Retinoic acid‐loaded liposomes induce human mucosal CD103+ dendritic cells that inhibit Th17 cells and drive regulatory T‐cell development in vitro

The active vitamin A metabolite, all‐trans‐retinoic acid (RA), primes precursor dendritic cells (DCs) into a mucosal phenotype with tolerogenic properties characterized by the expression of integrin CD103. CD103+ DCs can counteract pathogenic Th1 and Th17 in inflammatory bowel disease (IBD) or celiac disease (CD). Tolerogenic manipulation of DCs using nanoparticles carrying tolerogenic adjuvants and disease‐specific antigens is a valuable treatment strategy to induce antigen‐specific mucosal tolerance in vivo. Here, we investigated the effects of RA‐loaded liposomes on human DC phenotype and function, including DC‐driven T‐cell development, both during the generation of monocyte‐derived DCs (moDCs) as well as by priming immature moDCs. RA liposomes drove CD103+ DC differentiation as well as ALDH1A2 expression in DCs. Neutrophil‐dependent Th17 cell development was reduced by RA‐liposome‐differentiated and RA‐liposome‐primed DCs. Moreover, RA liposome treatment shifted T‐cell development toward a Th2 cell profile. Importantly, RA liposomes induced the development of IL‐10‐producing and FoxP3+ regulatory T cells (Tregs) of various Treg subsets, including ICOS+ Tregs, that were potent inhibitors of bystander memory T‐cell proliferation. Taken together, RA‐loaded liposomes could be a novel treatment avenue for IBD or CD patients.


Introduction
The inner mucosal lining of the gut is continuously exposed to microbial pathogens but also to harmless antigens such as nutrients from our diet or commensal bacteria required for digestion.Therefore, the intestine requires tight immune regulation to prevent the invasion of pathogens whilst maintaining tolerance to commensal bacteria and autoantigens.Maintaining intestinal homeostasis requires, among others, anti-inflammatory regulatory T cells (Tregs) to dampen the immune response and the activity of proinflammatory Th1 and Th17 CD4 + T cells, which are indispensable for the defense against pathogens and an intact gut barrier [1,2].Dendritic cells (DCs) orchestrate these immune responses, mounting both protective immunity and tolerance to harmless antigens.Amongst the various DC subsets present in the intestinal lamina propria [3], CD103 + DCs form a specialized subset characterized by an enhanced ability to generate retinoic acid (RA), a vitamin A metabolite, via expression of ALDH1A2, encoding for a retinal dehydrogenase [4].RA-secretion by CD103 + DCs, in turn, induces gut-homing T cells via the chemokine receptor CCR9 [5,6].Immune dysfunction and the resulting lack of mucosal immune homeostasis can result in inflammatory bowel disease (IBD) or celiac disease (CD) [1,2,7].IBD is a chronic relapsing disorder of the gastrointestinal tract that can be specified as either ulcerative colitis or Crohn's disease, characterized by abdominal pain, fever, diarrhea, and bloody stool [1,2].CD is caused by a distorted immune response against gluten with similar symptoms to IBD [7].
While Th1 and Th17 cells are indispensable for intestinal homeostasis by protection against pathogens, an excess of these cells and their cytokines is found in the intestinal mucosa of IBD and CD patients, contributing to pathogenesis [1,[7][8][9].Th17 cells, maintained by DC-derived IL-23 [10], are pathogenic by their production of proinflammatory cytokines, including IL-17 and TNF [1,2,11].IL-17 and TNF stimulate intestinal myofibroblasts to produce damaging metalloproteinases [11] and mediate the recruitment of inflammatory cells, including neutrophils, to the gut [2,12,13].The abundant recruitment of neutrophils to the intestine participates in IBD pathogenesis by reactive oxygen species production, degranulation, and neutrophil extracellular trap formation, amongst others, which leads to collateral tissue damage [13][14][15].Hence, patients with IBD or CD could benefit from shifting the pathogenic intestinal Th1/Th17 axis to Tregs to restore immune homeostasis.
RA, which is produced by gut epithelial cells lining the lamina propria, stimulates the differentiation of CD103 + DCs that contribute to the establishment of immune homeostasis in the gut [4,16].As demonstrated in mouse studies, RA-regulated CD103 + intestinal DCs instructed the development of FoxP3 + Tregs [4,16,17].We previously showed that in the human setting, RAinduced CD103 + DCs polarize naïve T cells to IL-10-producing FoxP3 − Tregs equipped with the gut-homing receptor CCR9 [18].Given the anti-inflammatory properties of RA, this vitamin compound appears a suitable candidate for DC modulation in patients with chronic inflammatory disorders in order to restore immune balance.Several clinical trials are applying ex vivo DC therapy by generation of tolerogenic DCs and reinfusion into patients, as reviewed in [19], for example, intraperitoneal administration of tolerogenic DCs generated with vitamin A and dexamethasone in patients with refractory Crohn's disease [20].Alternatively, DCs could be targeted in vivo using vaccine-containing nanoparticles, circumventing this costly ex vivo therapy, as reviewed in [21,22].Liposomes, nanostructures consisting of a phospholipid bilayer, are easily adjustable to accept various compounds [23] and can, therefore, contain disease-relevant antigens, RA, and DCtargeting moieties.This strategy reduces off-target side effects and yields an opportunity to ameliorate mucosal inflammation in an antigen-specific manner [22], for example, loading nanoparticles with flagellin antigens for Crohn's disease patients or gluten antigens for patients with CD [7,24].Of note, the charge of liposomes affects the internalization efficiency by DCs, as we previously demonstrated that liposomes containing the anionic lipid 1,2-distearoyl-sn-glycero-3-phosphoglycerol (DSPG) were internalized more effectively by DCs compared with two other cationic formulations [25].Anionic DSPG liposomes have recently been loaded with RA and human proteoglycan-derived peptide, an arthritis-inducing antigen widely used in murine studies [26].When injected in mice, this liposomal formulation reduced the percentage of activated splenic DCs compared with the injection of free peptide and RA while inducing the development of antigen-specific Tregs.However, little is known about the effect of RA-loaded nanoparticles on human DCs and T cells.
Here, we investigate the effects of RA loaded in DSPG liposomes on DCs and their capacity to promote specific T-cell development in comparison to soluble RA.We studied the impact of RA liposomes on the differentiation of monocyte-derived DCs (moDCs) as well as on established moDCs.We demonstrate that RA liposomes drive the differentiation of mucosal-like CD103 + DCs, which express ALDH1A2, comparable to soluble RA.Notably, neutrophil-dependent Th17 cell development is reduced by these RA-liposome-differentiated DCs as well as by DCs primed with RA liposomes.Concomitantly, RA liposomes foster tolerogenic DCs that induce the development of IL-10-producing, functional Tregs that suppress bystander T-cell proliferation, similar to soluble RA.Taken together, these data support the use of RA-loaded liposomes in future in vivo applications, that is, testing the concept of shifting the distorted immune balance from Th1/Th17 cells toward Treg polarization in disease models of IBD and CD.

RA liposomes drive human CD103+ DC differentiation in vitro
Since in vivo targeting of DCs with nanoparticle vaccines could be a valuable future treatment strategy for patients with immunemediated inflammatory diseases, we investigated whether RAloaded anionic DSPG liposomes could drive the differentiation of precursor DCs (monocytes) into mucosal-like tolerogenic CD103 + DCs.The mean loading efficiency (LE) of RA with the rigid head lipid 1,2-distearoyl-sn-glycero-3-phosphocholine (DSPC) in the DSPG liposome formulation was merely 36 %, and therefore, the head lipid was replaced by 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC), leading to improved LE of 69 % (Table 1).Consequently, we continued with DSPG liposomes containing DOPC as the head lipid and added RA liposomes, soluble RA, vehicle control dimethyl sulfoxide (DMSO), or empty DSPG(−) liposomes to monocytes that were differentiated to moDCs.Differentiation of DCs in the presence of DSPG(−) RA liposomes induced the expression of CD103, which was significantly increased compared with soluble RA (Fig. 1A and B).Interestingly, empty liposomes alone already increased CD103 expression, which was significantly enhanced by the incorporation of RA into liposomes.Furthermore, mRNA expression of ALDH1A2 was significantly increased in RA-liposome-differentiated DCs, as well as by soluble RA, but no effect of empty liposomes on ALDH1A2 expression was observed (Fig. 1C).RA-differentiated DCs retained the capacity to mature upon stimulation with IL-1β and TNF-α (this combination is referred to as maturation factor or "MF") and LPS (Fig. 1D-F).Our data indicate that RA administered in liposomal form induces the differentiation of mucosal-like CD103+ DCs that express ALDH1A2.

RA-liposome-induced CD103+ DCs inhibit human Th17 cell development in favor of Tregs
Although the induction of mucosal-like DCs by RA is established, it is unknown how these DCs affect the development of human Th17 cells from naïve T cells.As we previously showed, DCdriven Th17 cell development from human naïve CD4+ T cells is neutrophil-dependent [27].Therefore, we cocultured immature DCs with autologous naïve T cells and neutrophils on Candida albicans hyphae, an antigen-specific system that requires autologous cells (Fig. 2A).DCs differentiated in presence of RA liposomes, or soluble RA, showed less Th17 cell polarization (Fig. 2B and C, Supporting information Fig. S1A), as measured by intracellular IL-17A expression, while Th1-and Th2-cell development was unaffected, as seen by similar intracellular IFN-γ and IL-13 expression, respectively (Fig. 2D and E, Supporting information Fig. S1B and C).In contrast, we observed that RA-liposomedifferentiated DCs significantly induced IL-10-producing T cells (Fig. 2F, Supporting information Fig. S1D), as measured in super-natant from restimulated T cells, and FoxP3+CD127low Tregs (Fig. 2G, Supporting information Fig. S1E).Additionally, RA-and RA-liposome-conditioned DCs induced T cells expressing the guthoming marker CCR9 (Fig. 2H).Taken together, our data suggests that Th17 cell development is inhibited by mucosal-like DCs versus control DCs, while Treg development is enhanced.

RA-liposome-treated DCs induce functional Tregs and shift Th1/Th17 cell development to Th2 cells
Besides the role of RA in the generation of tolerogenic DCs from precursors, RA and especially RA liposomes may also affect differentiated DCs.Therefore, we activated immature DCs with a combination of MF + LPS in the presence of RA liposomes or relevant controls and analyzed the expression of DC surface markers within HLA-DR + cells (Supporting information Fig. S2A).DC maturation marker expression was unaffected by soluble RA or RA liposome treatment compared with the mature DC control condition (Fig. 3A-C, Supporting information Fig. S2B-D).Expression of the tolerogenic marker ILT3 was increased in soluble RAtreated DCs compared with immature DCs and empty liposometreated DCs (Fig. 3D), while the remaining tolerogenic markers were either unaffected by RA treatment or only significantly altered compared with immature DCs (Fig. 3E, F; Supporting information Fig. S2E-H).After 48-h maturation with RA liposomes, close to 100 % of DCs internalized DSPG liposomes, as measured using the fluorescent liposome label 1,1-Dioctadecyl-3,3,3,3-tetramethylindodicarbocyanine (DiD; Supporting information Fig. S3), indicating that DCs received equal concentrations of soluble RA and liposome-loaded RA.Next, we observed DC capacity to induce specific T-cell responses upon coculture with allogeneic naïve CD4 + T cells, as for this antigen-independent read-out, we need allogeneic cell stimulation to achieve T-cell proliferation (Fig. 4A).A hallmark of Tregs is suppression of bystander memory T cells.Indeed, we observed the suppressive capacity of emerging effector T cells on bystander memory T cells (Fig. 4B and C).Importantly, both RA-liposome-and soluble RA-treated DCs induced suppressive T cells compared with the control (Fig. 4C).RA-liposome-or RAtreated DCs equally induced suppressive T cells with no significant difference between the conditions.We further investigated the phenotype of DC-primed allogeneic T cells by assessing IL-10 production and FoxP3 expression.Compared with the medium   control of MF + LPS-activated DCs, RA-liposome-and RA-treated DCs significantly induced IL-10-producing T cells (Fig. 4D).RAliposome-treated DCs also significantly generated IL-10-producers compared with the empty liposome control.Additional to IL-10-producers, RA-liposome-or RA-treated DCs (at an RA concentration of 3 µM) significantly stimulated the development of FoxP3 + CD127low T cells, compared with the activated DC medium control or the empty liposome control condition (Fig. 4E, Supporting information Fig. S4A and B).Induction of Tregs can be paralleled by shifts in effector T-cell polarization.Hence, we analyzed Th1-and Th2-cell polarization in the effector T cells primed by RA-liposome-treated DCs.In the allogeneic coculture setting, RA-liposome-or soluble RA-treated DCs significantly reduced Th1-cell polarization while stimulating Th2-cell polarization (Fig. 4F and G).Moreover, to measure the effect on Th17cell development, immature control DCs were primed for 2 h with RA liposomes, soluble RA, or vehicle controls and cocul-tured as depicted in (Fig. 2A (right panel)).Th17 cell development was significantly reduced by RA-liposome-or RA-priming of DCs (Fig. 4H).Thus, our data indicate that RA-liposome-or RA-treated DCs induce suppressive Tregs, which are both IL-10producing and FoxP3+CD127low, while T helper-cell polarization is shifted from Th1/Th17 cells to Th2 cells.

RA-liposome-treated DCs induce Tregs with a distinct footprint of tolerogenic markers
Since IL-10-producing and FoxP3 + T cells encompass a heterogeneous population of T cells, we set out to analyze the phenotype of Tregs induced by RA-liposome-treated DCs using a flow cytometry panel containing various Treg markers (Supporting information Fig. S5).CTLA-4 + cells were significantly induced within the FoxP3 + CD127low T-cell population primed by RA-liposome-or soluble RA-treated DCs (Fig. 5A).Furthermore, FoxP3+ CD127low T cells coexpressing inducible costimulatory (ICOS) and CTLA-4 (known as ICOS+ Tregs) were also increased within this population, by RA-liposome-or RA-treated DCs (Fig. 5B).Additionally, the FoxP3+ T cells promoted by these DCs featured a higher expression of the coinhibitory marker Tcell immunoglobulin and ITIM domain (TIGIT) (Fig. 5C).Next, we observed the expression of CD39, which is an ectoenzyme that converts ATP to adenosine, thereby depleting resources from effector T cells [28].Interestingly, FoxP3 + T cells raised by RAliposome-or RA-treated DCs more frequently coexpressed CD39, underlining the functional Treg identity of these FoxP3 + T cells (Fig. 5D).As the T-cell activation marker CD69 may accentuate suppressive function of Tregs [29], we examined the expression of FoxP3 by CD69 + and CD69 -populations.Activated CD69+ T cells demonstrated a significantly higher FoxP3+ coexpression in the soluble RA-treated DC condition, while a trend of increased CD69 + FoxP3 + T cells was observed in the RA-liposome-treated DC condition (Fig. 5E).However, CD69 -FoxP3 + T cells were significantly induced by RA-liposome-and soluble RA-treated DCs, indicating that not all induced FoxP3 + T cells possessed an activated phenotype (Fig. 5F).With respect to the total population of T cells, programmed cell death protein 1 (PD-1) expressing T cells were significantly elevated in RA-liposome-or RA-treated DC conditions (Fig. 5G), while no significant changes in expression of this immune checkpoint marker were noted within the FoxP3 + population (data not shown).As IL-10 production by T cells was clearly stimulated by RA-treated DCs, we examined the presence of peripheral Treg type 1 cells (Tr1s) using flow cytometric staining for LAG-3 and CD49b, two markers typifying the Tr1 subset [30].Even though Tr1s are classified as major IL-10producing Tregs [31], only soluble RA-treated DCs increased the frequency of Tr1 cells in two out of three donors.No other differences were found (Fig. 5H), indicating that the primary source of IL-10 was not attributed to Tr1 Tregs in our cocultures.Taken together, our data suggest the induction of FoxP3 + T cells coexpressing the functional Treg markers CD39 and TIGIT by RA-and RA-liposome-treated DCs, as well as the elevated frequency of ICOS + Tregs.

Discussion
In this study, we explored the use of RA complexed in a nanoparticle for the future treatment of chronic inflammatory diseases.
We demonstrate that RA-loaded liposomes and soluble RA are potent inducers of DC-driven Treg development, either during the generation of CD103 + DCs from monocytes or upon modulation of established DCs in vitro.Furthermore, neutrophil-dependent human Th17-cell development is inhibited by RA liposomes as well as soluble RA, whereas a shift from Th1-to Th2-cell development occurs only when DCs are activated in the presence of RA liposomes or RA but not when DCs are generated with these compounds.In-depth analysis of RA-liposome or RA-induced Tregs demonstrated that within this heterogenous population, distinct Treg subsets arise, including ICOS + Tregs, CD39 + FoxP3 + , and TIGIT + FoxP3 + Tregs.Importantly, these T cells possess the ultimate functional Treg characteristic: the capacity to suppress the proliferation of bystander cells.
For this study, we used anionic DSPG(−) liposomes loaded with RA since we previously showed that anionic formulations are more efficiently taken up by DCs than cationic liposomes [25].Anionic formulations, such as DSPG, were previously described as tolerogenic toward DCs [32,33].However, in our recent study, treatment of DCs with empty liposome formulations, including DSPG(−), did not yield significant modulation of DC function, for example, IL-10 production or maturation marker expression [25].In contrast, here we show that upon differentiation with empty DSPG(−) liposomes, CD103 expression was significantly increased on DCs, to a similar degree as by soluble RA.This effect of empty liposomes possibly resulted in a significantly increased induction of CD103 upon DSPG(−) RA treatment, compared with soluble RA.We and others previously demonstrated that differentiation of monocytes to DCs with RA generated CD103 + mucosal-like DCs that express ALDH1A2 [18,34,35].Indeed, CD103 and ALDH1A2 were both induced in RA-liposome-and RAdifferentiated DCs, while ALDH1A2 expression was not increased in empty liposome-differentiated DCs.Therefore, it is unclear whether the elevated CD103 expression on these DCs has functional implications, especially since autocrine DC RA production, as indicated by unchanged ALDH1A2 expression, was not enhanced and IL-10-producing Tregs were not induced by emptyliposome-differentiated DCs.
We analyzed the capacity of CD103 + mucosal-like DCs to induce neutrophil-dependent Th17-cell development from naïve T cells and found that Th17-cell polarization is reduced by RAliposome-or RA-differentiated DCs.Bene et al. [34] previously demonstrated that microbiota-induced Th17-cell polarization by human RA-differentiated CD103 + moDCs is reduced compared with control moDCs, but this was tested with peripheral blood lymphocytes, comprising both memory and naïve T cells.Therefore, these data most likely reflect Th17-cell activation of memory T cells rather than the initiation of de novo Th17-cell development from naïve precursors.We expect that RA-liposome-differentiated DCs subsequently also produce RA due to their ALDH1A2 expression and our previous data on increased enzymatic aldehyde dehydrogenase activity of RA-DCs [18].Therefore, we speculate that DC-derived RA may directly affect T cells in our cultures.

RA reduces the differentiation of T cells to Th17 cells in mice by
blocking IL-1β, IL-6, and IL-23 signaling in T cells [36,37].Of note, human Th17-cell development is different from mice, but IL-1β, IL-6, and IL-23 are also involved in the differentiation or expansion of human Th17 cells, in addition to TGF-β and IL-21 [10,[38][39][40].RA was also shown to inhibit IL-1β and IL-6 production, among others, in T-cell receptor-stimulated human peripheral blood mononuclear cells and to directly inhibit human Th1 and Th17 cell development from naïve CD4+ T cells cultured under Th1-or Th17-polarizing conditions [35,41].In addition to CD103 + DCs, RA-liposome-or RA-treated DCs also reduced Th17cell development compared with controls.Collectively, our data and existing literature demonstrate that neutrophil-dependent human DC-driven Th17-cell development is inhibited by RA liposomes or soluble RA, which could be a result of the direct effects of DC-derived RA on T cells.
Besides Th17 development, we also investigated the Th1/Th2 cytokine profiles induced by the mucosal-like CD103+ DCs, as well as by RA-liposome-or RA-treated DCs.RA-liposome-or RA-treatment of established DCs demonstrated a clear, significant shift to Th2 (IL-13 + ) cells and a reduced Th1 (IFN-γ + ) cell polarization.Indeed, RA was previously shown to induce a shift from Th1 to Th2 cells in antibody-stimulated T cells [41].However, we observed no differences between Th1 cell polarization stimulated by RA-generated mucosal-like CD103 + DCs or control DCs, as previously described [18].A possible explanation for this discrepancy may be that IL-12 production by RAdifferentiated DCs is not reduced compared with control DCs, as established in our previous study [18].In contrast, 48-h treatment of established DCs with RA may inhibit the LPS-stimulated production of IL-12, as this was previously observed in mouse macrophages [42], and thereby result in the skewing that we observed.
Both Th1 and Th17 cells can be pathogenic in IBD and other chronic inflammatory diseases such as spondyloarthritis, and the holy grail for restoration of immune homeostasis is the induction of Tregs that can suppress the development and function of T-helper cells [1,2].Th17 and Tregs reciprocally control proliferation to maintain an equilibrium [43], and TGF-β and RA play a dictating role here.TGF-β and RA both increase the polarization of human naïve CD4 + T cells to FoxP3 + Tregs, and the combination of TGF-β and RA is superior [18].TGF-β is also required for Th17 cell differentiation, although a high dose inhibits the expression of transcription factor RORγT critical for Th17-cell development [2,39].Moreover, RA displayed dosedependent effects as well on Th17-cell development in mice: 10 µM inhibited Th17-cell induction by lamina propria (mucosal) DCs, while a low dose of 1 nM actually stimulated Th17-cell development by spleen DCs [44].Hence, the concentration range of adjuvant RA dictates the outcome of the ensuing immune response, and optimal RA-loading of liposomes should be thoroughly investigated and tightly controlled prior to clinical applications.
The only tolerogenic modulation we observed by RA treatment of established DCs was an increased frequency of ILT3 + DCs upon soluble RA treatment, which may explain some of the T-cell-dampening effects of RA-treated DCs.Unexpectedly, no other maturation or tolerogenic markers were altered by RA or RA liposomes, contrasting with a recent murine study where DSPG RA liposome treatment decreased BMDC expression of CD86 and HLA-DR [26].Despite observing only a few changes in established DC phenotype, we found that RA liposome or RA treatment of DCs significantly induced both IL-10-producing and FoxP3 + CD127low Treg cells, which was also observed with RA-liposome-differentiated DCs cocultured with autologous naïve T cells.This is in contrast to our previous results with RA-differentiated DCs, where no enhanced FoxP3 expression was observed on allogeneic T cells [18].Interestingly, when looking at LAG-3 and CD49b, the markers identifying Tr1-type Tregs [30], we did not observe induction by RAtreated DCs.This suggests that IL-10-producing Tregs induced by RA-treated DCs possess a different phenotype.A subset of the induced FoxP3 + T cells were ICOS + Tregs, which are known to produce substantial amounts of IL-10 [45,46].Furthermore, CD69 + FoxP3 + Tregs were induced by RA liposomes and soluble RA.This subset is also described to produce IL-10, which attenuated inflammation in experimental IBD in mice [29].Of note, FoxP3 + T cells promoted by RA-liposome-or RA-treated DCs coexpressed TIGIT.Besides selectively inhibiting Th1 and Th17 cell responses [47], TIGIT + Tregs were shown to induce IL-10 production by DCs [48], providing an anti-inflammatory feedback loop.The enhanced IL-10 production by T cells can thus be explained by the significant induction of various FoxP3 + Tcell subsets in the allogeneic cocultures.These FoxP3 + T cells also expressed CD39, an ectoenzyme that converts ATP to immunosuppressive adenosine, thereby depleting resources from effector T cells and limiting their function.This CD39 + FoxP3 + coexpression further strengthens the observation that RA-liposome-or RAtreated DCs indeed induce an array of functional Tregs equipped with various mechanisms of action to suppress T-cell proliferation.
In this study, we demonstrate that RA liposomes have tolerance-inducing properties, both on human monocytes differentiating toward DCs and on established DCs via induction of Tregs.The advantage gained by packaging RA in liposomes is the possibility to also add disease-specific antigens and DC-targeting moieties to a single particle, which could increase treatment specificity and reduce off-target effects.Hence, we provide in vitro precedent for a vaccination strategy using RA-loaded liposomes to treat (autoimmune) chronic inflammatory diseases to induce immune tolerance via DCs.

Data limitations and perspectives
The aim of our study was to establish the effectiveness of RAloaded liposomes as vesicles capable of modulating DC function and inducing DC-mediated T-cell tolerance.As we tested our compounds with human cells, we used exclusively in vitro-generated DCs.Given the nature of this in vitro work with purified human cells that are directly accessible and consist of our cells of interest (i.e., DCs), a superior effect of the liposomes was not observed, but also not expected.However, we did demonstrate that RAloaded liposomes are equally capable of modulating DC function, thereby underscoring the potential of these nanocarriers to target RA to DC and modulate their function in vivo.To this end, future valuable additions to our study would be the in vitro testing of RAloaded liposomes in human tissues containing naturally occurring DCs (such as in gut tissue obtained from IBD patients or skin tissue obtained from plastic surgeries).Once the combination of adjuvant, disease-specific antigen, and targeting molecules to reach specific DC subsets in vivo is optimized, we expect a clear superior effect of all the compounds being delivered together in one nanostructure compared with separately in soluble forms.Transfer of in vitro-induced Tregs to an IBD mouse model to determine the stability of Treg function would also constitute a valuable addition to the current findings before moving into clinical trial settings in patients.It would also be of interest to explore the mechanism of Treg induction further by monitoring the fate of different induced T-effector phenotypes.

Cell isolation
Blood was collected into heparin tubes (Greiner Bio-One, Alphen a/d Rijn) from healthy donors after informed consent.Healthy volunteers were recruited for blood sampling in accordance with study protocols reviewed and approved by the Institutional Review Board of the Amsterdam University Medical Centers (METC 2015_074) and by TbDC protocol NL73819.018.21.Monocytes and peripheral blood lymphocytes were isolated from the peripheral blood mononuclear cell fraction after density gradient centrifugation on Lymphoprep and neutrophils from the erythrocyte pellet, as described earlier [27].CD4 + T cells were purified from peripheral blood lymphocytes by negative magnetic selection using the MACS CD4 + T-cell isolation kit (Miltenyi Biotec), and CD4 + CD45RA + naïve cells were separated from memory cells by negative selection on CD45RO expression by using CD45RO-PE (DAKO) and magnetic-labeled anti-PE beads (Miltenyi Biotec).The purity of the naïve T cells always exceeded 98 %.While T cells were stored in liquid nitrogen, autologous neutrophils (purity ≥ 98 %) were isolated fresh on the day when a coculture was initiated.Cell purities were assessed by flow cytometry.

Liposome preparation
Anionic liposomes consisting of DSPG were made using the thin film dehydration-rehydration method, as previously described [25,32].For the RA-loaded formulations, 300 µg all-trans-RA (Sigma-Aldrich) dissolved in chloroform as 2.5 mg/mL was added to approximately 3 mg of lipids in the lipid mix.For fluorescent labeling, the lipophilic tracer DiD (Thermo Fisher Scientific) was added to the lipid mix in a molar percentage of 0.1 %.RA-loaded liposomes were dialyzed overnight using a Spectra-Por Float-Alyzer dialysis kit (MWCO 100,000 Da) against 400 mL 10 mM phosphate buffer (PB) pH 7.4 to separate nonencapsulated vitamins.Liposomes were stored at 4 °C in PB and used for further experiments within 3 months.

Quality control of liposomes
To assess the lipid concentration of the formulations and RA concentration in the liposomes, reversed-phase ultraperformance liquid chromatography (Waters ACQUITY UPLC) was used, as described [25].RA was detected by absorbance at 252 nm using an ACQUITY UPLC TUV detector (Waters).LE of RA was calculated as LE (%) = RA concentration after dialysis RA concentration before extrusion × 100.
All formulations had a size (Z-average) of less than 200 nm and a PdI of less than 0.2, indicating monodispersity (Table 1).Measured ζ -potential corresponded to the expected anionic surface charge of the formulations.Quality control of formulations was performed as previously described [25].For stability testing, formulations were monitored over the course of 3 months, and measurements were repeated each month after liposome preparation (Table 2).Formulations were stable throughout their use.

Stimulation and analysis of CD4 + T cells
For the allogeneic cocultures, 5000 MF + LPS-activated DCs that were previously matured for 48 h with or without liposomeloaded RA or soluble RA were cocultured in a flat-bottom 96-wells plate with 20,000 naïve CD4 + T cells in 200 µL IMDM with 10 % FCS and gentamycin in the presence of 10 pg/mL Staphylococcus aureus enterotoxin B (Sigma-Aldrich), as described before [51].DCs were washed three times with 3 mL medium prior to use in culture.On day 4-5 of coculture, effector T cells were trans-ferred to a 24-well plate and refreshed with rhIL-2 supplemented medium, leaving plastic-adherent moDCs behind.When resting on day 10-12 of culture, a maximum of 500,000 effector T cells was restimulated for 5 h in IMDM/10 % FCS for assessment of intracellular cytokines IL-13 and IFN-γ [50].In addition, 100,000 CD4 + T cells were stimulated in triplicate with soluble anti-human CD3 and CD28 in 200 µL of IMDM/10 % FCS for analysis of IL-10 in 24-h supernatants, as done previously [50].For phenotyping of Tregs, a panel of antibodies was used as recently described [49].

T-cell suppressor assay
For suppressor assays, 300,000 naïve CD4 + T cells from an allogeneic donor were cocultured for 6 days with 30,000 MF + LPSactivated and soluble RA or RA-liposome-treated moDCs in 1 mL total volume of IMDM/10 % FCS medium.After 6 days, the developing effector T cells were gently harvested, irradiated at 30 gray to prevent proliferation, and washed three times with 3 mL IMDM/10 % FCS medium.Subsequently, 50,000 effector T cells were cultured in triplicate with 1500 allogeneic MF + LPS-matured DCs and 25,000 autologous memory CD4 + T cells labeled with CFSE (0.5 mM; Molecular Probes) as target cells in 200 µL IMDM/10 % FCS.After 5 days, the proliferation of CFSElabeled CD4 + memory T cells was evaluated on a FACS Canto machine (BD Biosciences) by acquiring 5000 cells per sample.

Statistics
Flow cytometric analyses were performed using FlowJo software (version 10.

Figure 1 .
Figure 1.RA liposomes drive differentiation of mucosal-like CD103 + DCs.Monocytes were differentiated to DCs in presence of vehicle control DMSO, RA, DSPG(−) empty liposomes, or DSPG(−) RA liposomes and analyzed by FACS and qPCR at day 5. (A) CD103 gating strategy and expression are shown of one representative donor, and (B) pooled data of n = 9 independent donors are depicted as mean fluorescence intensity (MFI).(C) Relative mRNA expression of ALDH1A2 to DMSO control is shown of n = 5 independent donors.(D, E, F) Expression levels (MFI) of CD83, CD86, and HLA-DR of differentiated DCs are shown upon 48-h stimulation with MF + LPS (conditions DMSO, RA, DSPG(−), DSPG(−) RA), or no stimulation (−) in example histograms and n = 4 independent donors.Statistical significance was determined using one-way ANOVA with Šidák's post hoc test for multiple comparisons (B, D-F) or Friedman test with Dunn's post hoc test (C).Error bars indicate mean ± SEM. *p ≤ 0.05.**p ≤ 0.01.***p ≤ 0.001.Individual data points: number of donors tested per condition (independent experiments).

Figure 2 .
Figure 2. RA-liposome-differentiated DCs reduce human Th17 cell development while inducing Tregs.Monocytes were differentiated to DCs in presence of vehicle control DMSO, RA, DSPG(−) empty liposomes, or DSPG(−) RA liposomes and subsequently cocultured with autologous naïve T cells and autologous neutrophils on C. albicans hyphae as antigenic stimulus.After 10-13 days, T-cell markers and cytokines were assessed via flow cytometry or sandwich ELISA.(A) Schematic model of using RA-or RA-liposome-differentiated DCs (left panel) for coculture of immature moDCs with autologous naïve T cells and neutrophils on C. albicans hyphae as model antigen (right panel).(B) Flow cytometry plots of a representative donor are shown of IL-17 and IFN-γ (Th17 vs. Th1 cells) and (C) pooled Th17-cell percentages of n = 6.(D) Frequencies of Th1 (IFN-γ + ) and (E) Th2 (IL-13 + ) cells are shown, n = 6.(F) Relative production of IL-10 by T cells to DMSO control as measured by ELISA (mean ± SD = 0.27 ± 0.22 ng/mL) is shown of n = 5 independent experiments.(G) Frequencies of FoxP3 + CD127low T cells are shown of n = 5. (H) Frequencies of CCR9 + T cells are shown of n = 3. Statistical significance was determined using Friedman test with Dunn's post hoc test for multiple comparisons (C, F) or one-way ANOVA with Šidák's post hoc test (D, E, G, H).Error bars indicate mean ± SEM. *p ≤ 0.05.**p ≤ 0.01.Individual data points: number of donors tested per condition (independent experiments).

Figure 3 .
Figure 3. RA-treated moDCs have increased expression of the tolerogenic marker ILT3, while their maturation is unaffected by RA liposome treatment.MoDCs were left immature (−) or matured for 48 h using MF + LPS (medium) in the presence or absence of RA liposomes, soluble RA, or vehicle controls.After maturation, the cells were measured for expression of maturation and tolerogenic markers using spectral flow cytometry.(A, B) Frequencies of CD83 + and CD86 + HLA-DR + moDCs are shown.(C) Frequencies of HLA-DR + moDCs are shown.(D-F) Frequencies of ILT3 + , ICOS-L + , and PD-L1 + HLA-DR + DCs are shown.N = 5 independent experiments.Statistical significance was determined using one-way ANOVA with Tukey's post hoc test for multiple comparisons.Error bars indicate mean ± SEM. *p ≤ 0.05.***p ≤ 0.001.****p ≤ 0.0001.Individual data points: number of donors tested per condition (independent experiments).

Figure 4 .
Figure 4. RA-liposome-treated DCs shift Th1/Th17-cell polarization to Tregs and Th2 cells.(A) Schematic model of the experimental set-up of the coculture of DCs matured with MF and LPS (medium) in presence of RA liposomes, soluble RA, or vehicle controls (left panel) prior to coculture with allogeneic naïve T cells (right panel).For characterizing Tregs and T-cell polarization, naïve T cells were cocultured with DCs in the presence of SEB for 10-12 days, with IL-2 supplemented every 2 days.When resting, T cells were characterized by flow cytometry to determine FoxP3 + Treg induction and intracellular cytokines.IL-10 was measured in T-cell supernatants.For observing suppressive capacity, naïve T cells were cultured with DCs for

Figure 5 .
Figure 5. DCs treated with soluble or liposome-loaded RA induce T cells with functional Treg markers.Allogeneic T cells were cocultured with DCs matured with MF and LPS (medium) in presence of soluble RA, RA liposomes, or vehicle controls.(A) Frequencies of CTLA-4 + T cells within the FoxP3 + CD127low population of T cells.(B) Frequency of ICOS + Tregs (ICOS + CTLA-4 + ) within FoxP3 + CD127low T cells is depicted.(C) TIGIT + FoxP3 + T cells (of total T cells) are shown.(D) Frequencies of FoxP3 + total T cells expressing the ectoenzyme CD39.(E) CD69 + FoxP3 + T cells are depicted.(F) Frequencies of non-activated CD69 -FoxP3 + T cells.(G) Frequencies of PD-1 + cells in the total T-cell population.(A-G) N = 10.(H) Frequencies of T cells coexpressing the markers CD49b and LAG-3 (Tr1 cells) are shown of n = 3. Statistical significance was determined using one-way ANOVA with Šidák's post hoc test.Error bars indicate mean ± SEM. *p ≤ 0.05.**p ≤ 0.01.Individual data points: number of donors tested per condition (independent experiments).
7.1 for Windows), and statistical analyses were performed using GraphPad Prism software (version 8.3.0 for Windows).D'Agostino & Pearson test was used to test normality of data if n ≥ 8, otherwise the Shapiro-Wilk test was used.Paired analyses with multiple comparisons were done by using one-way ANOVA, or Friedman tests if the distribution of data was not normal.p-values of 0.05 or less were considered significant.

Table 2 .
Stability measurements of the liposome formulation DSPG(−) RA.Formulation stability was measured over the course of 3 months using Z-average diameter, PdI, and ζ -potential.Mean ± SD of n = 3 liposome batches are shown. Note: