The TLR9 ligand CpG ODN 2006 is a poor adjuvant for the induction of de novo CD8+ T-cell responses in vitro

Toll-like receptor 9 (TLR9) agonists have gained traction in recent years as potential adjuvants for the induction of adaptive immune responses. It has nonetheless remained unclear to what extent such ligands can facilitate the priming events that generate antigen-specific effector and/or memory CD8+ T-cell populations. We used an established in vitro model to prime naive precursors from human peripheral blood mononuclear cells in the presence of various adjuvants, including CpG ODN 2006, a synthetic oligonucleotide TLR9 ligand (TLR9L). Unexpectedly, we found that TLR9L induced a suboptimal inflammatory milieu and promoted the antigen-driven expansion and functional maturation of naive CD8+ T cells ineffectively compared with either ssRNA40 or 2′3′-cGAMP, which activate other pattern recognition receptors (PRRs). TLR9L also inhibited the priming efficacy of 2′3′-cGAMP. Collectively, these results suggest that TLR9L is unlikely to be a good candidate for the optimal induction of de novo CD8+ T-cell responses, in contrast to adjuvants that operate via discrete PRRs.

Scientific RepoRtS | (2020) 10:11620 | https://doi.org/10.1038/s41598-020-67704-0 www.nature.com/scientificreports/ by HLA-A*02:01 (abbreviated from hereon as HLA-A2), which are readily accessible in standard preparations of peripheral blood mononuclear cells (PBMCs) 12,13 . Healthy donors were recruited for this study to minimize the likelihood of naturally occurring memory responses to EV10. We found that lower frequencies of EV10-specific CD8 + T cells were generated after 10 days in the presence of TLR9L compared with other adjuvants, namely a standard cocktail of inflammatory cytokines, ssRNA40 (TLR8L), or the STING ligand 2′3′-cGAMP (Fig. 1A,B). A similar pattern was observed on day 7, excluding the possibility of early induction and subsequent cell death, and on day 15, excluding the possibility of late induction ( Supplementary Fig. S1A). Equivalent results were also obtained using a different neoantigen, HIV-1 Nef 138-147 (RF10), restricted by a different allotype, HLA-A*24:02 (Fig. 1C). Moreover, EV10-specific CD8 + T cells primed in the presence of TLR9L expressed lower levels of the cytolytic molecules granzyme B and perforin compared with EV10-specific CD8 + T cells primed in the presence of TLR8L or 2′3′-cGAMP (Fig. 1A,B). Recent studies have shown that low expression levels of T-bet and high expression levels of Eomes are associated with functionally impaired CD8 + T cells 14,15 . In line with these findings, EV10-specific CD8 + T cells primed in the presence of TLR9L exhibited lower T-bet/Eomes ratios compared with EV10-specific CD8 + T cells primed in the presence of TLR8L or 2′3′-cGAMP (Fig. 1A,B and Supplementary  Fig. S1B). All of these effects were abolished at lower concentrations of TLR9L ( Supplementary Fig. S2).
To assess the relationship between priming efficacy and the inflammatory milieu, we quantified various chemokines and cytokines secreted by PBMCs after overnight exposure to TLR9L, again using the comparators TLR8L and 2′3′-cGAMP ( Fig. 2A,B). TLR9L induced the secretion of several inflammatory factors, including interleukin (IL)-8, interferon (IFN)-γ-induced protein (IP-10), and monocyte chemoattractant protein (MCP)-1, at levels equivalent to or greater than those induced by TLR8L or 2′3′-cGAMP. In contrast, key effector molecules, namely IFN-γ, tumor necrosis factor (TNF), and granzyme B, were secreted in much lower amounts after stimulation with TLR9L versus stimulation with either TLR8L or 2′3′-cGAMP. This pattern was replicated for cytokines with costimulatory effects known to play important roles in priming events, such as IL-1α, IL-1β, and, more selectively, IL-12. In addition, higher levels of IL-1RA were induced by TLR9L versus 2′3′-cGAMP. This soluble factor can dampen the activation of antigen-specific T cells 16 . www.nature.com/scientificreports/ To extend these findings, we tested the adjuvant effects of TLR9L in combination with 2′3′-cGAMP. These experiments were predicated on earlier work in mice, which showed that TLR9 and STING agonists acted synergistically to enhance various innate and adaptive immune responses 17 . In line with the induction of a suboptimal inflammatory milieu and previous studies in various experimental systems [7][8][9][10][11] , but counter to the notion of cooperative activity, we found that TLR9L suppressed the expansion and functional maturation of EV10-specific and RF10-specific CD8 + T cells adjuvanted by 2′3′-cGAMP ( Fig. 3A-C). Although direct signaling via the STING pathway can inhibit cell proliferation, the priming efficacy of 2′3′-cGAMP relates primarily to the induction of type I IFNs via effects on DCs 13 . This latter process was likely impeded by TLR9L. We also found that TLR9L enhanced the production of IL-10 among otherwise subtle effects on 2′3′-cGAMP-induced patterns of chemokine/cytokine secretion in overnight assays with PBMCs (Fig. 3D). Of note, TLR9L has been shown to regulate plasmacytoid dendritic cell (pDC) responses to other immunostimulants via the induction of IL-10 18 , and high levels of IL-10 have been shown to inhibit the priming activity of 2′3′-cGAMP 13 . TLR9L can also upregulate the immunomodulatory enzyme indoleamine 2′3′-dioxygenase (IDO) 19,20 . However, our attempts to block these intermediaries using a purified anti-IL-10 monoclonal antibody or the IDO inhibitor D-1-methyltryptophan (D-1MT), respectively, failed to enhance the priming activity of TLR9L (Supplementary Fig. S3A).  www.nature.com/scientificreports/ To explore other potential mechanisms of suppression, we predepleted CD19 + cells, which include a population of DCs that acquire IDO-dependent regulatory functions in response to TLR9L 21 . In addition, we predepleted BDCA-3 + pDCs, which produce IFN-α in response to TLR9L 22 , and attempted to block the activity of IFN-α, which regulates many critical parameters involved in the genesis, maturation, and sustention of antigenspecific CD8 + T-cell populations 23 . None of these interventions enhanced the priming efficacy of TLR9L (data not shown and Supplementary Fig. S3B).
Collectively, these data show that TLR9L is a relatively poor adjuvant in vitro, at least compared with a standard cocktail of inflammatory cytokines, TLR8L, or 2′3′-cGAMP, all of which promoted the antigen-driven expansion and functional maturation of naive CD8 + T cells more effectively under otherwise identical conditions. TLR9L also inhibited the priming efficacy of 2′3′-cGAMP. This effect appeared to be independent of IDO, IFN-α, and IL-10. It should be noted that CpG ODN 2006 is a type B TLR9L. Further studies are therefore warranted to test the effects of type A and type C TLR9Ls, which are more potent inducers of IFN-α 3,4 . Moreover, our study was confined to the in vitro setting, which does not fully embrace the anatomical and physiological complexities of priming events in vivo, and further limited to a human system, which likely explains some of the discrepancies reported in mice 3 . Indirect effects may be especially pertinent in vivo. For example, TLR9L can stimulate B cells and CD4 + T cells to produce IgG2a and IFN-γ, respectively, which may promote the activation and expansion of CD8 + T cells 3,4 . In addition, we deliberately focused our investigations on the naive pool, excluding a formal evaluation of secondary responses, which may be preferentially amplified in the presence of TLR9L. Our results nonetheless suggest that CpG ODN 2006 is unlikely to be a good candidate for the optimal induction of de novo CD8 + T-cell responses, in contrast to TLR8L or 2′3′-cGAMP.

Materials and methods
Ethics. The use of human material was approved by the Comité de Protection des Personnes of the Pitié Salpétrière Hospital (France) and the Ethical Committee of Kumamoto University (Japan). Written informed consent was obtained from all donors in accordance with the Declaration of Helsinki.

Measurement of soluble factors.
Chemokines and cytokines were measured using a Luminex T200 instrument in combination with a human Bio-Plex Immunoassay Kit (Bio-Rad). All concentrations were determined as the mean of two replicates after background subtraction.
Statistics. Univariate statistical analyses were performed using nonparametric tests in Prism software version 8.4.2 (GraphPad). Significance was assigned at p < 0.05.