Human Regulatory Dendritic Cells Develop From Monocytes in Response to Signals From Regulatory and Helper T Cells

Dendritic cells (DCs) are powerful antigen presenting cells, derived from bone marrow progenitors (cDCs) and monocytes (moDCs), that can shape the immune response by priming either proinflammatory or tolerogenic immune effector cells. The cellular mechanisms responsible for the generation of DCs that will prime a proinflammatory or tolerogenic response are poorly understood. Here we describe a novel mechanism by which tolerogenic DCs are formed from monocytes. When human monocytes were cultured with CD4+FoxP3+ natural regulatory T cells (Tregs) and T helper cells (Th) from healthy donor blood, they differentiated into regulatory DCs (DCReg), capable of generating induced Tregs from naïve T cells. DCReg exhibited morphology, surface phenotype, cytokine secretion, and transcriptome that were distinct from other moDCs including those derived from monocytes cultured with Th or with GM-CSF/IL-4, as well as macrophages (MΦ). Direct cell contact between monocytes, Tregs and Th, along with Treg-derived CTLA-4, IL-10 and TGF-β, was required for the phenotypic differentiation of DCReg, although only IL-10 was required for imprinting the Treg-inducing capacity of DCReg. High ratios of Treg:Th, along with monocytes and DCReg similar in function and phenotype to those induced in vitro, were present in situ in human colorectal cancer specimens. Thus, through the combined actions of Tregs and Th, monocytes differentiate into DCs with regulatory properties, forming a positive feedback loop to reinforce Treg initiated immune regulation. This mechanism may contribute to immune tolerance in tissues such as tumors, which contain an abundance of Tregs, Th and monocytes.


INTRODUCTION
As one component of the "mononuclear phagocyte system" (MPS), monocytes constitute approximately 10% and 4% of the leukocytes in human and murine peripheral blood, respectively (1). Circulating monocytes can infiltrate into mucosal (2), inflammatory (3), cancer tissues (4), or draining lymph nodes (LNs) (5) and differentiate into either macrophages (M ) or dendritic cells (moDCs) (6). Monocytes are highly plastic and their differentiation is subject to the signals that they receive (7), which enables the cells to acquire distinctive features to promote or hamper immune responses (1). We previously observed that human CD4 + Th cells and monocytes frequently interact with one another in inflamed tissues of patients with autoimmune and allergic disease, and demonstrated that such interactions result in the differentiation of monocytes into pro-inflammatory immunogenic moDCs (DC Th ), which in turn induce the formation of Th effector cells from naïve T cells (8,9). The differentiation of DC Th from monocytes occurs in a cell contact, GM-CSF and TNFα dependent manner (8).
Although numerous inflammatory moDC subsets have been identified in inflammatory environments (8)(9)(10)(11), little is known about the signals required to induce tolerogenic/immunoregulatory moDCs. Since regulatory T cells (Tregs) play a central role in promoting immune tolerance and maintaining immune homeostasis (10), we and others considered the possibility that they might directly induce regulatory moDCs (DC Reg ) from monocytes in a manner analogous to the induction of DC Th by Th cells. However, when monocytes are cultured with activated Tregs alone, they become macrophages (M ) (12). Since Tregs most often exert direct immune suppression on Th cells (11,13,14), we hypothesized that both Tregs and Th might be required for the generation of immune regulatory DCs from monocytes. To evaluate this hypothesis, we cultured classical human CD14 + monocytes with activated natural Tregs and Th from healthy donors. The results show that under these conditions monocytes differentiate into regulatory DC Reg with the capacity to induce the formation of immune suppressive CD4 + FoxP3 + Tregs. DC Reg are distinctive in their morphology, phenotype, cytokine secretion, and transcriptome. DC Reg similar in phenotype and function to those induced in vitro were present in situ in colorectal cancer (CRC), along with an abundance of monocytes, Tregs and Th cells. Therefore, our study reveals a novel mechanism by which Tregs can inhibit the immune response by inducing the generation of DC Reg .

Mixed Leukocyte Reaction (MLR)
CD14 + monocytes were cultured as indicated earlier and stimulated with 1 µg/ml LPS on day 3. After 18 h, cells were washed 3x in PBS and HLA-DR + CD2 − DCs were purified by FACS and incubated with allogeneic naïve CD4 + T cells (10 5 /well) at a DC to T cell ratio of 1:2 in the presence or absence of 2 µg/ml anti-TGFβ. The CD4 + T cells for these assays were purified from PBMCs using a RosetteSep CD4 + Human T cell Isolation Kit (Stem Cell Technologies) followed by magnetic purification (>95% CD2 + CD4 + CD45RA + cells by flow cytometry) using a Naïve CD4 + T Cell Isolation Kit II (Miltenyi Biotec) and subsequently labeled with CFSE. After 6 days, responder CD4 + T cells were analyzed for CD25 and FoxP3 expression. CD2 + CD4 + CD25 + CFSE − cells were further isolated by FACS and cocultured with 10 5 CFSE labeled allogeneic CD45RA + CD4 + naïve responder T cells in a new MLR with irradiated autologous DCs (5 × 10 4 ), which had been purified with human CD11c microbeads (Miltenyi) and anti-CD3 (0.5 ng/ml, plate-bound).

Luminex Assays
Human Luminex 63-plex assays were performed by the Human Immune Monitoring Core at the Stanford Institute for Immunity, Transplantation and Infection. Samples were run in duplicate, and data were analyzed with GraphPad Prism6 (GraphPad-PRISM, Inc). Error bars represent SEM. Statistical differences for the mean values are indicated as follows: * P < 0.05; * * P < 0.001; * * * P < 0.0001; ns, not significant.

Microarray Analysis
DC Th , DC Reg , M Treg and DC GM were sorted on a FACSAria (BD) and RNA was extracted with an RNeasy Micro Kit (Qiagen). Total RNA samples were sent to Stanford Functional Genomics Facility and microarray was performed on GeneChip Human Gene 2.0 ST Array from Affymetrix. Microarray data were analyzed using the oligo (13) and annotated with data base hugene20sttranscriptcluster.db (14) in Bioconductor. limma package (15) was applied to identify the differentially expressed genes among cell populations. P-values were adjusted with false discovery rate (FDR) and genes with FDR adjusted P value < 0.05 were selected to be differentially expressed genes. Principal component analysis on the differentially expressed genes was performed using prcomp in R and plotted with ggplot2 package (version 2.2.1) (16). Heatmaps were generated using unsupervised clustering in the pheatmap package (version 1.0.8) (17). Biological functional gene ontology analysis was done with topGO (18). All the above data analyses were performed in R (version R 3.3.2 1 ). Pairwise gene set enrichment analyses (GSEA) were applied for assessment of the similarity of DC Th , DC Reg , M Treg and DC GM with the well-defined reference gene signatures by pairwise transcriptomes comparison. GSEA was done using Bubble Map module of Bubble GUM version 1.3.19 (19). The cell-specific gene fingerprints from previous published reports were selected as references: six gene sets of human DC subsets (DC1, DC2, DC3, DC4, DC5, DC6) from Science (20); one gene set of Macrophage reference gene signature generated by Segura et al. (21); and another gene set of VITD3 DCs (22). The BubbleMap returned a bubble map of pairwise GSEA results to show enrichment of a given gene set (reference) in a pairwise comparison. The normalized enrichment score (NES) and corrected P-value (FDR) of each bubble were also generated 1 https://www.r-project.org/ (19). Enhanced Volcano and ggplot2 were used for volcano plots using gene expression and adjusted p value.

Processing of Human Colon Cancer Samples
Fresh CRC tissues were obtained from the Stanford Tissue Bank in accordance with IRB protocol 6304 following surgical resection of primary tumors. Some of the tissues were minced with surgical scissors and transferred to 15ml scintillation vials containing 2 mg/ml collagenase type IV (Worthington Biochemical Corporation) and 50 U/ml DNase I (Roche) for 30min at 37 • C with constant agitation. Samples were subsequently filtered through a 70 µm filter and resuspended in PBS containing 2% human serum and 2mM EDTA. CD4 + T cells were isolated from the cell suspension with CD4 MicroBeads (Miltenyi Biotec) and further purified by FACS. DCs were FACSsorted from the CD4 negative population. Samples of the same tissues were sectioned, stained and analyzed as described. The primary antibodies used were rabbit polyclonal anti-CD14 (1:200; Atlas), mouse monoclonal anti-FoxP3 (1:100) and rabbit antimouse T-bet (1:100; Santa Cruz Biotechnology).
Human blood collection to obtain PBMCs and CRC collection were approved by the Stanford Research Compliance Office and performed according to institutional guidelines under Stanford IRB protocol 6304, "Dendritic and T cell Signaling in Gastrointestinal Cancer." All sequencing data are publicly available from NCBI's Gene Expression Omnibus at GEO accession GSE148114.
See Supplemental Experimental Procedures for transmission electron microscopy, endocytosis and phagocytosis, CFSE labeling, and RNA isolation and quantitative RT-PCR.

RESULTS
In the Presence of Both Tregs and Th, Human Monocytes Differentiate Into Regulatory DCs That Induce CD4 + FoxP3 + Tregs To test our hypothesis, freshly isolated CD14 + monocytes from the blood of healthy donors were cultured with Th alone, Tregs alone or with Tregs and Th cells at a ratio of 1:1. Prior to culture, each population had been sorted to > 99% purity (Supplementary Figure S1A) and the Tregs shown to suppress Th proliferation (Supplementary Figure S1B). After 3 days of culture, the myeloid cells (HLA-DR + CD2 − ) were sorted (>99% purity), activated with LPS and incubated for 6 days with CFSE-labeled naïve CD4 + T cells. While monocytes that had been cultured with Tregs and Th at 1:1 elicited weaker CD4 + T cell proliferative responses (Figures 1A-C) compared with those cultured with Th or Tregs alone, a significant proportion of T cells that proliferated developed into CD25 high FoxP3 high Tregs (Figures 1D-F). In contrast, monocytes cultured with Th or Tregs alone failed to induce CD25 high FoxP3 high Tregs (Figures 1D-F), although Th cultured monocytes induced vigorous CD4 + T cell proliferation and CD25 expression. Treg  (G) CD14 + monocytes were cultured with Tregs and Th at a ratio of 10:1:1 in the presence of anti-IL-10, 2 µg/ml; anti-TGFβ, 2 µg/ml or isotype control antibody, 5 µg/ml. After 72 h, 1 µg/ml LPS was added, and 16 h later FACS-purified myeloid cells (DR + CD2 − ) were further cultured with MACS sorted CFSE-labeled naïve allogeneic CD4 + T cells. The percentage of FoxP3 high cells in CD2 + CD4 + CFSE − cells was determined 6 days later. Mean ± SEM, n = 5. Similar results were obtained in two additional experiments. *P < 0.05; **P < 0.001; ***P < 0.0001; ns, not significant. Data were analyzed with one-way ANOVA, followed by Dunnett's test for multiple comparisons. (H,I) 10 5 CFSE labeled CD45RA + CD4 + T cells were cultured with 5 × 10 4 allogeneic myeloid cells purified from 3-day cultures of monocytes, Tregs and Th at 10:1:1. After 6 days the cells were analyzed by flow cytometry. Responder CD4 + T cells were gated as CD2 + CD4 + . differentiation was dependent on TGFβ, but not IL-10, since addition of neutralizing anti-TGFβ antibody, but not neutralizing anti-IL-10 antibody, during the MLR abrogated FoxP3 expression by proliferating T cells ( Figure 1G).
To confirm that the FoxP3 + T cells induced from monocytes cultured with Th and Tregs at 1:1 were immunosuppressive, we FACS purified CD25 high CFSE − T cells (∼50% FoxP3 + ) (Figures 1H,I) and tested their ability to inhibit the activation of naïve CD4 + T cells (Supplementary Figure S2) stimulated with autologous DCs and anti-CD3 mAb. The purified CD25 high CFSE − CD4 + T cells suppressed naïve T cell proliferation in a dose-dependent manner (Figures 1J-L), whereas purified CD25 + CFSE − CD4 + T cells obtained from the MLRs containing Th or Tregs alone had no suppressive effect ( Figure 1M). Collectively, these data demonstrate that in the presence of Th, Tregs from healthy donors promote the formation of CD4 + FoxP3 + Treg-inducing myeloid cells from monocytes.

Regulatory DCs Derived From Monocytes Following Their Culture With Tregs and Th Are Phenotypically and Functionally Distinct From DC Th and M Treg
Given the stark functional differences between the cells derived from monocytes cultured with Th alone and those cultured with mixtures of Th and Tregs, we decided to study these populations in greater detail. Consistent with a previous study (8), monocytes that had been cultured with Th alone developed prominent dendrites within 24h and these DC-like changes peaked at day 4; hence, we refer to these cells as DC Th . The addition of Tregs to fresh cultures of monocytes and Th, at a ratio of Th:Treg of 1:1, resulted in the formation of cells with fewer and shorter dendrites that were otherwise similar in appearance to the DC Th . Based on their DC-like appearance and FoxP3 + Treg-inducing capacity (Figures 1A-F), we refer to these cells as DC Reg (Figure 2A). Examination of the cells by electron microscopy confirmed that both DC T h and DC Reg had typical DC morphology with large nuclei, dense cytoplasm and numerous dendrites. In contrast, monocytes cultured with Tregs in the absence of Th developed numerous intracellular vesicles and a small nucleus (Figure 2B), characteristic of classical M , and we refer to these cells as M Treg .
We next analyzed the molecular phenotype of each cell population as an additional means to assess the differences and similarities between DC Reg , DC Th and M Treg . As expected, M Treg expressed high transcript levels of the M -specific receptor MerTK (23-25) ( Figure 2C) and high levels of CD14 and CD163 proteins. In contrast, DC Th expressed low levels of these molecules (Figures 2D,E). DC Reg exhibited low expression of MerTK and CD163 and intermediate expression of CD14 and CD209 (Figures 2D,E).
To analyze the cytokine secretion profiles of DC Reg , DC Th , and M Treg , each population was stimulated overnight with LPS, and selected cytokines were measured in the supernatants. LPS-stimulated DC Reg secreted nearly 1000-fold more of the antiinflammatory cytokine IL-10 than DC Th , but less than M Treg ( Figure 3F). Conversely, DC Reg produced significantly less IL-1β, IL-6 and TNFα than DC Th but more than M ( Figure 2F). Real-time PCR analysis of FACS purified myeloid cells from these cultures indicated that the RNA expression profiles matched the cytokine secretion patterns, with DC Reg expressing mainly IL-10 (Supplementary Figure S3 and Figure 4D) and only small amounts of inflammatory cytokines (Supplementary Figure S3).
To assess the endocytic capacity of these cells, we incubated each population with DQ-OVA, a self-quenched conjugate of OVA that emits green fluorescence after receptor-mediated internalization and proteolysis (26,27). After a 60 min incubation, M Treg had the highest level of DQ-OVA (green fluorescence) followed by DC Reg and DC Th (Figure 2G; left). Furthermore, when the phagocytic activity of these cells was assessed on the basis of E. coli particle uptake, DC Reg displayed strong particle uptake, second only to that of M Treg (Figure 2G; right).
Taken together, these results indicate that the phenotype, cytokine secretion profile and endocytic/phagocytic  (Figure 3B). DC Reg exhibited a transcriptome that was intermediate between that of DC Th and M Treg and closer to that of DC Th than M Treg (Figure 3B), which is consistent with their DC morphology ( Figure 2B) and cellsurface phenotype (Figures 2D,E).
We further analyzed the three myeloid populations by selecting differentially expressed genes corresponding to the biological functions of antigen processing (GO 0019882) (Figure 3C), chemotaxis (GO 0006935) (Figure 3D), and endocytosis (GO 0006897) (21) (Figure 3E). Hierarchical clustering of these gene sets showed that the antigen-processing signature of DC Reg was more closely related to that of DC Th than M Treg (Figure 3C). Also, DCs or M generated from T cell-monocyte cocultures were distinct from DC GM in their antigen-processing signature (Figure 3C), which is consistent with their respective total transcriptome analysis ( Figure 3B). By contrast, the chemotaxis ( Figure 3D) and endocytosis ( Figure 3E) transcriptomic pathways of DC Reg were more similar to those of M Treg than to DC Th , and distinct from those of DC GM (Figures 3D,E).
Recently, single-cell RNA sequencing (scRNA-seq) led to the identification of six human DC subsets (20): DC1 (CD141 + Clec9A + ), DC2 (CD1c + _A), DC3 (CD1c + _B), DC4 (CD1c − CD141 − CD11c + ), DC5/AS DCs (AXL + SIGLEC6 + ) and DC6 (pDCs). To address whether and where DC Reg fit in this new DC taxonomy, we used the method of Segura et al. (21) to compare the gene signatures of DC Reg , DC Th , M Treg and DC GM with those of these six DC subsets and macrophages (MACRO) (21) (Figure 3F). To display the data, we transformed the information for each pairwise gene set enrichment analysis (GSEA) (12) into a dot whose color corresponds to the cell in which the gene signature was more represented. The dot area size is proportional to the NES (Normalized Enrichment Score), and the color intensity indicates the p-value (a darker, larger dot indicates a stronger enrichment). The results show that DC Reg were relatively enriched for a DC1 (CD141 + Clec9A + ) signature when compared with M Treg , and a DC2 (CD1c + _A) signature when compared with DC Th . As expected, compared with DC Reg and DC Th , M Treg were more like conventional macrophages (MACRO) (21). Interestingly, DC Th were also enriched for the DC6 (pDCs) signature when compared with the other three populations. M Treg , DC Th , DC Reg and DC GM all showed variable similarity to DC1-4 and DC6 (Figure 3F), in contrast to 1,25-dihydroxyvitamin D3 (1,25(OH)2D3) induced tolerogenic DCs (VITD3DC) (12,22) (Figure 3F). When analyzing differentially expressed genes (DEG) between tolerogenic DC Reg and other T cell-induced DC Th and M Treg , we found that TLR4 was overexpressed in the DC Reg (Figure 3G). Interestingly, the DC Reg maintained their surface phenotype (Supplementary Figure S4), even when subjected to subsequent stimulation with LPS.
These results confirm that DCs differentiated from monocytes in the presence of CD4 + T cells are transcriptomically distinct from DCs differentiated from monocytes with GM-CSF and IL-4 (DC GM ), and further, that DC Reg are distinct from DC Th .

Cell Contact, TGFβ, IL-10, and CTLA-4 Mediate DC Reg Formation
To investigate whether direct contact between monocytes and Tregs or Th is required for DC Reg differentiation, we cultured these populations on opposite sides of 0.4 µm transwell membranes, which prevent cell-cell interaction but permit diffusion of soluble molecules (Figure 4A). Separation of Tregs from monocytes and Th resulted in the induction of DC Thlike cells (Figure 4B, Transwell 5). However, when Th were separated from monocytes and Tregs, M -like cells were induced ( Figure 4B, Transwell 7). Monocytes cultured in transwells and separated from Th or Tregs (Figure 4B, Transwell 6&8) retained their monocyte-like phenotype. Moreover, when monocytes were cultured in transwells and separated from both Th and Tregs, the monocytes became M Treg -like (Supplementary Figure S5). Thus, both Tregs and Th require direct contact with monocytes to mediate their effects on DC differentiation and the contribution of activated Th cells could not be replaced by exogenous GM-CSF and IL-4 (Supplementary Figure S5).
We next sought to determine whether Treg-derived molecules that are known to impact DC function in other settings are necessary for DC Reg differentiation (28). DC Reg induction cultures were performed in the presence of neutralizing mAbs against CTLA-4, IL-10 or TGFβ. Each of these mAbs partially prevented DC Reg formation, as indicated by the increased frequency of cells exhibiting a DC Th -like morphology with more extensive dendrites in these cultures compared to cultures containing isotype control antibody (Figure 5A). When all 3 neutralizing mAbs were used simultaneously (anti-all), the resultant cell morphology was comparable to that of DC Th (Figure 5A). We also analyzed the effects of each blocking antibody individually on the expression of DC-associated surface molecules. Blocking TGFβ affected the expression of several of these molecules, but only after blocking all 3 molecules was the surface phenotype of the resulting DCs comparable to that of DC Th (Figure 5B). These mAbs also altered surface marker expression in M induction cultures (Supplementary Figure S6). Taken together, these findings suggest that TGFβ, IL-10 and CTLA-4 govern the formation of DC Reg . To determine the impact of TGFβ, IL-10 and CTLA-4 on the acquisition of DC Reg function, we analyzed the cytokine secretion profile and Treg-inducing potential of FACS purified DC Reg formed under conditions in which one or more of these factors were neutralized. Neutralization of TGFβ or IL-10 increased the expression of proinflammatory cytokines IL-1β, IL-6 and TNFα by the DCs, while decreasing their expression of IL-10 ( Figure 5C). Importantly, neutralization of IL-10, but not TGFβ prevented the resultant DCs from acquiring the capacity to induce Foxp3 high T cells ( Figure 5D). Thus, IL-10 is required for the formation of functionally active DC Reg , while functionally active DC Reg utilize TGFβ to induce new FoxP3 + Tregs.

Human Colorectal Cancers (CRC) Contain DC Reg , and T Cells From the Same Tumors Induce DC Reg Formation From Monocytes
Since Tregs (29) and monocytes (30) are often abundant in human tumors, including CRC, where their phenotypes and functions are often altered (31), we tested the hypothesis that interactions between these cells might result in the generation of DC Reg . We utilized immunohistochemistry to identify Tregs and CD14 + cells in human CRC, and found that these cells were often in close apposition to one another ( Figure 6A; left), and conventional Th (T-bet + ) cells were also identified in the same regions (Figure 6A; right). Flow cytometric analysis of single cell suspensions prepared from 5 CRC patients indicated that CD25 + CD127 low Tregs comprised approximately 25% of total CD45 + CD2 + CD4 + T cells (Figure 6B), which is similar to that reported previously for CRC (20) and much higher than in surrounding uninvolved tissue (Supplementary Figure S7) or healthy donor peripheral blood (Figure 6C).
Given the relative ratios of monocytes, Tregs and Th in these tumors, we hypothesized that the conditions would favor formation of DC Reg . As a first step to investigate this possibility, we evaluated the ability of unfractionated tumor CD4 + T cells to induce DC Reg from monocytes, in vitro. Thus, FACS purified CD4 + T cells from CRC tissues were cultured with peripheral blood monocytes at a 1:10 ratio in the presence of an agonistic anti-CD3 monoclonal antibody (mAb). HLA-DR + CD2 − cells with DC morphology appeared within 4 days of culture initiation (Figure 6D). Consistent with the phenotype of DC Reg (Figures 2D,E), these cells expressed characteristic DC markers such as CD209, CD274, CD40, and CD86, and also expressed monocyte-defining markers including CD14 and CD163 (Figure 6E). When these DCs were cultured for 1 week with allogeneic naïve CD4 + T cells (> 99% purity; Supplementary Figure S2), > 30% of the responder T cells expressed FoxP3 (Figure 6F). Moreover, in accord with our data from healthy donors (Figure 5B), neutralization of TGFβ, IL-10 and CTLA-4 inhibited DC Reg formation ( Figure 6G).
Given that CD4 + T cells isolated from tumor tissue were enriched for Tregs, we hypothesized that the ratio of Tregs to Th in tumors would correlate with the capacity of the resultant DCs to induce FoxP3 + Tregs. To investigate this possibility, we quantified the percentage of Tregs in CD4 + T cells in tumor specimens and co-cultured total tumor CD4 + T cells with peripheral blood monocytes. Subsequently, we analyzed the capacity of the resultant DCs to elicit FoxP3 + Treg differentiation from naïve T cells stimulated in an MLR. The results indicate that the percentage of tumoral Tregs strongly correlated with the capacity of the ensuing DCs to induce FoxP3 expression, R 2 = 0.9509 ( Figure 6H).
Studies of the same tumor specimens revealed that the majority (∼80%) of tumor-associated DCs (TADCs), defined as CD45 + CD11c + HLA-DR + cells, expressed CD14, which indicates their monocytic origin and likely inclusion among the CD14 + cells associated with Tregs (Figures 6I,J). Staining for additional DC surface markers showed that they expressed CD40, CD80, CD86 and CD274 ( Figure 6K). When these CRC mo-DCs were incubated with naïve CFSE-labeled naïve allogeneic CD4 + T cells (> 99% purity; Supplementary Figure S2) in a mixed leukocyte reaction (MLR), a high proportion (> 20%) of the responder T cells expressed FoxP3 (Figure 6L). In summary, our findings suggest that not only are all of the cellular requirements for generating DC Reg present in these tumors, but also DC Reg similar to those induced from monocytes in vitro are present in situ in the same tumors.

DISCUSSION
Human monocytes can be induced to differentiate into tolerogenic DCs upon exposure to growth factors, cytokines, or pharmacological agents, in vitro (32). However, if and how human Tregs impact DC differentiation from monocytes has not been described. Here, we have demonstrated the ability of natural Tregs to promote the formation of DC Reg directly from monocytes, and thereby reinforce an  On day 3, 1 µg/ml LPS was added to selected wells, and 16h later FACS-purified DC subsets were further cultured with MACS sorted CFSE-labeled naïve allogeneic CD4 + T cells. After another 6 days, CD2 + CD4 + T cells were analyzed by flow cytometry. Mean ± SEM of 4 donors. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, ns, not significant. Data were analyzed with one-way ANOVA, followed by Dunnett's test for multiple comparisons. In the presence of anti-CD3 mAb, CD14 + monocytes were cultured with bulk CD45 + CD2 + CD4 + T cells sorted from CRC, in the presence of a mixture of anti-CTLA-4 (5 µg/ml), anti-IL-10 (2 µg/ml), anti-TGFβ (2 µg/ml), or isotype control antibody (5 µg/ml). At day 4, cells in culture were analyzed by flow cytometry for the indicated surface markers. (H) CD14 + monocytes were cultured with bulk CD45 + CD2 + CD4 + T cells sorted from CRC (red) or healthy donor peripheral blood (black). CD25 + CD127 low Treg percentages were analyzed before coculture. The graph shows the correlation between the percentage of Tregs in CD2 + CD4 + T cells prior to their initial culture with monocytes (X axis) and the percentage of CFSE − FoxP3 + T cells among the responder CD4 + T cells cultured with the DCs generated in the initial culture (Y axis). Linear regression was determined by prism. immunosuppressive environment through the induction of increased numbers of induced FoxP3 + Tregs. Our study reveals a novel mechanism whereby Tregs and Th collaborate in the induction of human regulatory DCs, thereby inducing immunosuppression and potentially contributing to "infectious tolerance" (33,34).
By studying the effects of Tregs and Th, separately and in combination, on monocytes, we were able to analyze the mechanism responsible for DC Reg formation. In agreement with our prior findings, cultures containing only activated Th and monocytes led to the formation of DC Th that secreted substantial amounts of IL-6 and TNFα and little or no IL-10 (8). Consistent with a previous report, monocytes cultured with Tregs alone differentiated into cells that are indistinguishable from macrophages (35). However, when equal numbers of Tregs and Th were added at the initiation of these cultures, the resultant morphology, surface phenotype and functionality of the monocyte-derived cells (DC Reg ) were distinct. Based on their large nuclei, dense cytoplasm, presence of surface dendrites and low expression of the M marker MerTK (36), these monocyte-derived cells are more similar to DCs than M . However, their surface phenotype and phagocytic activity indicate that they have features of both DC Th and M . Their secretion of IL-10 but not proinflammatory cytokines, combined with their ability to induce the formation of FoxP3 + Tregs from naïve T cells via TGFβ, suggests that their major function is to suppress the immune response.
Monocytes are highly mobile and plastic and, therefore, ideally equipped to respond to inflammatory or immunoregulatory signals. Our studies show that the signals present during monocyte differentiation determined not only whether DCs or M develop, but also the functions of the differentiated cells. DC Reg differentiated from monocytes were morphologically, functionally and transcriptomically distinct from other moDCs. DC Reg were also distinct from the previously described Tregtreated DC (Treg-DC) in their phagocytic capacity and cytokine secretion profile, although both can induce FoxP3 + Tregs (37). Although DC Reg expressed comparatively higher levels of TLR4 compared with DC Th and M Treg , their phenotype remained relatively stable upon LPS stimulation. Thus, DC Reg appear to be differentiated DCs that function mainly to induce Foxp3 + Tregs from CD4 + naïve T cells. These results are consistent with our previous finding that monocytes induced to differentiate into DC Th by different Th subsets develop into relatively stable DC populations that promote the polarization of the same Th subsets responsible for DC differentiation (8).
Tregs promoted the generation of DC Reg from monocytes, but they did so only in the presence of activated Th, and only if all 3 cell types were in direct contact. Interestingly, IL-10 but not TGFβ was required for the formation of functionally active DC Reg , while TGFβ but not IL-10 was required for FoxP3 + T cell induction by DC Reg . M derived from culturing monocytes with Tregs alone secreted large amounts of IL-10 (35). However, unlike DC Reg , they failed to induce FoxP3 + Tregs, indicating that the formation of DC Reg requires DC differentiation signals from Th, in addition to polarizing signals from Tregs.
The FoxP3 + Tregs induced by DC Reg are, by definition, inducible Tregs (iTregs) as opposed to thymus-derived natural Tregs. Whereas expression of FoxP3 is considered a definitive marker of murine Tregs, this is not the case in humans. Indeed, one study showed that activation of human T cells with anti-CD3 and anti-CD28 in the presence of TGFβ resulted in the generation of FoxP3 + T cells that were not suppressive and produced high levels of effector cytokines (38). Our findings differ from this study in that DC Reg induced the development of FoxP3 + T cells that are functionally suppressive. Interestingly, although both IL-10 and TGFβ contributed to the morphology and surface phenotype of DC Reg induced by Tregs, IL-10 but not TGFβ was required for the formation of functionally active DC Reg . On the other hand, TGFβ but not IL-10 was required for FoxP3 + T cell induction by DC Reg .
We also found DC Reg -like cells in CRC specimens, and they are likely present in a wide range of human tumors given the high frequency of Tregs, monocytes and DCs in many tumor types. Their expression of the monocyte/macrophage related markers CD64 and MerTK on CD11b + TADCs described by Broz et al. also suggests their mixed ontogeny (39). Indeed, a significant portion of the myeloid cells in these and other tumors are macrophages. Comprehensive gene expression profiling at single-cell resolution will provide a more accurate indication of the frequency and location of DC Reg in tumors relative to other myeloid cells. Nonetheless, in addition to DC Reg , the CRC specimens contained a high proportion of Tregs relative to Th, and when monocytes were cultured with total CD4 + T cells from these tumors, they differentiated into phenotypically and functionally similar DC Reg . As conditions favoring the development of DC Reg are present in many tumors, these cells are likely important contributors to tumor immune tolerance across a broad range of tumor types. Moreover, we speculate that regulatory DCs similar to those described here are likely present in tissues, in addition to tumors, that contain an abundance of myeloid cells along with Th and Tregs (29,40).

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online