Lost therapeutic potential of monocyte-derived dendritic cells through lost tissue homing: stable restoration of gut specificity with retinoic acid

Human monocyte-derived dendritic cells (DC) (MoDC) are utilized for immunotherapy. However, in-vitro immunological effects are often not mirrored in vivo. We studied the tissue-homing potential of MoDC. Circulating monocytes and DC expressed different tissue-homing markers and, during in-vitro development of MoDC, homing marker expression was lost resulting in a ‘homeless’ phenotype. Retinoic acid (RA) induced gut-homing markers (β7 and CCR9) and a regulatory phenotype and function [decreased human leucocyte antigen D-related (HLA-DR) and increased ILT3 and fluorescein isothiocyanate (FITC-dextran uptake) in MoDC]. RA-MoDC were less stimulatory and primed conditioned T cells with a gut-homing profile (β7+CLA−). Unlike the normal intestinal microenvironment, that from inflamed colon of ulcerative colitis (UC) patients did not induce regulatory properties in MoDC. However, RA-MoDC maintained their regulatory gut-specific properties even in the presence of UC microenvironment. Therefore, MoDC may be ineffectual for immunotherapy because they lack tissue-homing and tissue-imprinting specificity. However, MoDC rehabilitation with gut-homing potential by RA could be useful in promoting immunotherapy in pathologies such as UC.


Introduction
Dendritic cells (DC) are unique in their ability to generate primary immune responses, determining both the nature and type of immune responses [1,2]. Manipulation of DC phenotype and function are promising strategies to perform immunotherapy. However, due their low numbers in human tissues, 'naturally occurring' human DC are not the usual source of DC for human studies. Given their easy accessibility and large numbers, monocyte-derived DC (MoDC) generated from fresh circulating monocytes during 5-7-day in-vitro culture are the standard human DC model [3]. They may be reinjected in vivo to elicit their function. This strategy has been proposed for individualized immunotherapy in several pathologies [4][5][6][7]. Generation of tolerogenic DC may help to restore homeostasis lost in the gastrointestinal tract in patients with inflammatory bowel disease (IBD) [7][8][9][10]. However, despite their in-vitro efficacy, DC immunotherapy is far from ideal when translated to the clinics. Factors affecting their success include prepa-ration, source and dose of DC, route of administration, antigen-pulsing strategy, maturation status and even the selected antigen itself [5]. However, we consider that there is an essential missing factor which may explain the lack of efficacy in vivo: the compartmentalization of immune responses through DC homing marker expression and migration.
Immune responses are elicited in a tissue-/organ-specific way. Immune compartmentalization is controlled by DC which imprint tissue specific homing markers on stimulated T cells, restricting their migration to target tissues [11][12][13][14]. The requirement for tissue specificity for inducing inflammation in the human gut has been demonstrated in patients with Crohn's disease where, despite systemic lymphocyte sensitization to antigens encountered through the gut, inflammation was induced only by challenge with these antigens in the gut, and not by skin challenge [15]. Therefore, appropriate tissue specificity is likely to be required for immunotherapy. However, information about the expression of homing markers on human DC themselves is scarce.
We demonstrated recently that human peripheral circulating DC display a multi-homing profile which may be necessary for them to populate different tissues. However, within skin and gut, DC express tissue-specific markers due to the effect of the surrounding microenvironment [16]. Thus, multi-homing circulating human DC can acquire a phenotype characteristic of gut tissue-specific DC when cultured in the normal tissue microenvironment in vitro [17][18][19][20]. Retinoic acid (RA), the active form of vitamin A, mediates several tolerogenic properties of intestinal DC [14,[21][22][23], including the acquisition of gut-homing properties on human DC and their capacity to imprint guthoming properties on stimulated T cells [16]. However, when human blood DC are cultured in vitro in the absence of tissue factors and/or RA, DC lost expression of homing markers [16].
We hypothesized that MoDC may also lose homingmarker expression upon in-vitro culture during development. Therefore, in this study we have (i) characterized the homing profile of human circulating monocytes and subsequent MoDC in comparison with those of fresh blood DC; and (ii) studied whether we could generate stable gutspecific regulatory MoDC. Our findings revealed that human MoDC displayed a homeless phenotype, but RA rehabilitated them with a regulatory gut-homing profile which was maintained even when exposed to an inflamed microenvironment from ulcerative colitis (UC) patients.
Our findings may open new doors for performing targeted tissue-specific immunotherapy.

Colonic samples and UC biopsy conditioning
Colonic biopsies from inflamed areas of three UC patients were collected in ice-chilled complete medium and cultured within 1 h in complete medium in 24-well culture dishes (one biopsy/well/ml) for 24 h, as explained previously [16]. Complete medium (1 ml) was added to empty wells and incubated for use as a basal condition. Media were centrifuged and either medium-only or cell-free UC colonic biopsy supernatants (UC-SN) were used to condition MoDC for 24 h. All experiments used freshly obtained cultured supernatants or medium. MoDC were also pulsed (t = 2 h) with and without RA (10 -6 M) and washed twice before being conditioned.

Phagocytic activity of DC
Phagocytic activity of MoDC was determined by uptake of FITC-dextran (molecular weight 40 kDa) (Sigma-Aldrich), following the manufacturer's instructions. Briefly, after conditioning of MoDC with different doses of RA, MoDC were incubated with 1 mg/ml of FITC-dextran for 2 h either at 37°C or on ice (internal control), and then washed twice in PBS. Flow cytometry acquisition was performed subsequently.

Flow cytometry and data analysis
Data were acquired on a FACSCalibur cytometer (BD Biosciences) and analysed using WinList version 5·0 software (Verity, ME, USA). The proportion of cells positive for a given marker was determined by reference to staining with an isotype-matched control antibody. For single-parameter analysis, WinList was used to subtract the normal cumulative histogram for isotype control staining from a similar histogram of staining with the test antibody using the superenhanced Dmax (SED) normalized subtraction [24] to obtain the percentage of positive cells for a given marker and the intensity ratio (IR) of the staining of cells identified as positive by the subtraction compared with that of the isotype.

Statistical analyses
One-or two-way paired analysis of variance (anova) corrected for multiple comparisons, or paired t-test, were applied as stated in the figure legends. The level of significance was fixed at P < 0·05.

CCR9 and CCR7 were the predominant homing markers on circulating monocytes
Human circulating monocytes were identified as CD14 + cells by flow cytometry (Fig. 1a). Homing-marker expression was determined by subtraction of isotype control staining from the corresponding test staining (Fig. 1b) to avoid any subjectivity in interpretation of the total percentage of cells which were positive for a given marker. No gender differences in any of the assayed homing markers (12 females and eight males) were detected (data not shown).
Human circulating DC displayed a multi-homing profile characterized mainly by co-expression of gut-homing b7 integrin and skin-homing CLA molecule, as described previously [16]. Conversely, gut-and lymph node-homing markers [chemokine receptors (CCR)9 and CCR7, respectively] were the predominant homing markers expressed by circulating monocytes from healthy controls (Fig. 1c). CCR5 (involved in migration to sites of inflammation) and skin-homing molecule CLA were also expressed on some circulating monocytes. CCR10 (associated with migration to skin and mucosal sites), b7 integrin (gut-homing), CCR4 and CCR8 (both skin-homing) were virtually absent on circulating monocytes (Fig. 1c).

Monocytes lost homing marker expression upon differentiation into MoDC
Human monocytes were cultured for 5 days in a classic in-vitro culture in the presence of IL-4 and GM-CSF to generate MoDC as assessed by up-regulated HLA-DR and loss of expression of the monocyte marker CD14 with time ( Fig. 2). During in-vitro differentiation, monocytes lost expression of all assayed homing markers, leaving day 5 MoDC with a 'homeless' profile ( Fig. 2).

RA induced an immature, gut-homing phenotype on MoDC
We next studied whether RA could induce a gut-homing profile on MoDC. None of the assayed RA concentrations had any effect on cell survival, as determined by total cell numbers (data not shown).
When cultured with different doses of RA, MoDC acquired expression of both b7 and CCR9 gut-homing markers in a dose-dependent manner (Fig. 3a). We also assessed the maturation status of RA-conditioned cells. Because all MoDC expressed HLA-DR and the inhibitory receptor ILT3 (data not shown), these markers were studied by IR compared to the isotype control. Expression of both markers increased in a dose-dependent fashion following RA treatment (Fig. 3b), while RA also increased the phago-cytic capacity of MoDC, a characteristic of immature DC for sampling antigen (Fig. 3b).

RA-conditioned MoDC had decreased T cell stimulatory capacity and primed T cells with a gut-homing profile
The stimulatory capacity of MoDC for allogeneic CFSElabelled T cells and the homing profile of stimulated T cells were also studied. Consistent with the induced immature phenotype (Fig. 3b), RA decreased the stimulatory capacity of MoDC in a dose-dependent manner (Fig. 4a). We also studied the homing-imprinting property of MoDC on stimulated T cells (CFSE low ). MoDC-stimulated T cells (CFSE low ) expressed CLA only, b7 only or were doublepositive or double-negative for both homing markers. However, RA-conditioned MoDC, which had acquired a gut-homing phenotype (Fig. 3a), also induced a guthoming phenotype on stimulated T cells via decreased expression of CLA, rendering most of the stimulated T cells potentially gut-homing with a b7 + CLAphenotype (Fig. 4b). T cells stimulated by RA-conditioned MoDC were mainly b7 + CLA -, so we also studied the intensity of b7 expression via IR, proving that b7 expression was also increased upon stimulation with RA-treated MoDC (Fig. 4c).

RA-MoDC maintain their decreased stimulatory and b7 T cell imprinting capacity in the presence of LPS
We next studied if the decreased stimulatory capacity of RA-MoDC (Fig. 4a) was maintained in the presence of a proinflammatory stimulus such as LPS. Also, as b7 was expressed on both resting T cells and most dividing T cells (irrespective of MoDC conditioning), we determined whether b7 expression was imprinted by MoDC or due to the selective survival of b7 + T cells.
b7 -T cells were enriched from total PBMC (Fig. 5a) and stimulated with MoDC that had been conditioned previously for 24 h with/without RA. Monocyte-derived DC had also been conditioned with LPS (0·1mg/ml) for 24 h with/ without previous pulsing with 10 -6 M RA for 2 h. Pretreatment of MoDC with RA abrogated the increased stimulatory capacity induced by LPS (Fig. 5b). RAconditioned DC also abrogated CLA expression on b7 -T cells and induced b7 expression, even following MoDC     LPS conditioning after RA treatment (Fig. 5c,d). Thus, RA specifically confers the ability of DC to prime gut-homing T cells.

RA-MoDC maintained their gut-homing, immature phenotype in the presence of colonic biopsy supernatants from UC patients
The ability of RA to induce gut-homing and gutimprinting properties on MoDC may enable their use in immunotherapy strategies to treat inflammatory pathologies of the gastrointestinal tract, such as UC. However, DC phenotype and function can be altered by the tissue microenvironment [25]. Thus, although RA-MoDC maintained their regulatory capacity even in the presence of LPS (Fig. 5b), once DC enter the inflamed mucosal sites in vivo they could potentially acquire proinflammatory properties and even contribute to the ongoing inflammation [26,27]. Supporting this effect of the inflamed microenvironment on DC, when human blood-enriched DC are exposed colonic microenvironments from healthy controls [16] or non-inflamed areas from UC patients [28] they acquire an immature phenotype. However, when conditioned with inflamed areas from UC patients, these properties are not acquired. In this case, DC do not decrease their stimulatory capacity and, furthermore, DC acquire an increased capacity to prime responding T cells with a skin-homing profile [28].
To address the effect of the UC microenvironment on MoDC, cells -previously pulsed with and without 10 -6 M of RA -were conditioned with supernatants from colonic biopsies of UC patients (inflamed areas; UC-SN). Similar to blood-enriched DC [28], MoDC conditioned with UC-SN did not differ in their stimulatory capacity from basal MoDC (Fig. 6a). However, when MoDC had been pulsed previously with 10 -6 M RA, their stimulatory capacity remained decreased compared with both basal (P < 0·05 at 3%) and no-RA supplemented with UC-SN MoDC (P < 0·05 at 2% and P < 0·01 at 3% (Fig. 6a). RA-treated MoDC retained their capacity to induce a gut-homing profile on conditioned T cells following UC-SN conditioning with virtually none of the stimulated T cells expressing CLA, unlike those MoDC conditioned with UC-SN without previous RA pulsing (Fig. 6b,c).
Finally, we studied the cytokine profile of stimulated T cells. Pulsing MoDC with RA prior to UC-SN conditioning enabled MoDC to stimulate T cells with no proinflammatory cytokine profile; the T cells had reduced IFN-g production and increased IL-10 production compared with T cells stimulated by MoDC conditioned with UC-SN only corrected for multiple comparisons was applied on pooled data from three independent experiments. P-value below 0·05 was considered statistically significant; *P < 0·05; **P < 0·01.
( Fig. 6d,e). Therefore, RA-conditioned MoDC exposed to an inflamed microenvironment showed no evidence of reversion towards a proinflammatory phenotype.

Discussion
Human circulating DC display a multi-homing b7 + CLA + profile [16] which was not shared by monocytes where CCR9 and CCR7 were the main homing markers. The differences between homing properties of DC and monocytes may be determinants of their different tissue distributions. Monocytes differentiated in vitro in the presence of IL-4 and GM-CSF lost expression of their homing markers and failed to express those found on normal circulating DC, rendering MoDC with a homeless phenotype. The complete loss of homing markers studied in MoDC contrasts with the multi-homing potential of circulating DC [16]. The homeless phenotype of MoDC is likely to indicate incapacity to home appropriately to skin or mucosal surfaces by these in-vitro-generated DC. MoDC also failed to induce any tissue-specific homing profile of stimulated T cells as opposed to tissue-resident DC [16]. This blanket loss in homing potential may limit the in-vivo usefulness of these cells for immunotherapy. Human circulating monocytes and DC have different immunological properties but also different homing marker expression. Although both populations have 'mixed' or multi-homing profiles, the homing markers that they express are different. Circulating DC are mainly b7 + CLA + [16] (gut-and skin-homing). In contrast, both homing markers are hardly expressed by circulating monocytes, which are mainly CCR7 + CCR9 + (lymph node-and small bowel-homing markers) (Fig. 1). These differences might account for different tissue distributions in vivo. During in-vitro culture, in the absence of external factors monocytes lose their homing marker expression (Fig. 2), rendering MoDC with a homeless phenotype different from that of circulating DC. We cannot exclude expression of other homing markers by MoDC. However, it is significant that all studied homing markers were lost in time following in-vitro differentiation to MoDC and that similar loss of homing markers has been described in cultured blood DC [16]. Therefore, the loss of homing markers following in-vitro culture in the absence of tissue and/or homing-inducing factors may be a general mechanism in antigen-presenting cells. Utilizing DC as therapeutic agents is effective in autoimmune and inflammatory diseases, and DC are now even being used to enhance survival rates of solid organ transplants [5,7]. Practical issues of the small numbers of DC obtainable from tissues predicate the use of MoDC for immunotherapy in humans. However, we demonstrate here that, unlike human blood and/or tissue DC, MoDC are devoid of homing marker expression. Thus, when reinjected in vivo into patients the lack of homing marker expression may abrogate their trafficking into effector sites, providing an explanation for the limited success of DC immunotherapy strategies in vivo [3].
DC control the migration of antigen-specific T cells to target tissues by imprinting their own homing markers on T cells that they stimulate [16], so antigen-specific T cells are directed back to the target tissue where the antigen had been sampled. The last aim of immunotherapy approaches utilizing MoDC is to trigger T cell specific responses. Subcutaneous immunization is the most clinically relevant administration route. However, MoDC must enter lymph nodes and present antigen, so MoDC used in immunotherapy are usually matured with cytokine cocktails [29] to induce CCR7 expression so that they acquire the capacity to migrate to the lymph nodes and are able to stimulate a T cell response. This approach is likely to abrogate the immature phenotype of MoDC. In addition, it is not only a matter of triggering immune responses, but also of directing them to the target tissues; although these DC may enter lymph nodes directly and stimulate T cells, they are unlikely to give tissue-homing potential to those stimulated cells, as it is the source of DC and not the lymph node cells or stroma that determines the homing property of the stimulated T cells [11]. Thus, even if MoDC enter the lymph and generate antigen-specific T cells, the lack of homing markers of MoDC may cause the failure to induce a tissuespecific homing profile on T cells.
An alternative approach would be to induce homing markers directly on MoDC, so that they would migrate to the target tissue and once there, following conditioning by the tissue microenvironment [16], enter the respective lymph nodes (mesenteric and/or Peyer's patches in the gut) to perform antigen presentation and imprint their own homing profile on responding T cells. However, intestinal tissue microenvironment not only induces CCR7 expression on DC [16], it also affects their phenotype and function; when blood DC are conditioned with inflamed colonic microenvironment from UC patients, DC lose their homeostatic properties and imprint a skin-homing profile on T cells in the absence of RA [28]. Therefore, a major concern regarding use of DC for immunotherapeutic purposes in IBD is that RA-MoDC could potentially adopt a proinflammatory phenotype and function in vivo in the inflamed colonic microenvironment, contributing to the ongoing inflammation [26,27]. However, RA-MoDC maintained their regulatory phenotype even in the presence or an inflamed microenvironment derived from UC patients (Fig. 6); a gut-specific immunomodulation might thus be achieved with such cells, negating the necessity for systemic immunosuppression produced with current treatments. However, a recent study demonstrated that RA-conditioned DC adopt proinflammatory properties in the presence of IL-15, abrogating oral tolerance to dietary antigens [30]. Several differences between these studies may account for the discrepancies. The major difference is the use of human material for this study compared with murine models; data from animal models may not be transferrable directly into human mechanisms of disease [31]. Also, the MoDC in this study were exposed to an inflamed colonic microenvironment derived directly from human UC colonic biopsies rather than IL-15; our previous studies suggest that IL-6 is the dominant cytokine in UC pathogenesis [28]. Finally, higher RA concentrations were used to pulse MoDC in this study. These factors may explain why RA-pulsed DC were not recruited towards a proinflammatory phenotype when exposed to an intestinal microenvironment from UC patients.
RA is a key player of immune homeostasis in the gut controlling immunoglobulin (Ig)A switching of B cells, the development of T cells with a regulatory phenotype/ function and induction of gut-homing properties of stimulated B and T cells [14,[21][22][23]32]. RA also mediates acquisition of gut-homing properties on human DC and regulates their ability to imprinting homing properties onto stimulated T cells [16,33], while decreasing their T cell stimulatory capacity [34], presumably through secondary transforming growth factor (TGF)-b induction [35]. We are aware that we did not study the capacity of RA-MoDC to induce regulatory T cells. That is because our recent evidence has shown that forkhead box protein 3 (FoxP3) expression may not be a valid marker to identify T cells with a regulatory phenotype; because its expression is dynamic, dependent upon T cell densities, it is subsequently lost in time following removal of the stimulant, and FoxP3 + T cells developed in vitro do not perform a suppressive activity [36]. However, we have shown that RA-MoDC displayed a decreased stimulatory capacity in a dose-dependent fashion (Fig. 4), even in the presence of LPS (Fig. 5) or an inflamed microenvironment from UC patients (Fig. 6), and T cells that they stimulated decreased their IFN-g production and increased their IL-10 content (Fig. 6). Together, these results suggest that RA had the capacity to rehabilitate homeless MoDC with an immature and gut-homing phenotype and the capacity to generate gut-homing T cells that were not proinflammatory.
In summary, in addition to factors usually considered for utilizing DC in immunotherapy, the homing profile of DC themselves and, more importantly, their capacity to imprint homing profiles on the T cells they stimulate will influence a DC immunomodulatory approach. Our data suggest that RA conditioning may increase the efficacy of MoDC as therapeutic agents, providing tissue specificity and sustained homeostatic effects for DC-specific therapy targeted at mucosal inflammatory diseases with tissue compartmentalization. Therapeutic approaches for IBD would, ideally, involve gut-specific prevention of dysregulated inflammation without compromising immunogenicity towards harmful or invading pathogens; our data provide a potential method of doing so while minimizing systemic immunosuppression, which is a side effect of almost all current IBD therapies. Our findings could also be extended to other gastrointestinal autoimmune disease such as Crohn's disease and/or coeliac disease. Future studies may identify ways of inducing other tissue-specific homing markers in order to develop tissue-homing specific MoDC to perform tissuespecific immunotherapy.