The Notch ligand DLL4 defines a capability of human dendritic cells in regulating Th1 and Th17 differentiation

Abstract

Notch signaling regulates multiple helper CD4⁺ T cell programs. We have recently demonstrated that dendritic cells (DCs) expressing the Notch ligand DLL4 are critical for eliciting alloreactive T cell responses and induction of graft-versus-host disease (GVHD) in mice. However, the human counterpart of murine DLL4⁺ DCs has yet to be examined. We report here the identification of human DLL4⁺ DCs and their critical role in regulating T helper (Th)1 and Th17 differentiation. CD1c⁺ DCs and plasmacytoid DCs (pDCs) from the peripheral blood of healthy donors did not express DLL4. In contrast, patients undergoing allogeneic hematopoietic stem cell transplantation (HSCT) had a 16-fold more DLL4⁺CD1c⁺ DCs than healthy donors. Upon activation of toll-like receptor signaling, healthy donor-derived CD1c⁺ DCs dramatically upregulated DLL4, as did pDCs to a lesser extent. Activated DLL4⁺DCs were better able to promote Th1 and Th17 differentiation than unstimulated PB DCs. Blocking DLL4 using a neutralizing antibody decreased Notch signaling in T cells stimulated with DLL4⁺ DCs, and reduced the generation of Th1 and Th17 cells. Both NFκB and STAT3 were crucial for inducing DLL4 in human DCs. Interestingly, STAT3 directly activated DLL4 transcription and inhibiting STAT3 alone was sufficient to reduce DLL4 in activated PB DCs. Thus, DLL4 is a unique functional molecule of human circulating DCs critical for directing Th1 and Th17 differentiation. These findings identify a pathway for therapeutic intervention for inflammatory disorders in humans, such as GVHD after allogeneic HSCT, autoimmunity and tumor immunity.

INTRODUCTION

Dendritic cells (DCs) have the unique capacity to elicit primary T-cell immune responses.(1–3) DCs process and present antigen peptides, activate naïve T cells, and promote activated T cell expansion and survival through the expression of costimulatory molecules. DCs also produce effector-polarizing cytokines that are crucial in directing effective T cell differentiation.(4–8) However, emerging evidence indicates that DCs can drive effector differentiation independent of cytokines.(9, 10) Our studies and others suggest that Notch ligands expressed on the surface of DCs are important in promoting the generation of different lineages of effector T cells.(11–14) Notch ligands (Dll1, Dll3, Dll4, Jagged1 and Jagged2) interact with Notch receptors (Notch 1, 2, 3, and 4),(15–17) triggering the release of intracellular Notch and the subsequent transcription of Notch target genes(15–17). Using mouse models of graft-versus-disease (GVHD), a life-threatening immune complication of allogeneic hematopoietic stem cell transplantation (HSCT), we found that DC-derived Notch ligand DLL4 regulates effector differentiation of alloreactive CD4⁺ T cells.(14) DLL4-positive DCs (DLL4⁺ DCs) more strongly promote CD4⁺ T helper(Th)1 and Th17 cell differentiation than DLL4-negative (DLL4⁻) DCs.(14) Blocking DLL4 reduces production of IFN-γ and IL-17 in mice receiving allogeneic HSCT and inhibits the development of GVHD.(14) The human counterpart of mouse DLL4⁺ DCs has yet to be examined, and their effects on human Th1 and Th17 cell differentiation remain unknown.

Our understanding of human DCs is derived predominantly from studies of cells isolated from peripheral blood (18). Under steady state condition, human PB DCs are defined as cells that lack lineage (Lin) markers (i.e., CD3, CD15, CD19, CD14, CD20 and CD56) and constitutively express HLA-DR (referred as Lin⁻DR⁺ pan-DCs)(6). Human PB DCs are broadly categorized into two major subsets: conventional DCs (cDCs) and plasmacytoid DCs (pDCs). cDCs are characterized as Lin⁻HLA-DR⁺CD11c⁺ cells, whereas pDCs are Lin⁻HLA-DR⁺CD11c⁻CD123^highcells(6, 19). In blood, cDCs can be further classified into at least two subsets: CD1c⁺ DCs and CD141⁺ DCs(20). The former comprises the predominant cDC subset, whereas the latter is a relatively small population. At least three lines of evidence indicate that these three subsets of DCs may have different functions in mediating T cell immune responses. CD1c⁺ DCs express toll-like receptor (TLR)4 and TLR7, CD141⁺ DCs have high expression of TLR3, and pDCs express TLR7 and TLR9 and lack TLR4.(6, 7, 21–23) In addition, when activated, CD1c⁺ DCs produce high levels of IL-12, IL-6, IL-23 and IL-1β, whereas CD141⁺ DCs secrete IL-12 and IFN-β. In contrast, pDCs produce IFN-α and IFN-β.(23–29) Finally, while all three DC subsets can elicit primary T cell responses, CD141⁺ DCs can most efficiently cross-present antigens.(25, 30, 31) Thus, individual DC subsets may have differential impact on T cell immunity in response to inflammatory environmental cues. DCs are important for mediating T cell inflammatory disorders such as chronic infection and autoimmune diseases. (32, 33) One unique clinical instance is the induction of GVHD in individuals receiving allogeneic HSCT: DCs activate donor T cells, generate alloreactive T cells that produce high levels of effector cytokines (e.g., IFN-γ, IL-4, IL-17, and TNF-α) and cytolytic molecules (e.g., granzyme B, perforin and Fas ligand), and ultimately lead to target tissue damage. We asked whether PB CD1c⁺ DCs and pDCs produce high levels of DLL4, and whether DLL4 derived from human DCs also regulate Th1 and Th17 differentiation.

MATERIALS AND METHODS

Healthy donors and patients

PB from healthy donors of deidentified and patients undergoing allogeneic HSCT were collected in this study after obtaining informed consent. The characteristics of allogeneic HSCT recipients (21 cases) are shown in Table 1. PB was obtained early after HSCT (day 21–39). This study was approved by the Institutional Review Boards of Temple University and University of Pennsylvania.

Table 1.

Characteristics of patients undergoing allogeneic HSCT (n=21)

Patients
Recipient age in years, mean	55 (26–69)
Recipient gender - male, %	38
Diagnosis
Acute myeloid leukemia, n	10
Acute lymphoblastic leukemia, n	4
Chronic myelogenous leukemia, n	1
Chronic myelomonocytic leukemia, n	1
Non-Hodgkin lymphoma, n	1
Myelodysplastic syndrome, n	3
Myelofibrosis, n	1
Conditioning intensity
Reduced-intensity, n	16
Myeloablative, n	5
Graft source
Peripheral blood stem-cells, n	21
GVHD prophylaxis regimen
Tacrolimus+methotrexate, n	19
Tacrolimus+mycophenolatemofetil, n	2
Donor type
Related, n	13
Unrelated, n	8
HLA Matching
9/10, n	3
10/10, n	18
Acute GVHD, n (%)	12 (57%)
Relapse, n (%)	5 (24%)

Antibodies and flow cytometric analysis

DCs were identified by multi-color flow cytometric analysis using a BD LSRII flow cytometer. Dendritic cells were characterized as negative for lineage markers (CD3, CD14, CD15, CD16, CD19, CD20 and CD56) and positive for HLA-DR. Dead cells were excluded using Fixable Viability Dye from eBioscience (San Diego, CA). DLL4 expression was determined by surface staining and analyzed on each DC subset after exclusion of dead cells. All other antibodies were purchased from Biolegend (San Diego, CA) unless otherwise mentioned.

Generation of monocyte derived DCs

Monocytes were purified using CD14 microbeads from Miltenyi according to the manufacturer’s instruction. Monocyte derived dendritic cells (moDCs) were derived after culture of monocytes in RPMI medium containing 10% human serum AB, 20ng/ml recombinant human granulocyte macrophage colony-stimulating factor (rhGM-CSF, Shenandoah biotechnology) and 10ng/ml rhIL-4 (Shenandoah biotechnology) for 6 days followed by further maturation with 10ng/ml rhTNF-α (Shenandoah biotechnology) for 2 days. Cells were further stimulated with LPS (Sigma-Aldrich, MO) and R848 (Invivogen, CA) overnight to induce maturation.

Real-time RT- PCR

Total RNA was extracted using TRIzol (Invitrogen Life Technologies). cDNA was quantified through quantitative real-time polymerase chain reaction (PCR) using a SYBR Green PCR mix on a Master cyclerrealplex (Eppendorf). Thermocycler conditions included an initial holding at 95°C for 2 min; this was followed by a three-step PCR program, as follows: 95°C for 30s,55°C for 30s, and72°C for 30s for 40 cycles. Transcript abundance was calculated using the ΔΔCt or ΔCt method (normalization with 18S). The primer sequences are listed in Supplementary Table 1.

Mixed lymphocyte reaction (MLR)

Naïve CD4⁺ T cells were obtained from PBMCs through negative selection using human naïve CD4 T cell isolation kit II from MiltenyiBiotec (San Diego, CA). DCs from unrelated healthy donors were co-cultured with CFSE (Life technologies, CA)-labeled naïve CD4⁺ T cells in a 1:10 ratio in a round-bottom 96-well plate in RPMI medium containing 10% human serum albumin and 5ng/ml rhIL-2. Cells were collected on day 7 of co-culture and stained for surface CD4⁺ and intracellular cytokine production using an intracellular staining kit from BD (BD Cytofix/Cytoperm™ Kit, BD Biosciences, CA) and antibodies from Biolegend. In some experiments, neutralizing anti-DLL4 antibody (MHD4–46(34)) was added into the co-culture system at the beginning.

Western blot analysis

Western blot on cell lysates was performed as described.(35) Cell lysates were examined by routine western blotting. The blots were incubated with Abs against STAT3 (9139, Cell Signaling Technology), phosphorylated STAT3 (p-STAT3, 9138, Cell Signaling Technology), NFκB P65 (8242, Cell Signaling Technology), and phosphorylated P65 (p-P65, 3033, Cell Signaling Technology), or actin, and subsequently incubated with HRP-conjugated anti-rabbit or mouse IgG (Vector Laboratories) in TBS containing 3% BSA (bovine albumin) and 0.05% Tween 20. The final reaction was developed with a chemiluminescent system (36).

DC preparation

PB mononuclear cells (PBMCs) from healthy donors and HSCT patients were isolated using Ficoll-paque. Lin⁻DR⁺ pan-DCs were enriched from PBMCs of healthy donors using pan-DC enrichment kit purchased from MiltenyiBiotec (San Diego, CA) according to the manufacturer’s instruction. DCs were cultured in RPMI containing 10% human serum with or without TLR agonists.

Luciferase reporter assay

The human DLL4 promoter region ranging from −1.5 kb to +0.4 kb of the transcription start site (TSS) was cloned to pGL3 luciferase reporter vector to generate a DLL4-specific reporter (named pGL3-DLL4 reporter). STAT3-null PC3 cells were cotransfected with pGL3-DLL4 reporter plasmid and pcDNA3.1 plasmid encoding human STAT3 or empty pcDNA3.1 plasmid. Cells were harvested 72 h after transfection and analyzed with the Dual Luciferase system (Promega).

Statistical analysis

Comparisons between group means were performed using two-tailed Student t test. P<0.05 was considered statistically significant.

RESULTS

Human circulating DCs upregulate DLL4 under inflammatory conditions

To identify human DLL4⁺ DCs and characterize their biological properties, we obtained PB from healthy donors. Flow cytometric analysis revealed that Lin⁻DR⁺ pan-DCs contained three subsets: pDCs (CD11c⁻CD1c⁻CD123^hi), CD1c⁺ DCs (CD11c⁺CD1c⁺CD123^low), and CD11c⁺CD1c⁻CD123^low cells (Fig. 1A). We found that only a small fraction of CD1c⁺ DCs (2.2% ± 0.7%) and pDCs (0.8% ± 0.2%) from PB of healthy donors expressed low levels of DLL4 on their surface (Fig. 1B). CD11c⁺CD1c⁻CD123^low cells, which accounted for approximately 35% of Lin⁻DR⁺ pan-DCs, were also negative for DLL4 (Fig. 1B). Thus, under steady-state conditions, most PB DCs do not express surface DLL4.

Fig 1. Circulating blood DCs from HSCT patients upregulate DLL4.

PBMCs from healthy donors and HSCT patients were obtained and directly stained for DC surface markers. (A) Plots show the fraction of pan-DCs and DC subsets from healthy donors (n=14) and HSCT patients (n=21). Percentage of each DC subset among Lin⁻HLA-DR⁺ cell population from these two groups was calculated. (B) Histograms and graphs show the expression of DLL4 on the surface of each DC subset from healthy donors (n=14) and HSCT patients (n=21). (C) Graphs show the expression of DLL4 on the surface on each DC subset from HSCT patients at the time of GVHD onset vs. non-GVHD patients. **P<0.01; ***P<0.001.

To determine if under inflammatory conditions circulating DCs might upregulate DLL4, we obtained PB from patients undergoing allogeneic HSCT between 21 and 39 days after transplantation when these patients were fully engrafted. HSCT recipients had lower proportion of CD1c⁺ DCs and pDCs than healthy donors (Fig. 1A). This is consistent with previous observations of decreased circulating DCs in HSCT patients(37–39). Despite this, allogeneic HSCT recipients (n=21) had a 16-fold more DLL4⁺CD1c⁺ DCs (32.4% ± 6.0% vs. 2.2% ± 0.7%) as compared to healthy donors (n=14)(Fig. 1B). Both pDCs (3.2% ± 1.0%) and CD11c⁺CD1c⁻CD123^low DCs (2.1% ± 0.7%) in HSCT patients contained low frequency of cells expressing DLL4 (Fig. 1B). These results indicate that upregulation of DLL4 on the surface of CD1c⁺ DCs is uniquely associated with HSCT. Expression of DLL4 on the surface of these DCs did not distinguish GVHD patients from non-GVHD patients, as there was no significant difference in DLL4⁺ DC frequency between GVHD (12 cases) and non-GVHD (9 cases) patients (Fig. 1C).

Activation of TLR signaling induces DLL4 in human CD1c⁺DCs and pDCs

Previous studies have demonstrated that activation of TLR signaling is important for inducing Notch ligands in murine antigen-presenting cells (APCs).(12, 13, 40, 41) To determine the stimulus capable of inducing high levels of DLL4 in human DCs, we isolated PBMCs from healthy donors and cultured them with a variety of TLR agonists. Pam3 (TLR1/2 stimulus), Poly I:C (TLR3 stimulus), LPS(TLR4 stimulus) and R848 (TLR7/8 stimulus) induced high levels of DLL4 expression on the surface of 50% to 80% of CD1c⁺ DCs, whereas IFN-α (pro-inflammatory cytokine) and CD40L (signal from activated T cells) did not (Fig. 2A,B). CpG oligodeoxynucleotides (TLR9 agonists) did not increase DLL4 in CD1c⁺ DCs (Fig. 2A,B), consistent with the absence of TLR9.(42, 43) pDCs increased DLL4 expression when activated by R848 (16.0% ± 2.7%) and to a lesser extent by CpG oligodeoxynucleotides (8.6% ± 0.8%) (Fig. 2A,B). These results demonstrate that activation of TLR signaling induces high levels of DLL4 in CD1c⁺DCs and pDCs, with R848 being the most potent stimulus. R848 also induced DLL4 on the surface of approximately 30% of CD141⁺ DCs (data not shown).

Fig 2. Induction of DLL4 in human CD1c⁺DCs and pDCs.

(A) PBMCs from healthy donors were cultured in RPMI1640 containing 10% human serum and TLR ligand Pam3CSK4 (5.0µg/ml), Poly IC (20.0µg/ml), LPS (100.0ng/ml), R848 (100.0ng/ml), CpG ODN (1.0µM), IFN-α (100.0ng/ml) or CD40L (1.0µg/ml). PBS was added as control. After overnight stimulation, cells were collected and stained to measure live/dead cell dye, DC surface markers and DLL4. The expression of DLL4 was examined by gating live DC subsets. (B) Graphs show the percentage of DLL4-positive DCs (n=4). (C) PBMCs from HSCT patients (n=10) and healthy donors (n=10) were stimulated with R848. Twenty-four hours later, cells were collected to measure the expression of DLL4 on the surface of DC subsets. (D) CD1c⁺ DCs and pDCs were highly purified using FACS sorter to measure gene expression using real-time RT-PCR. Data shown are representative of three different donors. *P<0.05; **P<0.01; ***P<0.001. Error bars indicate mean ± SD.

We next focused on assessing biological properties of PB CD1c⁺ DCs and pDCs that were activated by R848. We observed no significant difference in fraction of DLL4-expressing CD1c⁺ DCs and pDCs between healthy donors and HSCT patients after R848 stimulation (Fig. 2C). This indicates that CD1c⁺ DCs and pDCs from healthy donors and HSCT patients have similar capacity to increase DLL4 upon activation of TLR7/8 signaling.

To further assess whether these DCs increased DLL4 at the transcriptional level, we isolated PB Lin⁻DR⁺ DCs from healthy donors and incubated them with R848 for 24 hours to induce DLL4. Highly pure populations of CD1c⁺ DCs and pDCs were then obtained through FACS cell sorting. The purity of sorted CD1c⁺ DCs and pDCs was consistently more than 98% (Fig. S1). Real-time RT-PCR validated that activated pDCs expressed significantly higher levels of IFNA and IFNB, their signature genes (7)(Fig. 2D). Notably, activated CD1c⁺ DCs expressed higher levels of DLL4 than pDCs (Fig. 2D), which was consistent with the observation that there was about 4-fold more in frequency of CD1c⁺ DCs capable of upregulating cell surface DLL4 compared to pDCs (Fig. 2B,C). Taken together, these results show that activation of TLR signaling induces high levels of DLL4 at transcriptional levels in both CD1c⁺ DCs and pDCs.

DLL4 is critical for human CD1c⁺DCs and pDCs to induce Th1 and Th17 cells

To assess whether DLL4 is critical for human DCs to induce Th1 and Th17 differentiation, we purified CD1c⁺DCs and pDCs and added them into MLR cultures containing CD4⁺ naïve T cells (T_N) from allogeneic healthy donors. CD4⁺ T_N were labeled with CFSE to monitor cell proliferation. As compared to unstimulated CD1c⁺DCs and pDCs, R848-activated CD1c⁺DCs and pDCs induced greater proliferation of allogeneic CD4⁺ T_N (Fig. S2A) and production of more IFN-γ- and IL-17-producing effector cells (Fig. S2B,C). These data suggest that activation of TLR7/8 signaling induced functional maturation of both CD1c⁺DCs and pDCs.

We next determined whether DLL4 was required for R848-activated CD1c⁺DCs and R848-activated pDCs to promote Th1 and Th17 differentiation. Blocking DLL4 using its neutralizing Ab did not affect activated T cell division (Fig. 3A), but caused slightly decreased expansion of allogeneic CD4⁺T cells (Fig. 3B). As compared to IgG control, addition of anti-DLL4 reduced production of Th1 cells in cultures stimulated by CD1c⁺DCs and pDCs by 2- and 3-fold less, respectively (Fig. 3C,D, n=8). Similarly, blocking Dll4 resulted in significant reduction of Th17 cells in cultures stimulated with CD1c⁺DCs and pDCs (2.4- and 1.7-fold, respectively, Fig. 3E,F). These data suggest that DLL4 derived from mature human DCs is critical for induction of both Th1 and Th17 cells in the setting of MLR. This is consistent with our previous findings in experimental mouse GVHD models.(14)However, CD1c⁺ DCs induced 3-fold more in frequency of Th17 cells than pDCs (Fig. 3E,F), in which the underlying mechanism has to be determined.

Fig 3. DLL4 is required for CD1c⁺DCs and pDCs to induce human Th1 and Th17CD4⁺T cells.

Alloreactive CD4⁺ T_N were labeled with CFSE and co-cultured with R848-activated CD1c⁺ DCs and pDCs at a ratio of 10:1, respectively. MLR cultures containing T cells but not DCs were served as controls. Neutralizing anti-DLL4 Ab (20 µg/ml) was added to the culture, with IgG Ab as control. Seven days after coculture, cells were collected to measure: (A) T cell division and (B) T cell expansion (recovery rate). Data shown are representative of three independent experiments. (C–F) Intracellular staining was performed to examine production of IFN-γ and IL-17 at the cellular level by gating on proliferating CD4⁺ T cells (CFSE^low cells). Dot plots show the fraction of cytokine-producing T cells (C, E), and graphs show the percentage of IFN-γ- and IL-17-producing T cells (D, F). *P<0.05; **P<0.01; ***P<0.001. Error bars indicate mean ± SD.

DLL4⁺ pDCs have greater ability than DLL4⁻pDCs to induce Th1 cells

We next examined whether DCs expressing high levels of DLL4 were better able than DLL4⁻DCs to promote Th1 and Th17 cell differentiation. When being activated with high concentration of R848, Lin⁻DR⁺ pan-DCs gave rise to approximately 90% of DLL4⁺CD1c⁺ DCs and 35% of DLL4⁺ pDCs (Fig. 4A). This allowed us to isolate pDCs into two subpopulations: DLL4⁺pDCs and DLL4⁻pDCs (Fig. 4B). When cocultured with allogeneic CD4⁺ T cells, we found that DLL4⁺ pDCs induced 2-fold more Th1 cells than DLL4⁻pDCs (Fig. 4C). However, both DLL4⁺ pDCs and DLL4⁻ pDCs induced similarly low levels of Th17 cells (Fig. S3).

Fig 4. DLL4⁺pDCs have greater capability than DLL4⁻pDCs to induce human Th1 cells.

(A) Lin⁻HLA-DR⁺ pan-DCs from healthy donors were stimulated with different doses of R848 overnight to induce DLL4 expression. Graph shows the percentage of DLL4⁺ CD1c⁺ DCs and pDCs in cultures (n=3). (B) Pan-DCs from healthy donors were stimulated with R848 (100.0 ng/ml) overnight. DLL4⁺pDCs and DLL4⁻pDCs were highly purified using FACS sorter. (C) CD4⁺ T_N were cocultured with allogeneic DLL4⁺pDCs and DLL4⁻pDCs. Seven days later, cells were collected to measure T cell division and IFN-γ production (n=3). (D) CD4⁺ T_N were cultured in the presence of anti-CD3Ab (0.5 µg/ml), with addition of autologous DLL4⁺pDCs and DLL4⁻pDCs, respectively. Neutralizing anti-DLL4 antibody was added, with IgG as control. Seven days later, cells were collected to measure T cell division and IFN-γ production. (E) mRNA was extracted from DLL4⁺ pDCs and DLL4⁻ pDCs of healthy donors (n=3) to measure gene expression using real-time RT-PCR. (F) Histograms show the expression cell surface markers. Results are representative of three independent experiments. *P<0.05; and **P<0.01. Error bars indicate mean ± SD.

Distinct DC subsets may have differential capabilities to activate TCR signaling,(5) thereby influencing their induction of effector T cells. To rule out this possibility, we added DLL4⁺ pDCs and DLL4⁻ pDCs to autologous CD4⁺ T cells cultured in the presence of anti-CD3Ab, with or without addition of anti-DLL4 Ab (Fig. 4D). This would allow us to assess the precise role of DLL4 in regulating Th1 cells triggered by identical strength of TCR signaling. We confirmed our preceding experiments (Fig. 3D) that DLL4⁺ pDCs induced significantly more Th1 cells than DLL4⁻ pDCs and this effect was dependent on DLL4 (Fig. 4D).

DLL4 identifies functionally distinct pDC subsets

These observations suggested that there existed two functionally distinct pDC subsets, distinguished by the surface expression of DLL4. Real-time RT-PCR analysis showed that both pDC subsets expressed similar levels of pDC signature genes IFNA and IFNB, and TCF4, a transcription factor important for pDC development(2). Both DC subsets expressed similar levels of TLR7 and TLR8 (Fig. 4E), excluding the possibility that the expression of TLR7/8 alone decreased the capacity of DLL4⁻ pDCs to express DLL4.Interestingly, as compared to DLL4⁻ pDCs, DLL4⁺ pDCs had higher levels of IL6, but were lower in IL1B expression (Fig. 4E), two inflammatory cytokines important for Th17 cell differentiation(28). In addition, DLL4⁺ pDCs expressed 2-fold higher IRF8, a transcription factor important for pDC development,(44) compared to DLL4⁻ pDCs (Fig. 4E). This difference seemed not to be the result of pDC activation, because both subsets showed similar levels of surface markers related to mature DCs (e.g., CD40, CD80, CD86 and CD83) (Fig. 4F). These data suggest that DLL4⁺ pDCs have differential biological properties and function from DLL4⁻ pDCs.

DLL4⁺ DCs activate Notch signaling in DC-stimulated CD4⁺ T cells

We examined whether Notch signaling mediated the Th1 and Th17-promoting effect of DLL4⁺ DCs. We found that CD4⁺ T_N stimulated by R848-activated CD1c⁺ DCs and pDCs upregulated Notch targets HES1 and DTX1 compared to T cells activated by unstimulated DCs (Fig. 5A,B). This effect of activated DCs was abrogated by addition of neutralizing anti-DLL4 Ab (Fig. 5A–B). Gamma secretase inhibitor (GSI) inhibited Notch signaling and impaired T cell proliferation and effector differentiation.(17, 45–47) Addition of GSI dramatically reduced the production of Th1 and Th17 cells in cultures stimulated with either activated CD1c⁺ DCs or pDCs (Fig. 5C). Thus, DLL4 induction of Th1 and Th17 involves the activation of Notch signaling in CD4⁺ T cells activated by allogeneic CD1c⁺ DCs or pDCs.

Fig 5. Evidence for DLL4 activation of Notch in activated human CD4⁺ T cells.

(A, B) mRNA from CD4⁺ T cells cocultured with non-stimulated DCs, R848-activated DCs, with or without anit-DLL4 Ab. Cells were collected to extract RNA and measure gene expression using real-time RT-PCR. (C) CD4⁺T_N were cocultured with R848-activated CD1c⁺ DCs or R848-activated pDCs in the presence of Notch inhibitors GSI (10.0 µM). PBS was added as control for GSI. Seven days later, cells were collected to measure their production of IFN-γ and IL-17. Results are representative of three independent experiments. *P<0.05. Error bars indicate mean ± SD.

NFκB only is not sufficient for DLL4 induction in human DCs

We explored the molecular mechanism that regulated DLL4 in human DCs. NFκB is a critical pathway downstream of TLR signaling(48). We found that the NFκB inhibitor PDTC completely blocked DLL4 induction in both DC subsets (Fig. 6A,B), suggesting the important role of NFκB in inducing DC expression of DLL4. MoDCs are believed to represent a subset of DCs of particular importance under inflammatory conditions.(6, 49) We found that stimulation of monocytes with R848 + LPS, which is known to activate NFκB signal in these cells,(50) activated NFκB as evidenced by increased expression of phosphorylated P65 (p-P65) (Fig. 6C), but induced low levels of DLL4 on the surface of monocytes (Fig. 6D). Real-time RT-PCR analysis further revealed that activated monocytes expressed 2 to 5-fold less DLL4 transcripts compared to pDCs and CD1c⁺ DCs (Fig. 6E). Furthermore, moDCs derived from cultures in GM-CSF and IL-4 had elevated p-P65 following stimulation by R848 + LPS (Fig. 6F) and upregulated the expression of costimulatory molecules (e.g., CD40, CD80, CD83, and CD86) (Fig. 6G), but were DLL4 negative (Fig. 6G). These data suggest that activation of NFκB is important but not sufficient for inducing DLL4 in human DCs.

Fig 6. NFκB is critical but not sufficient for inducing DLL4 in PB DCs.

(A, B) Pan-DCs from healthy donors (n=3) were pretreated with the NFκB inhibitor PDTC (1.0 µM) for 2 hours, followed by stimulation with R848 (100.0 ng/ml). PBS was added as control of PDTC. Twenty-four hours later, cells were collected to examine DLL4 expression. Histograms show the expression of DLL4 (A) and graphs show the percentage of DLL4⁺ DCs treated by doses of PDTC treatment (B). (C, D) Monocytes were magnetically purified from PBMCs of healthy donors (n=3) using CD14 microbeads, activated with or without LPS + R848 for 24 hours. Cells were harvested to assess NFκB using Western blot analysis (C), and expression of DLL4 using flow(D). (E) mRNA was extracted from CD1c⁺ DCs, pDCs and monocytes, which had been cultured with or without the presence of R848 (100 ng/ml, for CD1c⁺ DCs and pDCs) or LPS + R848 (both 100 ng/ml, for monocytes) for 24 hours, to measure DLL4 expression using real-time RT-PCR.(F, G) MoDCs were activated with or without LPS (100.0 ng/ml) + R848 (100.0 ng/ml) overnight. Cells were collected for Western blot analysis (F) and flow cytometric analysis (G). Results are representative of at least three independent experiments. *P<0.05; **P<0.01; ***P<0.001. Error bars indicate mean ± SD.

STAT3 is critical for inducing DLL4 in PB CD1c⁺ DCs and pDCs

STAT3, which is a transcription factor that regulates genes involved multiple cell processes,(51) is essential for fate decisions towards Flt3L-dependent DCs rather than monocytes and moDCs.(52) To determine the effect of STAT3 on DC expression of DLL4, we first examined the expression of STAT3 in DC subsets. We found that R848-activated CD1c⁺ DCs and pDCs expressed 8- to 10-fold more STAT3 transcript than both monocytes and moDCs that were activated by R848 plus LPS (Fig. 7A). Our preceding experiments showed that DLL4⁺ pDCs differed from DLL4⁻ pDCs (Fig. 4). As expected, DLL4⁺ pDCs expressed significantly higher level of STAT3 than DLL4⁻ pDCs (Fig. 7B).

Fig 7. STAT3 mediates DLL4 expression in PB CD1c⁺ DCs and pDCs.

(A) CD1c⁺ DCs, pDCs, monocytes and moDCs were activated with or without R848 (100 ng/ml, for CD1c⁺ DCs and pDCs) or LPS + R848 (both 100 ng/ml, for monocytes and moDCs) overnight to measure their expression of STAT3. Graphs show the level of STAT3 mRNA in these cells. Data are representative of 3 independent experiments. (B) Highly purified DLL4⁺ pDCs and DLL4⁻ pDCs were examined for their expression of STAT3 mRNA. Data are representative of 4 independent experiments. (C, D) Pan-DCs from healthy donors (n=3) were pretreated with different doses of STAT3 inhibitor S31–201 for 2 hours, followed by stimulation with R848 (100.0 ng/ml). Flow cytometric analysis was performed to measure the protein levels of p-STAT3 (C) and DLL4 (D) in DCs activated by R848 and monocytes activated by LPS plus R848 for 24 hours. Data are representative of 3 independent experiments. (E, F) Western blot analysis shows the expression of p-STAT3 and total STAT3 in pan-DCs treated with or without S31–201, and real-time RT-PCR showed the expression of genes in these DCs. One representative of 3 independent experiments. *P<0.05; **P<0.01; ***P<0.001. Error bars indicate mean ± SD.

To further evaluate the effect of STAT3 on DLL4 expression in DCs, we examined the amount of phosphorylated STAT3 (p-STAT3) in both CD1c⁺DCs and pDCs. Flow cytometric analysis showed that both activated CD1c⁺ DCs and pDCs possessed significantly higher levels of pSTAT3 than monocytes (Fig. 7C). Addition of the STAT3 inhibitor S31–201, which blocks STAT3 phosphorylation and dimerization(53), dramatically reduced DLL4 expression in both CD1c⁺DCs and pDCs (Fig. 7D). Using Lin⁻DR⁺ pan-DCs, we confirmed that S31–201 treatment decreased p-STAT3 in these cells (Fig. 7E). This was accompanied with a reduction of DLL4, IFNA and IFNB (Fig. 7F), suggesting that STAT3 inhibition may have a broad effect on DC function. Interestingly, S31–201 also significantly reduced expression of STAT3 transcripts and protein (Fig. 7E,F). This indicates that STAT3 may self-regulate, consistent with previous observations(54). Thus, in addition to NFκB, activation of STAT3 is critical for inducing DLL4 in PB DCs.

To define whether STAT3 can activate DLL4 transcription, we identified two STAT3 binding site located at the promoter region of DLL4 gene using the MotifMap (http://motifmap.ics.uci.edu/) (Fig. 8A). This regulatory region of the DLL4 promoter (ranging from −1.5kb to +400bp of the TSS) was cloned into the pGL3 luciferase reporter vector to generate pGL3-DLL4 reporter plasmid. We then co-transfected PC3 cells, a STAT3-null cell line(55), with pGL3-DLL4 reporter and pcDNA encoding STAT3 (Fig. 8B). The pcDNA plasmid lacking the STAT3 insert was used as control. Transfection of PC3 cells with STAT3-pCDNA resulted in overexpression of both STAT3 and to a less extent p-STAT3 (Fig. 8B), and induced DLL4 transcription (Fig. 8C). Furthermore, STAT3-overexpressing PC3 cells stimulated with R848 had enhanced DLL4 transcription compared to unstimulated PC3 cells (Fig. 8C). Interestingly, NFκB was constitutively activated in these PC3 cells and overexpression of STAT3 did not affect the expression of active NFκB (Fig. 8B), suggesting that overexpression of STAT3 does not activate NFκB in this context. Collectively, our results suggested that STAT3 is required for activating DLL4 transcription.

Fig 8. Overexpression of STAT3 activates DLL4 transcription.

(A) Schematic diagram of STAT3 binding site on −1,102 bp and −570 bp upstream of TSS in DLL4 gene promoter. Putative binding sequence was shown. (B, C) PC3 cells were transfected with pcDNA3.1 empty vector (control vector) or STAT3 encoding pcDNA3.1 vector (STAT3 vector), together with pGL3-DLL4 reporter plasmid and Renilla. Forty-eight hours after transfection, R848 (100.0 ng/ml) was added into the cell culture. PBS was added as control. Seventy-two hours after transfection, PC3 were collected to measure protein expression using Western blot (B) and luciferase activity (C). Results are representative of three independent experiments. **P<0.01. Error bars indicate mean ± SD.

DISCUSSION

Our previous studies have demonstrated the importance of DLL4 derived from DCs in regulating murine T cell immune response.(13, 14) Here we identified human DLL4⁺ DCs of mouse counterparts and the impact of DLL4 on DCs in human T cell responses. We identified markedly increased human DLL4⁺ DCs in patients undergoing allogeneic HSCT than in healthy donors. When activated by TLR ligands, both human CD1c⁺ DCs and pDCs from healthy donors upregulated DLL4, enabling alloreactive human CD4⁺ T cells to produce high levels of IFN-γ and IL-17. Blocking DLL4 abrogated this effect suggesting that DLL4 expression defines the activity of human inflammatory DCs in the regulation of T-cell immune responses.

Our findings may have significant implications in better understanding GVHD in patients undergoing allogeneic HSCT. For example, experimental studies of mouse GVHD models have demonstrated that blocking DLL4 protects mice from lethal GVHD while retaining GVL activity (9). In addition, blocking Notch signaling in donor T cells reduces GVHD while retaining GVL effects in leukemic mice undergoing allogeneic HSCT.(56) However, these studies did not examine if human DCs express DLL4 and if DLL4 derived from human DCs play a role in instructing Th1 and Th17 cell differentiation. Inhibiting Notch signaling also decreases effector differentiation of activated CD8⁺ T cells.(56) Data from the present study identify the critical role of DLL4 by human DCs in regulating human alloreactivity. Our findings further provide evidence that modulating DLL4 and Notch signaling might achieve a potential effect on reducing GVHD while preserving GVL in humans.

Data from experimental mouse studies indicate that Notch orchestrates multiple helper CD4⁺ T cell programs.(9, 10) DLL4 induces Th1 and Th17 cells through activating T cell Notch signaling.(9, 12, 13, 47, 56, 57) Notch signaling can directly activate transcription factors T-BET and RORγ, which is important for Th1 and Th17 differentiation respectively (40, 58). We found that donor CD4 T cells from Notch-deficient mice in HSCT recipient mice had decreased production of multiple cytokines including IFN-γ, IL-17 and IL-4.(56) In the present study, we identified the critical role of Notch signaling in regulating human effector CD4⁺ T cell differentiation. Inhibiting DLL4 decreased the expression of Notch target genes (e.g., DTX1 and HES1) in human CD4⁺ T cells stimulated by DLL4-expressing CD1c⁺ DCs and pDCs. Addition of GSI to the culture dramatically reduced production of IFN-γ and IL-17 by human CD4⁺ T cells activated by allogeneic PB CD1c⁺ DCs and pDCs. Thus, it is unlikely that reduced production of IFN-γ and IL-17 by CD4 T cells upon DLL4 blockade is a result of compensatory effects of other T cell subsets such as Th2 and regulatory T cells. Furthermore, identification of DLL4 in mediating human DC regulation of Th1 and Th17 cell differentiation are of great interest given that DC-based vaccine strategies can boost immune responses against tumor cells and infections, and may have significant implications in T cell-mediated autoimmune diseases.

These observations help explain previous observations that the reciprocal differentiation of Th1 and Th17 causes tissue-specific GVHD in mice of allogeneic HSCT.(59) In their experiments, donor CD4⁺ T cells lacking IFNG showed enhanced Th2 and Th17 cell differentiation and tissue damage in the lung and skin in mice, whereas donor T cells lacking IFNG and IL17 had augmented production of Th2 cells and lung injury.(59) Given the direct and multiple roles of Notch signaling in regulating alloreactive T cell responses, inhibiting individual lineage-polarizing cytokines may not affect Notch signaling regulation of Th cell differentiation during GVH reaction.

Given the important role of alloreactive Th1 and Th17 cells in mediating host tissue injury during GVHD, it will be interesting to understand the underlying mechanisms by which HSCT patients had significantly higher frequency of DLL4⁺CD1c⁺ DCs compared to healthy donors. Using animal models, we have previously demonstrated that preparative conditioning (i.e., irradiation) induced the production of DLL4⁺ DCs in mice undergoing HSCT.(14) DLL4⁺ DCs also occurred in mice without receiving allogeneic T cell-depleted HSCT.(14) It is likely that preparative conditioning, which causes increased production of damage-associated molecular pattern molecules and entry of LPS,(60) is sufficient to induce DLL4. While we have identified the human functional counterpart it appears that human PB DLL4⁺ DCs differ from murine DLL4⁺ DCs. We found that DLL4⁺ DCs from HSCT mice resembled pDCs but not cDCs, as evidenced by expression of PDCA-1 and B220.(14) In humans, while the majority of CD1c⁺ DCs (~90%) upregulated DLL4 upon R848 stimulation, only about one third of pDCs increased DLL4 after activation (Fig. 4A). Nevertheless, increased circulating DLL4⁺ DCs in patients undergoing allogeneic HSCT suggest they are important indicator of an inflammatory reaction in the host.

Our experiments do not show whether DLL4⁺ DCs are derived from host or donor in allogeneic HSCT recipients. Due to limitations of our blood samples and since circulating DCs are rare in the peripheral blood, we were unable to explore this further. However, previous studies have demonstrated that approximately 80% of circulating DCs were of donor origin 14 days after transplantation, and increased to 95% by 56 days after transplantation.(61) Therefore, it is likely that in our studies that utilized blood samples from patients between 21 days and 39 days after transplantation, DLL4⁺ DCs were primarily of donor origin. Studies in mice indicate that while host APCs are important for alloreactive T cell responses,(14, 62–68) donor APCs enhance GVH reactions and tissue injury.(65, 66, 69) Under inflammatory conditions, donor tissue resident APCs are uniquely able to promote alloreactive T cell responses.(65, 66, 69) These data indicate that regardless of origin, human DLL4⁺ DCs may contribute to GVH reactions in patients undergoing allogeneic HSCT.

It is interesting to note that while DLL4⁺CD1c⁺ DCs were critical for induction of Th1 and Th17 cell differentiation, their presence in the circulating PB could not distinguish GVHD patients from non-GVHD patients. This is in agreement with previous observations of our studies and others that DCs eliciting alloreactive T cell responses are short-lived in allogeneic HSCT.(14, 62, 63, 65, 66, 70, 71) After priming donor T cells, these inflammatory DCs rapidly diminished in mice, whereas activated donor T cells undergo a programmed proliferation and differentiation to produce sufficient amount of alloreactive T cells.(60, 62) However, our measurement of circulating blood DCs may not reflect tissue-resident DCs. pDCs are present in the skin and gut in acute GVHD, and correlate with the presence of Th17 cells.(31, 72) In experimental GVHD, we have observed DLL4⁺ DCs in the skin and gut early after transplantation.(14) Thus, future studies should investigate whether DLL4⁺ DCs present in the skin and intestine correlate with the development of severe GVHD in allogeneic HSCT patients.

We demonstrate here that the ability to upregulate and activate STAT3 in response to inflammatory stimuli is likely a key determinant for induction of DLL4 in human DCs. Many studies have demonstrated that STAT3 plays multiple roles in both innate and adaptive immunity(51, 73–75). For example, STAT3 is essential for production of indoleamine-2,3-dioxygenase (IDO) by murine antigen-presenting cells,(74) thereby repressing T cell response. Conversely, data from clinical studies indicate that about half of these patients treated with a STAT3 inhibitor experienced pathogen-mediated diarrhea(75), suggesting that STAT3 is required for protective mucosal immunity. In mice, loss of STAT3 results in deficiency of common DC precursors and their DC progenies, but has no effect on monocyte/macrophage differentiation of hematopoietic precursor cells(52). We found that activation of TLR7/8 signaling increased STAT3 and p-STAT3 in human PB CD1c⁺ DCs and pDCs. Inhibiting STAT3 led to dramatic decrease in expression of DLL4 transcripts and proteins in circulating DCs. Promoter reporter assay revealed that STAT3 activated DLL4 transcription. In contrast, monocytes or moDCs expressed low levels of STAT3 mRNA and active p-STAT3 protein in response to R848 and LPS. This may explain the inability of monocytes and moDCs to produce high level of DLL4. Thus, DC expression of DLL4 may have been developmentally programmed through a mechanism of permitting or repressing STAT3 transcription.

It is intriguing that inhibiting either NFκB or STAT3 caused a decrease in DLL4 expression in circulating CD1c⁺ DCs and pDCs, suggesting that an interaction of NFκB and STAT3 is required for these DCs to activate DLL4 transcription. It has been shown that the interaction between NFκB and STAT3 plays an important role in regulating inflammatory immune cells.(73)This may explain our observation that both NFκB and STAT3 are required for the induction of DLL4⁺ DCs. Future studies will investigate the molecular mechanisms that control the repressive and permissive conditions of STAT3 during DC development and maturation, and how NFκB and STAT3 cooperate to activate DLL4 transcription in human DCs.

In summary, we have identified a critical and previously uncharacterized role for DLL4 in mediating human DC-regulated Th1 and Th17 cell differentiation. Future studies will investigate the cellular and molecular mechanisms that regulate the development of DLL4-expressing CD1c⁺DCs and pDCs. Future studies will also focus on the role of DLL4⁺ DCs in the alloreactive responses that mediate host tissue injury and eliminate leukemic cells in allogeneic HSCT patients. Given the critical role of DLL4 as an immune modulator, our studies provide proof of concept for both DLL4⁺ DCs and DLL4 as promising therapeutic targets. DLL4 has further potential to be employed in cellular vaccines to augment anti-tumor and infection immunity. This study may have significant implications in better defining other inflammatory disorders, such as alloimmunity and autoimmunity.