b.ii. pDC innate functions in tumor immunity

IFN-I is an important molecule for the induction of potent anti-tumor immune responses, although it has a dichotomous effects, since prolonged signaling of this molecule can lead to immune dysfunctions (Demoulin et al., 2013; Dunn et al., 2004; Minn, 2015; Snell et al., 2017;

Zitvogel et al., 2015). Several cancer immunotherapies are efficient only in the presence of IFN-I, which has been associated with favorable outcomes (Demoulin et al., 2013; Diamond et al., 2011;

Rautela et al., 2015; Wang et al., 2011; Zitvogel et al., 2015). As previously mentioned, pDCs are professional IFN-I-producing cells (Reizis et al., 2011a). The ability of TA-pDC to secrete IFN-I is impaired in many types of cancers (Demoulin et al., 2013; Hartmann et al., 2003; Labidi-Galy et al., 2011; Perrot et al., 2007; Sisirak et al., 2012; Terra et al., 2018). TA-pDC capacity to secrete other cytokines, including TNF-α, macrophage inflammatory protein (MIP)-1β, IL-6 and regulated on activation, normal T cell expressed and secreted (RANTES), was also found altered in tumors (Labidi-Galy et al., 2011). Furthermore, pDCs were shown to directly interact with MM cells, leading to the production of IL-3 and subsequent increased MM growth, which in return also stimulates pDC survival (Ray et al., 2017).

However, pDCs that are not immersed in the TME and are activated in vitro/ex vivo and injected directly into tumors retain the ability to induce antitumor responses (Liu, 2008). Moreover, pDCs can have a direct tumoricidal activity, they have the potential to directly induce the apoptosis of tumor cells (Drobits et al., 2012; Lombardi et al., 2015; Tel et al., 2014; Tel et al., 2012a).

III.1.b.iii. pDC antigen-presenting functions in tumor immunity

MHC-I and MHC-II antigen presentation by pDCs in the tumor context

Whether pDCs were shown to induce the differentiation or expansion of T cells with a regulatory phenotype, or the suppression of effector T cell proliferation, pDCs associated with tumors and their MHC-I- and MHC-II-mediated antigen-presenting functions have been associated with poor clinical outcomes (Lombardi et al., 2015).

In regard with MHC-I-mediated antigen-presenting functions, pDCs isolated from peripheral blood have been shown to have the ability to acquire membrane patches from cancer cells that harbored tumor antigenic peptide/MHC-I complexes, a mechanism resembling trogocytosis (Bonaccorsi et al., 2014). This occurred in a cell-cell contact-dependent manner and led to the presentation of exogenous antigens to CD8⁺ T cells. In addition, pDCs extracted from human colon cancer exhibited epithelial cell markers at their surface, suggesting this membrane transfer also happened in vivo (Bonaccorsi et al., 2014).

62

In human ovarian cancer, TA-pDCs have been suggested to promote IL-10-producing CD8⁺ Tregs (Wei et al., 2005). In addition, tumor-specific CD8⁺ T cells exhaustion has been correlated with high CD86⁺ pDC counts, in chronic myeloid leukemia (Schutz et al., 2017). The pDC expression of CD86, the ligand for CD28 (co-stimulatory) and CTLA-4 (co-inhibitory) molecules expressed by T cells, has been found to predict risk of disease recurrence after treatment (Schutz et al., 2017). Although it remains to be elucidated whether CD86⁺ pDCs migrate to the tumor and induce tumor-specific CD8⁺ T cell exhaustion, such a mechanism might occur, via CD86-CTLA-4 interaction.

Plasmacytoid DCs from the BM of MM patients express high levels of PD-L1, compared to normal BM pDCs (Ray et al., 2015; Ray et al., 2014). Co-culture of pDCs from MM patients with autologous CD4⁺ or CD8⁺ T cells, in the presence of anti-PD-1 blocking antibody, led to increased T cell proliferation, compared with controls in absence of the antibody, suggesting a mechanism of T cell suppression by pDCs via PD-1/PD-L1 interaction (Ray et al., 2015).

The accumulation of pDCs in tumors and TdLNs from melanoma patients and melanoma-bearing humanized mice is associated with a poor clinical outcome (Aspord et al., 2013, 2014a).

These cells were shown to promote Th2 cells, producing IL-5, IL-13 and TNF-α, in an OX40-L-dependent manner (Aspord et al., 2013, 2014a). Another study, in the context of different breast cancer types, showed tumor-derived supernatant-induced activation of pDCs isolated from peripheral blood mononuclear cells (PBMCs), due to GM-CSF contained in the supernatants, which induced the upregulation of CD80, CD86 and ICOS-L (Ghirelli et al., 2015). GM-CSF-activated pDCs primed naïve T cells, leading to the differentiation of Th2 cells with a regulatory phenotype, producing IL-4, IL-5, IL-10, IL-13 and TNF-α, but low levels of IFN-γ. In vivo, high levels of GM-CSF and pDC infiltration correlated with the most aggressive breast cancer type (Ghirelli et al., 2015). In addition, TA-pDCs from ovarian cancer patients induced the production of IL-10 by allogeneic CD4⁺ T cells in ex vivo cultures (Labidi-Galy et al., 2011).

Several studies, in human cancer and tumor mouse models, have shown an accumulation and colocalization of pDCs and CD4⁺ Tregs, correlating with poor prognoses (Le Mercier et al., 2013;

Sisirak et al., 2012). A study in human breast cancer, although it did not demonstrate a direct role for pDC antigen presentation, showed that the decreased production of IFN-I by TA-pDCs was implicated in the expansion of Tregs (Sisirak et al., 2012). Furthermore, Treg expansion in breast and ovarian cancers is dependent on the expression of the immunosuppressive molecule ICOS-L by pDCs (Conrad et al., 2012; Faget et al., 2012). In both cases, pDC-induced Treg expansion was associated with disease progression and poor outcome. Moreover, TA-pDCs from melanoma patients induce the differentiation of IL-10-producing Tregs, in an ICOS-L-dependent

63

manner (Aspord et al., 2013, 2014a). In addition, the presence of Tr1 cells (Foxp3^- IL-10⁺ IL-13^-) in hepatocellular carcinoma or liver metastases from colorectal cancer was linked with pDC infiltration (Pedroza-Gonzalez et al., 2015). Increased IL-10 secretion by Tr1 cells was dependent on the expression of ICOS-L by pDCs.

Finally, a subset of pDCs in the TdLNs of melanoma-bearing mice has been found to express the immunosuppressive molecule IDO, which led to the inhibition of T cell responses in vitro (Munn et al., 2004). In vivo, adoptive transfer of cells from TdLN led to T cell anergy, which was abrogated by IDO inhibitor administration. These IDO⁺ pDCs were subsequently demonstrated to directly activate Tregs in an IDO-dependent manner (Sharma et al., 2007). IDO⁺ pDCs have also been observed in the peripheral blood of late-stage melanoma patients (Chevolet et al., 2015). Whether they infiltrate the tumor and induce Treg expansion remains to be determined.

Harnessing the potential of antigen-presentation by pDCs to enhance anti-tumor immune responses

Despite TA-pDCs having immunosuppressive properties, the potential of antigen presentation by pDCs that are not immersed in the TME, such as pDC lines, pDCs from peripheral blood or pDCs in non-invaded distal LNs, can be harnessed in order to enhance anti-tumor responses (Aspord et al., 2012; Guery et al., 2014; Lombardi et al., 2015; Tel et al., 2013a; Tel et al., 2012b).

Aspord and colleagues used GEN2.2, a pDC line that harbor the MHC-I allele HLA-A*0201 (Aspord et al., 2010). Co-cultures of irradiated GEN2.2 loaded with tumor-derived antigens with PBMCs or tumor-infiltrating lymphocytes (TILs) from HLA-A*0201-matched melanoma patients led to enhanced CD8⁺ T cell degranulation and IFN-γ production and increased cytotoxicity towards patient tumor cells (Aspord et al., 2010; Aspord et al., 2012; Charles et al., 2018). The strategy of injecting tumor-derived peptide-pulsed GEN2.2 in vivo is now tested in clinical trial for late-stage melanoma patients⁴.

In addition, Tel et al. performed therapeutic vaccination using pDCs isolated from melanoma patient peripheral blood (Tel et al., 2013a). Autologous pDCs were activated ex vivo by tick-borne encephalitis virus vaccine and loaded with tumor-antigenic peptides, inducing IFN-I secretion and upregulation of MHC-I, MHC-II, CD80, CD83 and CD86 expression. These cells were subsequently injected intranodally into patients, leading to enhanced CD4⁺ and CD8⁺ T cell responses (Tel et al., 2013a). Furthermore, in tumor mouse model, antigen delivery to pDCs via BST2, along with TLR agonists, led to antigen-presentation by pDCs in vivo, inducing a strong anti-tumoral response and tumor growth inhibition (Loschko et al., 2011b). Finally, Liu et al.,

4 http://pdc-line-pharma.com/clinical-trials/

64

injected ex vivo TLR9L (CpG-A)-activated pDCs in B16 melanoma-bearing mice, directly into the tumor, which led to regression of injected and untreated tumors (Liu et al., 2008a).

In an attempt to study a potential MHC-II-mediated antigen-presenting functions of pDCs in anti-tumor immunity, our group previously performed vaccination at a distal site from the tumor, i.e. subcutaneous injection in the right flank, while the tumor was established in the left flank (Guery et al., 2014). In this study, we used the TLR9 agonist CpG-B, along with a MHC-II-restricted peptide. Indeed, as mentioned previously, there are distinct classes of CpG, with different outcomes on pDCs (Gilliet et al., 2008). CpG-A forms aggregates that bind to TLR9 in the early endosome, leading to the production of IFN-I and IFN-III via IRF7 activation (Gilliet et al., 2008) (Fig. 14). In contrast, CpG-B, which exists in a monomeric form, traffics to the late endosome/lysosome, inducing the secretion of pro-inflammatory cytokines, such as IL-6 and TNF-α, and to the upregulation of the expression of genes involved in antigen presentation, such as MHC-II and the co-stimulatory molecules CD40, CD80 and CD86 (Gilliet et al., 2008) (Fig.

14). Finally, the binding of CpG-C on TLR9 leads to the expression of genes activated by CpG-A and of CpG-B-activated genes (Fig. 14) (Gilliet et al., 2008; Vollmer and Krieg, 2009). Therefore, CpG-B appeared to be the most appropriate class of CpG to analyze pDC antigen-presenting functions.

Following vaccination with CpG-B and MHC-II-resticted peptide, pDCs in distal LNs acquired an immunogenic phenotype and primed Th17 cells (Fig. 15) (Guery et al., 2014). Indeed, the use of mice in which MHC-II expression was selectively abrogated in pDCs allowed our group to conclude that Th17 differentiation was dependent, in this context, on MHC-II-mediated antigen-presenting functions of pDCs (Guery et al., 2014). Th17 cells subsequently migrated into the tumor, which led to immune cell infiltration, including CTLs, promoting tumor cell killing and inhibition of tumor growth (Fig. 15) (Guery et al., 2014). Our study, therefore, highlighted an anti-tumor role of pDCs, when activated at a distal site from the tumor, through their MHC-II-mediated antigen-presenting functions.

65 Figure 14. Signaling pathways of type A and B CpG oligodeoxynucleotides in early and late endosomes in plasmacytoid dendritic cells.

Upper panel. In pDCs, the binding of CpG-A, which is aggregated, to Toll-like receptor (TLR)-

9 takes place in the early endosomal compartment in which markers, including early endosomal antigen 1 (EEA1) and transferrin receptor (TfR), are expressed. The prolonged interaction between CpG-A and TLR9 leads to the activation of myeloid differentiation primary response gene 88 (MyD88) and subsequently to interferon-regulatory factor 7 (IRF7), promoting a strong secretion of type I interferon (IFN-I).

Lower panel. On the contrary, CpG-B (monomeric) bound to TLR9 traffic rapidly through the early endosome and enter the acidified late endosome/lysosome, in which LysoTracker and lysosomal-associated membrane protein 1 (LAMP1) are expressed. This leads to the activation of genes different from the ones activated by CpG-A, such as nuclear factor-κ B (NF-κB), mitogen-activated protein kinases (MAPK) and IRF5. It induces a different outcome for pDCs, with the upregulation of co-stimulatory molecules including CD80, CD40 and CD86, the production of pro-inflammatory cytokines such as tumor necrosis factor (TNF) and interleukin (IL)-6, and chemokines.

BTK, Bruton's tyrosine kinase; IRAK, interleukin-1-receptor-associated kinase; OPN, osteopontin; TAK1, transforming-growth-factor-β-activated kinase 1; TRAF, tumor-necrosis factor (TNF) receptor-associated factor.

Adapted from Gilliet*, Cao* & Liu, Nat Rev Immunol, 2008 (Gilliet et al., 2008).

66 Figure 15. Priming of Th17 cells by activated antigen-presenting plasmacytoid dendritic cells leads to tumor growth control: working model.

1. pDCs in tumor and tumor-draining lymph nodes (TdLNs) are rendered tolerogenic (purple) by the tumor microenvironment.

2. Nonetheless, tumor-bearing mice vaccination with CpG-B together with a MHC-II-restricted tumor antigenic peptide at a distal site from the tumor leads to the priming of Th17 cells by activated antigen-presenting pDCs (red).

3. Th17 cells migrate towards the tumor and induce immune cell infiltration into the tumor, including conventional DCs (cDCs), Th1 cells, regulatory T cells (Tregs) and cytotoxic T lymphocytes (CTLs).

4. Increased tumor antigen-specific CTL infiltration promotes tumor cell killing and tumor growth inhibition.

Adapted from Guéry and Hugues, Oncoimmunology, 2015 (Guery and Hugues, 2015a), using data from Guéry et al., Cancer Res, 2014 (Guery et al., 2014).

Using TLR ligands to reprogram tolerogenic TA-pDCs

Whether tolerogenic TA-pDCs can be reprogrammed using TLR agonists, in order to enhance their immunogenicity, is still a matter of debate. Topical administration of imiquimod (TLR7L) on skin neoplasms led to the infiltration of IFN-I-producing pDCs in lesions, inducing an important tumor regression (Stary et al., 2007; Urosevic et al., 2005). In addition, administration of imiquimod in a melanomal model of humanized mice induced activation of pDCs, with

67

increased cytotoxic functions and IFN-I production, and subsequent inhibition of tumor growth and metastasis (Aspord et al., 2014c). In another model of melanoma, topical imiquimod administration induced the recruitment of pDCs with direct tumoricidal activity thanks to granzyme B and TRAIL, leading to tumor growth control (Drobits et al., 2012). Finally, using an orthotopic mouse model of breast cancer, TLR7L intratumoral delivery led to a reprogramming of tolerogenic pDCs and to inhibition of tumor growth, which was dependent on IFN-I signaling (Le Mercier et al., 2013).