Discussion

TLR9 recognizes bacterial and viral CpG DNA and initiates innate immune signaling. In this study, we focus specifically on ectodomain-level structural rearrangements associated with ligand recognition. These mechanistic features cannot be inferred from static structures alone and emerge only when the human TLR9 ectodomain is examined dynamically, highlighting the value of a simulation-based approach. Because our analysis relies on a homology model of human TLR9, some structural uncertainty is unavoidable. However, the unusually high sequence identity and conserved LRR architecture shared with mice and equine TLR9 substantially reduce this uncertainty. Accordingly, our conclusions emphasize relative conformational changes and interaction patterns, which are robust across independent simulations, rather than absolute atomic positions.

TLR9 is highly conserved in its interior, non-exposed residues, which likely contribute to overall structural stability. In contrast, residues at the dimer interface show lower conservation. Several cross-monomer interaction residues identified in our simulations such as H612, H641, and K690 are not conserved in mouse or equine TLR9 and therefore cannot be inferred from existing crystal structures. This underscores the need for human-specific modeling. The 70%–80% sequence identity between human TLR9 and its orthologs enabled reliable homology modeling, and the modeled TLR9_apo structure remained stably dimeric throughout the simulations. This behavior is consistent with mouse and equine TLR9, which also exist as prefusion dimers.21,24 DCCM analysis further supported this prefusion arrangement, revealing positive correlations between the N terminuses of the monomers. Although the apo dimer is stable, it exhibits substantial global motions, as reflected in both DCCM and PCA analyses. Previous structural and biochemical studies have shown that Z-loop cleavage is required for TLR9 activation21 and our simulations are consistent with this model. In the apo state, the Z loop remains solvent-exposed and flexible, with residues in LRRs 15–23 contributing to a stable dimer interface. These conserved LRR regions form a robust interaction surface dominated by positively charged and hydrophobic residues, supporting prior observations that electrostatic and hydrophobic complementarity stabilize TLR9 dimerization. Although the apo dimer is stable, DCCM and PCA analyses reveal substantial global motions, indicating that the ectodomain samples multiple conformational substates that may prime the receptor for ligand engagement.

When CpG ssDNA from organisms like viruses or bacteria binds to TLR9, the Z loop in the TLR9_CpG complex fluctuates approximately 2-fold more than in TLR9_apo, suggesting increased accessibility for proteolytic cleavage. CpG binding also induces downstream conformational changes: LRR18 and LRR19 become negatively correlated with the N-terminal regions of the monomers, while the N-terminal regions of TLR9/TLR9∗ remain positively correlated. This pattern suggests that LRR18 and LRR19 move apart to accommodate IM binding. Several residues in LRR19–20 (I587, N608, L610, W614) exhibit high CB despite lacking direct interactions, indicating potential functional relevance. Additionally, interactions present in the apo dimer (e.g., H641 and T642) are reassigned to the ssDNA ligands. CpG engages both the N-terminal LRRs of one monomer and the C-terminal LRRs of the opposing monomer, which likely contributes to the reduced anti-correlated motion observed between residues 100–300 of TLR9 and 700–800 of TLR9. LRR22 of one monomer also interacts with ssDNA bound to the partner monomer, stabilizing the complex. Because the number of TLR9-TLR9 contacts remains similar to the apo state, we propose that dimer interactions are redistributed to accommodate ligand binding.

Our findings align with mutational studies identifying Q346, R348, and Q562 as functionally important residues in human TLR9. Pohar et al.29 demonstrated that substitutions at R348 and Q562 impair signaling, highlighting their roles in ligand recognition. These residues correspond to R346, K348, and K563 in mouse TLR9 and to H346, K348, and R562 in equine TLR9, where they contribute to species-specific DNA preferences, specifically the substitution of Q346 in human TLR9 for R346 in mouse TLR9.24 Structural studies of equine TLR9 bound to CpG and 5′-xCx DNA show that H346 and K348 contact the third and fourth nucleotides of the 5′-xCx motif, while R562 forms part of the base of the 5′-xCx binding pocket.25 R562 in equine TLR9 lies within the loop region of LRR18 and forms part of the base of the 5′-xCx DNA-binding pocket. The residue projects toward the luminal face of the opposing protomer in the TLR9 dimer rather than directly contacting the 5′-xCx DNA itself. This positioning suggests that R562 (or Q562 in humans) may contribute to recognizing the 3′ extension of the DNA bound at the CpG-DNA binding site. Consistent with this idea, alanine substitution of Y345, F375, R377, F402, D534, Y536, and G560 in human TLR9 markedly reduces NF-κB activation in response to stimulation with the CpG oligodeoxynucleotide ODN10104.28,29 These observations support the functional relevance of the corresponding human residues and reinforce the mechanistic interpretations derived from our simulations.

TLR7, which shares mechanistic similarities with TLR9, adopts open and closed conformations with intermonomer distances of 6.93 nm and 3.65 nm, respectively.33 In contrast, we observe a modest 0.45 nm reduction in intermonomer distance between TLR9_apo and TLR9_CpG_IM. This difference is substantially smaller than the 3.2 nm shift reported for TLR7, likely reflecting the limited timescales accessible to atomistic simulations, which may not capture large-scale rearrangements in macromolecular receptors. Nevertheless, we detect subtle dynamic changes consistent with experimental findings. Many of these motions appear intrinsic to the receptor, with ligand binding suppressing inherent fluctuations and increasing rigidity. Most interactions observed in TLR9_CpG are preserved in the TLR9_CpG_IM complex. We observe that the binding of CpG and IM shifts several dimer-stabilizing interactions toward the ligands, resulting in a more stable dimer than in the apo form. The reduced intermonomer distance further supports ligand-induced stabilization of dimers. Together, these results support a stepwise activation mechanism in which CpG binding increases Z-loop flexibility, redistributes intermonomer contacts, and exposes the IM binding pocket.

The DNA IM used in this study represents a synthetic modulatory sequence whose relevance is supported by extensive structural and functional work. Equine and mouse TLR9 structures reveal a conserved 2:2 receptor-ligand arrangement, and human TLR9 activation requires two additional CpG motifs, indicating cooperative engagement of two DNA elements i.e., human TLR9 requires both CpG and 5′-xCx motifs for activation, whereas mouse TLR9 responds to CpG alone.26,27,29 Consistent with this, our results show 5′-xCx DNA enhances TLR9 dimerization in the presence of CpG DNA. Previous studies have shown that short 5′-xCx sequences can potentiate activation21,24 and more recent work shows that short 5′-xCx DNA potentiates TLR9 activation when co-present with CpG DNA.28 In this context, the IM provides one of the two required CpG motifs, with the second supplied by the ssDNA ligand. The resulting 2:2:2 configuration observed in our simulations aligns with known synthetic TLR9 modulators and experimentally supported stoichiometries. IM binding is stabilized by LRRs flanking the Z loop in both monomers, consistent with the opening of LRR18–19 upon CpG binding. The reduced Z-loop dynamics in TLR9_CpG_IM suggest that increased Z-loop flexibility in TLR9_CpG facilitates IM engagement. Elevated CB in LRR26 across apo and ligand-bound states indicates altered long-range communication within the ectodomain. Similarly, the Y180/K181 pair also shifts from weak inter-monomer contacts with E616∗ in the apo form to interactions that contribute to dimer-interface stabilization (Y180) and CpG stabilization (K181) in the ligand-bound states. Because our simulations are limited to the ectodomain, any linkage between these rearrangements and downstream signaling remains inferential and will require validation in full-length receptor systems.

While these findings highlight potential allosteric nodes within the human TLR9 ectodomain, they should be viewed as testable predictions rather than definitive mechanistic determinants. Experimental interrogation of these residues, particularly H612, H641, T642, and K690, will be essential for establishing their roles. Furthermore, mutations at H612 or K690 could weaken CpG stabilization despite their lack of contact in the apo structure and disruption of Y180–E616∗ or C507-mediated interface contacts could impair the CpG-induced rearrangement required for IM binding. Another testable hypothesis is the residues with large shifts in CB, such as T642 and K738, which are predicted to mediate long-range allosteric communication between CpG and IM binding sites. These predictions provide a framework for targeted mutagenesis studies in human TLR9. Overall, our results indicate that the TLR9 dimer is intrinsically stable and that LRRs flanking the Z loop play key roles in coordinating CpG ssDNA and IM binding.

Limitations of the study

This study modeled human TLR9 in its apo, CpG ssDNA-bound, and IM-bound forms using a homology model based on highly conserved TLR9 orthologs. While the model is strongly supported by high sequence identity and structural similarity, homology modeling inherently introduces uncertainty. As a result, our conclusions focus on relative conformational changes and interaction patterns, rather than absolute structural predictions. The key interactions we observe are consistent with experimentally resolved mouse and equine TLR9 structures, but human-specific nuances may not be fully captured. Our simulations support a 2:2 stoichiometry for TLR9-CpG ssDNA and a stable 2:2:2 arrangement for TLR9-CpG-IM, consistent with available crystal structures. However, we did not explore alternative stoichiometries or ligand orientations, which is a limitation of this study. Similarly, the diverse sequence specificity of CpG and IM ligands and the opposite IM orientation were not examined here. The rearrangement of interactions between monomers and ligands highlights residues important for CpG and IM recognition and stabilization. These findings suggest a mechanism in which CpG binding relaxes the LRR region around the Z loop, enabling IM engagement and promoting ligand-induced dimer stabilization. Nonetheless, our simulations capture only ectodomain-level rearrangements; the full multi-step activation process of TLR9, including TIR-domain reorganization, remains unresolved. Finally, this study lacks direct experimental validation. While the observed interactions align with known structural and biochemical data from multiple species, future experimental work will be essential to confirm the mechanistic insights proposed here.