A Novel Adipose-Specific Gene Deletion Model Demonstrates Potential Pitfalls of Existing Methods

Abstract

Adipose-specific gene deletion in mice is crucial in determining gene function in adipocyte homeostasis and the development of obesity. We noted 100% mortality when the Hdac3 gene was conditionally deleted using Fabp4-Cre mice, the most commonly used model of adipose-targeted Cre recombinase. However, this surprising result was not reproduced using other models of adipose targeting of Cre, including a novel Retn-Cre mouse. These findings underscore the need for caution when interpreting data obtained using Fabp4-Cre mice and should encourage the use of additional or alternative adipose-targeting Cre mouse models before drawing conclusions about in vivo adipocyte-specific functions.

Spatial and temporal conditional gene targeting with the use of cell type-specific promoter sequence driving the expression of Cre recombinase is a powerful tool in the interrogation of gene function in mice (1). Adipose selective gene targeting with the use of a fragment of the Fabp4 promoter has been reported by multiple groups (2–4), and of these, the Fabp4-Cre mouse line originally developed and described by He et al. (2) is the most widely used tool to generate adipose-specific knockout mice (Table 1) (5–27).

Table 1.

A selection of floxed genes that have been targeted using Fabp4-Cre mice originally reported by He et al. (2) illustrating widespread implementation of this model in the current literature (the general phenotype reported is briefly summarized)

Targeted gene	Reported phenotype	Reference
Acc1	Growth retardation and reduced lipid accumulation	Mao et al., 2009 (5)
Acsl1	Increased adipose mass and impaired cold tolerance	Ellis et al., 2010 (6)
Ar	Hyperleptinemia	Yu et al., 2008 (9)
Arfrp1	Defective lipid droplet formation and lipodystrophy	Hommel et al., 2010 (10)
Atg7	Decreased adipose mass and increased insulin sensitivity	Singh et al., 2009 (7)
Enpp2	Increased fat mass associated with a high-fat diet	Dusaulcy et al., 2011 (11)
Fas	Increased insulin sensitivity and decreased adipose inflammation	Wueest et al., 2010 (12)
Lrp1	Reduced lipid transport and increased insulin sensitivity	Hofmann et al., 2007 (13)
Mapk8	Increased insulin sensitivity	Sabio et al., 2008 (14)
Ncor	Increased obesity, improved glucose tolerance, and decreased adipose inflammation	Li et al., 2011 (15)
Pnpla2	Impaired thermogenesis, increased insulin sensitivity and defective lipolysis	Ahmadian et al., 2011 (16) and Wu et al., 2012 (17)
Ppard	Impaired cold tolerance	Pan et al., 2009 (8)
Pparg	Lipodystrophy and hepatic steatosis	He et al., 2003 (2)
Ppargc1a	Decreased body weight and loss of rosiglitazone mediated induction of UCP-1 in brown adipose	Pardo et al., 2011 (18)
Raptor	Less adipose and increased insulin sensitivity	Polak et al., 2008 (19)
Rictor	Increased body size and defective glucose and lipolysis	Cybulski et al., 2009 (20) and Kumar et al., 2010 (21)
Scl27a4	Increased adipose mass	Lenz et al., 2011 (22)
Spry1	Increased adipose mass and decreased bone mass	Urs et al., 2010 (23)
Stat3	Increased adipose mass and impaired lipolysis	Cernkovich et al., 2008 (24)
Sufu	Lack of white adipose tissue	Pospisilik et al., 2010 (25)
Trp53	Decreased adipose inflammation and increased insulin sensitivity	Minamino et al., 2009 (26) and Shimizu et al., 2012 (27)

We have been systematically examining the function of class I histone deacetylase 3 (HDAC3), a component of a multiprotein transcriptional corepressor complex, in various tissues using the Cre-LoxP system (28–31). Here we report that Fabp4-Cre targeted deletion of HDAC3 leads to unexpected lethality in mice. We demonstrate that this phenotype cannot be due to adipose-specific loss of HDAC3 because we do not observe the same phenotype when we ablate HDAC3 to a similar degree in adipose tissue using two other independent models, including a novel Retn-Cre mouse line described for the first time here. The dramatically contrasting phenotypes obtained when using these three different adipose targeting Cre lines highlight the need for careful interpretation of the data obtained using adipose-specific Cre-LoxP models and should be considered when embarking on new studies of adipose physiology.

Materials and Methods

Mice

Floxed Hdac3 mice were generated previously (29). Fabp4-Cre and Rosa26 mice were purchased from Jackson Laboratories (Bar Harbor, ME) [B6.Cg-Tg(Fabp4-cre)1Rev/J, stock 005069 and B6.129S4-Gt(ROSA)26Sortm1Sor/J, stock 003474]. AdipoQ-Cre mice were a kind gift from Dr. Evan Rosen (Beth Israel Deaconess Medical Center, Boston, MA) and are now available from Jackson Laboratories [B6;FVB-Tg(Adipoq-cre)1Evdr/J, Stock 010803]. Retn-Cre mice were generated using bacterial artificial chromosome (BAC) modification methods. The structure of Retn was based on the annotated mouse genomic DNA sequences from the University of California, Santa Cruz, genome browser (http://genome.ucsc.edu/). An NLS-Cre-polyA fragment was introduced just in front of the start codon of Retn on a BAC clone using a method described by Gong et al. (32). Start sites of included neighboring genes were mutated to prevent aberrant transcription. A DNA fragment encompassing approximately 23 kb upstream and 10 kb downstream from the transcriptional start site of Retn was retrieved from the modified Retn-Cre BAC into a pBR322 vector (New England Biolabs, Beverly, MA) using a method described by Liu et al. (33). The construct was linearized by NotI digestion, separated in agarose gel and recovered by electroelution. The DNA fragment was purified using Elutip-d column (Whatman, Middlesex, UK), followed by dialysis, and microinjected into pronuclei of fertilized C57BL/6 embryos using standard method at the University Pennsylvania Transgenic Mouse Core Facility. Potential founders were identified by genomic PCR for Cre recombinase. Positive founders were then mated to C57BL/6 mice and offspring were tested for Cre expression (specificity) and used for subsequent experiments with ROSA26 mice (efficiency). The line that demonstrated the best specificity and efficiency is referred to herein as the Retn-Cre line.

Mice were genotyped by PCR using the following primer sequences: Cre, forward, 5′-GCA TTA CCG GTC GAT GCA ACG AGT GAT GAG-3′, reverse, 5′-GAG TGA ACG AAC CTG GTC GAA ATC AGT GCG-3′; Rosa26, oIMR8052, 5′-GCG AAG AGT TTG TCC TCA ACC-3′; oIMR8545, 5′-AAA GTC GCT CTG AGT TGT TAT-3′; and oIMR8546, 5′-GGA GCG GGA GAA ATG GAT ATG-3′ as described by Jackson Laboratories (jaxmice.jax.org); and HDAC3^f/f, forward, 5′-GCA GTG GTG GTG AAT GGC TT-3′, reverse 5′-CCT GTG TAA CGG GAG CAG AAC TC-3′. Mice were maintained in a 12-h light, 12-h dark cycle and had free access to food and water. The Institutional Animal Care and Use Committee at the University of Pennsylvania approved all protocols for animal use and euthanasia.

Adipose fractionation

Adipose tissue was finely minced and then incubated in media containing 1.5 U/ml Collagenase D (Roche, Indianapolis, IN) and 2.4 U/ml of Dispase (Roche) while shaking at 37 C for 30 min. After centrifugation, the floating adipocyte fraction and the pelleted stomal-vascular fraction were separated, washed once in media, and then isolated for subsequent RNA isolation.

β-Galactosidase staining

Tissues were prefixed in 2% paraformaldehyde and 0.1% gluteraldehyde, washed and stained with 1 mg/ml x-gal, 4 mm potassium ferricyanide, and 4 mm potassium ferrocyanide at 37 C for 3–16 h. The reaction was stopped when background staining began to appear in tissues from control littermates without Cre. Stained tissues were then washed in PBS and postfixed in 4% paraformaldehyde overnight at 4 C before photographing.

Gene expression

RNA was isolated from tissue using the RNeasy kit (QIAGEN, Valenica, CA). cDNA was generated using the high-capacity cDNA reverse transcription kit (Applied Biosystems, Foster City, CA). Quantitative PCR was performed using Power SYBR Green PCR master mix on the 7900HT Fast real-time PCR system (Applied Biosystems). Data were analyzed using a standard curve and normalized to Arbp expression. Primer sequences used are as follows: Arbp, forward, 5′-GGC ACC GAG GCA ACA GTT-3′, reverse, 5′-TCA TCC AGC AGG TGT TTG ACA-3′; Cre, forward, 5′-GCA TTA CCG GTC GAT GCA ACG AGT GAT GAG-3′, reverse 5′-GAG TGA ACG AAC CTG GTC GAA ATC AGT GCG-3′; and Hdac3, forward, 5′-TTG GTA TCC TGG AGC TGC TT-3′, reverse, 5′-GAC CCG GTC AGT GAG GTA GA-3′.

Statistical analyses

Comparison of survival curves was performed using Graph Pad Prism (Graph Pad Inc., San Diego, CA), and statistically significant differences were determined using the Log-rank (Mantel-Cox) test.

Results

Fabp4-Cre-mediated deletion of HDAC3 is lethal

To determine the in vivo function of adipose HDAC3, we attempted to delete its expression in adipocytes by mating the commercially available Fabp4-Cre line originally generated by He et al. (2) with floxed Hdac3 mice (HD3^f/f) (29). Surprisingly, no viable adult Fabp4-Cre⁺, HD3^f/f mice were obtained. To determine whether the Fabp4-Cre⁺, HD3^f/f mice were born normally, mice were genotyped at birth and monitored until weaning. Importantly, Fabp4-Cre⁺, HD3^f/f mice were born at the expected Mendelian ratio (Fig. 1A). However, no Fabp4-Cre⁺, HD3^f/f mice survived past 25 d of age (Fig. 1B). More than 50% of the mice died before postnatal d 5, and visual monitoring of the remaining mice revealed a failure to thrive and wasting phenotype before death (Fig. 1C). Importantly, adipose tissue developed and was morphologically normal at postnatal d 4 (Fig. 1D). To date, we are unaware of any conclusive reports of adipocyte-intrinsic defects leading to death. In fact, severely lipodystrophic mice, although metabolically abnormal, survive until adulthood (2, 34–38). Therefore, we rationalized that it was unlikely that the early death in the Fabp4-Cre⁺, HD3^f/f mice was due to an adipose autonomous defect. Consistent with this hypothesis, there have been several reports of nonadipose Cre activity under the control of the 5.4-kb fragment of the Fabp4 promoter (39–41).

Fig. 1.

A Cre-LoxP recombination strategy to delete HDAC3 using Fabp4-Cre mice is lethal. Fabp4-Cre⁺, HD3^f/f mice were born at the expected frequency. A total of 397 pups from Fabp4-Cre⁺, HD3^f/+ x HD3^f/+ breedings were genotyped between postnatal d 1 and 3 (A) and monitored for survival until weaning (B). C, Two representative Fabp4-Cre⁺, HD3^f/f mice at postnatal d 7 are shown in comparison with littermate controls. D, Inguinal white adipose depots (arrows) at postnatal d 4.

Development of Retn-Cre mice for adipose selective targeting

To address the possibility that the early lethality in the Fabp4-Cre⁺, HD3^f/f mice was due to loss of HDAC3 in a nonadipose tissue, we sought to delete floxed Hdac3 using an alternative Cre line selectively targeting adipose tissue. We generated a BAC transgenic mouse in which Cre recombinase was placed within a 33-kb fragment of Retn, the gene encoding resistin, an adipokine exclusively expressed in murine adipose tissue (Fig. 2A) (42). This construct contains a previously described adipocyte-specific enhancer (43).

Fig. 2.

Development and validation of Retn-Cre mice for adipose-selective targeting. A, A BAC construct containing a 33-kb fragment surrounding the transcriptional start site of Retn was generated with NLS-Cre+polyA inserted in exon 2. Start sites of included neighboring genes were mutated to prevent aberrant transcription. B, Adult tissues from ROSA26 heterozygous (R26^+/−) mice with and without Retn-Cre, Fabp4-Cre and AdipoQ-Cre were stained for β-galactosidase activity. BAT, Brown adipose tissue; IWAT, inguinal white adipose tissue; EWAT, epididymal white adipose tissue. C, Cre expression was examined by quantitative PCR in adipose, fractionated adipocytes (ADS), SVFs, PECs, and spleen isolated from Retn-Cre⁺ and control Cre⁻ adult male mice and is shown normalized to the control gene arbp (n = 2–3).

To determine the efficiency and specificity of Cre expression, the Retn-Cre⁺ mice were crossed with a ROSA26 reporter strain in which recombinase activity leads to removal of a floxed stop codon and permanent expression of the lacZ gene (44). A comprehensive panel of tissues was harvested and stained for β-galactosidase activity (Fig. 2B and data not shown). Efficient recombination and removal of the floxed stop codon in the ROSA26 reporter mice was detected in all fat depots examined including epididymal and inguinal white adipose as well as intrascapular brown adipose (Fig. 2B). No activity was observed in other metabolic tissues including the liver, skeletal muscle, and pancreas, whereas scattered punctate staining was observed in some regions of the brain (Fig. 2B and data not shown). This was in stark contrast to the results obtained when crossing the Fabp4-Cre⁺ mice to the same reporter strain that revealed detectable Cre activity in all tissues examined (Fig. 2B and data not shown). Furthermore, Cre expression was negligible in the adipose stromal-vascular fraction (SVF), peritoneal exudate cells (PECs), and spleen, all of which contain a significant percentage of macrophages, an immune cell previously reported to have detectable Cre activity in Fabp4-Cre⁺ mice (Fig. 2C) (5). Of note, the specificity of Cre expression in the Retn-Cre⁺ mice was similar to that observed in mice previously generated expressing Cre under the control of the AdipoQ promoter (Fig. 2B) (45). Therefore, although expression of Cre from this construct is not identical with endogenous resistin expression (which is not detectable in brown adipose or brain), this Retn-Cre⁺ model does serve as an alternative tool for adipose-selective gene inactivation.

Adipose selective targeting of HDAC3 with Retn-Cre

We next generated Retn-Cre⁺, HD3^f/f mice and assessed them for efficient deletion of Hdac3 and survival. Detailed phenotypic analyses of these mice will be described elsewhere. Importantly, Hdac3 expression was markedly reduced in epidydmal, inguinal, and brown adipose tissues but not in adipose, SVF, PECs, or spleen (Fig. 3A). The adipose depletion of Hdac3 in Retn-Cre⁺, HD3^f/f mice was comparable with that in adipose tissue of Fabp4-Cre⁺, HD3^f/f mice (Fig. 3B). The residual expression of Hdac3 is consistent with the fact that adipose tissue contains other cell types in addition to adipocytes, including preadipocytes, macrophages, and endothelial cells (46).

Fig. 3.

Depletion of HDAC3 in adipose tissue using Retn-Cre. A, Adipose Hdac3 expression determined by quantitative PCR is reduced in adult epididymal white adipose tissue (EWAT), inguinal white adipose tissue (IWAT), and brown adipose tissue (BAT) but not in the SVF of adipose, PECs, or spleen of Retn-Cre⁺, HD3^f/f mice compared with littermate controls. B, A similar reduction of Hdac3 expression in adipose tissue of Fabp4-Cre⁺, HD3^f/f mice is observed in comparison with littermate controls at postnatal d 4 (n = 2–4).

Depletion of HDAC3 in adipose tissue is not lethal

Adipose tissue selective Retn-Cre⁺, HD3^f/f mice were born in an expected Mendelian ratio and, in remarkable contrast to the Fabp4-Cre⁺, HD3^f/f mice, postnatal mice displayed no overt abnormalities and all survived until adulthood (Fig. 4A). The difference in survival between the Fabp4-Cre⁺, HD3^f/f and Retn-Cre⁺, HD3^f/f mice coupled with the widespread recombinase activity detected in the Fabp4-Cre⁺, ROSA26 mice suggests that the early lethality observed when Hdac3 is deleted under the control of the Fabp4 promoter is due to loss of HDAC3 in cells other than adipocytes. In support of this conclusion, crossing our floxed Hdac3 mice with those expressing Cre under the control of the AdipoQ promoter (45) led to similar results to those obtained from the Retn-Cre⁺, HD3^f/f mice, that being no reduction in survival (Fig. 4B). In a longer follow-up, Retn-Cre⁺, HD3^f/f and AdipoQ-Cre⁺, HD3^f/f mice have shown no increase in mortality up to 6 months (data not shown).

Fig. 4.

Deletion of adipose HDAC3 is not lethal. A, Normal early life expectancy of Retn-Cre⁺, HD3^f/f mice. B, Normal early life expectancy of AdipoQ-Cre⁺, HD3^f/f mice. Retn-Cre⁺, HD3^f/f (A) and AdipoQ-Cre⁺ HD3^f/f (B) pups and corresponding littermate controls (HD3^f/f) were genotyped between postnatal d 1 and 3 and monitored for survival until weaning (n = 60 total pups in A and n = 39 total pups in B).

Discussion

Tissue-specific gene targeting in mice has become one of the most powerful tools in biomedical research in the last 2 decades (1). In contrast to global knockout models, it allows functions of gene products to be assigned to specific cell types. In particular, adipose specific gene deletion allows for a more detailed understanding of the peripheral effects of aberrant adipocyte lipid metabolism and storage and the endocrine actions of adipokines and hormones of which adipocytes can be both effector and/or target cells. Fabp4 is a gene highly expressed in adipocytes, and multiple groups have used a fragment of its promoter for transgenic expression of Cre recombinase in mouse adipose tissue (2–4). Despite several reports of nonadipose expression of Fabp4-Cre (39–41), the commercially available Fabp4-Cre mouse model originally described by He et al. (2) is the most widely used line in studies in which adipose-selective gene targeting is desired. Consistent with previous reports, we have detected extraadipose Cre activity under the control of the Fabp4 promoter, suggesting that it is not sufficient for adipose-selective gene targeting. More recently alternative adipose-targeting Cre lines have been generated using the regulatory sequences of Adipoq, another gene highly expressed in adipose tissue (45, 47). Unlike Fabp4-Cre and consistent with our findings, expression of Adipoq-Cre in cells other than adipocytes has not been reported.

Resistin is an adipokine whose production is limited to white adipose tissue in the mouse (42). In an attempt to express Cre recombinase in a similarly tissue-restricted manner, we generated a transgenic mouse with Cre inserted into a 33-kb fragment surround the Retn transcriptional start site that includes a previously reported adipocyte-specific enhancer (43). Using ROSA26 reporter mice, sufficient Cre activity was detected in white adipose tissue. However, unlike endogenous Retn expression, Retn-Cre activity was also detected in punctate regions of the brain and in brown adipose tissue. Therefore, although expression of Cre in this model is not white adipose restricted as anticipated, it is considerably more limited than that observed in the Fabp4-Cre model. Thus, this novel mouse line can serve as an additional resource for independent confirmation of results obtained with other adipose-selective systems, such as Adipoq-Cre.

Most importantly, we report here drastically different phenotypes upon deletion of floxed Hdac3 to similar levels with Fabp4-Cre, Retn-Cre, and Adipoq-Cre. Whereas deletion of Hdac3 with Fabp4-Cre results in unexpected postnatal lethality, Retn-Cre⁺, HD3^f/f and Adipo-Cre⁺, HD3^f/f mice exhibit survival rates comparable with their control littermates. These findings, in conjunction with the detection of nonadipose activity of Cre in Fabp4-Cre mice, clearly demonstrate that the observed dramatic phenotype in the Fabp4-Cre⁺, HD3^f/f mice is not due to loss of HDAC3 in adipocytes.

The data presented here underscore the need for careful validation of the specificity of target gene deletion before drawing conclusions about adipose-specific function. In particular, great caution should be taken when interpreting data based on use of the commercially available Fabp4-Cre mouse line (2). Indeed, important conclusions already in the published literature ought to be verified using alternative models such as an AdipoQ-Cre line (45, 47) or the Retn-Cre mice introduced here. Moreover, given the imperfections of all of these models, we suggest that to avoid future errors of interpretation, conclusions about in vivo adipocyte-specific functions should be validated in at least two independent mouse models, particularly if the results are surprising and difficult to explain by a gene deletion that is restricted to adipose tissue.