewhere in the database, cDNA primed from A-rich internal sequences in mRNA or heterogeneous nuclear RNA, or genes not yet represented in the EST databases or GenBank. Five of the bands represented perfect copies of EST sequences derived from repetitive sequences. The perfect match in these cases suggests that the genomic template has been identified.

We obtained 48 nonrepetitive sequences that had no match in the gene databases. About half of these had a perfect polyA signal or a hexanucleotide that differed by a single base from the consensus AAUAAA sequence.32 Such deviation is commonly seen in mRNAs for known genes, so it is likely a large fraction of these represent polyadenylated RNA. Approximately 8% of the sequences corresponded to repetitive sequence, and most of them did not precisely match anything in the database. These frequently lacked even an approximate polyA addition signal. However, 4 different specific repetitive sequences were strongly induced in the neutrophils by exposure to bacteria. Increased transcription of repetitive sequences has been noted in stimulated cells and may have a physiologic role.33

### Clustering neutrophil gene expression patterns

We grouped the neutrophil genes according to their expression profiles under 4 conditions in the following order: control, incubation with E coli K12 for 2 hours, and incubation with KIM5 (pCD1+) and KIM6 (pCD1−) strains of Y pestis. We clustered the genes according to their similarity to idealized expression patterns. For instance, the expression pattern of an ideal gene that is overexpressed (high) for the virulent KIM5 condition and under-expressed (low) for the control, E coli K12, and avirulent KIM6 conditions would be low-low-high-low (described as “LLHL”). Overall we have 24 − 2 idealized patterns excluding HHHH and LLLL. The Pearson correlation was used as the measure of similarity of each gene expression pattern,x = (x1,x2,x3,x4) to each of the 14 idealized patternsy = (y1,y2,y3,y4). The order of the entries for each gene expression vector xor y is control, E coli, KIM5, and then KIM6. The query gene is assigned to a cluster designated by the idealized pattern that has the maximal correlation with that gene. Figure3 shows 2 representative normalized gene expression patterns of neutrophils, LHLH (upper) and HLHL (lower).

Fig. 3.

Comparison of the expression patterns of 2 clusters of genes from neutrophils stimulated with various bacteria.

Patterns LHLH (upper, 48 genes) or HLHL (lower, 44 genes) correspond to those described in Table 4. Each set of 4 bands (control, E coli K12 and KIM5 and KIM6 strains of Y pestis; Figure2, right panel) in adjacent lanes with the same electrophoretic mobility in a differential display gel was quantified by its intensity and normalized so that the average of 4 bands equals zero, and the variance of 4 bands equals one. Each line on these graphs corresponds to one dot in Figure 4 and represents one gene in Table 4.

Fig. 3.

Comparison of the expression patterns of 2 clusters of genes from neutrophils stimulated with various bacteria.

Patterns LHLH (upper, 48 genes) or HLHL (lower, 44 genes) correspond to those described in Table 4. Each set of 4 bands (control, E coli K12 and KIM5 and KIM6 strains of Y pestis; Figure2, right panel) in adjacent lanes with the same electrophoretic mobility in a differential display gel was quantified by its intensity and normalized so that the average of 4 bands equals zero, and the variance of 4 bands equals one. Each line on these graphs corresponds to one dot in Figure 4 and represents one gene in Table 4.

To show the affinity between genes classified to the same cluster, principal components analysis (PCA) was performed. Genes tend to coalesce in homogeneous clusters determined by their similarity to an ideal expression pattern (Figure 4). Thus, our criterion for classifying genes according to their similarity to predetermined idealized expression patterns allows us to recognize well-separated clusters. We note that this is equivalent to the first iteration of the standard k-means clustering technique.34There are 2 differences from k-means: (1) our clustering method does not require reassignment of new centers for all clusters, as is done at each k-means iteration step, and (2) the centers are predetermined by the idealized expression profiles as opposed to a random centers' initialization, which is the first step of the k-means algorithm.

Fig. 4.

Gene clusters separated by PCA.

PCA allows us to present the multidimensional data (in this case, 4-dimensional data of each gene expression pattern) in a simple 2-dimensional graph. First we derived the 4 principle components, which are a linear combination of the original variables (certain gene expression intensities of neutrophils of control or stimulated with one of 3 bacteria: E coli K12 and KIM5 and KIM6 strains ofY pestis). Then we found that the first 2 principal components capture most of the variation of the data (95.2% in our case). Therefore the data can be displayed (with a minor loss of information) in a 2-dimensional graph, with these 2 largest key principal components as the x- and y-axes. The axes titles “cn1” and “cn2” represent the first 2 principal components. The label of each cluster is the same as those in the “Expression pattern” row of Tables 4 and 5. As can be seen, a large fraction of the total differences in expression patterns of the genes can be visualized in this 2-dimensional graph.

Fig. 4.

Gene clusters separated by PCA.

PCA allows us to present the multidimensional data (in this case, 4-dimensional data of each gene expression pattern) in a simple 2-dimensional graph. First we derived the 4 principle components, which are a linear combination of the original variables (certain gene expression intensities of neutrophils of control or stimulated with one of 3 bacteria: E coli K12 and KIM5 and KIM6 strains ofY pestis). Then we found that the first 2 principal components capture most of the variation of the data (95.2% in our case). Therefore the data can be displayed (with a minor loss of information) in a 2-dimensional graph, with these 2 largest key principal components as the x- and y-axes. The axes titles “cn1” and “cn2” represent the first 2 principal components. The label of each cluster is the same as those in the “Expression pattern” row of Tables 4 and 5. As can be seen, a large fraction of the total differences in expression patterns of the genes can be visualized in this 2-dimensional graph.

### Genes differently expressed in neutrophils exposed toY pestis

We also compared the effects of 2 strains of Y pestis, the causative agent of plague, on neutrophils. The high virulence of this pathogen is in part due to its ability to prevent the accumulation of neutrophils at foci of infection early in the course of disease.15,35,36 An important contribution of the type III secretion system to suppressing neutrophil accumulation is the inhibition of cytokine production.37,38

The most common pattern of mRNA change was a substantial increase in response to E coli or KIM6, but there was no change in response to KIM5 (Table 4, LHLH). Most of the cytokines we identified showed this pattern. A second common pattern is that mRNAs present in the control and KIM5-treated cells were depressed in the cells treated with E coli and KIM6 (Table 4, HLHL). This pattern also confirms that most of the cells received a stimulus as a result of exposure to the bacteria. A smaller number of mRNAs were induced or substantially up-regulated only by KIM5 (Table 4, LLHL). Overall the effects of nonpathogens on genes listed in Table 4 were quite parallel, presumably because the bacteria present common stimuli.

Table 4.

Genes expressed by neutrophils differently regulated by KIM5 (pCD1+) and KIM6 (pCD1) strains ofY pestis

 KIM5-responsive KIM6-responsive E coliK12 NC NC NC (slightly up-regulated) Up-regulated Down-regulated KIM5 Up-regulated Down-regulated NC (slightly up-regulated) NC (slightly up-regulated) NC KIM6 NC NC Up-regulated Up-regulated Down-regulated Cytokines GRO1 — — IL1B SCYA3 GRO2 IL1A LTB S100A9 S100A8 IL8 OSM SCYA4 IL1RN PBEF SCYA20 Receptors IFNGR2 — — HM74 PLAUR OLR1 CD44 IFNGR1 LILRA2 EDG6 IL8RB FPR1 CD97 GPR44 Transcription modulators CROC4 — — COPEBHDAC3 NFKB1TRIP8 XIP BRF2 DSIPI ZNF220BTF3 TOM1 Apoptosis regulators CFLAR — — PPIFGADD34 IER3 PIG7 BCL2A1MCL1 Proteases PPGB — — PSMC4 CLN2 GCL Protein kinases CAMKK2 — — — MKNK1CLK1 RPS6KA1 MAP3K8 Other signal transducers TPD52L2 — — — ARHGDIBHCLS1S100A11 ICB-1 Membrane trafficking ATP6L — — ATP2B1 RAB1 AQP9 RAC1 LOC51312 VPS35 RAB5A FLOT1 SLC7A5 ATP6F SECTM1 Oxidases — — — PTGS2 NCF2NCF1 NCF4 Others BRI3 HLA-C ADAM8 FTL G0S2PLEK FTH1 ACTB DIFF48 H3F3A MYO1E B2MFACL1 DAF NPM1 SATHLA-A TMSB4X VDUP1 ARPC1B GCSH TPM3 PRG1 NS1-BP SOD2 EVI2B TALDO1 WBSCR1 EIF4A1 HNRPC KIAA0415 MCP NBS1 UBE2B SUI1 Gene expression pattern LLHL HHLH LLLH LHLH HLHL
 KIM5-responsive KIM6-responsive E coliK12 NC NC NC (slightly up-regulated) Up-regulated Down-regulated KIM5 Up-regulated Down-regulated NC (slightly up-regulated) NC (slightly up-regulated) NC KIM6 NC NC Up-regulated Up-regulated Down-regulated Cytokines GRO1 — — IL1B SCYA3 GRO2 IL1A LTB S100A9 S100A8 IL8 OSM SCYA4 IL1RN PBEF SCYA20 Receptors IFNGR2 — — HM74 PLAUR OLR1 CD44 IFNGR1 LILRA2 EDG6 IL8RB FPR1 CD97 GPR44 Transcription modulators CROC4 — — COPEBHDAC3 NFKB1TRIP8 XIP BRF2 DSIPI ZNF220BTF3 TOM1 Apoptosis regulators CFLAR — — PPIFGADD34 IER3 PIG7 BCL2A1MCL1 Proteases PPGB — — PSMC4 CLN2 GCL Protein kinases CAMKK2 — — — MKNK1CLK1 RPS6KA1 MAP3K8 Other signal transducers TPD52L2 — — — ARHGDIBHCLS1S100A11 ICB-1 Membrane trafficking ATP6L — — ATP2B1 RAB1 AQP9 RAC1 LOC51312 VPS35 RAB5A FLOT1 SLC7A5 ATP6F SECTM1 Oxidases — — — PTGS2 NCF2NCF1 NCF4 Others BRI3 HLA-C ADAM8 FTL G0S2PLEK FTH1 ACTB DIFF48 H3F3A MYO1E B2MFACL1 DAF NPM1 SATHLA-A TMSB4X VDUP1 ARPC1B GCSH TPM3 PRG1 NS1-BP SOD2 EVI2B TALDO1 WBSCR1 EIF4A1 HNRPC KIAA0415 MCP NBS1 UBE2B SUI1 Gene expression pattern LLHL HHLH LLLH LHLH HLHL

Genes were sorted by their expression patterns as follows: first, by the difference between minimum and maximum band intensities; then by their maximum band intensity; and lastly by the alphabetical order of gene symbols. Genes in boldface showed a minimum difference between H and L of more than 5-fold (ie, 3-step increase/decrease in 8 as maximum) in band intensity. Others were detected with the difference of 3- to 5-fold.

NC indicates no change. Abbreviations of gene names were taken from the gene symbols in the Locuslink database of NCBI, where applicable, and are listed in Appendix  A1.

The expression of a smaller number of mRNAs appeared to be influenced by Y pestis, regardless of pCD1 but not by E coli(Table 5, LLHH and HHLL). Some genes were affected only byE coli (LHLL and HLHH), and a number were regulated alike by all 3 bacteria (LHHH and HLLL). Many of the changes in the levels of mRNA could be interpreted in terms of the known behavior of neutrophils. Activation of neutrophils by the nonpathogenic gram-negative bacteria induced expression of a variety of cytokines and receptors. Several known cytokines have not been previously associated with neutrophils or were first described in this context after the present analyses were completed. These include SCYA20 (LARC/MIP3A), oncostatin M, GRO1, and GRO2.

Table 5.

Genes similarly regulated by both KIM5 (pCD1+) and KIM6 (pCD1) strains of Y pestis

E colinonresponsiveE coli responsive
Y pestis responsiveY pestisnonresponsiveY pestis responsive
E coliK12 NC NC Up-regulated Down-regulated Up-regulated Down-regulated
KIM5 Up-regulated Down-regulated NC NC Up-regulated Down-regulated
KIM6 Up-regulated Down-regulated NC NC Up-regulated Down-regulated
Receptors ITGB2 — FCAR FCER1G TNFAIP6 FCGR3A
IL8RA PTPRC
Chaperones — — — — HSPCB —
Transcription modulators — — ETS2 NACA — MNDA
Apoptosis regulators — — — — GADD45B —
Proteases CTSD — TIMP1 — CPD —
Protein kinases — — — GPRK6 MAP2K3 —
Other signal transducers RALGDS
ICAM3 LSP1
RGS2 EHD1 PTPN6
TYROBP
MACS HPCAL1
RGS14
Membrane trafficking — — — — RAB7 VATD ATP6S1
Oxidases — CYBA — — — —
Others RPL18A ULK1 — EMD LOC51669BB1 HMOX1 H1F2 DNM2
HMG17  KIAA1100
KIAA0370
Gene expression pattern LLHH HHLL LHLL HLHH LHHH HLLL
E colinonresponsiveE coli responsive
Y pestis responsiveY pestisnonresponsiveY pestis responsive
E coliK12 NC NC Up-regulated Down-regulated Up-regulated Down-regulated
KIM5 Up-regulated Down-regulated NC NC Up-regulated Down-regulated
KIM6 Up-regulated Down-regulated NC NC Up-regulated Down-regulated
Receptors ITGB2 — FCAR FCER1G TNFAIP6 FCGR3A
IL8RA PTPRC
Chaperones — — — — HSPCB —
Transcription modulators — — ETS2 NACA — MNDA
Apoptosis regulators — — — — GADD45B —
Proteases CTSD — TIMP1 — CPD —
Protein kinases — — — GPRK6 MAP2K3 —
Other signal transducers RALGDS
ICAM3 LSP1
RGS2 EHD1 PTPN6
TYROBP
MACS HPCAL1
RGS14
Membrane trafficking — — — — RAB7 VATD ATP6S1
Oxidases — CYBA — — — —
Others RPL18A ULK1 — EMD LOC51669BB1 HMOX1 H1F2 DNM2
HMG17  KIAA1100
KIAA0370
Gene expression pattern LLHH HHLL LHLL HLHH LHHH HLLL

For abbreviations, see Table 4 and Appendix  A1.

Putative membrane trafficking regulators were up-regulated in a functionally coordinate fashion. Thus mRNAs encoding the 3 small guanosine 5′-triphosphatases (GTPases), RAB1, RAB5A, and RAB7, were all up-regulated by E coli and KIM6. Interestingly, KIM5 slightly up-regulated RAB1 and RAB5A, but strongly up-regulated RAB7, a small GTPase implicated in transport from late endosomes to lysosomes. ARHGDIB, a guanine nucleotide dissociation inhibitor that would presumably delay reconversion of the GDP-bound form to the active GTP-bound form of these proteins, was down-regulated by nonpathogens, while KIM5 did not change its expression (Table4, HLHL).

Apoptosis of neutrophils in vitro is delayed by various activating stimuli. Examination of the RNAs up-regulated by the nonpathogens offers several potential mechanisms for this effect. BCL2A1 was strongly induced, as previously reported for activated monocytes.39 BCL2A1 has antiapoptotic properties, but unlike BCL2, it does not inhibit the accumulation of differentiated myeloid cells from the 32D cell line.40MCL1 is another strongly up-regulated product implicated as an antiapoptotic protein. IER3 is a p53 responsive gene that protects cells from FAS- or TNF-induced apoptosis.41 One of the most strongly up-regulated mRNAs was that for PPIF (cyclophilin F). This protein is a mitochondrial peptidyl-prolyl isomerase implicated in mitochondrial pore structure and perhaps permeability transitions.42 This is intriguing because cytochrome c release from mitochondria is a component of a caspase-activating system central to many forms of apoptosis. The mRNA encoding certain subunits of the vacuolar adenosine 5′-triphosphate (ATP)-dependent H+ pump, another potential downstream antiapoptotic factor, were also up-regulated. KIM5 had little effect on most of the above genes.

### Changes in neutrophil gene expression were asynchronous

The changes in mRNA expression patterns at short time intervals following the addition of E coli K12 were also analyzed. Many of the striking increases in mRNA levels seen at 2 hours after exposure to bacteria were not reflected by changes in levels of the corresponding mRNA within the first 60 minutes, although the levels of some mRNAs progressively increased beginning within 30 minutes (Table 6, LHL and LHH). Display ofBglII-cut cDNAs prepared 3 and 4 hours after exposure toE coli showed a pattern that was for the most part similar to the 2-hour pattern (data not shown).

Table 6.

Time-course profile and functional classification of genes expressed by E coli K12–stimulated neutrophils

Up-regulated
EarlyLateDown-regulated
TransientSustainedEarlyLate
Cytokines GRN OSM IL1RN SCYA4 IL8 SCYA20 S100A9 S100A8
SCYA3 IL1B GRO2PBEF
Receptors — PLAURHS1-2 CCRL2 ADORA2A SELL IL8RA IL8RB
TNFRSF1A HM74 CD44 FPR1 CD97 FCGR3A
ITGA5 TNFRSF10B  IFNGR1
LILRA2
PTPRC
Membrane trafficking — CLIC1ATP5E
FLOT1
ATP6J RAB1ATP2B1
NAF1 RAB7SLC16A3
ATP2A3 —
SSR2 SLC11A2 NAPA
Apoptosis regulators — TNFAIP3
PPIF BCL2A1 GSTTLp28 CFLAR PPIA PIG7
MCL1
Transcription modulators FOS ZFP36
ETR101 NFKBIA
HIF1A HMGIY DDIT3 NFE2L2
NFKB1 NFKBIE GC20
NFE2CEBPB
BTF3
DSIPI MNDA
ZNF220BRF2
EGR1 FOSBIRF1
ZNF148 ATF4
PSCDBP
Others BTG2HLA-E WBP2 G0S2 SGK MAP2K3 CPD EHD1 GCL ARPC1B EVI2B AMPD2
S100P TPT1 TNFAIP2 PLEK CD48 DUSP6 FTH1 LOC51669 DIFF48 CLN2
DYRK1APCOLN3 PTGS2 MACS PFN1 RTN4 TALDO1 MKNK1
GCP2 RPN2 MME HSPA10 CDKN1A LIMK2 PRG1 ERV1 MYO1E KIAA0370
HLA-C HEM1 ICAM1NPM1 SH3BP5 ACTG1 ANPEP KIAA0446 ARHGDIB
HSPF1 NUMB HLA-A SAT HEF1 HNRPC NS1-BP  IFITM1 VDUP1
EIF4A1 LCP2 MAP3K8 NDUFV2  ICB-1 NCF4
HSPCASUI1   S100A11 MCP
IRAK1
Gene expression pattern LHL LHH LLH HLL HHL
Up-regulated
EarlyLateDown-regulated
TransientSustainedEarlyLate
Cytokines GRN OSM IL1RN SCYA4 IL8 SCYA20 S100A9 S100A8
SCYA3 IL1B GRO2PBEF
Receptors — PLAURHS1-2 CCRL2 ADORA2A SELL IL8RA IL8RB
TNFRSF1A HM74 CD44 FPR1 CD97 FCGR3A
ITGA5 TNFRSF10B  IFNGR1
LILRA2
PTPRC
Membrane trafficking — CLIC1ATP5E
FLOT1
ATP6J RAB1ATP2B1
NAF1 RAB7SLC16A3
ATP2A3 —
SSR2 SLC11A2 NAPA
Apoptosis regulators — TNFAIP3
PPIF BCL2A1 GSTTLp28 CFLAR PPIA PIG7
MCL1
Transcription modulators FOS ZFP36
ETR101 NFKBIA
HIF1A HMGIY DDIT3 NFE2L2
NFKB1 NFKBIE GC20
NFE2CEBPB
BTF3
DSIPI MNDA
ZNF220BRF2
EGR1 FOSBIRF1
ZNF148 ATF4
PSCDBP
Others BTG2HLA-E WBP2 G0S2 SGK MAP2K3 CPD EHD1 GCL ARPC1B EVI2B AMPD2
S100P TPT1 TNFAIP2 PLEK CD48 DUSP6 FTH1 LOC51669 DIFF48 CLN2
DYRK1APCOLN3 PTGS2 MACS PFN1 RTN4 TALDO1 MKNK1
GCP2 RPN2 MME HSPA10 CDKN1A LIMK2 PRG1 ERV1 MYO1E KIAA0370
HLA-C HEM1 ICAM1NPM1 SH3BP5 ACTG1 ANPEP KIAA0446 ARHGDIB
HSPF1 NUMB HLA-A SAT HEF1 HNRPC NS1-BP  IFITM1 VDUP1
EIF4A1 LCP2 MAP3K8 NDUFV2  ICB-1 NCF4
HSPCASUI1   S100A11 MCP
IRAK1
Gene expression pattern LHL LHH LLH HLL HHL

For abbreviations, see Table 4 and Appendix  A1.

Some genes were transiently up-regulated, peaking at 30-60 minutes, but returning to baseline levels by 2 hours after treatment. Among the earliest response mRNAs for known genes was that for the dual-specificity protein kinase DYRK1A.43,44 This is the human homolog of Drosophila minibrain and potentially a homolog of the S cerevisiae gene YAK1, a possible negative regulator of growth and cell cycle progression.45By 60 minutes after activation, the pattern changed with down-regulation of some mRNAs and strong up-regulation of others, among which was the mRNA for ETR101,46 a proline-rich cytoplasmic protein known as a sometimes unstable early activation protein in other systems.

## Discussion

The current study demonstrates that neutrophils are capable of extensive, rapid, and complex changes in gene expression. The changes in mRNA levels include both genes that are expressed and regulated in many cell types and genes that are expressed in a limited range of cells. Few of the regulated genes were strictly neutrophil-specific.

Activation of neutrophils by bacteria is a complex process that delivers multiple types of exogenous and endogenous signals to the cell. The bacterial lipopolysaccharide itself interacts with a specific receptor on the cell surface, and bacterially derived formyl peptides interact with the FMLP receptor (FPR1). Immunoglobulins and complement components associated with the bacteria stimulate an array of receptors present on neutrophils. An early consequence of neutrophil activation is the production of reactive oxygen species, and these in turn elicit a stress response from the cells. Neutrophil production of IL-1 or granulocyte-macrophage colony-stimulating factor (GM-CSF) presumably activates the corresponding receptors on the cell surface. The relative kinetics of induction of IL-8 and down-regulation of its receptors offer another potential for feedback effects on neutrophil activation.

In our study many known genes were induced on neutrophil activation including G0S2, ZFP36 (TTP/G0S24), PBEF (G0S9),ETR101, COPEB, FOSB (G0S3),FOS (G0S7), and the urokinase receptor(PLAUR). These corresponded to mRNA appearing in many other cell types during the transition from G0- to S-phase of the cell cycle or after other modes of activation. Other genes for widely used stress-response proteins, such as the heat shock products (HSPA10, HSPCA, HSPCB, and HSPF1) and the protein kinase MAP2K3, were also activated.

Two groups recently reported array analysis of changes in gene expression in fibroblasts in response to PDGF receptor47 or serum48 stimulation. For cells induced by receptor stimulation, more than 40% of the genes induced within 4 hours of stimulation were induced in neutrophils by 2 hours. This is an impressive overlap, as the fibroblasts were a transformed murine cell line held in 0.5% serum, whereas the neutrophils are postmitotic normal human cells maintained in high serum. This overlap emphasizes a commonality of very early response transcripts in mammalian cells, but suggests that quantitative considerations of the time and level of mRNA production may be central to understanding the differences in behavior of cell types.

None of the signal transduction or cell cycle genes induced in fibroblasts47 was regulated in neutrophils. However, more than 40% of immediate-early transcription factors were also up-regulated in neutrophils, and 7 of 8 genes classified as inflammation-related were also up-regulated in neutrophils. Itoh et al49 analyzed the 3′-end sequences of 1142 cDNA clones from neutrophils that were not intentionally activated and obtained sequences for 748 independent species. They listed 46 named genes for which they recovered 3 or more clones. In the present study we found that 90% of these genes were up-regulated on neutrophil activation.

Our data indicate that activated neutrophils are a source of physiologically significant trans-cellular signaling molecules. Measurements of IL-8 protein accumulation have shown that neutrophils produce IL-8 at about 1 ng per million cells per hour after exposure toE coli (J.D.G. and Y.V.B.K.S., unpublished results, February 1994). This corresponds to approximately 105molecules of IL-8 per cell per hour. In vitro, the cellular activating effects of IL-8 reach half-saturation levels at a concentration range of 0.5-1.0 nM. In vivo, human neutrophil counts commonly rise above 10 million cells per mL blood, enough to raise the concentration of IL-8 to physiologically effective levels within 1-2 hours. At sites of infection, tissue neutrophils are considerably more concentrated. Therefore, the levels of IL-8 production by neutrophils are physiologically very significant.

The levels of induced mRNA for a number of intracellular proteins are comparable to those for the more abundant cytokine mRNAs. This strongly suggests that the intracellular molecules are produced at levels that are physiologically significant, although the possibility of concomitant negative control of translation rate of specific mRNAs has not been investigated. More caution is necessary in interpretation of down-regulation of mRNA. The down-regulation will only correspond to changes in protein level if the protein normally has a short half-life or is specifically degraded following activation of the neutrophils. Some of the down-regulation is undoubtedly due to stopping transcription of relatively short-lived mRNAs. This change would not produce synchronous effects on all mRNA both because they have differing half-lives and because transcription may not be down-regulated simultaneously on different genes. Some mRNAs that are stable in cells treated with 5,6-dichloro-1-β-d-ribofuranosylbenzimidazole (DRB) disappear rapidly after exposure to bacteria (data not shown). Studies with actinomycin D indicate that the mRNA for certain chemokine receptors is destabilized on LPS activation of cells,50and this destabilization is blocked by simultaneous, but not by delayed, addition of the transcription inhibitor. In any case, the events leading to destabilization are heterogeneous.

CC-chemokines, such as SCYA3, SCYA4, and SCYA20, were up-regulated. CXC chemokines, such as IL-8, GRO1, and GRO2, were also up-regulated. Although GRO1 and GRO2 share 90% identity at the deduced amino acid level, and both have melanoma growth-stimulating activity, their expression patterns were different. GRO1 was induced by KIM5 more strongly than by nonpathogenic bacteria, but the induction of GRO2 seen with KIM6 did not occur with KIM5.

Although both CXCR1 (IL-8RA, a receptor that is relatively specific for IL-8) and CXCR251 (IL-8RB, a receptor activated by other CXC chemokines including GRO1) were down-regulated, KIM5 fully inhibited gene expression of CXCR1, but not CXCR2. CXCR1 and CXCR2 are regulated in different modes by CXC chemokines and play diverse roles in mediating the inflammatory process.52 The putative G protein–coupled receptors CCRL2 (HCR) and HM74 were prominently up-regulated. HCR was previously identified in public databases as CCR6 (a receptor for SCYA20/LARC/MIP3A), but recently it has been described as a distinctive receptor, CCRL2. The sequence of CCRL2 in GenBank Accession no. U95626 is identified as CCR6, but differs from the sequence of U68030 CCR6 mRNA, so it remains uncertain whether CCRL2 is the receptor of SCYA20/LARC/MIP3A. The presence of both SCYA20/LARC/MIP3A and its receptor on the same cells would imply an autocrine loop. The strong induction of HM74 in human neutrophils suggests its utility as a clinical parameter and/or a drug target in inflammatory disorders. Overall, the responses to some stimuli were down-regulated, and new response pathways could be established. Whether these maintain or modulate the active state or have other functions remain to be determined, but they probably play important roles in the early evolution of the inflammatory process. A suggestion to explain the virulence of KIM5 is that the loss of production of the primary activating and chemo-attractant cytokine IL-8 would decrease the possibility that neutrophils which have ingested bacteria would attract additional neutrophils to sites of inflammation. The net effects of up-regulation of IL-1 and its receptor antagonist IL-1RN are uncertain, but they could provide an additional measure of feedback.

The balance between apoptotic and necrotic cell death in neutrophils plays an important role in the control of inflammation. Neutrophils accumulate in large numbers at sites of inflammation, forming tissue infiltrates and pus. Necrotic death of these cells releases toxic granule contents, such as elastase and collagenase, whereas removal of apoptotic neutrophils by macrophages protects surrounding tissues from such damage.53,54 However, inhibition of neutrophil apoptosis may augment host defense against infection by prolongation of functional longevity of the cells.55 When cultured in vitro, neutrophils rapidly undergo apoptosis, which is preceded by intracellular acidification.56 G-CSF and a variety of inflammatory mediators delay programmed cell death, in part by up-regulation of expression of Bcl-X1, but not other Bcl-2 family members.53,56-58 The current data suggest that other proteins, such as BCL2A1, MCL1, PPIF, TNFAIP3, and perhaps spermidine/spermine N1-acetyltransferase (SAT), may be important for the regulation of neutrophil apoptosis in response to infection.

Increases in mRNA for genes regulating transcription or translation were observed 2 hours after activation. These include theCOPEB gene, which is reported to stimulate expression of genes lacking a TATA box. In cells exposed to KIM5, 12 of 14 transcription-modifying genes examined were present at levels more similar to those of control neutrophils than of the neutrophils stimulated by the other bacteria.

In time-course studies we found that NFKBIA (IκBα) was induced byE coli K12 in 30 minutes, but NFKB1 (NF-κB) induction was observed after 60 minutes. In contrast to NFKBIA, NFKBIE (IκBε) was activated rather later. IκB is known as a negative regulator of NF-κB by formation of stable IκB/NF-κB complexes. This retains NF-κB in the cytoplasm until the NF-κB activation signal is received.59 This asynchronous activation of reciprocal transcription factors presumably reflects a transient activation of NFκB-dependent genes.

Activation by nonpathogens, but not by the pathogenic KIM5, caused down-regulation of mRNAs for some antibacterial products, including the phagocyte oxidase (PHOX) system (NCF1, NCF2, and NCF4) generating reactive oxygens,60 and the calgranulins (S100A8 and S100A9). In contrast, the free radical scavenging enzyme SOD2 was up-regulated by the nonpathogens. The reactive oxygen system is regulated by external stimuli61 and is auto-cytotoxic for neutrophils. Its down-regulation may contribute to the prolongation of life span of activated cells in inflammation.

Decay-accelerating factor (DAF) was up-regulated by nonpathogen, but MCP,62 a cofactor of serine protease factor I for inactivation of complements C3b and C4b, was down-regulated. Although both play a protective role in host cells against homologous complement, MCP is also the receptor for various viruses and bacterial pathogens. CD97,63 the receptor for DAF, is regulated oppositely to DAF, so that the cells may become desensitized to DAF.

Overall, the patterns of induction or disappearance of mRNAs for genes of known function can largely be rationalized in terms of the biologic role of neutrophils. Several different antiapoptotic mechanisms are set in play in an asynchronous fashion. This response would allow neutrophils that ingested nonpathogenic material to survive longer, potentially migrating to restricted tissue areas and also degrading ingested material. Additional defensive changes in the neutrophils include production of DAF. The cells change their own cytokine responsiveness and begin producing a range of new cytokines. These would not only transmit inflammatory signals and recruit unstimulated neutrophils, but they would also further stimulate activated neutrophils, contributing to the congregation of activated neutrophils and hence to abscess formation. Changes in the levels of intracellular signaling molecules might well change the responses to stimulation of pre-existing receptors. Membrane trafficking is accelerated perhaps related to ingestion of bacteria and discharge of preformed granules. There is also a previously unappreciated transition from early to delayed responses at the level of mRNA production.

In summary, nonpathogenic gram-negative bacteria induce a marked change in the patterns of gene expression in neutrophils, indicating massive changes in cytokine output and prolongation of cell survival. These changes imply that neutrophils are important effectors of the progression of the cellular inflammatory response. Interruption of these changes by pathogens, such as Y pestis KIM5, could be, at least in part, responsible for the failure to contain the infectious process.

Supplementary information is available on our Web site.67

We thank Andrea M. Neuman, Carolyn Padden, Angela Plette, Anne-Marie Quinn, and Connie Whitney for technical assistance and Dov Greenbaum for Web site set-up.

### GENE SYMBOLS FOR TABLES 4, 5, AND 6

Y.V.B.K.S. and S.Y. contributed equally to this work.

Submitted August 1, 2000; accepted November 30, 2000.

Supported by grants AI22176 (J.D.G.), CA42556 (S.M.W.), and DK54369 (P.E.N.) from the National Institutes of Health, Bethesda, MD, and by a research grant (P.E.N.) from the Arthritis Foundation.

## References

References
1
Newburger
PE
Parmley
RT
Neutrophil structure and function.
Hematology: Basic Principles and Practice.
Hoffman
R
Benz
EJ
Jr
Shattil
SJ
Furie
B
Cohen
HJ
Silberstein
LE
1995
738
750
Churchill Livingstone
New York, NY
2
Ward
PA
Lentsch
AB
The acute inflammatory response and its regulation.
Arch Surg.
134
1999
666
669
3
Wiseman
H
Halliwell
B
Damage to DNA by reactive oxygen and nitrogen species: role in inflammatory disease and progression to cancer.
Biochem J.
313
1996
17
29
4
Lindemann
A
Riedel
D
Oster
W
Ziegler-Heitbrock
HW
Mertelsmann
R
Herrmann
F
Granulocyte-macrophage colony-stimulating factor induces cytokine secretion by human polymorphonuclear leukocytes.
J Clin Invest.
83
1989
1308
1312
5
Cassatella
MA
Gasperini
S
Russo
MP
Cytokine expression and release by neutrophils.
832
1997
233
242
6
Lord
PC
Wilmoth
LM
Mizel
SB
McCall
CE
Expression of interleukin-1 alpha and beta genes by human blood polymorphonuclear leukocytes.
J Clin Invest.
87
1991
1312
1321
7
Strieter
RM
Kasahara
K
Allen
R
Showell
HJ
Standiford
TJ
Kunkel
SL
Human neutrophils exhibit disparate chemotactic factor gene expression.
Biochem Biophys Res Commun.
173
1990
725
730
8
Miller
EJ
Nagao
S
Carr
FK
Noble
JM
Cohen
AB
Interleukin-8 (IL-8) is a major neutrophil chemotaxin from human alveolar macrophages stimulated with staphylococcal enterotoxin A (SEA).
Inflamm Res.
45
1996
386
392
9
Grenier
A
Dehoux
M
Boutten
A
et al
Oncostatin M production and regulation by human polymorphonuclear neutrophils.
Blood.
93
1999
1413
1421
10
Kasama
T
Strieter
RM
Lukacs
NW
Lincoln
PM
Burdick
MD
Kunkel
SL
Interferon gamma modulates the expression of neutrophil-derived chemokines.
J Investig Med.
43
1995
58
67
11
Hachicha
M
Rathanaswami
P
Naccache
PH
McColl
SR
Regulation of chemokine gene expression in human peripheral blood neutrophils phagocytosing microbial pathogens.
J Immunol.
160
1998
449
454
12
Ohta
TM
Kasama
T
Hanyuuda
M
et al
Interleukin-13 down-regulates the expression of neutrophil-derived macrophage inflammatory protein-1 alpha.
Inflamm Res.
47
1998
361
368
13
Prashar
Y
Weissman
SM
Analysis of differential gene expression by display of 3′ end restriction fragments of cDNAs.
Proc Natl Acad Sci U S A.
93
1996
659
663
14
Goguen
JD
Yother
J
Straley
SC
Genetic analysis of the low calcium response in Yersinia pestis mu d1(Ap lac) insertion mutants.
J Bacteriol.
160
1984
842
848
15
Une
T
Brubaker
RR
Roles of V antigen in promoting virulence and immunity in yersiniae.
J Immunol.
133
1984
2226
2230
16
Fetherston
JD
Schuetze
P
Perry
RD
Loss of the pigmentation phenotype in Yersinia pestis is due to the spontaneous deletion of 102 kb of chromosomal DNA which is flanked by a repetitive element.
Mol Microbiol.
6
1992
2693
2704
17
Forsberg
A
Rosqvist
R
Wolf-Watz
H
Regulation and polarized transfer of the Yersinia outer proteins (Yops) involved in antiphagocytosis.
Trends Microbiol.
2
1994
14
19
18
Perry
RD
Fetherston
JD
Yersinia pestis—etiologic agent of plague.
Clin Microbiol Rev.
10
1997
35
66
19
Cornelis
GR
Boland
A
Boyd
AP
et al
The virulence plasmid of Yersinia, an antihost genome.
Microbiol Mol Biol Rev.
62
1998
1315
1352
20
Perry
RD
Straley
SC
Fetherston
JD
Rose
DJ
Gregor
J
Blattner
FR
DNA sequencing and analysis of the low-Ca2+-response plasmid pCD1 of Yersinia pestis KIM5.
Infect Immun.
66
1998
4611
4623
21
Subrahmanyam
YV
N
Newburger
PE
Weissman
SM
A modified method for the display of 3′-end restriction fragments of cDNAs: molecular profiling of gene expression in neutrophils.
Methods Enzymol.
303
1999
272
297
22
Mentzer
SJ
Guyre
PM
Burakoff
SJ
Faller
DV
Spontaneous aggregation as a mechanism for human monocyte purification.
Cell Immunol.
101
1986
312
319
23
Ginsburg
D
Handin
RI
Bonthron
DT
et al
Human von Willebrand factor (vWF): isolation of complementary DNA (cDNA) clones and chromosomal localization.
Science.
228
1985
1401
1406
24
Sambrook J, Fritsch EF, Maniatis T. Molecular Cloning: A Laboratory Manual. Vol. 2. 2nd ed. Cold Spring Harbor Laboratory Press; 1989: 7.53-7.55.
25
Gatti
RA
Concannon
P
Salser
W
Multiple uses of Southern blots.
BioTechniques.
2
1984
148
155
26
Katz
RA
Erlanger
BF
Guntaka
RV
Evidence for extensive methylation of ribosomal RNA genes in a rat XC cell line.
Biochim Biophys Acta.
739
1983
258
264
27
Johnson
CV
Singer
RH
Lawrence
JB
Fluorescent detection of nuclear RNA and DNA: implications for genome organization.
Methods Cell Biol.
35
1991
73
99
28
Subrahmanyam
YVBK
Yamaga
S
Newburger
PE
Weissman
SM
A modified approach for the efficient display of 3′ end restriction fragments of cDNAs.
Differential Display: A Practical Approach. Practical Approach Series.
Leslie
RA
Robertson
HA
2000
101
129
Oxford University Press
Oxford, England
29
Altschul
SF
TL
Schaffer
AA
et al
Gapped BLAST and PSI-BLAST: a new generation of protein database search programs.
Nucleic Acids Res.
25
1997
3389
3402
30
Gordon
D
Abajian
C
Green
P
Consed: a graphical tool for sequence finishing.
Genome Res.
8
1998
195
202
31
Rettenmier
CW
Roussel
MF
Sherr
CJ
The colony-stimulating factor 1 (CSF-1) receptor (c-fms proto-oncogene product) and its ligand.
J Cell Sci Suppl.
9
1988
27
44
32
Wilusz
J
Pettine
SM
Shenk
T
Functional analysis of point mutations in the AAUAAA motif of the SV40 late polyadenylation signal.
Nucleic Acids Res.
17
1989
3899
3908
33
Schmid
CW
Does SINE evolution preclude Alu function?
Nucleic Acids Res.
26
1998
4541
4550
34
Struyf
A
Hubert
M
Rousseeuw
PJ
Integrating robust clustering techniques in S-PLUS.
Comput Stat Data Anal.
26
1997
17
37
35
Sodeinde
OA
Subrahmanyam
YV
Stark
K
Quan
T
Bao
Y
Goguen
JD
A surface protease and the invasive character of plague.
Science.
258
1992
1004
1007
36
Welkos
SL
Friedlander
AM
Davis
KJ
Studies on the role of plasminogen activator in systemic infection by virulent Yersinia pestis strain C092.
Microb Pathog.
23
1997
211
223
37
Nakajima
R
Brubaker
RR
Association between virulence of Yersinia pestis and suppression of gamma interferon and tumor necrosis factor alpha.
Infect Immun.
61
1993
23
31
38
Schesser
K
Spiik
AK
Dukuzumuremyi
JM
Neurath
MF
S
Wolf-Watz
H
The yopJ locus is required for Yersinia-mediated inhibition of NF-kappaB activation and cytokine expression: YopJ contains a eukaryotic SH2-like domain that is essential for its repressive activity.
Mol Microbiol.
28
1998
1067
1079
39
Perera
LP
Waldmann
TA
Activation of human monocytes induces differential resistance to apoptosis with rapid down regulation of caspase-8/FLICE.
Proc Natl Acad Sci U S A.
95
1998
14308
14313
40
Lin
EY
Orlofsky
A
Wang
HG
Reed
JC
Prystowsky
MB
A1, a Bcl-2 family member, prolongs cell survival and permits myeloid differentiation.
Blood.
87
1996
983
992
41
Wu
MX
Ao
Z
KV
Wu
R
Schlossman
SF
IEX-1L, an apoptosis inhibitor involved in NF-κB-mediated cell survival.
Science.
281
1998
998
1001
42
Marzo
I
Brenner
C
Zamzami
N
et al
The permeability transition pore complex: a target for apoptosis regulation by caspases and bcl-2-related proteins.
J Exp Med.
187
1998
1261
1271
43
Kentrup
H
Becker
W
Heukelbach
J
et al
Dyrk, a dual specificity protein kinase with unique structural features whose activity is dependent on tyrosine residues between subdomains VII and VIII.
J Biol Chem.
271
1996
3488
3495
44
Becker
W
Joost
HG
Structural and functional characteristics of Dyrk, a novel subfamily of protein kinases with dual specificity.
Prog Nucleic Acid Res Mol Biol.
62
1999
1
17
45
Kassis
S
Melhuish
T
Annan
RS
et al
Saccharomyces cerevisiae yak1p protein kinase autophosphorylates on tyrosine residues and phosphorylates myelin basic protein on a C-terminal serine residue.
Biochem J.
348
2000
263
272
46
Shimizu
N
Ohta
M
Fujiwara
C
et al
Expression of a novel immediate early gene during 12-O-tetradecanoylphorbol-13-acetate-induced macrophagic differentiation of HL-60 cells.
J Biol Chem.
266
1991
12157
12161
47
Fambrough
D
McClure
K
Kazlauskas
A
Lander
ES
Diverse signaling pathways activated by growth factor receptors induce broadly overlapping, rather than independent, sets of genes.
Cell.
97
1999
727
741
48
Iyer
VR
Eisen
MB
Ross
DT
et al
The transcriptional program in the response of human fibroblasts to serum.
Science.
283
1999
83
87
49
Itoh
K
Okubo
K
Utiyama
H
Hirano
T
Yoshii
J
Matsubara
K
Expression profile of active genes in granulocytes.
Blood.
92
1998
1432
1441
50
Xu
L
Rahimpour
R
Ran
L
et al
Regulation of CCR2 chemokine receptor mRNA stability.
J Leukoc Biol.
62
1997
653
660
51
Murphy
PM
Neutrophil receptors for interleukin-8 and related CXC chemokines.
Semin Hematol.
34
1997
311
318
52
Feniger-Barish
R
Ran
M
Zaslaver
A
Ben-Baruch
A
Differential modes of regulation of cxc chemokine-induced internalization and recycling of human CXCR1 and CXCR2.
Cytokine.
11
1999
996
1009
53
Haslett
C
Granulocyte apoptosis and inflammatory disease.
Br Med Bull.
53
1997
669
683
54
Savill
J
Apoptosis in resolution of inflammation.
J Leukoc Biol.
61
1997
375
380
55
Lee
A
Whyte
MK
Haslett
C
Inhibition of apoptosis and prolongation of neutrophil functional longevity by inflammatory mediators.
J Leukoc Biol.
54
1993
283
288
56
Gottlieb
RA
Giesing
HA
Zhu
JY
Engler
RL
Babior
BM
Cell acidification in apoptosis: granulocyte colony-stimulating factor delays programmed cell death in neutrophils by up-regulating the vacuolar H(+)-ATPase.
Proc Natl Acad Sci U S A.
92
1995
5965
5968
57
Hannah
S
Cotter
TG
Wyllie
AH
Haslett
C
The role of oncogene products in neutrophil apoptosis.
Biochem Soc Trans.
22
1994
253S
58
Weinmann
P
Gaehtgens
P
Walzog
B
Bcl-Xl- and Bax-α-mediated regulation of apoptosis of human neutrophils via caspase-3.
Blood.
93
1999
3106
3115
59
Huxford
T
Huang
DB
Malek
S
Ghosh
G
The crystal structure of the IkappaBalpha/NF-kappaB complex reveals mechanisms of NF-kappaB inactivation.
Cell.
95
1998
759
770
60
Babior
BM
Blood.
93
1999
1464
1476
61
Yamaga
S
Okamura
S
Otsuka
T
Niho
Y
Effect of granulocyte-macrophage colony-stimulating factor on chemiluminescence of human neutrophils.
Int J Cell Cloning.
7
1989
50
58
62
Seya
T
Hirano
A
Matsumoto
M
Nomura
M
Ueda
S
Human membrane cofactor protein (MCP, CD46): multiple isoforms and functions.
Int J Biochem Cell Biol.
31
1999
1255
1260
63
Qian
YM
Haino
M
Kelly
K
Song
WC
Structural characterization of mouse CD97 and study of its specific interaction with the murine decay-accelerating factor (DAF, CD55).
Immunology.
98
1999
303
311
64
Johnston
RF
Pickett
SC
Barker
DL
Electrophoresis.
11
1990
355
360
65
National Center for Biotechnology Information (NCBI), National Library of Medicine (NLM). LocusLink Web site. Available at:http://www.ncbi.nlm.nih.gov/LocusLink/ Accessed November 9, 2000.
66
National Center for Biotechnology Information, National Library of Medicine. Unigene Web site. Available at:http://www.ncbi.nlm.nih.gov/ UniGene/ Accessed November 9, 2000.
67
The Department of Molecular Biophysics and Biochemistry, Yale University, and the Department of Genetics, Yale University School of Medicine. Available after April 15, 2001 at:http://bioinfombb.yale.edu/expression/neutrophil/

## Author notes

Sherman M. Weissman, Department of Genetics, Boyer Center for Molecular Medicine, Yale University School of Medicine, 295 Congress Ave, New Haven, CT 06536-0812; e-mail:sherman.weissman@yale.edu.