Loss of TSC complex enhances gluconeogenesis via upregulation of Dlk1-Dio3 locus miRNAs

Significance

Organisms maintain a constant level of blood glucose. mTOR, an atypical serine/threonine kinase, plays a major role in glucose metabolism. This is evident in mouse models where disruption of mTOR signaling can cause a phenotype similar to a prediabetic state. However, how mTOR signaling affects blood glucose levels is incompletely understood. Here we show that sustained activation of mTORC1 enhances expression of miRNAs of the Dlk1-Dio3 cluster to induce gluconeogenesis and increase blood glucose levels. Enhanced miRNA expression correlates to a reduction in levels of DNA methylation at the Dlk1-Dio3 locus. Our findings reveal a connection between miRNAs, mTOR signaling, and glucose metabolism, and may provide a new strategy in the treatment of metabolic disease.

Abstract

Loss of the tumor suppressor tuberous sclerosis complex 1 (Tsc1) in the liver promotes gluconeogenesis and glucose intolerance. We asked whether this could be attributed to aberrant expression of small RNAs. We performed small-RNA sequencing on liver of Tsc1-knockout mice, and found that miRNAs of the delta-like homolog 1 (Dlk1)–deiodinase iodothyronine type III (Dio3) locus are up-regulated in an mTORC1-dependent manner. Sustained mTORC1 signaling during development prevented CpG methylation and silencing of the Dlk1-Dio3 locus, thereby increasing miRNA transcription. Deletion of miRNAs encoded by the Dlk1-Dio3 locus reduced gluconeogenesis, glucose intolerance, and fasting blood glucose levels. Thus, miRNAs contribute to the metabolic effects observed upon loss of TSC1 and hyperactivation of mTORC1 in the liver. Furthermore, we show that miRNA is a downstream effector of hyperactive mTORC1 signaling.

Mammalian target of rapamycin (mTOR) signaling integrates inputs from growth factors, nutrients, and intracellular cues to regulate cell growth and metabolism (1, 2). mTOR forms two structurally and functionally distinct complexes, mTOR complex 1 (mTORC1) and mTORC2 (3, 4). Dysregulation of either complex can lead to metabolic disease and cancer (5, 6). A major negative regulator of mTORC1 is the tumor suppressor TSC complex that is composed of tuberous sclerosis 1 (TSC1), TSC2, and TBC1 domain family member 7 (TBC1D7) (7). Disruption of the TSC complex leads to increased mTORC1 signaling and disease (8, 9). There are a number of mouse models with aberrant mTOR signaling, virtually all of which exhibit dysregulation of glucose metabolism (10). Models in which mTOR signaling is up-regulated develop tumors (10). For example, up-regulation of mTORC1 signaling in mouse liver, achieved by deleting Tsc1, leads to impaired glucose homeostasis within weeks and to liver cancer within ∼12 mo (11–13).

The known effectors downstream of the mTOR complexes are proteins. Little is known about noncoding RNAs (ncRNAs), in particular microRNAs (miRNAs), that may mediate downstream effects of mTOR. miRNAs are a class of short noncoding RNAs involved in many aspects of cell growth and metabolism, and, like mTOR, affect glucose metabolism and tumorigenesis (14). For example, dysregulation of a single miRNA, miR-122, in the liver affects liver metabolism and causes tumorigenesis (15, 16). miR-122 and other miRNAs are involved in liver development and homeostasis and also in liver disease (17). Any given miRNA represses many targets, and, as such, alterations in expression of one or a small number of miRNAs can impact many downstream processes (18). The coordinated control of many targets by a single miRNA is a means to mediate complex processes such as cellular metabolism.

miRNAs are encoded throughout the genome, including at imprinted loci from which expression is regulated by CpG methylation and heterochromatin maintenance in an allele-specific manner (19). The largest imprinted miRNA cluster is part of the Dlk1-Dio3 locus, found in mouse chromosome 12. This is a nearly 1-Mb imprinted locus that carries three paternally expressed protein-coding genes (Dlk1, retrotransposon-like protein 1 [Rtl1], and Dio3) and several maternally expressed noncoding genes. The latter include long noncoding RNA (lncRNA) genes, such as maternally expressed 3 (Meg3), Anti-Rtl1, RNA imprinted and accumulated in nucleus (Rian), and miRNA containing gene (Mirg), small nucleolar RNA (snoRNA) genes, Piwi-interacting RNA (piRNA) genes, and more than 50 miRNA genes (20, 21). Like all imprinted loci, expression at the Dlk1-Dio3 locus is controlled by CpG methylation of so-called imprinting control regions (ICRs). ICRs determine whether the paternal or maternal allele is expressed (21). The Dlk1-Dio3 locus miRNAs, expressed from the maternal chromosome, are studied mainly in brain and placenta and are reported to be involved in metabolic disease and tumorigenesis (20, 22, 23).

It is well documented that miRNAs affect proteins involved in mTOR signaling (24, 25). However, only a few reports describe mTOR-mediated regulation of miRNA expression (26–28). Moreover, to our knowledge, there is no evidence that mTOR signaling can affect specific miRNAs to control cellular metabolism. To determine whether miRNAs play a role in the metabolic alterations seen upon dysregulation of mTOR signaling, we examined miRNA expression in liver-specific Tsc1-knockout mice, hereafter referred as L-Tsc1KO mice. We report that, upon loss of Tsc1, aberrantly constitutive mTORC1 signaling up-regulates expression of miRNAs of the imprinted Dlk1-Dio3 locus to enhance gluconeogenesis. These findings, show that mTORC1 engages miRNAs to regulate metabolism.

Results

Noncoding RNAs of the Dlk1-Dio3 Locus Are Up-Regulated in L-Tsc1KO Mouse Liver.

We utilized the L-Tsc1KO mouse model to investigate how miRNAs may contribute to the metabolic phenotypes observed upon constitutively high mTORC1 activity (SI Appendix, Fig. S1A). We performed small-RNA sequencing on livers of L-Tsc1KO animals and littermate controls at 14 wk of age, when L-Tsc1KO mice display disrupted glucose metabolism but no sign of tumorigenesis. We found 31 and 6 miRNAs up-regulated and down-regulated, respectively, in L-Tsc1KO mice compared to controls (Fig. 1A and SI Appendix, Table S1). qPCR confirmed the changes in expression in all cases examined (Fig. 1B). Twenty-six of the 31 up-regulated miRNAs are encoded by the Dlk1-Dio3 locus (SI Appendix, Table S1, miRNAs in bold).

Fig. 1.

Dlk1-Dio3 locus noncoding genes are up-regulated in L-Tsc1KO mouse livers. (A) miRNA landscape in L-Tsc1KO mice shown as log2 fold change over control levels. Black dots represent miRNAs that are statistically different (adj.P < 0.05) between L-Tsc1KO and control mouse livers. (B) qPCR validation of a few significantly changing miRNAs. U6 snoRNA is used as a control. (C) Levels of three lncRNAs, Meg3, Rian, and Mirg, as well as the protein-coding genes of the Dlk1-Dio3 locus measured by qPCR shown as fold change to control. (D) Levels of other noncoding genes found at the Dlk1-Dio3 locus, ICR-ncRNA (a noncoding transcript associated with increased transcription), snoRNA-MBII-343, and two pre-miRNAs (pre-miR-127 and pre-miR-541) shown as fold change over control. *P < 0.05; **P < 0.01.

DNA methylation at three distinct CpG regions, DLK1 differentially methylated region (DLK1-DMR), intergenic differentially methylated region (IG-DMR), and MEG differentially methylated region (MEG-DMR), controls expression of genes at the Dlk1-Dio3 locus (29–31) (SI Appendix, Fig. S1B). The noncoding genes of the locus are expressed, from the maternal chromosome, as one transcript (21, 32, 33). Consistent with a single transcript, the Meg3, Rian, and Mirg long noncoding RNAs were up-regulated in L-Tsc1KO mice (Fig. 1C). Four other Dlk1-Dio3 locus transcripts expressed from the maternal chromosome, the small nucleolar RNA MBII-343, the noncoding transcript ICR-ncRNA, and two pre-miRNAs, were also up-regulated (34) (Fig. 1D). In contrast, expression of the Dlk1 and Dio3 protein-coding genes did not change (Fig. 1C and SI Appendix, Fig. S1C). Thus, transcription of noncoding genes from the Dlk1-Dio3 locus is up-regulated in the liver of L-Tsc1KO mice.

Up-Regulation of Dlk1-Dio3 Locus miRNAs Is mTORC1-Dependent.

The miRNAs of the Dlk1-Dio3 locus are expressed in pluripotent stem cells and in brain and placenta in adult mice (35). Furthermore, they are expressed in all tissues during embryogenesis and repressed after birth in liver and other tissues (36, 37). In contrast, we detected sustained expression of the miRNAs in L-Tsc1KO liver at 14 wk of age. To determine whether the miRNAs are up-regulated throughout postnatal development, we measured hepatic expression of 3 Dlk1-Dio3 miRNAs at 2, 4, and 8 wk of age. Mature miRNA levels were similar in control and L-Tsc1KO mice up to 4 wk of age (SI Appendix, Fig. S2A). However, at 8 wk, the miRNAs were significantly more abundant in L-Tsc1KO mice (SI Appendix, Fig. S2A). This was due primarily to loss of miRNA expression in control littermates at 8 wk, i.e., aberrantly sustained expression in the L-Tsc1KO mice (Fig. 2A). The L-Tsc1KO livers even exhibited a small increase in miRNA levels at 8 wk compared to 4 wk (Fig. 2A). We further observed a clear down-regulation of the three miRNAs at 8 wk in pure C57BL/6J (wild type) mice compared to 4-wk-old mice (Fig. 2B). These results demonstrate a physiological down-regulation of Dlk1-Dio3 locus miRNAs at 4–8 wk of age. Furthermore, loss of Tsc1 prevents this down-regulation.

Fig. 2.

Sustained mTOR signaling up-regulates expression of Dlk1-Dio3 locus noncoding genes in 8-wk-old L-Tsc1KO mice. (A) qPCR showing miRNA levels of 4- and 8-wk-old livers from control or L-Tsc1KO mice. Values are shown as fold change to the control 4-wk-old livers. (B) qPCR showing mature miRNA levels in C57BL/6J (wild type) mice at 4 and 8 wk of age. Values are shown relative to 4-wk-old values. (C) qPCR showing mature miRNA levels upon treatment with vehicle or rapamycin (“Rapa” bars). Values are shown as fold change comparisons of L-Tsc1KO (vehicle treated) and L-Tsc1KO Rapa (rapamycin treated) to control mice (vehicle treated). To assess significance, L-Tsc1KO values are compared to control values and L-Tsc1KO Rapa values are compared to L-Tsc1KO values. (D) qPCR showing mature miRNA levels upon treatment with vehicle or INK128 (“INK128” bars). Values are shown as fold change comparisons of L-Tsc1KO (vehicle treated), L-Tsc1KO INK128 (INK128 treated), and control INK128 (INK128 treated) to control mice (vehicle treated). To assess significance, L-Tsc1KO values are compared to control values and L-Tsc1KO INK128 values are compared to L-Tsc1KO values. (E) qPCR showing Meg3, Rian, and Mirg levels of 4- and 8-wk-old mouse livers from control or L-Tsc1KO mice. Values are shown as fold change to the control 4-wk mouse livers. (F) qPCR showing Meg3, Rian, and Mirg levels upon treatment with vehicle or rapamycin (“Rapa” bars). Values are shown as fold change comparisons of L-Tsc1KO (vehicle treated) and L-Tsc1KO Rapa (rapamycin treated) to control mice (vehicle treated). To assess significance, L-Tsc1KO values are compared to control values and L-Tsc1KO Rapa values are compared to L-Tsc1KO values. (G) qPCR showing Meg3, Rian, and Mirg levels upon treatment with vehicle or INK128 (“INK128” bars). Values are shown as fold change comparisons of L-Tsc1KO (vehicle treated), L-Tsc1KO INK128 (INK128 treated), and control INK128 (INK128 treated) to control mice (vehicle treated). To assess significance, L-Tsc1KO values are compared to control values and L-Tsc1KO INK128 values are compared to L-Tsc1KO values. *P < 0.05; **P < 0.01.

To determine if high mTORC1 activity in L-Tsc1KO mice is responsible for the observed dysregulation of Dlk1-Dio3 locus miRNAs, we treated control and L-Tsc1KO mice with rapamycin, an allosteric mTORC1 inhibitor, or INK128, an mTOR kinase site inhibitor, from 4 to 8 wk of age. Long-term rapamycin and INK128 treatment partly reduced expression of the miRNAs at 8 wk in L-Tsc1KO mice (Fig. 2 C and D and SI Appendix, Fig. S2 B and C). INK128 also had a mild inhibitory effect on the already low levels of Dlk1-Dio3 locus miRNAs in control mice at 8 wk (Fig. 2D). Also, we note that mTOR signaling was not altered in untreated control mice at 4 and 8 wk of age (SI Appendix, Fig. S2D). Thus, although the physiological inhibition of the Dlk1-Dio3 locus miRNAs in wild type mice appears to be independent of mTOR signaling, elevated mTORC1 activity in L-Tsc1KO mice maintains hepatic expression of Dlk1-Dio3 locus miRNAs beyond 4 wk of age.

To determine if the down-regulation of miRNAs from 4 to 8 wk is transcriptional, we assessed the hepatic levels of the lncRNAs of the Dlk1-Dio3 locus at 4 and 8 wk. Levels of Meg3, Rian, and Mirg were comparable in 4-wk-old control and L-Tsc1KO livers but significantly increased in 8-wk-old L-Tsc1KO livers compared to controls (SI Appendix, Fig. S3 A and B). Mirroring the level of miRNAs, the level of the lncRNAs was significantly reduced in control animals at 8 wk when compared to 4-wk-old mice. Moreover, 8-wk-old L-Tsc1KO animals displayed a further increase in levels of Meg3, Rian, and Mirg when compared to 4-wk-old control or L-Tsc1KO livers (Fig. 2E). Similar to the miRNAs, long-term rapamycin or INK-128 treatment prevented the high expression of Meg3, Rian, and Mirg in L-Tsc1KO livers (Fig. 2 F and G). Thus, constitutively high mTORC1 signaling sustains expression of Dlk1-Dio3 miRNAs and other noncoding RNAs beyond 4 wk of age.

mTOR Prevents CpG Methylation at the MEG-DMR of the Dlk1-Dio3 Locus.

CpG methylation of the IG and MEG DMRs that lie between the Dlk1 and Meg3 genes controls expression of the entire Dlk1-Dio3 locus (21, 38). Methylation at the IG-DMR is thought to determine imprinting of the locus, while methylation of MEG-DMR appears to play a secondary role in controlling expression (39, 40). Bisulfite pyrosequencing revealed that CpG methylation at IG-DMR and MEG-DMR was significantly increased in control mice at 8 wk of age compared to 4 wk (Fig. 3A), which likely explains the reduced expression of miRNAs and other ncRNAs of the Dlk1-Dio3 locus. Interestingly, L-Tsc1KO livers at 8 wk of age showed significantly reduced CpG methylation of MEG-DMR, but not of IG-DMR, compared to controls (Fig. 3A and SI Appendix, Fig. S3C). Both DMRs exhibited similar CpG methylation levels in L-Tsc1KO and control samples at 4 wk of age (SI Appendix, Fig. S3C). To confirm the methylation results obtained by bisulfite pyrosequencing, we assessed methylation changes using an assay based on cleavage by the HpaII and MsaI restriction enzymes, hereafter referred to as the restriction assay. We measured CpG methylation at sites adjacent to IG-DMR and MEG-DMR as described in the Materials and Methods section. Similar to the pyrosequencing results, the restriction assay showed an increase in CpG methylation at both DMRs in 8-wk-old control mice as compared to 4-wk-old (Fig. 3B). As before, CpG methylation was reduced only at the MEG-DMR in 8-wk-old L-Tsc1KO mouse livers compared to controls (Fig. 3 C and D).

Fig. 3.

CpG methylation and histone occupancy is reduced in L-Tsc1KO livers at 8 wk of age. (A) CpG methylation, measured via bisulfite pyrosequencing, at IG-DMR and MEG-DMR in 4-wk and 8-wk livers of control and L-Tsc1KO mice. Values are shown as log2 fold change over 4-wk-old liver samples of the respective genotype. The median and the range in CpG methylation are shown for each sample. Nlme R-package was used to determine significance. (B) Percent methylation measured at IG-DMR and MEG-DMR for 4- and 8-wk-old control mice. (C) Percent methylation measured at IG-DMR and MEG-DMR for 4-wk-old control and L-Tsc1KO mice. (D) Percent methylation measured at IG-DMR and MEG-DMR for 8-wk-old control and L-Tsc1KO mice. (E) Percent CpG methylation measured at IG-DMR and MEG-DMR via restriction assay using livers from 8-wk-old control and L-Tsc1KO mice that were treated with vehicle or rapamycin (“Rapa” labeled) for 4 wk from 4 wk of age. To assess significance L-Tsc1KO values are compared to control values and L-Tsc1KO Rapa values are compared to L-Tsc1KO values. (F) Percent CpG methylation measured at IG-DMR and MEG-DMR via restriction assay using livers from 8-wk-old control and L-Tsc1KO mice that were treated with vehicle or INK128 (“INK128” labeled) for 4 wk from 4 wk of age. To assess significance, L-Tsc1KO values are compared to control values and L-Tsc1KO INK128 values are compared to L-Tsc1KO values. (G) CpG methylation, measured via bisulfite pyrosequencing, at MEG-DMR in 8-wk-old control and L-Tsc1KO mouse livers treated with vehicle or INK128 for 4 wk starting at 4 wk of age. Values are shown as log2 fold change difference of L-Tsc1KO mouse livers, either vehicle or INK128 treated, over control mouse livers of the same treatment. The median and the range in CpG methylation are shown for each sample. Nlme R-package is used to determine significance. (H) Percent CpG methylation measured at the MEG-DMR using livers from 14-wk-old control and L-Tsc1KO mice. (I) ChIP assay showing occupancy of KAP1 and SETDB1 at MEG-DMR regions of control and L-Tsc1KO mice. Data are shown as fold change compared to occupancy of respective control samples. (J) ChIP assay showing occupancy of H2A, H3, and H4 histone proteins at MEG-DMR regions of control and L-Tsc1KO mice. Data are shown as percent of the input. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, not significant.

Next, we used the two methylation assays to determine the CpG methylation status of the MEG-DMR upon long-term rapamycin or INK-128 treatment. Treatment was started at 4 wk of age, and the livers were examined at 8 wk. The restriction assay revealed that rapamycin-treated L-Tsc1KO animals showed an increase in CpG methylation at MEG-DMR as compared to vehicle-treated controls (Fig. 3E). INK-128 treatment had a more pronounced effect in preventing loss of MEG-DMR methylation in L-Tsc1KO liver compared to controls (Fig. 3 F and G). The partially increased CpG methylation in drug-treated 8-wk-old livers suggests that constitutively high mTORC1 signaling in L-Tsc1KO mice prevents CpG methylation of MEG-DMR and thereby sustains miRNA expression.

As the level of Dlk1-Dio3 locus miRNAs was sustained in L-Tsc1KO animals even at 14 wk of age, we assessed the CpG methylation of MEG-DMR at this time point. L-Tsc1KO mice exhibited reduced CpG methylation at MEG-DMR at 14 wk of age, as observed at 8 wk, compared to controls (Fig. 3H). Thus, the failure to CpG-methylate MEG-DMR is also observed in older mice.

Methylation at many imprinted sites, including the Dlk1-Dio3 locus, is maintained by the zinc finger protein 57 (ZFP57)–KRAB-associated protein 1 (KAP1)–SET domain bifurcated 1 (SETDB1) complex. This complex recruits DNA methyl transferases (DNMTs) for de novo methylation, maintains already methylated CpG sites, and recruits other chromatin remodelers, such as nucleosome remodeling histone deacetylase (NURD) complex and heterochromatin protein 1 (HP1) to maintain a repressive chromatin environment (41, 42). Chromatin immunoprecipitation revealed decreased occupancy of KAP1 and SETDB1 at MEG-DMR in L-Tsc1KO liver (Fig. 3I and SI Appendix, Fig. S3 D and E). As the ZFP57–KAP1–SETDB1 complex maintains heterochromatin at CpG sites, we assessed histone occupancy at MEG-DMR in L-Tsc1KO liver at 14 wk. We observed a reduction in overall histone occupancy in L-Tsc1KO liver compared to controls (Fig. 3J and SI Appendix, Fig. S3F). These findings suggest that the reduced presence of the ZFP57–KAP1–SETDB1 complex at MEG-DMR accounts for the decreased methylation at the Dlk1-Dio3 locus in L-Tsc1KO liver.

Taken together, these experiments support a model where aberrantly sustained mTORC1 activity disrupts CpG methylation and heterochromatin silencing of the Dlk1-Dio3 locus that occurs starting at 4 wk of age. Lack of both CpG methylation and heterochromatin silencing in L-Tsc1KO livers is maintained as mice age and contributes to the continuous expression of the noncoding genes from the Dlk1-Dio3 locus. Our findings also suggest that MEG-DMR, rather than IG-DMR, is the main site mediating transcriptional repression of noncoding genes. Constitutively high mTORC1 activity prevents CpG methylation and heterochromatin silencing at MEG-DMR, resulting in sustained expression of miRNAs and the lncRNAs of the Dlk1-Dio3 locus.

Aberrant Expression of Dlk1-Dio3 miRNAs Causes Metabolic Defects.

L-Tsc1KO mice exhibit increased gluconeogenesis and glucose intolerance (10, 43, 44). To determine if this phenotype is due to aberrantly high expression of Dlk1-Dio3 miRNAs, we introduced a deletion of the so-called miR-379/miR-410 region of the Dlk1-Dio3 locus into the L-Tsc1KO mouse (SI Appendix, Fig. S4A). The miR-379/miR-410 region includes 41 miRNA genes of the Dlk1-Dio3 locus. We note that this deletion does not change mTORC1 signaling (SI Appendix, Fig. S4B). As shown previously, full body deletion of the miR-379/miR-410 region, hereafter referred to as miRKO, causes partial (30%) litter mortality upon weaning due to a defect in glucose metabolism, in particular low blood glucose (45). Like miRKO mice, L-Tsc1KO miRKO mice displayed partial litter mortality and no obvious phenotype as adults. We performed glucose tolerance tests (GTTs) and pyruvate tolerance tests (PTTs) on control, L-Tsc1KO, miRKO, and L-Tsc1KO miRKO mice. L-Tsc1KO mice exhibited glucose intolerance and increased gluconeogenesis, whereas miRKO mice displayed neither, as compared to control mice (Fig. 4 A and B and SI Appendix, Fig. S4 C and D). Interestingly, L-Tsc1KO miRKO mice showed improved glucose tolerance and reduced gluconeogenesis compared to L-Tsc1KO mice (Fig. 4 C and D). These observations suggest that aberrantly increased levels of the Dlk1-Dio3 locus miRNAs play a role in the metabolic phenotype of L-Tsc1KO mice.

Fig. 4.

Dlk1-Dio3 locus miRNAs play a role in gluconeogenesis and glucose metabolism. In this figure, at least four male mice 14–16 wk of age were used per condition. (A) Glucose tolerance test (GTT) of control and L-Tsc1KO mice. Area under the curve (AUC) is shown (Right) to determine significance. (B) Pyruvate tolerance test (PTT) of control and L-Tsc1KO mice. Area under the curve (AUC) is shown (Right) to determine significance. (C) Glucose tolerance test (GTT) of L-Tsc1KO versus L-Tsc1KO miRKO mice. Area under the curve (AUC) is shown (Right) to determine significance. (D) Pyruvate tolerance test (PTT) of L-Tsc1KO versus L-Tsc1KO miRKO mice. Area under the curve (AUC) is shown (Right) to determine significance. (E) qPCR showing transcript levels of enzymes involved in the gluconeogenesis pathway. The comparison between L-Tsc1KO mice to control mice is shown. (F) qPCR showing transcript levels of enzymes involved in the gluconeogenesis pathway. The comparison between L-Tsc1KO miRKO mice to L-Tsc1KO mice is shown. (G) Blood glucose levels measured after 18 h of starvation. The comparison of L-Tsc1KO mice to control mice is shown. (H) Blood glucose levels measured after 18 h of starvation. The comparison between L-Tsc1KO miRKO and L-Tsc1KO mice is shown. *P < 0.05; **P < 0.01; ***P < 0.001.

Glucose-6-phosphatase (G6P), fructose-1,6-bisphosphatase (FBP), phospho-enol-pyruvate carboxykinase (PEPCK), and pyruvate carboxylase (PC) are enzymes that catalyze gluconeogenesis. The genes expressing these enzymes were up-regulated in L-Tsc1KO mice compared to control mice and reduced in L-Tsc1KO miRKO mice compared to L-Tsc1KO mice (Fig. 4 E and F). Further, rapamycin treatment showed a trend in down-regulation of expression of the gluconeogenetic genes (SI Appendix, Fig. S4E) in a manner, as shown earlier, that correlates with decreased levels of Dlk1-Dio3 locus miRNAs, in L-Tsc1KO mice. It was previously shown that the increased rate of gluconeogenesis in L-Tsc1KO mice, compared to controls, is due to reduced protein kinase B (PKB; also known as AKT) signaling that in turn leads to derepression of the forkhead box protein (FOXO) transcription factors (11, 44). We did not see a significant difference in phosphorylation of AKT and FOXO in L-Tsc1KO liver compared to L-Tsc1KO miRKO liver (SI Appendix, Fig. S4F). Thus, Dlk1-Dio3 miRNAs promote gluconeogenesis via transcriptional activation of gluconeogenic genes, and this effect is independent of the AKT-FOXO axis.

Gluconeogenesis is turned on during periods of fasting to maintain blood glucose levels. After 18 h of fasting, blood glucose levels in L-Tsc1KO mice were high compared to controls (Fig. 4G). Interestingly, the increased fasting blood glucose levels were normalized in L-Tsc1KO miRKO mice as compared to L-Tsc1KO mice (Fig. 4H). These results suggest that miRNAs of the Dlk1-Dio3 locus, in particular the miRNAs in the miR-379/miR-410 region, maintain blood glucose levels during fasting. Taken together, our findings suggest that pathologically high mTOR signaling induces expression of miRNAs to promote gluconeogenesis. Furthermore, we propose that high mTORC1 signaling induces miRNA expression, most likely by preventing CpG methylation of the Dlk1-Dio3 locus (Fig. 5).

Fig. 5.

Loss of TSC1 regulates Dlk1-Dio3 locus miRNAs to affect gluconeogenesis. Changes with regard to Dlk1-Dio3 locus miRNAs are depicted in 4-wk- versus 8-wk-old animals in control and L-Tsc1KO mice. CpG islands at IG-DMR and MEG DMR are shown with filled (methylated) or empty (unmethylated) circles, and miRNAs are shown with vertical lines.

Discussion

Here we show that constitutively high mTORC1 signaling due to loss of the tumor suppressor Tsc1 up-regulates miRNAs of the Dlk1-Dio3 locus to possibly increase gluconeogenesis. Furthermore, we report that the developmentally expressed miRNAs of the Dlk1-Dio3 locus are transcriptionally silenced in the liver of wild type mice at 4 to 8 wk of age, but remain expressed in L-Tsc1KO livers. Sustained expression of the Dlk1-Dio3 locus miRNAs upon high mTOR activity correlated with lower CpG methylation of the Dlk1-Dio3 locus.

How does pathologically high mTORC1 activity (loss of Tsc1) prevent CpG methylation of the Dlk1-Dio3 locus to maintain expression of miRNAs? In wild type mouse livers, we observed more CpG methylation at IG-DMR and MEG-DMR of the Dlk1-Dio3 locus at 8 wk of age compared to 4 wk. Loss of Tsc1 prevented CpG methylation at MEG-DMR but not at IG-DMR. The interplay between IG-DMR and MEG-DMR CpG methylation and downstream consequences are still being unraveled (29, 46–48), but our data suggest that methylation at the IG- and MEG-DMRs are independently controlled. TSC1 appears to promote methylation at MEG-DMR without affecting methylation at IG-DMR. How loss of TSC1 prevents CpG methylation at MEG-DMR is unclear. CpG methylation is maintained by the ZFP57–KAP1–SETDB1 complex. We observed reduced histone and ZFP57–KAP1–SETDB1 complex occupancy at MEG-DMR upon loss of TSC1. The reduced binding by the complex could lead to decreased recruitment of the histone-depositing enzyme HP1, which would in turn account for less histones and increased miRNA expression. mTORC1 may phosphorylate members of the ZFP57–KAP1–SETDB1 complex, accounting for the increase in miRNA expression upon loss of TSC1. Interestingly, KAP1 was detected as a potential mTOR substrate in large-scale screens (49–51). Alternatively, mTORC1 may phosphorylate a factor directly involved in transcription of Dlk1-Dio3 locus miRNAs. Sustained mTOR activity may promote recruitment of this factor to MEG-DMR, thereby preventing methylation of the site. Proteins such as DPPA3, AFF3, or CTCF have been shown to act at the MEG-DMR (52–55). It would be interesting to determine if they are targets of mTOR.

How do the miRNAs of the Dlk1-Dio3 locus activate gluconeogenesis? Upon deletion of the miR-379/miR-410 region in L-Tsc1KO mice, we observed reduced transcription of gluconeogenic genes. It is possible that Dlk1-Dio3 locus miRNAs promote gluconeogenesis by inhibiting one or more negative regulators of gluconeogenic enzymes. AKT negatively regulates gluconeogenic genes by phosphorylating and inhibiting FOXO transcription factors. Loss of TSC1 reduces, via a negative feedback loop, AKT activity, but deletion of the miR-379/miR-410 region in the L-Tsc1KO background did not restore phosphorylation of AKT and FOXO. Thus, the effect of the Dlk1-Dio3 locus miRNAs on gluconeogenesis appears to be independent or downstream of the AKT-FOXO axis. Also, gluconeogenic flux is controlled posttranscriptionally via increased levels of AMP and/or changes in the NAD⁺/NADH ratio (56–59). It is possible that increased expression of Dlk1-Dio3 locus miRNAs alters the level of AMP and/or NAD⁺/NADH ratio, which may in turn affect gluconeogenesis. It remains to be determined how up-regulation or deletion of the Dlk1-Dio3 locus affects gluconeogenesis and fasting blood glucose levels. We also note that the miR-379/miR-410 region of the Dlk1-Dio3 locus is deleted in the entire body, and, as such, the phenotype seen with the L-Tsc1KO miRKO mice could potentially originate outside of the liver.

Our findings add another dimension to mTOR signaling and how it affects glucose metabolism. This report shows that mTORC1 signals through miRNA, and does so to impact glucose metabolism. Sustained expression of Dlk1-Dio3 locus miRNAs can explain the pathological increase in gluconeogenesis and blood glucose levels upon loss of TSC1. Dlk1-Dio3 locus miRNAs may be necessary to maintain elevated glucose levels in blood and tissues during mouse embryogenesis and periods of lactation. However, in adult mice, aberrantly increased levels of these miRNAs may cause high blood glucose levels that can produce a prediabetic phenotype. The Dlk1-Dio3 locus found in mouse chromosome 12 is conserved in humans in chromosome 14 and is linked to disease upon dysregulation. Consistent with our findings, patients with Temple syndrome, characterized with increased expression of Dlk1-Dio3 locus miRNAs, are obese and may develop diabetes (60, 61). The targets of Dlk1-Dio3 locus miRNAs and how they affect gluconeogenesis remain to be identified.

Materials and Methods

RNA Isolation.

Total RNA was isolated from 50 mg of mouse liver and was homogenized in 1 mL TRIzol (Sigma) using lysing matrix D tubes (Q-Biogene) and a bead beater. Subsequent steps were performed as per TRIzol manufacturer’s instructions.

Small RNA Sequencing and Analysis.

Total RNA was isolated as described earlier, and 1 µg of total RNA was used to generate small RNA libraries for sequencing as described in ref. 62. Libraries were sequenced on an Illumina HiSeq-2500 deep sequencer. For analysis, the raw miRNA sequencing data were uploaded to the ClipZ server (63) for adapter removal and annotation. Differentially expressed miRNAs were extracted using DESeq (64) implemented as a Bioconductor package. A cutoff at 0.05 for the adjusted P value provided by DESeq was used to filter the hits.

Quantitative Reverse Transcriptase PCR.

For mRNAs, quantitative PCR was performed from 200 ng of total RNA, isolated as described earlier. DNase digestion was performed using the Roche RNase free DNase kit (Roche) as per manufacturer’s instructions. cDNA synthesis was performed using iScript (Bio-Rad) as per manufacturer’s instructions. Samples were diluted 10 times prior to usage as a template in qPCR reactions together with SYBR Green Mix (Applied Biosystems) and primers. An Applied Biosystems StepOnePlus Real-Time PCR System was used. Relative expression levels were determined by normalizing each CT value to TBP expression using the ΔΔCT method. Sequences of the primers used are in SI Appendix, Table S3.

For miRNAs, quantitative PCR was performed from 2 pg of total RNA that was used as a template together with the Applied Biosystems miRNA kit (Applied Biosystems) as per manufacturer’s instructions to generate cDNA. miRNA TaqMan primers from Invitrogen for miR-127, miR-34a, miR-376b, miR-381, miR-455, miR-541, and miR-802 (cat no. 4427975) were used to perform the PCR, together with TaqMan Universal 2× MasterMix (Applied Biosystems) as per manufacturer’s instructions. An Applied Biosystems StepOnePlus Real-Time PCR System was used. Relative expression levels were determined by normalizing each CT value to U6 expression using the ΔΔCT method. Sequences of other primers used are shown in SI Appendix, Table S3.

Protein Isolation and Western Blotting.

Livers from mice were rapidly dissected and flash-frozen in liquid nitrogen. For protein extraction, liver tissues were homogenized using a Polytron machine (Polytron) in ice-cold T-PER lysis buffer (Invitrogen) supplemented with protease and phosphatase inhibitors (Sigma). Lysates were cleared by centrifugation at 10,000 × g for 15 min at 4 °C. Total protein concentration was assessed using a BCA Protein Assay kit (Invitrogen) as per manufacturer’s instructions. Protein (10 µg) was loaded on a gradient SDS/PAGE gel and transferred to a nitrocellulose membrane for blotting. Antibodies used are shown in SI Appendix, Table S2. Samples were probed overnight with primary antibody and 1 h with secondary antibody conjugated to HRP. A SuperSignal West Pico chemiluminescent kit (Thermo Fisher) was used to detect staining in a Fusion X machine (Vilber).

Rapamycin and INK128 Treatments.

For chronic mTOR inhibition, 4-wk-old mice were injected intraperitoneally daily for 4 wk with rapamycin at 1 mg/kg, INK128 at 0.5 mg/kg, or the respective vehicle. Rapamycin was dissolved in 5% PEG-400, 4% ethanol, and 5% Tween 80. INK128 was dissolved in 5% 1-methyl-2-pyrrolidinone, 15% polyvinylpyrrolidone K30, and 80% water. In all cases, the injection volume did not supersede 100 µL.

Statistical Analysis.

Unless stated otherwise, statistical significance was measured using a Student’s unpaired t test to determine differences among groups. The differences were considered to be significant if P < 0.05. Data are presented as mean ± SEM. The asterisks in the figure legends represent the degree of significance (*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001).

Glucose and Pyruvate Tolerance Tests.

For glucose tolerance tests (GTTs), mice were fasted for 18 h and administered 2 g/kg body weight glucose by intraperitoneal injection (i.p.). Blood glucose levels were measured in 10-min intervals for 180 min. For pyruvate tolerance tests (PTTs), mice were fasted for 18 h before administering 2 g/kg body weight sodium pyruvate via i.p. injection. Blood glucose levels were measured in 10-min intervals for a duration of 150 min. For each test, at least 6 male mice 14 to 16 wk of age per condition were used.

Mouse Husbandry.

L-Tsc1KO mice and miRKO mice were generated as described previously (45, 50). L-Tsc1KO miRKO mice were generated during this work by crossing L-Tsc1KO mice to miRKO mice. Littermate controls were used in all experiments. To determine the genotype, PCR was done as described previously (45, 65). Mice were maintained under temperature- and humidity-controlled conditions with 12-h light and 12-h dark cycles with free access to food and water. All experiments were performed in accordance with federal guidelines and were approved by the Kantonales Veterinaeramt of Kanton Basel-Stadt.

Pyrosequencing Analysis.

Tissues were ground in dry ice, and DNA was isolated using a DNA blood and tissue kit (Qiagen) as per manufacturer’s instructions. Five hundred nanograms of high-purity, intact DNA was used for bisulfite conversion using an EZ DNA Methylation-Gold kit (Zymo Research) by following standard protocols. Bisulfite-converted (BSC) DNA quality and concentration was determined using an RNA Pico 6000 Kit on a Bioanalyzer 2100 instrument (Agilent Technologies) and Nanodrop 2000 (Thermo Scientific). Bisulfite-converted (BSC) samples were normalized to 10 ng/µL. Primers were designed according to recommendations in ref. 66: IG-DMR_5prime FW-5′-GTGGTTTGTTATGGGTAAGTTT-3′ and RV-5′(Btn)-CCCTTCCCTCACTCCAAAAATTA-3′; MEG-DMR_5prime FW-5′-AGTTTGGGATTTAAAATTAAGGTTT-3′ and RV-5′(Btn)-CTAAAAACTATCACCCCCACAT-3′ (Microsynth). Promoter fragments were amplified using an AmpliTaq Gold Kit from Applied Biosystems (Life Technologies). PCR was done in 30-µL reactions containing the following: 1× PCR buffer II, 300 µM deoxynucleotide triphosphates, final 3.5 mM MgCl₂, 200 µM of each primer, and 20 ng of BSC DNA. We used the following cycling conditions: 95 °C, 15 min; 50 × (95 °C, 30 s; 55 °C, 30 s; 72 °C, 30 s); 72 °C, 10 min. PCR products were purified and sequenced using a PyroMark ID System (Biotage) following the manufacturer’s suggested protocol and the following sequencing primers: for IG-DMR_5prime 5′- GTTTTATGGTTTATTGTATATAATG-3′; for MEG-DMR_5prime 5′-TTTTTGTTTTAATAATGTTAAATTT-3′ (MicroSynth). Controlling for PCR temperature bias was done with a series of calibrator samples of known methylation levels. Briefly, unmethylated standards were prepared by using two rounds of linear whole-genome amplification with an Ovation WGA System Kit (NuGene), starting from 10 ng of DNA, as recommended by the manufacturer. Methylated standards were made using CpG methyltransferase assay with M.SssI (New England Biolabs) starting from 2 µg of purified DNA, following the standard protocol. Bisulfite conversion of standard samples was done as described earlier. All samples were analyzed in quadruplicates. Differential methylation analysis was performed using linear models and the limma R package, and multiple linear regression was used to determine significance (67). DNA methylation levels were averaged across CpGs in the examined promoter regions. To account for multiple testing, we applied Bonferroni correction.

Restriction Assay.

DNA from tissues was isolated using a DNA blood and tissue kit (Qiagen) as per manufacturer’s instructions. Restriction assays were performed using the EpiJet 5-mC analysis kit (Thermo Scientific) as per manufacturer’s instructions. The following primers were used for IG-DMR, 5′-CTGCAGCCGCTATGCTATG-3′ and 5′-CAGCTAACCTGAGCTCCATG-3′, and for MEG-DMR, 5′-GACGAAGAGCTGGAATAGAG-3′ and 5′-CATGTCCAGGAGGACGGAG-3′

Measuring Fasting Blood Glucose Levels.

In order to measure the fasting blood glucose levels, mice were starved for 18 h, after a small tail puncture an Aviva glucometer device (Accu-Check) was used to measure the concentration of glucose in a drop of blood. For each test, at least 6 male mice 14 to 16 wk of age were used.

Chromatin Immunoprecipitation.

Mouse livers were homogenized in cold PBS solution and cross-linked with 1% formaldehyde for 15 min at room temperature. Cross-linking was stopped with 0.125 M glycine, and homogenates were centrifuged at 600 × g for 15 min. The cell pellets were resuspended in lysis buffer (5 mM Pipes, 85 mM KCl, 0.5% Nonidet P-40) supplemented with protease inhibitors (Roche) and incubated at room temperature for 30 min. Cell lysates were centrifuged at 5,000 × g for 15 min, and nuclei were resuspended in nuclei lysis buffer (50 mM Tris⋅HCl, 10 mM EDTA, 1% SDS) for 30 min at room temperature. Chromatin was sonicated three times for 15 s using a Branson probe sonicator at 80% power setting. A total of 50 µg of DNA was incubated with 1 µg of antibody in the cold room overnight and further incubated with a mix of A/G beads (Pierce) for 2 h. Beads were washed three times in the cold room as per manufacturer’s instruction and eluted with 0.1 M NaHCO₃, 1% SDS buffer. Cross-linking was reversed by incubation at 65 °C for 6 h. DNA was purified using phenol chloroform-isoamyl alcohol (25:24:1) extraction, and ethanol precipitated. Recovered DNA was used in qPCR reactions. Antibodies used are shown in SI Appendix, Table S2, and primers used to amplify the DNA are 5′-GACGAAGAGCTGGAATAGAG-3′ and 5′-CATGTCCAGGAGGACGGAG-3′ for MEG-DMR; for the transcribed region, primers that map at GAPDH (Active Motif 71016) were used, and for the nontranscribed region, primers that map to a gene desert (Active Motif 71011) were used.

Data Availability.

Small-RNA sequencing data have been deposited in the Gene Expression Omnibus (GEO) database, https://www.ncbi.nlm.nih.gov/geo/ (accession no. GSE141361).