Human MDDCs do not fully mature when exposed to HTLV-1-infected T cells
To determine whether chronically infected T cells could directly or indirectly manipulate human DC functions in vitro, we first addressed whether infected T cells were sensed by DCs. We used human monocyte-derived DCs (MDDCs) cultured in vitro with HTLV-1-infected and transformed T cell lines, and assessed MDDCs maturation by flow cytometry after 24h or 48h of coculture, using the MDDC-specific CD11c marker to identify MDDCs in the coculture (see Fig. 1a for the experimental settings, and Supplementary Fig. 1 for the gating strategy). At these time points, less than 1% of MDDCs have integrated the viral genome and they do not release virions11, indicating that they are not productively infected. In contrast to MDDCs exposed to measles virus (MeV), MDDCs exposed to HTLV-1 transformed C91-PL cells, failed to fully upregulate CD86, used in this first experiment as a surrogate of DC maturation (Fig. 1b, left panel). Repeated experiments using MDDC samples from independent donors showed that while around 90% of MDDCs upregulated CD86 upon exposure to MeV, regardless of their ability to express MeV (Fig. 1b, right panel, compare total versus GFP− and GFP+ MDDCs), as well as upon stimulation by the TLR-4 agonist LPS (Fig. 1b, right panel), the percentage of activated MDDCs after coculture with HTLV-1-transformed C91-PL cells remained low (Fig. 1b, right panel). To exclude a cell line-specific effect, and to thoroughly characterize MDDCs maturation status, MDDCs were exposed to several control uninfected T cell lines (Supplementary Fig. 2a-f; Jurkat, CEM or Molt-4 T cell lines, green bars), as well as to several HTLV-1-transformed T cell lines (Supplementary Fig. 2a-f; C91-PL, MT-2 or Hut102; blue bars), and the regulation of both maturation markers CD86, CD83, CD80, and CD40, (Supplementary Fig. 2a-d) and inhibition markers ICOSL and PD-L1 (Supplementary Fig. 2e-f) at the cell surface was monitored by flow cytometry. All the HTLV-1-transformed T cell lines induced either no change, or only minor changes, in the expression (percentage of positive cells and normalized mean signal intensity, MFI) of maturation or inhibition markers at the surface of cocultured MDDCs (Supplementary Fig. 2a-f). Increasing the duration of coculture to 48h did not impact these observations (Supplementary Fig. 2a-f), excluding a delay in maturation. The inefficient upregulation of MDDC maturation markers was also accompanied by a low ability to secrete TNF-α or IFN-I (Fig. 1c and Supplementary Fig. 2g), strengthening the notion that HTLV-1-infected T cells do not induce a typical maturation program in MDDCs, and thus might not be efficiently sensed by human DCs, in contrast to observations made with cell-free virus in a mouse model15. Since similar observations were made with all HTLV-1-infected T cell lines tested, and with the tested maturation markers and cytokines, only the C91-PL T cell line and CD86 were used in some of the following experiments.
Previous work from our laboratory showed that MDDCs efficiently capture HTLV-1 after coculture with HTLV-1-infected T cell lines11, suggesting that in our experimental conditions, HTLV-1 capture itself is inefficient in driving MDDC maturation. To formally demonstrate this, we assessed HTLV-1 capture in MDDCs following coculture, by staining the HTLV-1 Gag p19 structural protein. As expected, MDDCs were efficient at capturing HTLV-1 (Fig. 1d), irrespective of the infected T cell line used in the coculture (Supplementary Fig. 2h), with around 30% of capture observed in a representative coculture experiment with C91-PL cells (Fig. 1d, left), and reaching up to 80% of capture observed in repeated experiments using independent donors (Fig. 1d, right). Maturation was then compared in MDDCs that had not captured HTLV-1 (Gag p19-negative MDDCs, cyan, Fig. 1e), or had captured HTLV-1 (Gag p19-positive MDDCs, magenta, Fig. 1e). Although CD86 expression was consistently higher in p19-positive MDDCs (Fig. 1e, right, compare cyan and magenta bars), the percentage of Gag p19-positive cells remained low compared to fully matured LPS-exposed MDDCs (Fig. 1e, compare magenta and red bars), highlighting that maturation remained inefficient even in MDDCs that had capture HTLV-1. No significant correlation was detected among repeated experiments between the level of HTLV-1 capture and the efficiency of MDDC maturation (Supplementary Fig. 2i, p = 0.1412), further confirming that HTLV-1 capture itself is inefficient in driving MDDC maturation.
Taken together, these observations, recapitulated in Fig. 1f, demonstrate that cell-associated HTLV-1 fails to fully mature MDDCs, suggesting that it is poorly sensed by these human innate cells.
Exposure to HTLV-1-infected T cells induces a unique transcriptional signature in MDDCs
Interestingly, our previous work demonstrated that in contrast to MDDCs, plasmacytoid dendritic cells (pDC) do efficiently sense cell-associated HTLV-120. Inefficient sensing of HTLV-1-infected T cells by MDDCs could be the result of either a stealth behavior of cell-associated virus towards MDDCs, resulting in a bona fide lack of response specifically in this cell type, or, alternatively, to the active manipulation of MDDC functions by HTLV-1-infected T cells, resulting in a specific response that differs from a typical maturation program. To discriminate between both these scenari, we aimed at determining the complete transcriptomic landscape of MDDCs after exposure to HTLV-1-infected T cells, or to control uninfected T cells, using bulk RNA-seq (Fig. 2). After 24h of coculture, MDDCs were magnetically separated from C91-PL or control Jurkat T cells, based on the exclusive expression of CADM1 on T cells (Supplementary Fig. 3a). Note that although the levels of CADM1 were higher at the surface of C91-PL cells compared to Jurkat cells (Supplementary Fig. 3a), a similar MDDC enrichment yield of around 99% was observed after separation in both conditions, with no significant T cell contamination (Supplementary Fig. 3b). LPS-stimulated MDDCs were also included in the RNA-seq analysis, as a reference for fully matured MDDCs. RNA was extracted from each MDDC sample, and submitted to quality control followed by sequencing.
First, to show global similarities in gene expression between samples, unsupervised hierarchical clustering and principal component analysis (PCA) were performed on normalized transcript count tables. Analysis of sample clustering along the two first PCs (Fig. 2a, PC1 and PC2), that explain 81 and 8% of the total variance, respectively, showed that the main source of variance in the dataset was the stimulation by LPS, followed by the infection status of cocultured T cells, while the identity of the MDDC donor contributed only weakly to the global variance. In agreement with flow cytometry data shown in Fig. 1, this confirmed, at the transcriptional level, the inefficient activation of MDDCs after exposure to HTLV-1-infected T cells, which contrasts with the typical maturation program of fully activated MDDCs. This also indicates that exposure to HTLV-1-infected T cells does induce a detectable, yet subtle, transcriptional response, when compared to exposure to uninfected T cells.
To characterize the changes in gene expression in MDDCs induced by exposure to HTLV-1-infected T cells, we used DESeq2 to identify differentially expressed genes (DEGs) between C91-PL- and Jurkat-exposed MDDCs. As a reference for a typical maturation program, we identified DEGs between LPS- and Jurkat-exposed MDDCs. Using a fold-change cut-off of 2, a total number of 474 DEGs were obtained between C91-PL- and Jurkat-exposed MDDCs (padj < 0.05), with 435 genes found significantly upregulated upon HTLV-1 exposure, and 39 genes found significantly downregulated (Fig. 2b and Supplementary Table 1). A heatmap of the 474 DEGs showed clustering between samples, as expected (Supplementary Fig. 4a).
To gain insight into the processes modulated in MDDCs after exposure to HTLV-1-infected T cells, we performed gene ontology analysis on the set of 435 upregulated genes. Over-represented KEGG pathways were retrieved and filtered to reduce redundancy (see Supplementary Table 2), leading to a final list of 7 pathways (Fig. 2c and Supplementary Tables 2 and 3). Strikingly, exposure to HTLV-1-infected T cells was characterized by significant over-representation of upregulated genes involved in lipid biosynthesis and metabolism (Fig. 2c). Interestingly, genes involved in several pathways linked with response to viral infection were also over-represented, including genes involved in innate immune sensing (sub-pathways: RIG-I or Toll-like receptors and NOD-like receptors), in viral infection (sub-pathways: Influenza A, Measles, Hepatitis C, Hepatitis B, Human Papillomavirus, Herpes simplex virus 1, Coronavirus, HIV, EBV), and in the NF-κB signaling pathway. Taken together, these results confirm that exposure to HTLV-1-infected T cells does induce a transcriptional response in MDDCs, indicating that sensing does occur to some extent, but does not culminate in a typical maturation and anti-viral program.
To further determine the extent to which this transcriptional response is distinct from a typical maturation signature, we compared the genes affected by LPS stimulation of Jurkat-exposed MDDC to those affected by MDDC exposure to C91-PL. Among the 474 DEGs, 373 genes were also differentially expressed after LPS stimulation (Fig. 2d), representing only 5% of all LPS-modulated genes. Among these shared genes, 7 genes annotated as involved in the NF-κB signaling pathway were retrieved, including CCL19, TNFSF13B, TRIM25, BCL2L1 and BCL2A1 (Supplementary Fig. 4b). However, the level of up-regulation of these shared NF-κB-related genes was strikingly lower after HTLV-1 exposure compared to LPS stimulation (Supplementary Fig. 4b). This observation of a lower magnitude of differential expression (either up- or downregulation) was general to most of the 373 shared DEGs (Fig. 2e). Of note, the RELA, RELB and NFKB1 genes that were upregulated by LPS stimulation were not upregulated upon C91-PL exposure (Supplementary Fig. 4c), possibly contributing to the lower activation of the NF-κB-dependent genes observed above (Supplementary Fig. 4b). In addition, among the 377 Interferon-Stimulated Genes (ISG) listed in Supplementary Table 1, only 121 were retrieved in the set of significantly upregulated genes, and again, their up-regulation was lower to that induced by LPS stimulation (Supplementary Fig. 4d). These observations strengthen the notion that while sensing of HTLV-1-infected T cells does occur to some extent, the magnitude of the maturation and antiviral transcriptional response remains very limited, which is consistent with the inefficient maturation of MDDCs observed at the protein level (see Fig. 1).
Interestingly, among the 474 DEGs, 101 genes were not present in the list of DEGs between LPS- and Jurkat-exposed MDDCs (Fig. 2d and f), defining a unique transcriptional signature associated with the exposure to HTLV-1-infected C91PL cells. Due to the limited number of genes included in this specific signature, gene ontology analysis was not feasible. Interestingly however, several genes retrieved in Fig. 2c as those involved in lipid biosynthesis and metabolism, were also included in this specific signature, such as ELOVL3 (Elongation of Very Long Fatty Acid Elongase 3), FADS1 (Fatty Acid Desaturase 1), and SLC27A6 (Solute Carrier Family 27 Member 6, a member of the fatty acid transport protein family) (Fig. 2g). Altogether, these transcriptomic analyses demonstrate that exposure to HTLV-1-infected T cells is not completely silent in MDDCs, but rather results in a unique transcriptional response that differs from a typical maturation program. This supports the notion that HTLV-1-infected T cells might actively manipulate MDDC functions to limit their functional response, possibly by rewiring lipid biosynthesis and metabolism.
Pre-exposure to HTLV-1-infected T cells dampens the responsiveness of MDDCs to subsequent stimulation
We next aimed at investigating whether this unique transcriptional response observed upon exposure to HTLV-1-infected T cells indeed translates into functional defects in MDDCs, beyond their inefficient maturation. To this end, we addressed whether pre-exposure to HTLV-1-infected cells influenced MDDC responsiveness when exposed to a subsequent stimulation. After 24h of coculture with HTLV-1-infected T cell lines, MDDCs were restimulated with strong inducers of MDDCs maturation, in the form of LPS (TLR-4 ligand) or R848 (TLR-7/8 ligand, Fig. 3a), for another 24h. When compared to MDDCs pre-exposed to uninfected control T cell lines (red bars), MDDCs pre-exposed to HTLV-1-transformed T cells (dark blue bars) upregulated CD86 (as well as other maturation markers) to a significantly lower extent when stimulated by LPS or R848 (Fig. 3b, Supplementary Fig. 5a-d). In addition, TNF-α secretion by MDDCs upon LPS stimulation was also significantly reduced by pre-exposure to HTLV-1-transformed T cells (Fig. 3c, left panel, Supplementary Fig. 5e), while IFN-I secretion was not affected (Fig. 3c, right panel, Supplementary Fig. 5e). These results indicate that pre-exposure to HTLV-1-transformed T cells indeed influences the responsiveness of MDDCs to subsequent stimulation, by specifically dampening their pro-inflammatory response (monitored here through the upregulation of maturation markers and through TNF-α secretion), without hampering their antiviral response (monitored here through IFN-I secretion) (recapitulated in Fig. 3d). This suggests that HTLV-1 might specifically manipulate the responsiveness of certain signaling pathways in MDDCs, such as the NF-κB pathway upstream of the pro-inflammatory program, while leaving the responsiveness of others unaffected, such as the IRF3 pathway upstream of the antiviral program.
Pre-exposure to HTLV-1-transformed T cells alters the transcriptional response of MDDCs to subsequent stimulation
To obtain a broader overview of how pre-exposure to HTLV-1-transformed T cells affects the responsiveness of MDDCs, a second bulk RNA-seq analysis was conducted on MDDCs pre-exposed to an HTLV-1-infected T cell line (C91-PL), or to a control uninfected T cell line (Jurkat), and then restimulated by LPS (Fig. 4). To address whether viral capture was required for this manipulation of MDDC responsiveness, we also included in the RNA-seq analysis MDDCs cocultured with the HTLV-1-infected C8166 T cell line that does not produce viral particles21. As expected, no viral particle was captured by C8166-exposed MDDCs, in contrast to C91-PL-exposed MDDCs (Supplementary Fig. 6).
PCA analysis of samples showed that both the donor and the infection status of cocultured T cells were the main sources of variance in the dataset (Fig. 4a), confirming that pre-exposure to HTLV-1-transformed T cells does alter the transcriptional response of MDDCs to subsequent stimulation. In contrast, the ability to capture HTLV-1 did not contribute to a visible extent to the global variance, suggesting that viral capture might not be required for this alteration.
We then retrieved the lists of genes differentially expressed between the different experimental conditions. In agreement with the PCA analysis, 1324 genes were found differentially expressed between LPS-stimulated C91-PL-pre-exposed MDDCs, and LPS-stimulated Jurkat-pre-exposed MDDCs (Supplementary Table 4), with a total of 403 DEGs being downregulated and 921 DEGs being upregulated (Fig. 4b). Gene ontology analysis was conducted on these sets of downregulated and upregulated genes, respectively (Fig. 4c, left and right panels, respectively). The set of downregulated genes was characterized by a significant over-representation of genes annotated as being involved in TNF-α or NF-κB signaling and innate immune sensing (Fig. 4c, left panel, arrows, see Supplemental Table 5). Conversely, the set of upregulated genes was characterized by a significant over-representation of genes annotated as being involved in the TGF-β signaling pathway (Fig. 4c, right panel, arrow, see Supplemental Table 6). This confirmed that pre-exposure to HTLV-1-transformed T cell does influence the transcriptional response to LPS stimulation.
To further identify the expression patterns of these genes differentially expressed between LPS-stimulated C91-PL-pre-exposed MDDCs, and LPS-stimulated Jurkat-pre-exposed MDDCs, we classified these genes based on their differential expression across conditions (detailed in Supplemental Fig. 7a, see also Supplementary Table 4) as follows: (i) Genes whose responsiveness to LPS (be it repression or induction) is specifically conferred by pre-exposure to infected cells (Fig. 4d, upper panel). (ii) Genes whose responsiveness to LPS (be it induction or repression) is exacerbated by pre-exposure to infected cells (Fig. 4d, middle panel). (iii) Genes whose responsiveness to LPS is attenuated or (iv) abolished by pre-exposure to infected cells (Fig. 4d, lower panel). This classification demonstrates that pre-exposure to HTLV-1-transformed T cells induces both a change in the identity of genes that transcriptionally respond to LPS stimulation, defining a unique LPS-induced transcriptional signature; and a change in the magnitude of the transcriptional response of genes that are normally responsive to LPS.
In line with the gene ontology analysis presented in Fig. 4c, the responsiveness of genes involved in NF-κB signaling, such as MYD88 (Fig. 4d, lower panel, see graph), or the pro-inflammatory genes CCL2, IL12B, TNF, CXCL10 and CXCL11 (Supplementary Fig. 7b, upper panel), was found drastically attenuated by pre-exposure to infected cells (C91-PL or C8166). This could account for the inefficient maturation and production of pro-inflammatory cytokines by HTLV-1-pre-exposed MDDCs after LPS stimulation, which was observed in Fig. 3. In addition, in line with the gene ontology analysis presented in Fig. 4c, genes of the TGF-β signaling pathway, including BMP6, BMP7, and IL13 (Supplementary Fig. 7b, middle panel), which do not respond to LPS in normal conditions, were found to be responsive to LPS when MDDCs were pre-exposed to HTLV-1-transformed cells, while the responsiveness of TGFB2, which is repressed upon LPS stimulation in normal conditions, was abolished by pre-exposure to HTLV-1-infected cells (Fig. 4d, lower panel, see graph). Interestingly, these genes encode cytokines known to participate in the induction of a tolerogenic immune microenvironment, with Treg and TH2 responses, which are inefficient at controlling viral infection. Finally, pre-exposure to HTLV-1-transformed cells also conferred responsiveness to DDIT4 (Fig. 4d, upper panel, see graph), a gene reported in other contexts of tolerogenicity22. Of note, and in agreement with the efficient induction of IFN-I after LPS stimulation in both Jurkat- or C91PL-pre-exposed MDDCs (see Fig. 3c), the responsiveness of ISGs such as ISG15, IFI44 and IRF2 was not affected by pre-exposure to HTLV-1-infected cells (Supplementary Fig. 7b, lower panel).
Altogether, this transcriptomic analysis indicated that pre-exposure to HTLV-1-transformed T cells influences the responsiveness of MDDCs to subsequent stimulation, by specifically dampening their pro-inflammatory response at the transcriptional level. In addition, it uncovered the fact that pre-exposure to HTLV-1-transformed T cells allows a unique set of genes to respond to LPS stimulation, triggering a biased, pro-tolerogenic response of MDDCs upon subsequent stimulation.
Neither HTLV-1 viral capture, nor cell-cell contact with infected T cells, are strictly required to dampen the responsiveness of MDDCs to subsequent stimulation
As stated above, RNA-seq analysis using the C8166 infected cell line suggests that viral capture might not be required to alter the transcriptional response of MDDCs to a secondary stimulation. To confirm this notion at the functional level, we repeated the coculture experiment followed by LPS stimulation, and analyzed MDDC maturation profile by flow cytometry (Fig. 5a). Despite the lack of viral particle capture by C8166-exposed MDDCs (see Supplementary Fig. 6), C8166 pre-exposure still dampened the responsiveness of MDDCs to LPS and R848 stimulation, as monitored through CD86, CD83 or CD80 upregulation, similar to C91-PL pre-exposure (Fig. 5a and Supplemental Fig. 8a), confirming that viral capture is not strictly required. Although C8166 do not produce viral particles, they might still engage in cell-cell contacts with MDDCs23, which could be the trigger of the dampening of MDDC responsiveness. We addressed this hypothesis by performing the cocultures in transwells, in which MDDCs were physically separated from infected T cells by a permeable membrane (Fig. 5b). Of note, reduced but detectable levels of viral capture were detected in MDDCs physically separated from C91-PL (Supplementary Fig. 8b), most probably because the virus is preferentially associated to the cell surface of infected cells and might be released as large adhesive viral aggregates24 poorly able to cross the 0.4µm pores of the permeable membrane. The absence of physical contact between MDDCs and infected T cells, either producing (C91-PL) or not producing viral particles (C8166), did not restore a fully efficient maturation after LPS stimulation (Fig. 5b), indicating that cell-cell contacts are not strictly required to allow HTLV-1-infected T cells to manipulate MDDC responsiveness. Of note however, C8166 cells appeared less efficient than C91PL cells in dampening MDDC responsiveness upon transwell coculture. This could result from additive effects of mechanisms dependent on viral capture on the one hand, and of cell-cell contact on the other hand: indeed, in the presence of cell-cell contacts (Fig. 5a), the contribution of viral capture might be negligible; while in the absence of cell-cell contacts (Fig. 5b), such contribution might become relatively more important. Consistently, comparison of MDDC responsiveness in paired experiments of MDDC pre-exposed to infected cells with (coculture) or without (transwell) physical contacts (Fig. 5c) showed that preventing physical contacts between MDDCs and infected cells resulted in a slightly higher, yet not significantly different, MDDC responsiveness (Fig. 5c). This indicates that viral capture and/or cell-cell contacts may participate, but are not strictly required to allow HTLV-1-infected T cells to manipulate MDDC responsiveness, and suggests the contribution of a distantly acting set of soluble mediators.
A molecular dialogue between HTLV-1-infected T cells and MDDCs induces the release of soluble tolerogenic mediators
Since neither viral particles nor cell-cell contacts are strictly required to manipulate MDDC responsiveness, we then tested the conditioning ability of the supernatant of HTLV-1-infected T cells. More specifically, we hypothesized that IL-10 produced by HTLV-infected T cells25 could be a candidate mediator, as it was reported to have tolerogenic properties26. However, except for the supernatant of the MT-2 infected T cell line, IL-10 was not detected in the supernatant of any of the other infected T cell lines (Supplementary Fig. 9a). More surprisingly, none of these supernatants were sufficient to manipulate MDDC responsiveness, as MDDCs pre-incubated in these supernatants still efficiently matured upon LPS stimulation (Supplementary Fig. 9b). This suggested that MDDC manipulation requires a molecular dialogue between HTLV-1-transformed T cells and MDDCs that would lead to the production of soluble tolerogenic mediators upon coculture. To test this hypothesis, we collected the conditioned medium of MDDCs cocultured with infected T cells for 24h, and used it to culture fresh, naïve MDDCs derived from the same monocyte donor (autologous cells) for 24h, before LPS stimulation for another 24h (see Fig. 6a for the experimental settings). In contrast to supernatant of infected T cells alone (Supplementary Fig. 9b), the supernatant from the co-culture was still able to dampen the responsiveness of MDDCs (Fig. 6b), although with less potency compared to coculture with infected T cells, as observed by analysis of paired experiments (Fig. 6c). This is consistent with the tendency observed upon transwell cultures (see Fig. 5c), and confirms the release of soluble tolerogenic mediators following a molecular dialogue between infected T cells and MDDCs upon coculture, which could cooperate with other mechanisms dependent on viral capture and/or cell-cell contacts.
Next, we addressed the kinetics of the molecular dialogue required for the release of these mediators. Conditioned medium of MDDCs cocultured with infected T cells were collected over time, and used to culture fresh autologous MDDCs before the addition of LPS (Fig. 6d). A lowered responsiveness of MDDCs was only observed with supernatants collected at least 18h after coculture (Fig. 6e), suggesting that the mediators is produced and released after a transcriptional response. Alternatively, we tested the kinetics required for the mediators to manipulate MDDC responsiveness. Conditioned medium of MDDCs cocultured with infected T cells was collected after 24h, and used to culture fresh autologous MDDCs for a varying duration before the addition of LPS (Fig. 6f). A lowered responsiveness of MDDCs was only observed when MDDCs were cultured for at least 18h with the conditioned medium before adding the LPS (Fig. 6g). The manipulation of MDDC responsiveness might thus also rely on a transcriptional control, which is consistent with the unique transcriptional signature induced by exposure to HTLV-1-infected T cells observed by RNA-seq.
Altogether, our results (recapitulated in Fig. 6h) show that upon coculture, a molecular dialogue is initiated between HTLV-1-transformed T cells and MDDCs, which results in the transcriptionally-controlled release of a set of soluble tolerogenic mediators that manipulates MDDC responsiveness, in cooperation with additional mechanisms dependent on viral capture and/or cell-cell contacts.