miR‐34a−/− mice are susceptible to diet‐induced obesity

Objective MicroRNA (miR)−34a regulates inflammatory pathways, and increased transcripts have been observed in serum and subcutaneous adipose of subjects who have obesity and type 2 diabetes. Therefore, the role of miR‐34a in adipose tissue inflammation and lipid metabolism in murine diet‐induced obesity was investigated. Methods Wild‐type (WT) and miR‐34a−/− mice were fed chow or high‐fat diet (HFD) for 24 weeks. WT and miR‐34a−/− bone marrow‐derived macrophages were cultured in vitro with macrophage colony‐stimulating factor (M‐CSF). Brown and white preadipocytes were cultured from the stromal vascular fraction (SVF) of intrascapular brown and epididymal white adipose tissue (eWAT), with rosiglitazone. Results HFD‐fed miR‐34a−/− mice were significantly heavier with a greater increase in eWAT weight than WT. miR‐34a−/− eWAT had a smaller adipocyte area, which significantly increased with HFD. miR‐34a−/− eWAT showed basal increases in Cd36, Hmgcr, Lxrα, Pgc1α, and Fasn. miR‐34a−/− intrascapular brown adipose tissue had basal reductions in c/ebpα and c/ebpβ, with in vitro miR‐34a−/− white adipocytes showing increased lipid content. An F4/80high macrophage population was present in HFD miR‐34a−/− eWAT, with increased IL‐10 transcripts and serum IL‐5 protein. Finally, miR‐34a−/− bone marrow‐derived macrophages showed an ablated CXCL1 response to tumor necrosis factor‐α. Conclusions These findings suggest a multifactorial role of miR‐34a in controlling susceptibility to obesity, by regulating inflammatory and metabolic pathways.


Introduction
The growing incidence of obesity worldwide has presented new therapeutic challenges, making it critical to understand the underlying mechanisms of obesity-associated pathology. In 2014, the World Health Organization estimated that 12.9% of adults have obesity (BMI 30 kg/m 2 , age 18 years) worldwide, with the Americas (26.8%) and Europe (23%) showing the highest percentage. Obesity is associated with increased low-grade, chronic inflammation that predisposes individuals to a number of comorbidities, e.g., type 2 diabetes (T2D), dyslipidemia, cardiovascular diseases (CVDs), and cancers (1,2).
The main proponent of weight gain during obesity is white adipose tissue (WAT). However, far from being just a lipid storage vessel, WAT can modulate the inflammatory response through secretion of inflammatory cytokines, including CCL2, CXCL8, tumor necrosis factor (TNF)-a, interleukin (IL)21b, and IL-6 (1-3). In opposition, brown adipose tissue (BAT) expends fatty acids (FA) to produce heat through mitochondrial action and uncoupled protein 1 (UCP1), maintaining body temperature. Peroxisome proliferator-activated receptor (PPAR)-c coactivator (PGC)21a is important for BAT adaptive thermogenesis, increasing mitochondrial biogenesis and UCP1 expression and promoting brown adipocyte differentiation (4).
MicroRNAs (miRNAs/miRs) are small (19-25 nt), noncoding RNAs, which primarily bind the 3' untranslated region of target mRNA, mediated by Argonaute proteins within the cytoplasmic RNA-inducing silencing complex (RISC). Before RISC loading, miRNAs form a duplex consisting of a mature (5p) and star (3p) strand, but usually only the 5p is loaded. mRNA is silenced through inhibition of translation or degradation (5).
The miR-34 family consists of 34a, 34b, and 34c, with miR-34a encoded on chromosome 1p36.22 and miR-34b and c on chromosome 11q23.1 (6). Interestingly, miR-34a appears to be dysregulated during obesity, with increased serum miR-34a transcripts found in patients who have excess weight (BMI 25 kg/m 2 ) with T2D, compared with patients who have excess weight without diabetes (7). In addition, studies have shown a correlation with increasing BMI and miR-34a transcripts in human subcutaneous WAT (scWAT) and increased miR-34a transcripts over subcutaneous adipocyte differentiation in vitro (8,9). Further to its pro-diabetes role, miR-34a transcripts have been positively associated with destruction of primary b-islet cells and cell lines, upon inflammatory stimulation, with decreased insulin secretion, through inhibition of VAMP2 (10,11). Inflammatory roles of miR-34a have been proposed, with lipopolysaccharide-induced inflammation reducing miR-34a expression in macrophage cell lines, and miR-34a mimics reducing IL-6 and TNF-a expression (12). Furthermore, miR-34a has been shown to target the soluble IL-6 receptor (13).
Adipose tissue macrophages (ATMs) represent a major immune infiltrate into the obese WAT (predominantly visceral), increasing from around 12% to 41% (14). Increasing inflammatory cytokines, particularly CCL2, recruits monocytes into the WAT that differentiate into ATMs (15). A high percentage are proinflammatory F4/80 1 CD11b 1 CD11c 1 M1 ATMs in high-fat diet (HFD)-fed rodents, and depletion of these cells has shown a reversal in inflammation and insulin resistance (16). Other immune cells, such as neutrophils, have also been shown to contribute to insulin resistance and obesity-induced inflammation (17).
Therefore, the aims of this study were to examine the metabolic and inflammatory role of miR-34a in a murine, diet-induced obesity model, using miR-34a 2/2 mice; specifically, examining miR-34a's role in adipocytes and macrophages during obesity. Our findings show that miR-34a 2/2 mice are susceptible to weight gain, likely through basal metabolic gene changes in adipose caused by dysregulation of PGC-1a.

Results
Expression of miR-34a in obese adipose tissue and in vitro bone marrow-derived macrophages In situ hybridization showed miR-34a distribution within visceral WAT was ubiquitous throughout the epididymal (e)WAT of WT mice on chow and HFD and omental adipose from bariatric surgery patients with obesity ( Figure 1A). Similarly, no change was observed in miR-34a or miR-34a* transcripts between WT chow and HFD groups ( Figure 1B). However, we observed increases in both miR-34a and 34a* transcripts within the liver (P 5 0.0088, P 5 0.0247) and intrascapular (i)BAT (P 5 0.0433, P 5 0.0399) during HFD feeding ( Figure 1C). Furthermore, we observed an increase in miR-34a (P 5 0.0086) and miR-34a* (P 5 0.0461) transcripts in WT bone marrowderived macrophages (BMDMs) when stimulated with TNF-a ( Figure  1D). We examined the other miR-34 family members but found no expression changes (Supporting Information Figure S1).
However, ex vivo investigations revealed a significant increase in eWAT weight in HFD-fed miR-34a 2/2 mice (0.61 6 0.10 g to 1.80 6 0.08 g; P < 0.0001) not observed in WT controls ( Figure  2e). Liver weight remained unchanged (Supporting Information Figure S2D). Histological examination of the eWAT revealed smaller, more densely packed adipocytes within the miR-34a 2/2 chow tissues, compared with WT ( Figure 2F, G). When stressed with a HFD, the eWAT from miR-34a 2/2 mice showed a significant increase in cell area (P 5 0.0030) and decrease in cell number per field (P 5 0.0013) not seen in WT, bringing them to WT levels.
To examine alterations in glucose homeostasis, we measured monthly fasting blood glucose and performed a GTT at week 24. However, we did not observe any changes between HFD-fed WT and miR-34a 2/2 mice, both showing hyperglycemia from week 12 (P < 0.001) (Supporting Information Figure S2E) and similar glucose handling (Supporting Information Figure S2F). This was further compounded by no change in fasting serum insulin or resistin between HFD groups at week 24 (Supporting Information Figure  S2F). Additionally, there was no change in fasting serum cholesterol Obesity miR-34a 2/2 Mice are Susceptible to Obesity Lavery et al. and triglyceride measurements between WT and miR-34a 2/2 groups on either chow or HFD (Supporting Information Table S1).
eWAT macrophages from HFD-fed miR-34a 2/2 mice expressed an F4/80 high phenotype, with an increase in type 2 cytokines Given that miR-34a transcripts were increased in WT BMDMs stimulated with TNF-a, they may contribute to the in vivo metabolic profile. Therefore, using FACS we gated on the larger, more granular F4/80 1 cells from WT and 34a 2/2 eWAT to identify macrophages, after 24 weeks on chow or HFD ( Figure 4A). These cells were CD45 1 , CD11b 1 , MHCII 1 , and CD86 1 . Further analysis revealed a population of F4/80 high cells within the miR-34a 2/2 HFD group not observed in WT ( Figure 4B). We observed a significant increase in F4/80 surface expression on miR-34a 2/2 cells, when fed HFD (P 5 0.0497), and a lower F4/80 1 macrophage content within the eWAT of miR-34a 2/2 chow mice, compared with WT chow (P 5 0.0078) ( Figure 4C). There were no changes in the other macrophage M1/M2 surface markers examined (Supporting Information Figure S4A), or M1/M2 genes within the eWAT, except a basal increase in miR-34a 2/2 chow eWAT Nos2 (P 5 0.0500) and decreased HFD-fed miR-34a 2/2 eWAT Retnla (P 5 0.0006) transcripts, over WT diet controls (Supporting Information Figure S4B). To check whether an increase in cytosolic lipid could contribute to the F4/80 high macrophage phenotype in miR-34a 2/2 HFD eWAT, we back-gated on this population for side-scatter, a measurement of internal complexity, but observed no difference in side-scatter MFI between any of the groups (Supporting Information Figure S4C).

In vitro miR-34a 2/2 adipocytes showed changes in lipid content
To examine the function of miR-34a in adipose tissue, we differentiated primary white and brown adipocytes in vitro. We observed significant upregulation of miR-34a transcripts (P 5 0.0330) by day 8 of differentiation in WT white adipocyte precursors ( Figure 5A), with upregulation of miR-34b* and 34c* at day 4 (P < 0.05), which decreased by day 8 (Supporting Information Figure S6A). Therefore, we examined differentiation genes in WT and miR-34a 2/2 white adipocytes. Examination of Cebpa, Ppar-c, and Fabp4 over the 8-day differentiation revealed only a small increase in Cebpa  Figure 5F).
Surprisingly, miR-34a 2/2 white adipocytes had increased supernatant leptin at day 0 (P 5 0.0458), with lower levels at day 6 (P 5 0.0009) of differentiation, compared with WT ( Figure 5G). Furthermore, supernatant insulin was decreased at day 2 (P 5 0.0041) of    miR-34a 2/2 white adipocyte differentiation, compared with WT ( Figure 5H). We observed no change in resistin (Supporting Information Figure S6B). In day 8 brown adipocytes, we did not observe any changes in leptin, insulin, or resistin (Supporting Information Figure S6C). Additionally, obesity proteome array analysis of day 8 supernatants from white and brown adipocyte cultures did not reveal any other cytokine or adipokine differences (Supporting Information Figure S7).  To further explore the metabolic changes in miR-34a 2/2 iBAT, we examined miR-34a expression over in vitro differentiation of WT brown adipocyte precursors, but found no change in miR-34a ( Figure 6A). However, we observed increases in miR-34b* at day 4 (P 5 0.0305) and both miR-34b* and 34c* at day 6 (P < 0.05) and day 8 (P < 0.01) of differentiation ( Figure 6B).
WT and miR-34a 2/2 day 8 brown adipocytes were stimulated with noradrenaline to examine the effect of b-adrenergic stimulation on Ucp1 and Pgc1a transcripts. However, we observed no difference in Ucp1 and Pgc1a transcript expression ( Figure 6C, D). Additionally, we quantified cellular mitochondrial content by FACS but found no difference ( Figure 6E, F). However, FACS analysis identified an increase in the percentage of noradrenaline-stimulated, miR-34a 2/2 lipid-positive brown adipocytes (P 5 0.0445) ( Figure 6G), with no difference in cellular lipid content ( Figure 6H).

Discussion
The regulation of chronic, low-grade inflammation of adipose tissue during obesity is poorly understood and this study aimed to address the role of miR-34a in this process. Our findings showed that miR-34a 2/2 mice were heavier than WT controls and gained more weight when challenged with a HFD, suggesting they were more susceptible to weight gain. miR-34a 2/2 eWAT showed smaller, more numerous adipocytes, with a lower tissue mass on chow that increased to WT levels on HFD. It is surprising that there was no difference in these characteristics between HFD-fed miR-34a 2/2 and WT mice, despite a clear difference in body weight. With no change in liver weight, this could suggest increases in other WAT depots (e.g., scWAT), BAT, or muscle tissue. One gene that was increased in all the miR-34a 2/2 tissues examined was Pgc1a, suggesting that it could have broad regulation by miR-34a. During preparation of this article, Fu et al. showed that miR-34a could increase acetylation of PGC1a, decreasing its activation, by inhibiting SIRT1 activity and FGF21 signaling in HFD-fed murine eWAT and 3T3L1 adipocytes, promoting adipose browning (19). Correlating with our results, the authors observed increased transcripts of Pgc1a in eWAT and BAT after lentiviral-mediated inhibition of miR-34a in WAT and BAT of male, HFD-fed BALB/c and C57BL/6 mice. However, they also observed a reduction in body weight, and visceral (epididymal and peritoneal) and scWAT, with improved glucose tolerance. Earlier studies have also shown anti-miR-34a administration to HFD-fed mice improved systemic glucose and insulin tolerance, and liver phenotype (20,21). All these studies used acute miR-34a inhibition in established murine obesity, which is milder than the 24-week HFD feeding used here. These differences with our model are likely due to compensatory effects of using a fullbody KO. It seems paradoxical that miR-34a ablation causes an upregulation in Pgc1a transcripts, but at the same time promotes weight gain. However, full-body Pgc1a KO mice were shown to lose weight, rather than gain weight as expected, and skeletal muscle-specific murine Pgc1a overexpression (MPGC-1a TG) caused muscle adipose accumulation, atrophy, and decreased ATP synthesis and content, indicative of mitochondrial respiratory uncoupling (22,23).
Respiratory uncoupling by UCPs has been linked to increased FA oxidation (24). It may seem counterproductive that in miR-34a 2/2 eWAT we saw basal increases in genes involved in lipid uptake: Cd36, de novo lipogenesis: Fasn, and cholesterol biosynthesis: Hmgcr, which caused increases in adipocyte total and free cholesterol, associated with omental adipocyte hypertrophy in primates (25)(26)(27). However, MPGC-1a TGs have increased expression of Cd36 and Fasn, with Fasn expression driven by PGC1a's increase in Lxra expression, and increased intracellular lipid (28). The authors suggest that this response "refuels" the cell's lipid stores for continued FA oxidation. Therefore, there may be an imbalance in lipolysis/FA oxidation vs. lipid uptake/lipogenesis, pushing metabolism to the latter in adipocytes of miR-34a 2/2 mice, due to chronic PGC1a upregulation, which predisposes these mice to weight gain when stressed with a HFD. Additionally, this would explain why miR-34a 2/2 mice are heavier before HFD feeding. In fact, when MPGC-1a TGs were fed HFD they developed insulin resistance and were not protected from diet-induced obesity, showing increased lipid uptake and lipogenesis genes in skeletal muscle, including Cd36, and no change in resting serum insulin, triglycerides, cholesterol, glucose, and FFAs (29). The increased lipid content observed in mature in vitro miR-34a 2/2 white adipocytes supports this for RT-qPCR, represented as 1/DCt. Statistics were calculated using a one-way ANOVA, with Bonferroni's multiple comparisons post-test, or an unpaired Student's t-test for FACS measurements. *P < 0.05, **P < 0.01, ***P < 0.001. (I) A diagrammatic representation of the theoretical mechanism in miR-34a 2/2 (KO) mice, predisposing them to diet-induced obesity. The imbalance in WAT metabolic genes caused by chronic overexpression of peroxisome proliferator-activated receptor-y coactivator (PGC)-1a and mitochondrial dysfunction promotes lipid uptake/storage and weight gain when stressed with a high-fat diet (HFD). As the tissue increases in size there is an increase in inflammatory cytokines, such as tumor necrosis factor (TNF)-a, which polarize macrophages to an M1 phenotype. However, the KO macrophages are unresponsive to TNF-a induction of CXCL1 and possibly other pro-M1 genes, through overexpression of PGC1a inhibiting the NF-jB subunit p65, suggesting an M2 phenotype. This inhibits the recruitment of neutrophils which could promote an M1 phenotype and inhibit lipogenic processes. The M2 phenotype is further promoted by interleukin (IL)-5, which can induce eosinophils to produce pro-M2 IL-4. IL-10 likely produced by macrophages can then stimulate adipocytes to be more insulin sensitive and dampen inflammatory processes, reducing pro-obesity processes. Inhibition of key thermogenic genes in BAT predisposes the mice to obesity development, with overcompensation by the induction of PGC1a during HFD feeding. metabolic imbalance. Additionally, the increase in miR-34a in WT adipocytes over the time course of differentiation suggests that it may regulate these processes in mature adipocytes, due to the lack of change in differentiation makers in miR-34a 2/2 white adipocytes. The decreases in Ppar-c, Fasn, and Acc-a eWAT transcripts upon HFD feeding in miR-34a 2/2 mice, could be evidence of the balance being shifted back to FA oxidation to regain lipid homeostasis, by other regulatory systems. This could be assisted by the upregulation of Pgc-1a mRNA in HFD-fed miR-34a 2/2 BAT, which could compensate for the reduced c/ebpa and b mRNA. Both c/EBPa and b transactivate UCP1, and c/EBPb is essential for brown adipocyte differentiation and thermoregulation (30,31). Unfortunately, we did not see these differences in vitro with b-adrenergic stimulation, but other stimuli could be involved in vivo. Furthermore, PGC1a has been shown to be a predicted target of miR-34a in humans, but not mice, using miRWalk (32). Therefore, the in vivo phenotype of miR-34a 2/2 mice could be attributed to PGC1a being a direct target of miR-34a or indirectly through SIRT1/FGF21.
Our study showed that miR-34a 2/2 ATMs have an altered phenotype, and deficiency in inflammatory TNF-a responses in vitro. Flow cytometric analysis identified F4/80 high CD45 1 MHCII 1 CD11b 1 CD86 1 macrophages within eWAT of HFD-fed miR-34a 2/2 mice and a reduced F4/80 1 macrophage population in chow-fed miR-34a 2/2 mice. F4/80 is a glycoprotein specific for murine macrophages, with expression increasing as macrophages mature, but its specific function remains unknown. However, Lin et al. showed that F4/80 has a role in generating CD8 1 regulatory T cells (33). F4/80 high ATMs have previously been observed in murine WAT, but have been shown to be in equal proportions F4/80 low :F4/80 high in obese db/db WAT and greater proportions of F4/80 high ATMs in diet-induced obese WAT (18,34). Bassaganya-Riera et al. showed that these F4/80 high ATMs expressed higher levels of MHCII, CX3CR1, CD11c, CCR2, and PPAR-c in db/ db mice, with increased intracellular IL-6, TNF-a, CCL2, and IL-10, which increased further upon LPS-stimulation (18). These reports support our observations of increased IL-10 mRNA in HFD-fed miR-34a 2/2 eWAT, correlating with the ATM F4/80 high phenotype. The increase in IL-10 suggests an M2-like phenotype. Corollary, the in vitro data showing ablated upregulation of CXCL1 in TNF-astimulated miR-34a 2/2 BMDMs, suggests a deficiency in TNF-a signaling. Additionally, the upregulation of miR-34a transcripts in TNFa-stimulated BMDMs, supports the possibility of miR-34a regulating the TNF-a pathway. This has been previously suggested by Jiang et al. with contradictory results using LPS-stimulated RAW264.7 macrophages, showing miR-34a inhibits TNF-a expression by targeting the NF-jB p50-subunit activator NOTCH1 (12). However, we showed that Pgc1a was upregulated in unstimulated miR-34a 2/2 in vitro macrophages, which others have shown represses TNF-a-induced proinflammatory cytokine production by reducing RelA/p65 phosphorylation, inhibiting NF-jB signalling in murine skeletal muscle (35). Paradoxically, Bassaganya-Riera et al. suggested that F4/80 high macrophages are M1-like, despite expressing high levels of IL-10 and PPAR-c. Furthermore, we observed a reduction in the M2 marker Retnla in HFD-fed miR-34a 2/2 eWAT and an increase in the M1 marker Nos2 in chow-fed miR-34a 2/2 eWAT. Therefore, it is unclear whether miR-34a 2/2 macrophages have a pro-or anti-inflammatory phenotype, but their phenotype could be altered by dysregulation of PGC1a.
Systemically, we observed characteristics of a type 2 immune response in miR-34a 2/2 mice with increased serum IL-5 protein during HFD feeding and a reduction in splenic neutrophils observed in both miR-34a 2/2 groups. Interestingly, neutrophil-specific elastase KO mice have increased M2-like ATMs and increased transcripts of Acc-a and Fasn within WAT (17). However, we did not see the same decrease in eWAT neutrophils. Importantly, a type 2 immune response has been reported to promote increased insulin sensitivity, glucose uptake, and anabolism in white adipocytes, but an overall protective effect from obesity and T2D (16,36,37). Therefore, a type 2 immune response in HFD-fed miR-34a 2/2 mice may be initiated as a compensatory effect during overnutrition, to regain metabolic homeostasis. A study limitation is that we did not examine miR-34a expression in BMDMs from HFD-fed mice or TNF-astimulated adipocytes.
In summary, this study has demonstrated that miR-34a regulates the development of diet-induced obesity in mice, with miR-34a 2/2 mice showing a susceptibility to weight gain, likely through dysregulation of PGC1a ( Figure 6I). It is clear that the role of miR-34a in metabolism is complex and likely to vary depending on context of treatment and model, with potentially contrasting roles in different cell types. miRNAs' fine-tuning of gene expression results in subtle changes in protein synthesis, making it complex to follow up in vivo observations with complete mechanistic data in vitro, as seen here. However, this study has highlighted that chronic miR-34a inhibition, as a treatment for obesity, may have undesirable metabolic and immunoregulatory consequences. Further research is required to fully elucidate the metabolic and inflammatory pathways regulated by miR-34a.O