Microarrays have been used to evaluate the expression of thousands of genes in various tissues. However, few studies have investigated the change in gene expression profiles in one of the most easily accessible tissues, whole blood. We utilized an acute inflammation model to investigate the possibility of using a cDNA microarray to measure the gene expression profile in the cells of whole blood. Blood was collected from male Sprague-Dawley rats at 2 and 6 h after treatment with 5 mg/kg (ip) LPS. Hematology showed marked neutrophilia accompanied by lymphopenia at both time points. TNF-α and IL-6 levels were markedly elevated at 2 h, indicating acute inflammation, but by 6 h the levels had declined. Total RNA was isolated from whole blood and hybridized to the National Institute of Environmental Health Sciences Rat Chip v.3.0. LPS treatment caused 226 and 180 genes to be differentially expressed at 2 and 6 h, respectively. Many of the differentially expressed genes are involved in inflammation and the acute phase response, but differential expression was also noted in genes involved in the cytoskeleton, cell adhesion, oxidative respiration, and transcription. Real-time RT-PCR confirmed the differential regulation of a representative subset of genes. Principal component analysis of gene expression discriminated between the acute inflammatory response apparent at 2 h and the observed recovery underway at 6 h. These studies indicate that, in whole blood, changes in gene expression profiles can be detected that are reflective of inflammation, despite the adaptive shifts in leukocyte populations that accompany such inflammatory processes.
microarray technology has become increasingly useful for gaining insights into biological and disease processes. For animal and clinical studies, minimally invasive procedures would be particularly useful for obtaining samples for cDNA microarray analysis. Whole blood is a logical candidate as an accessible biofluid, and circulating leukocytes would contain informative transcripts as a first line of immune defense and sentinels for many disease processes (48). Therefore, gene expression changes in leukocytes could possibly serve as early warnings of potential health threats. Yet, to date, few microarray studies have been performed with whole blood, largely because of logistical and technical challenges that accompany collection, storage, processing, and isolation of high-quality RNA from this tissue. Recently, blood RNA isolation reagents have been developed that immediately stabilize RNA upon collection and produce high-quality RNA (38). One such method involves the use of PAXgene blood RNA tubes (PreAnalytiX/QIAGEN, Hilden, Germany). This approach avoids the complications inherent in other approaches such as Ficoll density gradient isolation of white cells with or without subsequent primary culture before RNA isolation. Such time-laden manipulations can significantly alter gene expression patterns compared with data obtained from RNA isolated relatively quickly from freshly collected stabilized whole blood (4, 9, 34).
To assess the utility of whole blood as a study tissue, we chose a well-characterized rodent model for short-term acute inflammation induced by lipopolysaccharide (LPS) (42). LPS is a component of the bacterial cell wall of gram negative bacteria that activates toll-like receptor (TLR)-4 in mammals. TLR-4 activation results in production and release of cytokines such as tumor necrosis factor (TNF)-α, interleukin (IL)-1β, and IL-6 by macrophages and neutrophils (polymorpholeukocytes), which serve as endogenous mediators of inflammation by receptor-mediated interactions with a variety of target cells (5). LPS-induced cytokine release results in a classic “left shift” in hematological parameters observed during bacterial infection (14). This shift, in rats, consists of an initial neutropenia with lymphophilia within ∼30 min of treatment followed by a rapid reversal to a neutrophilic/lymphopenic state, with peaks of neutrophilia at 2 and 6 h (44). Activated neutrophils, together with macrophages, produce and release reactive oxygen species, lipid mediators, and cytokines in response to LPS. In addition, they acquire adhesion and phagocytotic properties, while also recruiting immune cells to amplify and regulate the immune response. Although LPS-mediated responses were previously believed to be mediated by release of preformed pools of immune reactive substances, recent research has demonstrated a robust gene expression response of leukocytes to acute inflammatory agents, indicating that the molecular biology of innate immunity may be more complex than previously thought (28, 29). Yet these studies have been performed utilizing isolated or cultured cells with the noted logistical and biological complications.
In the current study, we report an approach that allows for immediate stabilization of RNA from all blood cells in a rapid and simple manner. Here we present data that demonstrate, for the first time, that exposure of animals to a known acute inflammatory agent elicits a temporal change in the gene expression profile obtained from whole blood that could be helpful for assessing the health effects of agents or diseases that involve inflammatory processes.
Animals and treatments.
Twelve-week-old male Sprague-Dawley rats were obtained from Taconic Laboratories (Germantown, NY) and acclimated for 14 days. The animals were housed at three per cage in polycarbonate cages (Lab Products, Maywood, NJ) with Sani-chips (PJ Murphy Forest Products, Montville, NJ). The animal rooms were maintained at 21–22°C and 48–53% relative humidity, with a 12:12-h dark-light cycle. NIH07 diet and tap water were provided ad libitum. Groups of six animals were dosed at ∼8 AM with either 0.9% saline (control) or LPS (Sigma-Aldrich; catalogue no. L-2637) by intraperitoneal injection at 5 mg/kg. Animals were euthanized either 2 or 6 h after LPS exposure by CO2 asphyxiation, immediately followed by the drawing of ∼8 ml of blood from the anterior vena cava. Protocols were approved by the National Institute of Environmental Health Sciences (NIEHS) Committee on Animal Care, in compliance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals.
ELISA assays for TNF-α, IL-6, and IL-β1 were performed as recommended by the manufacturer (Biosource International, Camarillo, CA). Briefly, 100 μl of the standard diluent buffer were added to the zero wells. Blank wells had only chromogen and stop solution added at the end of the assay. A standard curve was generated with a recombinant version of the cytokine of interest. The serum samples were diluted with the standard buffer or used at a 1× concentration, depending on the cytokine to be tested. The serum was incubated with a biotinylated antibody against the cytokine of interest for at least 1 h, washed to remove nonspecifically bound proteins, and incubated with a streptavidin-horseradish peroxidase (HRP) working solution as a secondary antibody. The plate was then incubated for 30 min at room temperature, washed again to remove nonspecific binding of proteins, and incubated with stabilized chromogen solution in the dark for 30 min. A stop solution was then added, and the absorbance at 450 nm was detected via the use of a plate reader. The final concentration of the samples was extrapolated from the standard curve.
The blood samples were assayed using the Technicon H*1 hematology analyzer (Bayer, Tarrytown, NY). The instrument performs white blood cell (WBC) counts by two independent methods, with the values between the two methods agreeing within 10%. Linearity for the WBC channels was 99.0 × 103/μl. Manual WBC counts and smear estimates were used to confirm values. To monitor red cell parameters, spun microhematocrits were performed for each specimen for comparison with the automated hematocrit [which is calculated from the directly measured red blood cell (RBC) and mean corpuscular volume (MCV)]. Control samples were assayed after every 10th specimen. Wright-Giemsa-stained blood smears were prepared, and manual WBC differential counts were performed on at least the first and last sample in each treatment group. RBC and platelet morphology were also examined. The smears were stained using the Ames Hema-Tek II automated slide stainer (Bayer, Ames Div., Elkhart, IN).
Blood (2.5 ml/tube, 2 tubes/rat) was collected in PAXgene vacutainer tubes (PreAnalytiX/QIAGEN). RNA was isolated according to the manufacturer’s protocol, with minor modifications. Briefly, samples were allowed to remain at room temperature for no longer than 2 h and were processed with no intervening cold storage. Also, centrifugation time after proteinase K digestion was increased from 3 to 20 min to obtain a tighter debris pellet. RNA quality was assessed with an Agilent Bioanalyzer (Palo Alto, CA) to ensure that samples with intact 18S and 28S ribosomal RNA peaks and a low degradation factor were used for microarray analysis (3). Equal amounts of the highest-quality RNA from three time-matched controls were pooled together for gene expression analysis.
cDNA microarray chip preparation.
Six thousand seven hundred rat clone cDNAs (Research Genetics, Huntsville, AL) coding for 5,364 known genes and expressed sequence tags (ESTs; http://dir.niehs.nih.gov/microarray/chips.htm) were printed as previously described (11, 20). The methods used to produce the chips are available at http://dir.niehs.nih.gov/microarray/methods.htm.
cDNA microarray analysis.
The cDNA microarray analysis has been described in detail (20). Briefly, the cDNA targets were prepared from 35 μg of total RNA by oligo(dT)-primed polymerization, using Superscript II RT (Life Technologies, Gaithersburg, MD). Analyses were performed twice per sample, using a dye-reversal procedure in which mRNA from the time-matched control pool was labeled with Cy3 and mRNA from individual treated rats was labeled with Cy5. In the second analysis, control mRNA was labeled with Cy5 and mRNA from treated rats was labeled with Cy3. This dye reversal helps to minimize error due to fluor-associated bias. Fluorescent intensities were measured with an Agilent DNA Microarray scanner (Palo Alto, CA). For each individual chip, Array Suite v.2 (Scanalytics, Fairfax, VA) software was used for data acquisition and image analysis. Data were normalized by use of the linear regression model from the Array Suite software. After data from the dye-reversed chip pairs were averaged, intensity values corresponding to each gene on the cDNA microarray chips from the Cy3 and Cy5 channels were represented as a ratio of LPS-exposed and time-matched, vehicle-treated control blood. Genes with altered transcript levels in the treated samples relative to the pooled controls were determined at the 99% confidence level (7). To assure reproducibility of the observed gene expression changes between samples, we considered only genes that were determined to be differentially expressed in at least two of the three animals for further analysis (based on a binomial distribution, using MicroArray Project System; P ≤ 0.00000015) (6). Cluster and TreeView were used to visualize gene changes occurring at each time point (13). Principal component analysis was used as a mathematical tool to present the complex data in a two-dimensional space (23) (http://dir.niehs.nih.gov/microarray/datamining). The microarray data have also been submitted to the Gene Expression Omnibus (GEO) database. Accession numbers are as follows: platform, GPL1344; samples, GSM28457, GSM28539, GSM28540, GSM28541, GSM28542, GSM28543, GSM28544, GSM28545, GSM28546, GSM28547, GSM28548, and GSM28549; and series, GSE1658.
Quantitative real-time RT-PCR was performed using FAM-MGB probes and primers, either predesigned (Applied Biosystems, Foster City, CA) or custom designed using the Primer Express 1.5 software (Applied Biosystems) and ordered from Applied Biosystems custom oligo synthesis service. Elongation factor-1α forward primer, reverse primer, and probe sequences are as follows: 5′-GTCTGGTGATGCTGCCATTG-3′, 5′-GGAGGGTAGTCAGAGAAGCTTTCA-3′, and 5′-Fam-TGACATGGTCCCTGGCAAGCCC-Mgb-3′. Likewise, for ribosomal protein S6 they are 5′-AAGAAGCCCAGGACCAAAGC-3′, 5′-TGCAGCCTCCTCCTTGTTTTT-3′, and 5′-Fam-CCAAGATTCAGCGTCTTGTTACTCCCCG-Mgb-3′. For MRP14 they are 5′-TGCTGATGGGAAAGTTGATCTTT-3′, 5′-GATGTTACATGGCGACCTCTTAATT-3′, and 5′-Fam-TGAGAAGCTGCATGAGAACAACCCAC-Mgb-3′. For tissue inhibitor of metalloproteinase (TIMP)2 they are 5′-CTTGACATCGAGGACCCGTAA-3′, 5′-ACCCTCAGAGGCTTTTCAATTG-3′, and 5′-Fam-CTGACAGAGCCCC-Mgb-3′. Total RNA was reverse transcribed (125 ng/μl) using Superscriptase II RT (Invitrogen, Carlsbad, CA). Briefly, RNA was incubated with 1× RT buffer, dithiothreitol (50 μM), and RNasin (0.3 U/μl) in 40-μl reactions at 25°C for 10 min, followed by 42°C for 2 min. RT (0.3 U/μl) was added, and the samples were incubated at 42°C for an additional 50 min, followed by a 70°C incubation for 15 min to inactivate the enzyme. The cDNA (2 μl) was then used in subsequent real-time RT-PCR reactions prepared with appropriate primers (900 nM) and probes (100 μM) and a PCR master mix (PE Applied Biosystems) in a final volume of 20 μl. Thermal cycling and real-time detection of the fluorescence were carried out using an Applied Biosystems 7700 Sequence Detection system, using the following amplification parameters: denaturation at 94°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 60 s. Cycle threshold (CT) values were determined as the cycle where the ROX-normalized fluorescence over background was significantly above background levels. Fold induction was calculated using the formula 2−ΔΔCt, where the ΔΔCT = average ΔCT (treated) − average ΔCT (control), where ΔCT is the difference between the CT of the gene of interest in either the control or LPS-treated sample and the CT of the 18S ribosomal gene in either the control or LPS-treated sample; ΔΔCT is the difference between the ΔCT of control and LPS-treated samples.
Data are presented as means ± SE. Student’s t-test (2 sided, equal variance) was used to compare ELISA data from control and treated rats at each time point. ANOVA was used to compare hematology data across samples. A P value ≤ 0.05 was considered statistically significant.
LPS treatment caused a marked neutrophilia, lymphopenia, and thrombocytopenia (Table 1). Absolute neutrophil counts increased 103% at 2 h and 183% at 6 h, whereas lymphocyte counts decreased 63% at 2 h and 78% at 6 h (P ≤ 0.01). Monocyte counts decreased 88% at 2 h and 81% at 6 h. Overall, total white cell counts decreased at both time points, 26 and 37%, respectively. Platelet counts decreased 17 and 52% at 2 and 6 h, respectively. Only the decrease in lymphocytes was statistically significant between 2 and 6 h. These shifts could have a significant effect on interpretation of differential gene expression and will be addressed in the discussion.
Serum TNF-α, IL-6, and IL-1β concentrations were measured in all control (n = 6) and treated (n = 6) animals. Serum TNF-α concentrations were undetectable in control animals at either time point, but LPS treatment caused TNF-α concentrations to increase to 62 ± 24 ng/ml (P ≤ 0.03) at 2 h before decreasing to the limits of detection at 0.10 ± 0.06 ng/ml at 6 h. IL-6 concentrations were also undetectable in control animals and increased to 16.3 ± 4.0 ng/ml (P ≤ 0.003) in LPS-treated animals 2 h before being reduced to 11.2 ± 3.7 ng/ml at 6 h (P ≤ 0.02). LPS treatment produced a small, statistically insignificant increase in IL-1β serum concentrations after 2 h compared with controls, which was also observed after 6 h (data not shown). These results are congruent with a previous study that investigated the temporal profiles of serum cytokine protein levels in the rat after LPS treatment (24).
Isolation of high-quality RNA.
All total RNA was of high quality and acceptable yield for microarray analysis. A representative electropherogram (Fig. 1) shows that the 18S and 28S peaks were intact, with minimal signs of degradation (28S-to-18S ratio of 1.70 ± 0.03). Agilent’s degradometer software returned low degradation factors, with only one sample being flagged with a cautionary “yellow” (3). This sample was not used in the differential gene expression studies. Total RNA yields ranged from 10 to 58 μg/ml of whole blood and averaged 32.4 ± 2.4 μg/ml across all samples (Table 2).
Patterns of gene expression associated with LPS exposure.
Because one of the goals of this study was to determine the plausibility of utilizing whole blood to measure gene expression changes due to acute inflammation, we used only RNA from blood of the three highest responders to LPS treatment at each time point (n = 6/group), as assessed by elevated TNF-α and IL-6 levels. Overall, there was a robust gene expression response in blood cells during this acute inflammation event, both at 2 h, when the serum biochemical indicators of the LPS were most elevated, and at 6 h, when the animals were beginning to recover. For the entire study, a total of 297 differentially expressed genes were identified at a 99% confidence interval (Fig. 2). Differences and similarities were noted between the time points: 188 genes were differentially expressed at only one of the time points, whereas 109 genes were differentially expressed at both time points. No transcripts changed from being significantly more abundant relative to controls in the LPS-treated rats at 2 h to significantly less abundant at 6 h, or vice versa. Of the genes determined to be significantly more abundant in the LPS-treated rats, 27 transcripts were in common between the time points, 75 transcripts were significant only at 2 h, and 36 transcripts were significant only at 6 h (Fig. 2, left). Of the genes determined to be significantly less abundant in the LPS-treated rats, 82 were in common between the time points, 42 transcripts were significant only at 2 h, and 35 transcripts were significant only at 6 h (Fig. 2, right).
To evaluate the gene expression patterns at both time points, we performed two-dimensional unsupervised hierarchical clustering on the 297 differentially expressed genes and ESTs from both time points (Fig. 3). This tool groups genes into clusters based on the similarity of their expression patterns (13). When applied to gene expression data, it provides a measure of similarity of the pattern of gene expression both among individual animals and among the individual genes. With the use of this analysis approach, the individual animals from each time point clustered together, indicating that gene expression profiles of the animals within each time point were more similar to one another than those from the other time point.
To further investigate the utility of differential gene expression data to distinguish between the two time points, we utilized principal component analysis to determine which of the expressed genes were most discriminatory between the treatment groups (Fig. 4). Principal component no. 1, which accounted for 50% of the variance of the gene expression data, clearly separated the rats based on the time after LPS treatment, whereas principal component no. 2, which accounted for only 17% of the variance, separated the individual animals at each time point. The top 25 genes that contributed to principal component no. 1 are shown in Table 3 and will be elaborated on in the discussion.
To gain insight into the biological changes in whole blood caused by LPS treatment, we focused on selected known genes for further analysis of gene expression (Table 4). The chosen genes have either been associated with LPS-induced biological phenomenon or grouped into other functional categories. All differentially regulated genes and ESTs can be accessed in Supplemental Table S1 (available at the Physiological Genomics web site).1 As expected, LPS treatment caused differential expression of numerous genes in cells from whole blood that have been found to be associated with immune function, oxidative stress, and inflammation. Many of these genes were differentially expressed at both time points, although to different magnitudes, but always in the same direction (more or less abundant), whereas others were of different abundance at only one time point. Additionally, we found differential expression in several other functional categories, including genes involved in protein synthesis (ribosomal proteins, eukaryotic elongation factors), which were strikingly less abundant and accounted for the highest percentage of differentially expressed genes in the experiment. Other genes differentially expressed included those involved in the cytoskeleton, cell adhesion, cell cycle, cell signaling, transcription factors, anion transport, phosphatases, and oxidative respiration processes.
Quantitative real-time RT-PCR validation of differentially expressed genes.
Real-time RT-PCR analyses for a representative subset of individual transcripts were consistent with microarray results (Fig. 5). These genes were chosen based on their relevance to inflammation and/or WBC population changes. Five genes (IL-1β, Calgranulin B, ribosomal protein S6, elongation factor-1α, Scya5) demonstrated the same direction and general magnitude of differential expression seen in the microarray data. The same was true for TIMP2 at 2 h. However, PCR indicated a robust decrease at 6 h, whereas the decrease according to the microarrary was small and insignificant. We also examined TNF-α gene expression by real-time RT-PCR, even though the clone was not contained on our in-house cDNA chip. As expected, TNF-α expression was greater in blood from LPS-treated rats compared with controls at both time points: 2.5-fold at 2 h and 8-fold at 6 h.
With the advent of microarray technology, researchers have been able to measure genome-wide changes in gene expression in a single experiment. To date, few studies have used microarrays to measure the changes in gene expression that can be obtained from cells in whole blood. Profiling gene changes in the cells obtained from whole blood is beneficial for several reasons. First, because blood is a highly accessible tissue, it can be used to obtain gene signatures that distinguish between disease states in clinical samples without the need of obtaining tissue biopsy samples. Gene expression profiling of peripheral leukocytes or monocytes has been used to differentiate humans exposed to tobacco smoke from those not exposed (26), humans with treated or untreated hypertension (8), humans on the basis of IgA nephropathy activity (35), and human kidney transplant recipients from controls (51). Most relevant to this study, a few microarray studies have explored the effects of LPS on cultured human leukocytes to identify genes involved in acute inflammation (15, 16, 28). Although isolated white cells in culture can serve as a useful model, as with cell culture from other tissues, the changes in gene expression may be vastly different than what is observed in vivo. Gene expression profiling of whole blood thus may better elucidate the true biology of inflammation and other diseases.
In this study, we report for the first time large-scale gene changes in blood cells from rats treated with LPS. We collected rat blood into PAXgene tubes, which stabilized the RNA immediately on blood collection, providing us with RNA of high quality and sufficient quantity to perform microarray experiments. In contrast to other studies, no additional ex vivo procedures were required to isolate specific cell types, such as using Ficoll gradient separation to isolate peripheral blood leukocytes. Thus we believe that we observed a truer “snapshot” of gene expression changes in our study (4). Because the RNA was stabilized immediately on blood collection, this method should minimize the changes in expression of those genes that are sensitive to the length of time of ex vivo isolation and incubation (9). In addition, it should be noted that this method for isolating RNA from blood cells has the advantage of minimizing potential RNA degradation and losses that accompany leukocyte isolation protocols (34). Because we isolated a sufficient quantity of RNA (roughly 150 μg RNA/rat), we were able to perform hybridizations on samples from each individual LPS-treated rat.
Although use of the PAXgene tubes to isolate RNA from whole blood proved to be simple and produced high-quality RNA, their use does involve additional considerations. Because no further isolations are performed on the blood, each of the cell populations in the blood contribute to the RNA isolated. Therefore, a complete blood count is necessary for each sample to account for hematological changes that can influence relative RNA abundance and subsequent data interpretation. We did, in fact, see the expected changes in leukocyte populations that are characteristic of LPS-induced inflammation (22). As discussed below, we present evidence that we believe indicates an ability to distinguish between many of the gene abundance changes that are due to changes in cell populations and those due to the inflammatory response. A further consideration is that, although mature RBCs are anucleated and largely transcriptionally quiescent, containing only insignificant amounts of remnant RNA, a large amount of globin mRNA is present in reticulocytes, the immature RBCs (10). Globin transcripts can thus account for up to 70% of total RNA isolated from whole blood (1). There is a possibility that these abundant, largely uninformative transcripts could significantly decrease the chances of seeing lower abundant, more informative leukocyte messages. However, observations from our laboratory using rat blood and cDNA chips and the data from this study indicate that this is not the case. Therefore, we are still able to detect robust changes in leukocyte gene expression despite the abundant globin background.
The feasibility of whole blood gene expression profiling in preclinical toxicogenomic studies was tested in rats, using an LPS-induced inflammation model. In our study, acute systemic inflammation was evidenced by characteristic hematological changes and elevations in serum cytokine protein levels that are consistent with a previous study of the temporal profiles of these proteins after LPS exposure in the rat (24). TNF-α concentration was elevated at 2 h after LPS administration but had subsided by 6 h. Also, IL-6 levels were similarly high at 2 h but remained elevated at 6 h. Only a small increase in IL-1β protein serum concentration was observed in this study, in which the peak level may have occurred earlier than 2 h. Given the dramatic rise in serum TNF-α and IL-6 and clinical symptomatology, we are confident that we induced an acute inflammatory state.
In support of the hypothesis that we are indeed seeing gene changes due to LPS treatment and not just to changes in composition of the leukocytes, we observed gene expression changes similar to those observed in neutrophils isolated from human blood and cultured with LPS for 4 h (16, 28). It should be noted that these results are from different species, cell and RNA isolation methods, and microarray platforms. Not withstanding, as in those studies, we saw an increased relative transcript abundance for IL-1β, annexin1, guanylate binding protein-2, interferon-inducible transmembrane protein-3, 14-3-3 eta, and interferon-inducible protein variant-10 at 2 and/or 6 h after LPS treatment. Likewise, as in these cultured human neutrophil studies, we found decreased transcript abundance for cathepsin E, S100 calcium-binding protein A4, and S100-related protein, clone 42c (Table 4). We are unaware of any similar gene profiling studies of blood from rodent species.
Beyond these parallels in inflammatory gene expression, we also observed additional changes in gene expression that differed markedly from these previous studies. For example, we observed both increased and decreased transcript abundance for numerous cytoskeletal protein genes, suggesting a selective cell structural response to LPS. One of these, syntenin, has been identified as being induced by TNF-α in human umbilical cord arterial endothelial cells (43). In the human cultured neutrophil studies, only repression of a very few structural genes was seen (29). Although some differences in results may relate to variation in species and microarray platform, we propose that they also suggest that RNA immediately isolated from whole blood is more representative of in vivo conditions compared with treatments of in vitro cultured leukocytes.
In addition to identifying gene changes that have previously been associated with acute inflammation in the study of cultured human leukocytes, we also observed novel gene changes in the present study of whole blood from the rat. Transcripts from neutrophil-associated genes lipocalin-2 and Calgranulin B were more abundant at both 2 and 6 h after LPS exposure, suggesting both the proliferation and activation of neutrophils (33). In fact, the transcript for lipocalin-2 showed the largest fold change at either time point. Lipocalin-2 is the rat homolog of human neutrophil gelatinase-associated lipocalin, which has been proposed as a specific marker for neutrophils in bacterial infection (50). Similarly, the mouse homolog, 24p3/uterocalin, has been proposed as a marker for the acute-phase response (36). Here, we report for the first time the association of rat lipocalin-2 with inflammation, demonstrating the utility of microarray approaches as a tool for identification of possible biomarkers for further study. Other novel changes included an increase in the relative transcript abundance for well-recognized oxidative stress-related genes. At 2 h, these include transferrin, aminolevulinic acid synthetase, TIMP2, and selenium-binding protein. Other notable stress-related genes with increased relative transcript abundance in the blood at both 2 and 6 h after LPS treatment include two cytochrome P-450s, 2F1, and leukotriene B4-ω hydroxylase. This is in contrast to what is observed in the liver, where cytochrome P-450s are downregulated during acute inflammation (32). Finally, we observed gene expression changes in blood in a number of other genes that have been associated with acute inflammation in other tissues. The transcript level of the glycolytic enzyme aldolase C was elevated after LPS treatment, which is consistent with increased glycolysis during sepsis (49). Also at 6 h, the transcript for neuropeptide Y, a hormonal signal involved in the cross talk between the immune and nervous systems (37), was less abundant. Two transcription factors that have been identified as playing a role in TNF-α-mediated gene expression, Nf-E2 and fos-like antigen (46), showed increased transcript abundance at 2 h. Other functional categories with differentially expressed genes included members of the anion transport, phosphatase, and cell signaling (PKC) families. Although all of these genes have been found to be expressed under some circumstance of inflammation or oxidative stress, we report these as novel changes in differential gene expression from whole blood during acute inflammation.
In contrast to those changes in gene expression we believe to be due to LPS activation of leukocytes, we also observed other altered transcript levels that may simply reflect the notable decrease in lymphocytes, monocytes, and platelets and the accompanying rise in neutrophils associated with acute inflammation. Although no such studies exist for rodents, we compared our gene expression changes with lists of genes found to be associated with various leukocyte cell types in human whole blood gene expression studies (9, 48). Genes found to be preferentially associated with lymphocytes in those studies the abundance of which was decreased in our own consisted of a number of antigen recognition messages (especially the major histocompatibility complex II RNAs) and transcripts for ribosomal proteins and elongation factors. Other genes not found in the human studies that we suspect also fit this category include the lymphocyte-associated CD37 and a T-cell receptor (C-region β-chain). The decreased expression of cyclophilin A at 2 h after LPS exposure is perhaps related to the decrease in monocyte count, since it is actually a proinflammatory secretory product of LPS-stimulated macrophages (41). Similarly, the genes found to be associated with unstimulated neutrophils in the human studies that we found to be upregulated in our rat study were arachidonate 12-lipoxygenase, carboxypeptidase Z, collagen (type 1, α1), exportin-1, IL-1 receptor (type II), IL-6 receptor, and solute carrier family 2 (member 4). Very little has been done to characterize the transcriptional activity of platelets, although one recent publication attempted to characterize the transcript profile in humans (18). Two rat homolog transcripts that were decreased in our study, cofilin-1 and prothymosin-α, were among their list of the top 50 expressed transcripts, suggesting these changes may be due simply to the decreased platelet count. Taken together, the observed changes in transcript levels that reflect both the intrinsic biological activity and shifts in cell populations that contribute to the RNA pool provide strong evidence that knowledge of hematological parameters is required for proper interpretation of microarray analysis of whole blood.
Selected changes in gene expression identified by the cDNA arrays were validated by quantitative real-time RT-PCR. Five of the six genes selected for real-time RT-PCR confirmation (IL-1β, MRP14, ribosomal protein S6, EF1α, and Scya5) showed good agreement with microarray results. The increased abundance at 2 h of TIMP2 was confirmed, whereas a small nonsignificant decrease seen for this gene at 6 h, using the microarray, appeared as a robust decrease in abundance with real-time RT-PCR. In sum, however, all directional changes in gene expression observed by microarray analysis were confirmed by real-time RT-PCR.
We also found it interesting that whereas serum TNF-α protein concentration approached control levels 6 h after LPS exposure, RNA expression levels remained elevated 6 h after LPS exposure, probably reflecting the complex regulation of this important acute-phase cytokine. TNF-α synthesis is induced at both a transcriptional and a translational level in LPS-stimulated macrophages, but control of TNF-α RNA stability and translation is not coupled in response to LPS (30). Furthermore, serum TNF-α protein levels reflect the contribution of multiple cellular and organ sources compared with mRNA levels in WBCs alone.
Aside from gaining insights into the molecular mechanisms involved in acute inflammation, we also wanted to determine whether gene expression profiling could separate the samples based on time after LPS exposure. By employing two different approaches to sort animals based on gene expression, we were able to discriminate between the rapidly changing inflammatory events produced after LPS treatment (i.e., high TNF-α at 2 h and normal levels at 6 h). Importantly, both hierarchical clustering and principal component analysis (PCA) could distinguish between interindividual variability and treatment effects of LPS time course data (Figs. 3 and 4). As an unsupervised data reduction technique, hierarchical clustering successfully grouped animals based on time after LPS treatment. PCA additionally allowed for identification of major sources of variability in the gene expression data. The difference in treatment time accounted for the majority of the variability in the gene expression data, as seen in principal component no. 1, whereas a much smaller proportion of the variability was due to the expression differences between the individual animals, as seen in principal component no. 2. Among those genes that contributed to the separation in principal component no. 1 are genes known to be involved in the inflammatory response. In addition, there are genes in other functional categories and a number of ESTs that merit further investigation to determine their role in the inflammation process (Table 3). The ability of gene expression profiling of whole rat blood to separate the samples based on time after LPS exposure strengthens our supposition that whole blood transcriptional profiling can be used to discriminate biologically disparate samples despite the complication of shifting WBC populations.
In summary, we have demonstrated the feasibility of using whole rodent blood to obtain useful gene expression data during inflammation, an approach that we believe is preferable to previously reported methods. This study also demonstrates the complexity of interpreting gene expression data obtained from such a highly dynamic tissue as the blood. Infection, toxicity, or disease may alter the concentrations and relative proportions of blood cells so that concurrent hematological measurements are essential for biologically meaningful interpretation of gene profiling data in experimental animal studies (14). Similarly, the conduct of clinical blood profiling studies that will likely encounter significant interindividual human variation would be greatly facilitated by knowledge of hematological parameters and improved bioinformatics techniques to deconvolute experimental effects from hematological response (27, 48).
We acknowledge the very helpful editorial comments of Drs. Dori Germolic and Steven Kleeberger of NIEHS.
↵1 The Supplemental Material for this article (Supplemental Table S1) is available online at http://physiolgenomics.physiology.org/cgi/content/full/00190.2004/DC1.
Article published online before print. See web site for date of publication (http://physiolgenomics.physiology.org).
Address for reprint requests and other correspondence: R. Fannin, MD D2-04, National Institute of Environmental Health Sciences, PO Box 12233, Research Triangle Park, NC 27709 (E-mail:).
- Copyright © 2005 the American Physiological Society