Role of the Beta and Gamma Isoforms of the Adapter Protein SH2B1 in Regulating Energy Balance

Abstract

Human variants of the adapter protein SH2B1 are associated with severe childhood obesity, hyperphagia, and insulin resistance—phenotypes mimicked by mice lacking Sh2b1. SH2B1β and γ isoforms are expressed ubiquitously, whereas SH2B1α and δ isoforms are expressed primarily in the brain. Restoring SH2B1β driven by the neuron-specific enolase promoter largely reverses the metabolic phenotype of Sh2b1-null mice, suggesting crucial roles for neuronal SH2B1β in energy balance control. Here we test this hypothesis by using CRISPR/Cas9 gene editing to delete the β and γ isoforms from the neurons of mice (SH2B1βγ neuron-specific knockout [NKO] mice) or throughout the body (SH2B1βγ knockout [KO] mice). While parameters of energy balance were normal in both male and female SH2B1βγ NKO mice, food intake, body weight, and adiposity were increased in male (but not female) SH2B1βγ KO mice. Analysis of long-read single-cell RNA seq data from wild-type mouse brain revealed that neurons express almost exclusively the α and δ isoforms, whereas neuroglial cells express almost exclusively the β and γ isoforms. Our work suggests that neuronal SH2B1β and γ are not primary regulators of energy balance. Rather, non-neuronal SH2B1β and γ in combination with neuronal SH2B1α and δ suffice for body weight maintenance. While SH2B1β/γ and SH2B1α/δ share some functionality, SH2B1β/γ appears to play a larger role in promoting leanness.

The adapter protein Src Homology 2 (SH2) B1 has been implicated as an important regulator of energy balance, feeding behavior, and glucose homeostasis. In humans, the role of SH2B1 in regulating energy balance is supported by genome-wide association studies of a wide range of racial groups, including European (1, 2), African (3), and Chinese (4) populations. It is also supported by gene deletions in which SH2B1 is the gene most likely to be responsible for the observed obesity (5), as well as rare SH2B1 variants in children or adolescents with severe obesity and hyperphagia (6-10).

Consistent with SH2B1 playing an important role in energy balance, male and female mice lacking SH2B1 (SH2B1 KO mice) are obese and hyperphagic (11). Transgenically reintroducing SH2B1β to both male and female SH2B1 KO mice using the neuron-specific enolase promoter largely restores the lean phenotype (12), suggesting that the lack of neuronal SH2B1β is primarily responsible for the obese phenotype in SH2B1 KO mice. A critical role for neuronal SH2B1 in the regulation of energy balance is consistent with the hypothesis that the brain is the central driver and regulator of both aspects of energy balance—food intake and energy expenditure (13).

SH2B1 is known to enhance cellular responses to hormones and neurotrophic factors that regulate energy balance. SH2B1β has been shown, in response to the satiety factor leptin, to be recruited via its SH2 domain to phosphotyrosyl residues in the leptin-activated tyrosine kinase JAK2, forming a complex that includes JAK2, the leptin receptor and insulin receptor substrate proteins (14). SH2B1 is also recruited to the receptors that bind brain-derived neurotrophic factor (BDNF) (15), nerve growth factor (NGF) (16, 17), fibroblast growth factor-1 (18), and glial cell–derived neurotrophic factor (GDNF) (19). Reduction of endogenous SH2B1 expression using RNAi reduces NGF, BDNF, and GDNF-induced neurite outgrowth and NGF-induced gene expression in PC12 cells expressing the relevant neurotrophic factor receptor (15, 16, 20, 21). Similarly, studies using isolated hippocampal neurons from SH2B1 KO mice (22) and cortical neurons treated with shRNA to SH2B1 (15) reveal roles for SH2B1 in neuronal architecture.

There are 4 known isoforms of SH2B1—α, β, γ, and δ—the products of exon skipping and alternative 5′ splice sites (23, 24). These isoforms share their first 631 amino acid residues, which contain dimerization, pleckstrin homology, and SH2 domains, several proline-rich regions, a nuclear localization sequence, and a nuclear export sequence (Fig. 1A). The C-termini of each isoform has a unique amino acid sequence. In humans, the β and γ isoforms are expressed ubiquitously, whereas the α and δ isoforms are largely restricted to the brain (6), raising the question of whether the α and δ isoforms are the key isoforms responsible for maintaining body weight. However, Cote et al (25) found that mice unable to make the SH2B1α and δ isoforms (SH2B1αδ KO mice) are not obese. In fact, both male and female SH2B1αδ KO mice are leaner than their wild-type (WT) littermates, suggesting that the obese phenotype in SH2B1 KO mice might be caused by the lack of the β and/or γ isoforms. In this study, we test more directly the hypothesis that neuronal expression of the β and/or γ isoforms of SH2B1 is critical for maintaining normal (nonobese) body weight.

Figure 1.

Generation and validation of neuron-specific SH2B1βγ NKO mouse model. (A) Schematic of SH2B1 isoforms. The isoform-specific C-terminal tails are denoted by the rectangles after residue 631. Numbers indicate amino acid residues in mouse and human sequences. P, proline-rich domain; DD, dimerization domain; NLS, nuclear localization sequence; NES, nuclear export sequence; PH, pleckstrin homology domain; SH2, Src homology domain. (B) Schematic showing region of Sh2b1 gene to be deleted to prevent the expression of the β and γ isoforms. The shaded region was targeted for deletion. The letters A and B denote the proposed Cas9 cut sites. Stop codons are labeled “stop.” Exons are labeled “Ex.” Locations of WT genotyping primers are indicated by horizontal arrows beneath the schematic. (C) Schematic of the overall strategy to insert loxP and restriction sites at the proposed Cas9 cut sites in the regions flanking exon 9 of Sh2b1 gene. Locations of genotyping primers are indicated by horizontal arrows beneath the schematic. (D) Genomic DNA was purified from mouse tails and analyzed by PCR using primers indicated in B and C and Table 3. (E) Genomic DNA was purified from brains and liver of βγ NKO mice or their littermate Syn1-Cre control mice and analyzed by PCR using primers as in (D). (F) Lysates from whole brain or pancreas from βγ NKO mice or their littermate βγ fl/fl control mice (18 weeks old) or SH2B1 KO mice were immunoblotted with αSH2B1 and reprobed with αβ-actin. The migration of molecular weight standards is shown on the right. The expected migrations of the different isoforms of SH2B1 and β-actin are indicated on the left. A band in the blots of pancreases from βγ fl/fl and βγ NKO mice migrates just below the SH2B1α band in the blots of brains from βγ fl/fl and βγ NKO mice. This band is thought to be a nonspecific band because of its faster migration and its presence in the pancreases of SH2B1 KO mice. IB, immunoblot; n.s., non-specific band. The band corresponding to SH2B1β/γ in 6 littermate pairs was quantified and normalized to the intensity of the SH2B1β/γ fl/fl band. Statistical difference was assessed using a paired 2-tailed t-test. Mean and SEM are shown; **P < .01.

Material and Methods

Animal Care

Mice were housed in ventilated cages at ∼22 °C on a 12-hour light/dark cycle (∼6 Am-6 Pm) in a pathogen-free animal facility at the University of Michigan. Food and tap water were available ad libitum except as noted. Experimental mice were fed a standard chow (20% protein, 9% fat; LabDiet cat. no. 5058). Experiments were approved by the University of Michigan Institutional Animal Care & Use Committee in accordance with Association for Assessment and Accreditation of Laboratory Animal Care and National Institutes of Health guidelines.

Mouse Models, Breeding, and Genotyping

CRISPR/Cas9 editing was used to produce mice with either whole body or tissue-specific deletion of Sh2b1 exon 9 that codes for the β and γ isoforms of SH2B1 (Fig. 1A-1C) with the help of the Michigan Diabetes Research Center Molecular Genomics Core and the University of Michigan Transgenic Animal Model Core. The reverse complement of the genomic Sh2b1 sequence in C57BL/6J mice (accession number NC_000073, GRCm38) was used to design the reagents for CRISPR. RNA guides were selected using the prediction scores from the website described in Ran et al (26) and CRISPOR (27).

Two sets of guides were chosen for full-scale embryo targeting. The cutting efficiency of the guides was tested in mouse blastocysts. Two templates were then designed to insert loxP sites into introns 8 and 9 away from sequences implicated in splicing: (1) a 200 bp template encoding a 5′ loxP site and diagnostic BamH1 and EcoR1 restriction sites, and (2) a 200 bp template encoding a 3′ loxP site (Integrated DNA Technologies, Coralville, Iowa). Guides A and B (Table 1) and the oligo templates (Table 2) were injected into C57BL/6J × SJL F2 oocytes and the injected oocytes were implanted into C57BL/6J × SJL F2 mice by the University of Michigan Transgenic Animal Model Core. Eight mice containing at least 1 allele with both loxP sites, the BamH1 and EcoR1 restriction sites, and no other mutations in the coding region (Sh2b1βγ^fl mice), and 5 mice containing at least 1 allele lacking exon 9 (Sh2b1^DELβγ) (βγ KO mice) were identified by Topo cloning, polymerase chain reaction (PCR) (primers listed in Table 3), and DNA sequencing (University of Michigan DNA Sequencing Core). These mice included a mosaic mouse (3816) containing 1 Sh2b1βγ^fl allele and 1 Sh2b1^DELβγ allele that was used as the founder for both the Sh2b1^DELβγ and Sh2b1βγ^fl mouse lines. A few experiments were also performed with a founder mouse (795) that was homozygous for Sh2b1^DELβγ. Germline transmission was confirmed for the 3 mouse lines, which were backcrossed against C57BL/6J mice. C57BL/6J mice used to invigorate our C57BL/6J colony came from The Jackson Laboratory (Bar Harbor, ME). Nonsibling Sh2b1^DELβγ/+ mice from backcrosses N2, N3, and N5 (line 3816) or N2 (line 795) were bred to produce experimental SH2B1βγ KO mice and WT littermates. The Sh2b1^DELβγ (795) line lacked nt 6878-7227 and had a TG insert. The Sh2b1^DELβγ (3816) line lacked nt 6877-7229.

Table 1.

Guide RNAs

Guide	Nucleotide nos.a	Strand	Sequence (5′→3′)
A	6884-6903	Antisense	CGCGGAGGGTCCAATGG^AAT
B	7211-7230	Sense	GATTCATCTTCCATGGA^GGG

^a Nucleotide numbers in mouse Sh2b1 are from the reverse complement of accession number NC_000073.

^Cas 9 cut site.

Table 2.

Templates for inserting loxP sites

Template	Total length	Strand	Sequence (5′→3′)a
Template to introduce loxP + EcoR1 + BamH1 sites into intron 8 at cut site directed by guide A	200 nt	Sense	Sh2b1 nt 6811–6881 (71 nt)- ccattATAACTTCGTATAGCATACATTATACGAAGTTATgaattctgGGATCCccatt-Sh2b1 nt 6892-6962 (71 nt)b
	200 nt	Antisense	Reverse compliment of above oligo
Template to introduce loxP site into intron 9 at cut site directed by guide B	200 nt	Sense	Sh2b1 nt 7145–7222 (78 nt)- atggaATAACTTCGTATAGCATACATTATACGAAGTTATggggg-Sh2b1 nt 7233-7310 (78 nt)c
	200 nt	Antisense	Reverse compliment of above sense oligo

^a Nucleotide numbers in mouse Sh2b1 are from the reverse complement of accession number NC_000073.

^b LoxP site and BamH1 site are in caps, EcoR1 site is underlined. The 5 base pairs flanking the inserted sequence are in lower case.

^c LoxP site is in caps. The 5 base pairs flanking the inserted sequence are in lower case.

Table 3.

Primers used for genotyping

Primer	Direction	Nucleotide nos.a	Sequence (5′→3′)
Sh2b1DELβγ and Sh2b1βγfl	Forward	6796-6817	GACCTAGAAGGGAAGCAGAAGG
	Reverse	7331-7355	GGATAAGAATGGTGAGAAGACATTG
Syn1-Cre	Forward	—	CCGGTGAACGTGCAAAACAGGCTCTA
	Reverse	—	CTTGCATGATCTCCGGTATTGAAACTCCAG
Sh2b1 b	Forward 1	6280-6303	GGAGGCACTGGCTCCCATGGTGTC
	Reverse	6460-6483	GCAGGATGACAAGTGAGGTGGGAG
Sh2b1-neob	Forward 2	—	ATTCCTCCCACTCATGATCTATAGATC

^a Nucleotide numbers in mouse Sh2b1 are from the reverse complement of accession number NC_000073.

^b Primers used for genotyping SH2B1 KO mice.

To obtain mice with the β and γ isoforms of SH2B1 specifically deleted in neurons, male Sh2b1βγ^fl/+ mice backcrossed for 2 to 6 generations against C57BL/6J mice were crossed with transgenic female mice expressing Cre recombinase regulated by the neuronal synapsin 1 promoter (Syn1-Cre) on a C57BL/6 background (28). Because Syn1-Cre is also expressed in the testis (29), maintenance and experimental mice were obtained by crossing female Syn1-Cre; Sh2b1βγ^fl/fl or Syn1-Cre; Sh2b1βγ^fl/+ mice with male Sh2b1βγ^fl/+ or Sh2b1βγ^fl/fl mice. Homozygous floxed, Cre-positive (Syn1-Cre; Sh2b1βγ^fl/fl) knockout mice (βγ NKO mice) and control homozygous floxed mice (Sh2b1βγ^fl/fl) (βγ fl/fl) mice were used in experiments. Mice were excluded if genotyping revealed that the exon 9 deletion had gone germline. Experimenters were blind to genotypes.

For genotyping, genomic DNA was isolated (30) from tails prior to weaning or occasionally ear snips after weaning. Diagnostic fragments were amplified by PCR using Q5 High-Fidelity DNA polymerase (New England BioLabs) (Sh2b1βγ^fl and Sh2b1^DELβγ mice) or DreamTaq DNA polymerase (Thermo Fisher Scientific) (Syn1-Cre and Sh2b1 KO mice). The primers used are listed in Table 3. For Fig. 1E, genomic DNA was isolated from frozen tissue.

Body Weight and Food Intake

Mice were individually housed, and body weight was assessed weekly. Food intake was monitored once or twice weekly. Food intake was determined by weighing the food remaining in the cage and subtracting the value from the weight of the food initially added to the food hopper. Following determination of food intake, old food was removed and fresh food was added.

Insulin and Leptin Levels

Blood was collected between 9 Am and 11 Am from mice given free access to food. Insulin and leptin levels in tail vein blood were measured using Crystal Chem ELISA kits for mouse insulin (cat. no. 90080; RRID:AB_2783626) or leptin (cat. no. 90030; RRID:AB_2722664).

Glucose and Insulin Tolerance Tests

Mice were subjected to a 4-hour (glucose tolerance test) or 6-hour (insulin tolerance test) morning fast followed by intraperitoneal (IP) injection of D-glucose (2 mg/kg of body weight) or human insulin (1 IU/kg body weight) as described previously (8). Blood glucose was measured in duplicate or triplicate via tail vein bleeding using a Bayer Contour glucometer. Replicate values for each timepoint were averaged for each mouse.

Tissue Harvest

Mice were anesthetized by isoflurane between 10 Am and 2 Pm and decapitated. Tissues, including brain, fat (gonadal white adipose), liver, muscle (quadriceps femoris), and/or pancreas, were dissected, weighed, cryopreserved in liquid nitrogen, and stored (−80 °C) for RNA or protein extraction.

Immunoblotting

Frozen tissues were lysed and homogenized using a glass Dounce homogenizer containing lysis buffer, as described previously (8). Protein concentrations were determined using BCA protein assay (Thermo Fisher). Equal amounts of protein in tissue lysates were subjected to immunoblotting overnight with mouse monoclonal antibody to SH2B1 (1:1000) (cat. no. sc-136065; Santa Cruz; RRID:AB_2301871) in 10 mM Tris, 150 mM NaCl, pH 7.4 (tris-buffered saline [TBS]) containing 0.1% Tween 20 and 1% fish gel (cat. no. G7041; Sigma), followed by IRDye 680LT goat antimouse antibody (1:15 000) (cat. no. 926-68020; Li-Cor; RRID:AB_10706161) in TBS containing 0.005% sodium dodecyl sulfate for 1 hour at room temperature. Blots were reprobed with rabbit polyclonal antibody to β-actin (1:20 000) (cat. no. AC026; ABclonal; RRID:AB_621843) for 4 hours followed by IRdye 800CW goat antirabbit antibody (1:20 000) (cat. no. 926-32211; Li-Cor; RRID:AB_2768234) in TBS. Relative expression levels of SH2B1 isoforms were quantified using Li-Cor Odyssey 3 or Li-Cor Image Studio (version 5.2.5). Small variations in loading were taken into account by normalizing SH2B1 isoform expression levels to levels of β actin or nonspecific bands.

Statistical Methods

Statistics were performed using GraphPad Prism 8. Body weights and cumulative food intake were compared by linear regression analysis of the linear parts of the curve. All other measurements were compared using unpaired or paired 2-tailed student t tests as noted. P < .05 was considered to be statistically significant. Outliers were determined using the ROUT method (Q = 1%) (31) in GraphPad Prism 8 and are listed in the figure legends.

Results

Generation of Mice Lacking SH2B1β and SH2B1γ in Neurons

The findings that SH2B1 KO mice are obese (11) and that transgenic expression of SH2B1β driven by the neuron-specific enolase promoter in SH2B1 KO mice largely reverses the obese phenotype (12) suggested that neuronal SH2B1β plays a major role in the control of energy balance. To better understand the contributions of SH2B1β and SH2B1γ to energy balance, we generated Syn1-Cre; Sh2b1βγ^fl/fl mice (βγ NKO mice) in which both the β and γ isoforms of SH2B1 are specifically deleted in neurons. It is not possible to delete the β and γ isoforms individually because the β and γ isoforms share alternative splice sites with the α and δ isoforms in exon 8 and intron 9 and utilize 2 separate reading frames in exon 9 (Fig. 1B). However, the β and γ isoforms exhibit similar behavior and subcellular distribution in cell culture studies and migrate similarly in immunoblots (24, 25, 32) (data not shown). To make the βγ NKO mice, we first used CRISPR Cas9 technology to insert loxP sites in the introns surrounding exon 9 to generate Sh2b1β^fl mice (Fig. 1C). A founder mouse containing an allele with 2 loxP sites was identified, germline expression confirmed, and heterozygous progeny crossed to establish the Sh2b1βγ^fl line (Fig. 1D). The Sh2b1βγ^fl mouse line was then crossed with transgenic mice expressing cre driven by the neuronal synapsin 1 promoter (Syn1-Cre mice) (28). Because Syn1-Cre is expressed in the testis (29), experimental mice were obtained by crossing female Syn1-Cre; Sh2b1βγ^fl/fl or Syn1-Cre; Sh2b1βγ^fl/+ mice with male Sh2b1βγ^fl/+ or Sh2b1βγ^fl/fl mice. To verify the neuronal specificity of Syn1Cre expression, we examined whether β/γ isoform–specific exon 9 was deleted in the brain and present in a nonbrain tissue (liver) of βγ NKO mice. Genotyping revealed that exon 9 was not deleted in the liver (band comigrating with the Sh2b1 βγ^fl/fl genotype in Fig. 1E, lane 2). However, in the brain of βγ NKO mice, Sh2b1 with exon 9 deleted (band comigrating with the size expected for the Sh2b1βγ KO genotype) was detected (Fig. 1E, lane 1). The small amount of exon 9-containing Sh2b1βγ^fl/fl remaining in the brain is presumed to be due to Sh2b1βγ^fl/fl in non-neuronal cells, which make up ∼35% of the cells in mouse brain (33). We next looked at the protein expression of SH2B1 isoforms in both the brain and a nonbrain tissue (pancreas) (Fig. 1F). As expected, in the pancreas of βγ NKO mice, expression of SH2B1 β/γ protein was unchanged compared with expression in control littermate Sh2b1βγ^fl/fl (βγ fl/fl) mice. In the brain, expression of SH2B1 β/γ protein was marginally decreased, consistent with a substantial non-neuronal population of cells in the brain.

Lack of SH2B1β and γ in Neurons Does Not Affect Body Weight or Adiposity in Male or Female Mice

We first assessed whether deleting the β and γ isoforms of SH2B1 from neurons influences energy balance. We predicted that the βγ NKO mice would be obese, like the total SH2B1 KO mice. We fed both male and female βγ NKO mice and their control βγ fl/fl littermates standard chow and measured their body weight weekly from weeks 7 to 16. Neither male (Fig. 2A) nor female (Fig. 2B) βγ NKO mice gained more weight than their control βγ fl/fl littermates. There was also no difference in adiposity, assessed by weight of perigonadal adipose tissue, in either male or female βγ NKO mice compared with their control βγ fl/fl littermates (Fig. 2C and 2D). There was also no difference in the weight of brain or liver of male or female mice (Fig. 2C and 2D).

Figure 2.

Male and female βγ NKO mice exhibit no change in body weight or adiposity. (A) Body weight of male mice was measured weekly from 7 to 16 weeks (n: βγ fl/fl = 12, βγ NKO = 11). (B) Body weight of female mice was measured weekly from 7 to 16 weeks (n: βγ fl/fl = 9, βγ NKO = 9). Statistical differences were assessed using linear regression. (C) Perigonadal adipose tissue, brain and liver tissues of 19- to 28-week-old male mice were dissected and weighed (n: βγ fl/fl = 13; βγ NKO = 11). (D) Perigonadal adipose tissue, and brain and liver tissues of 19- to 24-week-old female mice were dissected and weighed (n: βγ fl/fl = 9; βγ NKO = 8). Statistical differences were assessed using unpaired 2-tailed t-test. For all panels, means and SEM are shown.

Single-Cell Analysis Reveals the Absence of the β and γ Isoforms of SH2B1 in Neurons in the Hippocampus and Prefrontal Cortex

The lack of increased body weight or adiposity in the βγ NKO mice combined with the minimal reduction of expression of the β and γ isoforms of SH2B1 in the brain of the βγ NKO mice led us to wonder whether the β and γ isoforms of SH2B1 in the brain might reside primarily in non-neuronal cells. To gain insight into the distribution of the different SH2B1 isoforms in neuronal vs non-neuronal cell types, we examined the data in the isoform atlas of the postnatal mouse brain of Joglekar et al (34), published after our phenotyping of the βγ NKO mice. These investigators used long reads to help identify differential isoform expression in the various cell types of the hippocampus and prefrontal cortex in postnatal day 7 (P7) of C57BL/6NTac mice. Figure 3 reveals that in these 2 regions of the brain, the SH2B1β and γ isoforms are expressed almost exclusively in the neuroglial cells (eg, radial glia–like cells, astrocytes). In contrast, the α and δ isoforms of SH2B1 are expressed almost exclusively in neurons (both excitatory and inhibitory). The 2 exceptions were oligodendrocyte progenitor cells (OPCs), which expressed similar amounts of Sh2b1α and Sh2b1β mRNA in both the hippocampus and prefrontal cortex, and proliferating OPCs in the hippocampus, which expressed only Sh2b1α and Sh2b1δ mRNAs. The apparent absence or low level of expression of the β and γ isoforms of SH2B1 in neurons is consistent with our finding that deleting the β and γ isoforms from neurons did not affect body weight or adiposity of the βγ NKO mice.

Figure 3.

Single cell expression of SH2B1 isoforms in neuronal and non-neuronal cells in mouse hippocampus and prefrontal cortex. (A, B) SH2B1 isoforms are indicated on the left. Each isoform appears more than once in the list because there are multiple mRNA transcripts that encode each isoform. Numbers in the boxes represent the number of unique molecules identified. The colors in the boxes represent the prevalence (% isoform) of that particular isoform in that cell type. The “% isoform” color gradient at the bottom of the table defines the percentages associated with each color. Gray boxes (“N/A”) indicate that there were very low reads across the entire cell type. For each cell type, the column adds up to 100%. 73aa, 73 amino acid variant; NMD, nonsense-mediated decay; lncRNA, long noncoding RNA; HIP, hippocampus; PFC, prefrontal cortex; InNeuron, inhibitory neuron; ExNeuron, excitatory neuron; Immat, immature; GuleNB, dentate gyrus granule neuroblasts; RGL, radial glia like; Astro, astrocyte; Epen, ependymal; ChorPlexEpith, secretory choroid plexus epithelial; OPCs, oligodendrocyte precursor cells; Prol, proliferating; COPs, committed oligodendrocyte precursors. The single cell expression data for the SH2B1 isoforms was accessed from the Brain mRNA Isoform Atlas (https://isoformatlas.com/; http://creativecommons.org/licenses/by/4.0/) of Joglekar et al (34). An isoform designation (eg, α, β, γ, δ) was assigned based on the mRNA sequence corresponding to its Ensembl transcript ID or the mRNA schematic.

Deleting SH2B1β and γ Throughout the Body Causes Isoform Switching

Because we did not see an obese phenotype with the SH2B1βγ NKO mice and single-cell sequencing suggested that the β and γ isoforms were primarily expressed in non-neuronal cell types, we tested whether deleting the β and γ isoforms of SH2B1 from all cells in the body would cause an obese phenotype. When we made our Sh2b1βγ^fl mice using CRISPR/Cas9, we also obtained founder mice in which Sh2b1 was cut, but nonhomologous end-joining occurred and exon 9 was deleted (Fig. 4A). Germline transmission was confirmed in these mice, and heterozygous progeny were crossed to establish the Sh2b1^DELβγline. Genotyping (Fig. 4B) and DNA sequencing (data not shown) of founder animals and their progeny identified mice containing a correctly edited Sh2b1^DELβγ allele. Sh2b1^DELβγ/+ intercrosses produced pups with Sh2b1 genotypes at the expected Mendelian ratio (data not shown). Immunoblots confirmed that SH2B1β/γ isoforms were absent in the brain, pancreas, and liver of the Sh2b1^{DELβγ/DELβγ} (βγ KO mice) (Fig. 4C). Deleting the β and γ isoforms of SH2B1 resulted in modestly increased levels (87 ± 26%, mean ± standard error of the mean [SEM], n = 5, P < .05) of a protein that comigrates with SH2B1α in the brain. In liver and pancreas, a protein that comigrates with SH2B1α appeared, consistent with some degree of isoform switching in the βγ KO mice. These latter results indicate that in the absence of exon 9, the cells in at least some tissues switch to making some α isoform.

Figure 4.

Generation and validation of whole-body SH2B1βγ KO mouse model. (A) Schematic showing the Sh2b1 gene with exon 9 deleted due to nonhomologous enjoining (NHEJ) occurring as a byproduct of CRISPR/Cas9 editing designed to insert loxP sites flanking Exon 9. The arrow denotes the location of the NHEJ. (B) Genomic DNA was purified from mouse tails and analyzed by PCR using primers indicated in Panels 1B, 4A, and Table 3. (C) Proteins in lysates of whole brain, pancreas, or liver from WT or SH2B1 βγ KO (18 weeks old) littermates or SH2B1 KO mice were immunoblotted with αSH2B1 and reprobed with αβ-actin as indicated. The migration of molecular weight standards is shown on the right. The expected migrations of the different isoforms of SH2B1 and β-actin are indicated on the left. IB, immunoblot; n.s., nonspecific band.

Deleting SH2B1β and γ Throughout the Body Modestly Increases Body Weight, Adiposity, and Food Intake in Male Mice

We next assessed whether deleting SH2B1β and γ isoforms in all tissues results in an obese phenotype. We fed both male and female βγ KO mice and their WT littermates standard chow and measured their body weight weekly from weeks 4 to 19 (males) or 4 to 16 (females). Male βγ KO mice gained modestly more weight than their WT littermates starting around 10 weeks of age with statistically significant increases detected by 15 weeks (Fig. 5A); female βγ KO mice did not differ in weight from their WT littermates (Fig. 5B). Perigonadal white adipose tissue weight was increased to a statistically significant extent in male (Fig. 5C), but not female (Fig. 5D), βγ KO mice. There was no statistically significant difference in the weight of the brain, liver, pancreas, or quadriceps femoris muscle in male or female mice (Fig. 5C and 5D). Consistent with their increased adiposity, male βγ KO mice exhibited elevated levels of leptin (Fig. 5E); female βγ KO mice showed no statistically significant difference in leptin levels compared to their WT littermates (Fig. 5F).

Figure 5.

Male SH2B1βγ KO mice exhibit modest increases in body weight, perigonadal fat, leptin, and food uptake. (A) Body weight of male mice was measured weekly from 4 to 19 weeks (n: WT = 12, βγ KO = 15). Statistical differences were assessed by linear regression analysis of weeks 6 to 16 (linear portion of the curves). (B) Body weight of female mice was measured weekly from 4 to 16 weeks (n: WT = 10, βγ KO = 11). (C) Tissues of 20- to 24-week-old male mice were dissected and weighed (n: WT, βγ KO = 12 or 13 for perigonadal white adipose tissue (adipose), brain, liver; 9 or 10 for pancreas, quadriceps femoris muscle). (D) Tissues of 20- to 24-week-old female mice were dissected and weighed (n: WT = 9-11; βγ KO = 12-14). Statistical differences were assessed using unpaired 2-tailed t-test. Two Rout outliers were removed, 1 each from male WT brain and female βγ KO adipose. (E, F) Blood was collected between 9 Am and 11 Am from (E) 21- to 32-week-old male mice (n = 3) or (F) 21- to 33-week-old female mice (n = 7) fed ad libitum. Serum leptin levels were measured by enzyme-linked immunosorbent assay. Means and SEM are shown. Statistical differences were assessed using unpaired 2-tailed t-test, *P < .05. (G, H) Food intake was measured weekly in (G) male mice from 4 to 19 weeks (n: WT = 11, βγ KO = 14) and (H) female mice from 4 to 15 weeks (n: WT = 11, βγ KO = 12). Statistical differences were assessed by linear regression. For all panels, means and SEM are shown, *P < .05.

To gain insight into whether the increased body weight might be due to increased food consumption, we measured cumulative food intake weekly from 4 to 19 weeks. Compared with WT littermates, male βγ KO mice consumed modestly more food than their littermates (Fig. 5G), suggesting that the increased body weight of the males was due, at least in part, to increased food intake. Consistent with their lack of increased weight gain, female βγ KO mice did not exhibit altered food intake (Fig. 5H).

Glucose and Insulin Sensitivity Are Normal in Mice Lacking SH2B1β and γ

Impaired glucose homeostasis is detected in SH2B1 KO mice (35), which, like their obesity, is largely reversed by transgenically overexpressing SH2B1β in neurons via the neuron-specific enolase promoter (12). Furthermore, a subset of humans with SH2B1 variants have higher insulin resistance than expected for their degree of obesity (6, 8). We therefore examined whether deletion of the β and γ isoforms of SH2B1 in mice would impair glucose homeostasis by performing both oral glucose tolerance tests and insulin tolerance tests. While a trend toward impaired glucose tolerance was detected in male βγ KO mice (Fig. 6A), the impairment did not achieve statistical significance. An insulin injection did not suppress blood glucose in male βγ KO mice (Fig. 6C). Female βγ KO mice were similar to their WT littermates for both of these measures (Fig. 6B and 6D). Insulin levels in ad libitum–fed male and female βγ KO mice were modestly elevated, although only the elevation in females achieved statistical significance (Fig. 6E and 6F). These results suggest that glucose homeostasis may be modestly impaired in βγ KO mice, but that neither the β nor the γ isoform of SH2B1 per se is essential for glucose homeostasis. Our finding of some isoform switching occurring when the β and γ isoforms of SH2B1 are deleted raises the possibility that any contribution of non-neuronal β/γ isoforms to glucose homeostasis may be compensated for by the newly expressed α and/or δ isoforms.

Figure 6.

SH2B1βγ KO mice show no alteration in glucose homeostasis. (A, B) Glucose tolerance tests (GTTs) were performed on (A) male (n: WT = 11, βγ KO = 11) and (B) female (n: WT = 13, βγ KO = 15) 17.5- to 22-week-old mice following IP injection with glucose (2 mg/kg body weight). Area under the curve for each animal GTT was calculated using a baseline of y = 0. (C, D) Insulin tolerance tests (ITTs) were performed on (C) male (n: WT = 9, βγ KO = 7) and (D) female (n: WT = 9, βγ KO = 11) 18.5- to 23-week-old mice following IP injection with insulin (1 IU/kg body weight). Inverse area under the curve (AUC) for each animal ITT was calculated using a baseline of y = 100. Statistical differences in AUC (GTT) and inverse AUC (ITT) were assessed using unpaired 2-tailed t-test. For all panels, means and SEM are shown, *P < .05. (E, F) Serum insulin levels were assessed by ELISA in blood collected from the same (E) male and (F) female mice used in Fig. 5E and 5F. Means and SEM are shown. Statistical differences were assessed using unpaired 2-tailed t-test, *P < .05.

Discussion

In this work, we investigated the role of the β and γ isoforms of SH2B1 in regulating energy balance. We found that deleting the β and γ isoforms of SH2B1 in neurons of male or female βγ NKO mice had no effect on body weight or adiposity. However, deletion of the β and γ isoforms of SH2B1 in all tissues of the total SH2B1βγ KO mice caused a modest increase in body weight, adiposity, and food intake in male mice. The modest increase contrasts with the lean phenotype observed in previous studies when the largely brain-specific α and δ isoforms of SH2B1 are deleted (25) and the dramatic early-onset obesity in mice seen when all isoforms of SH2B1 are deleted (SH2B1 KO mice) (11). The finding that male βγ KO mice are slightly obese suggests that metabolic signaling downstream of the β/γ isoforms decreases body weight. In addition, our finding that tissues that express only the β and γ isoforms of SH2B1 in WT mice switch to expressing the α isoform in SH2B1βγ KO mice suggests that the effect of eliminating specific isoforms might be diminished by compensation. However, the finding here that SH2B1βγ KO mice are modestly obese when combined with the previous finding that SH2B1αδ KO mice are lean raises the possibility that the α/δ and β/γ isoforms each make unique contributions to the regulation of energy balance.

Interestingly, the modest increase in weight, adiposity, and food consumption seen in male SH2B1βγ KO mice was not seen in female SH2B1βγ KO mice. When exposed to diets of varying fat content, male mice are known to have a greater propensity to gain weight and become obese than female mice, a difference that is lost with ovariectomy (36). The greater weight and adiposity in male mice is thought to be due at least in part to greater sensitivity to leptin (37), a hormone whose responses are enhanced by SH2B1 (11). It would be interesting to determine whether the difference in effects between male and female SH2B1βγ KO mice on weight, adiposity, and food intake are mediated indirectly by an effect of estrogen on leptin sensitivity or by some other mechanism and whether estrogen differentially affects the expression or effectiveness of the different isoforms of SH2B1.

In support of specific SH2B1 isoforms being able to compensate for the absence of other isoforms, the different isoforms are known to share some cellular functions relevant to energy balance. All 4 isoforms share an SH2 domain (23, 24) that enable them to be recruited to activated neurotrophic factor receptors (TrkA, TrkB, Ret) and leptin receptor–activated JAK2 (15-17, 19, 21, 38). Additionally, all 4 isoforms increase dendritic branching of hippocampal neurons as assessed by Sholl analysis (22). The β, γ, and δ isoforms of SH2B1 promote NGF-induced neurite outgrowth using the PC12 cell neuronal differentiation model (16, 17, 22) and SH2B1β and δ have both been shown to promote NGF-induced gene expression in PC12 cells and BDNF-induced gene expression in primary hippocampal neurons (20, 22). In addition, in non-neuronal cells, all 4 isoforms enhance a variety of functions, including cell proliferation in response to various growth factors (insulin-like growth factor-1, insulin, platelet-derived growth factor) (39) and insulin receptor catalytic activity (32).

However, other studies have revealed differences in the cellular actions of the different SH2B1 isoforms. In contrast to SH2B1β, γ and δ, SH2B1α does not stimulate NGF-induced neurite outgrowth or gene expression in PC12 cells (40) (data not shown). In fact, coexpression of SH2B1α inhibits the ability of SH2B1β to enhance NGF-induced neurite outgrowth. Mutation of a single tyrosine that is a substrate of TrkA and unique to the C-terminal tail of SH2B1α restores the ability of SH2B1α to stimulate NGF-induced neurite outgrowth and gene expression, indicating that the stimulatory effect of SH2B1α is regulatable. The δ isoform is the only SH2B1 isoform that localizes to nucleoli, and in isolated hippocampal neurons δ is the only isoform found to increase total neurite length to a statistically significant extent (22). Additionally, rare SH2B1 variants have been identified in children and adolescents with severe obesity, hyperphagia, and in some instances, insulin insensitivity greater than expected for the degree of obesity (6-8). Most of these variants are found in the N-terminal 631 amino acid residues shared by all isoforms of SH2B1 (eg, P322S, P90H, T546A) but others are found in the C-termini unique to specific isoforms (eg, R680C in δ/A663V in α, A723V in α, G638R in β). The finding that rare variants associated with extreme obesity are found in the unique C-termini in α, δ, and β isoforms suggest that the unique C-termini of these isoforms play important roles in metabolic signaling. Indeed, using a hippocampal neuron architecture assay (Sholl analysis) to assess the effect of human SH2B1 variants in the context of SH2B1δ, the R680C variant in the unique C-terminal tail of SH2B1δ increased dendritic complexity, whereas the P90H and P322S variants within the amino acid residues shared by all isoforms modestly decreased dendritic complexity (22).

While it has been known that the β and γ isoforms are ubiquitously expressed in tissues throughout the body while the α and δ isoforms of SH2B1 are expressed primarily in the brain, our analysis of the postnatal day 7 brain region– and cell type–specific isoform atlas data of Joglekar et al (34) reveals the surprising finding that SH2B1β and γ are not expressed in neurons in the hippocampus or prefrontal cortex to any great extent. Rather, only the α and δ isoforms are expressed in neurons while the β and γ isoforms are expressed primarily in various populations of glial cells. In other studies, SH2B1β and/or γ mRNA and/or protein have been detected in all regions of the brain examined (6, 25), as well as in cultured hippocampal neurons isolated from E16 rats (41) and P0-P1 mice (22). However, in these latter studies, it is not clear whether the β/γ isoforms resided in neurons or primarily in non-neuronal cells present in the preparations. Because the mice used in the Joglekar study and the current study have a similar background (C57BL/6), we think it likely that the isoform distribution in our mice is similar to that reported by Joglekar et al (34). Furthermore, our finding that deleting the β and γ isoforms specifically in neurons using cre recombinase under the control of the highly neuron-specific synapsin 1 promoter does not affect body weight or adiposity would tend to support the relative absence of the β and γ isoforms in neurons.

A finding that neurons express unique isoforms of SH2B1 would not be surprising given that alternative splicing of precursor mRNAs is particularly widespread in the nervous system (42). Neuron-specific splicing factors such as Nova2 have been identified (43). Additionally, mutations in neural RNA-binding proteins that cause aberrations in neural alternative splicing patterns have been linked to multiple neurological disorders and diseases (44). Interestingly, in the SH2B1β/γ KO mice that do not express the β and γ isoforms of SH2B1, an SH2B1 antibody recognizes a protein that comigrates with SH2B1α on sodium dodecyl sulfate-polyacrylamide gel electrophoresis gels (Fig. 4D). This finding suggests that while peripheral tissues preferentially produce SH2B1β/γ (6), they can make SH2B1α. In contrast, we did not detect a compensatory increase in the levels of the β and γ isoforms of SH2B1 in the brains of SH2B1α/δ KO mice (25). Because decreasing the ratio of SH2B1α/δ to SH2B1β/γ appears to favor leanness, it will be important to identify the mechanism by which Sh2b1 precursor mRNA is spliced and whether splicing can be artificially regulated to favor the β/γ isoforms in obese individuals or the α/δ isoforms in anorexic individuals. One can also envision interventions that target specific isoforms or the stability of their mRNA.

Our finding, that deleting SH2B1β/γ throughout the body causes modest weight gain but deleting SH2B1β/γ only in neurons has no effect, suggests a role for SH2B1β/γ in non-neuronal cells. The finding in Fig. 3 that although β/γ isoforms are relatively abundant, they are primarily expressed in glial cells suggests a possible role for glial cell SH2B1β/γ in regulating body weight. Glial cells make up ∼35% and 50% of cells in the brain of mice and humans respectively (33, 45). Our RNA seq analysis of hypothalami isolated from SH2B1α/δ KO mice determined that deleting SH2B1 α/δ had the greatest impact on genes associated with immune system cells, including microglia. Most of these microglia-related genes contribute to complement-mediated synaptic pruning (46-50), raising the possibility that microglia contribute to the lean phenotype of SH2B1αδ KO mice via complement-mediated synaptic pruning of appetite-regulating neuronal synapses. Other studies have suggested a role for microglia in feeding behavior and the regulation of body weight. One of the more compelling is the work of Gao et al (51), who found that male mice lacking LepR specifically in myeloid cells, including microglia and macrophages, showed increased body weight and food intake, and a decrease in projections from proopiomelanocortin neurons to the paraventricular nucleus. Proopiomelanocortin neurons are leptin-responsive neurons that, when activated, suppress appetite (52). In further support of non-neuronal SH2B1 having a positive impact on neural circuits in the brain, Miyamoto et al (53) recently reported that deleting SH2B1 specifically from Schwann cells in mice reduces myelin segment numbers, decreases myelin thickness, and decreases the percentage of myelinated axons in Schwann cell-neuron co-cultures. SH2B1 is believed to have these effects by increasing the activity of the guanine nucleotide exchange factor protein cytohesin-2 by binding it and thereby protecting it from dephosphorylation by PTP4A1. Increasing the activity of cytokesin-2 would be consistent with reports that SH2B1 may regulate the actin cytoskeleton at least in part by interacting with small GTPases and/or their guanine nucleotide exchange factor and GTPase-activating protein regulators (54).

Our findings that the βγ NKO mice are not obese and that SH2B1β/γ are largely excluded from neurons raises the question of why transgenic expression of SH2B1β driven by the neuron-specific enolase promoter in SH2B1 KO mice restores the lean phenotype (12). One possibility is that some of the restorative effect detected following transgenic expression of SH2B1β driven by the neuron-specific enolase promoter in SH2B1 KO mice is mediated via SH2B1β expression in non-neuronal cell types. Indeed, Kugler et al (55) reported that when adenovirus encoding enhanced green fluorescence protein (eGFP) driven by the human neuron-specific enolase promoter is injected into the striatum of rats and eGFP expression is assessed 2.5 months later, about a third of the eGFP is expressed in non-neuronal cells. Furthermore, in in vitro studies, non-neuronal cells (eg, astrocytes) were found to express reporters driven by the neuron-specific enolase promoter (56). In contrast, eGFP fluorescence was detected exclusively in neurons after transduction with adenovirus encoding eGFP driven by the Synapsin 1 promoter (55). Thus, in the βγ NKO mice, which use the Synapsin1 promoter to drive the Cre, the deletion of the β/γ isoforms is likely to be restricted to the neurons, while in the SH2B1 KO mice, the SH2B1β driven by the enolase promoter may be able to restore a lean phenotype to SH2B1 KO mice because of expression in non-neuronal cell types. An alternative possibility is that whatever functionality is shared by the 4 different isoforms is sufficient for transgenic expression of SH2B1β to compensate for the lack of the other 3 isoforms. In other words, the important factor is that some SH2B1 is present in the neurons. However, if the α/δ isoforms have the same function as the β/γ isoforms regarding energy balance, it is not clear why in the SH2B1αδ KO mice deleting what appears to be all of the neuronal SH2B1 without compensation results in lean rather than obese mice.

The findings that the SH2B1αδ KO mice are lean while the SH2B1 KO and SH2B1βγ KO mice are obese would seem to indicate that the α/δ isoforms have some different effects from the β/γ isoforms regarding energy balance. It is therefore possible that multiple isoforms may be needed in the right ratio in the right cells at the right developmental stage for optimal body weight. In support of this idea, adding back any single isoform in excess was insufficient to overcome the decrease in the length of the longest neurite observed when all 4 isoforms were absent from cultured hippocampal neurons isolated from SH2B1 KO mice (22). Since the longest neurite is thought to reflect axon length, a combination of at least 2, and perhaps all 4, isoforms may be required to obtain the right number and types of neuronal connections that regulate feeding behavior and/or energy expenditure.

Summary

In summary, our finding that mice lacking the β and γ isoforms of SH2B1 specifically in neurons have normal energy balance contradicts the prevailing notion that neuronal β/γ isoforms play a critical role in energy balance. Instead, our findings here with the βγ KO and βγ NKO mice, combined with previous findings regarding the αδ KO and total KO mice, indicate that the full complement of SH2B1 isoforms is required for body weight maintenance. They also indicate that the most likely reason mice lacking neuronal SH2B1β and γ exhibit unaltered body weight is because neurons express predominantly the α and δ isoforms of SH2B1. Therefore, the obesity shown when β/γ isoforms are deleted throughout the body suggest that SH2B1 in non-neuronal cells (eg, neuroglial cells) plays an important role in regulating energy balance. Our results also show that non-neuronal cells may be able to compensate for the lack of SH2B1β/γ by isoform switching to the α/δ isoforms. Finally, our studies highlight the importance of determining SH2B1 isoform expression and function in the multiple regions of the brain implicated in the regulation of body weight and in the various cell types (eg, leptin, BDNF-receptor and MC4R-expressing neurons, glia) found within those regions.

Abbreviations

BDNF

brain-derived neurotrophic factor

βγ fl/fl mice

Sh2b1βγ^fl/fl mice

βγ NKO mice

Syn1-Cre; Sh2b1βγ^fl/fl mice

GDNF

glial cell–derived neurotrophic factor

IP

intraperitoneal

KO

knockout

NGF

nerve growth factor

OPC

oligodendrocyte progenitor cells

PCR

polymerase chain reaction

SH2

Src Homology 2

WT

wild type

Funding

This work was supported by the National Institutes of Health (R01-DK-054222 and R01-DK-107730 to C.C.-S.; R01-DK-056731 to M.G.M.) and a predoctoral fellowship from the Horace H. Rackham School of Graduate Studies, University of Michigan (to J.L.C.). Mouse body composition was partially supported by the Michigan Mouse Metabolic Phenotyping Center, University of Michigan (NIH U2C-DK-110678). Generation of the CRISPR mice was partially supported by the Molecular Genomics Core of the Michigan Diabetes Research Center, University of Michigan (NIH P30-DK-020572).

Author Contributions

Conceptualization: L.S.A., C.C.-S., M.J.M; Methodology: L.S.A., A.F., J.L.C.; Validation: C.C.-S., L.S.A.; Formal analysis: L.S.A., C.C.-S.; Investigation: A.F., N.S., M.E., J.M.C., C.R., L.S.A., J.L.C.; Data curation: L.S.A.; Writing—original draft: C.C.-S.; Writing—review & editing: C.C.S., L.S.A., M.G.M.; Visualization, supervision, and project administration: L.S.A., C.C.-S; Funding acquisition: C.C.-S., L.S.A., M.G.M.

Disclosures

Each author responded to Dr. Carter-Su that they had no conflict of interest.

Data Availability

Some or all datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.

Current affiliations

A.F.: Department of Biology, California Baptist University, Riverside, CA, 92504, USA. J.L.C.: Department of Cell Biology & Physiology, University of North Carolina School of Medicine, Chapel Hill, NC 27599, USA.