mTORC1 suppresses PIM3 expression via miR-33 encoded by the SREBP loci

The mechanistic target of rapamycin complex 1 (mTORC1) is a central regulator of cell growth that is often aberrantly activated in cancer. However, mTORC1 inhibitors, such as rapamycin, have limited effectiveness as single agent cancer therapies, with feedback mechanisms inherent to the signaling network thought to diminish the anti-tumor effects of mTORC1 inhibition. Here, we identify the protein kinase and proto-oncogene PIM3 as being repressed downstream of mTORC1 signaling. PIM3 expression is suppressed in cells with loss of the tuberous sclerosis complex (TSC) tumor suppressors, which exhibit growth factor-independent activation of mTORC1, and in the mouse liver upon feeding-induced activation of mTORC1. Inhibition of mTORC1 with rapamycin induces PIM3 transcript and protein levels in a variety of settings. Suppression of PIM3 involves the sterol regulatory element-binding (SREBP) transcription factors SREBP1 and 2, whose activation and mRNA expression are stimulated by mTORC1 signaling. We find that PIM3 repression is mediated by miR-33, an intronic microRNA encoded within the SREBP loci, the expression of which is decreased with rapamycin. These results demonstrate that PIM3 is induced upon mTORC1 inhibition, with potential implications for the effects of mTORC1 inhibitors in TSC, cancers, and the many other disease settings influenced by aberrant mTORC1 signaling.

is coupled to mTORC1 signaling via the transcription factors SREBP1 and 2 (sterol regulatory element-binding proteins 1 and 2). SREBP transcriptional activity is induced by mTORC1 via its stimulation of SREBP processing from an endoplasmic reticulum-bound inactive form to a mature nuclear form 18,[23][24][25][26] . The mTORC1-mediated increase in SREBP transcriptional activity induces the transcript levels of its own gene products Srebf1 and Srebf2 18 , and we find that this regulation causes a concomitant increase in the levels of its intronic microRNA, miR-33 27 . We demonstrate that induction of miR-33 downstream of mTORC1 suppresses PIM3 expression. These findings highlight a previously unappreciated mechanism of gene repression downstream of mTORC1, and demonstrate an unexpected outcome of treatment with mTORC1 inhibitors in the induction of the PIM3 proto-oncogene.

PIM3 expression is repressed downstream of mTORC1 and induced by mTOR inhibitors.
Through analysis of gene expression array data from Tsc1 −/− and Tsc2 −/− mouse embryonic fibroblasts (MEFs), which have constitutive mTORC1 activation, Pim3 transcript levels were found to be repressed relative to untreated wild-type MEFs and induced in a time-dependent manner over 24 hours of treatment with the mTORC1 inhibitor rapamycin (Fig. 1a). PIM3 protein levels increased with similar kinetics, peaking with prolonged rapamycin treatments of 12 or 24 hours (Fig. 1b). We confirmed that both PIM3 mRNA and protein levels were decreased in Tsc2 −/− MEFs and found that mTORC1 inhibition with either rapamycin or the mTOR kinase domain inhibitor Torin1 restored PIM3 expression to wild-type levels (Fig. 1c,d). The ability of rapamycin to induce PIM3 mRNA and protein levels was also observed in ELT3 cells, a cellular model of TSC derived from a TSC2 −/− uterine leiomyoma arising in the Eker rat (Fig. 1e,f) 28 . Even suppression of the low basal mTORC1 signaling in wild-type MEFs with rapamycin led to a further increase in PIM3 expression, demonstrating that this effect is not specific to genetic settings with loss of TSC2 (Fig. 1g).
In vivo, mTORC1 signaling is highly sensitive to feeding status, especially in the liver, being repressed during fasting and acutely activated upon feeding. Thus, to determine whether physiological control of mTORC1 signaling influenced PIM3 expression, PIM3 mRNA and protein levels were measured in liver samples from mice that had been fasted or fed with or without rapamycin pretreatment. PIM3 transcript and protein levels were highest in the fasted state and were strongly suppressed upon feeding, coincident with activation of mTORC1 signaling, indicated by phosphorylation of its direct downstream target S6K1 (Fig. 1h,i). Importantly, the feeding-induced suppression of PIM3 was largely dependent on mTORC1 activation, as PIM3 mRNA and protein levels remained elevated in liver samples from mice treated with rapamycin just prior to feeding.
To ascertain the broader applicability of these findings, we determined the effects of rapamycin on PIM3 in human cancer cell lines that exhibit growth factor-independent mTORC1 activation and varying levels of PIM3 (Fig. 2). In human glioblastoma cells (U87MG) with activated mTORC1 signaling downstream of PTEN loss, PIM3 transcript levels were increased upon mTORC1 inhibition (Fig. 2a). In U87MG cells expressing a doxycycline-inducible PTEN (U87MG-iPTEN), repression of mTORC1 signaling either with rapamycin or PTEN re-expression via doxycycline treatment led to elevated PIM3 (Fig. 2b). PIM3 levels were also increased to varying degrees by rapamycin treatment in the hepatocellular carcinoma (HCC) lines JHH-4, Hep-G2 and JHH-6 ( Fig. 2c-e), and the breast cancer cell lines MDA-MB-453 and MDA-MB-468 (Fig. 2f,g). Therefore, mTORC1 signaling suppresses PIM3 expression in a variety of mammalian settings, resulting in PIM3 induction by rapamycin.
A survey of mTORC1-regulated transcription factors identifies SREBP1 and 2 as upstream of PIM3. Due to the nature of our original transcriptional profiling experiment and the timescale over which PIM3 is induced by rapamycin, we hypothesized that a transcription factor downstream of mTORC1 influences PIM3 expression. We therefore tested siRNAs targeting a panel of transcription factors established to be downstream of mTORC1 for effects on PIM3 levels in Tsc2 −/− MEFs, including hypoxia-inducible factor 1 alpha (HIF1α), c-myc, activating transcription factor 4 (ATF4), transcription factor EB (TFEB), and sterol regulatory element-binding proteins 1 and 2 (SREBP1/2) (Fig. 3). None of these knockdowns affected mTORC1 signaling. Interestingly, only siRNA-mediated knockdown of SREBP1 and 2 resulted in a substantial increase in PIM3 levels (Fig. 3e). The effect of SREBP knockdown was similar to that observed with rapamycin treatment, and knockdown of both SREBP1 and 2 blocked further induction by rapamycin (Fig. 3f).
Full-length SREBP is retained as an inactive precursor form on the membrane of the endoplasmic reticulum (ER), and mTORC1 signaling promotes its proteolytic processing at the Golgi and subsequent nuclear localization of its mature form, which binds to sterol regulatory elements (SREs) in the promoters of the genes that it induces (Fig. 4a) 23 . Therefore, in Tsc2 −/− MEFs, mature SREBP protein levels are elevated, resulting in increased transcription of numerous SREBP target genes 18 including its canonical target SCD1 (Fig. 4b). Treatment with rapamycin decreased both the precursor and mature forms of SREBP1 and lowered expression of SCD1 to levels similar to wild-type MEFs. siRNA-mediated knockdown of SREBP1 and 2 increased PIM3 transcript levels, concurrent with a decrease in the mRNA levels of SREBP1, SREBP2, and SCD1 (Fig. 4c). Independent of mTORC1 signaling, SREBP processing and activation is sensitive to intracellular sterol levels 29 . Interestingly, treatment of Tsc2 −/− cells with 25-hydroxycholesterol (25-HC), which, like rapamycin, strongly inhibits SREBP processing and expression of SCD1, resulted in increased PIM3 mRNA and protein levels despite sustained mTORC1 activity (Fig. 4d,e). Thus, inhibition of SREBP with either siRNAs or sterols overrides the mTORC1-mediated suppression of PIM3 expression. miR-33, an intronic microRNA within the SREBP loci, targets PIM3 downstream of mTORC1. SREBP1 and 2 are primarily characterized as activators of gene transcription, although a few examples of transcriptional repression have been described 30,31 . However, there are no discernible SREs in the PIM3 promoter, indicating that SREBP1 and 2 are unlikely to directly repress PIM3 expression. Interestingly, the SREBP1 and 2 loci (gene names SREBF1 and 2) contain intronic microRNAs that are expressed upon transcription of SREBP ( Fig. 5a) 27 . These microRNAs encoded by the SREBP1 and 2 loci, respectively miR-33b and miR-33a, differ by just two nucleotides and thus widely target the same set of transcripts 32 . It should be noted that while humans express both miR-33 forms, mice only possess miR-33a 27 . Importantly, the PIM3 3′UTR contains a consensus target sequence for miR-33 that is conserved throughout mammals, including in mouse, rat, and human ( Fig. 5b), all species where rapamycin induces PIM3 expression (Figs 1, 2) 33 . Furthermore, a recent study found that miR-33 does indeed target the PIM3 transcript 34 . While mTORC1 signaling promotes the processing of SREBP1 and 2, it also induces transcription of the SREBP loci due to autoregulation from SREs present in both the SREBP1 and 2 promoters 18,23,35,36 . We confirmed that the transcription of SREBP1 and 2 is sensitive to prolonged rapamycin treatments of 12 to 24 hours (Fig. 5c). Importantly, this decrease in SREBP transcript levels resulted in corresponding decreases in miR-33a levels upon rapamycin treatment (Fig. 5d).
The PIM kinase PIM1 is highly homologous to PIM3 37 , and its mRNA transcript has also been shown to be a target of miR-33a 38 . Therefore, we investigated its expression levels in Tsc2 −/− MEFs upon mTORC1 inhibition with rapamycin treatment. Consistent with this shared regulatory feature with PIM3, PIM1 mRNA and protein levels were induced by rapamycin (Fig. 5e,f).
To determine whether mTORC1 signaling suppresses PIM3 expression through the induction of miR-33, we tested the effects of both an anti-miR inhibiting miR-33a and a miR-33a mimic in Tsc2 −/− MEFs. Introduction of the anti-miR targeting miR-33a induced a time-dependent increase in PIM3 expression, similar to the de-repression observed with rapamycin treatment (Fig. 6a). Conversely, a mimic of miR-33a decreased PIM3 levels and attenuated the ability of rapamycin to induce PIM3 at both the mRNA and protein levels (Fig. 6b,c). It is interesting to note that while we observed concomitant changes in PIM3 mRNA and protein levels with the miR-33a mimic, a previous study found that miR-33 mimics reduced PIM3 protein without detectable effects on its transcript levels 34 . As microRNAs are well known to affect both mRNA stability and translation 39 , it is possible that the inhibitory mechanism of miR-33 on the PIM3 transcript could vary at different timepoints and between different settings. Collectively, our data suggest that mTORC1 signaling suppresses PIM3 through the induction of SREBP transcription and a corresponding increase in miR-33, which blocks PIM3 expression (Fig. 6d). Discussion mTORC1 is aberrantly activated in the majority of human cancers, and increasing evidence has highlighted the vital role of sustained mTORC1 signaling in resistance to targeted therapies of upstream pathways 40,41 . While a number of studies have indicated that mTORC1 inhibition is necessary for therapeutic response to oncogene-targeted therapies, it is also recognized that mTORC1 inhibition alone is generally not sufficient for a robust anti-tumor response in most settings 2 . mTORC1 inhibitors induce autophagy, which can have pro-survival effects on tumor cells 42 , and also relieve feedback inhibition of receptor tyrosine kinase signaling leading to enhanced activation of the pro-survival kinase Akt 43,44 , suggesting cell-survival mechanisms are promoted by mTORC1 inhibition. A more thorough understanding of signal wiring and rewiring upon mTORC1 activation and inhibition could help explain the cytostatic effect of mTORC1 inhibitors and suggest promising candidates for the design of combination therapies. In this study, we demonstrate that a proto-oncogenic kinase, PIM3, is repressed downstream of mTORC1, and that its expression is induced upon treatment with the mTORC1 inhibitor rapamycin. Importantly, this induction of PIM3 expression is also observed under physiological control of mTORC1 in the liver with fasting and feeding. Furthermore, induction of PIM3 by rapamycin was also observed in a variety of human cancer cell lines. Our data suggest an additional mechanism limiting the effectiveness of mTORC1 inhibitors as single agent cancer therapies. There has been increasing interest in the PIM kinases in cancer due to their role as pro-survival kinases. Indeed, they phosphorylate a consensus sequence highly similar to that preferred by AGC family kinases such as Akt and S6K, and PIM kinases have overlapping substrates with these kinases that contribute to their pro-survival and pro-growth role in cells 43,45,46 . The PIM kinases are unique in that they are constitutively active and have a short half-life, such that their levels are proportional to their cellular activity 47 . While PIM levels are regulated primarily at the transcriptional and protein stability levels [48][49][50][51] , there are also reports implicating miRNAs as key regulators of PIM expression 34,38,52 . Consistent with these reports, we find evidence that both PIM3 and PIM1 expression are repressed downstream of mTORC1 via its induction of SREBP1 and 2 transcriptional activity and a subsequent increase in miR-33 levels. Interestingly, the inhibition of these kinases downstream of mTORC1 could reflect a novel negative feedback mechanism, as the PIM kinases have been suggested to induce mTORC1 signaling via TSC2 and PRAS40 phosphorylation in some settings 53,54 . Furthermore, a recent study found that PIM1 is upregulated in breast cancer cells that are resistant to the PI3K inhibitor BYL719 and that all three PIM kinases are capable of sustaining mTORC1 signaling and cell proliferation in the presence of the inhibitor in these resistant cells 55 . Whether upregulation of the PIM kinases contributes to the limited effectiveness of, or resistance to, mTOR inhibitors in some settings is an interesting area for future investigation. Several recent studies have found that mTORC1 inhibitors can cause pronounced changes in cellular miR-NAs 56,57 . This is partly due to mTORC1-mediated repression of the microRNA processing enzyme DROSHA 58 , which leads to a general increase in miRNA expression upon mTORC1 inhibition; however, a subset of microRNAs were found to be decreased upon rapamycin treatment. The ability of mTORC1 to modulate the expression of miRNAs provides new insights into the full molecular effects of mTORC1 inhibitors, and alterations in the expression of miR-33 may play a key role in the response to these inhibitors in some settings. A recent study suggested that miR-33a inhibition might decrease sensitivity to cisplatin treatment in HCC 59 . Perhaps this effect is partly due to increased levels of PIM3 in this setting, and other uncharacterized miR-33 targets might contribute to the pro-survival effect of miR-33 inhibition. However, other studies have suggested that high levels of miR-33 might be detrimental to cancer therapy 60,61 . These conflicting findings highlight that the role of miR-33 in cancer and therapeutic responses remains to be fully elucidated.
Consistent with their expression requiring transcriptional induction of the SREBP1/2 loci, the majority of miR-33 targets that have been characterized to date are involved in fatty acid and cholesterol metabolism 27,32,62 . Given that the PIM kinases are believed to have a high degree of functional redundancy, it is interesting to note that PIM1 appears to stabilize the cholesterol transporter ABCA1, which is also a canonical miR-33 target, providing a rationale for PIM targeting by miR-33 63 . While many canonical PIM targets are involved in cell survival, recent work has also implicated them in various metabolic pathways, including glycolysis and mitochondrial biogenesis [64][65][66][67] . Furthermore, transgenic mice with human PIM3 expression in the liver exhibited increased lipid droplet accumulation when challenged with a carcinogen 68 , indicating that the PIM kinases may have as-yet undefined roles in lipid regulation, and providing further rationale for their regulation downstream of SREBP and miR-33. The metabolic consequences of decreased miR-33 and subsequent increase in PIM3 and PIM1, therefore, might also influence the cellular and systemic responses to mTOR inhibitors in cancer and other disease settings. Interestingly, chronic and complete inhibition of mTORC1 in the liver has been found to enhance carcinogen-induced hepatocellular carcinoma 69 , an effect also observed with liver-specific overexpression of PIM3 68 .
Ethics Approval. All animal experiments in this study were approved by the Harvard Medical Area's Institutional Animal Care and Use Committee and were performed in accordance with the approved guidelines for animal experimentation at Harvard Medical School.
Statistical Analysis. All qRT-PCR data were analyzed with GraphPad Prism (La Jolla, CA, USA). P-values were calculated by an unpaired two-tailed Student's t-test, where appropriate. Data Availability. All data generated during this study are included in this published article and its Supplementary Information files. Raw qRT-PCR data from multiple independent experiments, represented in graphical form in the manuscript, are available from the corresponding author on request.