Leukocyte reduction system (LRS) cones and thrombocyte apheresis cassettes (TACs) were retrieved from healthy adults undergoing thrombocytapheresis at the Department of Transfusion Medicine and Hemostaseology of the University Hospital Erlangen. This study was performed with the informed written consents of all donors in accordance with the Declaration of Helsinki (approved by the local ethics committee [Ethikkommission der Friedrich-Alexander-Universität Erlangen-Nürnberg]; ethics vote 346_18 B). Peripheral blood mononuclear cells (PBMCs) were isolated from LRS cones and TACs as described before (46, 47). Briefly, the blood product was extracted and diluted with PBS. Subsequently, 20 ml of diluted blood product was overlaid onto 14 ml of Lymphocyte Separation Medium (Anprotec) and centrifuged for 20 min with 520 x g at room temperature without brakes. Then, the interphase containing the mononuclear cells was transferred to a 50 ml tube and washed twice with phosphate-buffered saline (PBS). After washing, cell numbers were determined using a Luna-FL Automated Fluorescence Cell Counter (Logos Biosystems) and the cells used for the experiments.

For experiments with human dendritic cell (DC) subpopulations, DCs were isolated from PBMCs of healthy adults by cell sorting as described before (16, 31, 46). Briefly, PBMCs were resuspended in PBS + 2% fetal calf serum (FCS) + 1 mM EDTA (EasySep Buffer) with a concentration of 1 x 10⁸ cells/ml. Up to 9 ml of cell suspension were transferred to 14 ml roundbottom tubes and enriched using the EasySep Human Pan-DC Pre-Enrichment Kit (Stemcell Technologies). Subsequently, enriched DCs were stained with a panel of fluorochrome-coupled antibodies and stained on ice for 30 min. After washing, cells were resuspended in EasySep Buffer + DAPI (100 ng/ml) and sorted into sterile FACS tubes using a BD FACSAria II cell sorter with a 70 µm nozzle. DCs were defined as Lin^- (CD3/CD19/CD20/CD56 and CD14/CD16) and sorted as cDC1 (HLA-DR⁺CD141⁺CD11c^intCD1c^-CD123^-), cDC2 (HLA-DR⁺CD1c⁺CD11c⁺CD64^-CD163^-CD123^-), DC3 (HLA-DR⁺CD1c⁺CD11c⁺CD64⁺CD163⁺CD123^-), and pDC (HLA-DR⁺CD123⁺CD141^intCD1c^-CD11c^-) as shown in Figure 1 .

For in vitro treatment with ECP, sorted human primary DC subpopulations were resuspended in DC medium (RPMI-1640 + 10% human serum type AB + 1% L-Glutamax + 1% Penicillin/Streptomycin + 1% Na-Pyruvat + 1% non-essential amino acids + 1% HEPES) and seeded in sterile 96-well plates (V-bottom). Then, either 400 ng/ml 8-MOP or solvent control (equal volume ethanol) were added and the cells incubated at 37°C for 30 min. After incubation, plates were either irradiated with 2 J/cm² UV-A light (BIO-LINK BLX-365 irradiation chamber) or MOCK treated (48). After centrifugation for 10 min at room temperature with 300 x g, supernatant was removed and the cells resuspended either in DC medium (steady state) or in DC medium + 5 µg/ml R848, 5 µg/ml pIC, 100 ng/ml CRX-527, or 2.5 µM CpG (ODN2216) (inflammatory conditions). Cells were cultured for different time points (18 h, 42 h) before flow cytometric analysis and harvesting of supernatants for cytometric bead assay (CBA) analysis. For flow cytometric analysis, cells were stained with BV605-coupled anti-CD1c (clone: L161, BioLegend), PE/Cy7-coupled anti-CD11c (clone: 3.9, BioLegend), A647-coupled anti-CD64 (clone: 10.1, BioLegend), BV510-coupled anti-CD123 (clone: 6H6, BioLegend), BV421-coupled anti-CD141 (clone: M80, BioLegend), PE/Dazzle 594-coupled anti-CD163 (clone: GHI/61, BioLegend), and APC/Cy7-coupled anti-HLA-DR (clone: L243, BioLegend) for identification of DC subpopulations. In addition, they were stained with either A700-coupled anti-CD40 (clone: 5C3, BioLegend), FITC-coupled anti-CD86 (clone: Bu63, BioLegend), BV650-coupled anti-PD-L1 (clone: 29E2A3, BioLegend) or respective isotype controls for 30 min on ice. After washing, cells were stained with PE-coupled Annexin V (BioLegend) and 7-AAD (BioLegend) for 20 min on ice. After washing, cells were acquired using a Cytoflex S (Beckman Coulter) and analyzed using FlowJo Software (V10). Supernatants of the cells were stored at -80°C until analysis for secreted cyto- and chemokines by CBA using the LEGENDplex Human Macrophage/Microglia Panel (BioLegend). Supernatants were thawed and concentrations of the cytokines IL-12p70, TNF-α, IL-6, IL-4, IL-10, IL-1β, Arginase, CCL17 (TARC), IL-1RA, IL-12p40, IL-23, IFN-γ, and CXCL10 (IP-10) determined as described by the manufacturer. Subsequently, samples were acquired using a Cytoflex S (Beckman Coulter) and analyzed using the LEGENDplex Data Analysis Software Suite (Qognit). Then, the data were normalized based on the highest measured value (median of six/five individual values for each condition) in the complete dataset for each cytokine and plotted as heatmap. The maximal measured median value was given as reference in the figure. Based on the color code (% of max. values) and the max. value measured in the dataset, the median concentration for each subset and condition can be determined.

In order to perform co-cultures of DC subpopulations and autologous memory T cells, cDC1, cDC2, DC3, and pDC were sorted from healthy blood donors as described above. Autologous memory T cells were isolated from the same blood donor by negative enrichment using the MojoSort Human CD3 T Cell Isolation Kit (BioLegend) with addition of biotinylated anti-CD45RA antibody (clone: HI100, BioLegend) to deplete naïve T cells. To measure proliferation of T cells, isolated memory T cells were labeled with 5 µM CFSE (BioLegend) for 15 min at 37°C prior to the co-culture.

Before the co-culture, DCs were incubated with 400 ng/ml 8-MOP or solvent control for 30 min at 37°C. Then, DCs were irradiated with 2 J/cm² UV-A light (BIO-LINK BLX-365 irradiation chamber) or MOCK treated. After centrifugation for 8 min with 300 x g at 30°C, supernatant was removed and DCs resuspended in medium either with 1 nM CEFT peptides for presentation on MHC-I and MHC-II molecules (JPT petide technologies), 1 nM CEFT peptides + 1 µg/ml R848, or 10 CFU/DC heat-killed E. coli (Invivogen). After 18 h at 37°C, autologous CFSE⁺ memory T cells were added in a 1:10 DC:T cell ratio and co-cultured for five days. After the co-culture, supernatants were stored at -80°C until analysis for cytokine secretion using the LEGENDplex Hu Th Cytokine Panel (BioLegend). The T cells were analyzed by flow cytometry for proliferation (dilution of CFSE signal) as well as phenotype (activation and exhaustion markers). Therefore, cells were stained with BV510-coupled anti-CD3 (clone: OKT3, BioLegend), APC/Cy7-coupled anti-CD4 (clone: OKT4, BioLegend), PE/Cy5-coupled anti-CD8a (clone: HIT8a, BioLegend) as well as either APC-coupled anti-CD25 (clone: BC96, BioLegend), PE-coupled anti-CD71 (clone: CY1G4, BioLegend), A700-coupled anti-CD197 (clone: G043H7, BioLegend), BV605-coupled anti-CD223 (clone: 11C3C65, BioLegend), as well as PE/Cy7-coupled anti-CD178 (clone: NOK-1, BioLegend) or respective isotype controls for 30 min on ice. After washing, cells were stained with 100 ng/ml DAPI (Carl Roth) for 5 min on ice. Then, cells were acquired using a Cytoflex S (Beckman Coulter) and analyzed using FlowJo Software.

Statistical analysis was performed in GraphPad Prism (V10) using 2way ANOVA for grouped data with Dunnett’s multiple comparisons tests as posthoc test. Individual symbols were used for each donor. In figures showing data from the same timepoint, donors can be traced by these individual symbols (all Figures for 18h time point, all Figure for 42h time point as well as all Figures and all Supplementary Figures for memory T cell assay).

ECP is an immunomodulatory treatment inducing apoptosis in lymphocytes as well as modulating the immune response towards TH1 in CTCL or Tregs in GvHD. In contact hypersensitivity in mice, ECP-treated DCs are sufficient to transfer the tolerogenic effect of ECP (45). However, it is unclear how primary human DCs, which are present in the photopheresate of patients, respond to the treatment with 8-MOP and UV-A light irradiation. As DCs are the main regulators of T cell responses, we speculated that ECP affects DCs directly. Due to the scarcity of DCs, we decided to perform in vitro ECP of sorted human primary blood DCs from healthy blood donors. Therefore, we enriched all human DC subpopulations from blood of healthy donors by negative enrichment followed by cell-sorting as described recently (31, 46). For cell sorting, enriched DCs were stained with a panel of fluorochrome-coupled antibodies ( Table 1 ) and sorted into CD141⁺CD11c^int cDC1, CD1c⁺CD11c⁺CD64^-CD163^- cDC2, CD1c⁺CD11c⁺CD14^-CD64⁺CD163⁺ DC3 (CD14^- DC3), and CD123⁺CD141^intCD11c^- pDC ( Figure 1 ). When we cultured the DCs for 18h at 37°C, the phenotype of the DC subpopulations was stable: cDC1 showed expression of CD141 and CD11c but lacked markers of the DC2 lineage ( Supplementary Figure 1 ). Both cDC2 and CD14^- DC3 remained CD1c⁺CD11c⁺. While they still could be differentiated based on CD163 and CD64 expression, the signal was weaker compared to freshly isolated DC3 ( Supplementary Figure 1 ). Further, pDC could be identified by CD123 and CD141 expression as during cell sorting and did not express DC2-associated markers such as CD1c ( Supplementary Figure 1 ). For the in vitro treatment with ECP, the purified DC subpopulations were incubated with 400 ng/ml 8-MOP or ethanol as solvent control for 30 minutes at 37°C followed by irradiation with 2 J/cm² UV-A or mock treatment. Subsequently, the medium containing 8-MOP or ethanol was replaced with fresh medium to mimic reinfusion of the photophoresate into the patients resulting in strong dilution of the photophoresate. The DCs were then either cultured in presence of the TLR ligand R848, as a model for DC activation in inflammatory conditions, or in medium to mimic steady state conditions. Since ECP induces apoptosis in lymphocytes such as T cells, we analyzed the induction of cell death using flow cytometry by staining the cells with Annexin V and 7-AAD to distinguish between early (Annexin V⁺/7-AAD^-) and late apoptosis (Annexin V⁺/7-AAD⁺) as well as necrosis (Annexin V^-/7-AAD⁺) ( Supplementary Figure 2 ). After 18 h of culture, we observed induction of apoptosis in cDC1 and cDC2 in steady state conditions as well as in cDC1, cDC2, and CD14^- DC3 after TLR stimulation ( Figure 2 ). Since ECP has been shown to influence the phenotype of immune cells such as monocytes (49, 50), we were interested in how in vitro ECP would modulate the expression of co-stimulatory and -regulatory molecules and the secretion of cyto- and chemokines in steady state as well as upon TLR stimulation. Therefore, we analyzed the living DCs (Annexin V^-/7-AAD^-) for the expression of co-stimulatory (CD40, CD86) and -regulatory (PD-L1) molecules by flow cytometry ( Supplementary Figure 3 ), while the supernatants of the cells were collected for analysis of secreted cyto- and chemokines. After 18h, we observed only minor changes in the expression of co-stimulatory and -regulatory molecules ( Figure 3 ). In steady state, 8-MOP/UV-A-treated cDC1 showed enhanced expression of CD40, whereas upon TLR stimulation CD86 and PD-L1 were enhanced on in vitro ECP-treated cDC1 ( Figure 3 ). The phenotype of CD14^- DC3 was more pro-inflammatory under ECP conditions both in steady state (CD86) as well as after stimulation with R848 (CD40 and CD86), whereas ECP-treated pDC showed enhanced expression of the immunoregulatory molecule PD-L1 after TLR stimulation ( Figure 3 ). The surface phenotype of cDC2 remained largely unchanged except for increased CD40 expression after ECP in presence of the TLR ligand R848 ( Figure 3 ). In order to analyze whether these changes were specific to stimulation of TLR7/8, we used the TLR3 ligand pIC for cDC1, the TLR4 ligand CRX-527 for cDC2 and CD14^- DC3, and the TLR9 ligand CpG for pDC. Then, we analyzed the induction of cell death as well as co-stimulatory and -regulatory molecule expression as before. As with R848, we observed induction of apoptosis in cDC1 and CD14^- DC3 after TLR3 and TLR4 stimulation, respectively, whereas the cell death of cDC2 were rather dependent on UV-A light irradiation ( Supplementary Figure 4A ). Except for upregulation of PD-L1 on DC3 and pDC, we did not observe changes in the expression of co-stimulatory or -regulatory molecules ( Supplementary Figure 5A ).

As ECP was reported to induce spontaneous release of IL-10 by myeloid CD1c⁺ DCs isolated from the photophoresate of refractory chronic GvHD patients (51), we were interested in how in vitro ECP would affect the secretion of cyto- and chemokines by sorted DCs. We determined the concentration of secreted cyto- and chemokines in the supernatants of the treated DCs by CBA using the LEGENDplex Human Macrophage/Microglia Panel. As expected, cDC1 were the main producers of IL-12 family members, i.e., IL-12p70, IL-12p40, and IL-23, after stimulation with R848, while cDC2 and CD14^- DC3 excelled in the secretion of the pro-inflammatory cytokines IL-6, TNFα, and IL-1β as well as the anti-inflammatory cytokines IL-10 and IL-1RA ( Figure 4A ). In contrast, pDC were the main producers of the chemokine CXCL10 (IP-10) that is dependent on signaling of interferon response factors (IRFs) ( Figure 4A ). When analyzing the impact of in vitro ECP, we did not observe induction of pro- or anti-inflammatory cytokines in steady state conditions and only slight changes on the secretion of cyto- and chemokines after TLR stimulation (cDC2: TNFα↑, IL-1β↑, IL10↑, IL-1RA↑; CD14^- DC3: IL-10↓; pDC: CXCL10↓; Figure 4A ).

In order to monitor how ECP would affect DCs at a later time point after reinfusion of the photophoresate, we performed the same analyses after 42 h of incubation at 37°C. The majority of DCs were apoptotic or necrotic in steady state conditions, but ECP reduced further the survival of CD14^- DC3 and pDC ( Figure 5A ). The low viability of the different DC subpopulations is in accordance with data on the circulating lifespan of human DCs in vivo showing a shorter lifetime of cDC1 (1.3 days) compared to DC2 and DC3 (2.2 days) (52). However, after TLR stimulation the survival of cDC1 (R848 and pIC) and pDCs (R848 and CpG) was strongly increased in control conditions, which was completely abrogated when cDC1 and pDC were treated with ECP in vitro ( Figure 5B , Supplementary Figure 4A ). In contrast, we did not observe enhanced cell death in cDC2 and CD14^- DC3 after in vitro ECP after stimulation with R848 ( Figure 5B ). As R848 can activate the NLRP3 inflammasome in human CD14^- DC3 (31), we used the TLR4 ligand CRX-527 to avoid the induction of cell death by the TLR ligand alone. Notably, in vitro ECP strongly induced cell death in CRX-527-activated CD14^- DC3 and to a minor extend in cDC2 ( Supplementary Figure 4B ). Thus, in vitro ECP seems to reduce the viability of cDC1 and pDC in general in response to TLR ligands, whereas it depends on the stimulated pattern recognition receptor for human cDC2 and CD14^- DC3. We also measured co-stimulatory and -inhibitory molecule expression by flow cytometry but did not observe strong changes on ECP-treated DCs compared to the control treatment ( Figure 6 ). cDC1 showed enhanced expression of immunoregulatory PD-L1 in steady state conditions, whereas PD-L1 was enhanced on pDC after in vitro ECP upon stimulation irrespective of the TLR ligand ( Figure 6 , Supplementary Figure 5B ). CD14^- DC3 showed slightly increased expression of the co-stimulatory molecule CD86 after R848 stimulation, when they were treated prior with ECP ( Figure 6 ). The analysis of cytokine secretion showed that anti-inflammatory IL-1RA released by cDC2 and CD14^- DC3 and IL-10 released by cDC2 were increased by ECP upon TLR stimulation ( Figure 4B ). Overall, changes on phenotype and cytokine secretion by DCs due to ECP were low. Thus, in vitro ECP of human DC subpopulations showed minor effects on the phenotype of the DCs but strongly induced apoptosis in cDC1 and pDCs after activation by TLR ligands as well as in CD14^- DC3 dependent on the used TLR ligand.

Since it has been reported that ECP induces amelioration of different T cell-mediated diseases, such as GvHD, we were interested in whether in vitro ECP would influence the T cell stimulatory capacity of human DCs. As ECP affects preexisting immune responses in patients, we analyzed the T cell stimulatory capacity of DCs by co-culturing them with autologous CD4⁺ and CD8⁺ memory T cells in the presence of common viral and bacterial antigens. DCs were sorted after negative enrichment and memory T cells were purified from PBMCs of the same donor by magnetic bead-based enrichment kits. Then, sorted DCs were treated with in vitro ECP as described above. They were either loaded with a pool of pre-processed MHCI- and MHCII-specific peptides derived from common antigens (CEFT: CMV, EBV, Influenza, and Tetanus Toxoid) which can be loaded on a broad array of HLA molecules with or without additional stimulation with the TLR ligand R848 as it strongly activates all human DC subpopulations. As loading with peptides is a passive process that does not require processing of antigens, we were interested in how in vitro ECP would influence the capacity to process antigens. Here, the sorted DCs were incubated with heat-killed E. coli that have to be phagocytosed and processed in order to present peptides on HLA molecules to T cells. Further, as E. coli is a complex pathogen, it is able to activate diverse pattern recognition receptors, such as TLR2, TLR4, and TLR8 (53–57). Thereby, influence of the activated TLR on cell death induction might be minimized. After 18 h of culture with the peptides or E. coli, CFSE-labelled memory T cells were added and co-cultured with the sorted DCs for five days. Then, T cells were analyzed by flow cytometry for proliferation (dilution of CFSE signal), activation (expression of CD25 and CD71), and phenotype (expression of CD178, CD223, and CCR7) ( Supplementary Figure 6 ). When DCs were loaded with CEFT peptides in steady state conditions, the DCs could hardly activate memory CD4⁺ and CD8⁺ T cells and we did not observe any influence of in vitro ECP on the induction of memory T cell proliferation ( Figure 7A ). Simultaneous stimulation and peptide loading (CEFT + R848) boosted the capacity of cDC1 and pDC to induce memory T cell proliferation but in vitro ECP showed only a minor reduction by cDC1 and pDC to activate memory T cells ( Figure 7B ). However, when DCs were cultured with heat-killed E. coli, which have to be phagocytosed and processed in order to restimulate E. coli-reactive memory T cells, we observed a strong decline in proliferated memory CD4⁺ and CD8⁺ T cells when cDC1 were exposed to in vitro ECP ( Figure 7C ). ECP also reduced the capacity of pDC to activate memory CD4⁺ T cells ( Figure 7C ). While in vitro ECP did not influence the capacity of cDC2 and CD14^- DC3 to induce T cell proliferation ( Figure 7C ), we observed changes of the phenotype of the proliferated CD4⁺ T cells ( Figure 8 ). T cells stimulated by steady state cDC2 and CD14^- DC3 showed lower expression of the activation markers CD25 and CD71 as well as of the exhaustion marker CD223 (LAG-3), when DCs were treated with ECP prior to the co-culture ( Figure 8A ). This was also the case when CD14^- DC3 were stimulated with heat-killed E. coli prior to the co-culture with memory T cells ( Figure 8C ). As ECP modulates the polarization of T cells (8, 10), we were interested whether ECP-treated DCs influence the secretion of cytokines by activated T cells. Therefore, we analyzed the supernatants of DC:T cell co-culture for cytokines associated with different subsets of T helper cells by CBA assay. When DCs were only loaded with CEFT peptide without TLR stimulation, levels of secreted cytokines were low and not influenced by ECP-treatment ( Supplementary Figure 7 ). When DCs were simultaneously activated during peptide loading, we observed higher level of TH1-associated cytokines such as IFNγ but they were not influenced by the treatment with ECP ( Supplementary Figure 8 ). However, in accordance with the T cell proliferation data ( Figure 7C ), we observed a strong reduction in TH1- and TH17-associated cytokines (IFNγ, IL-22, IL-17A, and IL-17F) as well as IL-2, when cDC1 were stimulated with E. coli after pretreatment with in vitro ECP ( Figure 9 ). In contrast, in vitro ECP did not influence the secretion of cytokines when T cells were co-cultured with cDC2, CD14^- DC3 and only slightly with pDC ( Figure 9 ). Further, in vitro ECP did not influence the secretion of TH2- (IL-4, IL-5, IL-13) or Treg-associated cytokines (IL-10) irrespective of the DC subset and the antigen ( Figure 9 , Supplementary Figures 7 , 8 ). Thus, in vitro ECP of human primary blood DCs directly influences the capacity of cDC1 and pDC to induce memory T cell activation as well as changes the phenotype of memory T cells activated by cDC2 and CD14^- DC3. Further, it reduces the secretion of TH1- and TH17-associated cytokines by T cells, when restimulated with ECP-treated cDC1.