Introduction

Insulin is believed to be one of the main antigens targeted in the autoimmune process that destroys the beta cells in type 1 diabetes [1]. In recent years, it has been clearly demonstrated that the expression of self-antigens, such as insulin, in the medullary thymic epithelial cells (mTECs) is an essential component for central tolerance [1–4]. Thymocytes entering the thymus after their development in the bone marrow require signals to survive and exit to the periphery to join the properly formed circulating T lymphocytes [5, 6]. These signals are both positive and negative, the former requiring functional T cell receptors and the latter triggered by recognition of self-antigens [5, 6]. Therefore, it is widely thought that the ‘promiscuous’ gene expression of self-antigens in the thymus serves the purpose of assessing the responses of newly formed thymocytes to self-antigens [3]. This process is necessary in order to purge the self-reactive cells from the system and avoid autoimmunity.

The expression of self-antigens has been localised to mTECs [2–4] and is an essential player in the deletion of self-reactive T cells as well as the generation of T regulatory cells [7–9]. Although the expression of self-antigens in the thymic epithelial cells is well documented, not much is known about how their levels are regulated. The level of this thymic expression of tissue-specific autoantigens may be crucial to immune tolerance, at least in the case of insulin: the second strongest genetic association with type 1 diabetes (after HLA) is with a polymorphism upstream of the insulin gene, whose type 1 diabetes-predisposing alleles are associated with substantially lower thymic insulin expression [10–12]. This concept is supported by mouse models, which demonstrate a strong relationship between a tissue representation by ‘promiscuous’ gene expression in the thymus and the development of tissue-specific autoimmune response [13–18]. Most work in the past few years on the regulation of this atypical expression process has been on the autoimmune regulator (autoimmune polyendocrinopathy candidiasis ectodermal dystrophy) (AIRE) transcription factor, which has a major role in the expression of most self-antigens in the thymus. Homozygous loss of AIRE function in the thymus causes the debilitating autoimmune polyendocrinopathy–candidosis–ectodermal dystrophy (APECED) syndrome in humans [19], characterised by multiple-organ autoimmune responses involving mostly endocrine glands and including high risk of type 1 diabetes [20]. In mice, targeted Aire disruption leads to a marked decrease in self-antigen expression, almost eliminating the ‘promiscuous’ gene expression pattern [21]. These mice, like APECED patients, have multiple autoimmune responses directed against various organs [22]. A considerable amount of work has been done on this transcription factor, demonstrating its essential role for this process and its implications for autoimmune disease [19, 21–23]. However, it is not known whether gene-specific regulatory mechanisms of this thymic ‘promiscuous’ gene expression exist. Most studies on medullary thymic self-antigen expression and presentation do not focus on one particular self-antigen but, instead, on the entire pool of expressed genes [24]. This study approach addresses only common regulatory mechanisms, and cannot uncover particular gene-specific mechanisms.

The problem is that tissue-specific antigens are expressed in only a small fraction of mTECs (1–3% for insulin) that have not, until now, been used in isolation, to study regulatory mechanisms specific to the gene of interest [21]. Single-cell PCR has been used [25, 26], but its results are qualitative rather than quantitative; for physiological regulation studies a stable cultured cell line would be ideal. To this end, an in vitro model has been established in our laboratory based on an immortalised cell line of insulin-expressing mTECs; this model allows the specific study of insulin expression regulation in the thymus [27]. Here we report a more detailed investigation into the specific regulatory mechanism of Ins2 gene expression in these cells. The first question we addressed was whether glucose, the major known pancreatic stimulus for Ins2 gene expression, also regulates insulin in the thymus. Then, we asked whether Ins2 expression in mTECs is regulated by cytokines and whether such alterations may be mediated through effects on Aire expression. We chose to start with IFN-γ, a cytokine known to be produced in thymocytes and to have effects on TECs [28]. In addition, we explored possible effects on Ins2 expression by co-culturing mTECs with isolated primary thymocytes.

Methods

In vivo glucose stimulation

Female C56 mice (5–6 weeks of age) were purchased from Charles River Laboratories (Wilmington, MA, USA) and used for all in vivo stimulation experiments. Dextrose (50%, wt/vol.) (Hospira, Lake Forest, IL, USA) was used for intraperitoneal injections at 3 g/kg body weight. Mice were injected intraperitoneally and killed 1 h later. All mice pairs were killed on the same day. The thymus and the pancreas tissue were extracted from each mouse, corresponding to injected (stimulated) and non-injected (non-stimulated) mice. RNA from tissues was extracted as described in the protocol for the High Pure Tissue Kit (Roche, Mannheim, Germany). RNA was quantified and reverse transcribed using Superscript II as described in the manufacturer’s protocol (Invitrogen, Carlsbad, CA, USA).

In vitro glucose stimulation

INS(+) mTEC clones were stimulated with high glucose at 17 mmol/l glucose for 1 h. mTECs incubated with 17 mmol/l glucose media were incubated in T25 flasks for 1 h and control flasks remained in regular mTEC media (5.5 mmol/l glucose). The cells were then collected for RNA extraction. RNA was extracted using the RNeasy Plus Mini Kit (Qiagen, Mississauga, ON, Canada). RNA was reverse transcribed using Superscript II Enzyme as described in the manufacturer’s protocol (Invitrogen). cDNA was then used in real-time PCR for quantification.

Real-time PCR

Ins2 was quantified in the pancreatic tissue by SYBR Green Real-Time PCR (Roche) using the primers and the conditions described by Evans-Molina et al. [29]. The primers used spanned the intron 2 and the exon 3 ensuring the quantification of newly transcribed mRNA. Because of the high abundance and stability of insulin mRNA in beta cells, short-term changes in transcription are best examined at the level of unspliced heteronuclear RNA.

All other quantification of mRNA samples was performed with real-time PCR, using Taqman Gene Expression Assays (Applied Biosystems, Foster City, CA, USA). Probe and primer sets Mm00731595_gH, Mm00477462_m1 and Mm01250092_g1 were used for the quantification of Ins2, Aire and Irf8, respectively. Because Ins2 is the main source of thymic insulin in both whole thymus [10] and our mTECs [27] with Ins1 barely detectable, we confined experiments to the measurement of Ins2. All samples were normalised to endogenous 18S RNA level using the eukaryotic primer and probe set 4333760F. All reactions were performed with TaqMan Universal PCR Mix (Applied Biosystems) as described in the manufacturer’s protocol, and in the Mx4000 apparatus (Strategene, La Jolla, CA, USA). For the cytokine stimulations, as well as the co-culture experiment, the quantification of insulin and Aire gene expression was performed on the same cDNA, with the exception of the addition of a few replicates, as needed for statistical power.

Cytokine stimulation assay

In vitro work was performed on INS(−) (non insulin-expressing) and INS(+) (insulin-expressing) mTEC cell lines, as described by Palumbo et al. [27]. Recombinant mouse IFN-γ (Roche) was used at a final concentration of 100 U/ml in both INS(–) and INS(+) mTEC clones. In both stimulations, cells were incubated overnight and RNA was extracted and reverse transcribed as described above.

Co-culture experiment

INS(−) and INS(+) mTECs were co-cultured overnight with primary thymocytes in a ratio of about 1:1, as well as with media conditioned by primary thymocytes. Ins2 heterozygotic knockout animals of either sex and between the ages of 8 and 12 weeks were used as the source of primary thymocytes. The thymus was extracted from animals and diced into smaller pieces in PBS. The pieces were stirred for 10 min on a magnetic stir plate in PBS and the supernatant fraction, containing the thymocytes, was collected. The supernatant fraction was centrifuged at 700×g for 10 min and the resulting supernatant fraction decanted. The cell pellet was resuspended in 10 ml fresh DMEM media (as used for mTEC clones [27]) and counted. Primary thymocytes (5 × 106) were plated into T25 flasks (BD Biosciences, Franklin Lakes, NJ, USA) already containing 5 × 106 mTEC cells. This represented a 1:1 ratio in 5 ml of fresh media per flask of mTEC clone. The two cell types were co-cultured overnight and all thymocytes were washed off with PBS three times before harvesting the mTEC clones. In addition, 5 × 106 primary thymocytes (per flask experiment) were placed alone into a T25 flask overnight and spun down at 700 g the following day. The supernatant fraction (thymocyte media—no cells included) was applied over another set of mTEC flasks, in both INS(−) and INS(+) clones, and incubated overnight. mTECs were harvested and RNA was extracted and reverse transcribed as described above.

Statistical analysis

Individual data points represent distinct incubation experiments and are each the average of duplicate or triplicate real-time assays. The paired Student’s t test was used to compare treated cells with untreated controls on the same day. As the difference between treated and untreated cells was not normally distributed, the test was performed after log10 transformation, which assured normality in all cases. The same statistical tests and log10 transformations were used in the in vivo experiments.

Results

In vivo glucose stimulation

It was very challenging to quantify insulin transcript increases with glucose stimulation accurately due to high levels of stable insulin mRNA already present in the pancreatic beta cells. Therefore, it was important to obtain a quantification specific to the newly transcribed insulin mRNA. Using modified conditions and primers described by Evans-Molina et al. [29], this quantification in the pancreatic tissue was successful. The quantification was not performed in the thymic tissue, as newly transcribed insulin mRNA does not accumulate in this tissue. As previously reported in the literature [30], the newly synthesised pancreatic Ins2 mRNA increased significantly 1 h after the glucose load (Fig. 1a). Insulin mRNA levels were 0.50 unit/unit 18S after stimulation, an increase from 0.008 unit/unit 18S at resting levels before stimulation. Conversely, in the thymus, no significant changes were seen upon stimulation with glucose. The resting levels of Ins2 mRNA were 5.52 fg/pg 18S; the levels were at 4.98 fg/pg 18S 1 h after glucose injection (not significantly different) (Fig. 1b). The in vitro model INS(+) mTECs were also stimulated with a high concentration of glucose at 17 mmol/l, for 1 h, demonstrating no significant changes between unstimulated and stimulated samples (Fig. 1c).

Fig. 1
figure 1

In vivo glucose stimulation. Ins2 mRNA was quantified in pancreatic and thymic tissues 1 h following intraperitoneal glucose injection. The newly transcribed, unspliced mRNA was measured by SYBR Green Real-Time PCR in the pancreatic tissue (a) and was upregulated (n = 4), while no significant changes were seen in the thymus (n = 5), measured by Taqman Real-Time PCR (b). Insulin transcripts were not affected in the INS(+) mTEC cell line (c) with the high glucose concentration

Full size image

In vitro cytokine stimulation

In view of the fact that the insulin gene expression in the thymus is not regulated by glucose, as it is in the pancreas, and has an assumed immune function, it was important to test immune cytokine action on the expression levels of Ins2 mRNA. The in vitro model of insulin-expressing mTEC clone was used to quantify the effect of cytokine stimulations on these clones on Ins2 mRNA. IFN-γ has been implicated in the negative selection process occurring in the thymus for the maintenance of self-tolerance [31, 32]. For this reason, this cytokine was used to stimulate the mTECs and Ins2 quantification was used to assess its effects. The use of the cytokine did not affect the morphology, recovery or proliferation characteristics of the cell lines.

Overnight incubation with 100 U/ml of IFN-γ caused a significant decrease in the Ins2 mRNA content (Fig. 2a), with a mean difference of 0.67 fg/pg 18S. As was previously established in recent literature [21, 33, 34], the AIRE transcription factor has a major role in the expression of the Ins2 gene in the thymus. Therefore, it was important to assess whether this decrease in the Ins2 mRNA was due to a decrease in the expression of Aire, which was also quantified by real-time PCR. Aire mRNA levels were increased in the INS(−) and INS(+) clones, increasing by 0.0025 fg/pg 18S and 0.24 fg/pg 18S, respectively (Fig. 2b).

Fig. 2
figure 2

In vitro IFN-γ cytokine stimulation. INS(−) and INS(+) clones were incubated with 100 U/ml of IFN-γ overnight, and insulin (a) (p = 0.015) and Aire (b) (p = 0.032) mRNA transcripts were quantified and compared with control incubations. Insulin transcript decreased significantly (n = 8), while Aire transcript increased in both INS(−) and INS(+) clones (n = 6 and n = 10, respectively)

Full size image

Moreover, we quantified another transcription factor, interferon regulatory factor 8 (IRF8), which is implicated in interferon signalling and has been shown to act in conjunction with Aire for the expression of some tissue-specific antigens in the thymus [35]. As expected, the Irf8 mRNA levels were significantly elevated in the samples treated with IFN-γ (Fig. 3a). The levels rose by a mean difference of 6.78 fg/pg 18S. Similar changes were also seen in the negative clone (Fig. 3b).

Fig. 3
figure 3

Irf8 quantification in IFN-γ cytokine stimulation. Transcription factor Irf8 mRNA was quantified after IFN-γ stimulation. mRNA levels increased in both the INS(−) clone (a) (n = 3; p < 0.01) and the INS(+) clone (b) (n = 3; p < 0.01)

Full size image

Thymocyte and mTEC co-culture

It is possible that the induction of central self-tolerance involves close contact of the medullary thymic epithelial cells with maturing thymocytes. To see whether insulin expression in mTECs is regulated as a result of a two-way interaction with thymocytes, we co-cultured insulin-expressing mTECs with primary thymocytes isolated from Ins2 knockout mice corresponding to the genotype of the mTEC clones. The direct co-culturing of these two cell types induced a significant increase in the insulin-expressing clone (Fig. 4a), raising the mRNA levels by a mean difference of 0.44fg/pg 18S. Aire expression was also quantified in the co-cultured mTEC samples to assess whether a correlation could be established for the upregulation of insulin mRNA. The co-culturing did in fact upregulate Aire mRNA as well in both INS(−) and INS(+) clones (Fig. 4b) by a mean difference of 0.045fg/pg 18S and 0.015fg/pg 18S, respectively [31].

Fig. 4
figure 4

Direct co-culture of thymocytes and mTECs. Isolated thymocytes and INS(−) and Ins (+) mTECs were co-culture overnight at a 1:1 ratio. Insulin (a) (p = 0.0105) and Aire (b) transcripts were quantified. Both insulin (n = 11) and Aire expression increased in both INS(−) (p < 0.01) and INS(+) (p < 0.01) clones (n = 9 and n = 11, respectively)

Full size image

The effect of Ins2 gene expression upregulation by co-culturing thymocytes with mTECs could be a possible outcome of direct cell–cell contact or secreted factors by thymocytes acting on mTEC cell receptors. Therefore, the action of thymocyte-conditioned media was assayed by adding it to insulin-expressing mTECs. No significant effect on mRNA levels was observed for either insulin or Aire in either mTEC clone (data not shown).

Discussion

The insulin gene is tissue-specific and thought to be expressed only in the pancreas. In recent years, however, it has been shown that it is expressed in the thymus, along with many other tissue-restricted antigens [3]. AIRE is necessary for the expression of most of these antigens, but the possibility of additional quantitative regulation by factors specific for each antigen needs to be explored. This question is especially important in the case of thymic insulin expression, whose importance in antigen-specific immune tolerance has been widely demonstrated [10, 36–38].

In the pancreas it is well established that metabolic substrates, principally glucose, have a key role in insulin expression regulation [39]. Considering the distinctive function of insulin in the thymus, the regulatory mechanisms controlling this ‘promiscuous’ gene expression are likely to differ from those of the pancreas. Our group has presented indirect evidence for such differences in the past. First, the transcriptional effect of the diabetes-associated polymorphism is seen mainly in the thymus, being minimal and in the opposite direction in pancreas [11, 40]. Second, in knockout mice for Ins1 and/or Ins2, carrying fewer than the four copies of the insulin gene, insulin expression in the pancreas remains normal through upregulation of the remaining copies [10], presumably via the known regulation of insulin expression in the pancreas by metabolic feedback, glucose in particular [10, 39]. This compensation was not observed in thymus, with transcript levels decreasing linearly with decreasing copy numbers [10]. Finally, genes known to be crucial in regulation of pancreatic insulin were not expressed in isolated insulin-positive mTEC clones [27]. The insulin gene, therefore, is expected to be differentially regulated in the thymus compared with the pancreas. In this paper we present a direct demonstration that glucose in vivo and in vitro had no significant effects on thymic Ins2 expression, in contrast with its known effect in pancreas. This is consistent with the different purpose of expression in the thymus (self-antigen presentation) compared with the pancreas (metabolic regulation).

In order to begin to elucidate the process(es) regulating Ins2 expression in the mTECs, immune cytokine stimulation was used on the in vitro cell line model. Aire is known to be necessary for insulin expression. However, Aire is a very general master switch and, if the need arises to regulate specific antigens, other mechanisms may be required. Cytokines, produced by various cell types, such as thymocytes, dendritic or cortical epithelial cells, would be prime candidates for such a role. In addition, cytokines may have very different roles in the thymus compared with their actions in the periphery. This expectation is consistent with the negative effect on mTEC Ins2 expression by IFN-γ which is produced in thymocytes, affects TECs [28, 41] and may be important in negative selection [31, 42]. This decrease was not accompanied by a decrease in the AIRE transcript, which, if anything, was upregulated. Therefore, the mechanism whereby IFN-γ regulates Ins2 expression in the thymus appears to act through a pathway that does not involve Aire alone.

Further evidence that specific processes act in concert with Aire to modulate individual thymic autoantigens comes from our observation that Irf8 mRNA was dramatically increased by IFN-γ stimulation, as previously reported [35]. IRF8 upregulates the transcription of other self-antigens in the thymus, such as CHRNA1, encoding the myasthenia gravis auto-antigen. [35] Although IRF8 levels increased with IFN-γ stimulation, levels of Ins2 mRNA actually decreased.

Taking into account the importance of physical contact of mTECs with thymocytes for the process of negative selection, it was of interest to see that this interaction results in an overall increase in Ins2 transcript levels. Interestingly, we could not reproduce this effect with thymocyte-conditioned media alone. This suggests that the Ins2 increase seen in the mTECs requires the direct interaction between the two cell types and this effect may be mediated by Aire, whose expression increased coordinately with Ins2. Lymphotoxin beta receptor (TNFR superfamily, member 3) (LTβR), known to upregulate both Aire and Ins2 [27, 43] may play a role here, in conjunction with interactions that may involve additional co-receptors between mTECs and thymocytes.

Whether these observations in mouse thymus cells also apply to humans remains to be confirmed. It is worth noting that the variable-number tandem repeats that flank the human insulin promoter and have allelic effects on its thymic expression are absent in the mouse. In addition, we do not know to what extent our mTEC lines represent the in vivo situation, especially given the cell-contact effect that we demonstrated. Further work is needed to answer these questions that may be crucial in the pathogenesis of such an important autoimmune disease as type 1 diabetes.