Krüppel-like Factor 3 (KLF3/BKLF) Is Required for Widespread Repression of the Inflammatory Modulator Galectin-3 (Lgals3)*

The Lgals3 gene encodes a multifunctional β-galactoside-binding protein, galectin-3. Galectin-3 has been implicated in a broad range of biological processes from chemotaxis and inflammation to fibrosis and apoptosis. The role of galectin-3 as a modulator of inflammation has been studied intensively, and recent evidence suggests that it may serve as a protective factor in obesity and other metabolic disorders. Despite considerable interest in galectin-3, little is known about its physiological regulation at the transcriptional level. Here, using knockout mice, chromatin immunoprecipitations, and cellular and molecular analyses, we show that the zinc finger transcription factor Krüppel-like factor 3 (KLF3) directly represses galectin-3 transcription. We find that galectin-3 is broadly up-regulated in KLF3-deficient mouse tissues, that KLF3 occupies regulatory regions of the Lgals3 gene, and that KLF3 directly binds its cognate elements (CACCC boxes) in the galectin-3 promoter and represses its activation in cellular assays. We also provide mechanistic insights into the regulation of Lgals3, demonstrating that C-terminal binding protein (CtBP) is required to drive optimal KLF3-mediated silencing. These findings help to enhance our understanding of how expression of the inflammatory modulator galectin-3 is controlled, opening up avenues for potential therapeutic interventions in the future.

Lectin, galactoside-binding, soluble 3 (Lgals3) encodes galectin-3, a 35-kDa protein noted for its diverse molecular roles in pre-mRNA splicing (1), macrophage activation (2), tumorigenesis (3), apoptosis (4), and other important cellular processes. Recent studies at the level of the whole organism have shown galectin-3 to be important in heart disease and fibrosis (5,6), and it is now used routinely as a prognostic biomarker in heart failure patients (7). Furthermore, galectin-3 is an essential regulator of inflammation in metabolic tissues, able to protect pancreatic ␤ cells from interleukin 1␤ cytotoxicity (8) and to neu-tralize inflammatory factors known as advanced glycation end products (9). However, despite our extensive knowledge of its significance in various biological processes and disease states, little is known about how the expression of Lgals3 is controlled. Developing a more thorough understanding of how Lgals3 is regulated should provide important information on the biological pathways upon which it operates and may ultimately provide strategies for therapeutic interventions.
While studying genetically modified mice with a mutation in the gene encoding the zinc finger transcription factor Krüppellike factor 3 (KLF3/BKLF), we noted that the Lgals3 gene was consistently up-regulated. KLF3 is one of 18 members of the Krüppel-like factor (KLF) 4 family of DNA-binding proteins, responsible for controlling gene expression at the transcriptional level in a diverse range of biological settings. All KLF family members possess three evolutionarily conserved Cys 2 -His 2 zinc fingers that facilitate binding to CACCC boxes and related GC-rich motifs in regulatory regions of DNA (10,11). KLFs are able to function as potent activators or repressors of transcription through interaction with protein co-factors via their highly variable N-terminal functional domains. KLF4, for instance, has been shown to both silence and activate gene expression depending on promoter availability and the presence of specific co-factors (12,13). Because of their inherent DNA-binding homology, KLFs are known to share target genes and operate within transcriptional networks (14). For instance, KLF feedback repression has been observed where KLF3 acts to silence a series of KLF1 target genes during erythropoiesis (15). KLF3 itself was first identified in a screen for factors in erythroid cells with homology to the DNA-binding domain of the founding family member KLF1 (16) but is now known to show a broad tissue expression profile (11). KLF3 is notable for being primarily a repressor of transcription. Upon binding control regions of its target genes, KLF3 recruits the co-repressor C-terminal binding protein (CtBP) via the N-terminal functional domain (17), which subsequently facilitates the assembly of a potent repressor complex of histone-modifying enzymes and other co-repressors to silence transcription (18,19). Post-translational modifications of KLF3 by phosphorylation and sumoylation have also been implicated in enhancing its repressive activity (20,21). KLF3 has been implicated in a host of biological roles spanning from the regulation of B cell development (22) and erythropoiesis (15) to adipogenesis (23,24) and, more recently, in heart function (25).
Here we set out to assess whether Lgals3 is a bona fide target of KLF3-mediated repression at the transcriptional level. We found that Lgals3 is consistently up-regulated in mouse-derived cell lines lacking KLF3, prompting a more rigorous examination of Lgals3 expression across KLF3-deficient tissues. Lgals3 was found to be highly derepressed in both subcutaneous and visceral white adipose depots as well as in the heart and pancreas of Klf3 Ϫ/Ϫ mice. Given its importance in fat homeostasis, closer examination of Lgals3 deregulation in subcutaneous adipose tissue was undertaken by immunohistochemistry, illustrating a distinct up-regulation in both the adipocytes and stromal vascular compartment in the absence of KLF3. The ability of KLF3 to bind to regulatory elements within the Lgals3 gene was confirmed by electrophoretic mobility shift assays and chromatin immunoprecipitation experiments, suggesting that KLF3 occupies Lgals3 regulatory regions in vivo. We also demonstrated that KLF1-driven activation of Lgals3 is suppressed with the addition of KLF3 and that CtBP participates in the mediation of repression. Together, these results represent the first evidence that Lgals3 is a direct target of KLF3 binding and repression.
Interestingly, mice lacking Lgals3 display increased fat mass and systemic inflammation (26 -28), the inverse of the phenotype seen in Klf3 Ϫ/Ϫ mice, which are obesity-resistant with reduced adiposity (23,24). The observation that KLF3 directly represses the galectin-3 gene and that the two proteins play what are essentially opposing roles in metabolic homeostasis is consistent with contributing to a regulatory circuit that controls both inflammation and metabolism.

Results
Lgals3 Is Up-regulated in the Absence of KLF3-In studies of the role of KLF3 in hematopoiesis and red blood cell development, microarrays performed on Ter119 ϩ fetal liver cells lacking KLF3 revealed that the Lgals3 gene was consistently up-regulated in knockout animals (15). Because of the importance of galectin-3 in a number of biological settings, we undertook a fuller analysis of whether the expression of Lgals3 was altered in a range of mouse tissues in the absence of KLF3. Lgals3 mRNA levels were assessed in cultured murine embryonic fibroblasts (MEFs) as well as a series of primary tissues from wild-type and Klf3 Ϫ/Ϫ mice by quantitative real-time PCR. In primary and immortalized Klf3 Ϫ/Ϫ MEFs, Lgals3 mRNA was up-regulated 4.7-and 4.3-fold, respectively, compared with wild-type expression (Fig. 1A). Importantly, Lgals3 levels were also found to be elevated in a number of primary tissues dissected from KLF3-deficient mice (Fig. 1B). Derepression was most evident in Klf3 Ϫ/Ϫ subcutaneous (6.7-fold) and epididymal (3.3-fold) white adipose depots and in the heart (6.6-fold) and pancreas (4.2-fold).
Following the demonstration that Lgals3 is derepressed in Klf3 Ϫ/Ϫ tissues at the mRNA level, we next sought to determine whether this up-regulation was reflected at the protein level. Whole cell protein extracts were prepared from wild-type and Klf3 Ϫ/Ϫ fat depots and spleens, and the expression of galectin-3 protein was assessed by Western blotting (Fig. 1, C-F). Significant elevation of galectin-3 protein was evident in Klf3 Ϫ/Ϫ white adipose depots, with up-regulation being 6.7-fold in subcutaneous white adipose tissue (scWAT) (Fig. 1D) and 7.6-fold in epididymal white adipose tissue (epiWAT) (Fig. 1E). Expression was also modestly but significantly up-regulated in the brown adipose tissue (BAT) (Fig. 1C) and spleen (Fig. 1F) of animals lacking KLF3 (3.5-and 2.4-fold, respectively). These observations correlate well with the degree of up-regulation observed at the transcript level and, together, point to widespread derepression of Lgals3 in the absence of KLF3.
As galectin-3 expression was found to be markedly derepressed in Klf3 Ϫ/Ϫ mouse adipose tissue at the mRNA and protein levels, we decided to determine in which specific cell types this was occurring using independent methodologies. We performed immunofluorescent staining of galectin-3 in scWAT ( Fig. 2A). In accordance with mRNA and protein expression results, galectin-3 was found to be considerably up-regulated in Klf3 Ϫ/Ϫ scWAT, as calculated by corrected total fluorescence (Fig. 2B). Klf3 Ϫ/Ϫ scWAT in pairs 1 and 2 registered 9-and 10.5-fold up-regulation of galectin-3 expression, respectively. In addition to the examination of cellular galectin-3 levels, an ex vivo analysis of galectin-3 secretion from epiWAT explants was conducted by enzyme-linked immunosorbent assay, revealing 2.7-fold higher levels of secretion from Klf3 Ϫ/Ϫ fat (Fig. 2C).
Closer inspection of the scWAT by immunofluorescence showed increased galectin-3 staining in both adipocytes and cells of the stromal vascular compartment (the non-adipocytes) in scWAT lacking KLF3 (Fig. 2D). To better quantify the differential expression of Lgals3 in these subfractions of the scWAT, quantitative real-time PCR was conducted, confirming that Lgals3 mRNA is significantly up-regulated in both the adipocytes (5-fold) and stromal vascular cells (3.8-fold) of Klf3 Ϫ/Ϫ scWAT (Fig. 2E). Given that galectin-3 is important in macrophage activation (2) and that macrophages are an integral cell type of the stromal vascular compartment, we assessed Lgals3 transcript levels in bone marrow-derived macrophages (Fig.  2F). We found that Lgals3 expression is modestly but significantly elevated in the absence of KLF3 in these cells, showing a 1.4-fold up-regulation. Taken together, these results demonstrate that galectin-3 expression is significantly derepressed at both the mRNA and protein levels in several different cell types and tissues in mice lacking KLF3.
KLF3 Binds the Lgals3 Promoter-To investigate whether KLF3 regulates Lgals3 expression directly, the ability of KLF3 to bind to the Lgals3 locus was assessed. We first explored the mechanism by which KLF3 might directly repress Lgals3 by focusing on the proximal promoter. Indeed, KLF3 has been shown in several previous cases to silence gene expression by directly binding to promoter elements through recognition of CACCC motifs and related GC-rich regions (15,23,29). We examined the sequence between Ϫ60 and ϩ20 bp relative to the Lgals3 transcription start site and observed several KLF3 consensus binding sites (16, 29) (shown as boxes in Fig. 3A). To assess whether KLF3 could bind these sites in vitro, we prepared DNA probes corresponding to these motifs and evaluated direct protein binding by EMSA (Fig. 3B). KLF3 bound to each probe with the addition of an anti-KLF3 antibody, resulting in a supershift confirming the identity of probe-bound KLF3. The probes centered around Ϫ5 bp (Gal-3A) and Ϫ46 bp (Gal-3C) from the transcription start site displayed the most robust binding, suggesting that concurrent binding and dimerization of two KLF3 molecules at the promoter may occur given the spacing between these motifs, as has been postulated previously for other KLF3 target genes (14,23). The middle site at Ϫ20 bp (Gal-3B) showed weaker binding.
Having established that KLF3 binds to motifs in the Lgals3 proximal promoter, we next analyzed ChIP-seq data of genome-wide KLF3 binding in MEFs rescued with a tagged KLF3 cDNA transgene, KLF3-V5 (29), and found significant enrichment in and immediately upstream of the Lgals3 gene (GEO accession no. GSE44748). Three significant peak clusters were identified: 19 kb upstream of the transcription start site Lgals3 expression was normalized with 18S rRNA levels and normalized again to the condition with the lowest mean Lgals3/18S expression, which was set to 1 (n ϭ 3 for A and n ϭ 4 -5 for B). C-F, Western blots showing galectin-3 protein expression in WT and Klf3 Ϫ/Ϫ BAT (C), scWAT (D), epiWAT (E), and spleen (F) with densitometry analysis (n ϭ 3-4). 15 g of total protein was loaded per lane, alongside a Rainbow molecular weight marker (Amersham Biosciences) used as a protein size standard (Size Std.). Negative controls (nuclear extract from untransfected COS cells) and positive controls (nuclear extracts from COS cells expressing Gal-3) were included to show the expected migration pattern and test the specificity of the antibody for Gal-3. Error bars represent mean Ϯ S.E., and Student's t tests were conducted to determine significance. *, p Ͻ 0.05; **, p Ͻ 0.01.
(site i), at the proximal promoter (site ii), and in the first intron (site iii) (Fig. 3C). These three prominent sites represent potential regulatory elements that are both accessible (as denoted by DNase-seq) and/or marked by active histone signatures (histone 3 lysine 4 trimethylation and histone 3 lysine 27 acetylation) in fibroblasts.
We also validated the ChIP-seq result with ChIP experiments followed by quantitative real-time PCR. Chromatin from wildtype and Klf3 Ϫ/Ϫ MEFs was immunoprecipitated with an anti-KLF3 antibody and subjected to amplification using primers targeted to the Ϫ19 kb promoter and intron 1 ChIP-Seq peaks (sites i-iii, respectively) from Fig. 3C. Previously confirmed KLF3 binding sites at the Fam132a promoter and the Klf8 promoter 1a were used as positive control loci, whereas primers targeting regions 30 kb downstream and 4.5 kb upstream of Klf8 promoter 1a served as negative control regions A and B, respectively (14,15). KLF3 enrichment at sites i, ii, and iii of Lgals3 in WT cells was increased 4.5-, 10.7-, and 7.3-fold, respectively, FIGURE 2. Galectin-3 levels are elevated in Klf3 ؊/؊ adipose tissue. A, immunofluorescent staining of scWAT to assess galectin-3 levels, showing two representative WT and Klf3 Ϫ/Ϫ pairs. DAPI and wheat germ agglutinin (WGA) were used to stain the nuclei and cell membranes, respectively. B, corrected total galectin-3 fluorescence using WT and Klf3 Ϫ/Ϫ images from the Gal-3 column in A. C, levels of galectin-3 secreted after 2 h from WT and Klf3 Ϫ/Ϫ epiWAT explants were measured using ELISA (n ϭ 7 for WT and 5 for Klf3 Ϫ/Ϫ ). D, the merged images from A were further magnified to assess galectin-3 up-regulation in adipocytes and the stromal vascular fraction of Klf3 Ϫ/Ϫ scWAT. Up-regulation in both compartments of the scWAT was confirmed by real-time quantitative PCR analysis of Lgals3 mRNA levels (n ϭ 3) E, Lgals3 levels were also assessed in cultured bone marrow-derived macrophages (n ϭ 4). SVF, stromal vascular fraction. F, Lgals3 expression was normalized with 18S expression levels and normalized again to the condition with the lowest mean Lgals3/18S expression, which was set to 1. For C, E, and F, error bars represent mean Ϯ S.E., and Student's t tests were conducted to determine significance. *, p Ͻ 0.05; ***, p Ͻ 0.001. compared with the Klf8 negative control B region (Fig. 3D). Negligible KLF3 enrichment was evident at any of the loci interrogated in Klf3 Ϫ/Ϫ MEFs, confirming the specificity of the anti-KLF3 immunoprecipitation.
We extended this analysis to investigate the binding in bone marrow-derived macrophages. These cells were chosen as they are important in metabolism, and Lgals3 is known to be a prominent macrophage marker (2, 30). Chromatin from WT and  (ϩ1) showing three KLF3 consensus binding sites (boxed) and their corresponding EMSA probes (Gal-3A, Gal-3B, and Gal-3C; black bars). The Lgals3 sequence was derived from the mm9 Mus musculus genome (NCBI reference sequence NM_001145953). B, nuclear extracts were prepared from COS-7 cells transfected with 5 g of pMT3-Klf3. These were used in EMSA to assess KLF3 binding of 32 P-radiolabeled probes Gal-3A, Gal-3B, and Gal-3C, each corresponding to KLF3 consensus motifs in the Lgals3 proximal promoter from A. The addition of anti-KLF3 antibody was used to confirm the identity of the protein bound to the probes by supershift. C, KLF3 binding at Lgals3 in MEFs from previously published KLF3-V5 ChIP-seq data available from GEO (accession# GSE44748) (29). KLF3 enrichment peaks are aligned with histone 3 lysine 4 trimethylation (accession no. GSM769029) and histone 3 lysine 27 acetylation (accession no. GSM1000139). ChIP-seq data from MEFs and a DNaseI hypersensitivity (accession no. GSM1014199) dataset were produced from murine fibroblasts (48 -50). Significant KLF3 binding peaks at Lgals3 are denoted by i, ii, and iii. D and E, in vivo occupancy of KLF3 at these sites (i, ii, and iii) in Lgals3 was assessed in WT and Klf3 Ϫ/Ϫ MEFs (D) and bone marrow-derived macrophages (E) (n ϭ 3-4 for WT and 2-3 for Klf3 Ϫ/Ϫ ). Sites i, ii, and iii from C were interrogated by quantitative PCR following chromatin immunoprecipitation with Pierce anti-KLF3 antibody (PA5-18030) or normal goat IgG. Fam132a and Klf8 promoter 1a were used as positive control sites and Klf8 negative controls A and B as negative control loci, as described previously (14,15,23). Error bars represent the mean Ϯ S.E. Student's t tests were used to determine significance. *, p Ͻ 0.05; **, p Ͻ 0.01; ***p Ͻ 0.001; WT versus WT Klf8-negative control B (D) or Klf8 negative control A (E). #, p Ͻ 0.05, WT versus Klf3 Ϫ/Ϫ (D and E).
Klf3 Ϫ/Ϫ macrophages was immunoprecipitated with anti-KLF3 antibody and then amplified by quantitative real-time PCR primers corresponding to the Lgals3 sites i, ii, and iii from Fig.  3C to assess KLF3 occupancy (Fig. 3E). As in Fig. 3D, Fam132a and Klf8 promoter 1a were used as positive control loci and Klf8 negative control regions A and B as sites where KLF3 does not bind (23). The results in macrophages reflected those seen in MEFs, with strong enrichment evident at the Ϫ19 kb site (3.9fold versus Klf8 negative control A), the promoter (3-fold), and intron 1 (3.3-fold) of Lgals3, confirming KLF3 occupancy in vivo and indicating that KLF3 resides at Lgals3 regulatory regions in multiple cell types, including primary cells.
KLF3 Repression of Lgals3 Involves Both CtBP-dependent and -independent Silencing-With the knowledge that KLF3 is able to bind consensus motifs in the Lgals3 promoter and occupies this element in vivo, cellular reporter assays were undertaken in SL2 cells to determine whether KLF3 is able to directly repress Lgals3 expression. Because of its reported role as a co-repressor for KLF3 (17), we investigated whether CtBP may be necessary for optimal silencing of Lgals3 by testing the repressive capacity of a KLF3 mutant unable to bind CtBP (KLF3-⌬DL) (29). SL2 cells are typically used for analyzing the functional roles of KLF proteins because of the fact that, unlike mammalian cells, they lack endogenous CACCC-binding proteins that may interfere with assays (31). To examine repression, it was first necessary to activate the Lgals3 promoter, and this was done by providing KLF1, a closely related activator of transcription that regulates a subset of KLF3 target genes and has been suggested to drive Lgals3 expression (15).
We showed that increasing the dosage of KLF1 was sufficient to activate the reporter construct (Fig. 4A), and the 250-ng dosage was chosen as an optimal amount to drive expression and to determine what effect increasing doses of KLF3 has in competitive assays. In these, 0-, 5-, 10-, and 25-ng amounts of KLF3 or KLF3-⌬DL were added to cells containing the Lgals3 promoter reporter construct. We have previously observed that the ⌬DL mutation does not adversely affect the expression of KLF3 in SL2 cells (17). Addition of 5 ng of KLF3 was sufficient to significantly reduce Lgals3 activation by 4-fold. 5 ng of KLF3-⌬DL also drove significant repression, but only 2.5-fold, suggesting the need for KLF3 to bind its co-repressor CtBP for effective silencing of Lgals3.
Given that the inability of KLF3 to bind CtBP resulted in impaired repression from our reporter assays, we sought to determine whether this affected Lgals3 expression in a series of rescued Klf3 Ϫ/Ϫ MEFs (Fig. 4B). Klf3 Ϫ/Ϫ cells rescued with Klf3-⌬DL displayed 2.5-fold higher expression of Lgals3 than cells rescued with wild type Klf3, confirming the observation from our reporter assays (Fig. 4A) that, without CtBP, KLF3 cannot optimally silence Lgals3. Furthermore, Lgals3 expression in Klf3 Ϫ/Ϫ cells rescued with the KLF3 DNA-binding domain only (Klf3-DBD) was 3.6-fold higher than those rescued with wild-type Klf3 and only modestly reduced compared with empty vector-rescued cells. Together, these results demon-

KLF3 Represses the Galectin-3 Gene
JULY 29, 2016 • VOLUME 291 • NUMBER 31 strate that KLF3 binds to and is able to directly repress Lgals3 expression and that elements in the KLF3 functional (non-DNA-binding) domain, including CtBP recruitment, are important for its capacity to silence target genes like Lgals3.

Discussion
Despite the importance of galectin-3 in a host of biological settings, little is known about how its gene is activated (32)(33)(34), and to our knowledge, there is no published work on the repression of Lgals3. Here we have shown that galectin-3 expression is up-regulated in the absence of KLF3, and we have demonstrated that KLF3 directly binds and represses the Lgals3 promoter in vivo. Furthermore, we have provided mechanistic insights into KLF3 repression of Lgals3. In reporter assays, a KLF3 mutant that is unable to bind the co-repressor CtBP showed a reduced ability to repress Lgals3. Analysis of the expression levels of Lgals3 in Klf3 Ϫ/Ϫ MEFs rescued with KLF3 or a KLF3 mutant unable to bind to CtBP also showed that KLF3 recruitment of CtBP is necessary for optimal Lgals3 repression. These two lines of evidence suggest that recruitment of the co-repressor CtBP is important for KLF3 repression of Lgals3 but that CtBP-independent mechanisms also exist. We also assessed the contribution of the KLF3 functional domain to repression in Klf3 Ϫ/Ϫ MEF rescue experiments. KLF3 DNA-binding domain only showed only a modest ability to rescue Lgals3 repression when introduced into Klf3 Ϫ/Ϫ MEFs, suggesting that the functional domain (where CtBP binds) is important and also that direct competition for DNA binding to the Lgals3 promoter between KLF3 and the activating KLF1 is not likely to be a major feature of the mechanism of repression.
Galectin-3 has been identified as an important regulator of inflammation in metabolic tissues (27). Its deficiency in mice is associated with increased adiposity, systemic inflammation, and an accumulation of inflammatory cells in metabolic tissues (26,28). This phenotype poses a striking contrast to that seen in mice lacking KLF3, which display reduced fat mass and are protected from diet-induced obesity and glucose intolerance (23). It has been proposed that the phenotype of KLF3-deficient mice may reflect the derepression of adipose tissue-derived hormones that regulate metabolism. For instance, we have reported that KLF3 represses the recently characterized adipokine adipolin (also known as Fam132a/CTRP12/C1qdc2) and suggested that the abundance of adipolin may in part explain the lean phenotype of KLF3-deficient mice (23). Given that galectin-3 has also been proposed to have a role in metabolism and is also derepressed in KLF3-deficient animals, it is possible that its up-regulation may also contribute to the KLF3 knockout mouse phenotype.
It is well established that obesity is accompanied by an accumulation of inflammatory immune cells, particularly macrophages, in adipose and other metabolic tissues (35,36). Interestingly, galectin-3 appears to have both pro-and antiinflammatory roles depending on the cellular and pathophysiological context. On the one hand, inhibition of galectin-3 in apolipoprotein E-deficient mice improves atherosclerotic progression (37). However galectin-3 has also been shown to bind and neutralize metabolic compounds known as advanced gly-cation end products, which arise from excessive nutrient accumulation and contribute to inflammation in type 2 diabetes (38). Elimination of advanced glycation end products protects tissue from sustained macrophage-mediated inflammatory signaling and associated metabolic complications, so galectin-3 is also regarded as a protective factor. Elevated galectin-3 levels in adipose tissue lacking KLF3 may contribute to ameliorating the onset of chronic inflammation that might otherwise accompany obesity, insulin resistance, and glucose intolerance. It is notable that these metabolic complications do not develop in Klf3 Ϫ/Ϫ mice even when they are maintained on a high-fat diet (23).
Additionally, we observed significant galectin-3 up-regulation in the pancreas of Klf3 Ϫ/Ϫ mice, where galectin-3 has been shown to protect ␤ cells from the inflammatory effects of interleukin 1␤ (8). We also found that Lgals3 expression is elevated in Klf3 Ϫ/Ϫ heart tissue, which may be important in the context of current research efforts focused on galectin-3 in cardiac function (5,39). These findings necessitate further study to elucidate the role of KLF3 in the pancreas and heart, such as fibrotic progression and advanced glycation end product accumulation in Klf3 Ϫ/Ϫ mice-areas that are presently lacking understanding.
Immunohistochemical staining of adipose sections from wild-type and Klf3 Ϫ/Ϫ mice revealed that galectin-3 is highly up-regulated in both adipocytes and in cells of the stromal vascular compartment. Macrophages are an important component of this stromal fraction and are heavily implicated in the onset of chronic inflammation in obese adipose tissue (36). During this process, macrophages undergo a functional switch away from the alternatively activated state seen under normal conditions toward a more invasive and pro-inflammatory character, accompanied by changes in cellular metabolism (40). Interestingly, CtBP, which we have shown participates in KLF3mediated repression of galectin-3, is known to act as a metabolic "sensor" through binding NADH, enhancing its propensity as a co-repressor (41,42). This may be relevant to its role in silencing galectin-3 expression in macrophages undergoing metabolic remodeling toward a pro-inflammatory phenotype.
Galectin-3 is crucial in sustaining alternative activation in macrophages through an anti-inflammatory axis involving interleukin 4 and CD98 (2), protecting against the inflammatory shift that can contribute to insulin resistance and other metabolic complications. Up-regulation of galectin-3 in adipose tissue macrophages lacking KLF3 may convey anti-inflammatory benefits and, in part, account for the obesity-protected phenotype evident in Klf3 Ϫ/Ϫ mice.
Accordingly, we investigated whether galectin-3 was regulated by KLF3 in macrophages themselves. We did observe binding of the KLF3 protein to Lgals3 regulatory regions in ChIP experiments (Fig. 3E) and also observed some up-regulation of galectin-3 in Klf3 Ϫ/Ϫ macrophages cultured from bone marrow (Fig. 2F). However, the up-regulation in macrophages was less striking than in other tissues, such as white adipose tissue and the heart. It remains to be determined whether this is a consequence of the fact that galectin-3 is already perhaps close to maximally expressed in wild-type murine macrophages. Indeed, galectin-3 was once known as Mac-2 because of its high abundance as a macrophage surface marker (30). The full functional network controlling metabolism will involve both macrophages and other cells, including adipocytes, and further work will be required to determine the contributions of each cell type to the murine phenotypes.
In conclusion, we have demonstrated that KLF3 binds and represses Lgals3, regulating its expression in vivo. We have shown that optimal KLF3 repression of Lgals3 is dependent on the functional domain of KLF3 and that both CtBP-dependent and CtBP-independent mechanisms are involved. Identification of Lgals3 as a bona fide KLF3 target gene allows us to better understand how its expression is controlled. Given the well documented importance of galectin-3 in inflammation and other disease states, future considerations of how its expression is regulated may open up avenues for therapeutic intervention strategies.

Experimental Procedures
Mouse Generation and Genotyping-Generation of Klf3 Ϫ/Ϫ mice on an FVB/NJ background and their genotyping have been reported previously (14,24). Approval for the use of animals was from the University of Sydney Animal Care and Ethics Committee (protocol L02/7-2009/3/5054) and the University of New South Wales Animal Care and Ethics Committee (protocol 12/150A). Mice were weaned at 3 weeks and had ad libitum access to standard chow and water until sacrifice at 12 to 14 weeks of age.
Tissue Collection-12-to 14-week-old male mice were anesthetized with isoflurane and euthanized by cervical dislocation. Tissues were harvested promptly and washed with cold PBS and then frozen in RNAlater (Sigma). For fat depot collection, scWAT was harvested from the inguinal pads, visceral white adipose tissue from the epididymal depot (epiWAT), and BAT from the interscapular region. Femora and tibiae were kept on ice and immediately transferred to tissue culture for bone marrow lavage.
Collagenase Digestion of Fat-Inguinal scWAT was minced with scissors and then digested in 0.75 mg/ml type II collagenase (Sigma) for 45 min of shaking at 37°C. The cell slurry was diluted in Hanks' buffered salt solution (Invitrogen) and then filtered through a prewet 100-m mesh prior to separation of the stromal vascular fraction from adipocytes by centrifugation. Following red blood cell lysis in the stromal vascular fraction, both fractions were kept at Ϫ80°C in preparation for RNA extraction.
Cell Culture-The wild-type, Klf3 Ϫ/Ϫ , and rescued Klf3 Ϫ/Ϫ MEF cell lines used here have been described previously (29) and were cultured at 37°C with 5% CO 2 in Dulbecco's modified Eagle's medium supplemented with 10% v/v heat-inactivated fetal calf serum and 1% v/v penicillin, streptomycin, and glutamine solution. COS-7 cells were cultured under the same conditions as MEFs. To generate macrophages, femora and tibiae from wild-type and Klf3 Ϫ/Ϫ mice were flushed with cold phosphate-buffered saline, and bone marrow-derived macrophages (BMDM) were cultured over 7-10 days at 37°C with 5% CO 2 in medium conditioned with M-CSF as described previously (43).
Protein Extraction and Western Blotting-Whole cell extracts were prepared in radioimmune precipitation assay buffer (50 mM HEPES (pH 7.5), 500 mM LiCl, 1 mM EDTA, 1% Nonidet P-40, and 0.7% sodium deoxycholate) with protease inhibitors from 20-mg portions of epiWAT, scWAT, BAT, and spleen. Total protein levels were assessed using a bicinchoninic acid protein assay (ThermoFisher). For control lanes, COS-7 cells were transfected with pMT3-Lgals3 using FuGENE6 (Promega) or left untransfected and then harvested 48 h later for nuclear extraction as described previously (16). Western blotting was performed according to standard methods. In brief, 15 g of protein was separated by SDS-PAGE and transferred to a nitrocellulose membrane, which was blocked with 3.5% skim milk in 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, and 0.05% Tween 20 (TBST). Membranes were probed for galectin-3 and ␤-actin protein for 1 h at room temperature in TBST with antigalectin-3 (AF1197, R&D Systems) and anti-␤-actin (A1978, Sigma) antibodies. The Immobilon Western chemiluminescent HRP substrate system (Millipore Corp.) was used for detection. ImageJ was used for densitometry quantification of Western blotting bands, where three background areas were used to normalize galectin-3 and ␤-actin band intensity.
RNA Extraction and cDNA Synthesis-MEFs and bone marrow-derived macrophages were harvested at 95% confluency, and total RNA was extracted using the RNeasy mini kit (Qiagen). The DNA-free kit (Ambion) was used to eliminate genomic DNA contamination. Tissues were harvested and suspended in TRI reagent (Sigma) or QIAzol lysis reagent (Qiagen) for lipid-rich adipose and brain samples. A TissueLyser II (Qiagen) was used to disrupt and homogenize tissues before extraction of RNA using the RNeasy mini or lipid tissue mini kit (Qiagen) according to the instructions of the manufacturer. For RNA extraction from stromal vascular and adipocyte fractions of scWAT, the RNeasy Plus mini kit (Qiagen) was utilized because of lower initial sample quantity. Residual genomic DNA was eliminated by treatment with the RNase-free DNase kit (Qiagen) prior to cDNA synthesis. RNA was converted into cDNA with the SuperScript VILO cDNA synthesis kit (Invitrogen) for use in quantitative real-time PCR.
Immunohistochemistry-Inguinal scWAT was harvested from 12-to 14-week-old male wild-type and Klf3 Ϫ/Ϫ mice, and portions were fixed overnight in 4% buffered formalin prior to paraffin embedding. 8-m sections of tissue were rehydrated with xylene and graded ethanol before treatment with citrate buffer, a heat-induced antigen retriever. Following this, sections were incubated with 2% w/v skim milk to block nonspecific background staining. The sections were then incubated for 90 min with primary goat anti-galectin-3 (AF1197, R&D Systems). Subsequently, a 45-min incubation was undertaken with secondary donkey anti-goat IgG conjugated to Alexa Fluor 488 (A-11055, ThermoFisher) as well as Alexa Fluor 633-conjugated wheat germ agglutinin (W21404, ThermoFisher) before treatment with 0.1% v/v Sudan black B to reduce autofluorescence. Stained sections were rinsed, and a DAPI coverslip was applied prior to imaging. Images were captured at ϫ20 magnification using a ScanScope FL (Aperio Technologies), and scanned slides were visualized with ImageScope software (Aperio Technologies). ImageJ was used to calculate the corrected total fluorescence of galectin-3 using a method described previously (46) by subtracting the selected area size times the average background fluorescence from the integrated density. Three areas of interest and background regions were randomly selected per section.
ChIP-and DNase-seq Datasets-A ChIP-seq dataset for V5-tagged KLF3 produced from murine embryonic fibroblasts was used to assess genome-wide KLF3 binding and was obtained from GEO (accession no. GSE44748) (29). ENCODE datasets for genome-wide histone 3 lysine 4 trimethylation and histone 3 lysine 27 acetylation were produced from murine embryonic fibroblasts by the Ren laboratory at the Ludwig Institute for Cancer Research. Both were downloaded from GEO (accession nos. GSM769029 and GSM1000139, respectively) (48,49). A DNase-seq dataset produced from murine fibroblasts by the Stamatoyannopoulos laboratory at the University of Washington was also downloaded from GEO (accession no. GSM1014199) (48,50).
Statistical Analysis-Data are presented as mean Ϯ S.E. Significance was determined using Student's t tests, with p Ͻ 0.05 being taken as statistically significant.