A novel plasma membrane-phillic graphene oxide nanocarrier for neuropeptide delivery to generate tolerogenic dendritic cells in GVHD Immunotherapy

The prevention and the treatment of graft-versus-host disease (GVHD) remains a barrier to allogeneic hematopoietic stem cell transplantation (allo-HSCT). Tolerogenic dendritic cells (TDCs) have become a hotspot in GVHD prevention despite the poor induction eciency. Herein, we designed a novel size-dependent platform of graphene oxide (GO) nanosheets for neuropeptide delivery for the purpose of TDC generation. GO with a lateral size(cid:0)1 µm (L-GO) showed strong anities to DC membrane, which effectively promoted the recognition between neuropeptide, urocortin (UCN) and its receptor CRHR2 and in turn beneted TDC generation through PKA C/CREB phosphorylation. Simultaneously, L-GO also elevated the expression of CCR7 and enhanced the migration ability of TDCs by mediating cytoskeleton reorganization. In vivo experiments offered direct evidence that TDCs inducted by UCN@L-GO exhibited ecient migration to lymph nodes, abundant generation of Treg, a signicant decrease of proinammatory cytokines and in turn excellent eciency in GVHD relief. In the current study, we proposed an innovative GO nanosheets based cytomembrane-targeted platform for neuropeptide delivery and subsequent TDC generation. In the meantime, the promoted mobility of TDC pulsed by GOs ensured the high homing eciency to secondary lymph nodes, attributing to GVHD treatment in vivo. Thus, such work provided a promising strategy that might be applicable more broadly to delivery systems for receptor-mediated drugs, and could meet the changing demand of nanotechnology-based immunotherapy.


Introduction
Allogeneic hematopoietic stem cell transplantation (allo-HSCT) has become a rst-line treatment for many types of malignancies, such as leukemia, lymphoma, and multiple myeloma 1 . Graft-versus-host disease (GVHD) is a serious and life-threatening complication that undergoes allo-HSCT with incidence rates ranging from 20 to 80% 2 , which limits the e ciency of allo-HSCT. Generally, GVHD arises from the activation of donor T lymphocytes after APCs presentation and attack the recipient's tissues, including the skin, liver, gastro intestinal tract and nervous system 3,4 . Based on the guidelines of American Society of Blood and Marrow Transplantation, corticosteroids are the mainstay treatments for GVHD. However, systemic corticosteroid showed limited overall complete response and a series of sequelae 2 , like high incidence of steroid-refractory 2 . So far, therapies targeting immune cells and immune cytokines have proved to be effective in GVHD treatment. In particular, adoptive immunotherapy of tolerogenic dendritic cells (TDCs), regulatory T cells (Tregs) and mesenchymal stem cells (MSCs) showed positive effect in the relief of GVHD by activating T cell deletion 5,6 . MSCs can interact with innate immune cells to downregulate immunity, but the lung-accumulated distribution of MSCs after transplantation limits its function. Tregs were also threatened by the targeted gene editing strategy 2,7 .
DCs are known as the strongest antigen-presenting cells (APCs) that accommodate immune response by eliciting primary T cell response and modulating immune tolerance 8- 10 . In the initial stage of GVHD, DCs presented host antigens which participated in donor T cell expansion, mediated GVHD and caused severe damage to the liver, skin and intestines 4,11 . DCs have excellent ability to lead T cells to exert immune cell response or induce immune tolerance. They can induce peripheral tolerance to relieve GVHD by deleting cytotoxic T cells and upgrading Treg expansion [12][13][14] . Preclinical data from murine BMT (bone marrow transplantation) model systems also showed that recipient DCs improved survival by decreasing the severity of GVHD. Therefore, induction of TDCs, especially recipient TDCs, may impair the immuneregulatory function and facilitate the generation of Tregs. Thus, TDC generation could be a potent candidate for GVHD treatment 15 .
Many protocols have been established to generate TDCs in vitro. Treatment of DCs with broad immunosuppressants, such as glucocorticoids, rapamycin and cyclosporine, could contribute to prevention in animal models of allograft rejection and autoimmune disorders. However, such agents may exert such adverse effects as increased infection, disordered metabolism, peptic ulcer and hypertension 16 . Cytokines, such as IL-10 and TGF-β, were also used in DCs pretreatment and showed suppressive e cacy in animal models. Unfortunately, these TDCs were not stable since they could produce proin ammatory cytokines upon introduction of a second stimuli 17,18 . However, even in such a semi-activation state, these TDCs showed slight elevation in mobility, which notably limited its performance.
Recently, neuropeptides found in neural tissues and produced by immune cells under in ammation have been rediscovered as potential candidates against autoimmune diseases and transplantation complications. Recent research demonstrated that neuropeptides, such as urocortin (UCN), vasoactive intestinal peptide (VIP), α-melanocyte-stimulating hormone(α-MSH), and adrenomedullin (AM), all had potent effects in inducing immune tolerance 19 . UCN, a member of the corticotropin-releasing hormone (CRH) family, acted as a broad neuromodulator and was demonstrated to be a potential balance factor in immune tolerance induction 20 through an indirect anti-in ammation effect mediated by the hypothalamus-pituitary-adrenal (HPA) axis 21 . AM is believed to act as an endogenous immunomodulatory factor with predominantly anti-in ammatory effects 22 . Experiments also showed that after treatment with VIP, DCs could induce a signi cant decrease in expressions of co-stimulatory molecules and in the production of anti-in ammatory cytokines. Notably, these TDCs kept their immature phenotype even when exposed to an in ammatory environment, which is of great importance for GVHD treatment 23 . Despite easy availability and broad regulation of autoimmunity, the application of neuropeptides as immunomodulators in vivo still is hindered by easy degradation and poor penetration.
In recent years, nanotechnology is fast-developing and provides a novel platform for biomedical applications. In particular, engineering nanomaterials for the delivery of cargos or regulation of the expression pro le of DCs during transplantation have been regarded as a potential means of immunomodulation for tolerogenic outcomes. Graphene oxide (GO), a typical 2D nanomaterial that was reported in 2004 for the rst time, has attracted much attention in biomedical research 24 . Despite the varying lateral dimensions, GO has a relatively large speci c surface area and processes excellent capability for bioactive molecule interaction. Therefore, besides the efforts of applying GO as biosensor assay in detecting infectious disease 25,26 , it is always utilized as a nanoscale vehicle for targeted drug delivery in vivo thanks to its limited cytotoxicity and high biocompatibility 27,28 . It is well documented that large-sized GO (normally 1µm or above) prefers to adhere onto the cell plasma membrane, unlike its small counterpart that tends to be internalized by immunocytes through endocytosis. Moreover, it is able to modulate the DC/T interactions 29 and shows some degree of immunosuppression or T cell tolerance. On the other hand, it is found that the most of the receptors of neuropeptides belongs to the G-proteincoupled receptor (GPCR) family, which located on the cell membrane 19 . Given the potential ability of antiin ammation and immunoregulation, it is intriguing to explore the combination of GO and neuropeptides in TDC induction. Therefore, here we designed a GO-based therapeutic platform to carry neuropeptides to the surface of immature DCs for TDC generation, and investigate its behavior and possible mechanism in GVHD treatment. After the most effective neuropeptides were screened out from their candidates, in-depth exploration indicated that GO loading improved the migration ability of DCs, which facilitated the homing of TDCs to secondary lymph nodes. An acute GVHD model was employed to investigate the e cacy of as-prepared TDCs. The Fluc+ T cells were used in our model for in-situ imaging, which ensured real-time observation of proliferation of donor T cells and e ciency of GVHD treatment. Our data indicated that the neuropeptide-GO complex could be used as a potent therapeutic candidate for the prevention of GVHD. More importantly, the current work proposed a cytomembrane-philic biomacromolecule delivery system with a novelty design.

Results
Characterization of GO and its a nity to cell membrane.
GO of different scales were synthesized as vehicles for the delivery of neuropeptides to DCs.
Transmission electron microscopy (TEM) observations showed that GOs were characterized by irregular polygon of different lateral sizes ( Figure 1A). The hydrodynamic radii of GOs were measured by dynamic light scattering (DLS) as 103.1 nm (marked as S-GO) and 1192.1 nm (marked as L-GO), respectively (Supplementary Figure 1). Atomic force microscopy (AFM) revealed the average thickness of 1-2 nm, which aligned with the structural characteristics of monolayer 2D materials ( Figure 1B). In order to determine the maximum allowable concentration of GOs in delivery, ow cytometry was performed to analyze the survival rates of imDCs after incubation with GOs for 48 h. When the concentration of GOs was set as 15.6 µg/ml, GO showed little toxicity with the survival rate of imDCs higher than 90% compared to the control group (Supplementary Figure 2).
GO was proved to have strong a nities to cell membrane and mediate contact through membrane interactions. According to these characteristics, GO was chosen as a good nanoscaffold for variety biomedical applications. Confocal Raman results showed that after co-culture with S-GO or L-GO for 48h, the accumulation of GO was observed in cell membrane, suggesting that both S-GO and L-GO had nonspeci c a nities to accumulate in cell-rich areas including RAW264.7, Hela and DC. L-GO were mostly adhered to the membrane while S-GO showed a preference of locating in cytoplasm ( Figure 1C). Cytomembrane is rich in transmembrane receptors, which is a critical component in cell properties and subtype switching. However, whether GO' a nities have any in uence on membrane receptors or on the combination of ligands and receptors are still unknown. Here we chose RBCs as a lipid bilayer model because it has no organelles or nucleus, thus simplifying the interaction model. The surface of RBCs was rst biotinylated to varying degrees to imitate receptors of different abundances on cytomembrane, and different concentrations of S-GO or L-GO were added to PE labeled avidin to allow physical adsorption.
The binding process of avidin-biotin was used to simulate the combination that was dependent on the ligand and receptor ( Figure 1D). The combining e cacy was re ected by the mean uorescence intensity (MFI) of PE monitored by FACS. It is worth noting that even in the non-biotinylated group, the introduction of GO caused nonspeci c a nity to RBCs (Supplementary Figure 3). Figure 1E gave direct evidence that GO-pretreatment could increase the binding a nity between PE-labeled avidin and biotinylated RBCs.
Despite the presence of nonspeci c binding, such improvement was much more signi cant in the case of L-GO than in its counterpart, especially in the low abundant biotin group (1 or 2 μg biotin added in the labelling system). The speci c binding of biotin-avidin was measured after deduction of the nonspeci c binding. When 8 μg GO was added to the pretreated avidin, FACS results showed that S-GO and L-GO both increased MFI at each concentration of biotin, especially at a lower concentration (especially 1 μg and 2 μg), and that L-GO showed stronger improvement in MFI than S-GO did ( Figure 1E). The results after nonspeci c binding deduced are depicted as Figure 1F, in which case the GOs just acted as a promoter in the course of speci c recognition of avidin and biotin. Take biotinylated RBCs (1 μg group) as an example. the addition of different concentrations (2, 4 and 8 μg) of S-GO showed 1.98-, 2.03-and 3.87-fold enhancements compared to the blank, respectively. Meanwhile, the corresponding results for the L-GO group were 2.58-, 3.87-and 4.71-fold, respectively. These results indicated that GOs' high a nity to cytomembrane facilitated the speci c combination between the ligand and receptor, and that the largesized ones showed stronger enhancement ( Figure 1F). As the current work was intended to explore TDCsbased immunotherapy in GVHD treatment, we then proceeded to the investigation of the interaction manner between GOs and DCs. Laser confocal was performed to validate the location of GOs on DC plasma membrane after 24 h of co-culture. S-GO and L-GO were labeled with FITC, and DCs were stained with rhodamine. Figure 1G shows that S-GO was mainly found in the cytoplasm of DCs, while L-GO was located at the edge of cells that were likely to correspond to the GOs associated with the plasma membrane, which was consistent with the phenomenon observed in the above RBCs-based model and literature. ( Figure 1G).
The screening of neuropeptides aiming at inducing TDCs Neuropeptides, mainly secreted by immune cells such as macrophages and monocytes, were considered capable of inducing immune tolerance with a broad anti-in ammatory effect. Neuropeptides UCN, α-MSH, VIP and AM were chosen as alternatives to induce the generation of TDCs rstly in a non-in ammatory microenvironment. Proin ammatory cytokines, such as IL-12p 70 , IL-6, IL-1β and TNF-α were measured by ELISA. It was found that neuropeptide incubation had little in uence on cytokine secretion of DCs as shown in Supplementary Figure 4. Given the systemic in ammation in case of GVHD, we next investigated the performance of the four candidates in an in ammatory environment established by low-dose LPS co-incubation for 48 h. In the case of DC LPS group, stimulatory DCs (SDCs) were obtained with elevated proin ammatory cytokines secretion ( Figure 2A). Meanwhile, in the presence of neuropeptide, the proin ammatory cytokines were found to be signi cantly decreased. Indeed, IL-12p 70 secreted by DCs co-cultured with UCN declined by 61.8% percent compared with the LPS stimulated ones (85.2±10.8pg/ml). IL-6 showed 80.3% decrease, while IL-1β was 64.8% and TNF-α was 82.4%, indicating that UCN exhibited more effective suppression on proin ammatory cytokines secretion than other candidates ( Figure 2A).
After UCN was selected as the most appropriate neuropeptides in the study, we stared to nd out whether UCN was capable of inducing TDCs. Balb/c derived CD8 + T cells and C57BL/6J derived DCs were isolated separately. CD8+ T cells were stained with carboxy uorescein diacetate succinimidyl ester (CFSE) and co-cultured with DCs at ratios of 2:1, 4:1 and 8:1 for 48h. Results showed that DCs with UCN treatment exhibited signi cant inhibition on cytotoxic T cell proliferation. And the strongest suppression was observed at the ratio of 4:1with a decrease of 52.6% compared to DC LPS . Cytotoxic T-cell proliferation test demonstrated that TDC UCN decreased CD8+ T cell proliferation and induced T cell anergy ( Figure 2B, 2C). CD40/80/86 were employed to characterize the maturity of DCs. It was found that after being cocultured for 48h, the TDC UCN group showed lower expressions of costimulatory molecules than the mature ones. FACS was signi cantly different between TDC UCN and DC LPS with CD40 (1.7-fold) and CD80 (1.5-fold), which demonstrated that UCN induced TDCs could preserve the tolerant phenotype in an in ammatory environment ( Figure 2D). Collectively, UCN was considered the most suitable one to induce TDCs.
Next, we attempted to construct a GO-based cell membrane targeted platform for the purpose of promoting the UCN delivery e ciency. UCN labeled by Cy5 incubated and physically adsorbed on to excessive S-GO or L-GO. Results of uorescence imaging showed that the conjugation e ciency of UCN with S-GO was (89.57 ± 1.53) %, compared with (86.33 ± 2.01) % with L-GO ( Figure 2E). The Fourier Transform Infrared (FT-IR) spectrometer was used to characterize the UCN@GO complex. A notable absorption peak of the peptide bond was observed compared to S-GO or L-GO alone ( Figure 2F), which demonstrated the neuropeptide UCN was successfully conjugated with GO nanosheets. Flow cytometry was performed to detect the survival rate of DCs after incubation with different concentrations of UCN@GO complex for 48h. The survival rates of DCs varied signi cantly in a dose-dependent manner according to Figure 2G, which was why 15.6 μg/ml was chosen as the working concentration, for the corresponding survival rates in UCN-S and UCN-L group were both above 90% (90.02±2.95%, 90.10±1.98%).

GO as nanocarriers for UCN delivery and TDCs generation
After being established for UCN, the GO delivery system was further evaluated in terms of TDCs generation. After incubation with UCN@GO for 48h, imDCs and the supernatant were collected separately. DCs were gated as CD11c+ population, CD40/80/86 showed no difference after incubation with free GO or UCN@GO (Supplementary Figure 5). ELISA measurements indicated that proin ammatory cytokines did not change signi cantly compared with imDCs after incubation with free GO or UCN@GO (Supplementary Figure 6). Further, the co-cultured DCs were added to CFSE labeled CD8+ T cells at a ratio of 1:4 for 48h. T cell proliferation rates of DC L-GO and DC UCN@L-GO were 9.9% and 7.87%, separately, compared with 10.94% of imDCs (Supplementary Figure 7). Therefore, we con rmed that neither free GO nor UCN@GO had any positive in uence on DC maturity, which facilitated TDCs induction.
To simulate the environment where GVHD occurs, UCN@GO pretreated TDCs were further challenged in an in ammatory microenvironment induced by low-dose LPS. As a member of the CRF family, UCN mainly binds to CRHR2 expressed on cell membrane and affects in ammation response. CRHR2, a G protein-coupled receptors, is mainly expressed on cell membrane. BioGPS, a gene annotation portal, can provide complete gene information, including patterns of gene expressions in different cells or tissues. Data from BioGPS showed that CRHR2 expressions in the immune system, like BDCA4+ DCs, CD19+ B cells, CD8+ T cells, CD56+NK cells and CD14+ monocytes, is under median / were below the median value (Supplementary Figure 8). It was speculated that the low abundance of CRHR2 expressions on DC membrane might be the key to restricting the e ciency of TDCs induction. Therefore, improving the enrichment of UCN on the cell surface and promoting the binding e cacy of UCN and CRHR2 seemed to be an alternative solution to the limited production of TDCs.
We went on to explore whether GO loading would bene t the recognition of UCN by CRHR2 as expected. UCN was labeled with 5-TAMER and coupled with GOs, respectively. After DC incubation, FACS results indicated that UCN@L-GO tended to be adhered to the cytomembrane compared with UCN@S-GO or free UCN (29.3% verse 20.0% and 14.8%) ( Figure 3A). 5-TAMER labeled UCN was further pre-conjugated with FITC-GOs, while DCs were stained with Cell Tracker TM Blue. Confocal imaging observation showed that most of the UCNs could be internalized in DCs cytoplasm without any GO carrier, which might fail to bind to the membrane receptor CRHR2. UCN@S-GO appeared to be in the cell cytoplasm, while UCN@L-GOs were mostly located on the surface of the cell membrane ( Figure 3B), which conformed with the manner in which the carrier GO alone interacted. These results suggested that UCN@S-GO experienced a normal endocytosis process during incubation with DCs, while the larger lateral diameter of L-GO might probably bene t the concentration of UCN molecules on plasma membrane and the binding event with its receptor.
The deduction was further veri ed by the co-location results of UCN and CRHR2 observed by the laser confocal (Supplementary Figure 9).
The downstream signaling molecular expressions were further examined by Western blot to explore the mechanism of TDC generation. Results showed that the presence of LPS had little in uence on PKA C or CREB phosphorylation, so did single S-GO or L-GO treatment. UCN and UCN@GOs complex exhibited a similar increase of PKA C phosphorylation to DC LPS . UCN incubation elicited more phosphorylation of the downstream CREB (1.81-fold), which indicated that UCN could effectively mediate TDCs production. And higher expressions of p-CREB (3.07-fold or 3.49-fold) were detected in the UCN@S-GO or UCN@L-GO group, suggesting that the enrichment of UCN on DC membrane, which resulted from GO coupling, could trigger cAMP/PKA pathway activation more effectively ( Figure 3C and 3D). The above results served as evidence that GO had positive effects on the targeted delivery of UCN and promoted the combination of the ligand and receptor on the plasma membrane of DCs.
Once the cAMP/PKA/CREB signal pathway was activated, the CBP (CREB binding protein, CBP) molecules were translocated into the nucleus and further activated NF-κB, resulting in a decrease of proin ammatory cytokine expressions and co-stimulatory molecular levels of DCs. Thus, the levels of cytokine secretion and CD40/80/86 phenotype of TDCs were examined after LPS challenge. It was observed that the concentration of IL-12p 70 in the supernatant showed no signi cant difference in S-GO or L-GO treated groups compared with mature DCs. The concentration of IL-12p 70 decreased to 54.85±4.99 pg/mL after UCN@S-GO pretreatment, and to 50.49±0.52 pg/mL after UCN@L-GO incubation, which was signi cantly lower than in the free UCN group (60.06±3.91 pg/mL) and mature DC group (85.97±2.19 pg/mL). The amounts of IL-1β, IL-6 and TNF-α secreted by DCs co-cultured with UCN@S-GO and UCN@L-GO also showed notable decreases compared to the free UCN group. These in ammatory cytokine levels demonstrated that the UCN-GO complex was more e cient in inducing TDCs than free UCN treatment ( Figure 3E). Although S-GO or L-GO had little in uence on expressions of costimulatory molecules, S-GO enhanced the e ciency of UCN in inhibiting the expression of surface costimulatory molecules CD40 (1.2-fold) and CD80 (1.3-fold) according to FACS results, compared with 1.5-fold, 1.7fold, and 1.1-fold after UCN@L-GO pretreatment, which exhibited a notable decrease in DCs maturity ( Figure 3F). Given the positive role that GO played in promoting ligand and receptor combination and delivery of UCN, it could be speculated that the UCN@GO process enhanced UCN's ability to induce TDCs. Furthermore, it was obvious that UCN@L-GO outperformed UCN@S-GO in TDC generation. Thus, UCN@L-GO was chosen as an inducer of TDCs for follow-up experiments.
TDC UCN@L-GO effectively deleted cytotoxic T cells and generated Tregs.
Then, the T cell subtype and function were detected after co-culture with TDC UCN or TDC UCN@L-GO . CD8+ T cells isolated from Balb/c mice were labeled with CFSE and incubated with different TDCs at the ratio of 4:1. After 48h of incubation, FACS results showed that both TDC UCN and TDC UCN@L-GO exhibited robust inhibition of cytotoxic T cell proliferation caused by LPS with a decrease of 39.4% and 25.1%, respectively ( Figure 4A). Next, the cytotoxic function of CD8+ T cells was evaluated by CD44, CD69 and CD107a. FACS demonstrated that CD8+ T cells lost their cytotoxic functions after incubation with TDC UCN or TDC UCN@L-GO , for TDC UCN@L-GO suppressed the CD44/69/107a expression with a decrease of 79%, 75% and 64% compared with mature DC, respectively ( Figure 4B).
According to primary research, despite the inhibition of CD8+ T cells through PD-L1 expression 6 , TDCs also mediated CD80/86 combination with CTLA4, leading to Treg generation. The proportion of Treg cells was further investigated in the population of CD4+ T cells. Notably, more CD4+ T cells obtained Treg phenotype after TDC UCN or TDC UCN@L-GO pretreatment, which was characterized by a remarkably increase in CD25 (1.38-fold and 1.63-fold) and CTLA4 (3.42-fold and 4.18-fold) compared with mature DCs ( Figure  4C). Collectively, these results indicated that UCN@L-GO complex was a promising candidate in TDC generation that led to anergy of cytotoxic T cells and Treg expansion.
UCN@L-GO complex triggered TDCs cytoskeleton rearrangement to promote the homing ability in vivo Another important feature of TDCs is the ability to migrate, for they have to home to lymph nodes to produce a notable in uence. Despite the lateral size, another important discrepancy in physicochemical property between L-GO and S-GO is the content of carbon free radicals. According to electron paramagnetic resonance (EPR) results, more carbon free radicals would be measured on the surface of L-GO compared to S-GO, which can be remarkably blocked by FBS (Supplementary Figure 10). Thus, it stands to reason to expect more ROS production in the DC L-GO or TDC UCN@L-GO co-incubation group, which has been further veri ed by FACS analysis as shown in Figure 5A. This indicated that ROS was accumulated during interactions between GOs and plasma membrane. Moreover, the expression of CCR7 on DC membrane showed the same tendency as ROS after co-incubation ( Figure 5B). Collectively, we deduced that the adherence of L-GO to DC membrane was crucial to the activation of Rho A/ROCK pathway which existed in the downstream of ROS and participated in cell migration. Thus, the phosphorylation of Rho A/ROCK signals was further evaluated. L-GO ampli ed the phosphorylated levels of ROCK and MLC of DCs by 2.04-fold and 2.40-fold compared to the control DC. Although free UCN signi cantly suppressed p-ROCK and p-MLC, L-GO binding exhibited robust reversal ( Figure 5C&D). Because of the remarkable changes in p-ROCK and p-MLC, L-GO was considered an effective initiator of cytoskeleton rearrangement. Simultaneously, we observed a signi cant decrease in p-IKBα and p-P65 in TDC UCN and TDC UCN@L-GO groups, suggesting that TDCs remained the tolerant phenotype even in the presence of LPS.
Since cytoskeletal is the executor of cell migration, the rearrangement of DC cytoskeleton was observed by immuno uorescence (IF) in advance.
To simulate the in ammatory microenvironment in the process of GVHD, low-dose LPS was added to the co-culture system as a stimulus to TDC challenge. Obviously, L-GO materials had positive effect on the rearrangement of DC cytoskeleton characterized by increased expressions of F-Actin and β-tubulin. In contrast, free UCN showed a notable decrease, which might have been brought about by in ammatory inhibition. However, after L-GO binding, TDC UCN@L-GO displayed remarkable restoration of F-Actin and βtubulin expressions. To be more speci c, the uorescence intensity of F-Actin and β-tubulin in the TDC UCN@L-GO group was 2.31-fold and 1.42-fold higher than that of TDC UCN , respectively ( Figure 5E&F). In the absence of LPS, a similar tendency was also observed as shown in Supplementary Figure 11. Thus, the strong migration ability of TDCs induced by UCN@L-GO complex was expected. Combined with the Western blotting results in Figure 5C, it could be concluded that the GO-based cytomembrane-philic delivery platform participated in cytoskeletal rearrangement via ROS generation and ROCK phosphorylation.
Living cell imaging was carried out to track and analyze a time-laps cell migration pathway and length. After 6h of observation, it was found that the L-GO had a stronger ability to induce DCs migration and longer displacement in contrast to UCN ( Figure 5G). And the complex of UCN@L-GO could signi cantly restore the migration ability of TDCs impaired by UCN. L-GOs treatment also accelerated cell migration velocity. Unlike the disordered movements of the control DCs, DCs incubated with UCN@L-GO showed faster track velocity with a dominant motile mode of classical amoebic movement, which displayed a more e cacy migration mode ( Figure 5H). Furthermore, we employed the "footpad injection model" to con rm the ability of DCs to home to lymph nodes (LNs) after being co-cultured with UCN and UCN@L-GO. Mice were observed at time points of 24 h, 48 h and 72 h after footpad inoculation of Fluc + DCs using the uorescence imaging system. The migration rate of DCs was measured by the ratio of uorescence signals in PLNs (popliteal lymph nodes, PLNs) to the signals remaining on the foot pad. LPS was employed as a positive control. After L-GO treatment, DCs acquired a much stronger ability to home to PLNs than immature DCs, while free UCN treatment contributed to promotion. TDC UCN@L-GO exhibited a more signi cant migration rate than TDC UCN . Comparisons between the groups demonstrated that in the TDC UCN@L-GO group, the ratio of DCs migrating to PLNs at 24 h, 48 h and 72 h were respectively 17.75 ± 4.11, 19.75 ± 5.09 and 25.24 ± 5.07, which was the highest among these groups. (Figure 6A&B).
We also investigated the distribution of Fluc + DCs after intravenous administration in vivo. Fluorescence signals were detected using the uorescence imaging system at 4 h, 24 h and 48 h post injection. Results showed that the entire uorescence intensities were hardly different between these groups at different time points ( Figure 6C&D). Organs were collected at 72 h post injection and imaged by the uorescence system. The LPS group exhibited much stronger migration to lymphoid tissue than the control did. Since L-GOs treatment also facilitated the migration of DCs, TDC UCN@L-GO showed more enhanced migration to lymphoid tissues than free UCN treatment, as was illustrated by analysis of uorescence intensities ( Figure 6E). Statistical analysis also demonstrated that UCN@L-GO treatment accelerated cell migration to lymphoid tissues more signi cantly than free UCN co-cultured ones ( Figure 6F).
Moreover, the isolated organs were separated into two groups, solid organs and lymphoid ones. Results con rmed that TDC UCN@L-GO exhibited stronger migration capacity to spleen and lymph nodes than TDC UCN , which probably bene tted the interaction and regulation with T cells ( Figure 6G).

UCN@L-GO treated TDCs regulated the phenotype of donor T cells and protected recipients from lethal GVHD
An acute GVHD mouse model was established to help nd out whether the TDC UCN@L-GO could notably inhibit cytotoxic T cell proliferation or generate adequate Treg cells in GVHD relief.
To determine the amount of L-GOs that model mice would be exposed to or caused by DCs immunotherapy, FITC labelled L-GO or UCN@L-GO were incubated with DCs for 48 h. The supernatant and cells were collected separately for uorescence detection. Results in Supplementary Figure 12 illustrated that the residual L-GOs that adhered onto the cytomembrane of DCs were less than 30% of the incubated amount with the excessive L-GO or UCN@L-GO mainly in the cell culture medium. A widely accepted visualized GVHD murine model was established by infusion of a cell mixture composed by T cell-deleted allogenic bone marrow cells and Fluc+ T cells. And the entire uorescence intensity could represent the proliferation level of T cells in vivo. The group labeled as "BM+T" meant no successive immunotherapy treatment. After transplantation, 4×10 6 DCs, DC L-GO , TDC UCN and TDC UCN@L-GO were infused intravenously respectively (Supplementary Figure 13). On the 14th day post transplantation, imaging analysis showed the "BM+T" group had the highest uorescence intensity, which indicated massive activation and acute proliferation of T cells. Injection of DC L-GO resulted in little difference compared with the DCs group, while TDC UCN displayed a 2.37-fold decrease in uorescence intensity.
Furthermore, the TDC UCN@L-GO group showed 2.8-fold reduction compared with the TDC UCN group, suggesting the positive effect of L-GOs binding in inhibiting T cell proliferation ( Figure 7A&B). The LNs, lung, thymus, spleen, pancreas, liver, intestine and kidney were isolated and imaged. Results showed that the BM+T group had extensive T cell proliferation in multiple tissues, especially in the spleen, mesenteric lymph nodes (MLNs), thymus and small intestine (SI), which indicated rapid progress of GVHD. The infusion of control DCs or DC L-GO displayed relief to a certain extent. As expected, TDC UCN and TDC UCN@L-GO groups both exhibited a signi cant decrease in organ uorescence intensities, especially the latter, which induced apoptotic deletion of donor CD8+ T cells ( Figure 7C&D). ELISA was performed to analyze the cytokine levels in serum. It was found that after transplantation, such proin ammatory cytokines as IL-1β, IL-6 and MCP-1 were robustly increased, while the anti-in ammatory cytokines of IL-10 and TGF-β both signi cantly decreased compared to the irradiation control, which was a sign of the explosive in ammation in vivo. Infusion of control DCs and DC L-GO had slight effect on cytokine secretion, while TDC UCN and TDC UCN@L-GO induced an obvious decrease in these proin ammatory cytokines and a notable increase in the anti-in ammatory ones. ( Figure 7E).
TDCs also play non-redundant roles in promoting expansion and function of Tregs. As primary studies have showed, TDCs can induce T cell tolerance broadly via two categories: intrinsic and extrinsic. TDCs produce PD-L1 and CD80/86, which bind to PD-1 and CTLA-4 respectively, leading to the deletion of cytotoxic T cells and expansion of Tregs that can obstruct the proliferation and survival of those activated T cells. Moreover, TDCs can suppress the antigen-reactive T cells and promote Treg generation through IDO production 6 . We further isolated the MLNs and spleens to evaluate the proportion of Tregs. FACS showed that the proportion of Treg increased dramatically both in the spleen and MLNs after TDC treatment. According to the FACS results, TDC UCN@L-GO injection mediated 1.67-fold and 9.23-fold increases of CD4+CD154+ and 66.7% and 73.7% decreases of CD4+CTLA4+ in the spleen and MLNs separately, compared to BM+T group, which both corresponded to the immune suppressive T cells ( Figure  7F).
Cytotoxic T cell apoptotic deletion and Tregs generated by TDCs both played a vital role in GVHD relief by mitigating tissue damage. Hematoxylin and eosin (HE) staining of the involved organs con rmed the positive effect. The BM+T group was attacked by the donor T cells in multiple organs and had massive lymphocyte in ltration, which were symptoms of severe GVHD. DCs and DC L-GO infusion both showed limited remission, while TDC UCN and TDC UCN@L-GO displayed notable improvement of multiple organs involved such as the spleen, liver, MLN, thymus and SI ( Figure 7G, Supplementary Figure 14), especially the latter. Also, mice which received TDCs treatment, especially those in TDC UCN@L-GO group, showed remarkable abrogation in symptoms caused by severe GVHD, including hunched posture, skin and ear inerythema, alopecia and weight loss. Altogether, these results indicated that UCN@L-GO treatment was extremely e cient for TDC generation, which was probably the key to altering T cell subtype and the resultant GVHD relief in vivo.

Discussion
Lymph node-resident immune cells are usually chosen as the targeting subpopulation for immunoregulation of allo-response in allogeneic bone marrow transplantation (BMT), organ transplantation rejection or autoimmune disease treatment. However, it is generally di cult to ensure the drug can reach the targeted cells or locations with the effective concentration because of the limitation of toxicity or low solubility, even under accurate control in drug delivery. Alternatively, local delivery methods that can generate regulatory DCs or T cells ex vivo under appropriate conditions could serve as a second solution. When the nanotechnology-based strategy was introduced for the delivery of immunomediators, the advantages of adoptive immunotherapy were magni ed to some extent, for this approach required a smaller quantity of nano-carriers or drugs that entered the circulatory system, increasing the chance of favorable biocompatibility. To be more speci c, only a small amount of GO nanosheets (0.8 µg per mouse, less than 30% of the incubation amount) would enter the organism along with TDCs infusion, which increased the likelihood of transformation from bench to bedside.
Herein, the bone marrow derived dendritic cells from the recipient were selected as the source of TDCs, given that host APCs, which were major histocompatibility complex (MHC)-matched with host antigens, were reported to provide more effective protection than donor DCs, which were host-mismatched ones, from lethal GVHD, with a remarkable increase of recipient survival 30,31 . Meanwhile, preclinical data of murine models showed that donor DCs, activated by donor engraft, secreted rising levels of in ammatory cytokines which exacerbated the lethal feed-forward GVHD cascade 32,33 .
UCN, one of neuropeptides produced by different cells and tissues, is associated with distinct physiological functions. It was activated in response to in ammation, injury and pathogen clearance. UCN binds to CRHR2, a G-protein-coupled receptor, and is expressed in different immunocompetent cells. CRHR2 is coupled to the activation of the cAMP/PKA/CREB pathway, which is related to downregulation of several transduction pathways associated with transcription of in ammatory mediators, like proin ammatory cytokines secretion and co-stimulatory molecule expression. However, it is worth noting that CRHR2 has a relatively low abundance on DC membrane, which is the major barrier to TDCs induction by UCN treatment. The excellent performance of the UCN@L-GO as an immunomodulator in inducing TDCs is due to the unique manner in which they interact with DCs, namely adhesion or adsorption onto the plasma membrane, which was known as the cytomembrane-philic property in the current work. Receptor-ligand interactions, which are crucial to various pathophysiological processes, proved to be a kinetic equilibrium of combination and dissociation. Thus, regulation of the dynamic process might improve the integration e cacy of receptors and ligands. As has been veri ed in the interaction model of a biotin-labeled RBCs system, the presence of L-GOs and their preferred location at the edges of plasma membrane contributed effectively to the encountering and binding of biotin and avidin, especially when the surface abundance of biotin was relatively low (corresponding to the biotin dosage of 1, 2, and 4 µg), much like the expression level of cytomembrane receptors of UCN, namely CRHR2, on the surface of DCs (Supplementary Figure 8). This led to the conclusion that L-GO could enormously enhance TDC generation through non-speci c UCN delivery and promoted UCN-CRHR2 binding.
The migration toward draining lymph nodes is critical to DCs in their induction and maintenance of adaptive immunity. Generally speaking, their high mobility is readily observed under in ammatory or activated conditions. In other words, tolerant or other mediate-phenotype DCs are not usually capable of homing to secondary lymph-nodes because of the low expression level of CCR7 and the non-organized cytoskeletons, which are often accompanied by low surface costimulatory molecules and proin ammatory cytokine, such as IL-6, IL-12p 70 , IL-1β and TNF-α ( Figure 3E). Traditional systemic use of immunosuppressive drugs is facing such challenges, for it will impair immune function and cause side effects. According to previous research, TDCs generated by rapamycine showed poor performance in cell movements 34 . In order to obtain TDCs with enhanced homing ability of lymph nodes, other researchers have attempted to employ adenovirus as a vector to overexpress CCR7 on DCs to preserve immune tolerance simultaneously. But the limited e cacy and security concerns stand in the way. In the case of UCN@L-GO pulsed DCs, it was a pleasant surprise to nd the high expression level of CCR7 ( Figure 5B) as well as the promotion of the dynamic reconstruction of F-actin and β-tubulin in cytoskeleton ( Figure 5E). This could be attributed to the abundance of carbon free radicals in L-GO in comparison with their small counterpart, which led to the more production of reaction oxygen species (ROS) and Rho A/ROCK pathway activation as shown in Figure 5A&C. Such a conclusion is also supported by literature that says the mechanism by which exogenous H 2 O 2 promotes migration is due to the up-regulation of CCR7 receptor on the surface of DCs and activation of the downstream PI3K/Akt and NF-κB pathways. Our results also demonstrated notable phosphorylation in both IKb/NF-κb and Rho A/ROCK pathway after L-GO or UCN@L-GO incubation 35 . As a result, strengthened homing ability of the current TDCs to secondary lymphoid tissues was also observed (Figure 6). This might be one of the special advantages brought about by nano-carriers, or more precisely, by GO nanosheets over molecular biology-based methods.
Engineering nanomaterials, such as PLGA 36-38 , have emerged as newly reported carriers for immune modulatory agent delivery in transplantation rejection treatment, making it necessary to address the overall advantages of employing GO as carrier over other nanoparticles. The high surface area of GO allows the accumulation of immunomodulators through coupling, and thus the high density of neuropeptide appears in the adjacent region of their receptor. Moreover, unlike other nanomaterials, largesized GO nanosheets mediated the interaction related to non-speci c adhesion or adsorption onto the surface of cell plasma membrane, which is by no means passive phagocytosis. As it is generally believed that phagocytosis would promote the differentiation from immature DCs to a more mature phenotype, this distinctive property of GO meets the particular requirements for the generation of tolerant DCs.
Moreover, as has been discussed previously, exogenous carbon free radicals from L-GO might probably be the main contributor to the overexpression of CCR7 and rearrangement of cytoskeleton, which might account for the promoted homing ability of lymph nodes and the inhibition of effector donor T cells.

Conclusion
In this study, we designed novel immune-inducer of TDCs composed of UCN and L-GO which were capable of facilitating TDCs generation, and the immunotherapy of TDCs remarkably inhibited cytotoxic T cell activation in acute GVHD. The overall process was schematically presented in Figure 8. Unlike their small counterparts, large-sized GOs proved to be cytomembrane-philic and effectively promote the recognition between UCN and its receptor CRHR2 expressed on the membrane. Moreover, L-GOs also contributed to the organization of cytoskeleton via Rho A/ROCK pathway and elevation of CCR7 levels, which could enhance cell mobility and homing e cacy even under the tolerant phenotype. Adoptive immunotherapy of TDC UCN@L−GO exhibited excellent e ciency with promoted migration to lymph nodes, remarkable inhibition of cytotoxic T cell proliferation and statistically signi cant generation of Tregs, thus preventing the progress of GVHD. Finally, our study provides an e cient TDC inducer initiated by UCN@L-GO, which can be a promising candidate in immunotherapy of GVHD.

Declarations
Identi cation of GO location on cell membrane 2×10 6 Red blood cells (RBCs) were biotined with gradient S-NHS (Abcam. USA) and incubated for 60 min at 37℃.The biotinalyzed-RBCs were washed and suspended with 1 mL PBS, GO pretreated PE-avidin (Biolegend, USA) was added to the biotinalized RBCs and incubated for 60 min in dark. Next, cells were washed again and suspended with 200 μL binding buffer. PE positive rate and mean uorescence intensity were analyzed by CellQuest software.
2×10 6 DCs were seeded in six-well plate and co-cultured with FITC-GO for 48h. After incubation, cells were xed and stained with rhodamine. Laser confocal were employed to investigate GO location on DC membrane.
GO and UCN conjugation.
Neuropepitide was purchased from Jier Biochemical (Shanghai) Co., LTD, (amino acid sequence was provided in Table 1). 2×10 6 DCs were cultured with complete RPMI 1640 medium and incubated with 10 -5 M neuropeptide for 48h in six-well plates. After incubation, cell medium and DCs were separately Characteristic of tolerogenic DCs.
To investigate the location of UCN@GO with DCs, laser confocal was used after co-culture for 48h. After incubation with FITC-GO and 5-TAMRA-UCN, cells were xed and stained with Cell Tracker TM Blue CMAC (Invitrogen, USA).
After being co-cultured with UCN@GO, DCs and cell medium were collected separately. Next, cell costimulatory marker and cytokines were detected as described above.
In vitro cell migration analysis DCs derived from GFP transgene mice were isolated and co-cultured with UCN@GO complex and timelapse cell migration observation was performed using a living cell imaging system (PerkinElmer, Massachusetts, USA). Then cell tracking data were analyzed by Velocity software including track length and track velocity.
Cytoskeleton rearrangement was measured after different treatments by immuno uorescence. Cells in each group were xed and permeabilized before stained with anti-F-actin (Abcam, USA) and anti-β-tubulin (Abcam, USA) overnight. Next, cells were washed and conjugated with AF-488 and AF-594 IgG (Abcam, USA) for 2h in dark, separately. The cells were then mounted using uoroshield mounting medium with DAPI (Abcam) and observed under a laser confocal microscope (PerkinElmer, Massachusetts, USA).

Bioluminescence imaging of DCs migration and distribution in vivo
We investigated bioluminescence imaging with an IVIS system (PerkinElmer, Inc.) to re ect in vivo cell migration and distribution after different treatment. For DCs migration measurements, Fluc+ DCs were injected through footpad. Imaging data were collected and cell migration rate was calculated as the signal intensity ratio of the PLN to the entirety. For DCs distribution, Fluc+ DCs were intravenously injected to wild type C57BL/6J mice and observed at 4h, 24h, 48h and 72h. Mice fur was depilated to improve light speci c. Mice were scari ed 72h after injection, tissues were dissected and imaged for 1 min to obtain photographs.

DCs treatment for allo-HSCT mouse model
An Allo-HSCT mouse model was established as an acute GVHD model to investigate different DC treatments. DCs were pretreated with UCN@GO complex for 48h. Fluc+ lymphocytes were isolated from the spleen of FVB Fluc+ Balb/c mice. Wild type C57BL/6J mice were chosen as recipients which were depilated and received 8Gy Co 60 irradiation. Fluc+ lymphocytes and DCs with different treatments were intravenously injected to the recipients while the activation of transplanted T cells was re ected by the uorescence intensity. Mice were sacri ced 14-day post injection, serum and tissues were collected for therapy e cacy evaluation. Tissues were dissected and imaged to monitor the activity of transplanted T cells in each organ. Next, all tissues were xed and sectioned for hematoxylin-eosin (HE) staining. Serum cytokines, like IL-1β, IL-6, MCP-1, IL-10 and TGF-β, were measured by ELISA according to the instruction.

Statistical analysis
Data were analyzed using SPSS (Version 20.0). Dunnett's t-test or one-way ANOVA analysis of variance was used to analyze the normally distributed data. Non-parametric testing was used to analyze the nonnormally distributed data. p< 0.05 was considered to indicate a signi cant difference. Table   Table 1 is not available with this version.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.