Necessity of oligonucleotide aggregation for toll-like receptor 9 activation.

Toll-like receptor 9 (TLR9), a member of the interleukin-1 (IL-1) family of pathogen-associated molecular pattern receptors, is activated by unmethylated CpG-containing sequences in bacterial DNA or synthetic oligonucleotides (ODNs) in the endosomal compartment. The stimulation of an IL-1 response is thought to require the aggregation of its receptor. By analogy, we postulated that the potency of a TLR9 ligand should depend first on its ability to enter cells and gain access to TLR9 and second on its capacity to form a multimeric complex capable of cross-linking these receptors. Previously, we selected from a random library a series of phosphodiester ODNs with enhanced ability to permeate cells. Here, we studied the structural requirements for these penetrating ODNs to elicit a functional TLR9 response, as assessed by cytokine production from bone marrow-derived mouse mononuclear cells. The presence of a prototypic murine immunostimulatory DNA hexameric sequence (purine-purine-CG-pyrimidine-pyrimidine) in the ODNs was not sufficient for stimulation. In addition, the TLR9-activating ODNs had to have the ability to form aggregates and often to form secondary structures near the core CpG motifs. Multimerization was promoted by the presence of a guanine-rich 3'-terminus. The phosphodiester ODNs with CpG motifs that did not aggregate antagonized the effects of the multimeric TLR9 activators. These findings suggest that an optimal TLR9 agonist needs to contain a spatially distinct multimerization domain and a receptor binding CpG domain. This concept may prove useful for the design of new TLR9-modulating agents.

The natural activating ligand for TLR9 is bacterial DNA containing unmethylated CpG dinucleotides, commonly referred to as immunostimulatory DNA or sequence (ISS) (15)(16)(17). Comparative studies of CpG oligonucleotides (ODNs) with different sequences have shown that the neighboring residues of the core CpG dinucleotides strongly influence immunostimulatory potency (18 -21). However, there are some differences between the optimal murine and human ISS. The central murine ISS motif for TLR9 activation has been shown to consist of the hexameric sequence purinepurine-CG-pyrimidine-pyrimidine in phosphodiester or phosphorothioate ODNs of at least 12 nucleotides (22). The prototypic ODN sequence for activation of human TLR9 contains a TCG sequence at the 5Ј-end of the molecule (23,24). Other sequence modifications influence the abilities of the ISS to activate selectively TLR9-expressing macrophages, dendritic cells, or B lymphocytes (25,26).
The initiating events in TLR9 activation have not yet been analyzed, in part because the recombinant protein is difficult to purify in quantity and to solubilize. However, the TLRs are structurally homologous to the interleukin-1 (IL-1) receptors, which have been studied in more detail. Activation of IL-1 type I receptors correlated with IL-1-dependent receptor aggregation (27). However, the IL-1 receptor antagonist, which blocks IL-1 function by competing for receptor binding, did not induce IL-1 receptor aggregation. If a similar signaling mechanism applied to TLR9, one could predict that self-aggregating ISS would strongly activate the receptor by promoting its multimerization, whereas nonaggregating ISS would antagonize these effects. According to this hypothesis, the higher structure of an ISS-ODN, rather than just its primary sequence, could dictate its ability to stimulate or to inhibit innate immune system activation.
To gain insight into this problem, we compared the physical properties and the TLR9-activating capacities of a series of phosphodiester ODNs that were selected from a random library for their capacity to be efficiently internalized by cells. Since all of the ODNs were selected for their ability to penetrate cells, their different functional activities were considered to be attributable mainly to TLR9 activation. The experimental results showed that a prototype ISS is not sufficient for TLR9 activation. Aggregating phosphodiester ODNs with ISS in rigid secondary structures activated the receptor effectively, whereas nonaggregating ODNs with CpG motifs functioned as receptor antagonists.

EXPERIMENTAL PROCEDURES
Oligodeoxynucleotides-Phosphodiester ODNs from a random library with scrambled insets of 40 nucleotides were selected for their ability to penetrate cells by a repetitive selection procedure that involved (a) incubation with viable cells, (b) extensive and stringent washing to remove all external ODN binding, and (c) asymmetric PCR amplification, as described previously (28). After 10 rounds of selection, the retained intracellular ODNs were amplified, cloned, and sequenced. Several 40-mer ODNs corresponding to the recovered sequences as well as a random ODN of the same length (random 40) were synthesized by Integrated DNA Technologies (IDT, Corvallis, OR) ( Table I). The prototype ISS-ODNs, 1018 and 1826, have been described (29 -32) and were synthesized with both phosphodiester and phosphorothioate backbones. Endotoxin contamination in ODNs was negligible as measured by a Limulus amebocyte lysate assay (BioWhittaker, Walkersville, MD).
Uptake studies were performed using fluorochrome-labeled ODNs to confirm their abilities to penetrate cells, as described (28). Briefly, viable cells were incubated in protein-free medium with 5Ј-Cy3-labeled ODNs at 0.5 M for 2 h at 37°C. Following washes with 3% fetal bovine serum in RPMI 1640 medium (Irvine Scientific, Santa Ana, CA), these cells were analyzed by flow cytometry using a Becton Dickinson FAC-Scaliber. Data analysis was carried out using FlowJo 3.4 software (Tree Star, Inc., Stanford, CA).
Isolation of Bone Marrow-derived Mononuclear Cells-BALB/c and C57BL/6 mice were purchased from The Jackson Laboratory (Bar Harbor, ME). MyD88 Ϫ/Ϫ and TLR9 Ϫ/Ϫ mice were a generous gift of Dr. S. Akira (Osaka University, Japan) (33). The mice were bred and maintained under standard conditions in the University of California, San Diego Animal Facility that is accredited by the American Association for Accreditation of Laboratory Animal Care. All animal protocols received prior approval by the institutional review board.
Bone marrow harvested from the femurs and tibias of various strains of mice were plated in non-tissue culture-treated Petri dishes with Dulbecco's modified Eagle's medium high glucose medium supplemented with 10% fetal bovine serum, L-glutamine, penicillin/streptomycin, all from Invitrogen, and 30% L929 cell-conditioned medium. Cells were grown at 37°C, 5% CO 2 for 7 days without replacing the medium. The bone marrow-derived mononuclear cells were harvested afterward by gentle scraping, counted, and replated in medium with different conditions described below.
ODN-stimulated Cytokine Release-For studies on cytokine production, 7-day-old bone marrow-derived mononuclear cells were seeded in 96-well plates at a density of 5 ϫ 10 4 cells/well and grown for another 3 days. These cells were then incubated with ODNs at a final concentration of 0.2, 0.5, or 1 M for 48 h without further supplement. In the competitive receptor binding study, the highly active ODN R10-60 (0.5 M) was premixed with various concentrations of the inactive ODN R10-9, R10-32, or R10-13 in serum-free medium, prior to addition to the cells. Culture supernatants were collected at the end of incubation and stored at Ϫ20°C for later determination of IL-12p40/70 by sandwich enzyme-linked immunosorbent assay (BD Biosciences).
In Vitro Kinase Assays-For kinase assays, the enriched mononuclear cells were dispersed in 6-well plates at a density of 1-2 ϫ 10 6 cells/ml/well and allowed to settle overnight. ODNs were then added to the mononuclear cells at 1 M and incubated for 0.5-2 h. The cells were quickly lysed in buffer A (20 mM Hepes, pH 7.9, 1 mM EDTA, 1 mM EGTA, 1% Nonidet P-40, 1 mM glycerophosphate, 2.5 mM sodium pyrophosphate, 1 mM sodium orthovanadate, 2 mg/ml aprotinin, 1 mM phenylmethylsulfonyl fluoride) with proteinase inhibitors and 1 mM dithiothreitol on ice and centrifuged at 12,000 ϫ g for 1 min. The aqueous phase containing cytoplasmic proteins was removed and saved. The nuclear pellet was lysed in buffer B (20 mM Hepes, pH 7.9, 1 mM EDTA, 1 mM EGTA, 0.4 M NaCl) and vortexed, and the nuclear supernatant was collected after centrifugation.
Specific kinases were immunoprecipitated from cytosolic proteins with either anti-IB kinase-␤ or anti-Jun NH 2 -terminal kinase 1 antibodies (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) at 4°C overnight. Afterward, the immune complexes were washed successively in buffer A containing 0.5 M NaCl followed by kinase buffer (25 mM Tris, pH 7.5, 10 mM MgCl 2 , 2 mM EGTA, 1 mM dithiothreitol, 1 mM sodium orthovanadate). IB kinase-␤ or Jun NH 2 -terminal kinase 1 kinase assays were performed using the respective recombinant glutathione S-transferase fusion protein with IB␣ or c-Jun as the respective substrates in the presence of 0.1 Ci of [␥-32 P]ATP at 37°C for 30 min, as described (34). The 32 P-labeled products were separated by SDS-PAGE and visualized by autoradiography.
DNA Secondary Structure Prediction-To predict the presence of secondary structures, the DNA mfold program (35) (available on the World Wide Web at www.bioinfo.rpi.edu/applications/mfold/old/dna/) was employed. The various sequences were submitted as linear DNA FIG. 1. Cellular uptake of selected ODNs. Bone marrow-derived mononuclear cells from TLR9 Ϫ/Ϫ and wild type control (B6 ϫ 129 F2 wild type mice from the colony) were incubated in protein-free medium with 5Ј-Cy3-labeled ODNs at 0.5 M for 2 h at 37°C. Following washes with 3% fetal bovine serum in RPMI 1640 medium, these cells were analyzed by flow cytometry. Examples of uptake by two selected fluorochrome-labeled ODNs and the random40 ODN are shown.

TABLE I
Oligonucleotide sequences and immunostimulatory activity CpG dinucleotides are in boldface type. Murine ISS motifs (5-purine-purine-CG-pyrimidine-pyrimidine) are underlined. IL-12p40/p70 production (ng/ml) was determined by ELISA using cell media of bone marrow-derived mouse mononuclear cells from different strains after a 2-day treatment with 0.5 M ODN or 5 ng/ml LPS. 1018* is mouse prototype ISS-ODN with phosphorothioate backbone. Data are representative of at least three separate experiments. ODN Sequence

Physical Characteristics of Immune Stimulators
and analyzed based on free energy using default program settings, assuming a temperature of 37°C, with ionic conditions of 150 mM Na ϩ and 0.5 mM Mg 2ϩ . Analysis of ODN Multimerization by Size Exclusion HPLC and PAGE-A TSK-Gel G2000SW XL HPLC column with a 5-m particle size (MAC-MOD Analytical, Montgomeryville, PA) was used to perform the size exclusion assay as previously described (28,36). Briefly, 50 l of a 50 M ODN solution in 30 mM NaCl was injected, and elution was carried out in buffer containing 10 mM sodium phosphate, pH 6.9, 0.3 M NaCl at a flow rate of 0.6 ml/min. The HPLC elution fractions were divided into a high molecular weight aggregate portion (retention time 9 -12.5 min) and a low molecular weight monomer portion (retention time 12.5-15 min). To address the association of ODN multimerization with immunostimulatory activity, the two fractions were collected, equal amounts were added to bone marrow-derived mononuclear cells, and the culture supernatants were collected 48 h later for enzymelinked immunosorbent assay, as described above.
For gel analysis, phosphorothioate and phosphodiester ODNs were mixed in RPMI 1640 with and without 2% tissue culture grade bovine serum albumin (Sigma) and incubated for 10 min at 37°C. 12 l of the mixture were then separated on a 4 -20% nondenaturing TBE polyacrylamide gel (Invitrogen). The oligonucleotides were visualized by staining with SYBR Green II (Molecular Probes, Eugene, OR) under UV light. The protein bands were then detected with Coomassie Blue staining.
Circular Dichroism Spectroscopy-Oligonucleotides were resuspended in 10 mM sodium phosphate buffer, pH 7.2, containing 0.1 M KCl at a final concentration of 10 M (final volume of 300 l), boiled for 5 min, and annealed at 60°C for 2 days (37). After slow cooling to room temperature, the samples were analyzed on an AVIV CD spectrometer (model 202, AVIV instruments, Inc., Lakewood, NJ) using a wavelength scan from 320 to 200 nm at 25°C. Spectra were collected over three scans at 1-nm bandwidth, 1-nm wavelength step, and an average 0.5-s response time for each sample. Data are presented as the average of three scans with integrated curve fitting performed by Prism software (version 3.0; GraphPad Software, Inc., San Diego, CA).

RESULTS
ODN Uptake Is Independent of TLR9 -We previously described the selection of phosphodiester ODNs with an average length of about 40 nucleotides that displayed improved cellular uptake compared with random sequence ODNs (28). Because the ODNs shown in Table I were selected for uptake by human B cells, it was necessary to confirm that they also effectively penetrated murine bone marrow-derived mononuclear cells. Experiments with fluorochrome-labeled ODNs showed that they were taken up 2-14-fold better than random sequence ODNs of the same length (examples are shown in Fig. 1). In addition, bone marrow-derived cells from TLR9 Ϫ/Ϫ mice displayed an uptake efficiency similar to cells from wild type mice.
Sequence Requirements for Activation of Bone Marrow-derived Mononuclear Cells-To characterize the mechanisms involved in activation by these penetrating ODNs and to study the structural and functional requirements for stimulation, we carried out studies on murine bone marrow-derived mononuclear cells from different strains. A panel of ODNs containing different numbers of CpG dinucleotides and murine ISS motifs were first compared for their abilities to induce IL-12p40/p70 secretion. As expected, ODNs without any CpG dinucleotides had no ISS activity (Table I, R10-5). Unexpectedly, however, no detectable IL-12 was released by cells treated with ODN R10-9, which contained the prototype ISS sequence motif AACGTT at the 5Ј terminus. Cells stimulated with ODNs that contained at least two sets of CpG dinucleotides produced detectable levels of IL-12. Furthermore, phosphodiester ODN R10-60 showed comparable or even better IL-12 stimulation than the positive control ODN 1018, with a more nuclease-resistant phosphorothioate backbone (Table I).
As little as 0.2 M ODNs R10-53, R10-60, R10-86, and D-R15-8 were sufficient to induce detectable IL-12, and the levels increased in proportion to the ODN concentration (Fig.  2). In contrast, R10-9 was not able to elicit any IL-12 secretion at concentrations up to 1 M. Together, these data demonstrated that phosphodiester ODNs can display equivalent immunostimulatory activity toward murine bone marrow-derived mononuclear cells as phosphorothioate ODN, and that a CpG motif is necessary but not sufficient for cell activation.
Role of the TLR-9 and MyD88 Pathways-Since there was no absolute correlation between an ISS motif and immunostimulatory activity among the selected ODNs, it was important to confirm that the ODNs signaled through the TLR9 and MyD88 pathway. No IL-12 production was observed from ODN-stimulated bone marrow-derived mononuclear cells from either TLR9 Ϫ/Ϫ (Table I) Table I

Physical Characteristics of Immune Stimulators
whereas bacterial lipopolysaccharide clearly activated these cells through the recently described MyD88-independent alternative pathway (Fig. 3). The kinetics of cell activation revealed a maximum at 2 h after phosphodiester ODN application.
Association of Multimerization with ISS Activity-As the primary ISS motif (e.g. in R10-9) was insufficient for immunostimulatory activity, we evaluated whether a higher structure of an ODN also could influence its biologic properties. Results of the DNA mfold program showed that the CpG dinucleotide sequences in the active ISS-ODN, at their predicted lowest free energy states, were often in or near rigid stem loop structures, whereas the CpG in R10-9 was not (Fig. 4A). Furthermore, point mutations of the CpG located within the predicted rigid loop structures of ISS-ODN also reduced their ability to activate murine bone marrow derived mononuclear cells (Table II, R10-53(T18G) and R10-60(T21G)). More importantly, size exclusion HPLC analysis showed that the active ISS-ODN formed multimers (Fig. 4B, R10-53, R10-60, and D-R15-8), whereas the inactive ISS-ODN R10-9 did not. Nondenaturing polyacrylamide gel fractionation of the ODNs also confirmed the presence of multimers in the biologically active ODN samples (data not shown). Removal of the guanine-rich sequences in the 3Јterminus of R10-60 and D-R15-8 (R10-60a and D-R15-8a) or near the 5Ј-end of R10-53 (R10-53(Ϫ7-14Pu)) abolished aggregate formation. Finally, circular dichroism spectroscopic analyses revealed absorption maxima that have been previously   FIG. 4. ODN secondary structure and multimerization. A, predicated structure derived from the DNA mfold program (available on the World Wide Web at www.bioinfo.rpi.edu/applications/mfold/old/dna/). Sequences were submitted as linear DNA with folding at 37°C and ionic conditions of 150 mM Na ϩ , 0.5 mM Mg 2ϩ , oligomer type corrections. CpG dinucleotides were boxed. B, ODN aggregate formation assessed by size exclusion HPLC. Samples were prepared at a concentration of 50 M in 30 mM NaCl, and 50-l aliquots were injected into the TSK column. High molecular weight aggregates eluted from the column starting at around 9 min, whereas monomers eluted at 13 min. R10-53 (Ϫ7-14Pu), 5Ј-TCTGCGCTACGTTACTAGTCGTGTGTCCGTG-3Ј. C, secondary structure formation measured by CD spectroscopy. The CD analysis was carried out using a wavelength scan from 200 to 320 nm at 25°C. Spectra were collected from three scans at 1-nm bandwidth, wavelength step of 1 nm, and at an average time interval of 0.5 s. An integrated curve was derived from the average of three scans using Prism software. The presence of a parallel type guanine quartet is suggested by a positive maximum at ϳ265 nm and a negative minimum near 240 nm (37,55).

Physical Characteristics of Immune Stimulators
associated with the presence of guanine quartets, which are known to form aggregated structures (Fig. 4C).
Although ODN multimerization correlated with enhanced immunostimulatory activity (Tables I and II), this observation did not prove that aggregation was responsible for the stimulation potency. To address the question directly, ODN fractions of different sizes were collected from the HPLC elutes, and equal amounts were added to bone marrow-derived mononuclear cells. The maximal IL-12 was produced by cells that were stimulated with ODN aggregates, whereas at least 5-10-fold less IL-12 was observed from cells stimulated with ODN monomers (Fig. 5). Modified ISS-ODN, which had their 3Ј-guanine tails removed to diminish multimerization, also lost stimulation activity (Table II).
Under physiological conditions, phosphorothioate ODNs also aggregated and formed multimers by binding to plasma proteins (Fig. 6). Phosphodiester and phosphorothioate ODNs of the same sequence were incubated in medium with or without bovine serum albumin and analyzed by nondenaturing TBE PAGE. In the presence of bovine serum albumin, the prototype phosphorothioate ISS-ODNs, 1018 and 1826, were retained in the gel at higher molecular weights than the oligonucleotides in unsupplemented medium (Fig. 6A). The Coomassie Bluestained bands suggested that these phosphorothioate ODNs co-migrated with the protein (Fig. 6B). In contrast, the phosphodiester counterparts of 1018 and 1826 were only visualized at the monomeric molecular weight and did not appreciably bind to protein. To determine whether ISS activity was retained in the absence of exogenous proteins, bone marrowderived mononuclear cells were extensively washed to remove plasma proteins and then cultivated in completely serum/protein-free medium with the different ODNs. In the absence of serum, R10-60-and 1826-stimulated cells released ϳ34 and 17% as much IL-12, respectively, as the cells stimulated in serum-containing medium (data not shown). However, ϳ40% of the cells died during a 24-h cultivation, making the exact interpretation of these results uncertain. Nonetheless, the data support that aggregating phosphodiester ODNs are better im-munostimulants than phosphorothioate ODNs under serumfree conditions.
Biological Activities of ODN Monomers-The previous experiments demonstrated that monomeric ODNs with an ISS motif failed to effectively stimulate mouse bone marrow-derived mononuclear cells. However, the question remained whether the nonaggregating ODNs displayed other biological activities, such as the ability to antagonize cell activation by multimeric ISS-ODNs. To test this possibility, cells were stimulated with a mixture of R10-60 that forms aggregates and the nonaggregating ODN, R10-9, R10-32, or R10-13, at different ratios. A 10:1 molar excess of the monomeric ODNs reduced IL-12 production by ϳ64, 55, and 46%, respectively, and even a 2:1 ratio had significant inhibitory activity (Fig. 7). The latter concentrations of nonaggregating R10-9 were insufficient to impede R10-60 uptake (results not shown).

DISCUSSION
Many oligonucleotides with CpG dinucleotide motifs have immunostimulatory activity. In order to activate mouse immune cells, an ISS-ODN generally must be at least 12 nucleotides long and should contain a prototypic ISS motif (purinepurine-CG-pyrimidine-pyrimidine) (22,32,38). However, there are many reported exceptions to this rule (39 -41). Akira and co-workers (16) discovered that cell activation by ISS-ODNs depends on the expression of the pattern recognition receptor TLR9. Functionally significant structural polymorphisms in TLR9 are not known. Therefore, it has been difficult to explain why minor sequence changes in ODNs with CpG motifs produce diverse activities toward dendritic cells and B lymphocytes, the two cell types that principally express TLR9. The results of the present experiments offer one possible explanation for these phenomena. Our data suggest that not just the primary structure of an ODN, but also its physical aggregation state, controls its immunostimulatory potency. Indeed, many ISS-ODNs with a uniform sequence are complex mixtures of monomeric and multimeric species, in which only the aggregates have clear-cut immunostimulatory activity (Fig. 5). More-

TABLE II
Effect of sequence modification on immunostimulatory activity CpG dinucleotides are in boldface type. Aggregate formation was determined by size exclusion HPLC analysis, and data are presented as percentage of peak area from two representative HPLC runs. Immunostimulatory activity was determined based on IL-12p40/p70 production (pg/ml) from B6 ϫ 129 F2 bone marrow-derived mouse mononuclear cells treated with 0.5 M ODN for 2 days. Data are normalized relative to the corresponding parent ODN set at 100% activity. Physical Characteristics of Immune Stimulators over, in some cases, the monomers can act as ISS antagonists (Fig. 7) (40). Thus, the net activity of an ISS-ODN is probably very complex, depending on the exact ratio of multimers to monomers at the endosomal sites of TLR9 activation. The two most common initiating events for activation of membrane receptors are aggregation and conformational change. The TLRs are members of the larger IL-1 receptor family. Fluorescence resonance energy transfer studies have demonstrated a close correlation between the ability of IL-1 to cause the aggregation of the receptor and the stimulation of a functional response (27). To determine the exact mechanism of TLR9 signaling will require equivalent direct studies of ligand-receptor interaction. However, our results strongly support the hypothesis that receptor aggregation is an important component of TLR9 activation, as has also been proposed for TLR4 signaling (42).
The natural ligand for TLR9 is thought to be bacterial DNA, which is a high molecular weight polymer containing multiple CpG motifs. Our experiments showed that phosphodiester ODNs with multiple CpG dinucleotides were more potent than ODNs with one or two CpG sequences (Table I and Fig. 2). However, there are ample reports showing that ODNs with a single ISS sequence can display potent immunostimulatory activity. With some exceptions, these ODNs usually have either a guanine stretch at the 3Ј-end of the molecule or a chemically modified backbone such as phosphorothioate (21,(43)(44)(45)(46)(47)(48). Single-stranded guanine-rich ODNs are known to have a propensity to form quadruplex structures and higher aggregates that are often essential for their biologic activities (48 -51). The presence of a guanine tail on phosphodiester ISS-ODNs thus may enhance activity both by enhancing cellular uptake and by facilitating their capacity to cross-link TLR9.
Guanine-rich stretches do not enhance the immunostimulatory activities of phosphorothioate ODNs. However, multimerization by aggregation might augment the potency of these ODNs. Our in vitro result and an in vivo disposition study have demonstrated that phosphorothioate ODNs bind extensively to plasma proteins within 15 min of administration (Fig. 6) (52). Thus, under physiologic conditions, phosphorothioate ODNs are almost certainly aggregated. In contrast, the less hydrophobic phosphodiester ODNs bind plasma proteins much less. Hence, the predominant in vivo state of phosphorothioate ISS-ODNs is most likely multimeric. The increased potency of phosphorothioate ODNs, compared with ODNs with a phosphodiester backbone, has been attributed to stability against degradation by nucleases. Our data showed that the specific immunostimulatory activity of strongly aggregating phosphodiester ODNs was nearly equivalent to a prototype phosphorothioate ISS-ODN. Future experiments will need to determine to what extent the enhanced potency of phosphorothioate ODNs is attributable to ligand multimerization mediated by plasma protein binding as opposed to nuclease stability.
Several investigators have described synthetic ODNs that inhibit the activity of ISS-ODNs, both in vitro and in vivo (40). Such ISS antagonists could block the uptake of the active ODNs or their ability to cross-link TLR9. Our results showed that several nonaggregating phosphodiester ISS-ODNs dosedependently diminished the in vitro ability of an aggregating ISS-ODN, R10-60, to activate bone marrow-derived mouse mononuclear cells (Fig. 7). The IL-1 receptor antagonist anakinra, which is used to treat rheumatoid arthritis, has been demonstrated to interact with the type I IL-1 receptor without causing receptor aggregation. By analogy, we postulate that a similar mechanism may underlie the biologic effects of some ISS-ODNs.
Ligand aggregation in an intact organism is a process that is very difficult to predict based on in vitro studies, which use much lower concentrations of proteins and other ligand-binding macromolecules. It may be preferable to use stable, preformed ODN multimers if optimal TLR9 activation indeed requires receptor cross-linking. Several studies have shown that ISS-ODNs conjugated to proteins are more potent immunostimulators than ISS protein mixtures (53,54). Recently, Fearon et al. (40) demonstrated that 5-nucleotide-long ODNs, with a human ISS motif (TCG), could potently stimulate interferon production by human peripheral blood mononuclear cells if adsorbed to the surface of cationic microparticles. They postulated that the multimerization of ISS motifs preferentially induced interferon ␣ secretion in human cells, compared with other ISS functions. It is likely that the density of TLR7 and TLR9 molecules in endosomal membranes, as well as their heterodimerization, varies among cell types and at different states of cell differentiation. In this setting, the physical state of TLR ligands may be a critical determinant in dictating the specificity and potency of cell activation.