In the Roadmap study (Kodama et al
., manuscript submitted), IL-4 was one of the very few cytokines whose gene expression was significantly changed over the course of NOD disease at the level of the total PLN. In accordance with many previous observations indicating a defect in IL-4 in both NOD mice and humans with T1D, we observed a deficiency in IL-4 expression in the NOD PLN relative to diabetes-resistant NOD.B10 PLN, at 12 wks of age (by RT-PCR), and during the period of disease onset spanning from 12 to 20 wks of age (Roadmap microarrays). The cause of this deficit in IL-4 expression only in the PLN among tissues examined in diabetes-prone mice relative to diabetes-resistant mice is difficult to explain, because an increase in Th1 cytokines was not detected during that time. Moreover, such deficit was not seen in younger mice (Kodama et al
., manuscript submitted). It is interesting that this reduction in IL-4 expression in the PLN at 12 wks of age precedes the period of onset of hyperglycemia, and follows a period (between 6 and 12 wks) previously characterized by a loss of Th2-priming ability by major β cell autoantigens [36
Several groups have reported successful prevention of diabetes by treatment of young NOD mice (4–5 wks of age) with DCs [37
]. In contrast, such non-modified DCs have not been shown to confer protection to old NOD mice with advanced insulitis. Later, Feili-Hariri et al.
] used DCs adenovirally-transduced to express IL-4 (Ad4.DCs) in NOD mice at 10 wks of age, and obtained significant protection, while control DCs (unmodified or expressing GFP) had no significant effect. Our study not only confirms these observations, but provides substantial additional information. First, adenovirus infection renders the DCs more mature and more immunogenic, whereas lentiviral transduction has no such effect [40
]. Feili-Hariri et al
. reported an enhanced CD40, CD80 and CD86 expression, while these markers, as well as MHC class II, were not or minimally changed in our case. Moreover, lentiviral transduction of DCs may enhance their differentiation, as they were found to express higher levels of CD11c, whereas adenoviral infection tended to reduce CD11c expression [28
]. In support of the immunogenicity issue, Feili-Hariri et al.
] reported that GFP+
Ad.DCs were detected in the spleen and PLN 24h after injection, but had disappeared by 72h, while we could still detect the presence of Luc+
DCs in these tissues for up to a week. Second, we were able to achieve the same level of significant protection as did Feili-Hariri et al.
], ~30% diabetic in treated, compared to ~80% in controls, using fewer administrations of DC/IL-4 in even older mice (one versus two injections; 12-wk old versus 10-wk old). Third, despite a similar level of protection achieved, there was an evident delay of onset seen with DC/IL-4, but not with Ad4.DCs. In both studies, the control DCs (Ad.DCs or DC/GFP) had a noticeable but insignificant protective effect.
Some studies involving transduced DCs in prediabetic NOD mice reported the presence of these cells in both the spleen and PLN after IV injection [28
], but did not address the selectivity of DC migration. We set out to determine whether all lymph nodes were equally targeted using such a systemic route or whether preferential homing specificity existed. To this end, DCs were lentivirally transduced to express luciferase and injected into 12-wk old NOD mice. We then followed the biodistribution of the DCs in live animals by bioluminescence imaging in vivo
, and by analysis of various tissues ex vivo
. Following IV injection, DCs rapidly homed to the most vascularized tissues (lungs, liver and spleen). Importantly, we show that bone marrow-derived DCs do not traffic equally to all lymph nodes. Migration to lung-draining mediastinal lymph nodes [41
] was not so surprising given the strong signal from Luc+
DCs measured in the lungs 24h after injection. However, rapid and specific migration to the PLN was more surprising to us, given that the tissues they are known to drain (pancreas and duodenum) did not appear to harbor Luc+
DCs. Moreover, we have observed PLN-specific homing after IV injection in absence of inflammation (in several non-diabetic strains) and in the absence of lentiviral infection (using DCs from Luc-transgenic mice; Creusot et al
., manuscript in preparation). It is possible that the PLN has an additional uncharacterized drainage or expresses a unique set of chemo-attractants. We are currently investigating what makes the PLN preferentially attract DCs in the absence of inflammation. Although PLN-specific homing appears to be a general phenomenon, it is a property worth exploiting for DC-mediated therapy of T1D, and possibly other pancreatic diseases.
Because of its selective targeting and known relevance to T1D, we have focused on the PLN tissue to address some of the cellular and molecular events that follow the migration of DCs and the delivery of IL-4. To this end, we first set out to determine by microarray analysis whether significant changes in gene expression occur in the NOD PLN during the course of disease, using a control from NOD.B10 PLN as a constant baseline (Kodama et al
., manuscript submitted). For this report, we focused on the genes that were the most over- or under-expressed in NOD PLN at 12 wks of age, compared to the NOD.B10 PLN control. We hypothesized that these changes in expression may constitute a signature associated with the beginning of destructive insulitis (since most of these particular changes were not seen in younger mice). Using an arbitrary threshold of ~3-fold difference in expression level (with a significance value of p<0.01), 221 genes were identified, the great majority of which (214) were under-expressed in NOD PLN (Supplementary data “Restricted list” in file “DC-IL4 IV analyzed.xls”). Using PLN samples obtained 3 days after IV injection of DC/IL-4 in NOD mice, we performed microarray analysis against the same NOD.B10 PLN control, in order to assess the effect of the treatment on the expression level of these genes. We showed that >85% of these 221 genes become significantly normalized (expression driven back towards NOD.B10 levels, with a p value <0.05 between untreated and treated). Less than 1% of the genes had their expression driven significantly further apart from the NOD.B10 levels. Similar results were seen by relaxing the threshold to a 2-fold difference in expression between NOD and NOD.B10 (p<0.01), which yielded 686 genes, >87% of which were significantly normalized by DC/IL-4 treatment (Supplementary data “Expanded list” in file “DC-IL4 IV analyzed.xls”). The possible relevance of these genes in the disease process is discussed elsewhere (Kodama et al
., manuscript submitted). However, it is interesting that few immune-associated genes appear in this list of 221 genes. One example is galectin-1 (Lgals1
), which is known to hinder proliferation and promote apoptosis of activated T cells [42
]. Galectin-1 was significantly under-expressed at 12 wks of age in the PLN, but its expression was corrected by treatment with DC/IL-4. This fits nicely with the report that DCs modified to express galectin-1 can delay the onset of diabetes [32
]. The genes that were normalized do not belong to particular ‘biological process’ or ‘molecular function’ categories (Gene Ontology data, not shown). However, the most under-expressed genes (214 genes) may have more in common in terms of their regulation. For example, a good number of them (Gatm
) are believed to be AIRE-regulated [43
]. All but one (S100a9
) had their expression up-regulated by treatment and half of them significantly, although the expression of Aire
was not changed (p=0.93). Furthermore, many among the most under-expressed genes in the PLN are islet-specific (Gcg
)), and may represent tissue-restricted antigens, regulated by AIRE and/or other factors. They were all up-regulated after treatment (all but Ins1
were significantly normalized). These observations point toward a possible defect in the maintenance of peripheral tolerance against islet antigens, which may be partially corrected by DC/IL-4 treatment. We are now further investigating the relevance of the changes affecting the expression of these genes.
We then analyzed all the genes, whose expression was changed in NOD PLN by DC/IL-4 treatment, regardless of whether they were over-, under- or normally expressed relative to NOD.B10 PLN. The number of genes significantly changed by treatment (p<0.01) was very high: about 1/8th of all the microarray features showed up-regulation, and another 1/8th showed down-regulation. It is at this point difficult to draw many conclusions until we can further dissect which cell types are responsible for these changes. This is complicated by the fact that most cell types in lymph nodes express IL-4Rα (data not shown). Our data are in agreement with a previously proposed Th2 deviation as mechanism of action, demonstrated by the significant increase in expression of some Th2 genes (Il4, Gata3, Ccl24), while expression of typical Th1 genes, such as Ifng and Tbx1, was not changed. The effect on IL-4 and IFN-γ expression was confirmed by RT-PCR, again at the level of the whole PLN tissue. This suggests that changes at the level of the responsible cell types may be even more dramatic.
Remarkably, DCs can be detected in the PLN within a few hours after injection (Creusot et al., manuscript in preparation) and transgene expression can be measured up to a week after injection. We have focused our analysis on day 3 after injection, a time at which DCs have cleared from most tissues, except the spleen and the PLN. As shown above, DC/IL-4 induced profound effects in the PLN, and as a consequence, the majority of treated mice were protected long-term. The effects on the spleen remain to be investigated. As seen in the Roadmap study (Kodama et al., manuscript submitted), genes that were significantly over- or under-expressed in 12-wk old NOD (compared to NOD.B10 control) did not overlap very much between spleen and PLN. In addition, we have studied the migration of Luc+ DCs and the therapeutic effect of DC/IL-4 after intraperitoneal (IP) injection. Although DCs homed more efficiently to the PLN after IP injection, their migration to the spleen was very poor compared to IV injection (Creusot et al., manuscript in preparation). We also observed that IP-injected DC/IL-4 significantly protected NOD mice, but not as efficiently as IV-injected DC/IL-4 (data not shown), suggesting that the spleen may also play an important role.
Finally, we have demonstrated that MHC class I and class II deficient DCs, transduced to express IL-4, fail to confer protection. A likely explanation is that the protection is T cell-mediated and that DCs are required to crosspresent antigens and to interact with T cells, while providing them with IL-4 in a paracrine manner. However, we are currently testing the possibility that MHC-deficient DCs may, to some extent, be impaired in their ability to efficiently home to the spleen and PLN.
Overall, our data demonstrate that therapeutic DCs (here, expressing IL-4) can effectively work at a later age than previously reported to prevent or delay the onset of hyperglycemia in NOD mice. Their effect required MHC expression, and in the PLN, was characterized by the normalization of many genes that were under-expressed during the late prediabetic stage of T1D (12-wk old NOD compared to NOD.B10). Th2 deviation may not be the only mechanism by which DC/IL-4 prevent disease; restoration of peripheral tolerance to islet antigen is another possibility that we are now investigating. Moreover, the restricted tissue (in particular PLN) homing specificity of these therapeutic DCs has provided implications for their use in clinical studies. Molecular imaging of monocyte-derived DCs, the human myeloid equivalent of murine bone marrow-derived DCs, will be required to confirm similar homing properties in man. High resolution and sensitive techniques such as positron emission tomography, using the Herpes Simplex Virus 1 thymidine kinase reporter gene and the 9-[4-[18
F]fluoro-3-(hydroxymethyl)butyl]guanine probe [45
], combined with computed tomography can safely be applied to human patients to address the possible use of modified DCs in terms of homing and therapeutic potential. Furthermore, the potential of DC/IL-4 to reverse overt disease (ongoing hyperglycemia) will be explored.