Mice were housed under specific pathogen-free (SPF) conditions. All animal experimentation procedures performed in this study were overseen and approved by the Institutional Ethical Committee for Animal Experimentation of the University of Lleida (CEEA) in accordance with European and U.S.A. regulations on Animal Experimentation. All results involving animal experiments were obtained with female mice. Mice heterozygous for the CDK11-deficiency (HTZ) with the mixed 129SvxC57BL/6 genetic background (22) were backcrossed to the NOD background for 15 generations to identify the Idd susceptibility loci (34). The null mutation of Cdk11 was detected by PCR (22) using the CPrev: CAAGAGAAGCCTGAGCAAATAG and the mp70: GAGATACTCTTTACATGCCAACC primers. Cyclin D3 deficient mice in the NOD background (D3KO) were genotyped by PCR as previously reported (10).

See Supplementary Methods .

Immunohistochemical analysis of pancreatic infiltration was performed on paraffin sections by hematoxylin & eosin counterstaining (Sigma-Aldrich, St. Louis, MO, USA) for the infiltration studies (Leica, Wetzler, Germany). To score the islet infiltration, the following values were assigned: 0 for non-infiltrated islets, 1 for peri-insulitic lesions, 2 for intra-insular insulitis with less than 50% of the islet area infiltrated, and 3 for insulitis with more than 50% of the islet area infiltrated. A minimum of eight sections per mouse at four different levels (>150 islets per genotype) were examined. For each individual mouse, the following formula was applied to calculate the infiltration score, and then, the mean ± SEM of each genotype was plotted:

All these procedures were performed as previously described (34).

Islets were isolated and disaggregated by trypsin digestion, and the resulting cell suspension was incubated with an anti-mouse CD45-phycoerythrin (PE) antibody (Becton Dickinson (BD), CA, USA), subsequently washed and incubated with the anti-PE antibody coupled to magnetic beads, and submitted to AUTOMACS magnetic negative selection according to the manufacturer’s instructions (AUTOMACS; Miltenyi Biotec, Bergisch Gladbach, Germany).

See Supplementary Methods .

The isolated islets were trypsinized, fixed and permeabilized prior to the cellular staining with the anti-mouse Glut-2 (beta-cell marker; R&D Minneapolis, MN, USA), anti-mouse Ki67 (proliferation marker; Dako, Glostrup, Denmark), phycoerythrin-conjugated CD45 (hematopoietic marker; BD), and 7AAD (cell death marker; BD) antibodies. The data were acquired using a FACSCanto II flow cytometer (BD).

Determination of late (TUNEL) apoptotic events was performed using the In Situ Cell Death Detection Kit, Fluorescein (ref, 11684795910, Roche-Sigma). Insulin was detected with anti-insulin antibody (A0564, Dako), and nuclear staining was performed with Hoechst 33342 (B2261, Sigma).

The islet-infiltrating lymphocytes (IILs) were isolated by trypsinization of the previously isolated islets using 500 µl of trypsin/EDTA (Lonza, Verviers, Belgium) for 5 min at 37°C to obtain single cell suspensions.

See Supplementary Methods .

Handpicked islets were cultured in supplemented DMEM in the presence or absence of a cytokine cocktail containing IFN-γ 100U/ml, TNF-α 1000 U/ml, IL1β 100U/ml, afterwards islets were harvested and lysed by hypotonic shock, stained with Propidium Iodide (Sigma) and acquired by flow cytometry to measure nuclei fragmentation (% ΔSubG1) as previously reported (35). Results were calculated as percentage on increment in SubG1 induction by cytokines as follows:

Handpicked islets from 5–7-week-old female mice donors were cultured in supplemented DMEM containing 11mM glucose, either in the presence or absence of 1μM Thapsigargin (Sigma) for the indicated time (either 24 h for ER-stress marker induction or 96 h for subG1 measurements).

Mann-Whitney U-tests were performed to evaluate the differences between pairs of groups. The log-rank test was applied to assess the difference between the diabetes cumulative incidence curves. The one-way ANOVA test was used to assess the differences in TUNEL studies. The two-way ANOVA test was used to analyze the differences in infiltration score studies, and SubG1 studies respectively. All analyses were performed using the GraphPad Prism (v5.0a) statistical package. The threshold for statistical significance was set at 0.05.

At 11 weeks of age, the NOD female pancreatic islets are heavily infiltrated, while those from age-matched NOD/SCID female mice are insulitis-free, because they lack lymphocytes (36, 37). Using the cDNA microarray technology, we observed that the mRNA expression levels of Cdk11 were 2.4-fold lower in the PECs from the 11-week-old pre-diabetic NOD female mice compared to those from age-matched NOD/SCID female mice (see Supplementary Table 1 ). We confirmed the microarray results by performing quantitative real time PCR (qRT-PCR) assays ( Figure 1A ) and found that the PECs from the 11-week-old NOD females exhibited a threefold reduction in the mRNA expression levels of Cdk11 compared to those from age-matched NOD/SCID females (0.223 ± 0.047 vs. 0.674 ± 0.09, respectively). CDK11^p130 protein expression was also reduced in PECs from NOD mice ( Figures 1B, C ). Nevertheless, CDK11^p58 protein levels were higher in NOD PECs compared to those from NOD/SCID, underscoring the role of inflammation in post-transcriptional regulation of CDK11 ( Figures 1B, C ).

We further aimed to determine whether the mRNA expression of Cdk11 was inversely correlated with the severity of the leukocyte infiltration into the islets. Therefore, we performed qRT-PCR to detect Cdk11 in the PECs of the NOD/SCID islets from the 11-week-old female mice that had been adoptively transferred with increasing amounts of total leukocytes from prediabetic NOD female donors 2 weeks before the islet isolation. We observed a significant inverse relationship between Cdk11 expression levels in PECs and the amount of transferred leukocytes ( Figure 1D ).

To address whether the downregulation of Cdk11 mRNA prior to T1D onset altered the NOD beta cell replicative activity, we assessed the beta cell proliferation rates using Ki67 staining of GLUT-2⁺ CD45^- (beta) cells from 11-week-old NOD and NOD/SCID mice and analyzed these rates by flow cytometry ( Figure 1E ). We found no differences in the beta cell proliferation rates between the groups.

To ascertain whether the mRNA downregulation of Cdk11 at the onset of T1D was causally related to beta cell death in vivo, we developed a NOD mouse model that was genetically hemideficient in CDK11 (N-HTZ) ( Figure 2A ), since the homozygous CDK11 deficiency is embryonically lethal (22). We also obtained NOD/SCID mice that were hemideficient in CDK11 (NS-HTZ) and checked protein expression levels in NS-HTZ versus NS-WT islet cells. CDK11^p58 protein levels were reduced while CDK11^p130 reduction was minimal in NS-HTZ compared to NS-WT mice ( Figures 2B, C ). This outcome revealed a differential translational regulation in CDK11 hemideficency favoring CDK11^p130 over CDK11^p58.

We monitored the spontaneous onset of T1D in the N-HTZ and N-WT littermate mice. In parallel we assessed the spontaneous incidence of diabetes in NS-HTZ compared to that in NS-WT littermates, in case an intrinsic defect in beta cell generation was related to CDK11 hemideficiency.

We found that the N-HTZ mice exhibited a significantly delayed kinetics of the disease and lower cumulative incidence of diabetes (70.8% vs. 81.4% respectively) compared to the N-WT littermates ( Figure 2D ). None of the NS-HTZ mice developed spontaneous diabetes (see Supplementary Table 2 ), which evidenced that no major intrinsic defect in beta cell mass and/or function is associated to the CDK11 hemideficiency.

To determine whether the protection against T1D shown by the N-HTZ mice was due to a milder insulitic attack compared to the N-WT mice, we first scored the islet leukocyte infiltration associated with both genotypes at three different time points prior to the onset of diabetes (5, 8, and 10 weeks of age). Dramatic differences were found in the severity of the insulitic attack between both genotypes independently of the age tested, being the infiltration score significantly lower in the N-HTZ genotype compared to the N-WT ( Figure 2E ).

We further assessed the diabetogenicity of the N-HTZ leukocyte repertoire compared to that of N-WT littermates by performing adoptive transfer of diabetes into NOD/SCID recipients. In total, 10 million cells isolated from Pancreatic Lymph Nodes (PLNs) from either 9-week-old N-HTZ or N-WT donors were transferred into NOD/SCID recipient mice ( Figure 2F ). The onset of the adoptively transferred diabetes was monitored, and no difference between both genotypes was observed in their diabetogenic capacity, which was supported by the unaltered representation of Tregs in the PLNs and insulitic infiltrate from N-HTZ females compared to the N-WT genotype ( Figures 2G, H respectively). We also evaluated the frequency of effector and naïve T cells in PLNs and in the insulitic infiltrates; the in vitro proliferative activity and apoptotic rates of PLN cells in front of stimulation by islet autoantigens ( Supplementary Figure 1 ); and no major difference was found between both genotypes. Moreover, in vitro Th1 response was not impaired in PLN cells from N-HTZ mice ( Supplementary Figure 2 ).

Therefore, the target tissue itself, the beta cell compartment, and no an impaired autoimmune repertoire in the N-HTZ mice, should account for the protection against diabetes exerted by the CDK11 hemideficiency. NOD macrophages are less efficient in engulfing apoptotic cells leading to inflammation (38). The release of smaller amounts of beta cell antigens by N-HTZ islets due the lower apoptosis rates, would not sustain islet inflammation as is fostered by the inefficient scavenger capability of N-WT macrophages, leading, hence, to a lower infiltration score in N-HTZ.

We further explored the relationship between the viability of beta cells and the expression levels of CDK11. We determined the number of late apoptotic beta cells using the TUNEL assay in WT and HTZ mice from both 5-week-old NOD and NOD/SCID strains ( Figure 3A ) ( Suplementary Figure 3 ). In the NOD strain, the N-WT beta cells exhibited a dramatic increase (10.8 times) in apoptotic activity compared to those from N-HTZ (173 × 10^-5 ± 69 × 10^-5 vs. 16 × 10^-5 ± 9 × 10^-5, respectively). The differences in the apoptosis rates between the WT and HTZ genotypes were abrogated in the NOD/SCID strain (3 × 10^-5 ± 2 × 10^-5 vs. 95 ± 74 × 10^-5, respectively) ( Figure 3A ). These results reveal that the protection against T1D exerted by the CDK11 hemideficiency in NOD mice is caused by a negative interference with cytokine (inflammation)-induced apoptosis, since, in the NOD/SCID strain, as expected, the apoptosis levels are negligible regardless of CDK11 expression levels, because there is no inflammation.

Altogether, these observations suggested a scenario in which the inflammation-driven cleavage of CDK11^p58 and CDK^p130 into CDK11^p46 and, the induction of CDK11^p58 in a mitotic effort to compensate for the initial beta cell loss induced by inflammation in T1D (30), would enhance beta cell apoptosis. Therefore, the physiological downregulation of CDK11 should have a protective effect against T1D onset.

In order to confirm the hypothesis that CDK11 hemideficiency protects from cytokine-induced apoptosis, we cultured pancreatic islets from either wild type (NS-WT) or CDK11 hemideficient (NS-HTZ) NODSCID female mice in the presence of a proinflammatory cytokine cocktail (TNFα+IL-1β+IFNγ) at either low (5mM), medium (11mM) or high (25mM) glucose concentration, respectively, for 24 h. Subsequently, the induction of DNA fragmentation was assessed by flow cytometry as a measurement of late apoptotic events ( Figures 3B–D ). The induction of apoptosis due to the exposure to pro-inflammatory cytokines was significantly lower in the NS-HTZ group compared to the NS-WT control group, independently of the glucose concentration. Therefore, an impairment in the pro-inflammatory trigger of apoptosis does account for the protection exerted by the CDK11 hemideficiency against beta cell apoptosis. This finding highlighted the role of CDK11 in promoting apoptosis once the proinflammatory milieu pervades the pancreatic islets during the progression of the autoimmune attack.

Since CDK11^p58 has been proven to be essential for sister chromatid cohesion (18, 19), we investigated whether the hemideficiency in the CDK11 resulted in impaired beta cell proliferation too. The percentage of proliferating beta cells (CD45^-Glut-2⁺Ki67⁺) was assessed by flow cytometry ( Figure 3E ), but, surprisingly, no differences were found in the beta cell proliferation rate between the genotypes, which implied that the role that CDK11 plays in beta cell viability is cell-cycle-independent.

In order to ascertain whether CDK11 hemideficiency also inhibits UPR-induced apoptosis in beta cells, we cultured pancreatic islets from either wild type (NS-WT) or CDK11 hemideficient (NS-HTZ) NODSCID female mice in the presence or absence of Thapsigargin, a drug that induces ER-stress. 24 h later, the islets were collected and the RNAs of all experimental groups extracted to quantify the relative expression levels of three key UPR markers: BiP (Binding-immunoglobulin protein, a chaperone) (31); Atf4 (Activating transcription Factor 4) (39); and Chop (C/EBP Homologous Protein) (32). ATF4 leads to CHOP expression, which leads to apoptosis in terminal UPR.

The results obtained showed a significant reduction in the upregulation of both, Chop (132.1 ± 39.61 vs. 411.6 ± 140.4) ( Figure 3F ) and Atf4 (23.79 ± 17.51 vs. 144.2 ± 40.17) ( Figure 3G ), upon ER-stress induction in the NS-HTZ compared to the NS-WT islets, respectively. In NS-HTZ, the induction of both transcription factors by ER-stress is severely impaired. However, ER-stress-induced BiP upregulation, was not significantly different between the NS-HTZ genotype compared to the NS-WT islets (1191 ± 277,4 vs. 2407 ± 647,4, respectively) (see Figure 3H ).

This outcome evidenced that CDK11 could have a role in promoting UPR-induced beta cell death, since CDK11 hemideficiency would protect beta cells from apoptosis by inhibiting UPR-induced CHOP expression. To further assess this hypothesis, we cultured pancreatic islets from both, NS-WT and NS-HTZ in either the presence or absence of Thapsigargin for 96 h (4 days), collected the islets, and measured %subG1 events by flow cytometry ( Figure 3I ). Surprisingly, no differences in the levels of Thasigargin-induced apoptosis were observed between both genotypes. This outcome revealed that CDK11 hemideficiency had not a significant role in inhibiting ER-stress induced apoptosis in beta cells.

Cdk11, similarly to Cyclin D3, is downregulated in beta cells following an autoimmune attack in T1D. However, while the downregulation of Cyclin D3 compromises the beta cell viability, the downregulation of CDK11 is protective against beta cell death. This is an intriguing finding since Cyclin D3 is a regulatory partner of CDK11^p58 (40–44), which, in turn, is required for cell cycle progression and is involved in apoptosis. Therefore, in the absence of Cyclin D3, CDK11^p58-promoted apoptosis should be impaired, and NOD mice that are deficient in Cyclin D3 should exhibit milder T1D due to its role in activating CDK11^p58. However, we previously showed (10) that the Cyclin D3 deficiency dramatically exacerbates T1D in NOD mice, while its overexpression in beta cells protects them against diabetes; furthermore, neither of these conditions, i.e., deficiency or overexpression, has any effect on beta cell replication (10), which strongly suggests that the role of Cyclin D3 in beta cells is independent of CDK11^p58.

Thus, we speculated about the relationship between CDK11 and Cyclin D3 regarding beta cell viability and whether a hierarchy exists among the respective signaling pathways that are engaged by these two molecules in beta cells. Our working hypothesis is that CDK11 may work as a decoy receptor for Cyclin D3, quenching Cyclin D3 anti-apoptotic activity in beta cells. In the N-HTZ genotype, the levels of free, unbound Cyclin D3 are higher than in N-WT, and hence, beta cell survival is promoted. Therefore, if this was the case, we would not expect any amelioration of the severity of the disease in Cyclin D3 deficient mice upon CDK11 hemideficiency.

We answered this question by obtaining NOD mice deficient in Cyclin D3 that were either hemideficent (K11HTZ-D3KO) or not (K11WT-D3KO) for CDK11.

We monitored the spontaneous incidence of diabetes in these two experimental groups and observed that regardless of the CDK11 hemideficiency, the Cyclin D3 deficiency had a dominant effect on the onset of diabetes and exacerbated the disease compared to normal NOD mice (K11WT-D3WT) since the CDK11 hemideficiency did not alleviate the T1D in the Cyclin D3 deficient mice ( Figure 4 ).

Altogether, this observation suggested that at least one of the signaling pathways responsible for the survival of beta cells in which Cyclin D3 is involved is independent of the CDK11 signaling pathway.

CDK11 is a pleiotropic kinase that is involved in gene expression, cell division and cell death. The function of CDK11 and its regulation in these processes remain unclear in beta cells. Beta cells have been shown to increase their proliferation rates prior to the onset of T1D (30), therefore, it is plausible that while the overall mRNA expression of Cdk11 is downregulated due to insulitis, the translation of CDK11^p58 is favored at the expense of CDK11^p130 in an attempt to subdue the loss of beta cells. However, this defense mechanism may be detrimental to beta cell survival because CDK11^p58 also signals for mitosis-coupled apoptosis (24) and becomes a substrate for the amplification of caspase-3, which would result in cytokine-induced beta cell death in T1D. Therefore, downregulation of Cdk11mRNA would prevent massive CDK11^p58 translation under inflammatory circumstances. In the spinal cord and astrocytes, the CDK11^p58 protein is also upregulated during inflammation. Moreover, in Schwann cells, a pro-inflammatory stimulus, such as LPS, could upregulate CDK11^p58, and the partnership CDK11^p58/Cyclin D3 inhibits of the cell cycle and promotes apoptosis (43). In this sense we found that the increased CDK11^p58 expression levels in NOD PECs prior to the onset of diabetes did not alter beta cell replication. Moreover, genetic CDK11 hemideficiency, (affecting mostly to the CDK11^p58 protein compartment, while CDK11^p130 remains roughly untouched in endocrine cells), protected beta cells from apoptosis in the NOD proinflammatory niche. Furthermore, when exposed to a powerful proinflammatory mileu NS-HTZ islets exhibited significantly lower apoptotic rates compared to those from the NS-WT control group, confirming the involvement of CDK11^p58 in cytokine-induced apoptosis. Therefore, the scenario in which CDK11 would be cleaved upon inflammation promoting apoptosis with a dominant effect over the other CDK11 roles, is evidenced as a mechanism responsible for protection against beta cell apoptosis in the CDK11 hemideficient NOD mice.

ER-stress and inflammation are responsible for most of the beta cell failure in T1D. Both, inflammation and ER-stress crosstalk and feed-back each other positively, resulting in exacerbation of beta cell death (33). It has been reported that interfering the UPR signaling reverses T1D in the NOD mouse model (45). Both, CDK11^p58 (24) and, CHOP (32), a molecule causally related to ER-stress-induced apoptosis, have been reported to promote apoptosis by interfering with the expression of Bcl-2, a key anti-apoptotic molecule in ER-stress (46). However, we have not observed an impairment in the induction of apoptosis by ER-stress associated to the CDK11 hemideficiency, which suggests that other (not dependent on CHOP) apoptotic cascades in terminal UPR, such as those downstream IRE1 (47), remain unaltered, and CDK11 would be mainly involved in cytokine-induced apoptosis.

The expression of the CDK11 promoter is hindered by inflammation, probably through Ets-1 transcription factor (48–51), resulting in a protection against apoptosis in beta cells. In the other cell types, cyclin D3 interacts with CDK11^p58 to promote cell division and/or apoptosis. In adult mouse beta cells, we did not observe changes in replication due to either Cyclin D3 or the CDK11 hemideficiency, which suggests that the protection against diabetes is not related to proliferation. Moreover, a priori, one would expect that the deficiency in Cyclin D3, which is a putative activator of CDK11^p58, should promote a phenotype that is similar to that observed in the CDK11 hemideficient mice, and the deficiency of both should add up to the same direction. However, this is not the case since both molecules have opposite effects on the onset of T1D. This observation shows that the downregulation of CDK11 exerts its protection against T1D in a Cyclin D3-dependent fashion. This outcome would be compatible with a scenario in which CDK11^p58 binds Cyclin D3 to prevent its signaling to promote beta cell survival, and, to activate CDK11^p58 pro-apoptotic activity. Therefore, the anti-apoptotic role of Cyclin D3 should be more prominent in non-cycling cells, a scenario when CDK11^p58 is not expressed.

The diabetogenicity of the autoimmune repertoire was not affected in N-HTZ mice. Moreover, the representation of the FoxP3⁺CD25⁺CD4⁺ Treg cell subset in either the regional lymph nodes or the islet infiltrate in not altered in CDK11 hemideficient mice. Therefore, the protection observed against T1D in N-HTZ cannot be explained by an altered diabetogenic repertoire in this genotype.