Phosphodiester backbone of the CpG motif within immunostimulatory oligodeoxynucleotides augments activation of Toll-like receptor 9

Toll-like receptor 9 (TLR9) stimulatory CpG-containing oligodeoxynucleotides (ODNs) with phosphorothioate backbones have successfully replaced the naturally occurring agonists of TLR9 in drug development due to their increased stability. Replacing the nonbridging oxygen with a sulfur atom in the phosphate linkage of ODNs has been accepted as having a minor impact on the chemical and physical properties of the agonists. Here, we report that the TLR9 binding site exhibits a strong bias in favor of a phosphodiester backbone over the phosphorothioate backbone of the CpG motif. Furthermore, we show that while single point mutations of W47, W96 and K690 within the TLR9 binding site retains full TLR9 activation by phosphodiester-based ODNs, activation by phosphorothioate-based ODNs is strongly impaired. The substitution of a phosphorothioate linkage for a phosphodiester linkage of just the CpG motif considerably improves the activation potency of a phosphorothioate-based oligonucleotide for human B-cells and plasmacytoid dendritic cells, as well as for mouse bone marrow-derived dendritic cells and macrophages. Our results highlight the functional significance of the phosphodiester linkage of a CpG dinucleotide for binding, which is important in designing improved immunostimulatory TLR9 agonists.

binding site discriminates between nonbinding oxygen and a sulfur atom in the phosphate bond of the CpG motif. Improvement in the activation efficiency of ODNs by introducing a PD backbone within the CpG motif is particularly important for the design of improved immunostimulatory synthetic hTLR9 agonists.

Results
Backbone of 5′CpG motif determines activation potency of ODNs for hTLR9. The substitution of nonbinding oxygen for a sulfur atom in a phosphate linkage confers a synthetic TLR9 agonist nuclease resistance and consequently improves the immunostimulatory potency of ODNs. To evaluate the impact of modified physical and chemical properties of this replacement on the activation efficiency of hTLR9, we designed synthetic agonistic ODNs with a mixed PTO and PD backbone. The minH75 (H75) was used as an ODN template and contains the minimal sequence motifs required to activate hTLR9 16 . In addition to the PD-and PTO-based H75 (H75 PD , H75 PTO ), variants of H75 (H75 variants) were designed with mixed backbone chemistry (Fig. 1a,b; oligonucleotide sequences can be found as Supplementary Table S1). We substituted the first, second, or both CpG motifs with PTO-based CpG motifs within H75 PD , generating H75 PD -CG 1 PTO , H75 PD -CG 2 PTO , and H75 PD -CG PTO , respectively. Using a similar strategy, H75 PTO -CG 1 PD , H75 PTO -CG 2 PD , and H75 PTO -CG PD were designed with PTO-based CpG motifs of H75 PTO substituted with PD-based CpG motifs of H75 PTO . Human B-lymphocytes (Ramos Blue cells) and human B-cells and pDCs isolated from PBMC were used to test the potency of H75 variants for the activation of TLR9. The majority of H75 variants, derived from the human-specific minH75, activated Ramos Blue cells (Fig. 1c,d); nevertheless, PTO-based H75 activated hTLR9 at approximately ten-fold lower concentrations than H75 PD , which can be ascribed to the protective nature of the PTO backbone against the nuclease degradation 14 . To our surprise, the H75 PTO variants with the PD-based first CpG motif (H75 PTO -CG 1 PD and H75 PTO -CG PD ) were significantly more potent activators of Ramos Blue cells compared to the PTO-based first CpG motif (H75 PTO and H75 PTO -CG 2 PD ) ( Fig. 1c). Similarly, the H75 PD variants with the PD-based first CpG were significantly better activators of hTLR9 than H75 PD -CG 1 PTO with the PTO-based first CpG motif (Fig. 1d). The relevance of a PD-based first CpG motif for the activation of hTLR9 was also confirmed on primary cells. We isolated untouched B-cells and enriched pDCs from human PBMCs and treated the cells with H75 PTO and H75 PTO -CG 1 PD (Fig. 1e,f). The secretion of hIL6 and hTNFα was significantly higher when B-cells and pDCs, respectively, were stimulated with H75 PTO -CG 1 PD compared to stimulation with H75 PTO . The hIFNα expression was not affected by H75 and its variants. These data establish that the backbone chemistry of the first CpG motif of H75 contributes considerably to the activation efficiency of hTLR9.
Furthermore, we examined whether the mixed backbone chemistry of H75 shows a similar effect on the activation of mTLR9. The H75 PTO weakly activated RAW-Blue macrophages, as already shown by Pohar et al. 20 (Fig. 1g). However, the substitution of PTO-with PD-based first, second, or both CpG motifs improved the activation efficiency of H75 PTO for mTLR9, with H75 PTO -CG 2 PD being the most potent activator of mTLR9. In contrast to H75 PTO , H75 PD effectively activated RAW-Blue macrophages (Fig. 1h), demonstrating that the sequence-specificity for PD-based agonists is less significant for the activation of mTLR9 15 . Contrary to hTLR9, H75 PD -CG 1 PTO and not H75 PD -CG 2 PTO was the most potent activator of mTLR9 (Fig. 1g,h). These results were subsequently confirmed using mouse bone marrow-derived dendritic cells (mBM-DCs) ( Supplementary  Fig. S1a,b). The most potent production of mTNFα and mIL6 was triggered by the stimulation of BM-DCs with the H75 PTO -CG 2 PD . Overall, for the activation of mTLR9, the second and not the first CpG motif can be used to fine-tune the potency of H75. This is not surprising, as we previously showed that the CpG motif positioned 4-6 nt from the 5′-end is important for the activation of mTLR9 15 .
Taken together, we demonstrated that the activation efficiency of H75 for hTLR9 can be significantly improved by the introduction of the PD-based first CpG motif and the PTO-based second CpG motif. To confirm that the PD substitution for the CpG motif improves the activation of other well-known B-type agonists, we selected PTO-based ODN2006 (h2006 PTO ) with four CpG motifs within ODNs. The activation of Ramos Blue cells was tested with h2006 PTO and its variant h2006 PTO -CG PD , having substituted the PTO-based first CG dinucleotide for the PD-based first CG dinucleotide. Similar to H75, the introduced PTO-based to PD-based substitution significantly improved the activation of hTLR9 ( Supplementary Fig. S1c).

Protection of 3′ end significantly improves activation of TLR9 by PD-based ODNs. Endonucleases
such as DNase I and exonucleases cleave naturally occurring ssDNA and dsDNA, producing short oligonucleotides 14 and mononucleotides; therefore, nuclease-resistant PTO-based TLR9 agonists are used in immunotherapeutic applications 21 . As the processing of exonucleases is 3′ to 5′ directional, we examined whether some protection could be achieved by modifying only the 3′-end of ODNs. To visualize the degradation dynamics and degradation products, the M80 PD was labeled at the 5′-and 3′-end with fluorescein isothiocyanate (FITC). In addition, an FITC label was used as a chemical modifier of thymine to protect the 3′-end against the exonucleases ( Supplementary Fig. S2). M80 PD labeled at the 3′-end with FITC (M80 PD_ 3′ FITC ) was protected from the nuclease degradation. For M80 PD_ 5′ FITC , however, degradation was already observed 4 h after incubation. No degradation of ODNs was observed in media lacking fetal bovine serum (FBS), demonstrating that the main source of nucleases is FBS. These results confirmed that the 3′-end protection of PD-based ODNs plays an important role in preventing the degradation of ODNs; therefore, it should consequently augment the activation of TLR9. An H75 PD variant with two 3′-end PTO-based thymines (H75s PD -TTT PTO ) was prepared, where the PTO linkage should protect the ODNs against the exonucleases. These modifications improved the TLR9 activation efficiency by H75 (Fig. 2a). Similar to the H75 variants (Fig. 1c), the 3′-protected H75 variant with PD-based first and PTO-based second CpG motifs (H75 PD -CG 2 PTO -TTT PTO ) was the most potent activator of Ramos Blue cells (Fig. 2b). This protected variant also activated Ramos cells more effectively in comparison to the unprotected H75 PD -CG 2 PTO (Fig. 1c). However, the PTO-based first CpG motif (H75 PD -CG 1 PTO -TTT PTO ) decreased the activation potency of H75 PD for hTLR9 (Fig. 2b). Taken together, the 3′-end protection of H75 improves hTLR9 activation; however, the impact of first CpG motif backbone chemistry on activation efficiency remains similar to that for H75.
The mixed backbone composition regulates potency of ODNs to activate mTLR9. We showed that careful positioning of PTO and PD bonds in human-specific H75 significantly improves the activation of hTLR9 and mTLR9 ( Fig. 1). Therefore, our next step was to verify whether the same rule applies for certain ODNs that specifically activate mTLR9. The mouse-specific minimal ODN, minM80 (M80) 15 and ODN1826 (m1826) and its variants were tested. A single CpG motif distant from the 5′-end by 4-6 nt is sufficient to activate mTLR9, as shown for the M80 15 . The m1826, on the other hand, possesses two CG motifs positioned at a distance of 8 and 17 nt from the 5′-end. As expected, M80 PTO failed to activate and M80 PD poorly activated human B-lymphocytes ( Fig. 3a,b) but effectively activated RAW macrophages (Fig. 3c,d) 15,16,20 . M80 PTO with the PD-based CpG motif (M80 PTO -CG PD ) was a more effective activator of RAW macrophages than M80 PTO (Fig. 3b), similar to H75 PTO -CG 2 PD activating mTLR9 (Fig. 1c). However, the PTO-based CpG motif within M80 PD (M80 PD -CG PTO ) reduced the potency to activate mTLR9 (Fig. 3d). The observed results were confirmed using mBM-DCs, with M80 PTO -CG PD significantly increasing the expression of mTNFα and mIL6 compared to M80 PTO (Fig. 3e).
Furthermore, we substituted the PTO-linkage of first CG dinucleotide of the ODN1826 PTO (m1826 PTO ) to PD-based, generating the m1826 PTO -CG PD variant. This m1826 variant activated the reporter cells better than m1826 PTO (Fig. 3f). The enhancement was not as strong as for the minimal ODNs (H75 and M80), which might be because the ODNs with the minimal required sequence have only the minimal required number of CpG motifs and substitution of a single motif can have a greater impact.
Taken together, the potency of PTO-based ODNs to activate TLR9 can be improved by substituting the PTO-based CpG motif backbone with the PD-based CpG motif backbone. The position of the CpG motif that backbone should be substituted from the PTO-to the PD-based differs for hTLR and mTLR9. For augmenting hTLR9 activation the 5′-CpG should be PD-based, and for mTLR9 activation the distant CpG should be PD-based. These could be due to the differences in the CpG recognition mechanism of hTLR9 and mTLR9 15,16 . The N-terminal region of the TLR9 ectodomain determines the sensitivity to the backbone chemistry of the first CpG motif. We showed that for the best activation of TLR9, the position of the PD-based CpG motif within the PTO-based H75 differs between hTLR9 and mTLR9 ( Fig. 1c,d vs. 1 g,h; Fig. 1e,f vs. Supplementary Fig. S1a,b). These results imply that the amino acid (aa) residues of the CpG binding site of TLR9 discriminate between the nonbridging oxygen and a sulfur atom in the phosphate bond, which motivated us to determine the region of hTLR9 that senses the backbone of the CpG motif. For that purpose, we prepared the TLR9 chimeras with different substituted segments of ECD (aa 26-201, 26-427, 26-503, 26-635, and 201-635) from h to mTLR9 and vice versa (detailed description of chimeras can be found as Supplementary Table S2).
The expression and localization of chimeras transiently transfected in HEK293 cells were similar to those of the wild type (wt-)TLR9 ( Supplementary Fig. S3a,b). The activation of the TLR9 chimera was tested using PTO-and PD-based mouse-and human-specific ODNs. Based on the results of the activation of the TLR9 chimera, they were classified into four groups. The chimeras h504-m-, h201-m635-h-, and m635-h-were inactive ( Supplementary Fig. S4a). To rule out the functional inactivation of inactive chimeras due to folding defects and to analyze whether the loss of function affects their ability to bind ODNs, we investigated the effect of the co-expression of mutants on wt-TLR9 activation ( Supplementary Fig. S4b). Inactive TLR9 chimeras that retain the ODN-binding ability should inhibit the activation of the wt-TLR9 due to the sequestration of ODNs, while folding defective mutants should have no effect. Indeed, all tested signaling-incompetent TLR9 chimeras were able to inhibit wt-TLR9 signaling. The second group comprises chimeras (h635-m-, m201-h635-m-, and m201-h-) whose response to H75 and M80 were as good or better than wt-hTLR9, regardless of the backbone chemistry of ODNs ( Supplementary Fig. S4a).
The third group of chimeras with the N-terminal part of hECD substituted with a mouse counterpart (m503-h-, m427-h-) could be activated with both PTO-and PD-based ODNs (Fig. 4a-c). The difference compared to wt-hTLR9 was that the chimeras were activated equally well by the PTO-based mouse-and human-specific ODNs. The wt-hTLR9 was only weakly activated by mouse-specific ODNs (M80 PTO , m1826 PTO ) 15 . The fourth group of TLR9 chimeras with the C-terminal part of hTLR9 ECD substituted with a mouse counterpart segment (h201-m-, h427-m-) were activated by PD-based ODNs but were signaling-incompetent for PTO-based ODNs (Fig. 4a,b). The chimeras apparently retained the ability to bind PTO-based ODNs, as they reduced the activation of the co-expressed wt-TLR9 with h2006 PTO (Supplementary Fig. S4b). These data imply that the C-terminal segment of hECD participates in sensing and discriminating the backbone chemistry of ODNs.
To further understand how the backbone chemistry of the CpG motif affects immuno-stimulation, the sensitivity of the first CpG motif to backbone chemistry was examined for the chimeras from the third and fourth groups. As expected, the responsiveness of the third group of chimeras to the mixed backbone chemistry of H75 was similar to that for the wt-hTLR9 (Fig. 4d,e), as these chimeras responded equally well to human-specific H75 PTO and mouse-specific M80 PTO . The chimeras (h201-m-, h427-m-) from the fourth group, which were signaling-incompetent to H75 PTO , gained responsiveness to H75 with the PD-based first CpG motif (H75 PTO -CG 1 PD ) (Fig. 4d,e). However, the H75 PD with the PTO-based first CpG motif (H75 PD -CG 1 PTO ) failed to activate these chimeras, even though the H75 PD activated receptors as effectively as wt-hTLR9. These data indicate that the chimeras with the N-terminal hECD are less responsive or not responsive to ODNs with the PTO-based first CpG motif, for example, H75 PTO , H75 PTO -CG 2 PD , and H75 PD -CG 1 PTO . This suggests that the properties of PD linkage within the CpG motif enable less restrictive activation of the immune response.
Identification of the amino acid residues within the binding site that sense the backbone chemistry. Having established that the chimeras with the N-terminal segment of hECD are biased in favor of ODNs with the PD-based first CpG motif, our next step was to corroborate whether the TLR9 binding site 12 participates in discriminating nonbinding oxygen over the sulfur atom in the phosphate backbone of agonists. Based on the structure of equine (Ec)TLR9 in combination with ODN1668_12nt 12 , we selected several aromatic and charged aa residues for point mutations (The EcTLR9 ODN1668 _12nt structure with highlighted positions of mutated residues are shown as Supplementary Fig. S5a). At the N-terminal site of the EcTLR9, the positive patch with K51 and R74 and aromatic residues W47, F49, and W96 were identified as important for binding the base moieties of ODN1668_12nt 12,23 and for the formation of a binding pocket with the C-terminus of the ECD TLR9 where aa residues mainly bind the backbone of the ligand. The aromatic residues W47, F49, and W96 at the N-terminal site were changed to alanine or tyrosine (W47 only). Positively charged H641 and K690 residues at the C-terminal site proximal to the N-site were changed to the negatively charged aspartic acid. Mutation W47A but not F49A moderately affected hTLR9 activity (Supplementary Fig. S5b). However, mutation W96A and mutations H641D and K690D led to the loss of TLR9 activation when stimulated with PTO-based ODNs. The expression level and localization of mutated hTLR9s were similar to those of wt ( Supplementary Fig. S5c,d). In addition, the signaling-incompetent mutants retained the ability to bind ODN PTO , determined through the inhibition of wt-TLR9 activation (Supplementary Fig. S5e). A binding assay revealed that the individual amino acid residue mutations are not sufficient to prevent ODN PTO binding, as no difference in the binding of biotinylated h2006 PTO between mutants and wt-TLR9 was observed ( Supplementary Fig. S5f).
All four selected mutants of hTLR9 when stimulated with H75 PTO , and its variants with the PTO-based first CpG motif (H75 PTO -CG 2 PD and H75 PD -CG 1 PTO ) failed to activate an NF-κB response (Fig. 5a,b). However, the substitution of the PTO-based backbone of the first CpG motif of H75 PTO with the PD-based backbone (H75 PTO -CG 1 PD ) restored the activation potency of ODNs for the mutants (Fig. 5a). Furthermore, ODNs with the PD-based first CpG motif (H75 PD , H75 PD -CG 2 PTO , and H75 PTO -CG 1 PD ) (Fig. 5b) activated mutants as effectively as wt-hTLR9. Overall, these data establish that the amino acid residues W47, W96, H641, and K690 of hTLR9 that correspond to the amino acid residues of the CpG binding site of EcTLR9 discriminate between the naturally occurring PD linkage of the CpG dinucleotide and the synthetic PTO linkage of the CpG dinucleotide.

Discussion
Our research focuses on evaluating the impact of a nonbinding atom in a phosphate bond of the CpG motif on the immunostimulatory potential of synthetic TLR9 agonists. We demonstrated that the backbone chemistry of the CpG dinucleotide of ODNs significantly affects TLR9 activation. Our results highlight that the CpG binding site of TLR9 distinguishes between the nonbridging oxygen and a sulfur atom of the CpG phosphate linkage, with a bias in favor of the oxygen atom. First, we found that the substitution of PTO with the PD-based 5′-end CpG motif significantly improved the hTLR9 activation efficiency of PTO-based human-specific ODNs. Furthermore, the PD-based CpG motif 6 nt separated from the 5′-end of mouse-specific ODN PTO improved the activation of mTLR9. With chimeric proteins between m-and hTLR9 and site-directed mutagenesis, we found that the N-and C-terminal parts of ECD forming a CpG-binding site sense the backbone chemistry. These data identify nonbridging oxygen of a phosphate bond within the CpG motif as superior to a sulfur atom and demonstrate that an improvement in ODN activation potency can be achieved with carefully positioned PD and PTO backbones within synthetic TLR9 agonists.
The replacement of nonbinding oxygen for a sulfur atom in a phosphate linkage renders oligonucleotides particularly nuclease-resistant 24 . Therefore, this conserved type of modification has been rapidly accepted in therapeutic applications due to the markedly enhanced immunostimulatory power 21 . However, the replacement of nonbridging oxygen by a sulfur atom changes the physical and chemical properties of the phosphate bond: increased Van der Waals radius (1.85 Å P = S vs. 1.44 Å P = O) with lengthening of the P-S bond (2.0 Å P = S vs. 1.5 Å P = O) and a changed partial charge distribution of the phosphate moiety 25,26 . This replacement also creates a chiral center ("Sp" or "Rp" conformation) at each phosphate bond, which generates multiple isomers of the oligo. These individual isomers have been shown to have different physical and chemical properties 27 ; therefore, it is not unexpected that the receptor response should be different. However, it is somewhat surprising that the effects of replacing PD with PTO are so localized. While our study highlights that the backbone chemistry of the CpG motif affects agonist potency to activate TLR9, it also raises the question of what the molecular mechanism of TLR9 is behind the favoring of nonbridging oxygen over sulfur in the phosphate linkage within particular CpG motifs. Based on the crystal structures of the EcTLR9 ectodomain in combination with 12 nt ODN1668 ligand 12 , the base moieties of the CpG motif are recognized by the lateral face of the N-terminus from LRRNT to LRR2, and LRR21 and LRR22 interact with the backbone of the CpG motif with the direct involvement of H641, H642, F667, and N694. Mutations of W47 and W96 at the N-terminus and of H641 and K690 at C-terminus resulted in reduced or abolished activation in response to PTO-based agonists. However, the ODNs with PD-based 5′-TCG (first CpG motif) effectively activated the hTLR9 mutants, indicating that phosphate bond chemistry significantly effects the binding of the unmethylated CpG motif to TLR9. We propose that smaller sized PD bond and partial charge distribution leading to a more positive phosphate atom and an overall increase in the negative charge density of other oxygen atoms 26 will make the interaction of a CpG dinucleotide with aa residues of the CpG binding pocket stronger and therefore less sensitive to point mutations within TLR9. It is difficult to completely exclude the impact of PTO bond chirality on the CpG motif-TLR9 binding mechanism, although it is interesting that a change as small as the replacement of a PTO linkage of a single CpG motif with a PD linkage of a single CpG motif significantly improves the activation potency of PTO-based ODNs.
Although the PD-based CpG motif within PTO-based ODNs improved the activation of TLR9, the optimal position of the PD-based CpG motif differs for h and mTLR9. We propose that the observed species-specific CpG position preferences originate from the fact that the sequence preferences of hTLR9 differ from those of mTLR9. It has become clear in recent years that the recognition of synthetic agonists by TLR9 is sequence-and SCientifiC RePoRtS | 7: 14598 | DOI:10.1038/s41598-017-15178-y species-dependent 11,15,16,18,20,[28][29][30] . We demonstrated previously that 5′-TCG trinucleotide is critical for the activation of hTLR9 16 . However, the CpG motif distant from the 5′-end critically affects the activation of mTLR9 15 . A mixed PTO-PD backbone sequence is characteristic for the A-class ODNs. The central PD-based palindrome with a single CpG motif is critical for TLR9 activation and the PTO-based poly G-tail at the 5′-and 3′-ends protects the ODN against exonuclease activity. For activation of TLR9, the A-class ODNs must first be cut with DNase II in endolysosomes 13 , where 11-12-mer fragments are generated. The backbone structure of the active 11-mer fragment resembles B-class ODNs with the PD-based TCG motif at the 5′-end and a PTO-based 3′-tail.
Replacing nonbridging oxygens with sulfur atoms changes the key properties of ODNs, making them more suitable for cell biology research applications and therapeutics. The PTO bonds integrated near the 5′-and 3′-ends protect the ODNs against exonucleases, and their incorporation throughout the ODN sequence protects them against the endonucleases but with some undesired consequences. PTO-modified ODNs trigger a strong multi-factor activation of nucleophiles and their adherence to collagen-coated surfaces irrespective of CpG content 31 . Moreover, PTO ODNs bind to platelets, leading to strong platelet CpG-independent activation 32 , and induce a CpG-independent respiratory burst, while PD ODNs do not 33 . However, this may be due to the much longer lifetime of PTO ODNs in the circulation. The binding site of the unmethylated CpG motif of TLR9 has been perfected during evolution to strongly detect naturally occurring agonists, making it insensitive to the effects of some point mutations. Therefore, PD-based synthetic agonists, particularly PD-based CG dinucleotides, might significantly improve the TLR9 activation efficiency of PTO-based ODNs.
Analyzing the effect of replacing PD to PTO linkage for the PD-based ODNs, the results are even more surprising. A phosphate linkage replacement of the 5′-end CpG motif with PTO, as expected, reduced the activation potency of the ODNs, while a PTO substitution of CpG motif distant from the 5′-end improved the activation of ODNs compared to PD-based ODNs with or without 3′-end protection. We propose that the CpG motif with PTO linkage is less favored for TLR9 binding, which results in binding only the 5′-end PD-based CpG motif and in improving TLR9 activation. It is important to realize that the PTO bond modification should not be incorporated at every available position in ODNs. For the optimal physiological effect, it should rather be introduced sparingly and with careful consideration of the impact each bond might have.
All appropriate measures were taken to minimize pain and discomfort in experimental animals collecting the femurs and tibiae. All of the procedures involving animals were performed according to the directives of the EU 2010/63 and were approved by the Administration of the Republic of Slovenia for Food Safety, Veterinary, and Plant Protection of the Ministry of Agriculture, Forestry, and Foods, Republic of Slovenia (Permit no. U34401-37/2015/5).

Human plasmacytoid dendritic cells, pDCs, and B-cells isolated from peripheral blood mononuclear cells (PBMCs).
The study was approved by the local Medical Ethics Committees (No. 142/10/13) and all methods were carried out in accordance with the relevant local regulatory requirements. A written informed consent was obtained from the blood donors. Human PBMCs were extracted from whole blood using the density gradient medium Lymphoprep and SepMate tubes (StemCell Technologies). Human plasmacytoid dendritic cells (pDCs) were isolated from PBMCs by the depletion of non-target cells using Plasmacytoid Dendritic Cell Isolation Kit II (Miltenyi Biotec, cat. no. 130-097-415). The LD column and QuadroMACS magnet were used for separation. The cells were seeded onto 96-well round-bottom plates (Corning) (0.25 × 10 5 cells/100 µl/well) in media containing RPMI1640, 10% (v/v) heat-inactivated FBS, 2 mM L-glutamine, penicillin (100 U/ml), streptomycin (0.1 mg/ml), and 50 mM 2-mercaptoethanol and were stimulated with ODNs (concentrations indicated on the graphs) for 20 h. The secreted hTNFα was determined using ELISA (human TNF alpha Ready-Set-Go ELISA, eBioscience).