If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
* This work was supported, in whole or in part, by National Institutes of Health Grant R01GM061761 from NIGMS and Grant R01CA088986 from NCI. [S] The on-line version of this article (available at http://www.jbc.org) contains supplemental Figures S1–S3 and supplemental Tables S1–S5.
Recent studies have indicated that lipopolysaccharides (LPS) isolated from particular bacterial strains can bias innate immune responses toward different signal transduction pathways thereby eliciting unique patterns of cytokines. Heterogeneity in the structure of lipid A (the active component of LPS) and possible contaminations with other inflammatory components have made it difficult to confirm these observations and dissect molecular motifs that may be responsible for modulatory properties. To address these issues, we have examined, for the first time, the ability of a range of well defined synthetic lipid As and isolated LPS and lipid A preparations to induce the production of a wide range of cytokines in three different mouse cell types. It was found that, for a given compound, the potencies of production of the various cytokines differed significantly. An additive model, in which a chemical change in the structure of a compound effects the potencies of all cytokines in the same manner, could describe the potencies of the cytokines for all compounds. Thus, no evidence was found that the structure of lipid A can modulate the pattern of cytokine production. In addition, the statistical analysis showed that the relative ordering of the potencies of the compounds was identical in the different cell types and that structural features such as the presence of a 3-deoxy-d-manno-octulosonic acid moiety, anomeric phosphate, lipid length, and acylation pattern were important for pro-inflammatory activity. Finally, it was found that transcriptional and post-transcription control mechanisms determine potencies and efficacies of cytokine production in cell-specific manners.
). It responds rapidly to unique molecules that are integral parts of pathogens and are perceived as danger signals by the host. Recognition of these molecular patterns is mediated by sets of highly conserved receptors (
), whose activation results in acute inflammatory responses. These responses include the production of a diverse set of cytokines and chemokines, direct local attack against the invading pathogen, and initiation of responses that activate and regulate the adaptive immune system (
Evidence is emerging that innate immune responses can be exploited for therapeutic purposes such as the development of adjuvants for vaccines and the treatment of a wide range of diseases, including asthma, infections, and cancer (
regulated on activation normal T cell expressed and secreted
monophosphorylated lipid A
analysis of variance
interferon-inducible protein 10.
are structural components of the outer surface membrane of Gram-negative bacteria that trigger innate immune responses through Toll-like receptor (TLR) 4/MD2. LPS consists of a hydrophobic domain known as lipid A, a non-repeating core oligosaccharide, and a distal polysaccharide (
Although LPS-induced cellular activation through TLR4/MD2 is complex as many signaling elements are involved, it appears that there are two distinct initiation points in the signaling process, depending upon the recruitment of the adaptor proteins myeloid differentiation primary response gene 88 (MyD88) or Toll-interleukin-1 receptor domain-containing adapter inducing IFN-β (TRIF). The MyD88-dependent pathway leads to early activation of the transcription factor NF-κB, the production of pro-inflammatory cytokines such as TNF-α, and Th1 cell responses. The TRIF-dependent pathway induces phosphorylation and dimerization of the transcription factor interferon regulatory factor-3, resulting in IFN-β production, which in turn activates the signal transducer and activation of transcription 1 pathway leading to the production of mediators such as nitric oxide (
An interplay between the MyD88 and TRIF signaling pathways has been demonstrated by microarray studies that examined LPS-induced changes in gene expression profiles of macrophages isolated from wild-type, MYD88−/−, TRIF−/−, and MYD88−/−TRIF−/− mice (
). Results from those studies demonstrated that LPS-inducible genes are completely abrogated in the MYD88−/−TRIF−/− mice; these genes can be classified into clusters based on their dependence on MyD88 and/or TRIF for expression. The cluster of genes regulated primarily by MyD88 included genes commonly associated with pro-inflammatory responses (e.g. TNF-α, IL-1β, and IL-6). The cluster that was dependent on both MyD88 and TRIF included IL-1α and IL-12A, and the cluster that was dependent primarily on TRIF included the IFN-inducible genes IFN-β, interferon-inducible protein 10 (IP-10), and RANTES.
Of the naturally occurring lipid A structures examined thus far, the vast majority contains a lipid backbone consisting of a β(1–6)-linked d-glucosamine or 2,3-diamino-2,3-di-deoxy-glucosamine in either homo- or hetero-dimeric combinations. However, striking structural differences in lipid As of different bacteria occur in the degree of phosphorylation and patterns of acylation. These structural differences account for the highly variable in vivo and in vitro responses to LPS (
) suggest that relationships between lipid A structure and biological responses are more complex, and cell culture, in vivo and ex vivo studies have indicated that LPS or lipooligosaccharide (LOS) isolated from particular bacterial strains can preferentially induce MyD88- or TRIF-dependent cytokines.
Heterogeneity in the structure of lipid A within a particular bacterial strain and possible contamination with other inflammatory components of the bacterial cell wall have complicated the use of either LPS or lipid A isolated from bacteria to dissect the molecular mechanisms responsible for the biological responses to specific lipid A molecules. To address this important issue, we have examined a range of well defined synthetic lipid As (
) derived from Escherichia coli, Salmonella typhimurium, S. minnesota, and Neisseria meningitidis LPS for the induction of mouse macrophages and dendritic cells, statistically analyzed the data to establish relative potencies, and uncovered a possible bias toward the production of MyD88- or TRIF-dependent cytokines. The particular synthetic compounds were selected because previous studies had implicated their LPSs in biasing innate immune responses toward MyD88- or TRIF-dependent pathways (
). Furthermore, synthetic chemistry makes it possible to obtain well defined lipid As that are free of other inflammatory components with complete control over fatty acid acylation and phosphorylation patterns. The synthetic compounds used in this study have structural modifications in the pattern and lengths of the lipid chains, presence of phosphate at the C-1 position, and presence of a dimeric 3-deoxy-d-manno-oct-2-ulosonic acid (KDO) moiety at the C-6′ position (see Fig. 1). LPS from E. coli, LOS from N. meningitidis, and isolated lipid A from S. minnesota R595 (MPLA) have also been examined.
E. coli 055:B5 LPS and E. coli 0111:B4 LPS (ultrapure) were obtained from List Biological Laboratories, meningococcal LOS was kindly provided by Dr. R. Carlson (Complex Carbohydrate Research Center, Athens, GA), and lipid A from S. minnesota R595 (MPLA) was obtained from Avanti Polar Lipids. The synthesis of synthetic lipid A derivatives 1–8 has been reported elsewhere (
). Compounds 1–5 and 7 were reconstituted in PBS with DMSO (10%) and compounds 6 and 8 in PBS with THF (10%). The monoclonal antibody to human/mouse IL-1β (3ZD) was from NCI-Frederick Cancer Research and Development Center, NIH, the monoclonal antibody to β-actin conjugated to HRP was from Abcam Inc., and the Mouse Innate and Adaptive Immune Responses RT2 Profiler PCR Array was from SABiosciences. The primer pairs for mRNA quantification were designed using the Universal ProbeLibrary (Roche Applied Science) and were obtained from Integrated DNA Technologies (supplemental Table S1).
RAW 264.7 γNO(−) cells, derived from the RAW 264.7 mouse monocyte/macrophage cell line, were obtained from ATCC and maintained in RPMI 1640 medium (ATCC) with l-glutamine (2 mm), adjusted to contain sodium bicarbonate (1.5 g/liter), glucose (4.5 g/liter), HEPES (10 mm), and sodium pyruvate (1 mm), and supplemented with penicillin (100 units/ml)/streptomycin (100 μg/ml, Mediatech) and FBS (10%, HyClone). BAC1.2F5 macrophages were kindly provided by Dr. E. R. Stanley (Albert Einstein College of Medicine, NY) and maintained in 75% MEM Alpha 1× (Cellgro) with Earle's salts and supplemented with penicillin (100 units/ml)/streptomycin (100 μg/ml) and FBS (10%) and 25% Eagle's minimum essential medium (ATCC) supplemented with penicillin (100 units/ml)/streptomycin (100 μg/ml) and FBS (10%) conditioned medium (as a source of growth factor CSF-1). Dendritic cells (DCs) were prepared from mouse bone marrow cultures as previously described (
) with some modifications. Briefly, freshly isolated bone marrow cells from C57BL/6J mice were grown for 7 days in RPMI 1640 medium (BioWhittaker/Cambrex) with l-glutamine (2 mm), adjusted to contain sodium pyruvate (1 mm), non-essential amino acids (1×), 2-mercaptoethanol (50 μm), and supplemented with penicillin (25 units/ml)/streptomycin (25 μg/ml), gentamicin (12.5 μg/ml), and amphotericin B (125 ng/ml). Furthermore, FCS (10%), rhFlt3L (25 ng/ml), rmIL-6 (25 ng/ml), and stem cell factor (25 ng/ml) were added to the medium. Next, the proliferated bone marrow cells were grown for 1 day in the same medium, but with the following additions: mouse serum (1%), recombinant granulocyte-macrophage colony-stimulating factor (10 ng/ml), rmIL-4 (10 ng/ml) to obtain mature DCs. All cells were maintained in a humid 5% CO2 atmosphere at 37 °C.
Cytokine Induction and Measurement
On the day of the exposure assay RAW 264.7 γNO(−) and BAC1.2F5 cells were plated as 2 × 105 cells/well in 180 μl and 3.4 × 104 cells/well in 180 μl, respectively, in 96-well tissue culture plates (Nunc). Cells were then incubated with different stimuli (20 μl, 10×) for 5.5 h and 24 h in replicates of six to give a final volume of 200 μl/well. Culture supernatants were then collected; two replicates of each sample were pooled to give final replicates of three and stored frozen (−80 °C) until assayed for cytokine production. After removal of the supernatant, cells were lysed by adding PBS containing Tween 20 (0.01%) and BSA (1%) in the same volume as that of the supernatant and sonicating for 5 min. In a similar way as that of the supernatants, the cell lysates of two replicates were pooled to give three final replicates and stored frozen (−80 °C) until assayed for cytokine production. For estimation of IL-1β secretion in BAC1.2F5 macrophages, supernatants were removed after 24-h incubation with compounds and replaced with the same volume of medium containing ATP (5 mm, Sigma). After re-incubation for 30 min supernatants were harvested. Similarly, on the day of the exposure assay mature DCs were plated as 4 × 106 cells/well in 1800 μl in 24-well tissue culture plates. Cells were then incubated with different stimuli (200 μl, 10×) for 5.5 h and 24 h in a final volume of 2 ml/well. Culture supernatants and cell lysates were then collected and stored frozen (−80 °C).
All cytokine ELISAs were performed in 96-well MaxiSorp plates (Nalge Nunc International). Cytokine DuoSet ELISA Development Kits (R&D Systems) were used for the cytokine quantification of mouse TNF-α, IP-10, RANTES, IL-1β, IL-6, and IL-10 according to the manufacturer's instructions. The absorbance was measured at 450 nm with wavelength correction set to 540 nm using a microplate reader (BMG Labtech). Concentrations of mouse IFN-β in culture supernatants were determined as follows. Plates were coated with rabbit polyclonal antibody against mouse IFN-β (PBL Biomedical Laboratories). IFN-β in standards (PBL Biomedical Laboratories) and samples was allowed to bind to the immobilized antibody. Rat anti-mouse IFN-β antibody (USBiological) was then added. Next, HRP-conjugated goat anti-rat IgG (H+L) antibody (Pierce) and a chromogenic substrate for HRP 3,3′,5,5′-tetramethylbenzidine (Pierce) were added. After the reaction was stopped, the absorbance was measured at 450 nm with wavelength correction set to 540 nm. Cytokine values are expressed as picograms of cytokine/ml or as picograms of cytokine/μg of total protein. The total protein content in the cell lysates was assessed using the bicinchoninic acid assay (Pierce Biotechnology). Cytokine values are presented as the means ± S.D. of triplicate measurements, with each experiment being repeated three times.
Preparation of RNA and Quantification of mRNA by Real-time PCR Analysis
RAW 264.7 γNO(−) and BAC1.2F5 cells were plated as 2.6 × 106 cells/well in 1800 μl in 6-well tissue culture plates (Nunc) and incubated for 3–4 h. Cells were then incubated with different stimuli (200 μl, 10×) for 90 min, after which cells were harvested and total RNA was isolated using Absolutely RNA Miniprep kit (Strategene) according to the manufacturer's instructions.
Isolated RNA (625 ng/50 μl) was subjected to cDNA synthesis using the TaqMan Reverse Transcription reagents (Applied Biosystems). The expression of selected genes in each cDNA sample was determined by quantitative real-time PCR using SYBR Green technology (Bio-Rad). Briefly, reaction mixtures (5-μl total volume) containing iQTMSYBR Green 2X Supermix (2.5 μl, Bio-Rad), forward and reverse primers (400 nm each, IDT), and cDNA (5 ng) were overlaid by Vapor-Lock (8 μl, Corbett Research), and the PCR reaction was carried out at 95 °C for 3 min and 40 cycles of 95 °C for 15 s and 60 °C for 1 min using a CFX96 Real-Time PCR Detection System (Bio-Rad). Melting curves were performed each time from 55 °C to 95 °C with increments of 0.5 °C for 5 s to ensure that only one amplification product was formed. For the quantitation of gene expression, 18 S rRNA, Gusb, Actin-β, and Hsp90ab1 were used as endogenous controls. Relative quantification with the obtained data was performed using the comparative (ΔΔCt) method, whereby the amount of target, normalized to an endogenous reference, and relative to a calibrator is given by the arithmetic formula: 2−ΔΔCt. cDNA samples were assayed in triplicates with each experiment repeated three times.
For the Mouse Innate and Adaptive Immune Responses RT2 Profiler PCR Array, cDNA was synthesized using the RT2 First Strand kit (SABiosciences) according to the manufacturer's instructions. Briefly, isolated RNA (1 μg) was subjected to genomic DNA elimination and cDNA synthesis using the RT mixture (10 μl) and incubations at 42 °C for 15 min and 95 °C for 5 min. Next, cDNA was added to a reaction mixture (final volume of 2550 μl) containing SABiosciences RT2 qPCR mastermix 2× (1275 μl). The reaction mixture was transferred to the PCR array (25 μl/well) and overlaid with Vapor-Lock (8 μl), and the PCR reaction was carried out as described above.
Statistical Data Analysis
Concentration-response data were analyzed using non-linear least-squares curve fitting in Prism (GraphPad Software, Inc). These data were fit with the following three-parameter logistic equation: Y = Emax/(1 + (EC50/X)), where Y is the cytokine response, X is the logarithm of the concentration of the stimulus, Emax is the maximum response, and EC50 is the concentration of the stimulus producing 50% stimulation.
The log(EC50) and S.E. values of log(EC50) calculated by Prism were used for further statistical analysis via an additive model: log[EC50(i,j)] = μ + α(i) + β(j) + e(i,j), where μ represents the grand mean, α(i) are deviations about μ due to effects of the compounds, β(j) are deviations about μ due to variations between the cytokines, and e(i,j) are random errors. The validity of the additive model was examined by computing Z-scores; Z(i,j) = (Obs(i,j) − Exp(i,j))/S.E.(i,j), for all pairs of compounds(i) and cytokines(j), where Obs(i,j) are the values observed in the experiment, Exp(i,j) are the expected values under the additive model, and S.E.(i,j) are the standard errors as determined by Prism. The data were analyzed by modified weighted ANOVA to calculate the least square means estimates for compounds and cytokines. The cytokines and compounds were separated into groups by significance calculated using a Fisher's protected least significant difference (LSD) procedure at α = 0.05.
For comparisons between two groups, the data were analyzed using the two-tailed Student t test with 95% confidence interval. A p value <0.05 was regarded as statistically significant.
Synthetic Lipid As Display Large Differences in Potencies of Cytokine Responses
We have employed the synthetic lipid A derivatives 1–8 (Fig. 1) corresponding to different bacterial species to study the structure-activity relationship for TLR4 activation. It was anticipated that analysis of potencies and efficacies of the mediators would establish whether structural differences in lipid A show a bias toward either a MyD88- or TRIF-dependent pathway. Lipid A derivatives 1 and 2, which are hexa-acylated in an asymmetrical manner and contain two phosphate residues, are derived from E. coli (
). Compounds 3 and 4 have phosphates at the C-1 and C-4′ positions, whereas 5 lacks a phosphate at the anomeric position. Furthermore, compounds 4 and 5 have several shorter lipids than lipid A 3. Lipid A 6 is a component of monophosphoryl lipid A derived from S. minnesota and has acyloxyacyl chains at C-2, C2′, and C-3′ (
). Synthetic lipid A 8 has a similar structure as 7 but is further linked to a KDO moiety at the C-6′ position. Two KDO molecules connect the lipid A molecule to the oligosaccharide core and are highly conserved in LPS structures investigated so far. In addition to the synthetic compounds, the isolated preparations E. coli 055:B5 LPS, E. coli 0111:B4 LPS, N. meningitidis LOS, and MPLA were examined. The inclusion of these preparations provides an opportunity to examine whether their behavior deviates from the well defined synthetic lipid A.
) and the isolated materials were exposed over a wide range of concentrations to RAW 264.7 γNO(−) and BAC1.2F5 macrophages cells. These cell lines were deemed attractive, because RAW 264.7 γNO(−) cells (
) do not produce nitric oxide upon treatment with IFN-γ alone but require LPS for full activation, thus making their behavior similar to primary macrophages from commonly used strains (e.g. C3H/HeN). BAC1.2F5 cells (
) are a splenic macrophage line derived from BALB/c × A.CA mice, which require CSF-1 for survival and proliferation, and adopt a morphology that closely resembles that of primary bone marrow-derived macrophages. Importantly, the use of these macrophage cell lines made it possible to examine the series of compounds over a wide range of concentrations on a sufficiently large scale to measure the production of a panel of cytokines and chemokines corresponding to the two distinct TLR4 activation pathways.
Thus, supernatants harvested after 5.5 and 24 h were examined for the presence of TNF-α, which is a prototypical MyD88-dependent cytokine, and IFN-β, which is a prototypical TRIF-dependent cytokine. Lipid As 1, 2, 4, 7, and 8, E. coli 055:B5 and 0111:B4 LPS, N. meningitidis LOS, and MPLA yielded clear dose-response curves with large differences in potencies (EC50, concentration producing 50% activity) (Fig. 2 and supplemental Fig. S1A). S. typhimurium lipid A 3 gave only a partial response at the highest concentration tested, whereas monophosphorylated lipid A 5 and S. minnesota lipid A 6 were inactive. In each case, TNF-α was induced at a somewhat lower concentration of agonist than IFN-β. Furthermore, the supernatants harvested after 5.5 h showed significant quantities of IP-10 and RANTES and no or very low concentrations of extracellular IL-1β, IL-6, and IL-10. Increasing the incubation time to 24 h showed, however, clear dose responses for the latter cytokines. Surprisingly, for each compound the potencies for the various cytokines differed substantially, and for example in each case TNF-α was induced at a much lower concentration of agonist than IL-10 and IL-1β.
The same range of compounds, except compounds 3, 5, and 6, which yielded partial or no response in RAW 264.7 γNO(−) cells, was also exposed over a wide range of concentrations to BAC1.2F5 macrophages (supplemental Fig. S1B). It was found that these cells do not secrete IFN-β and IL-10 at 5.5 or 24 h. Moreover, a short incubation time of 5.5 h led to detection of TNF-α, IP-10, and RANTES, and after 24 h dose-response curves were obtained for extracellular IL-1β and IL-6.
To examine whether primary cells elicit similar responses, a limited number of compounds (E. coli 0111:B4 LPS, 7, and 8) was exposed over a wide concentration range to primary mouse DCs, which were derived from bone marrow cells (supplemental Fig. S1C). Dose-response curves were obtained for all cytokines except for the anti-inflammatory cytokine IL-10, which was not detected at a sufficient high level to establish EC50 values.
To investigate the possible presence of intracellular cytokines, cell lysates of the cells exposed for 5.5 h to the various inducers were examined for the presence of intracellular TNF-α, IL-1β, IL-6, and IL-10. It was found that the cell lysates did not contain any TNF-α, IL-6, and IL-10 (data not shown), while a significant quantity of IL-1β was present intracellularly (supplemental Fig. S1).
The BAC1.2F5 macrophages required activation by ATP to secrete IL-1β, and a clear dose response was obtained after stimulating the cells with lipid A or LPS for 24 h, followed by the addition of ATP (5 mm) and re-incubation for 30 min (supplemental Fig. S2). The RAW 264.7 γNO(−) cells did not require activation with ATP for secretion of IL-1β, and ATP treatment did not affect levels of intracellular IL-1β. Optimum exposure times for intracellular and extracellular IL-1β were found to be 5.5 and 24 h, respectively, for both BAC1.2F5 and RAW 264 γNO(−) cells. For the RAW 264 γNO(−) cells, the potency of intracellular IL-1β decreased with longer incubations. Interestingly, the primary DCs did not require activation with ATP for secretion of IL-1β, but upon treatment with ATP the cells secreted an additional amount of IL-1β (supplemental Fig. S1C).
IL-1β is expressed as a pro-protein (pro-IL-1β), which is cleaved by caspase-1 into its active form IL-1β, which is then secreted (
). Therefore, Western blot experiments were performed to establish the nature of intracellular and secreted IL-1β after stimulation of RAW 264.7 γNO(−) and BAC1.2F5 cells with E. coli 055:B5 LPS for 5.5 and 24 h. A 33-kDa band corresponding to the pro-IL-1β was observed in the cell lysates, and, as expected, no mature IL-1β (17 kDa) could be detected (data not shown). We were not able to detect (pro)-IL-1β in supernatants due to the presence of large quantities of cell culture medium proteins.
The Cytokine Responses Can Be Described by an Additive Model, Which Does Not Show a Bias toward MyD88- or TRIF-dependent Responses
Potencies (EC50, concentration producing 50% activity) and efficacies (maximum level) for cytokine production were determined by fitting the dose-response data to a logistic equation using Prism software. As can be seen in Table 1, the synthetic compounds and LPS preparations exhibited large differences in potency for a given cytokine. Furthermore, for each compound, large differences in potencies for the various cytokines were observed.
TABLE 1Cytokine log (EC50) values (nm) of E. coli LPS, N. meningitidis LOS, lipid A derivatives 1, 2, 4, 7, and 8, and MPLA in (A) RAW 264.7 γNO(−), (B) BAC1.2F5, and (C) dendritic cells
To establish the relative order of potencies of the compounds and uncover a possible bias toward the production of MyD88- or TRIF-dependent cytokines, the log (EC50) and their S.E. values were statistically analyzed by fitting the data to an additive model. The results were analyzed by ANOVA, and significant differences in activities of compounds and cytokines were determined by a protected LSD procedure. In an additive model, the relative effects of the potency of cytokines (log (EC50)) are independent of the nature of the compound and vice versa. Possible deviations from such a model, which would signal immune modulation by a particular compound, were examined by computing Z-scores.
Thus, the experimentally determined log (EC50) and S.E. values were analyzed by an additive model in which log[EC50(i,j)] = μ + α(i) + β(j) + e(i,j); where μ is the grand mean (the weighted average of all observed EC50 values), α(i) represents the deviations about the grand mean due to the effects of the compounds, β(j) represents deviations about the grand mean due to the effects of the cytokines, and e(i,j) represents random errors (assumed to be normally distributed with mean 0 and unknown variance). The data were analyzed by an unweighted ANOVA (observed least squares), weighted ANOVA (weighted least squares (WLS)), and modified weighted ANOVA (adjusted WLS). Although the order of activity for the compounds and cytokines was consistent for the three models, the modified weighted ANOVA yielded the highest R2 values for the two cell lines: RAW 264.7 γNO(−) cells (R2 = 0.9652) and BAC1.2F5 cells (R2 = 0.9216). Hence, this model was employed to calculate the least square means estimates for compounds and cytokines. The expected values under the adjusted WLS model in log scale for the cytokine-compound combinations are shown in supplemental Table S2.
The cytokines and compounds were separated into groups by significance calculated using a protected LSD procedure at α = 0.05. As can be seen in Table 2, the compounds have more statistically significant differences than the cytokines, with the nine compounds being separated into seven distinct groups for the RAW 264.7 γNO(−) cells and into five distinct groups for the BAC1.2F5 cells.
TABLE 2Ordering of compounds and cytokines by their LSM into groups for (A) RAW 264.7 γNO(−), (B) BAC1.2F5, and (C) dendritic cells
The cytokines were separated into four statistically significant different groups for the RAW 264.7 γNO(−) cells, three for the BAC1.2F5 cells, and four groups for the DCs. Importantly, the relative ordering of the potencies of the nine compounds was almost identical for the various cell types, whereas this was not the case for the cytokines, and most notably, IL-6 was of low potency in the BAC1.2F5 cells and highly responsive in the DCs.
Z-scores were computed via Z(i,j) = (Obs(i,j) − Exp(i,j))/S.E.(i,j), for all pairs of compounds (i) and cytokines (j), where, Obs(i,j) represents the values observed in the experiment, Exp(i,j) represents the expected values under the adjusted WLS model, and S.E.(i,j) represents the Prism S.E.s. More than 5% of the Z-scores calculated fell outside the range of −2.5 < Z < +2.5 (supplemental Table S3), which indicates that there are some interactions between the compounds and cytokines with respect to values of log (EC50).
To validate the additive model, the same procedure was applied for one compound (E. coli 055:B5 LPS) in seven different assays for cytokine responses in RAW 264.7 γNO(−) cells. The analyses yielded a good fit and the same relative order of cytokines (data not shown).
A similar statistical analysis was performed for the plateau values (efficacies) of the dose-response curves. However, even with modifications, the additive model did not provide meaningful information for these values. The latter result is probably due to the fact that the differences in efficacies among the different compounds for a given cytokine are relatively small. Nevertheless, examination of differences in efficacies for the cytokines in the two cell lines revealed interesting correlations. When corrected for total protein, the RAW cells showed much higher efficacies of IP-10, RANTES, INF-β, and IL-10 as compared with BAC cells, while only the efficacy of the secretion of IL-1β was higher in the BAC cells (Table 3).
TABLE 3Cytokine Plateau values (pg/μg) of dose-response curves of E. coli LPS, N. meningitidis LOS, lipid A derivatives 1, 2, 4, 7, and 8, and MPLA in (A) RAW 264.7 γNO(−), (B) BAC1.2F5, and (C) dendritic cells
Potencies of Transcripts of the Cytokines Follow an Additive Model
Significant differences in the potencies of the various cytokine proteins was observed when cells were activated by a particular compound. To establish whether these differences in responsiveness are due to transcriptional processes, potencies of mRNA for the various cytokines were determined after stimulation with several compounds at a range of different concentrations and the results compared with similar responses for cytokine protein. Thus, mouse macrophages (RAW 264.7 γNO(−) cells) were exposed for 90 min over a wide range of concentrations to E. coli 0111:B4 LPS and the synthetic lipid A derivatives 7 and 8. At this time point, maximum or near maximum transcription was observed (supplemental Fig. S3). Next, cell lysates were prepared, total RNA was isolated, and reverse transcription PCR was performed to obtain cDNA, which was used for SYBR Green quantitative real-time PCR to determine the levels of mRNA induced for various cytokines. Potencies of mRNA were determined by fitting the dose-response curves to a logistic equation using PRISM software (supplemental Table S4, A–C).
The potencies of cytokine transcription were ranked in statistically significant groups by employing an additive model and then analyzed by weighted ANOVA. Furthermore, Z-scores were calculated to validate the additive model (supplemental Table S5). All the Z-scores calculated fell well within the range of −2.5 < Z < +2.5, indicating the absence of deviations from the additive model and hence absence of immune modulation by compound structure. The cytokines could be separated into four statistically significant groups by using a protected LSD procedure (Table 4). Interestingly, the order of mRNA and protein potencies was almost identical. The only exception was IFN-β, and in this case, mRNA was in a lower activity group compared with protein. A two-tailed, two-sample t test with 95% confidence intervals was performed over the log (EC50) values of each cytokine for protein and mRNA. As expected, a statistically significant difference was only observed for IFN-β (supplemental Table S4, A–C). Thus, except for IFN-β, transcriptional regulatory processes appear to control the different responsiveness (potencies) of cytokine protein production.
TABLE 4Ordering of cytokines (mRNA) induced by E. coli 0111:B4 LPS, 7, and 8 in RAW 264.7 γNO(−) cells by their LSM into groups
The Two Cell Lines Express Different Inflammatory Genes
To study the origin of the differences in cytokine production between the two cell lines, the expression of a hundred different innate and adaptive immune response genes, including transcription factors, adaptor proteins, cytokines, MAPKs, and proteins involved in the inflammasome complex was examined in RAW 264.7 γNO(−) and BAC1.2F5 cells after stimulation with N. meningitidis lipid A 8 at a concentration at which all cytokine proteins had reached a plateau. For each gene, the amount of mRNA in BAC1.2F5 cells relative to the amount in RAW 264.7 γNO(−) cells was calculated (Table 5), and the gene expression of BAC1.2F5 versus RAW 264.7 γNO(−) cells is depicted in a scatter plot (Fig. 3).
TABLE 5Expression of genes involved in innate and adaptive immune responses after induction by 8 in RAW 264.7 γNO(−) vs. BAC1.2F5 cells
Endogenous control 18 S rRNA was used to calculate the ΔCt for each gene for both cell lines. The ΔΔCt was calculated by subtracting the RAW ΔCt values from the BAC ΔCt values, and -fold change was calculated as 2−ΔΔCt. The RAW vs. BAC ΔCt values of genes showing more than 3-fold differences from three independent experiments were analyzed by t-test.
a Endogenous control 18 S rRNA was used to calculate the ΔCt for each gene for both cell lines. The ΔΔCt was calculated by subtracting the RAW ΔCt values from the BAC ΔCt values, and -fold change was calculated as 2−ΔΔCt. The RAW vs. BAC ΔCt values of genes showing more than 3-fold differences from three independent experiments were analyzed by t-test.
It was observed that the induction of TRIF-dependent cytokine IFN-β was >5-fold higher in RAW 264.7 γNO(−) cells as compared with the BAC1.2F5 macrophage cells, whereas the expression of MyD88-dependent cytokines, such as TNF-α, IL-6, and IL-1β, was higher in BAC1.2F5 macrophages. Moreover, expression of MYD88 was similar in both cell types, whereas Ticam2 was >10-fold higher expressed in BAC1.2F5 macrophages. Induction of NFKB, the transcription factor of MyD88-dependent cytokines, was higher in BAC1.2F5 macrophages, whereas IRF3, the transcription factor of TRIF-dependent cytokines, was similar in both cell types. ERK1, ERK2, JNK, TAB1, and TAB2 were induced higher in BAC1.2F5 macrophages than in RAW 264.7 γNO(−) cells, whereas other MAPKs studied (e.g. p38, MEKK1) were similarly induced in both cell types. Moreover, the genes of proteins involved in the inflammasome complex such as ASC/PYCARD and Nalp-1 were highly induced in the BAC1.2F5 macrophages.
Structural studies have demonstrated that the carbohydrate backbone, degree of phosphorylation, and fatty acid acylation patterns vary considerably among bacterial species and that these differences account for high variability in vivo and in vitro host responses to LPS (
). On the other hand, E. coli 055:B5 and Vibrio cholerae LPS, at the same picomolar concentration of lipid A, selectively induced the MyD88-dependent pathway, whereas S. typhimurium LPS primarily invoked the MyD88-independent pathway. In another study, it was found that bone marrow-derived macrophages exposed over a narrow range of concentrations to monophosphorylated lipid A (MPLA) from S. minnesota, which is a promising human vaccine adjuvant candidate, induced lower levels of IL-6 and similar levels of IFN-β and IP-10 as compared with S. minnesota LPS (
). These observations led to the conclusion that MPLA is a TLR4 agonist for the TRIF-dependent pathway. Recent gene expression analysis of bone marrow-derived dendritic cells indicated that a mono-phosphorylated lipid A derivative induced more strongly transcripts of MCP1 and COX than a biphosphorylated lipid A derivative (
). This protease is involved in the maturation of several pro-inflammatory mediators such as IL-1β and IL-18, and the reduced production of these cytokines may explain the low toxicity of MPLA. Yet another explanation for the low toxicity adjuvant functions of MPLA is that it stimulates higher levels of the anti-inflammatory cytokine IL-10 (
The importance of selective activation of the MyD88-dependent and -independent pathways has been underscored by the results of in vivo and ex vivo studies. For example, knock-out mice, lacking the ability to synthesize TNF-α, are resistant to lethal amounts of LPS from E. coli 055:B5, but remain as susceptible to S. typhimurium LPS as wild-type mice (
). In this study, the lethal effects of S. typhimurium LPS were mainly due to production of IFN-γ, IL-1, and IL-18.
Heterogeneity in lipid A of particular bacterial strains and possible contaminations with other inflammatory components of the bacterial cell wall has made it difficult to dissect specific molecular motifs that may bias an immune response toward the MyD88 or TRIF pathway. To address these issues, we have examined, for the first time, the ability to induce the production of a wide range of cytokines in two well established mouse macrophage cell lines (RAW 264.7 γNO(−) and BAC1.2F5) and primary DCs obtained by differentiation from mouse bone marrow cells by a range of well defined synthetic homogenous lipid As derived from E. coli, S. typhimurium, S. minnesota, and N. meningitidis (
). In addition, similar responses induced by isolated and heterogeneous LPS from E. coli 055:B5 and 0111:B4 LPS, LOS from N. meningitidis, and MPLA derived from S. minnesota R595 (MPLA) were examined. It was found that the synthetic compounds and LPS preparations displayed large differences in the potencies (EC50) of cytokine protein production. A statistical analysis showed the same order of potencies for the various compounds in both cell lines, which probably reflects their affinity for the MD2·TLR4 complex and the ability to dimerize such a complex to initiate cell signaling events.
It was found that the LPS preparations from E. coli were more potent than the LOS from N. meningitidis and the KDO-containing compound 8, which in turn were significantly more potent that the prototypic synthetic lipid As derived from E. coli (1) and N. meningitidis (7). It has long been thought that the inflammatory properties of LPS and LOS reside in the lipid A moiety (
). The data presented here, using well defined synthetic compounds, confirm that KDO significantly contributes to the potency of cytokine responses induced by LPS. Interestingly, an x-ray co-crystal structure of LPS and MD2 and TLR4 showed at least one hydrogen bond of KDO with TLR4 (
), and this interaction may account for a higher affinity and hence a greater potency of KDO-containing compounds.
The statistical analyses showed significant differences between the potencies of the various lipid As. Importantly, the prototypic lipid A from N. meningitidis (7) was significantly more potent than E. coli lipid A (2). The lipid A of N. meningitidis is hexa-acylated in a symmetrical fashion (7), whereas enteric bacteria have an asymmetrically hexa-acylated lipid A (2). Also, a number of the fatty acids of N. meningitidis lipid A are shorter compared with those of E. coli. Probably, these structural differences account for differences in pathobiological properties of enteric and meningococcal LPS (
It was also found that S. typhimurium lipid A 3 gave only a partial response at the highest concentration, whereas a similar compound having shorter fatty acids (4) was almost three orders of a magnitude more potent. Also, E. coli derived lipid A 2, which has shorter lipids than the parent compound 1, was significantly more potent, although the difference was not as large as for S. typhimurium lipid As 3 and 4. These findings indicate that shortening fatty acids results in an increase of potency of lipid A. It has been observed that pathogens that cause chronic infections such as periodontitis and peritoneal infections (
), have fatty acids much longer than what is typically observed in enteric bacteria, and hence this type of substitution may diminish detection by the innate immune system.
A surprising observation was that for each compound the potencies of expression of the various cytokine proteins differed substantially. For the RAW 264.7 γNO(−) cells, four statistically different groups could be identified, three for the BAC1.2F5 cell, and four for the DCs. The grouping, which did not follow a simple pattern of MyD88- and TRIF-related cytokines, differed between the three cell types. For example, IL-6, which was previously assigned as an MyD88-dependent cytokine (
), was in the lowest responding group in BAC1.2F5 cells, whereas it had an intermediate reactivity in the RAW 264.7 γNO(−) cells, and was the most potent cytokine in the DCs. On the other hand, the MyD88-dependent cytokine TNF-α (5.5 h) was secreted at a relative low concentration of agonist in all cell types.
An additive model could describe the potencies of the various cytokines. In such a model, a chemical change in a compound affects the potencies of all cytokines in the same manner and, hence, indicates the absence of immune modulation by compound structure. Examination of Z-scores, which define deviations on the observed and expected potencies in the additive model, revealed the presence of some interactions between compound and cytokine response. The largest Z-scores, which signify substantial deviations from the additive model, were associated with isolated LPS and lipid A preparations and the secretion of IL-1β and IL-10. Although the isolated materials were purified, it cannot be excluded that they contain impurities that selectively modulate the secretion of IL-1β and IL-10. Thus, this study did not find evidence that the structure of lipid A can modulate potencies of pro-inflammatory cytokine production.
Previously observed biases of expression of particular sets of cytokines (
) may be due to the presence of pro-inflammatory contaminants or comparisons of compounds at incorrect concentrations. In respect to the latter, firm conclusions about immune modulation by LPS or lipid A derivatives can only be drawn when activities of several agonists are examined over a sufficiently large range of concentrations. Also, it may be possible that immune modulation arises from differences in efficacy of cytokine production. A statistical analysis of the maximum responses for the compounds examined in this study did, however, not provide meaningful correlations. Finally, it cannot be excluded that previously observed biases may be due to the use of different cell types (e.g. primary cells versus cell-lines, different cell-lines) or differences in employed reagents.
The potencies of mRNA expression of the various cytokine differed significantly when induced by a given compound and could be described by an additive model underscoring the absence of immune modulation of potency by compound structure. Furthermore, the responsiveness of the cytokine transcripts and proteins followed the same order with INF-β being the only exception. Thus, the observed differences in potencies of cytokine protein production must in large measures be due to transcriptional control mechanisms. It is well known that cytokine production is tightly controlled by such mechanisms (
). Further research is, however, required to elucidate processes that control the potencies of cytokine production in cell-specific manners.
It was found that the two cell lines exhibit significant differences in the expression of genes involved in pro-inflammatory responses. For example, the BAC1.2F5 cells express significantly higher levels of mRNA of MyD88-dependent TNF-α, IL-1β, and IL-6. These cells also express higher levels of Irak2 and several MAPKs, which may be responsible for increases in mRNA expression of the cytokines (
). However, the BAC1.2F5 and RAW cells exhibited similar plateau values (maximum responses) for TNF-α, IL-1β, and IL-6 protein, and thus there is a lack of correlation between transcript and protein production. The latter is underscored by the finding that both cell lines express similar amounts of mRNA of IP-10, IL-10, and RANTES but produce significantly different amount of the proteins. The poor correlation between the mRNA expression and efficacies (maximum responses) of protein production highlights that post-transcriptional control mechanisms (
) play an important role in cytokine production. Detailed studies are required to establish the molecular mechanism of such process. It is, however, clear that transcript analysis is an unreliable approach to estimate protein production.
The gene expression data showed that RAW 264.7 γNO(−) cells express very low levels of ASC/PYCARD and Nalp-1, which are important components of the inflammasome. Activation of the inflammasome by engagement of cytotoxic T cells or activation of purinergic P2X7 receptor by a high concentration of extracellular ATP (
). On the other hand, activation of BAC1.2F5 cells by a short treatment of ATP leads to activation of P2X7, resulting in the subsequent activation of ASC, caspase-1, and secretion of mature IL-1β. A recent study (
) showed that the IL-1β release was reduced when the expression of ASC or NALP3 was inhibited by specific siRNA in primed THP-1 cells implicating post-translational processes in cytokine responses. The difference in concentration requirement of LPS or lipid A for intracellular synthesis and secretion of IL-1β in both cell lines indicates that LPS plays a role in the P2X7-dependent and -independent secretion of IL-1β.
The studies reported here show, for the first time, that dosage of LPS or lipid A and the differentiation state of macrophages and dendritic cells is an important determinant of the pattern of cytokine secretion. It has been shown that activation of TLRs of hematopoietic stem cells influences the differentiation of these cells into macrophages or monocytes (
). It is likely that the structure of lipid A and the administered dose will influence the differentiation of relevant cells, which in turn will bias pro-inflammatory responses. Future in vivo studies employing a wide range of concentrations of well defined lipid A derivatives will be required to test this hypothesis. Furthermore, recent studies suggest that C-type lectins such as DC-SIGN cooperate with other innate immune receptors, such as Toll-like receptors, to fine-tune and modulate immune responses (
). The experimental approach reported here is ideally suited to investigate such mechanisms.
We thank Dr. E. R. Stanley (Albert Einstein College of Medicine, New York) for the BAC1.2F5 macrophages, the NCI-Frederick Cancer Research and Development Center for the anti IL-1β antibody, and Drs. S. Gendler and V. Lakshminarayanan (Mayo Clinic College of Medicine, Scottsdale) for assisting with cellular activation studies using primary mouse DCs.