Fetal Pancreas Transplants Are Dependent on Prolactin for Their Development and Prevent Type 1 Diabetes in Syngeneic but Not Allogeneic Mice

C3H/HeNRJ (H-2^k), BALB/c × C3H/HeNRJ F1 (H-2^dxk), BALB/c × C57Bl/6 (B6) F1 (H-2^dxb), DBA/2 (H-2^d), BALB/c (H-2^d), B6, and BALB/c nude mice (pregnant or not) were obtained from Janvier Laboratories (Le Genest St. Isle, France). Ins-HA BALB/c transgenic mice expressing hemagglutinin (HA) under the control of the insulin promoter (15), TCR-HA BALB/c transgenic mice expressing a T-cell receptor specific for the HA₁₁₁ epitope presented by I-E^d (15) and neurogenin3-deficient mice (Ngn3^−/−) (16) (a gift from Dr. Ravassard, CNRS U975, Paris, France) were maintained under specific pathogen-free conditions in accordance with current European legislation. The Regional Ethics Committee approved all protocols.

FPs from E11 to E12 embryos were dissected using a microscope, as previously described (17). Some FPs were cultured overnight at 37°C in RPMI medium supplemented with 10% FCS. Adult pancreatic islets were purified as described previously (18). Recipient mice, anesthetized with a xylazine/ketamine mix, were transplanted with 1–10 FPs or 400 pancreatic islets under the left kidney capsule. Some FPs were grafted within a ring device placed under the kidney capsule as previously described (19).

Diabetes was induced in the ins-HA mouse model as previously described (20).

Mice were implanted subcutaneously with pellets releasing 5 mg of bromocriptine per day for 21 days (Innovative Research of America) to inhibit prolactin production, as previously described (21). Sham control mice received placebo pellets.

The graft was manually homogenized, and proteins were extracted on ice in an acid/ethanol solution (49 parts 95% ethanol, 1 part concentrated hypochlorous acid) and then sonicated. Insulin was quantified by ELISA (Mercordia).

FPs from E11 B6 embryos were cultured for 8 days, as previously described (22). When indicated, human prolactin (1 µg/mL; produced in our laboratory) and BrdU (10 µg/mL; Sigma-Aldrich) were added to the culture. Images were digitized and BrdU/insulin-positive cells were quantified blindly using Image J software (National Institutes of Health), as previously described (17).

Graft-bearing kidneys and FPs were either fixed in 4% or 10% formalin, embedded in paraffin, and cut in 4- to 5-μm sections, or they were frozen in optimal cutting temperature (OCT) compound, stored at −80°C, and cut in 8-μm sections. Paraffin sections were stained with hematoxylin-eosin. Paraffin (Figs. 1 and 3) or frozen sections (Fig. 5) were stained with anti-insulin (Sigma-Aldrich or DakoCytomation), antiglucagon (Immunostar), or anti-BrdU (Amersham Biosciences) antibodies, followed by fluorescent secondary antibodies (Jackson ImmunoResearch Laboratories or Beckman Coulter). Frozen sections were stained with phycoerythrin-, fluorescein isothiocyanate–, or allophycocyanin-labeled antibodies against CD4 (RM4-5), CD8 (53-6.7), CD11b (M1-70.15), CD11c (HL3), Gr-1 (RB6-8C5), and Foxp3 (FJK16s) (BD Biosciences). Nuclei were stained with Hoechst 33342 (Invitrogen). Tissues were observed by fluorescence microscopy (Leica Microsystems). The Metaview software was used for image analysis.

Statistical significance was calculated using a one-tailed Mann–Whitney nonparametric test or an unpaired Student t test.

We first determined the minimal number of FPs required for diabetes control. Because of the toxic impact of hyperglycemia on FP development (14), we set up a model in which we could specifically kill endogenous β-cells at any given time without affecting newly differentiated β-cells in the transplant. This was achieved by transplanting FPs from wild-type (WT) BALB/c mice into transgenic BALB/c mice expressing the HA model antigen specifically in β-cells. Without transplantation, we previously showed that diabetes was induced within 10 days in these ins-HA transgenic mice by injecting HA-specific effector T cells (Teffs) that killed only HA-expressing β-cells (20). In this setting, any β-cell derived from the WT FP graft would be spared. In a first set of experiments, we used FPs that were maintained overnight in culture medium and then transplanted under the kidney capsule within a ring device in order to more easily localize the graft for histological analysis. Two to three weeks later, which is the time needed for FPs to differentiate into insulin-producing cells (2,3), mice were injected with HA-specific Teffs. Nongrafted controls became diabetic within 5–10 days, whereas female recipients transplanted with five to seven E12 FPs, but not with less than five E12 FPs, were fully protected from diabetes induced by Teff injection (Fig. 1A and Supplementary Fig. 1). After removal of the graft-containing kidney 6 to 7 weeks after transplantation, mice became hyperglycemic within 24 h, confirming that euglycemia was maintained solely by insulin produced by the transplant. Typical structures of dense pancreatic islets, rich in insulin- and glucagon-producing cells, were observed within the graft by immunohistology (Fig. 1B). Surprisingly, when we performed the same experiment in male ins-HA recipients, five to six E12 FPs regulated glycemia in only one of nine mice (Fig. 1C). Five weeks after transplantation, the graft was much smaller in male compared with female recipients, with only small numbers of pancreatic islets (Fig. 1D).

We then modified the protocol in order to increase the efficiency of FP development. Moreover, we implemented quantitative measurements of the graft insulin content by ELISA 3 weeks after transplantation of one to five E12 FPs in syngeneic WT BALB/c females. Using fresh FPs and in the absence of a ring, each FP produced an average of 11.3 µg of insulin as compared with only 2.4 µg using the initial protocol. Both the overnight culture and the ring device altered the FP development into insulin-producing cells (Supplementary Fig. 2). Except where noted, we preferentially used this new protocol (fresh FPs and no ring) for the remainder of the study. In this setting, we observed that as few as two E12 FPs were sufficient to regulate T1D in ins-HA females, which were made diabetic by injection of HA-Teffs 3 to 4 weeks earlier (Fig. 1E). As expected, mice became hyperglycemic hours after ablation of the grafted kidney. We confirmed the sex difference using this protocol since up to three FPs were not able to control T1D in ins-HA males (Fig. 1F). Taken together, our results show that as few as two FPs were sufficient to control T1D in female recipients, provided that the right transplantation protocol was used.

To address whether FP development was impaired or just delayed in males, we measured the quantity of insulin produced by the transplant at various times after transplantation of one FP in WT BALB/c mice. In female recipients, the content of insulin per transplant was hardly detectable after 1 week, averaged 3.8 µg after 2 weeks, and reached a plateau of 8–15 µg after 3 weeks of development. Compared with females, insulin production was significantly lower in males after 2 and 3 weeks, but interestingly attained similar levels (∼9 µg) after 6 weeks of development (Fig. 2A). We then performed functional studies in the ins-HA diabetes model. Male ins-HA mice were grafted with two to five E12 FPs and injected with HA-specific Teffs 6 weeks later to take into account the delayed development of FPs in males compared with females. All mice were protected from diabetes induction, and the protection was due to the transplant since these mice became rapidly diabetic following ablation of the grafted kidney (Fig. 2B). Thus, FP development was delayed rather than impaired in male compared with female recipients.

The difference in the kinetics of FP development between males and females may be due to sex hormones since they appear to have an effect on proliferation, survival, and insulin production of β-cells (23). To address this question, we castrated male and female BALB/c mice before transplantation of one FP, eliminating hormones produced by gonads, and quantified insulin produced by the graft 3 weeks later. Interestingly, the difference in the insulin amount between males and females was maintained in castrated mice (Fig. 2C). The absence of a role for sex hormones on graft development was confirmed by the similar amount of insulin produced by the graft when comparing castrated mice and sham surgically operated controls in either males or females (Supplementary Fig. 3). Importantly, we could show that the sex of the donor graft did not impact FP development (Supplementary Fig. 4).

We then studied the role of prolactin because this hormone is more abundant in females than males, and it may improve β-cell development and proliferation (24). To address this question, we treated mice with bromocriptine, an inhibitor of prolactin production (25,26). The higher level of insulin in transplants of female compared with male recipients at 3 weeks was confirmed again in control mice receiving the placebo. In contrast, insulin levels were decreased in females upon bromocriptine treatment, which reached levels observed in treated males. Thus, the sex difference was abolished in mice treated with bromocriptine (Fig. 3A)

To further support a role for prolactin in FP development, we first assessed expression of its receptor in E18 FPs. Expression was clearly detected in fetuses from WT mice but was severely reduced in fetuses from Ngn3^−/− mice (Fig. 3B), which were genetically deficient for endocrine cell differentiation (16). This suggests that prolactin receptor was already expressed in β-cells during late fetal development. Then, we assessed the effect of prolactin in cultures of E11 FPs. The addition of prolactin after 7 days of culture increased the proportion of dividing β-cells (Fig. 3C), suggesting that this hormone increased β-cell proliferation during late fetal development.

Since prolactin levels appeared lower in immunodeficient nude females than WT females (27), we studied FP development 3 weeks after transplantation in these mice. Interestingly, insulin levels were significantly decreased in nude females compared with control females but not in nude males compared with control males. As a result, the sex-related difference was abolished in nude recipients (Fig. 3D). In conclusion, in a syngeneic context, FP development was optimal and more rapid in females compared with males, which seemed to be linked to a higher prolactin level and a functional immune system.

Since T1D patients treated with insulin often have short periods of hyper- or hypoglycemia, it was important to evaluate FP development in a comparable context. Diabetes was first induced in ins-HA mice by HA-Teff injection, followed by insulin treatment at diabetes onset, transplantation of FP from ins-HA mice 5 to 6 days later, and quantification of graft insulin 14 days later (Supplementary Fig. 5A). Graft insulin was hardly detectable in mice that had repeated periods of hypo- (<60 mg/dL) and/or hyperglycemia (>300 mg/dL), whereas it was readily detected in the one mouse that had fewer peaks of moderate hyperglycemia (200–300 mg/dL) (Supplementary Fig. 5B). The lower graft insulin content in this latter mouse compared with values observed in BALB/c recipients (Fig. 2A) may be due to low-grade FP rejection by HA-Teffs. To further assess the impact of islet-specific autoreactive T cells on FP development, we performed an experiment in NOD mice. Two weeks after FP transplantation in syngeneic 12-week-old NOD females, graft insulin contents were similar to the values obtained in the BALB/c model (Supplementary Fig. 5C). Thus, FP development appears to be impaired in mice with poor control of glycemia but not in the presence of islet-reactive T cells.

We then studied the development of E12 FPs in allogeneic female recipients in order to test their immunogenicity. To address this question in the context of different degrees of alloreactivity, we grafted C3H or B6 FPs into BALB/c mice (full major histocompatibility complex [MHC] disparity), BALB/c × C3H F1 or BALB/c × B6 F1 FPs into BALB/c mice (semi-MHC disparity), and DBA/2 FPs into BALB/c mice (minor histocompatibility [mHC] disparity). These experiments were first performed in ins-HA BALB/c females using the initial protocol (overnight cultured FPs and ring device). In this setting, initial experiments showed that five or more syngeneic FPs were able to regulate T1D (Fig. 1A). Surprisingly, we observed that mice transplanted with 5–10 allogeneic FPs became diabetic 5–10 days after Teff injection regardless of the degree of disparity between donor and recipient (fully MHC, semi-MHC, and mHC) and as quickly as nontransplanted controls (Fig. 4A). Histological analyses showed that the grafts were highly infiltrated in all three allogeneic conditions (Fig. 4B) but not in syngeneic control recipients (Fig. 1B). Even in the context of mHC disparity, immune cell infiltration was already detectable at day 7, had increased at day 10, and was massive by day 15. Developing pancreatic islet structures could be observed at days 7 and 10 but were completely destroyed by day 15 (Supplementary Fig. 6). Thus, diabetes could not be prevented by FP transplantation in various allogeneic contexts, probably because of the rapid rejection of the transplanted FPs.

Since the presence of a ring device may increase inflammation and precipitate allograft rejection, we addressed the immunogenicity of E12 FPs using the improved FP transplantation protocol (fresh FP and no ring). We performed a kinetic analysis of insulin produced by the transplant in various allogeneic contexts, as a quantitative indicator of graft rejection. One week after transplantation, FP grafts produced only an average of 0.08–0.1 µg of insulin in the context of mHC or fully MHC disparities versus 0.25 µg in the syngeneic condition (Fig. 5A). The difference in insulin content was higher at week 2 (mean of 3.8 µg in syngeneic vs. 0.14–0.76 µg in the three allogeneic conditions) and even increased at week 3 (mean of 10.6 µg in syngeneic versus 0.03–0.9 µg in the three allogeneic conditions). In the context of mHC, semi-MHC, and fully MHC disparities, the peak of insulin reached 0.9, 0.76, and 0.14 µg, respectively (Fig. 5A). In addition, the peak value was reached at later time points in the mHC condition as compared with the semi-MHC and fully MHC conditions. Thus, as expected, different intensities of graft rejection were observed depending on the allogeneic context, but even in the mHC context, rejection was strong enough to prevent proper and durable development of FPs. Immunohistology analyses confirmed these findings with only scarce and dispersed insulin- and glucagon-producing cells observed in the three allogeneic conditions at week 2 (Fig. 5B and C). At week 3, these cells were no longer detected, whereas they were abundant in the syngeneic condition (Fig. 5B and C). In conclusion, mouse E12 FPs are highly immunogenic and rapidly rejected in various allogeneic contexts, confirming the results obtained in the ins-HA model.

Previous studies have shown that during allograft rejection, the graft is infiltrated mostly by CD4 and CD8 T cells as well as myeloid cells (13,28). After validation of immunostaining on the spleen (Supplementary Fig. 7), we performed immunohistology of FP transplants to detect cells expressing CD4, CD8, Foxp3, CD11c, and Ly6C/G (Gr1) at days 7, 14, and 21 after transplantation. In control mice transplanted with syngeneic grafts, we detected very few infiltrating immune cells (Fig. 6A and B). In contrast, in mice grafted with allogeneic FPs, we observed few T cells at day 7 but a high number of CD4⁺, CD8⁺, and Foxp3⁺ T cells at days 14 and 21. Myeloid cells (CD11c⁺ and Gr1⁺ cells) were present in small to intermediate numbers in allogeneic conditions depending on the time point and individual mouse (Fig. 6A and B). No obvious difference in the type of infiltrating cells could be found between the various allogeneic contexts. As a control, we performed immunohistology of transplanted allogeneic adult pancreatic islets and observed similar T and myeloid cell infiltration (Fig. 6A and B). Thus, allogeneic E12 FPs were massively infiltrated with T and myeloid cells after transplantation.

When considering transplantation of FPs as a treatment for T1D, it is important to evaluate the minimal numbers of FPs required to control hyperglycemia. Previous studies addressing this question were performed in rats in which diabetes was induced by streptozotocin injection before transplantation. Since hyperglycemia was detrimental to the proper development of FPs (14), a finding confirmed in this study, recipient animals concomitantly received insulin administration. However, stable euglycemic condition could not be properly obtained in these animals because it would require repeated glycemia measurements and insulin injections. In this setting, at least six E17 to E18 FPs were required to control diabetes in syngeneic rats (2). Since normalization of glucose level is usually obtained in diabetic patients with only transient peaks of low hyperglycemia (200–300 mg/dL), it was essential to re-evaluate the question of the number of FPs required to control T1D in a context in which FPs develop in euglycemic animals. With this goal in mind, we developed a new mouse model allowing the exclusive destruction of the recipient’s endogenous β-cells without affecting the transplanted FPs. Our data showed that as few as two E12 FPs were sufficient to control T1D in mice. It is likely that this number could be reduced to only one FP by further optimizing the experimental process. For instance, it has been shown that T1D was better controlled when FPs were preincubated with some growth factors before transplantation (7) or by shunting the venous drainage from the transplants to the liver (29). Additionally, it would be more appropriate to propose this transplantation in diabetic patients with well-controlled glycemia.

Using this new model, we found a delay in the development of transplanted FPs in syngeneic males compared with females that was reproducibly observed in various settings (diabetes regulation or insulin quantification). To our knowledge, this sex discrepancy was not reported before despite the fact that many groups have performed syngeneic FP transplantation (14,28–32). This may be explained by the use of same-sex recipients, hyperglycemic recipients that do not present optimal conditions for FP development, or lymphopenic recipients in which we showed that the sex-related difference in graft development was abolished. Importantly, insulin production by transplanted islets also seems higher in women than men, suggesting that this sex-related difference might be specific to the endocrine pancreas. Alternatively, this difference might be also observed for other fetal tissues or tissues derived from embryonic or adult stem cells. This would be interesting to address since these different types of tissues may be transplanted in the future as part of the emerging field of regenerative medicine. Our study suggests that the environment and nature of the recipient have to be properly considered and may affect the function of the transplanted tissue. The normal physiological development of fetal tissues is taking place in pregnant females, which undergo major hormonal modifications such as elevated levels of prolactin (33). One may speculate that the differentiation of adult stem cells or their derivatives, which usually occurs physiologically in nonpregnant animals, is less dependent on hormones that are increased during gestation.

We tested different parameters that could have accounted for the delayed development of FPs in males compared with females. Sex hormones were interesting candidates because they have an effect on proliferation, survival, and insulin production of β-cells (23), and prolactin levels were reduced after ovariectomy (34). However, our data showing that the difference in FP development between females and males was maintained after castration suggested that it was not due to differences in sex hormones. In contrast, prolactin, for which serum levels are twofold higher in females than males in the steady state (34), seems to be involved in the sex difference in FP development. Indeed, we showed that this difference was abolished upon administration of bromocriptine, an inhibitor of prolactin production (25), and that prolactin promoted β-cell proliferation during late FP development. Of note, this observation does not exclude the possibility that this hormone also has an effect on β-cell precursors to favor their differentiation. Our observation is compatible with several previous reports (23–25,35): 1) overexpression of placental lactogen in β-cells, a hormone structurally related to prolactin that can bind the prolactin receptor and increases β-cell replication and insulin production; 2) mice deficient in prolactin receptor have reduced β-cell mass and islet density; 3) high levels of prolactin produced during gestation stimulate the proliferation of β-cells and increase insulin secretion in the maternal pancreas; and 4) β-cells express high levels of the prolactin receptor during late gestation and lactation. Thus, prolactin appears to promote the proliferation of β-cells, not only in the mature pancreas, as previously shown (24), but also during late fetal development, and higher levels of prolactin in females than males may underlie the more rapid development of FPs in females.

Although it is still controversial (36–39), it has been suggested that FPs at their early stage of development (E28 pig FPs or E15 chick FPs) were well-tolerated in xenogeneic rat or monkey recipients. We were thus surprised when we found that E12 mouse FPs were rapidly rejected in allogeneic recipients even in a context of minor histocompatibility disparities. One possible explanation to reconcile our data with these previous studies is that FPs at an early stage of development would be less immunogenic across a xenogeneic than allogeneic barrier. An alternative explanation is that, compared with rat or monkey recipients, mice are more prone to reject FPs. Indeed, xenogeneic early developed FPs (i.e., from E28 pigs or E15 chick FPs) were accepted in rats or monkeys (7–9), whereas they were rejected in mice (31,38). Thus, it is difficult to speculate on the immunogenicity of allogeneic or xenogeneic FPs after transplantation in humans. If performed in T1D patients, additional complications could potentially arise from the possible reactivation of autoreactive T cells that have destroyed the patient’s own β-cells, as has been observed after transplantation of allogeneic pancreatic islets (40). This question has been addressed previously by transplanting histocompatible E12 FPs in NOD mice that spontaneously develop autoimmune diabetes. The transplant developed into functional pancreatic islets when performed in prediabetic mice but was rejected in diabetic mice (32), suggesting that severe autoimmunity may increase the FP rejection.

Depletion of CD4 and/or CD8 T cells prevents rejection of allogeneic adult islets or E17 to E18 FPs, showing their critical role in the rejection (13,41). T cell–mediated rejection was substantiated by histology of the rejected transplant showing massive infiltration with CD4⁺ and CD8⁺ T cells as well as myeloid cells (13,28,41). We observed a similar population of infiltrating cells after transplantation of mouse E12 FPs, suggesting that rejection was also T cell–mediated. Nevertheless, although common mechanisms may operate in rejection of adult and fetal tissues, several reports have shown that a mild immunosuppressive treatment was sufficient to prevent rejection of FPs across xenogeneic or allogeneic barriers (30,31,38,42). It would be important to reanalyze the efficacy of these treatments in male compared with female recipients. In conclusion, whereas only two FPs were sufficient to control T1D in syngeneic condition, allogeneic FP transplants were rapidly rejected, precluding their capacity to control T1D, even in the context of minor histocompatibility disparities. Thus, FP transplantation needs to be further studied in preclinical models before one may envisage its translation to the clinics.

This work was supported by grants from the Juvenile Diabetes Research Foundation (1-2005-1056), the Agence Nationale de la Recherche (ANR-09-GENO-006-01), and the region Ile-de-France. G.F. was supported by the DIM STEM-Pôle Ile-de-France.

No potential conflicts of interest relevant to this article were reported.

G.F. performed research, analyzed data, contributed to discussion, and wrote the manuscript. B.M.C., V.G., and R.S. designed research, contributed to discussion, and edited the manuscript. S.G., A.B., L.R., and F.G. performed research. B.L.S. designed research, analyzed data, contributed to discussion, and wrote the manuscript. B.L.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The authors thank Pedro Carranza, supported by the CODDIM Ile-de-France, Sébastien Giraud (INSERM U945, Paris, France), Benoît Barrou (INSERM U945), Philippe Ravassard (CNRS U975, Paris, France), Yasmine Hazouz (CNRS U975), Bertrand Blondeau (INSERM U872, Paris, France), Marie-Christine Nou (INSERM U598, Paris, France), Natascha Pigat (INSERM U845, Paris, France), Florence Boutillon (INSERM U845), and Elodie Pliquet (INSERM U959, Paris, France) for advice and technical help; Helene Bour-Jordan (ImmunExpression, San Francisco, CA), Gilles Marodon (INSERM U959), Eliane Piaggio (INSERM U959), and Yenkel Grinberg (INSERM U959) for critical reading of the manuscript; and Christelle Enond, François Bodin, Olivier Bregerie, and Bocar Kane, all from the Centre d’Exploration Fonctionnelle, Université Pierre et Marie Curies, for expert care of the mouse colony.