The Oncogenic MicroRNA miR-21 Promotes Regulated Necrosis in Mice

MicroRNAs (miRNAs) regulate apoptosis, yet their role in regulated necrosis remains unknown. miR-21 is overexpressed in nearly all human cancer types and its role as an oncogene is suggested to largely depend on its anti-apoptotic action. Here we show that miR-21 is overexpressed in a murine model of acute pancreatitis, a pathologic condition involving RIP3-dependent regulated necrosis (necroptosis). Therefore, we investigate the role of miR-21 in acute pancreatitis injury and necroptosis. miR-21 deficiency protects against caerulein- or L-arginine-induced acute pancreatitis in mice. miR-21 inhibition using locked-nucleic-acid-modified oligonucleotide effectively reduces pancreatitis severity. miR-21 deletion is also protective in tumor necrosis factor-induced systemic inflammatory response syndrome. These data suggest that miRNAs are critical participants in necroptosis, and miR-21 enhances cellular necrosis by negatively regulating tumor suppressor genes associated with the death-receptor-mediated intrinsic apoptosis pathway and could be a therapeutic target for preventing pathologic necrosis.


Introduction
Recent advances in the understanding of the fundamental biology of cellular death have led to the identification of two distinct, but interrelated, forms of programmed cell deathapoptosis and regulated necrosis 1 . Apoptosis involves the engagement of intrinsic or extrinsic pathways that activate caspase proteases. Morphologically, necrotic cells differ significantly from apoptotic cells. For example, necrotic cells exhibit translucent cytoplasm, swelling of organelles, and disruption of the plasma membrane with release of endogenous molecules such as lactate dehydrogenases (LDH), ATP, ions, and DNA that directly trigger a pro-inflammatory response. On the other hand, cells undergoing apoptosis display chromatin aggregation, nuclear and cytoplasmic condensation, and partitioning of cytoplasm and nucleus into membrane bound-vesicles. Moreover, unlike necrosis, apoptosis is not usually associated with tissue inflammation because phagocytic cells efficiently clear apoptotic cells. Once regarded as an unscheduled and unregulated cell death induced by nutrient deprivation and traumatic or toxic injury, necrotic cell death in some contexts has been found to be a regulated process driven by defined molecular pathways 2 .
Among many forms of regulated necrosis, necrotic cell death that depends on receptorinteracting protein 3 (RIP3, also known as RIPK3) is termed as necroptosis, a cellular process integrated with the extrinsic apoptosis pathway. Extrinsic apoptosis is regulated by death receptors such as tumor necrosis factor (TNF)-α receptor 1 (TNFR1) and CD95 that activate cell death pathways when engaged by their cognate ligands such as TNF-α and FasL (also known as CD95L). This extrinsic death pathway is mediated by the adaptor protein Fas-associated protein with a death domain (FADD), which couples the receptor generated-signals to the death-executing protease enzyme, caspase 8. Activation of caspase 8 triggers downstream effector caspases to initiate the classical apoptosis cascade. In contrast, TNFR-mediated initiation of necroptosis requires the kinase activity of receptorinteracting protein 1 (RIP1; also known as RIPK1) and RIP3, and its execution involves the active disintegration of mitochondrial membranes 1,3,4 . RIP1 and RIP3 are components of the necrosome, and caspase 8 can directly cleave and inactivate both kinases to negatively regulate necroptosis induced by TNF-α [5][6][7] . Upon pharmacological inhibition of caspase, RIP1 and RIP3 are phosphorylated to trigger necroptosis. The interplay of apoptosis and necroptosis is further demonstrated in recent mouse genetic experiments showing that ablation of RIP3 or RIP1 protects caspase 8-null or FADD-null mice from embryonic lethality [6][7][8][9] . Necroptosis is associated with several pathologic conditions, including acute pancreatitis, ischaemia-reperfusion injury, neurodegeneration and infection 1 . In an experimental model of acute pancreatitis, mice treated with supramaximal doses of caerulein predominantly display necrosis with little apoptosis, and the degree of necrosis correlates with injury severity 10 . Embryonic fibroblasts from RIP3-null mice have been found to be resistant to necrosis, and RIP3-null mice are devoid of inflammation-inflicted tissue damage in caerulein-induced acute pancreatitis, indicating RIPs are important in necroptosis 3 . Over 1,000 human microRNAs (miRNAs) have been discovered so far 11 , and they are predicted to target up to 92% of all human genes 12 . The role of miRNAs in cell death, particularly in apoptosis, has been extensively investigated. miR-21 is the most studied miRNA, and data from various transgenic and/or deficient mouse models of miR-21 indicate that miR-21 exerts its oncogenic function predominantly through inhibiting cellular apoptosis [13][14][15] . However, the role of miR-21 or any other miRNA in necroptosis remains completely unknown.
In the present study, we report that miR-21 is overexpressed in a murine model of acute pancreatitis, a condition characterized by cellular necrosis. Genetic ablation of miR-21 in mice protects against caerulein-or L-arginine-induced acute pancreatitis. Pharmacological miR-21 inhibition significantly reduces pancreatitis severity in wild-type (WT) mice. In addition, miR-21 deficiency is protective in tumor necrosis factor-induced tissue injury. Our results suggest that miR-21 promotes TNF-α induced necroptosis and indicate miR-21 is a potential therapeutic target for drug development against diseases associated with cellular necrosis.

Role of miR-21 in caerulein-induced acute pancreatitis
To study the role of miR-21 in acute pancreatitis, we first measured the expression of miR-21 in WT mice treated with phosphate-buffered saline (PBS)-formulated caerulein using quantitative real-time PCR (qPCR). We found that compared with untreated or PBStreated controls, the ΔC t values of miR-21 normalized to a reference RNA (snoRNA292) were significantly lower (P=0.003; Fig. 1a) in pancreas from mice at day 1 after the last caerulein injection, when the severity of acute pancreatitis was at its peak 16 . The difference in ΔC t values represents a >16-fold upregulation of miR-21 (using either mean or median values). Correspondingly, at day 3 and day 7 post treatment, miR-21 expression was reduced (Fig. 1a).
Next, we evaluated the changes in pancreatic histology at 1, 3, and 7 days after the last caerulein injection. In WT mice, extensive acinar cell injury and necrosis in pancreatic tissue was found at day 1 ( Fig. 1b). At day 3, the pancreatic tissue morphology was in the process of recovering and at day 7, it appeared normal (Supplementary Fig. 1a-1b). This agrees with existing knowledge that pancreatic regeneration after acute pancreatitis induced by caerulein is largely completed within a week, when the pancreas histology is similar to healthy untreated controls 16 . In contrast, miR-21 −/− mice at day 1 had dramatically less pancreatic injury (Fig. 1b). Furthermore, histological analyses showed reduced pancreatic edema and acinar cell necrosis in miR-21 −/− mice ( Fig. 1c-1d). Pancreatic regeneration at days 3 and 7 in miR-21 −/− mice, however, was similar to that in WT mice ( Supplementary Fig. 1a-1b), likely because regeneration after caerulein-induced injury is partially completed by day 3 and by day 7 is completed in both WT and miR-21 −/− mice.
We performed immunofluorescence staining for amylase in pancreata and confirmed a substantial loss of exocrine cells in WT mice, but not in miR-21 −/− mice at day 1 post treatment ( Fig. 1e and 1f). Elevated amylase and lipase activity in the serum is a hallmark of acute pancreatitis in both humans and mice [17][18][19] . Compared with WT mice, both serum amylase and lipase activity in miR-21 −/− mice were significantly lower 6 h after the final dose of caerulein ( Fig. 1g and 1h). Taken together, these results suggest that miR-21 deficiency protects against acute pancreatitis induced by caerulein in mice.
In addition to acinar cell injury, another feature of acute pancreatitis is damage to distal organs such as the lung. As shown in Supplementary Fig. 2, lung injury in WT mice manifested as thickened alveolar membranes and neutrophil sequestration as reflected by higher myeloperoxidase (MPO) enzyme activity. In contrast, alveolar membrane thickening and MPO activity were markedly lower in miR-21 −/− mice treated with caerulein compared with WT mice. These observations indicate that genetic ablation of miR-21 reduces the severity of lung injury following caerulein-induced pancreatitis.

Regulation of inflammatory cell infiltrate by miR-21
Monocyte-derived macrophages are known to infiltrate the pancreata during acute pancreatitis and to act as inflammatory mediators that participate in the pathophysiology of pancreatitis 20,21 . Thus, we determined whether miR-21 deletion reduces inflammatory cell infiltration. Upon caerulein treatment, CD11b-positive cells (i.e., monocytes and macrophages) rapidly migrated to the injured pancreata of WT mice starting day 1 after caerulein treatment ( Fig. 2a and 2b); however, in miR-21 −/− mice the number of infiltrating CD11b-positive cells was much lower after treatment with caerulein. Immunostaining with F4/80, a marker of murine macrophages, confirmed that fewer macrophages were present in the pancreata of miR-21 −/− mice ( Fig. 2c and 2d). Similarly, tissue sections stained with antibodies against Gr1, a neutrophil marker, showed less neutrophil infiltration in miR-21 −/− pancreata compared with WT mice (Fig. 2e and 2f). These data support the notion that miR-21 −/− mice have less severe acute pancreatitis upon caerulein treatment than WT mice.
To determine whether the reduced severity of acute pancreatitis in miR-21 −/− mice was due to decreased macrophage activation, we profiled the pro-inflammatory gene expression in bone marrow-derived macrophages (BMDMs). Macrophages display remarkable plasticity and follow two distinct states of polarized activation: the classically activated (M1) phenotype and the alternatively activated (M2) phenotype 22 . Upon M1 stimuli such as lipopolysaccharides (LPS), macrophages secrete cytokines (including TNF-α, IL-1β, and IL-6) and enzymes such as prostaglandin-endoperoxide synthase 2 (Ptgs2) to promote an inflammatory response. In response to the T helper type 2 cytokines IL-4 and IL-13, M2 polarization of macrophages dampens inflammation and promotes tissue remodeling 22 . Studies of acute pancreatitis show an early and intense M1 activation of macrophages, characterized by high expression of TNF-α and other M1 genes 21,23,24 . When treated with LPS, only the levels of the chemokine Cxcl-9 and the enzyme gene Ptgs2 were lower in miR-21 −/− BMDMs compared with the WT BMDMs ( Supplementary Fig. 3a). On the other hand, upon IL-4 (an M2 stimulus) addition, expression of anti-inflammatory genes, such as Arg1, Csf1, Ccl22 and Ccl17, was decreased by miR-21 deletion (Supplementary Fig. 3b).
To test the relative role of miR-21 in bone marrow-derived inflammatory cells during pancreatitis, we performed bone marrow transplantation using WT mice (with CD45.1) as recipients and miR-21 −/− or WT mice (both with CD45.2) 25 as donors. Prior to bone marrow transplantation, we analyzed peripheral blood lineage cells and found no differences in cell populations between WT and miR-21 −/− mice. At 4 weeks after bone marrow transplantation, mice were analyzed for hematopoietic recovery and level of chimerism. Analysis of white blood cells in peripheral blood indicated normal hematopoietic recovery with no significant differences between mice transplanted with bone marrow isolated from WT or miR-21 −/− mice ( Supplementary Fig. 4). Flow cytometry analysis of peripheral blood mononuclear cells for the presence of donor-derived leukocytes expressing CD45.  Fig. 4 and Supplementary Fig. 5). We then treated animals with caerulein at 5 weeks after transplantation. The response to caerulein-induced acute pancreatitis in WT mice receiving miR-21 −/− bone marrow was similar to that in WT mice transplanted with WT bone-marrow ( Fig. 2g and 2h). Collectively, these data suggest that the phenotypic difference in acute pancreatitis between caerulein-treated WT and miR-21 −/− mice is unlikely to be due to differences in macrophages or other bone marrow-derived blood cells and more likely to depend on the intrinsic response of acinar cells.

The expression of necroptosis genes and miR-21 targets
The severity of experimental pancreatitis in rodents positively correlates with necroptosis but negatively with apoptosis 10 . In mice, necrosis is the major mechanism of acinar cell death in caerulein-induced acute pancreatitis 10 . We first determined whether deletion of miR-21 increases cell apoptosis in the pancreata at day 1 post treatment. In agreement with a previous report 10 , we found few apoptotic acinar cells as judged by terminal deoxynucleotidyl transferase nick end labeling (TUNEL) in the pancreata of WT mice, but based on cell morphology, some inflammatory cell apoptosis was inferred (Fig. 3a). Yet TUNEL staining of inflammatory cells and acinar cells in miR-21 −/− pancreata was elevated compared with the WT controls ( Fig. 3a and 3b). We also measured cleaved caspase-3 using immunohistochemistry and found more cells stained positively in tissues from miR-21 −/− mice than the WT controls ( Fig. 3c and 3d). Next, we analyzed the expression of RIP1, RIP3 and FADD in the pancreata of WT and miR-21 −/− mice as these proteins are part of the necrosome that plays a key role in caerulein-induced pancreatitis. In line with previous observations 3 , RIP3 expression level was dramatically elevated in the pancreata of caerulein-treated WT mice; however, RIP3 levels in miR-21 −/− pancreata were significantly lower than in WT controls post-treatment (Fig. 3e). In addition, FADD protein was lower in miR-21 −/− pancreata compared with WT controls. Levels of RIP1 were elevated by caerulein treatment in both WT and miR-21 −/− pancreata, yet miR-21 deletion produced little change (Fig. 3e). Lastly, we probed the expression of five reported miR-21 target genes (PDCD4, PTEN, SPRY2, FASL, and BCL-2). Only Pten and FasL expression was higher in miR-21 −/− mice than in WT mice (1.6 and 2.1 fold, respectively, Fig. 3f). The expression of Pdcd4, Spry2 and Bcl-2 remained largely unchanged. These data indicate that miR-21mediated target suppression maybe effective for some but not all miR-21 gene targets in pancreata.
Given the role of miR-21 targets in the extrinsic apoptosis pathway 3 , we hypothesized that up-regulation of miR-21 targets in the pancreas elevates caspase 8 activity, which in turn, downregulates RIP1 and RIP3 expression and acinar cell necrosis. Analyzing caspase 8 activity, first, we found that the expression of total caspase 8 was reduced in miR-21 −/− pancreata compared with WT controls post treatment, but the active form of caspase 8 (i.e., cleaved caspase 8 protein [p43]) was upregulated in miR-21 −/− pancreata (Fig. 3g). The expression of a downstream caspase (total caspase 3) was significantly upregulated in miR-21 −/− pancreata, but no cleaved caspase 3 products were detected (data not shown). Next, we isolated acinar cells to recapitulate necrosis in vitro. miR-21 expression was higher in WT cells treated with cholecystokinin-8 (CCK-8), which stimulates necrosis in vitro (Fig.  3h). CCK-8 induced significant necrotic cell death, as judged by LDH activity (an indicator of mitochondrial rupture and necrosis), in WT acinar cells -an effect inhibited by miR-21 genetic ablation (Fig. 3i). Caspase 8 activity, however, was higher in miR-21 −/− acinar cells than in WT cells at the basal level and with CCK-8 induction (Fig. 3j). Four h post CCK-8 treatment, RIP1 processing had significantly increased in miR-21 −/− acinar cells compared with WT controls (Fig. 3k), and Pten and FasL expression were elevated as well. (Fig. 3l). These results support the possibility that miR-21 deficiency leads to less necrotic cell death and higher caspase 8 activity in isolated acinar cells and in pancreata.
To determine whether CCK receptor expression is affected by miR-21 genetic depletion, we measured the expression of CCK receptors A and B (CCKAR and CCKBR) in pancreatic tissues from mice treated with caerulein, as well as in isolated acinar cells treated with CCK-8. We found no significant differences in CCKAR and CCKBR expression between WT and miR-21 −/− tissues ( Supplementary Fig. 6a) or between isolated WT and miR-21 −/− cells ( Supplementary Fig. 6b). In addition, we found that the amylase activity in the supernatant did not differ between in WT and miR-21 −/− acinar cells treated with CCK-8 for 1-4 h ( Supplementary Fig. 6c). These data indicate that miR-21 deficiency affects neither expression nor function of CCK receptors. The difference in serum amylase activities in Fig.  1g is likely due to the differential necrotic death rate of WT and KO acinar cells in pancreata.

Caspase inhibition promotes necrosis in miR-21 −/− mice
To test whether elevated caspase 8 activity is responsible for the seen protective effects of miR-21 deficiency against caerulein-induced acute pancreatitis, we pre-treated animals with z-VAD-fmk, a pan-caspase inhibitor. z-VAD-fmk pre-treatment significantly enhanced pancreatitis severity in miR-21 −/− mice as judged by morphology ( Supplementary Fig. 7a) and histological scores for necrosis ( Supplementary Fig. 7b). We also tested the efficacy of z-IETD-fmk, a caspase 8-specific inhibitor 26,27 , and similarly, z-IETD-fmk strongly sensitized miR-21 −/− mice to caerulein-induced pancreatitis (Supplementary Fig. 7c and 7d). In combination with our in vitro data, these results indicate that the potential mechanism underlying the observed reduction in acute pancreatitis severity in miR-21 −/− mice can be attributed to the increased activity of caspase 8, an endogenous negative regulator of necroptosis.

miR-21 deficiency attenuates L-arginine-induced pancreatitis
Because a high dose of L-arginine (L-Arg) also induces acute necrotizing pancreatitis in rodents 28,29 and potentially in humans 30 , we employed this model to test whether the protective effect of miR-21 ablation was reproducible and model-independent. Extensive pancreatic tissue injury was observed in WT mice 3 days after L-Arg treatment, yet in miR-21 −/− mice pancreata were largely intact ( Supplementary Fig. 8a). Serum amylase and lipase activity were significantly elevated in WT mice compared with miR-21 −/− mice. (Supplementary Fig. 8b and 8c). Histological analyses confirmed the reduced severity of acinar cell necrosis in miR-21 −/− mice compared with WT mice (Supplementary Fig. 8d). These data provide further evidence that genetic deletion of miR-21 protects pancreata from necrotic injury during acute pancreatitis.

miR-21 inhibition in WT mice ameliorates acute pancreatitis
Given the attenuation of acute pancreatitis in miR-21-deficient mice, we sought to determine whether inhibiting miR-21 with a locked nucleic acid (LNA)-based stable, nontoxic backbone-modified oligonucleotide complementary to miR-21 (anti-miR-21) could effectively treat or prevent caerulein-induced acute pancreatitis. In a treatment study to determine whether anti-miR-21 reverses ongoing acute pancreatitis, we injected anti-miR-21, 20 or 50 mg kg −1 , into WT mice after the last dose of caerulein. Neither dose reduced disease severity. In a prevention study, we delivered anti-miR-21 simultaneously with caerulein into WT mice. Compared with either control anti-miR or vehicle (PBS), anti-miR-21 administration significantly downregulated miR-21 (Fig. 4a) and upregulated its targets Pten and FasL (Fig. 4b). Consistent with gene expression, acute pancreatitis severity was consistently attenuated by anti-miR-21 ( Fig. 4c-4f). These observations complement our findings that miR-21 deletion in mice protects the pancreas and indicate that prophylactic anti-miR-21 administration protects animals from caerulein-induced pancreatic injury.

miR-21 deficient mice are protected from TNF-α-induced shock
A common complication of acute pancreatitis is a systemic inflammatory response syndrome (SIRS), which often leads to multiple organ failure. RIP1-and RIP3-mediated cellular necroptosis drives mortality during TNF-α-induced SIRS 31 . We hypothesized that miR-21 deletion protects against lethal SIRS. We administered TNF-α to mice to test the role of miR-21 in TNF-α-induced SIRS. We first evaluated animal survival using a lethal TNF-α injection. All WT mice died within 10 h, whereas only 10% of miR-21 −/− mice succumbed; 20% of miR-21 −/− mice eventually survived TNF-α shock (P<0.0001, n=10 per group, log rank test, Fig. 5a). We then examined systemic changes at 3 h after TNF-α injection in a separate set of mice. The serum LDH levels were significantly reduced in miR-21 −/− mice compared with WT controls (Fig. 5b). Ileum, kidney and liver were collected, and tissue damage was evaluated by histological analyses. We found that villi were more swollen and inflammatory infiltrates were greater in the ileum of WT than miR-21 −/− mice ( Fig. 5c and 5d). Moderately increased kidney tubule dilatation, degeneration, and necrosis also were found in WT compared with miR-21 −/− mice ( Fig. 5e  and 5f). Liver damage was slightly inhibited in miR-21 −/− mice compared with WT mice, but the difference was not statistically significant ( Fig. 5g and 5h). These data demonstrate that loss of miR-21 protects mice from TNF-α-induced tissue damage and mortality.

Fas/FasL and necrostatin-1 treatment in pancreatitis
The pathological aggravation of acute pancreatitis is reported to be related to the Fas/FasL pathway. First, FasL expression in pancreatic tissues from mice treated with caerulein was consistently lower than that from sham-treated animals 32 . Second, in patients with acute pancreatitis, FasL serum levels decreased significantly as acute pancreatitis severity increased 33 . Third, Fas deficiency exacerbates caerulein-induced pancreatitis 34 . To determine whether FasL, a miR-21 target, is responsible for the observed protection exerted by miR-21 deletion in caerulein-induced pancreatitis, we treated WT mice with caerulein plus recombinant mouse FasL (mFasL), finding that mFasL did not significantly influence pancreas injury (Supplementary Fig. 9a and 9b). We then applied MFL3, a FasL-inhibiting antibody, in caerulein-treated miR-21 −/− mice. We found that MFL3 did not affect the severity of pancreatitis in miR-21 −/− mice (Supplementary Fig. 9c and 9d). As controls, we also administered MFL3 to WT mice and mFasL to miR-21 −/− mice treated with caerulein and found no significant impact. We next sought to inhibit necroptosis using necrostatin-1 (Nec-1) in both caerulein-induced pancreatitis and TNFα-induced shock in WT and miR-21 −/− mice. Nec-1 is an allosteric inhibitor of RIPK1 kinase activity, and thus is a potential drug candidate against pathologic necroptosis. In the acute pancreatitis model, we found Nec-1 exacerbates pancreas damage in WT mice, which is in line with the published data 35 . Yet Nec-1 had no effect on miR-21 −/− mice treated with caerulein ( Supplementary  Fig. 10a and 10b). In the TNFα-induced shock model, we found that Nec-1 protected against liver damage moderately in WT or miR-21 −/− mice treated with TNFα, but not against ileum damage; kidney damage was increased significantly by Nec-1 in both WT and miR-21 −/− mice ( Supplementary Fig. 11a-11f). These experiments indicate that the blockade of necroptosis with Nec-1, as a drug candidate, did not adequately protect against acute pancreatitis or septic shock-induced tissue damage. Nonetheless, these results in combination with genetic ( Fig. 1) and pharmacologic data with anti-miR-21 (Fig. 4) suggest that inhibiting miR-21, rather than Nec-1 treatment or Fas/FasL manipulation, is effective in ameliorating tissue damage in acute pancreatitis induced by caerulein.

miR-21 in recruitment of the necrotic complex
To delineate the molecular mechanism of miR-21 in necrosis, we examined necrotic cell death using mouse embryonic fibroblasts (MEFs) as an in vitro cell model. In MEFs and many other cells, the combination of TNF-α and cycloheximide (TC) induces apoptosis, whereas the addition of z-VAD-fmk to this regimen (TCZ) promotes necrosis 36 . We found that upon TCZ treatment, miR-21 −/− MEFs exhibited less necrosis than WT cells; no differences in apoptosis were observed (Fig. 6a to 6d). The recruitment of RIP1 and RIP3 to FADD is essential to form the necrosome, which is required for necroptosis. Immunoprecipitation experiments with anti-FADD antibodies revealed that after treatment with TCZ, significantly less RIP1 and RIP3 protein was associated with FADD in miR-21 −/− MEFs compared with WT MEFs (Fig. 6e). We then determined the role of miR-21 target genes (Pten and FasL) in cellular necrosis of miR-21 −/− MEFs treated with TCZ. Notably, these two miR-21 targets were upregulated in miR-21 −/− MEFs 15 , in miR-21 −/− pancreata (Fig. 3c, and in miR-21 −/− acinar cells, Fig. 3i). Knockdown of either Pten or FasL via siRNA (Fig. 6f) significantly enhanced cellular death of TCZ-treated miR-21 −/− MEFs (Fig. 6g). As expected, recruitment of RIP1 and RIP3 to FADD increased when Pten or FasL expression was downregulated (Fig. 6h). These data indicate that upregulation of Pten and FasL is partially responsible for the reduced necrosis observed in cells and animals without miR-21.

Discussion
Recent studies have identified new participants in regulated necrosis 1, 2, 37 , yet the role of noncoding RNAs in necroptosis and pancreatitis remains unexplored. The major finding of this study is that miR-21 regulates necrotic cell death and that deletion of miR-21 prevents necroptosis during caerulein-induced acute pancreatitis. In mouse models of acute pancreatitis, acinar cell death is regarded as a form of regulated necrosis, specifically necroptosis that depends on the activity of two kinases, RIP1 and RIP3 3,4 . Because miR-21 is overexpressed in pancreata from mice treated with caerulein, we examined the potential pathological function of miR-21 in acute pancreatitis. Compared with WT animals, miR-21 −/− mice are protected against acinar cell necrosis and edema, demonstrating, for the first time, a novel pathogenic contribution of miR-21 in this acute pancreatitis model. Moreover, this result was validated in a second model of acute pancreatitis induced by L-Arg. Acute pancreatitis is the most common reason for hospitalization due to gastrointestinal disease. Necrosis is the final cause of pancreatic tissue injury during acute pancreatitis, and it precipitates several major pathobiologic processes such as pancreatic inflammation and edema and leads to systemic involvement of extrapancreatic organs 17 . We demonstrate that miR-21 inhibition using LNA-based oligonucleotides also protects mice from caeruleininduced pancreas injury. Given the paucity of treatment options for acute pancreatitis, inhibition of miR-21 may be a feasible therapeutic strategy against acute pancreatitis and its secondary clinical complications.
To dissect the mechanism of miR-21 action, we reasoned that the reduced severity of acute pancreatitis in miR-21-deficient mice could be due to less cell death caused by acinar cell necrosis, a limited pro-inflammatory response, or a combination of both processes. The bone marrow transplantation experiment provided evidence to exclude the possibility that inflammatory infiltration by macrophages or other leukocytes played a differential role in the less severe acute pancreatitis phenotype in miR-21 −/− mice. Thus, although we appreciate that inflammatory cells are essential for caerulein-induced pancreatitis injury in mice, our data indicate that the necrotic death of acinar cells is likely the predominant underlying cause of the phenotypic difference between miR-21 −/− and WT mice. This role for miR-21 was further confirmed in isolated acinar cells independent of the presence of inflammatory cells. When treated with CCK-8 as an in vitro stimulus of necrosis, miR-21deficient acinar cells displayed less necrotic cell death than CCK-8-treated WT cells. Similar protection was observed in primary miR-21 −/− MEFs undergoing stimulated necrosis. Notably, we found that miR-21 plays opposite roles in two forms of cell death -miR-21 −/− MEF cells display reduced necrotic cell death, but are more susceptible to apoptotic cell death stimulated by DNA damage 14,15 .
We further distinguished the mechanism of miR-21 in necroptosis by examining the expression of RIP1, RIP3, FADD, and caspase 8 in the pancreata of miR-21 −/− mice. Caerulein treatment elevates pancreatic RIP3 protein levels, yet miR-21 deletion significantly reduces the induction of this critical component of necroptosis. The expression of two endogenous inhibitors of necrosis -FADD and caspase 8 -are differentially affected. FADD is moderately downregulated, yet the activated form of caspase 8 is significantly upregulated by miR-21 deletion. We reasoned that elevated caspase 8 activity is likely insufficient to induce massive apoptosis but may be adequate to inhibit RIP3 expression and reduce necroptosis. We tested this hypothesis by pretreating miR-21 −/− mice with z-VAD-fmk (a pan-caspase inhibitor) or z-IETD-fmk (a caspase 8-specific inhibitor), which largely abolished the protective effect of miR-21 genetic ablation on acute pancreatitis. These data suggest that elevated caspase 8 activity is pivotal in protecting against acute pancreatitis mediated by miR-21deficiency. This is consistent with previous reports showing that deletion of caspase 8 elevates necroptosis 6,7 . The precise mechanism by which caspase 8 is engaged to prevent necroptosis without triggering apoptosis in acinar cells during acute pancreatitis remains elusive. miR-21 participates in the regulation of multiple gene networks, and it has been reported to directly repress the expression of more than 30 genes 38 . These genes include FasL and Pten, which are involved in death receptor-mediated extrinsic apoptosis pathways 38 (Fig.  6i). FasL is a type II transmembrane protein that belongs to the TNF family and induces necroptosis through a complex consisting of caspase 8 and FADD by binding its cognate receptor CD95 1 . It is reported that Pten promotes TNF-α-or FasL-induced apoptosis by facilitating caspase 8 activation that depends on FADD 25,39 . We found that FasL and Pten are upregulated in miR-21 −/− pancreata, and thus, we examined whether modulation of these miR-21 targets affects necrotic cell death. Both necrotic cell death and necrosome recruitment were enhanced in miR-21 −/− cells under necrotic stimuli when the expression of FasL or Pten was knocked down. Yet manipulating FasL alone cannot recapitulate the phenotype of protecting mice from caerulein-induced acute pancreatitis as does miR-21 genetic deletion or inhibition with anti-miR-21. These findings suggest that miR-21 modulates cell necroptosis through negative regulation of FasL and Pten and most likely many other targets and further supports a key role for miR-21 in a complex network of interrelated death receptor-mediated apoptosis and necroptosis (Fig. 6i).
Taken together, our data support the idea that miR-21 is a pro-necroptosis miRNA. In future studies, it will be critical to determine how miR-21 enhances RIP1/3-mediated necroptosis and thereby creates a pro-inflammatory microenvironment with a strong necrosis element contributing to many other life-threatening conditions.

Murine pancreatitis models
The miR-21 knockout mouse line (miR-21 −/− ) 15 was crossed back to the C57BL/6J genetic background through speed congenic breeding. Female mice (22-25 g, age 7-8 weeks) were used for caerulein-induced pancreatitis, and males were used for the bone marrow transplantation and L-arginine-induced pancreatitis models. We did not find sex bias in caerulein-induced acute pancreatitis in mice. Mice, randomized by body weight if there are two or more groups, were administered two series of 8 hourly intraperitoneal injections (separated by 40 h) of a supramaximally stimulating dose of caerulein (50μg per kg of body weight [kg −1 ], Sigma-Aldrich, St. Louis, MO). The final day of caerulein injection was defined as day 0 (d0). For the mechanistic studies of necroptosis, z-VAD-fmk or z-IETDfmk (BD Pharmingen; San Jose, CA) was dissolved in dimethyl sulfoxide at 10 mg ml −1 and administered intraperitoneally 15 min before the first dose of caerulein at 5mg kg −1 40 . For the L-Arg model, mice were injected (i.p.) with L-Arg hydrochloride (8%, pH 7.0 in PBS) in two doses of 4 g kg −1 , 1 h apart; animals were euthanized after 72 h 28 . No lethality was observed. All animal experiments were performed under protocols approved by University of Louisville Institutional Animal Care and Use Committee and in accordance with the Guide for the Care and Use of Laboratory Animals (NIH). To determine the water content, pancreata were obtained at 12 h after the last caerulein injection and the wet tissue weight determined. Tissues were then freeze-dried overnight and reweighed to obtain a dry weight; pancreatic water content was calculated as the difference between the wet and dry tissue weights and expressed as a percentage of the total (wet) weight.

TNF-α-induced SIRS model
Recombinant mouse TNF-α (Shenandoah Biotechnology, Warwick, PA) was injected intravenously via the tail vein into WT and miR-21 −/− mice in a volume of 0.2 ml (9 μg in endotoxin-free PBS); animals were observed hourly for survival. In a separate set of experiments, mice were sacrificed 3 h after TNF-α injection, and serum and tissue samples of ileum, kidney and liver were collected. Histopathology of tissue damage of ileum, kidney and liver were evaluated by a pathologist in a double-blind manner on paraffin sections on a scale ranging from 0 (completely unaffected tissue), 1 (negligible damage), 2 (minor damage), 3 (moderate damage), to 4 (severe damage).

Histology, immunohistochemistry (IHC) and apoptosis
Pancreatic tissues were collected and fixed overnight in 10% neutral-buffered formalin. Fixed tissue was paraffin embedded, sectioned (5 μm thick), and stained with hematoxylin for morphological analyses. Quantification of histological scoring for acute pancreatitis and necrotic death of acinar cells was performed as described 41 . Cells with swollen cytoplasm, loss of plasma membrane integrity, and leakage of organelles into interstitium were considered necrotic. The necrotic areas were automatically calculated based on hematoxylin staining for pancreas or hematoxylin and eosin staining for other tissues using Olympus cellSens software 42 . The extent of acinar cell necrosis was expressed as the percent of the total acinar tissue that was occupied by necrotic areas, meeting the criteria for necrosis: (i) the presence of acinar cell ghosts or (ii) vacuolization and swelling of acinar cells and the destruction of the histoarchitecture of whole or parts of the acini.
For IHC, primary antibodies against CD11b, F4/80 (Abcam, Cambridge, MA) and Gr1 (BD Pharmingen, San Diego, CA) were used to assess inflammatory cell infiltration. Deparaffinized and rehydrated sections were boiled in Na-citrate buffer (10 mM, pH 6.0) for 10 min for antigen retrieval. Tissue sections were developed using the Ultra Vision Detection System, anti-rabbit (Thermo Scientific). About 1,500 positive cells were counted per slide per animal (n=5 animals in each group) to calculate the percentage of positively stained cells. For immunofluorescence staining of amylase, pancreatic tissue sections were incubated with anti-α-amylase antibody (1:400, Sigma) for 2 h at room temperature. After being rinsed with PBS, the sections were incubated with Alexa Fluor 488 goat anti-rabbit (1:100, Invitrogen) for 1 h at room temperature. After another PBS rinse, the sections were incubated with 1.0 μg ml −1 4′, 6-diamidino-2-phenylindole (DAPI) for 5 min. Slides were analyzed under fluorescence microscopy. Morphometry of amylase density was based on computer-aided area measurement of amylase protein expression (excluding islets).
For the TUNEL assay to detect apoptosis, deparaffinized pancreatic sections were incubated with 20μg ml −1 protease K for 15 min at room temperature, then washed with PBS and incubated with the TUNEL reaction mixture (Roche Applied Science) for 60 min at 37°C in a humidified atmosphere. Tissue sections were incubated with converter-peroxidase (with anti-fluorescein antibody) and subsequently with the 3, 3′-diaminobenzidine substrate. Images were acquired using an Olympus IX51 microscope and the cellSens digital imaging software. Nuclei with condensed or fragmented chromatin (brown cells) were considered apoptotic; 2,000 acinar cells were counted per slide per animal (n=5 animals in each group). We also performed IHC staining for cleaved caspase 3 as condensed fragmented chromatin is not specific 43 .

Mouse primary cell preparation
Acinar cells were prepared by collagenase digestion following an established protocol 44 . Briefly, mice were fasted overnight, and the pancreas was removed, minced, washed with PBS with 0.01% soybean trypsin inhibitor (Sigma-Aldrich, St. Louis, MO). Tissues were rinsed with Ham's F12K medium (Invitrogen, Carlsbad, CA) with antibiotics and 0.1 mg ml −1 soybean trypsin inhibitor, 5 mg ml −1 bovine serum albumin faction V (Sigma-Aldrich) and 10% fetal bovine serum. The mixture was centrifuged at 500 g for 2 min, and the pellet was resuspended in 10 mL digestive solution containing 200 U ml −1 collagenase type III (Sigma-Aldrich) and 0.2 mg ml −1 bovine serum albumin fraction V. Digestions were performed at 37°C with shaking at 130 cycles per min for 30 min. The pellet was then collected by centrifugation at 500 g for 2 min, resuspended, and passed through a pipette several times. Large cell blocks were removed by filtration through a 70-μm cell strainer. The filtered cells were washed twice with Ham's F12K medium. Acinar cells obtained were maintained in Ham's F12K supplemented with 10% fetal bovine serum, 0.1 mg ml −1 soybean trypsin inhibitor and 5 mg ml −1 bovine serum albumin faction V and incubated at 37°C with 5% CO 2 . To stimulate necrosis, acinar cells were treated with CCK-8 (100nmol L −1 ) for 2-8 h 45 . Mouse embryonic fibroblasts (MEFs) were prepared according to an established protocol 15,46 . Primary WT and miR-21 −/− MEFs were treated with TNF-α (T, 20 ng ml −1 ), cycloheximide (C, 1 μg ml −1 ), and/or z-VAD-fmk (Z, 20 μM).
Bone marrow-derived macrophages (BMDMs) were prepared using an established protocol 47 . Mice (8 weeks old) were sacrificed and hind legs were removed. All muscle tissues from the bones were removed using scissors. The femur was separated and flushed with PBS using a 3-mL syringe and a 23-gauge needle. Bone marrow was passed through a pipette several times to make a single-cell suspension. Cells were washed with PBS twice before being plated onto a culture dish. Cells were maintained in DMEM supplemented with 10% fetal bovine serum and 10 ng ml −1 murine granulocyte macrophage colony stimulating factor (M-CSF, R&D systems, Minneapolis, MN) and incubated at 37°C with 5% CO 2 .

Amylase, lipase, caspase 8, MPO activity and necrosis assay
Blood was collected from mice through the tail vein and kept at 4°C for 1 h before centrifugation at 4,000 g for 20 min at 4°C. Serum was collected from the supernatant and amylase activity was measured by using EnzyChrom ™ α-Amylase assay kit (BioAssay systems, Hayward, CA) according to the manufacturer's protocol. Lipase activity was assayed using the Lipase Detection Kit (Colorimetric) (ab102524, Abcam). Caspase 8 activity was measured by homogeneous luminescent assay using the Caspase-Glo 8 kit (Promega, Madison, WI) according to the manufacturer's protocols. For MPO activity, lung samples were homogenized in 20 mM phosphate buffer (pH 7.4), and the pellet was resuspendend in 50 mM phosphate buffer (pH 6.0) containing 0.5% hexadecyltrimethylammonium bromide (Sigma). The suspension was sonicated for 30 s, centrifuged at 10,000 g for 5 min, and the supernatant was used for the MPO assay. The supernatant was mixed with 1.6 mM tetramethylbenzidine (Sigma), 80 mM sodium phosphate buffer (pH 5.4), and 0.3 mM hydrogen peroxide. This mixture was then incubated at 37°C for 110 s, the reaction was terminated with 0.18 M H 2 SO 4 , and the absorbance was measured at 450 nm. This absorbance was then corrected for the calculated dry weight of the tissue samples.

RNA preparation and qPCR
To measure M1 and M2 expression in BMDM, BMDM were isolated from WT and miR-21 −/− mice, and total RNA was extracted using Trizol (Invitrogen) and run on an Agilent Bioanalyzer 2100 to ensure an A260/A280 ratio in the range of 1.8-2.2 and an rRNA ratio (28S/18S) > 0.9. RNA was reverse-transcribed with the iScript ™ Reverse Transcription Supermix kit (BIO-RAD, Hercules, CA). Real-time PCR was performed using the iQ ™ SYBR Green Supermix (BIO-RAD) and CFX96 ™ Real-Tme PCR system (BIO-RAD) with primers designed based on Universal Probe Library (Roche) with ACTB as the reference (oligonucleotide sequences were provided in Supplementary Table 1).

Bone marrow transplantation
Twelve-week-old recipient WT CD45.1 mice (B6.SJL-Ptprc a Pep3 b /BoyJ; Jackson Laboratory, Bar Harbor, Maine; n=16) were conditioned with 950 cGy total body irradiation from a cesium source (Gamma-cell 40; Nordion, Ontario, Canada). At 24 h after irradiation, the mice were transplanted with bone marrow cells (1×10 7 cells per mouse in 0.1 ml PBS) isolated from tibias and femurs of 11-week-old CD45.2 WT and miR-21 −/− donor mice through the retro-orbital plexus with a 27-gauge needle. Six mice received WT bone marrow and 10 received miR-21 −/− bone marrow. After 4 weeks, recipient mice were characterized for hematopoietic recovery and chimerism by determining the relative percentages of CD45.2 cells in peripheral blood. Co-staining with markers of T-cells (CD3), B-cells (B220), granulocytes (Gr-1), and monocytes (CD11b) was used to determine the level of chimerism in lineage-specific cells. Peripheral blood was obtained through retro-orbital bleeding for whole blood cell counts using a HemaVet 950 system (Drew Scientific Group, Oxford, CT), and nucleated cells were analyzed by flow cytometry (LSRII, BD Biosciences) to determine the percentage of CD45.2 cells. Both the recipient and donor mice were maintained on normal chow throughout the experiment.

RNA interference and miR-21 inhibition
Low-passage MEFs were transfected with 25 nM siRNA duplexes targeting PTEN and FASL by using lipofectamine RNAiMAX reagent. After 48 h, the cells were collected and assayed to ensure silencing was achieved. siRNA duplex oligonucleotides, purchased from Dharmacon (Lafayette, CO), were: FASL (5′-UACCAGUGCUGACCAUUUA-3′), PTEN (5′-GAUCAGCAUUCACAAAUUA-3′), and a control (Non Targeting siRNA Pool #2). For miR-21 inhibition in cells, the Ambion Anti-miR miRNA Inhibitors (Life Technologies, Grand Island, NY) were used with the Negative Control #1 as the control for transfection, and cells were collected after 48 h. For miR-21 inhibition in murine acute pancreatitis, LNA-modified antisense to miR-21 (anti-miR-21) was mixed with caerulein and intraperitoneally injected into mice. The control (control-anti-miR) was the miRCURY LNA microRNA inhibitor control Negative Control A (Exiqon, Woburn, MA), and it has no more than 70% homology to any sequence in the mouse genome. The total dose for anti-miR-21 or control-anti-miR is 20 mg kg −1 of body weight.

Statistical analysis
Data are presented as mean with s.d.. The SPSS16.0 software (SPSS Inc. Chicago, IL) was used for all statistical analyses. An unpaired two-tailed student's t-test was performed for two-group comparisons. One-way ANOVA was performed for multiple group comparisons with one independent variable and two-way ANOVA for multiple group comparisons with two independent variables (miR-21 gene status, treatment, histology, etc.). For the survival analysis, Kaplan-Meier results were analyzed by log rank test. Statistical significance was set at *P<0.05 or **P<0.01.

Supplementary Material
Refer to Web version on PubMed Central for supplementary material.     mice were collected 3 h after challenge with 9 μg TNF-α, and LDH activities were analyzed (n=5 per group, a separate cohort). Representative histological sections of ileum (c), kidney (e), and liver (g) stained with hematoxylin and eosin 3 h after TNF-α injection in WT and miR-21 −/− mice. Scale bars=200 μm for ileum and 100 μm for kidney and liver. (n=5 for each group). The tissue damage score for ileum (d), kidney (f) and liver (h) was evaluated. *P<0.05 and ** P<0.01 indicate a significant difference between WT and miR-21 −/− mice treated with TNF-α (twoway ANOVA). NS: not significant.