Large-Scale Culture and Selective Maturation of Human Langerhans Cells from Granulocyte Colony-Stimulating Factor-Mobilized CD34⁺ Progenitors

Materials and Methods

Results

Before autologous stem cell transplantation, cancer patients are typically treated with G-CSF to mobilize large numbers of progenitor cells into the peripheral blood (24). Given the quantity and availability of these preparations from patients and normal volunteers, we investigated their utility as a source of DCs following in vitro differentiation using a modification of techniques previously described for the differentiation of CD34⁺ cells from cord blood (16). CD34⁺ cells from leukapheresis product were immunopurified to >95% homogeneity and then seeded in 24-well plates at 1 × 10⁴ cells/well in serum-free medium (X-VIVO 15) containing GM-CSF, TNF-α, TGF-β, stem cell factor, and Flt3L. Cells were grown for 7–10 days without further manipulation.

By day 7, total cell number had increased 50- to 100-fold with 35–60% of the cells being positive for CD1a. A representative profile of cell-surface expression for various markers at day 8 is shown in Fig. 1⇓A. Interestingly, and in contrast with the majority of other methods for producing DCs from CD34⁺ precursors, most of the cells in culture were HLA-DR^dim as expected for immature DCs. No monocyte- or granulocyte-related cells were detected in culture as indicated by the absence of CD14⁺ and CD66b⁺ cells; these markers were also not detected at earlier times of culture (days 2–6, data not shown), suggesting that there was not even transient expression of early granulocyte markers. The majority of the cells were positive for CD13, suggesting that they were myeloid in origin. Interestingly, most of the cells were CLA⁺, a feature consistent with their being LC-committed precursors (25).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

Expression of cell-surface markers by human DC cultures. A, CD34⁺ cells were seeded at 1 × 10⁴ cells/ml and cultured for 8 days in the described growth medium (serum-free X-VIVO plus cytokines). Cells were then harvested, and the expression of HLA-DR, CLA, CD13, CD1a, CD14, and CD66b was determined by flow cytometry. Thin lines represent control Abs. B, Undisturbed control cells were stained at days 10 and 12 (day 10, thin line; day 12, thick line). Alternatively (right two panels), day 10 cultures were harvested at day 10 and recultured for 2 days at the same density in fresh growth medium (X-VIVO plus cytokines) or RPMI 1640 plus 5% FCS. Cells were stained for CD1a on day 10 (thin line) and day 12 (thick line). Dotted lines represent isotype controls. CD1a expression, consistent with the LC phenotype, was not lost upon reculture.

An important feature of the CD1a⁺/CLA⁺ cells (presumptive LCs) was their phenotypic stability. As shown in Fig. 1⇑B, CD1a expression did not decrease with time in culture when day 10 cells were replated either in serum-free X-VIVO or in RPMI 1640 containing 5% FCS. However, the serum-fed cells exhibited a marked increase in surface HLA-DR (data not shown), suggesting that maturation had been induced by a cytokine or LPS contaminant in the serum. In any event, it is clear that the presumptive LCs could be maintained, and apparently in an immature state, for days if grown in a defined serum-free medium (see below).

We next performed a more detailed analysis of DCs generated from the mobilized CD34⁺ precursors. Specifically, we were interested in determining whether homogenous populations of presumptive LC precursors could be enriched from the cultures, maintained in an immature DC phenotype and then triggered to mature in a synchronous fashion. From day 4 on, loosely adherent aggregates of cells appeared in culture, which increased in size and number with time. At day 7, presumptive LCs were found only in multicellular clusters, as indicated by staining with CD1a and with the unique LC marker Lag (18) (Fig. 2⇓, A–D). The cluster cells also exhibited most of their MHC class II intracellularly with very little on the plasma membrane, typical of immature DCs (3). These features were highly reminiscent of the DCs produced from mouse bone marrow in which proliferating and differentiating DCs accumulate as immature DCs in analogous cell clusters (26).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2.

DCs within cell clusters exhibit a LC phenotype. Proliferating cultures (day 8) contain abundant multicellular aggregates of HLA-DR⁺ cells that express the LC markers CD1a and Lag as shown by immunofluorescence microscopy (B and D), while single cells outside of the clusters generally expressed only HLA-DR (A and C). Note how one cell in the center of the cluster shown in A and B, most likely a proliferating precursor, is negative for both differentiation markers. E, Day 10 cells were harvested and subjected to sedimentation on 7.5% BSA columns resulting in an enrichment of CD1a⁺ cells from 55 to 89%. Cells were then replated for 2 days in the same medium. The majority (≈90%) of cells after cluster purification and reculture are also Lag⁺. F, Electron microscopy analysis of LC in culture reveal the presence of abundant Birbeck granules (indicated by arrowheads). G, Only BGs (and not other structures such as multilamellar lysosomes, indicated by asterisks) were heavily decorated by anti-Lag Abs. However, the multilamellar structures did label with Abs to MHC class II molecules (not shown).

To enrich for CD1a⁺ cells, the clusters were purified by 1 × g sedimentation on 7.5% BSA columns (see Materials and Methods for details). FACS analysis revealed that this procedure produced excellent yields both in term of number of cells and purity (70–90% CD1a^+/HLA-DR^dim) (Fig. 2⇑E). The LC marker Lag was expressed by >70% of the cluster-purified cells. Electron microscopy of the purified cells revealed the abundant presence of pathognomonic LC-specific organelles, Birbeck granules (Fig. 2⇑F, arrows). The granules persisted for several days in immature LC in culture (data not shown). Birbeck granules were found by protein A-gold immunocytochemistry of cryosections to be specifically labeled by anti-Lag Abs (Fig. 2⇑G) but negative for MHC class II or Lamp 1 (not shown).

If left undisturbed, cells in the proliferating clusters exhibited this immature LC phenotype for up to 12 days. However, upon disruption of the clusters by vigorous repeated pipetting, the cells appeared to mature as indicated by increased surface expression of MHC class II and other markers of mature DCs (CD86, CD40, CD80) (Fig. 3⇓B). Yet, when clusters were purified and gently replated in fresh growth medium (X-VIVO plus cytokines), the immature phenotype was maintained and spontaneous maturation was not observed, as indicated by the low level of HLA-DR and costimulatory molecules expressed on the cell surface (Fig. 3⇓A). Previous attempts to cultivate human DCs from CD34⁺ precursors have yielded only mature cells with high levels of cell-surface MHC class II products and costimulatory molecules (14, 16). Conceivably, this difference may reflect the sensitivity of DCs to physical manipulation, which might disrupt and activate cell clusters.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3.

Maturation profile of human LC-type DCs. Purified clusters were replated at 5 × 10⁵ cell/well and cultured for 2 days. Replating was performed very carefully in order not to break the clusters (A) or after cluster disaggregation by repeated pipetting (B). Thin lines are cells stained immediately after cluster purification; thick lines are cells stained after 2 days in culture. Breakage of clusters induced spontaneous up-regulation of specific maturation markers. C, Unbroken clusters can be induced to mature by addition of 250 ng/ml LPS (thin line) or 12.5 ng/ml TNF-α (thick line) to the regular growth medium (dotted line). D, Clusters were cocultured with resting (thin line) or activated (thick line) platelets in RPMI 1640/5% FCS or simply replated in regular growth media (dotted line).

The controlled maturation of gently purified cluster LCs could be accomplished by using proinflammatory agents such as TNF-α, LPS, or CD40L. Fixed, activated human platelets were used as a source of membrane-bound CD40L (27). After 2 days, FACS analysis demonstrated that these agents greatly enhanced the expression of HLA-DR, costimulatory molecules, and CD83 on the plasma membrane (Fig. 3⇑, C and D and Fig. 5⇓E).

Thus, as previously shown for monocyte-derived DCs, epidermal LCs, and bone marrow-derived DCs, treatment with TNF-α, LPS, or CD40L induced a terminal DC differentiation from an immature to a mature phenotype. Although each of these treatments induced qualitatively similar differences when assayed by FACS, analysis by confocal microscopy revealed significant differences among the differently treated cells, in terms of both morphology and maturation stage. Cells from purified clusters, replated in regular growth medium, exhibited the phenotype typical of early DCs: class II is intracellularly distributed and perfectly colocalized with lysosomal markers (Fig. 4⇓, top row) (3). Following TNF-α treatment (12.5 ng/ml), nearly all of the cells exhibited the “intermediate” phenotype previously observed only in rat and mouse DCs. The intermediate DCs were characterized by the accumulation of MHC class II on the surface as well as in peripheral intracellular vesicles (class II vesicles or “CIIV”) devoid of lysosomal markers (Fig. 4⇓, second row). The lysosomes were largely depleted of MHC class II and clustered in the perinuclear cytoplasm. The accumulation of cells in this intermediate phenotype suggests a role for TNF-α in the early but not the late events of DC or LC activation. Thus, TNF-α may be unable to drive the maturation process to completion.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4.

Different proinflammatory agents induce morphologically distinct maturational states of human LCs. Purified clusters were cultured for 2 days in fresh growth media alone (top row) or in presence of 12.5 ng/ml TNF-α (second row), 250 ng/ml LPS (third row), or activated platelets (bottom row). Cells were then fixed and stained for confocal immunofluorescence microscopy. In the merged images, HLA-DR staining is shown in green and Lamp-1 (CD107a) staining in red.

LPS-treated LCs were also extremely homogeneous in morphology, but their appearance was quite different from TNF-α-treated cells. The maturation driven by LPS seemed to be more advanced; the cells appeared smaller, but with abundant phyllopodia and more of the MHC class II at the plasma membrane than in the TNF-α-treated cells (Fig. 4⇑, third row). Therefore, the LPS-treated cells were reminiscent of the mature or “late” stage identified in the murine system, although the structure of the lysosomal compartment in LPS-treated human DCs was much less compact, and indeed more similar to what was observed in intermediate cells.

Activated platelets, a source of CD40L, induced the most dramatic change in cell morphology (Fig. 4⇑, bottom row). Again, virtually all of the MHC class II was found at the plasma membrane, which was now organized into long dendrites. Class II-negative lysosomes were tightly clustered in the perinuclear region.

Importantly, these mature phenotypes were stable, with the LCs expressing high levels of surface HLA-DR and maintaining CD1a for days (data not shown). Indeed, simply plating clusters in RPMI 1640 supplemented only with FCS, as opposed to cytokine-supplemented X-VIVO, supported LC maturation. Together, these findings indicate that LCs prepared in this way are irreversibly committed to a DC phenotype.

We then asked if the observed differences in morphology and cell organization corresponded to differences in function. We tested the T cell stimulatory capacity of the differently treated DCs in an alloreaction using CD8⁺ T cells. DCs were cluster purified at day 7 and replated in regular growth media or in the presence of LPS or TNF-α; on day 9 DCs were mixed with CD8⁺ T cells. After 2–6 days, cells were harvested, washed extensively, and proliferation was evaluated. As shown in Fig. 5⇓A, LPS-treated DCs were the most efficient in stimulating CD8⁺ T cell proliferation when compared with control or TNF-α-treated DCs, although the differences were not very pronounced. In fact, even control DCs (i.e., cells not treated with TNF-α or LPS before assay) were able to support a T cell response.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 5.

LC maturation induces enhanced CD4 and CD8 T cell alloreaction. Representative data from three experiments is shown. A, Differentially treated DCs were harvested and cocultured with freshly isolated CD8⁺ T cells (see Results and Materials and Methods for details). Proliferation at day 2, 4, and 6 was measured by [³H]thymidine incorporation. B, Proliferation of freshly isolated CD8 and CD4 T cells were measured in response to fixed DCs. C, IL-2 production was also measured for CD4⁺ T cells. D, Proliferation at day 6 of CD4⁺ T cells (300,000 cells/well) was measured in response to the indicated amount of stimulating DC. E, The ability of TNF-α and LPS to induce differential T cell responses did not reflect differences in the quantitative expression levels of surface markers. Surface staining for MHC class I, MHC class II, CD58, and CD83 are shown for control cells (dotted line), TNF-α-treated cells (thick line), and LPS-treated cells (thin line).

Interaction with T cells alone is almost certainly capable of inducing DC maturation due to the presence of CD40L on T cells. Thus, it was likely that maturation of even the untreated control DCs occurred during the course of the proliferation assay, obscuring any possible differences in stimulatory capacity. To circumvent this problem, we decided to use for the assay previously fixed DCs to prevent further maturation due to T cell contact. As before, control, TNF-α-, and LPS-treated DC were harvested, but fixed with 0.5% PFA and washed extensively, before coculture with freshly isolated CD4⁺ and CD8⁺ T cells. Although the magnitude of the T cell proliferative response was markedly lower using fixed DCs, the results clearly indicated that both CD4⁺ and CD8⁺ responses were completely dependent on prior exposure of the DCs to a maturational stimulus with LPS-treated DCs being the most efficient (Fig. 5⇑B). Similar results were obtained when IL-2 production by CD4⁺ cells was assayed (Fig. 5⇑C). Despite the differences in stimulatory capacity between TNF-α- and LPS-treated DCs, both treatments caused equivalent up-regulation of MHC class I, MHC class II, LFA-3 (CD58), and the DC-maturation marker CD83 (Fig. 5⇑E). Although just 10% of the total T cell stimulatory activity was left upon DC fixation, the magnitude of these responses was still clearly dependent on the number of DCs added to the assay (Fig. 5⇑D). The T cell stimulatory activity of human DC seems to be much more affected by PFA fixation than what has previously been shown for mouse DCs (3). While this difference may reflect differences between the mouse and human DCs used, it should also be noted that the murine DC studies were conducted using T cell hybridomas as a read out, as opposed to the primary T cells used here.

One interesting feature of our culture system is that the cells are sensitive to LPS, although they do not express CD14, generally thought to be an important LPS receptor. Because the cells were grown in serum-free media, another potential mediator of LPS responsiveness, LPS binding protein, was also unlikely to play a role (28). Recently, a new family of molecules involved in the response to LPS, the Toll-like receptors (TLR1 through TLR5), has been cloned and partially characterized (29). Therefore, we investigated which, if any, TLRs were expressed by our in vitro-derived LCs and if their levels of expression level was affected by maturation. Using specific oligonucleotide primers for each of the human TLR molecules, RT-PCR was performed on cDNA obtained from cluster-purified cells treated for 48 h with or without TNF-α or LPS. As shown in Fig. 6⇓A, we observed distinct expression patterns for the five receptors depending on the type of inflammatory agent used. While the expression of TLR1 mRNA was almost unchanged regardless of whether TNF-α or LPS were added to the cultures, LPS but not TNF-α induced a dramatic down-regulation of TLR2. Conversely, expression of TLR3 and TLR4 were up-regulated to varying extents by both mediators, but more so by LPS than by TNF-α. Thus, TNF-α and LPS exert markedly different effects on the expression of individual members of the TLR family, further emphasizing that different mediators can elicit qualitatively different types of DC maturation.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 6.

A, Differential expression of TLR genes following maturation by LPS and TNF-α. Cluster-purified LCs were replated in X-VIVO (control, C) or X-VIVO containing LPS (L) or TNF-α (T) for 48 h, and expression of TLR1–5 was determined by RT-PCR as described in Materials and Methods. Amplification of actin mRNA was included as an internal control. B, LPS-matured, but not TNF-matured, LCs produce IL-12. Cells were matured (as above) with either 12.5 ng/ml of TNF-α or 10 μg/ml of LPS. Supernatants were assayed for IL-12 content by ELISA. Representative data from two experiments is shown.

Finally, we examined whether cytokine production by the LCs might be differentially regulated depending on the nature of the maturational stimulus. We assayed the production of IL-12, and found that only LPS-matured, but not TNF-α-matured, DCs produced significant levels of IL-12 (Fig. 6⇑B).

Discussion

It is becoming apparent that DCs are not a single, homogeneous cell type, but rather a system of cells of multiple lineages and activation states (30). Although different DC populations are becoming associated with different immunological outcomes, the cell biological basis for these differences remains poorly understood. In humans, only DCs derived from peripheral blood monocytes by growth in GM-CSF and IL-4 have been studied in any detail. Despite several attempts, it has thus far proved difficult to establish an analogous system to study the differentiation and early stages of maturation of human DCs all the way from CD34⁺ precursors. Several groups have differentiated DCs in vitro, even DCs of the LC type, from CD34⁺ cells isolated from cord blood. However, previous approaches yielded relatively small numbers of heterogenous cells that invariably exhibited a high level of MHC II expression by the end of the isolation procedure, precluding the possibility of analyzing events related to maturation (12, 14, 15). We have adapted this approach to permit the generation of larger numbers of virtually homogeneous DCs from CD34⁺ cells isolated from apheresis products. Based on their surface markers and abundance of intracellular Birbeck granules, the cells were indistinguishable from epidermal LCs. The cells were not only irreversibly committed to the LC lineage, but also arrested in an immature state (low surface MHC class II and accessory molecules) unless activated by proinflammatory mediators or physical disruption.

Maturation was induced by three different mediators, TNF-α, LPS, and CD40L. Each of these agents generated mature cells that were similar on the basis of their surface marker profiles as determined by flow cytometry. However, several striking differences were uncovered by more detailed cell biological and functional analysis. By confocal microscopy, each agent was found to produce cells exhibiting markedly different morphologies. Perhaps most important was the finding that TNF-α treatment drove the cells no further than the “intermediate” DC phenotype, characterized by an accumulation of nonlysosomal intracellular vesicles that, in mouse bone marrow-derived DCs, carry peptide-loaded MHC molecules to the plasma membrane (3).⁵ Thus, TNF-α may be capable of inducing only the earliest steps in DC maturation. This is in contrast to LPS and CD40L that can convert immature cells, which typically localize the bulk of the MHC class II products in late endosomes and lysosomes, to mature cells that express little if any intracellular class II (4, 11, 31).

The phenotypic differences observed were reflected, at least in part, at a functional level. Using an allogeneic response as an assay, diminished T cell stimulatory activity was consistently observed when TFN-α-treated rather then LPS-treated DCs were used as APCs. We also observed that the important Th1-driving cytokine, IL-12, was produced only by LPS-treated DCs and not by TNF-α-treated cells.

We also found that the cells were markedly different in another feature of likely functional importance, namely the expression of TLR family members (32, 33). Recent evidence strongly suggests that TLRs play critical roles in innate immunity by providing a means whereby cells of the innate immune system distinguish between self and invading microorganisms (34). LPS is thought to bind and activate at least two members of this family, TLR2 and TLR4 (35, 36). Thus, the expression of these two receptors and their relatives can be expected to play an important role in DC function. Interestingly, we found that LPS and TNF-α elicited markedly different patterns of TLR gene expression following induction of DC maturation. In particular, LPS induced the complete down-regulation of mRNA encoding one of its presumptive receptors TLR2, while TNF-α had no effect on TLR2 expression. On the contrary, TLR4 is completely absent in immature cells, and its expression is differentially induced upon maturation. That LPS induced higher levels of TLR3 and TLR4 than did TNF-α implies a further qualitative difference in the effects of these mediators, but also may reflect the ability of LPS to drive maturation to “completion” or at least beyond the intermediate phenotype elicited by TNF-α.

The functional consequences of TLR expression on DC or LC function are not yet known, although it does seem likely that these alterations will control one or more important activities. Cella et al. (37) have recently found that influenza virus infection or double-stranded RNA can cause the maturation of human monocyte-derived DCs in a fashion that favors their ability to selectively trigger the development of Th1 T lymphocytes. This observation also clearly illustrates that different DC activators can generate mature DCs with important functional differences.

The advent of a cell culture system that allows the differentiation and controlled maturation of DCs from uncommitted precursors itself has created some interesting opportunities. First, because the cells produced under the conditions used in our experiments appear identical with epidermal LCs, it should now be possible to apply the same cell biological and functional techniques to the study of LCs as have previously been used to study monocyte-derived or bone marrow-derived DCs. Cell fractionation has already enabled us to isolate highly purified fractions of Birbeck granules, which can now be analyzed biochemically in vitro (M.A.V., M.W.E., and I.M., unpublished observations).

In addition, our system for LC production from mobilized CD34⁺ cells may have good potential for clinical applications. Not only can large numbers of autologous cells be produced with reasonable ease, but the fact that the LCs generated remain immature in culture will allow them to remain in the state best suited for Ag accumulation. Also relevant to any potential clinical applications was the fact that once differentiated, the LCs we have produced are phenotypically stable. To be maximally effective, DCs must maintain their differentiated features upon reintroduction into the bloodstream or tissues, environments in which the cytokine composition is markedly different from the medium in which they had been cultured. DCs differentiated from peripheral blood monocytes, for example, may rapidly lose their DC features upon removal of IL-4 or GM-CSF, suggesting that they may not resemble DCs for long after in vivo reintroduction. While more work will be required to document the in vivo behavior of any reintroduced DC population, the in vitro stability of the CD34-derived LCs described here suggest that they may have critical advantages as potential immunotherapeutic vehicles.