Chloroquine Interferes with Lipopolysaccharide-Induced TNF-α Gene Expression by a Nonlysosomotropic Mechanism¹

Chloroquine (CQ)³ is an affordable generic drug with over 60 years of safe clinical use in the treatment of malaria and inflammatory disorders. However, despite a rich clinical history, the mechanism by which CQ mediates its anti-inflammatory effects has not been fully characterized. Its small lipophilic nature enables CQ to readily penetrate the lipid bilayer (1). Within the cell, this diprotic weak base is thought to diffuse across a pH gradient into acidic subcellular compartments such as endolysosomes (2), where it becomes diprotonated at lower pH. Diprotonated CQ concentrates within endolysosomes up to 10,000-fold higher than its extracellular concentration (1), resulting in therapeutically obtainable intracellular concentrations in the millimolar range (3).

A panoply of effects on cellular function have been attributed to CQ. This drug interferes with protein synthesis (4), processing (5), and degradation (6) by mechanisms thought to involve, at least in part, alkalization of endosomes and lysosomes. Other effects appear to be independent of the lysosomotropism of CQ. Thus, CQ can interact with DNA (7), alter its superhelical structure (4), and block DNA synthesis at high concentrations (8). CQ also interferes with generation of reactive oxygen species (9), inositol 1,4,5-triphosphate signaling (10), and protein phosphorylation (11).

CQ is known to inhibit TNF-α release from human and murine cells (5, 12, 13, 14, 15, 16, 17), although the mechanism by which this is accomplished is unclear. In a murine macrophage cell line (RAW 264.7), CQ (100 μM) interfered with posttranslational processing of TNF-α (5) by a mechanism postulated to involve alteration of cellular iron metabolism (13). However, in the same cell line, other investigators found that CQ (250 μM) primarily reduced TNF-α gene expression (12, 16), an effect possibly involving CQ cytotoxicity (5, 16). The aim of the present study was to elucidate the mechanism by which this enigmatic drug reduces LPS-stimulated TNF-α release from human PBMC. We found that CQ specifically reduced TNF-α release at the pretranslational level by reducing TNF-α gene expression without altering mRNA stability or signaling pathways upstream of NF-κB mobilization. Moreover, CQ appeared to mediate its effects by a nonlysosomotropic mechanism.

Reagents were obtained from Sigma (St. Louis, MO) unless stated otherwise. Experiments were designed to minimize endotoxin contamination. All plasticware was obtained prepackaged and endotoxin-free. RPMI 1640 and PBS were obtained from BioWhittaker (Walkersville, MD) and contained less than 0.005 U/ml endotoxin. RPMI 1640 was supplemented with l-glutamine, penicillin, and streptomycin. CQ was dissolved in PBS at a stock concentration of 100 mM, syringe-filtered (0.2 μM), aliquoted, and stored at −80° until use. Bafilomycin A₁ was solubilized in methanol at a stock concentration (100 μM), aliquoted, and stored at −80° until use. LPS from Escherichia coli O111:B4 was prepared as a stock solution (100 μg/ml) in PBS, stored in aliquots at −80°C, and used at a final concentration of 100 ng/ml. Pooled human serum (PHS) was obtained by combining sera of 10–15 healthy donors under conditions designed to preserve complement activity (18). CQ, bafilomycin A₁, LPS, and PHS were thawed immediately before use, and any unused portion was discarded. PBMC viability was measured 8 h after LPS stimulation of CQ (100 μM)-pretreated PBMC using the Live/Dead Viability/Cytotoxicity Kit (Molecular Probes, Eugene, OR). At this time point, LPS-induced TNF-α release has already peaked (19).

After informed consent, peripheral blood was obtained by venipuncture from normal healthy volunteers. Each donor was used no more than once per set of experiments. PBMC were purified using standard methods (19). Blood was anticoagulated with 5 U pyrogen-free heparin (Fujisawa, Deerfield, IL) per ml of blood and centrifuged at 500 × g for 15 min. Leukocyte-rich buffy coats were then subjected to Ficoll-Hypaque density gradient centrifugation before collection of PBMC from the light density fraction. Cells were washed three times with ice-cold PBS before counting by hemocytometer and resuspension in RPMI 1640.

PBMC (2 × 10⁵–1 × 10⁶) were preincubated in the absence or presence of inhibitor (CQ or bafilomycin A₁) for 2 h at 37° in 96-well polystyrene plates (Dynatech Laboratories, Chantilly, VA) containing 200 μl of RPMI 1640. Intracellular concentrations of CQ (3) and bafilomycin A₁ (20) reach equilibrium within this time period. Cells were then stimulated for 18 h with LPS (100 ng/ml) in the presence of 10% PHS. Preliminary experiments demonstrated that this concentration of LPS induced near-maximal stimulation of TNF-α release from PBMC. Inhibitors were not washed away before stimulation with LPS. Cell-free supernatant was collected and analyzed for human TNF-α, IL-1β, IL-6, and RANTES (21) by ELISA. Briefly, to assay TNF-α and RANTES, Ab pairs and recombinant cytokine (for use as standards) were obtained from R&D Systems (Minneapolis, MN), and the ELISA was performed according to the manufacturer’s instructions using HRP as the detection reagent. IL-1β and IL-6 concentrations were assayed using ELISA kits (Biosource International, Camarillo, CA) according to the manufacturer’s protocol.

PBMC (5 × 10⁶) were incubated in 6-well polystyrene plates in RPMI 1640 for 2 h at 37° in the absence or presence of CQ (10 or 100 μM). Cells were then stimulated with LPS for 2 h in the presence of 10% PHS. Total cellular RNA was extracted from PBMC using TRIZOL reagent (Life Technologies, Grand Island, NY) as previously described (19). Total RNA (5 μg) was separated on 1.2% agarose-formaldehyde gels, transferred to a nylon membrane (Immobilon-Ny+; Millipore, Bedford, MA), and analyzed by Northern blotting (22). Briefly, the 633-bp fragment of the EcoRI and HindIII double digest of plasmid encoding the full-length cDNA for human TNF-α (a gift from Leo Lina, Cetus, Emeryville, CA) was purified from an agarose gel slice using the JetSorb Gel Extraction Kit (Genomed, Research Triangle Park, NC). This fragment (25 ng) was ³²P-labeled by the random primed labeling method (Prime-a-Gene Labeling System; Promega, Madison, WI). Membranes were incubated in a commercial hybridization solution (ExpressHyb; Clontech, Palo Alto, CA) containing ∼2 × 10⁷ cpm of labeled probe per ml of solution. Hybridization was quantitated by phosphorimager analysis (Molecular Dynamics, Sunnyvale, CA). RNA integrity and equal loading were evaluated by ethidium bromide staining of agarose gels before transfer as well as by stripping and reprobing the membrane with a G3PDH cDNA probe (Clontech).

Nuclear translocation of the transcription factor NF-κB was assayed according to a standard protocol (22). Briefly, nuclear extracts from PBMC were prepared in the presence of protease inhibitors after 30 min LPS (100 ng/ml) stimulation of PBMC. Total protein content of the nuclear extracts was determined using a commercial kit (Bio-Rad Laboratories, Hercules, CA). An oligonucleotide containing the NF-κB consensus binding sequence was obtained (Promega) and end-labeled with [α-³²P]dATP and [α-³²P]dCTP using Klenow DNA polymerase (Promega). Unincorporated nucleotides were removed using a spin column (Microspin G-25; Amersham Pharmacia Biotech, Piscataway, NJ). Labeled probe (20,000 cpm) was incubated with nuclear extract (∼4 μg) for 30 min at room temperature in 1× band shift buffer (10 mM Tris-HCl (pH 7.5), 1 mM EDTA, 40 mM KCl) containing 100 μg/ml poly(dI-dC) and 10% glycerol. Samples were fractionated by electrophoresis over 4% native polyacrylamide gels, transferred to 3 MM filter paper (Whatman Laboratory Products, Clifton, NJ), dried, and analyzed by phosphorimager. A 250-fold excess of unlabeled probe was used as specific competitor. To determine the composition of NF-κB complexes by supershift, Ab specific for the p50 or p65 isoforms (Santa Cruz Biotechnology, Santa Cruz, CA) of NF-κB was added immediately after addition of nuclear extract to labeled oligonucleotide and then incubated for 30 min at room temperature.

PBMC (5 × 10⁶) were preincubated for 2 h in the absence or presence of CQ (100 μM) in 6-well polystyrene plates. Cells were then stimulated with LPS (100 ng/ml) in the presence of 10% PHS for 1 h. At this time point, 5,6-dichlororiboside imidazole and actinomycin D were added at 200 μM and 5 mg/ml final concentrations, respectively, to block transcription. Preliminary experiments established that transcription was terminated at these concentrations. At 0, 60, and 120 min after transcriptional blockade, total RNA was extracted, and 5 μg was analyzed by RNase protection assay (RPA) using a commercial kit (PharMingen, San Diego, CA). The mRNA content of specific bands was quantified by phosphorimager analysis.

PBMC were preincubated in the absence or presence of bafilomycin A₁ (100 nM) for 2 h before incubation with a fluorescent congener of CQ, QC (1 μM), for 2 h. Adherent cells were washed once with HEPES and visualized on a Zeiss LSM 510 laser confocal scanning microscope equipped with a ×63 oil objective lens.

Two-color staining to detect cell surface markers and intracellular TNF-α was performed according to the manufacturer’s protocol (R&D Systems). PBMC (1 × 10⁶) were incubated in the absence or presence of CQ (10, 30, or 100 μM) for 2 h at 37° in 24-well polystyrene plates. Cells were then stimulated for 6 h with LPS (100 ng/ml) in the presence of 10% PHS, washed twice with cold PBS, and stained with anti-CD14 PE conjugate (Caltag Laboratories, Burlingame, CA). After an additional two washes with cold PBS, PBMC were fixed with 2% paraformaldehyde, permeabilized with saponin, and stained with FITC-conjugated anti-TNF-α, which recognizes intracellular forms of TNF-α (R&D Systems). Stimulation with LPS was performed in the absence or presence of monensin (2 μM), an inhibitor of protein secretion (23). PBMC (1–2 × 10⁵) were counted on a FACScan flow cytometer (Becton Dickinson, San Jose, CA) and analyzed using CELLQuest software (Becton Dickinson). The monocyte population was gated on based upon forward and side scatter properties and CD14 staining (24).

Data are expressed as mean ± SE. Data sets were compared by the Student two-tailed, paired t test using a statistical software package (SigmaStat; Jandel Scientific Software, San Rafael, CA). The Bonferroni correction was utilized for multiple comparisons. Significance was considered achieved when the p value multiplied by the number of comparisons was <0.05.

Initial experiments sought to confirm data by others (5, 12, 13, 14, 15, 16, 17) that CQ inhibits TNF-α release. PBMC were pretreated for 2 h with a micromolar concentration range of CQ before an 18-h incubation with LPS. CQ inhibited TNF-α release, as measured by ELISA, in a dose-dependent fashion over the 1–100 μM concentration range tested (Fig. 1) with a calculated IC₅₀ value of 8.02 μM. The highest level of CQ examined, 100 μM, inhibited TNF-α release by >98 ± 2% (p = 0.00002; n = 3). CQ-induced inhibition of TNF-α was not secondary to cell death, in that viability of the untreated and CQ-treated PBMC was 95.7 ± 1.9% and 92.6 ± 3%, respectively (p = NS; n = 5).

To extend these findings and to confirm that CQ was not toxic to PBMC, levels of the proinflammatory cytokines IL-1β, IL-6, and RANTES were measured in the supernatant from LPS-stimulated PBMC pretreated with CQ (Fig. 2). The highest concentration of CQ examined (100 μM) nearly completely abrogated both IL-1β and IL-6 release. However, IL-6 release was not significantly affected by 10 μM CQ, unlike IL-1β secretion. Release of the β-chemokine RANTES was unaffected by CQ concentrations up to 100 μM. Thus, there is considerable variation in the concentrations of CQ required to inhibit proinflammatory cytokine release from PBMC.

Having confirmed that CQ inhibited TNF-α release without causing nonspecific cytotoxicity, the next set of experiments sought to establish at what level of cellular function CQ acted. To determine whether CQ interfered with TNF-α release pretranslationally, we investigated the effect of this drug on TNF-α mRNA accumulation as measured by Northern blotting and RPA. PBMC were preincubated in the absence or presence of CQ (10 or 100 μM) and then stimulated for 2 h with LPS. This time point was previously demonstrated to correspond with peak levels of LPS-induced TNF-α gene expression (25). Compared with unstimulated PBMC, LPS induced TNF-α mRNA levels by 12.0 ± 4.0-fold (p = 0.019; n = 8). CQ pretreatment antagonized TNF-α mRNA accumulation (Fig. 3) with 100 μM CQ reducing TNF-α mRNA levels by 61.2 ± 15.1% (range, 45–99%; p = 0.03; n = 8). CQ concentrations of 10 μM CQ reduced levels of TNF-α mRNA by 31.2 ± 21.5% (n = 5). CQ did not affect expression of the housekeeping gene G3PDH. These results demonstrate that CQ interferes with TNF-α release at the pretranslational level.

The mRNA transcripts of numerous proinflammatory cytokines, including TNF-α, contain distinct 3′ untranslated regions making them subject to regulation at the level of mRNA stability (26). To assess whether CQ reduced levels of TNF-α by destabilizing its RNA transcript, the rate of degradation of its mRNA was determined. PBMC were preincubated in the absence or presence of CQ (100 μM) and then stimulated with LPS for 1 h. At this time point, new transcription was blocked by treatment with 5,6-dichlororiboside imidazole and actinomycin D. After 0, 60, and 120 min of transcription inhibitor treatment, total RNA was extracted and analyzed by RPA (Fig. 4, inset). In the absence of CQ, LPS-stimulated TNF-α mRNA had a calculated half-life of 91 min, a value in close agreement with data from other investigators (27, 28). As expected based on our results using Northern blotting, CQ treatment of PBMC reduced the intensity of the TNF-α mRNA band at time 0 min. However, the rate of decay of TNF-α mRNA (half-life = 78 min) was not significantly different from that seen in the absence of CQ. Thus, CQ appears to prevent the initial production of cytokine transcript, rather than accelerating its degradation.

Nuclear translocation of the transcription factor NF-κB is known to potently induce the expression of numerous genes encoding proinflammatory cytokines, including TNF-α (29, 30). Because interruption of NF-κB activation is a mechanism known to reduce TNF-α gene expression (31), we evaluated the effect of CQ on NF-κB nuclear translocation and p50/p65 isoform composition in LPS-stimulated PBMC. After a 2-h incubation in the absence or presence of CQ (100 μM), PBMC were stimulated with LPS for 30 min. Nuclei were harvested from adherent cells, and NF-κB was quantitated in the extracts by EMSA. As described previously, unstimulated PBMC contained small amounts of NF-κB in the nucleus (Ref. 22 and Fig. 5). Stimulation with LPS led to 6.2 ± 4.5-fold induction of NF-κB translocation (p = 0.013; n = 3). CQ treatment did not inhibit NF-κB translocation. Rather, there was a trend toward slightly higher levels of NF-κB in the nucleus after CQ treatment of both resting and LPS-stimulated PBMC (Fig. 5). Although p50/p65 heterodimers potently activate transcription of TNF-α (32), p50 homodimers interfere with NF-κB transactivation by binding the NF-κB consensus site and preventing p50/p65 heterodimer binding (33). Thus, we analyzed the subunit composition of DNA-binding complexes by Ab supershift. NF-κB oligonucleotide-bound complexes were predominantly p50/p65 heterodimers (Fig. 5), in that antisera to either or both supershifted the complex. Bands were completely eliminated by a 250-fold excess of unlabeled NF-κB oligonucleotide probe (data not shown). These findings suggest that CQ does not reduce TNF-α gene expression by perturbing entry of p50/p65 NF-κB heterodimers into the nucleus or by promoting mobilization of inhibitory p50 homodimers.

We next sought to determine whether endolysosomal alkalization was necessary and/or sufficient to inhibit TNF-α gene expression. To mimic the alkalinizing effects of CQ, we treated PBMC with bafilomycin A₁, a specific inhibitor of lysosomal v-ATPase (20). Bafilomycin A₁ treatment alone had no effect on LPS-stimulated TNF-α mRNA levels (Fig. 6), suggesting that alkalization of lysosomes is not sufficient to reduce TNF-α gene expression. Nevertheless, it still remained possible that CQ accumulation in endolysosomal compartments could inhibit TNF-α gene expression by a pH-independent mechanism. To exclude this possibility, we preincubated PBMC with bafilomycin A₁ (100 nM) before CQ (100 μM) treatment. The rationale for this approach was that by alkalinizing endolysosomes, bafilomycin A₁ pretreatment would greatly reduce the pH-dependent accumulation of CQ within this compartment. PBMC sequentially treated with bafilomycin A₁ and CQ before LPS stimulation had less TNF-α mRNA than PBMC treated with CQ alone (Fig. 6).

We next utilized confocal microscopy to examine the distribution of QC within human monocytes preincubated in the absence and presence of bafilomycin A₁ (100 nM). QC is a fluorescent congener of CQ, and the two diprotic drugs would be expected to have similar subcellular distribution (1). In the absence of bafilomycin A₁, QC accumulated within cytoplasmic vacuoles of mononuclear phagocytes (Fig. 7, A and B). However, when added after bafilomycin A₁ preincubation, QC was concentrated within the nucleus with a consequent loss of cytoplasmic staining (Fig. 7, C and D). Taken together, these results argue that endolysosomal accumulation and alkalization are neither necessary nor sufficient to reduce TNF-α gene expression. Furthermore, these data suggest that CQ mediates its effects outside of the endolysosomal compartment.

Although the above experiments demonstrated that isolated endolysosomal alkalization did not affect TNF-α gene expression, it remained possible that raising the pH of this subcellular compartment could disrupt TNF-α protein processing and/or secretion (34, 35). To address this issue, PBMC were incubated in the absence or presence of bafilomycin A₁ (100 nM) for 2 h and then stimulated with LPS (100 ng/ml). Bafilomycin A₁ treatment reduced TNF-α release by 46 ± 10.6% (p = 0.0034; n = 5) relative to untreated, LPS-induced PBMC. The solvent alone (methanol) had no effect on LPS-induced TNF-α release (115 ± 16.8%; p = NS; n = 3). PBMC viability in the bafilomycin A₁-treated group was 88.5 ± 5.2% (p = NS compared with untreated PBMC; n = 5).

To determine whether CQ similarly interfered with TNF-α processing and secretion, flow cytometry was utilized to measure cell-associated TNF-α in LPS-stimulated PBMC. Initially, intracellular staining was performed in the absence of monensin, an inhibitor of secretion. The rationale for this approach was that if CQ disrupted posttranslational processing and/or secretion of TNF-α, it would be detected as an increase in intracellular TNF-α. However, if TNF-α was secreted normally in the presence of CQ, pretreatment would not be expected to increase intracellular staining above background levels. Consistent with the latter possibility, CQ did not increase intracellular TNF-α staining in the absence of monensin (Fig. 8, top panels). Subsequent experiments were performed in the presence of monensin which, by blocking Golgi apparatus function, causes accumulation of all translated protein within the cell. In the presence of monensin, CQ significantly inhibited accumulation of intracellular cytokine in a dose-dependent fashion (Fig. 8, bottom panels) with 10 and 30 μM CQ reducing cell-associated TNF-α levels by 31.8 ± 14.2 and 57.2 ± 7.9%, respectively (p < 0.0015 for both concentrations; n = 3). The highest CQ concentration tested (100 μM) reduced intracellular TNF-α levels by 87 ± 8.3% (p = 0.00021; n = 3), closely approximating our ELISA findings. These data argue that CQ primarily acts pretranslationally by reducing TNF-α mRNA levels.

The data presented herein confirm that CQ reduces TNF-α release from human PBMC and begin to dissect the complex intracellular mechanisms by which CQ exerts its inhibitory effects. CQ interfered with LPS-induced TNF-α release by reducing levels of its mRNA without affecting the half-life of its RNA transcript or nuclear translocation of p50/p65 NF-κB heterodimers. Moreover, the inhibitory effect of CQ on TNF-α gene expression appeared to be independent of its ability to alkalinize endolysosomal compartments. Bafilomycin A₁, which alkalinizes endosomes and lysosomes by a mechanism distinct from CQ, had no significant effect on LPS-induced TNF-α mRNA levels. Moreover, potent inhibition of TNF-α gene expression was observed when PBMC were sequentially treated with bafilomycin A₁ and then CQ. Such treatment would be expected to greatly reduce entry of CQ into endolysosomal compartments, as is evidenced by the confocal imaging studies using QC, a fluorescent congener of CQ.

Interference with LPS signaling by CQ is one possible explanation for the reduction of TNF-α mRNA levels. LPS forms a complex with LPS-binding protein and the glycosyl-phosphatidylinositol-linked receptor, CD14. The transmembrane LPS signal is transduced when this complex interacts with its signaling partner, a member of the Toll-like family of receptors (36). Once ligated, the cytoplasmic domain of Toll-like receptor interacts with the adapter protein MyD88 (37), which then interacts with and activates the IL-1R-associated kinase (38). IL-1R-associated kinase, via interaction with the adapter protein TNFR-associated factor-6, activates the NF-κB-inducing kinase to phosphorylate two I-κB (inhibitory protein that dissociates from NF-kB) kinases, IκKα and IκKβ (38, 39). These kinases phosphorylate cytoplasmic I-κB, targeting it for ubiquitination and proteasomal destruction (40). Free NF-κB is now able to translocate to the nucleus.

Our results demonstrating that CQ does not inhibit nuclear translocation of NF-κB argue that this LPS signal transduction pathway likely remains intact. Furthermore, by supershift analysis, we did not see an increase in inhibitory p50 homodimers (33). On the contrary, there was a trend toward increased nuclear p50/p65 heterodimers after CQ treatment. This increase is consistent with the capacity of CQ to inhibit cytosolic protease activity (41, 42). It is important to note that our gel shift findings do not reveal whether the p50/p65 NF-κB heterodimers appearing in the nucleus were capable of activating transcription. Additional proteins such as CREB-binding protein and p300, coactivators that bridge transcription factors with the transcriptional apparatus (43), are also required for transcriptional activation and could conceivably be inhibited by CQ.

NF-κB activation is necessary but not sufficient for TNF-α gene expression (44). Thus, it is possible that CQ interferes with the mobilization or activity of an essential transcription factor other than NF-κB. In support of this concept, CQ has been shown to interfere with binding of certain transcription factors to consensus binding sites on DNA (45). However, the interaction of NF-κB with DNA is insensitive to the presence of millimolar concentrations of CQ (46). Finally, CQ has been shown to bind to DNA (7), altering its structure (4). This could conceivably interfere with accessibility of chromatin or transcriptional activity, resulting in reduced levels of certain transcripts. Future studies will be directed at evaluating the molecular basis by which CQ inhibits TNF-α gene expression.

In addition to control at the transcriptional level, TNF-α is known to undergo extensive posttranslational regulation. After translation, the 26-kDa membrane-bound pro-TNF-α is cleaved at the cell surface by a matrix metaloproteinase, TNF-α converting enzyme (ADAM-17), releasing a soluble 17-kDa form of the cytokine (47). In our study, when PBMC were treated with bafilomycin A₁, LPS-induced TNF-α release was reduced without altering levels of its mRNA, suggesting that endolysosomal alkalization interfered with secretion or processing of TNF-α. However, we were unable to detect a similar pH-dependent effect after CQ treatment. Thus, flow cytometry in the absence of monensin did not reveal an accumulation of cell-associated TNF-α after CQ treatment. In support of this finding, the reduction in levels of TNF-α mRNA and release were comparable, arguing that CQ acts primarily by reducing TNF-α gene expression. However, these data must be interpreted cautiously because flow cytometry might lack the sensitivity to detect an effect on protein processing if the amount of protein present within human cells is already reduced by the pretranslational effects of CQ. Therefore, posttranslational inhibitory effects of CQ on LPS-stimulated TNF-α release from human PBMC cannot be totally excluded.

In contrast to our findings, a previous study using murine monocytic cell lines (RAW 264.7 and P388D1) demonstrated that CQ (100 μM) inhibited LPS-stimulated TNF-α release by interfering with TNF-α processing and release (5), causing pro-TNF-α to become lodged in the secretory apparatus. Only at toxic concentrations of CQ (250 μM) were reduced levels of TNF-α mRNA observed (5, 16). These findings are contrasted by our data and by a previous study evaluating the effect of CQ (100 μM) on TNF-α release from human whole blood (15), showing that CQ interferes with TNF-α gene expression in human cells at physiologic concentrations. Furthermore, we were unable to demonstrate effects of CQ on protein processing or secretion. Thus, it appears that CQ might differentially affect TNF-α gene expression in human and murine monocytic cells. This hypothesis is supported by the observation that transcriptional induction of TNF-α appears to differ in humans and mice (48). For example, NF-κB is very important for murine TNF-α gene expression, whereas its role in humans is controversial. It is possible that CQ interferes with the action of a transcription factor other than NF-κB that is involved in human, but not murine, TNF-α expression.

The inhibitory effects of CQ on LPS-stimulated TNF-α gene expression and release are highly unlikely to result from nonspecific cellular toxicity. First, although CQ also inhibited release of IL-1β and IL-6, numerous other cellular functions were unaffected by CQ, including nuclear translocation of NF-κB, expression of the housekeeping genes L32 and GAPDH, and release of the proinflammatory β-chemokine RANTES. Second, PBMC viability was not affected after treatment with the highest concentration of CQ studied (100 μM). This lack of toxicity is consistent with results from other investigators (13, 14, 15, 17) and is not surprising considering that in vitro treatment of leukocytes with 100 μM CQ results in intracellular concentrations of CQ comparable to those obtained in vivo during CQ therapy (3).

The clinical implications of our data are speculative. TNF-α is critical for the development of both the innate and adaptive immune response (49, 50). On the other hand, it has been argued that the pathogenesis of septic shock involves dysregulated production of this and other proinflammatory mediators. Therefore, numerous studies have postulated that immunomodulation of the proinflammatory cytokines TNF-α and IL-1β would reduce sepsis-related morbidity and mortality. Although certain animal models of LPS-induced septic shock argue that blockade of proinflammatory cytokine signaling is beneficial, clinical trials utilizing anti-cytokine strategies to block TNF-α or IL-1β function have failed to show any benefit and have occasionally worsened clinical outcomes in sepsis (51, 52). Nevertheless, in certain infectious and autoimmune diseases, neutralization of TNF-α with mAbs or soluble TNF-α receptors has improved clinical outcomes (53, 54). Despite their potentially deleterious properties, CQ and its congeners remain safe and affordable therapies for malaria and rheumatologic disorders. Interestingly, in severe falciparum malaria, the capacity of CQ to inhibit TNF-α release might contribute to its clinical efficacy (55, 56). Clearly then, caution must be exercised when utilizing drugs known to modulate the proinflammatory cytokine response, a “double-edged sword” at best. By further dissecting the mechanism by which CQ reduces TNF-α release, it might be possible to design congeners with either greater or lesser anti-inflammatory properties. Such drugs could prove more efficacious in certain disease states.

We thank Ryan Hastey for help with the confocal microscopy and Dr. Chao Huang for performing the RANTES ELISA.