STAT3- and STAT5-dependent pathways competitively regulate the pan-differentiation of CD34^pos cells into tumor-competent dendritic cells

Dendritic cells (DCs) are the most potent antigen-presenting cells in the body and are used in many tumor vaccine immunotherapy trials, rarely with therapeutic impacts.^1,2  DC preparations display a wide range of characteristics in vitro and in vivo, and it remains uncertain which individual properties may best promote successful immunotherapy.^3-6 

A variety of single agents, including CD40 ligand, Toll-like receptor (TLR) agonists, and calcium ionophore, can induce DC phenotypic maturation.^3,4,6-9  Such maturation includes pronounced expression of MHC and costimulatory molecules, CD40 and CCR7, and IL-8 secretion, but falls short of the DC's potential to achieve DC1 polarization, a highly effective state for promoting cell-mediated immunity.^4,6,7,9,10  DC1 polarization includes abundant production of IL12p70 heterodimer and IL-23, secretion of the chemokine MIP-1β, and preferential expression of Delta-4 Notch ligand.^4,6,7  Such DC1 products are highly associated with chemoattraction and activation of T1-type CD4⁺ and CD8⁺ T cells.^4,6,7  Furthermore, IL12p70 production is critical to sensitize high-avidity T cells that directly recognize and kill tumor targets.^4,6,7 

Although desirable for antitumor immunity, DC1 polarization is more easily signaled by infectious agents than by tumor exposure. Immature DCs use recognition of pathogen-associated molecular patterns to assess the likelihood of host infection and the appropriateness of DC1 polarization.^6,7  Individual TLRs signal DCs primarily through the MyD88 pathway (eg, TLR7-9) or TRIF pathway (eg, TLR3), with TLR4 evidencing pathway duplicity.⁷  Although activation of either pathway can induce elements of phenotypic DC maturation, dual pathway activation, or single pathway activation potentiated by exposure to IFN-γ or CD40 ligation, is required for robust DC1 polarization.^6,7 

Fresh mobilization of DC1 precursors has the theoretic potential to promote cross-presentation of tumor Ags within the tumor host. Recent studies confirm the capacity of stem cell–mobilizing treatments, notably granulocyte-macrophage colony-stimulating factor (GMCSF), granulocyte colony-stimulating factor, fms-like tyrosine kinase receptor-3 (Flt3 ligand, Flt3L), or combined Flt3L plus GMCSF to mobilize DC precursors,^5,11-13  but the potential of such precursors to achieve DC1 polarization is presently unclear. Flt3L plus GMCSF mobilization was recently reported to induce abundant infiltration of DCs into mouse tumors, but such DCs also activated regulatory T cells and promoted tumor tolerance.⁵  Immunosuppressive factors including IL-10, transforming growth factor beta (TGF-β), vascular endothelial growth factor (VEGF) and prostaglandin E₂ (PGE₂) are often produced by tumors, and may impede mobilized DC precursors from attaining optimal maturation and DC1 polarization.^14-17  However, mobilizing treatments themselves can also influence the later differentiation responses of stem cells.^5,11-13  We therefore postulated that particular stem cell proliferative treatments might provide conditioning signals that licensed rather than limited responsiveness to DC1 polarization stimuli.

Stem cell proliferation is inducible via multiple signaling pathways, including receptor-linked tyrosine kinases (Flt3L and stem cell factor [SCF]),^18-20  gp130 (eg, IL-6),^18,19  and the hematopoietin receptor superfamily (eg, GMCSF),²⁰  During the systematic testing of such signaling agents we identified that combined exposure to Flt3L plus IL-6 (Flt3L + IL-6) not only synergized for stem cell proliferation, but also licensed CD34^pos progenitor cells to forego multilineage differentiation in favor of STAT3-dependent, dedicated DC differentiation. Such STAT3-dependent DC differentiation proved highly conducive to DC1-type functional competence during later interactions with tumor. In contrast, exposure of CD34^pos cells to GMCSF dominantly promoted an alternative, STAT5-dependent pathway of DC differentiation that was less conducive to tumor interactions.

All experiments were performed under institutionally approved animal research committee protocols adhering to USDA guidelines. Female C57BL/6N, C3H/HeJ, and BALB/c mice were purchased from Biologic Testing Branch, National Cancer Institute (NCI, Frederick, MD). They were maintained pathogen-free under UDSA guidelines and studied at 8 to 12 weeks or as indicated. MCA-203, -105, and -205 fibrosarcomas and B16 melanoma, syngeneic to C57BL/6N mice; CT26 colonic adenocarcinoma, syngeneic to BALB/c mice; and the 888 mel human melanoma line were maintained as described previously.^21,22 

After extensive C57BL/6N backcross, STAT5a/b knockout (KO) mice²³  with hypomorphic STAT5 expression (N-terminal truncation²⁴ ) were obtained by breeding heterozygotes to yield viable STAT5ab^−/− pups. Mice were genotyped by polymerase chain reaction (PCR) of tail clip genomic DNA using primer pairs specific for STAT5a and/or STAT5b genes as previously described.²⁵ 

Following exhaustive C57BL/6N backcross, mice homozygous for a STAT3 allele with loxP sequences flanking critical exons 18 to 20 (F allele) were bred with Tie2-Cre mice (C allele) to generate mice with bone marrow (BM) conditionally knocked out for STAT3 (STAT3-CFF).²⁶  Tail clip PCR genotyping distinguished a 490-bp product of wild-type STAT3 from the 520-bp STAT-F product, as well as the presence or absence of a Tie2-Cre 300-bp product.

rhFlt3L (gift of Amgen, Thousand Oaks, CA); rhIL-2 (gift of Chiron, Emeryville, CA); rmGMCSF (gift of Immunex, Seattle, WA); rmCSF, rmIL-6, rmIL-10, rmVEGF, rmIL-3, rm thrombopoietin, rmIL-4, rmIFN-γ, rrhIL-7, and rhIL-15 (Peprotech, Rocky Hill, NJ); rhTGFβ1 (R&D, Minneapolis, MN); PGE₂, LPS (Escherichia coli 026:B6), poly I:C, and prostaglandin E₂ (Sigma-Aldrich, St Louis, MO); CpG (ODN 1826); and imiquimod (Invivogen, San Diego, CA) were used. Culture medium (CM) consisted of RPMI-1640 with 10% heat-deactivated FCS and conventional additives.²¹ 

Mouse BM suspensions were prepared from femurs and tibias.²⁷  Following red blood cell (RBC) lysis by ammonium chloride, 12 to 15 million BM cells (CD34^pos cell frequency 8%-12%) were cultured at 0.5 million/mL in 75-cm² flasks (Corning, Corning, NY) at 10% CO₂ in CM with the factors to be tested: hFlt3L 25 ng/mL (similar results 25-300 ng/mL); mIL-6 25 ng/mL (similar results 25-100 ng/mL; 100 ng/mL rhIL-6 also effective); mSCF 25 ng/mL (similar results 25-100 ng/mL; rSCF also effective); rmGMCSF 10 ng/mL (similar results 10-25 ng/mL); and rmIL-4 10 ng/mL (similar results 10-25 ng/mL). Recombinant thrombopoietin, IL-3, and fibronectin were also assessed, with negligible superimposed proliferative impacts (not shown). BM cells were normally cultured for 6 to 7 days, harvested, and washed twice in PBS before step 2 of culture.

Step 2 was initiated (0 hours) in CM, rmGMCSF, and optionally rmIL-4 in 24-well cluster plates, 4 million BM cells/well. DC1 polarization stimuli such as CpG (ODN1826, 5 μM) and lipopolysaccharide (LPS, 50 ng/mL) were added between 18 to 24 hours, and cells harvested at 40 to 44 hours for analyses. When included, rmIL-10, rmVEGF, rhTGFβ1, or PGE₂ was added between 0 to 24 hours of step 2 of culture. Alternatively, particulate tumor cells, either viable unirradiated, viable irradiated (10 000 cGy), or killed freeze-thawed lysate,²⁸  were added. Viable tumor cells were sometimes labeled with CFSE (Molecular Probes, Eugene, OR) to allow exclusion during fluorescence-activated cell sorting (FACS) analyses.²¹ 

Cells were cultured in anti-CD32 mAb plus normal mouse IgG to block FcR, then stained with fluorescently conjugated specific or isotype controls mAbs (BD-PharMingen, Mountain View, CA).²⁹  When additionally assessing IL-12 production at the cellular level, the last 16 hours of step 2 of culture were performed in monensin (Golgistop; BD-PharMingen). Then, after FcR block and surface molecule staining, cells were treated with CytoPerm/CytoFix (BD-PharMingen), then stained with PE anti–mouse IL12p40 (BD-PharMingen) or PE isotype control.

Intracellular staining for TLR was performed directly on fixed, permeabilized cells with conjugated mAb (TLR3, TLR4, TLR8, TLR9), or indirectly with unconjugated mAb (TLR7), and appropriate controls from Imgenix (San Diego, CA). Staining for intracellular IRF4 and IRF8 was performed with reagents from Santa Cruz Biotechnology (Santa Cruz, CA) and mouse adsorbed F(ab)′₂ fragments of donkey anti–goat Ab (Research Diagnostics, Concord, MA), following the method of Tamura et al.³⁰ 

To isolate CD34^pos and CD34^neg cells from freshly harvested BM, the latter were stained with FITC rat anti–mouse CD34 (dialyzed to remove sodium azide), then sorted on a FACSAria (BD Biosciences, San Jose, CA), yielding more than 96% pure CD34^pos and CD34^neg subpopulations.

In addition to PCR STAT determinations, pSTAT proteins were analyzed on the cellular level by FACS.³¹  Nuclei of cultured BM cells were permeabilized by sequential exposure to formalin (CytoFix) and 90% methanol. Cells were then stained with PE- or FITC-conjugated anti-pSTAT3 (pY705), anti-pSTAT5 (pY694), or isotype controls (BD Biosciences).

Supernatant contents of mIL12p70 heterodimer were quantified with BD-PharMingen reagents. mIFN-α and mIFN-β were quantified with PBL Biomedical Laboratories kits (New Brunswick, NJ).

Prior to T-cell harvest, DCs were prepared under various step 1 conditions in CM; step 2 was performed in 1% non–heat-deactivated mouse serum (MS) instead of FCS, during which BM cultures were exposed to viable irradiated tumor cells, then CpG plus LPS. On the day of DC harvest, T cells were freshly harvested from tumor-draining lymph nodes (TDLNs) of 12-day tumor-bearing mice. The L-selectin^low (tolerized pre-effector) fraction of T cells was isolated as previously described,^29,32  and cultured in CM-MS with immobilized anti-CD3^29,32  or variously conditioned DCs. Beginning on day 2 of T-cell culture, some groups also received rhIL-2 (24 IU/mL), or rhIL-2, rhIL-7 (50 ng/mL), and rhIL-15 (5 ng/mL). T cells were harvested for assays and adoptive therapy after 12 days of culture.

Cultured T cells were replated in fresh CM-MS at 2 million T cells/well in 24-well plates. Whole-cell irradiated tumor digests (5000 cGy) were added at 2 million cells/well as stimulators. Coculture proceeded for 18 hours, with monensin added the final 13 hours. At harvest, individual treatment groups were FcR blocked, then stained with FITC anti-CD4 and cychrome anti-CD8. Following fixation/permeabilization, cells were additionally stained with PE antimouse IFN-γ or isotype controls, then analyzed.

Viable tumor cells (1.5 million) were injected into healthy syngeneic mice to establish intradermal tumors.²⁹  Five or 10 days later, mice received conventional nonmyeloablative whole-body irradiation (WBI, 500 cGy),²⁹  followed by culture-activated T cells intravenously. Perpendicular bidimensional tumor measurements were performed twice weekly. Mice were killed when bidimensional product exceeded 225 mm².

Cultured BM cells were labeled with CFSE²¹  and injected intravenously into syngeneic mice bearing 10-day subcutaneous MCA-203 or MCA-105 tumors. Forty-eight hours later, mice were killed and tumors harvested and enzymatically digested to produce whole-cell digests, with spleen cell suspensions prepared in parallel.²¹  Preparations from individual mice were analyzed by FACS for CFSE^pos cell frequencies. Groups were then costained with PE-conjugated mAb against DC-associated surface determinants, and FACS gated to analyze the CFSE^pos subpopulation.

Survival among treatment groups was compared by Fisher exact test. Individual mice were scored for final treatment outcome (lethal tumor vs cure) and treatment groups compared. A 2-tailed P value less than .05 was deemed significant. Proliferative synergy was assessed by Wilcoxon signed rank test for paired data (proliferation following exposure to combined factors vs summed synchronous proliferations of the individual factors). Trafficking accumulation of CFSE-labeled cells and FACS pSTAT quantitations for cultured BM were assessed by Wilcoxon signed rank test for paired data. In all cases, a 2-tailed P value less than .05 was deemed significant.

We cultured fresh mouse BM in 2 steps, a 6- to 7-day proliferative culture (step 1) and a 48- to 72-hour postproliferative culture (step 2), followed in some experiments by T-cell coculture (Figure 1A).

Consistent with previous reports,^18-20  3 treatment pairings (Flt3L + IL-6, SCF + IL-6, or Flt3L + GMCSF) produced synergistic proliferation during step 1 of culture (Figure 1B). Such expansions represented a selective 35- to 80-fold numeric expansion of the CD34^pos cell subpopulation, with rapid dropout of initially CD34^neg cells (Figure 1C). CD34^pos cells continued brisk expansion for at least 1 additional week in culture if replenished with the same treatment pairings (not shown).

Even though Flt3L is myeloproliferative when administered to animals,^5,33  step 1 of culture with single-agent Flt3L produced poor yields, even when dosing extended up to 300 ng/mL (Figure 1B and not shown). This validated previous reports that single-agent Flt3L is poorly proliferogenic ex vivo unless BM cultures are seeded at sufficiently high density to confer natural IL-6 supplementation.³⁴ 

Prior to initial culture, freshly harvested CD34^pos BM cells rarely displayed DC or other lineage markers (not shown). By the end of step 1, however, cultures treated with either single-agent Flt3L or single-agent GMCSF displayed frequent differentiation into immature DCs, based on their dual positivity for CD11c and MHC class II and low expression of CD40 and B7.2 (Figure 1D; also see Figure S1A for full tested panel, available on the Blood website; see the Supplemental Materials link at the top of the online article). Conventional DCs (positive for CD11b^pos but negative for B220^neg) predominated in both instances (Figure S1B). Uniformly high MHC class II expression was a hallmark of Flt3L-induced DC differentiation, whereas GMCSF-induced DCs were heterogeneous and predominantly low in regard to MHC class II expression (Figure 1D).

Since Flt3L by itself was poorly proliferogenic (Figure 1B), we examined the superimposed impacts of IL-6 or GMCSF, since either agent produced proliferative synergy in conjunction with Flt3L (Figure 1B). BM proliferatively conditioned with Flt3L + IL-6 failed to acquire either CD11c or MHC class II expression, due to a pronounced antidifferentiative effect attributable to IL-6 (Figure 1D). In contrast, BM conditioned with either Flt3L + GMCSF or Flt3L + IL-6 + GMCSF developed heterogeneous differentiation that closely resembled treatment with GMCSF alone (Figure 1D). Therefore, even though IL-6 could exert a pronounced antidifferentiative effort upon Flt3L cultures, GMCSF could dominantly antagonize the conditioning impacts of both Flt3L and IL-6 during step 1 of culture.

Following step 1 conditioning, step 2 cultures were exposed to TLR agonists to examine their real-time potentials for DC1 polarization (Figure 1E).^6,7  Since both TLR9 and TLR4 were invariably expressed at the end of step 1 mouse BM cultures (not shown), we standardly used CpG (ODN 1826) and lipopolysaccharide (LPS) during step 2 of culture to elicit coordinate activation of MyD88 and TRIF pathways.^6,7 

Step 1 conditioning with single-agent GMCSF rendered a large proportion of cells hyporesponsive or unresponsive to subsequent TLR agonists, as evidenced by limited or absent up-regulation of MHC class II, CD40, and B7.2 expression (Figure 1E; Figure S2A). Gr-1 coexpression was a common feature of poorly responsive subpopulation(s) (Figure S2B).^35,36 

In contrast, step 1 conditioning with single-agent Flt3L licensed consistently high TLR responsiveness, manifested by nearly global phenotypic DC maturation and only scant Gr-1^pos elements during subsequent step 2 of culture (Figure 1E; Figure S2B). However, Flt3L's global licensing impacts were antagonized if GMCSF was also included during step 1 of culture (Figure 1E; Figure S2A). Such inhibition was not observed if initial exposure to GMCSF was deferred until step 2 of culture (not shown).

In contrast to GMCSF, IL-6 not only produced proliferative synergy in conjunction with Flt3L (Figure 1B,C), but also promoted Flt3L's global licensing impacts. Although an undifferentiated state persisted during step 1 conditioning in Flt3L + IL-6 (Figure 1D), nearly the entire expanded CD34^pos progenitor cell pool responded to subsequent TLR stimulation with DC differentiation, robust phenotypic maturation, and nearly uniform IL-12 production (Figures 1E; S2A). IFN-β was coelicited rather than IFN-α by all tested TLR agonist combinations, indicating that conventional DC differentiation dominated during step 2 of culture (Figure S2C).⁷  However, as for Flt3L, Flt3L + IL-6's global licensing impacts were abrogated if GMCSF was also included during step 1 of culture (Figure 1E; Figure S2A).

The functional consequences of each of these conditioning regimens proved age independent, with indistinguishable outcomes observed for BM obtained from mice aged 4 to 80 weeks (not shown). Similar responses were observed in all tested mouse strains.

Following lineage commitment, DCs typically remain immature unless they are exposed to exogenous signals such as CD40 ligation, calcium ionophore, or TLR agonists.^3,4,6-9  It was observed, however, that BM cells conditioned in single-agent Flt3L subsequently displayed spontaneous DC maturation, even when GMCSF was the only exogenous supplement provided during step 2 of culture (Figure 2B; Figure S3A). Similar spontaneous DC maturation was also observed following step 1 proliferative conditioning in Flt3L + IL-6, even though the inclusion of IL-6 had delayed the onset of DC differentiation until the onset of step 2 of culture (Figure 1D vs Figure 2B; Figure S3A). Spontaneous DC maturation following either Flt3L or Flt3L + IL-6 conditioning was even more vigorous when IL-4 was also provided during step 2 of culture (Figure S3B). Marked up-regulation of the endocytic C-type lectin receptor DEC-205 was a conspicuous component of such spontaneous maturation (Figure S3C).

For all conditioning treatments other than Flt3L or Flt3L + IL-6, spontaneous step 2 maturation was either highly attenuated or not observed. Importantly, Flt3L + IL-6 was the only factor pairing that resulted both in proliferative synergy during step 1 of culture and in spontaneous DC maturation during step 2 of culture (Figure 2B; Figure S3). Even when other conditioning treatments evoked limited elements of DC maturation during step 1 of culture (Figure 1D; Figure S1A), such elements spontaneously reverted during step 2 of culture unless exogenous maturational stimuli such as TLR agonists were also provided (Figures 1E vs 2B; full panels in Figures S2A vs S3). Flt3L or Flt3L + IL-6's capacity to license spontaneous DC maturation was abrogated, however, if GMCSF was also included during step 1 conditioning (Figure 2B; Figure S3).

We characterized expression elements of the conventional DCs that dominated step 2 of culture following Flt3L + IL-6 conditioning. Expression of TLR3, TLR4, TLR7, TLR8, and TLR9 remained uniformly detectable at all stages of culture, functionally confirmed by these cells' broad responsiveness to respective TLR agonists (Figures S2C and S4A). Interferon-regulatory factors IRF4 and IRF8 were dually expressed both by DC precursors and polarized DC1 (Figure S4B), demonstrating a largely homogeneous sequence of differentiation, maturation, and polarization following Flt3L + IL-6 conditioning.

We investigated whether Flt3L + IL-6 proliferative conditioning licensed uniform responsiveness to DC1 polarization stimuli even in the presence of putative immunosuppressive factors. We added factors at doses that equaled and exceeded those reported to inhibit the maturation of other DC preparations.^14-17  IL-10 and VEGF exposure had negligible impacts upon DC1 polarization, whereas TGFβ1 exposure paradoxically enhanced both phenotypic maturation and IL-12 secretion (Figure 2C,D). Only early PGE₂ exposure detectably inhibited TLR agonist-induced phenotypic DC maturation and IL-12 production (Figure 2C,D). Nonetheless, a large subpopulation of Flt3L + IL-6–conditioned BM cells resisted PGE₂ inhibition even at the beginning of step 2 of culture (Figure 2C, note the bimodal B7.2 expression), and such resistance became increasingly prevalent within hours of step 2 of culture (Figure 2D; Figure S5).

We tested the impact of exposing DCs to voluminous tumor burdens at 16 to 24 hours of step 2 of culture. To each well containing 4 million preconditioned BM cells, we added either 4 million freeze-thawed (killed) tumor cells; 3 million irradiated (10 000 cGy), trypan-excluding apoptotic tumor bodies; or 2 million unirradiated, actively proliferating tumor cells.

After step 1 of Flt3L + IL-6 conditioning, contact with any of these tumor materials accelerated DC phenotypic maturation, mimicking the stimulatory impact of exposure to a single TLR agonist (Figure 3A).^6,7  Combining such tumor exposure with IFN-γ treatment induced IL12p70 production (Figure 3A), mimicking the impact of combined exposure to IFN-γ plus a single TLR agonist.^4,7  When other step 1 conditioning conditions were compared, activating effects of tumor were either highly attenuated or completely absent (Figure 3A).

The capacity of Flt3L + IL-6–conditioned DCs to be activated by tumor contact proved strain independent, occurring even for BALB/c BM-derived DCs despite this strain's Th2-biasing tendency³⁷  (not shown). All tested tumor lines stimulated DC maturation after Flt3L + IL-6 conditioning, including MCA-205 and MCA-203 sarcomas and B16 melanoma derived from C57BL/6N mice, CT26 colon adenocarcinoma derived from BALB/c mice, and 888mel from a melanoma patient, and was not attributable to endotoxin content (not shown). Both freshly harvested whole-cell tumor digests and established tumor lines proved stimulatory, indicating that host stromal cells were unessential, and DC maturation was stimulated whether tumor was syngeneic, allogeneic, or xenogeneic (not shown). Although fully killed lysate was effective, viable tumor proved more effective. Sequestration of tumor from Flt3L + IL-6–conditioned DC precursors by Transwell membranes (Corning) abrogated the activating effects of tumor, underscoring a requirement for direct contact; in contrast, phagocytosis of latex beads did not accelerate maturation (not shown).

We examined the capacities of previously conditioned DCs to reverse tolerance in T cells harvested from mice bearing advanced tumors. L-selectin^low T cells from tumor-draining lymph nodes (TDLNs) are naturally sensitized to the relevant tumor but are also tolerized, consequent to the progressive upstream tumor burden.²⁹  In vitro exposure to anti-CD3, followed by IL-2 stimulation, can reverse tolerance and numerically expand antitumor effector T cells.²⁹  However, such polyclonal stimulation causes CD8⁺ T cells to overgrow CD4⁺ T cells, and lacks the element of antigen presentation needed to selectively promote the tumor-specific T-cell subset.^29,32 

Conditioned DC precursors were transferred to step 2 of culture, exposed to viable irradiated tumor, optionally further activated with CpG plus LPS, then cocultured with tolerized TDLN T cells from mice bearing the relevant tumor. Flt3L + IL-6–conditioned DCs efficiently reversed tolerance and stimulated robust T-cell proliferation, even when exogenous cytokines such as IL-2 were not added to coculture (Figures 3B and 4A). CpG plus LPS treatment enhanced but was unessential for such efficacy (not shown). In contrast, DCs prepared after other conditioning treatments typically proved lethal to T cells, and toxicity could not be prevented by exogenous IL-2, IL-15, and/or IL-7 (Figures 3B and 4A). Cocultures driven by Flt3L + IL-6–conditioned DCs displayed superior outgrowth of both CD4⁺ and CD8⁺ tumor-specific T cells compared with anti-CD3 treatment (Figure 4Bi,ii), and were highly potent when provided as adoptive therapy against early or advanced established tumors (Figure 4C; Figure S6).

Following various step 1 culture treatments, BM cells were CFSE labeled and administered to 10-day tumor-bearing mice. Cells injected immediately after step 1 conditioning with Flt3L + GMCSF or Flt3L + GMCSF + IL-6 displayed negligible trafficking into either tumor or spleen. In contrast, Flt3L plus IL-6–conditioned BM cells infiltrated both established tumors and spleen, achieving essentially uniform DC differentiation at either location (Figure 5A,B). Moreover, accelerated DC maturation was observed following entry into tumor compared with spleen (Figure 5B), consistent with observed stimulatory impacts of tumor contact in vitro during step 2 of culture (Figure 3A).

Knockout studies have demonstrated that Flt3 ligation transitions CD34^pos common precursors into committed DC precursors via a STAT3-dependent process, whereas GMCSF promotes STAT3-independent myeloid differentiation.³⁸  IL-6 signaling is known to induce gp130-mediated STAT3 activation in CD34^pos cells,³⁹  whereas GMCSF induces more complex STAT modulations, including STAT5 activation, at several stages of myeloid differentiation.^38,40,41  Because step 1 GMCSF consistently abrogated the unique conditioning impacts of Flt3L or Flt3L + IL-6, we examined whether STAT modulations played a pivotal role in GMCSF's apparently dominant regulation.

Step 1 IL-6 in the absence of GMCSF produced sustained activation of STAT3 but not STAT5, as evidenced by intranuclear staining for pSTAT3 (pY705) and pSTAT5 (pY694) (Figure 6A end step 1; Figure S7). In contrast, inclusion of GMCSF as a component of any step 1 regimen depressed STAT3 activation and produced biphasic up-regulation of pSTAT5 (Figure 6A end step 1; Figure S7). These distinctive pSTAT expression patterns modulated again during step 2 of culture, however, when cultures previously conditioned in Flt3L + IL-6 now displayed the greatest capacity for STAT5 activation (Figure 6A end step 2).

We next examined how STAT knockout BM preparations responded to Flt3L + IL-6 conditioning. Consistent with STAT3's putative obligate role in Flt3L-induced DC differentiation,³⁸  we observed that STAT3^KO BM could not survive Flt3L + IL-6 in vitro conditioning (not shown). In contrast, STAT5^KO BM responded to Flt3L + IL-6 conditioning with intact robust proliferation (not shown) and nearly uniform DC differentiation during subsequent step 2 of culture (Figure 6B row 2). Nonetheless, compared with wild-type littermates, Flt3L + IL-6–conditioned STAT5^KO DCs displayed submaximal DC maturation and IL12p70 production (Figure 6B row 2 vs row 1 and Figure 6D left). Therefore, the proliferative and differentiative impacts of Flt3L + IL-6 conditioning were absolutely STAT3-dependent, whereas subsequent phenotypic maturation and DC1 polarization were at least partially STAT5-dependent.

We also examined how STAT^KO BM preparations responded to GMCSF-containing step 1 regimens (GMCSF alone, Flt3L + GMCSF, or Flt3L + GMCSF + IL-6). All of these GMCSF-based conditioning regimens were strikingly ineffective for generating DCs from STAT5^KO BM, instead yielding predominantly Gr-1^pos, MHC class II^neg cells that displayed morphologic features of mature neutrophils (eg, Figure 6B row 4, and data not shown). In contrast, STAT3^KO BM displayed normal proliferative kinetics (not shown) and typical heterogeneous differentiation, including DC differentiation, in response to all GMCSF-based regimens (Figure 6C row 4). Following GMCSF-based conditioning, however, STAT3^KO DCs displayed an abnormally heightened maturational and IL12p70 response to TLR stimulation (Figure 6C row 4 vs row 3 and Figure 6D right). Therefore, all tested GMCSF-containing regimens required intact STAT5 for DC differentiation, and later phenotypic maturation and DC1 polarization displayed the capacity for negative regulation through STAT3.

Certain timely stimuli can condition DCs to adhere to an extended period of programming. For example, we recently reported that exposure of monocyte-derived human DCs to IFNγ + LPS not only stimulated an initial burst of IL12p70 secretion, but also licensed a second burst of IL12p70 secretion even days later if CD40 ligation was experienced.⁶ 

We report here that durable DC programming can also be secured at remarkably early stages of hematopoiesis, even prior to discernible phenotypic DC lineage commitment. Exposure of fresh mouse BM to Flt3L + IL-6 triggered multilog expansion of CD34^pos progenitor cells, and committed nearly all cells to subsequent DC differentiation. Such programming included subsequent spontaneous up-regulation of MHC/costimulatory molecules, as well as nearly uniform responsiveness to DC1 polarization stimuli. Moreover, proliferative conditioning with Flt3L + IL-6 conferred progressive resistance to many tumor-associated immunosuppressive factors, as well as the capacity to respond to either tumor contact or to TGFβ with facilitated DC1 polarization.

Flt3L + IL-6 treatment produced identical licensing whether performed upon unfractionated BM or upon sorted CD34^pos cells, indicating that more primitive (CD34^neg/Sca-1^pos) hematopoietic stem cells were not essential conditioning targets. Pronounced biasing toward ultimate DC differentiation was observed whether step 1 conditioning was performed in Flt3L + IL-6 or Flt3L alone. However, in the absence of coconditioning by IL-6, Flt3L-induced DC precursors remained scant in number and displayed early maturation, whereas the inclusion of IL-6 produced an immense proliferative pool of CD11c^neg/MHC class II^neg cells that were nonetheless already committed to subsequent DC differentiation. Such still phenotypically undifferentiated cells displayed the capacity both for wide distribution following intravenous injection and for spontaneous DC differentiation/maturation after entry into tumor.

IL-6 and its receptor transducing component gp130 have long been recognized to synergize for stem cell proliferation with receptor tyrosine kinase–activating stimuli, including both c-kit ligand (SCF) and Flt3-L,^18-20  but only combined Flt3L + IL-6 also preconditioned for nearly global differentiation of expanded CD34^pos cells into spontaneously maturing DCs. Paradoxically, although GMCSF also synergized with Flt3L to induce step 1 stem cell proliferation, GMCSF antagonized every DC-licensing effect attributable to Flt3L.

Our evidence supports the existence of at least 2 discrete pathways for conventional DC differentiation with strikingly divergent functional outcomes, a Flt3L-promoted STAT3-dependent pathway and a GMCSF-promoted STAT5-dependent pathway (schematized in Figure 7). The Flt3L-promoted pathway was highly potentiated by IL-6, most likely due to the latter's capacity to confer sustained STAT3 activation during extended CD34^pos cell proliferation, and biased for differentiation into DC precursors that were uniformly facilitated for DC1 polarization, spontaneous DC maturation, and activation by tumor contact. In contrast, the GMCSF-promoted, STAT5-dependent pathway gave rise to diverse myeloid elements, including DCs that failed to mature in the absence of further downstream driving signals such as TLR agonists.

We hypothesize that STAT5-dependent DC differentiation constitutes the normally dominant pathway, based on several observations: (1) the inclusion of GMCSF during CD34^pos progenitor cell proliferation blocked STAT3 activation and promoted STAT5-dependent myeloid differentiation regardless of whether Flt3L or Flt3L + IL-6 were also present; (2) recent studies indicate that STAT3-dependent DC differentiation is not normally detectable in adult mice.⁴²  Nonetheless, we have observed that BM even from much older mice remains highly responsive to exogenous Flt3L or Flt3L + IL-6 conditioning (not shown). STAT3-dependent preempting of normal hematopoiesis in favor of dedicated DC production may therefore reflect a reserved host response, perhaps to life-threatening infection of the BM itself. Consistent with this possibility, Flt3L and IL-6 are strongly induced by myelosuppression and BM inflammation, respectively.^43,44 

Despite the competing roles played by STAT3 and STAT5 in DC differentiation, it is also apparent that their respective impacts change before and after DC differentiation occurs. Pronounced sequential STAT3/STAT5 activation (ie, Flt3L + IL-6–conditioned DCs in Figure 6A) corresponded to the greatest observed attainment of DC tumor competence, and neither STAT3 nor STAT5 knockout BM responded optimally to Flt3L + IL-6 conditioning. The preemptive impacts of GM-CSF during step 1 of culture appear to be mediated through early combined STAT3 inhibition and STAT5 activation (Figure 6). Paradoxically, however, appropriately delayed exposure to GMCSF may promote maximal DC1 polarization through the identical STAT modulations (Figure 7).

The mechanism(s) causing Flt3L + IL-6 preconditioned DCs to respond to tumor as a maturational signal remains to be elucidated. A plausible mechanism involves DC activation via enhanced expression of lectin receptors such as asialoglycoprotein receptor and DEC-205 (Figure S3C),^3,45  since identical or similar carbohydrate receptors are used by tumoricidal macrophages to bind and kill tumor cells in an MHC-unrestricted and antigen-unrestricted manner.^46,47  Many tumor cells bind lectin more avidly than nontransformed cells, due to a chronically high density of exposed carbohydrates and a diminished presence of differentiation elements that normally mask such car-bohydrates.^46,47  Similarly, we have observed that exposure to albumin cross-linked with mannose or N-acetyl-glucosamine accelerates the maturation of Flt3L + IL-6–conditioned BM cells (not shown).

We are investigating why most proliferative conditioning treatments caused DC preparations to be highly toxic to T-cell cultures. Such lethality was unlikely attributable to activation of regulatory T cells or indoleamine dioxygenase-expressing plasmacytoid DCs (pDCs), since toxicity was not remedied by adding exogenous IL-2 to the T-cell cocultures.^48,49  There was, however, a correlation between the Gr-1^pos cell–inducing tendency of individual conditioning treatments and observed lethality. It is therefore possible that the Gr-1^pos subpopulation(s) mediate the untoward effects of many of the conditioning treatments.^35,36  Consistent with this possibility, heightened induction of Gr1^pos myeloid suppressor cells appears to be the basis of immune disruption by high GMCSF-producing vaccine formulations.³⁶  In contrast to GMCSF-conditioned cultures, Flt3L + IL-6–conditioned DC cultures were exceptional for their low Gr1^pos content and consistent absence of toxicity, even when added to T cells in high proportions.

The above experiments, as well as preliminary studies with human CD34^pos progenitor cells (not shown), demonstrate that Flt3L + IL-6 may provide an effective means to proliferate, condition, and mobilize highly therapeutic DC precursors for tumor therapy. It has long been appreciated that IL-6 has extremely potent therapeutic properties against established mouse tumors, even when administered as a single agent.⁵⁰  We postulate that the major mechanistic role and benefit of IL-6 therapy will be in tandem with Flt3L to proliferate and condition tumor-competent DC precursor populations in cancer patients.

We thank Dr Walter Storkus for his extremely helpful suggestions.

This work is supported by National Institutes of Health (NIH, Bethesda, MD) grants R01-CA089511, R01-CA103946, and RO1-CA129815.

National Institutes of Health

Contribution: P.A.C. wrote the paper, designed the experiments, performed experiments, and analyzed data; G.K.K., S.S., B.J.C., J.F.T., K.D.B., X.-Y.F., and C.S.C. participated in experimental design and data analyses; and Z.W., W.-J.Z., M.A., C.A.D., H.N., C.C.P., and T.D.J. participated in experimental performance and data analyses.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Peter A. Cohen, Center for Surgery Research, NE6, Cleveland Clinic Foundation, 9500 Euclid Avenue, Cleveland, OH 44195; e-mail: cohenp@ccf.org.