Truncation of Pik3r1 causes severe insulin resistance uncoupled from obesity and dyslipidaemia by increased energy expenditure

Objective Insulin signalling via phosphoinositide 3-kinase (PI3K) requires PIK3R1-encoded regulatory subunits. C-terminal PIK3R1 mutations cause SHORT syndrome, as well as lipodystrophy and insulin resistance (IR), surprisingly without fatty liver or metabolic dyslipidaemia. We sought to investigate this discordance. Methods The human pathogenic Pik3r1 Y657∗ mutation was knocked into mice by homologous recombination. Growth, body composition, bioenergetic and metabolic profiles were investigated on chow and high-fat diet (HFD). We examined adipose and liver histology, and assessed liver responses to fasting and refeeding transcriptomically. Results Like humans with SHORT syndrome, Pik3r1WT/Y657∗ mice were small with severe IR, and adipose expansion on HFD was markedly reduced. Also as in humans, plasma lipid concentrations were low, and insulin-stimulated hepatic lipogenesis was not increased despite hyperinsulinemia. At odds with lipodystrophy, however, no adipocyte hypertrophy nor adipose inflammation was found. Liver lipogenic gene expression was not significantly altered, and unbiased transcriptomics showed only minor changes, including evidence of reduced endoplasmic reticulum stress in the fed state and diminished Rictor-dependent transcription on fasting. Increased energy expenditure, which was not explained by hyperglycaemia nor intestinal malabsorption, provided an alternative explanation for the uncoupling of IR from dyslipidaemia. Conclusions Pik3r1 dysfunction in mice phenocopies the IR and reduced adiposity without lipotoxicity of human SHORT syndrome. Decreased adiposity may not reflect bona fide lipodystrophy, but rather, increased energy expenditure, and we suggest that further study of brown adipose tissue in both humans and mice is warranted.


INTRODUCTION
Adipose tissue is essential for metabolic health in the face of sustained positive energy balance. When adipose expansion is constrained, as in lipodystrophy, its capacity to sequester excess energy and serve as a physiological "energy buffer" is overwhelmed [1]. Injury and inflammation of any residual adipose tissue ensues, causing systemic inflammation and rerouting substrates to the liver, pancreas and muscle. Insulin resistance (IR), diabetes, aggressive fatty liver disease, and severely elevated plasma lipoprotein concentrations are thus common consequences of lipodystrophy [2]. SHORT syndrome (short stature, hyperextensibility of joints, ocular depression, Rieger anomaly of the iris, and teething delay) is a human monogenic syndrome including lipodystrophy [3e5]. This form of lipodystrophy is highly unusual -though associated with insulin resistant diabetes [3e5], this is uncoupled from fatty liver and dyslipidaemia [6]. SHORT syndrome also does not feature the suppressed plasma adiponectin concentration commonly seen in IR and other 1 lipodystrophies [7], proving that some elements of the "insulin resistance syndrome" are dissociable. SHORT syndrome is caused by mutations in PIK3R1; encoding components of the insulin signalling enzyme phosphoinositide 3-kinase (PI3K) [3e5]. The PI3K subtype required for insulin action comprises a p110a catalytic subunit bound to a regulatory subunit, three (p85a, p55a and p50a) of which are encoded by PIK3R1. PIK3R1 products stabilise catalytic subunits, mediating their recruitment to receptor tyrosine kinases or their substrates, including the insulin receptor and IRS1 [8,9]. Before the discovery of PIK3R1 mutations in SHORT syndrome, Pik3r1 had been intensively studied in mice. Selective p85a deficiency [10,11] or deletion of p50a and p55a [12] surprisingly enhances insulin sensitivity, which has been accounted for by compensation by other regulatory subunits, and/or altered stoichiometry of catalytic and regulatory subunits. A critical insight from the study of SHORT syndrome was that the insulin signalling role of PIK3R1 could be severely attenuated by neomorphic mutations disrupting the C-terminal SH2 domain [3e5]. Such dominant negative PIK3R1 alleles, which severely reduce insulininduced activation of PI3K without abolishing the expression of PI3K, offer the opportunity to interrogate, more precisely than has been possible to date, the in vivo pathophysiological roles of PI3K activity in IR. In human lipodystrophy, de novo lipogenesis contributes significantly to liver triglyceride and plasma lipid content [13]. In contrast, mutations in the INSR gene encoding the insulin receptor produce extreme IR with normal plasma lipid profile, normal liver lipid content, and no increase in hepatic de novo lipogenesis [13]. Furthermore, murine models with knockout of any one of several components of the insulin signalling pathway (Insr [14], Irs1 and Irs2 [15], p85a and b [16], Pik3ca [17], or Rictor [18]) demonstrate that liver insulin signalling drives hepatic de novo lipogenesis. We thus hypothesised that in SHORT syndrome, the liver is protected from increased insulin-mediated lipogenesis, which is relatively unaffected by common IR [19], and that this explains the absence of dyslipidaemia and fatty liver despite severe IR and lipodystrophy. We report a novel murine SHORT syndrome model caused by a human pathogenic allele, Y657*, that reproduces uncoupling of metabolic dyslipidaemia from severe IR. Surprisingly, hepatic lipogenic transcriptional programmes were minimally perturbed, although metabolic inflexibility with a dyscoordinated fasting response was seen. Adipose tissue accumulation was reduced on high-fat feeding. We suggest that this is related to increased energy expenditure, likely via brown adipose tissue, rather than lipodystrophy, with hypolipidemia attributable to increased fuel oxidation. This potential explanation for the unusual IR subphenotype of SHORT syndrome warrants testing in humans with SHORT syndrome or other proximal insulin signalling defects.

Mice generation and maintenance
The Pik3r1 Y657* mutation and a neomycin resistance cassette flanked by LoxP sites were introduced into C57Bl/6 embryonic stem cells by homologous recombination before injection into Bl/6J blastocysts (Supplemental Figure S1). Animals were kept on a C57Bl/6J background, backcrossed at least three times, and housed on a 12-hlight/12-h-dark cycle at 23 C, with ad libitum food and water access. Feeding was with chow (Catalogue no. 105, Safe diets) or 45% fat diet (45% fat, 35% Carbohydrate and 25% protein) (Catalogue no. D12451, Research diet, Inc). Experiments were carried out under the UK Home Office Animals (Scientific Procedures) Act 1986, following University of Cambridge ethical review.
2.2. Assessment of growth, body composition, and energy homeostasis Body composition was determined by time-domain NMR with Bruker's minispec analyser. For food intake and energy expenditure measurement, male mice were acclimatized for 1 week to single housing then given chow or 45% fat diet for 10 days. Food intake was measured daily and energy expenditure measured by indirect calorimetry [20]. Locomotor activity was quantified as beam breaks over 48 h. Faecal energy content was measured by combusting dried faeces from 48 h calorimeter runs in an IKA Calorimeters Oxygen Bomb calorimeter (IKA C1).

In vivo metabolic studies
In fasting/refeeding studies, mice were fed chow ad libitum until 16 weeks of age before 16 h fasting, then 6 h refeeding with chow. Tissues were harvested and snap frozen prior to overnight fasting, or after fasting or after refeeding. For the study of insulin action, 16 hfasted mice (16 weeks old) were injected with 2 U/kg insulin intraperitoneally and tissues collected after 10 min and snap frozen. Glucose tolerance testing was performed on overnight (16 h) fasted mice using 2 g/kg 20% glucose via oral gavage. Insulin tolerance testing was done by giving 0.5 U/kg intraperitoneal insulin to 6 hfasted mice. For lipid tolerance testing, 16 h-fasted mice were administered 10 mL/kg olive oil by gavage. Hyperinsulinemic euglycemic clamp studies and subsequent sample processing and analysis were conducted as described previously [21] on 16 week old mice after overnight fasting. The clamps used a priming dose of human insulin (0.7 mU), followed by constant insulin and [3-3 H]-D-glucose infusion at 7 mU/h and 0.72 mCi/h, respectively.

Study of tissue insulin responsiveness ex vivo
Measurement of insulin action on isolated Soleus and Extensor Digitorum Longus was undertaken in 16-week-old male mice as described [22]. For primary preadipocyte differentiation studies, fat pads from 8 week-old male mice were isolated and cultured as described [20], but without 5-Triiodo-L-thyronine.

Biochemical assays
Blood was collected by cardiac puncture and plasma snap frozen before assays detailed in Supplemental Table 3. Glycogen hydrolysis in 1 M HCl (2.5 h at 100 C), NaOH neutralization and fluorometric glucose analysis were used to determine the glycogen content of snapfrozen liver [23]. Hepatic triglyceride determination was used in the Triglyceride Colorimetric Assay kit (Cayman). Hepatic cholesterol was extracted from 10 mg of liver using a 400 mL chloroform:isopropanol:NP-40 mixture (7:11:0.1) before centrifugation, 30 min vacuum drying (50 C) and dissolving in phosphatebuffered saline (PBS). The samples were assayed by Siemens Healthcare Diagnostics kit.
2.6. Metabolomic studies Liver metabolites were extracted from tissue snap-frozen after dissection using a reported methanol/chloroform method [24]. Aqueous metabolites were further extracted as described in [24] and analysed using a Vanquish ultra-high performance liquid chromatography (UHPLC) system and TSQ Quantiva triple quadrupole mass spectrometer (Thermo Scientific) with compounds directly infused.
Parameter optimisation used 1 mM standard solutions in a chromatographic buffer. Optimal mass spectrometry parameters and mass transitions were generated by automatic MassLynxÔ (Version 1.4, Waters) protocols or, if standards were unavailable, deduced from known analogue parameters.

Histology
The sizes of at least 1000 adipocytes per genotype from 16-week-old chow-fed mice were measured in Formalin-fixed, paraffin-embedded (FFPE) tissue as described [26]. Adipose Treg cells were isolated and counted by FACS, also as reported [27].

Gene expression analysis
For protein studies fresh tissues were snap-frozen in liquid nitrogen, and homogenised using MD ceramic beads in RIPA buffer with proteinase/phosphatase inhibitors (liver, muscle) (Roche) or mortar and pestle in liquid nitrogen before dissolving in buffer (adipose). Protein concentrations were determined by BCA assay (BioRad). Western blotting employed the Novex gel system (Thermo Fisher Scientific). Antibodies are shown in Supplemental Table S4. Quantification employed the BioRad image system. Total RNA was extracted and reverse transcribed for qPCR as previously described [28]. Target expression was normalised to the geometric mean of 4 housekeeping genes (Ywhaz, Ppia, B2m and Eef1a1) (Supplemental Table S4 for primer/probe sequences). RNA sequencing protocols and analysis have previously been reported [29].
2.10. Statistical analysis D'Agostino & Pearson or ShapiroeWilk tests were performed to test for Gaussian distribution of data. For Gaussian data, unpaired two-tailed Student's t tests for two group comparison and ANOVA with post hoc for multiple group comparison were performed using GraphPad Prism (GraphPad Software). For non-Gaussian data ManneWhitney testing was used for two group comparison and KruskaleWallis and Dunn's multiple comparison testing were used for multiple group analysis using GraphPad Prism (GraphPad Software). Tissue weights, food consumption and energy expenditure were analysed by ANCOVA using XLSTAT (Addinsoft).

Data and resource availability
All transcriptomic data were deposited in GEO under accession number GSE153431. Other datasets, and the Pik3r1 WT/Y657 * mice generated during the current study are available from the corresponding author upon reasonable request.

Growth and development of Pik3r1 Y657* knock-in mice
Mice harbouring the Pik3r1 Y657* mutation were generated by homologous recombination-based gene targeting ( Figure S1A). Immunoblotting confirmed the expression of truncated p85a at greater amounts than full length protein in most tissues, likely due to the loss of an ubiquitylation motif located beyond the nonsense mutation in the C-terminal [30]. No compensatory change in PI3K p85b subunit was seen. p110a subunit expression was reduced in subcutaneous white adipose tissue, and p110b expression was reduced in liver, subcutaneous and epididymal white adipose tissue ( Figure S1B).
No Pik3r1 Y657 * /Y657 * embryos were identified beyond E11.5. At E11.5 they were smaller, with poorly developed limb buds and reduced eye pigmentation. Heterozygous embryos were smaller from E15.5 ( Figure S2AeE). Pik3r1 WT/Y657 * mice were born at the expected frequency, but showed impaired linear growth and body weight on chow (Figure 1AeC). At 16 weeks, there was a small decrease in whole body adiposity ( Figure 1D), however, no difference was found in epididymal or inguinal white adipose nor interscapular brown adipose depot weights when lean mass was taken into account ( Figure 1EeG). Plasma leptin concentrations were similar in male Pik3r1 WT/Y657 * and wild-type mice (3.3 AE 1.2 vs 2.6 AE 0.3 mg/L on fasting (n ¼ 9,10; NS); 10.3 AE 2.2 vs 15.9 AE 3.9 mg/L on ad libitum feeding (n ¼ 13,13; NS)).
Liver weights were indistinguishable ( Figure 1H), but hearts were heavier in Pik3r1 WT/Y657 * mice ( Figure 1I). No difference in food consumption nor energy expenditure was seen at 16 weeks on chow ( Figure 1J,K), however, the respiratory exchange ratio (RER) was higher during the light phase in Pik3r1 WT/Y657 * mice, blunting the circadian fluctuation ( Figure 1L). Similar differences were seen between heterozygous and wild-type females for all variables assessed in both sexes.

3.2.
Response of Pik3r1 WT/Y657 * mice to a high-fat diet To assess whether placing a greater load on adipose tissue in Pik3r1 WT/Y657 * mice would unmask lipodystrophy, mice were fed a 45% fat diet (HFD) from 8 weeks old. Pik3r1 WT =Y 657 * mice showed reduced weight gain and body fat over the 8 weeks of study ( Figure 2A,B). Epididymal white adipose tissue was most severely affected ( Figure 2C), with no significant difference in inguinal nor brown adipose tissue ( Figure 2D,E). Lean mass did not increase at a greater rate in Pik3r1 WT/Y657 * mice ( Figure S3A,B), and liver weights also did not differ, while heart weights were elevated in heterozygotes, as on chow ( Figure 2F,G). Consistent with reduced adiposity, Pik3r1 WT/Y657 * mice had lower serum leptin concentrations when fed (37.7 AE 7.3 vs 11.3 AE 2.0 mg/L (n ¼ 15,11; p ¼ 3 Â 10 À3 )), but not when fasted (15.9 AE 3.9 vs 10.3 AE 2.2 mg/L (n ¼ 13,13; not significant)), while adiponectin concentrations were unchanged (Table 1). Given reduced adipose expansion, we subsequently sought histological evidence of "adipose failure", namely adipocyte hypertrophy and inflammation, which would be expected in lipodystrophy when adipose expansion is pathologically constrained, causing adipose injury on positive energy balance. The histological appearance and adipocyte size distribution in epididymal adipose tissue was indistinguishable between genotypes ( Figure 2H,I), however with no evidence of inflammation, nor fibrosis. Consistent with the lack of histological evidence of adipose overload, ex vivo differentiation of preadipocytes from inguinal or epididymal depots was unaltered ( Figure 2J). Finally, given the role of Pik3r1 in lymphocyte function, and the important role of adipose-resident T regulatory (T reg ) cells in energy homeostasis, the adipose content of T reg cells was assessed by FACS. No major difference was observed between genotypes ( Figure 2K). Given these findings, which are at odds with a true lipodystrophy, we then assessed whether reduced adipose accretion rather reflected altered energy homeostasis. Pik3r1 WT/Y657 * mice exhibited increased food intake ( Figure 2L) and energy expenditure ( Figure 2M) when the size of the animals was taken into account, but locomotor activity was unchanged across 24 h ( Figure 2N). Markedly reduced accumulation of adipose tissue despite increased food intake, and without sustained hyperglycaemia, demonstrates severely reduced food efficiency. This indicates that increased food intake is a compensatory response to chronically increased energy expenditure rather than a primary phenomenon. As on chow, the circadian RER profile was blunted, with higher values in the dark phase and no discernible difference between light and dark phases for Pik3r1 WT/Y657 * mice ( Figure 2O), suggesting greater carbohydrate oxidation in the dark phase.
3.3. Pik3r1 WT/Y657 * mice are severely insulin resistant At 12 weeks old, neither male nor female Pik3r1 WT/Y657 * mice on chow were hyperglycaemic compared to controls, but plasma insulin concentrations were raised, more severely on feeding (Tables 1 and S1).
On high-fat feeding, male Pik3r1 WT/Y657 * mice remained hyperinsulinemia, but the difference between mutant and control mice was abolished by the greater increase in insulin concentrations in wild-type mice. Blood glucose concentration was paradoxically lower in fed Pik3r1 WT/Y657 * mice on HFD. Plasma adiponectin concentrations in fasted chow-fed Pik3r1 WT/Y657 * mice were lower than in wild-type controls, however no significant difference was seen in other conditions, despite the severe IR of heterozygous mice (Table 1). In Pik3r1 WT/Y657 * mice on HFD, adiponectin normalised to body fat was  higher than in controls, reminiscent of the preserved adiponectin seen in severely insulin-resistant patients with SHORT syndrome [7]. Normalised plasma leptin concentrations, in contrast, were lower ( Figure S3C,D). Hyperinsulinemic euglycemic clamps were undertaken on chow-and HFD-fed male mice. Chow-fed Pik3r1 WT/Y657 * mice required a steady state glucose infusion rate of only one eighth that required by wild-type mice, confirming severe IR ( Figure 3A). On HFD, Pik3r1 WT/Y657 * mice remained extremely insulin resistant, however the difference between genotypes was much smaller than on chow due to increased IR in wildtype animals ( Figure 3A). Further studies were thus carried out on chow. Glucose and insulin excursions on oral glucose tolerance testing showed a trend towards an increase in male Pik3r1 WT/Y657 * mice ( Figure 3B,C), but were increased in female mice ( Figure S4A,B), while the hypoglycaemic response to insulin was attenuated in both sexes ( Figures 3D and S4C). Clamp studies in males using isotopic tracers showed glucose disposal to be 19% lower in Pik3r1 WT/Y657 * mice ( Figure 3E) while suppression of hepatic glucose production by hyperinsulinemia was 49% lower compared to wild type littermates ( Figure 3F). Liver glycogen content was similar between genotypes ( Figure 3G). Insulin infusion lowered plasma free fatty acid concentrations in wild-type mice, but little suppression was seen in heterozygous mice, consistent with impaired suppression of adipose lipolysis ( Figure 3H). Intraperitoneal insulin in male mice strongly induced Akt phosphorylation in the liver, skeletal muscle and epididymal and inguinal adipose tissue of wild-type controls, as expected, and this was reduced in Pik3r1 WT/Y657 * mice (Figures 3IeL and S5AeD). As muscle insulin sensitivity was previously reported not to be reduced in Pik3r1 WT/R649W mice [31], insulin responsiveness of soleus and extensor digitorum longus (EDL) muscles was assessed ex vivo. Soleus and EDL muscle from Pik3r1 WT/Y657 * mice both showed a 1.3-fold reduction in insulinstimulated deoxyglucose uptake compared to wild-type muscle ( Figure 3M). Insulin-induced Akt phosphorylation was also markedly reduced in both types of muscle ex vivo to a much more significant degree than in vivo ( Figures 3N,O and S5E). This finding of muscle insulin resistance using several techniques, although contrasting with clamp studies reported previously, is consistent with the robust expression of Pik3r1 gene products in skeletal muscle.
3.4. Pik3r1 WT/Y657 * mice are hypolipidemic Having demonstrated that, as in SHORT syndrome, heterozygous Pik3r1 WT/Y657 * mice show severe IR and reduced adiposity, we assessed whether, like humans, they do not exhibit fatty liver nor metabolic dyslipidaemia. In fed and fasting states, on either chow or HFDs, Pik3r1 WT/Y657 * mice showed hypolipidemia, with lower plasma total cholesterol and HDL cholesterol (Table 1). This was most striking in fed mice on HFD, mirroring the lower glucose seen in Pik3r1 WT/Y657 * mice in this state. Plasma triglyceride concentrations were also lower in heterozygous animals in all states except when the fasting state after HFD. Plasma free fatty acid concentrations were lower in all states except in fed animals maintained on chow. Similar hypolipidemia was seen in chow-fed female mice on fasting (Table S1). No difference was seen in VLDL concentrations in chow-or high fat-fed mice (Table 1). Despite severe IR in Pik3r1 WT/Y657 * mice, no difference was seen in liver triglyceride content compared to controls either on chow or HFD, assessed by Oil-Red-O staining and biochemical quantification ( Figure 4A,B). Liver cholesterol content was also the same in chow-fed animals of both genotypes ( Figure 4C). Oral lipid tolerance testing revealed similar triglyceride excursion after lipid loading in both  Original Article genotypes ( Figure 4D), and bomb calorimetry of faeces on chow showed the same energy content ( Figure 4E), excluding abnormal intestinal lipid handling as the explanation for hypolipidemia and reduced adiposity.
Insulin stimulates hepatic de novo lipogenesis and cholesterogenesis via Srebp1 and Srebp2 respectively. In prevalent IR, hepatic de novo lipogenesis is enhanced, and the mismatch between this and the reduced hypoglycaemic action of insulin is evidence for "partial IR" in metabolic dyslipidaemia [19]. Liver PI3K mediates the action of insulin on lipogenesis, with hypolipidemia despite IR seen in numerous models in which hepatic insulin/Irs/PI3K/Akt2/mTORC2 signalling is impaired [14e16, 18,32]. Based on these findings, we hypothesised that the hypolipidemia seen in Pik3r1 WT/Y657 * mice would be accounted for by reduced hepatic de novo lipogenesis, as in SHORT syndrome due to PIK3R1 Y657* [7]. In keeping with this, no increased de novo synthesis of palmitate could be measured in heterozygotes despite fasting hyperinsulinemia and refeeding with a high carbohydrate diet ( Figure 4F). Liver mRNA expression of Srebp1c, Srebp2 and key target genes (Acc1, Fasn, Scd1 and Mvk, Nsdhl, respectively) was also determined in chow-fed animals in fed, fasting and refed states, and no difference was seen in the expression of any of these genes under any nutritional conditions (Figure 4GeM). Other transcriptional regulators relevant to hypolipidemia include carbohydrate response element binding protein (Chrebp) and Xbp1, with the latter binding p85a and facilitating its nuclear translocation [31,33,34]. A sentinel Chrebp-responsive gene (Pklr) showed mild but significant reduction of mRNA in the fasting state, but mRNA of two Xbp1 target genes (Acacb and Dgat2) showed no reduction during a Original Article 8 fasting refeeding cycle ( Figure S6AeC). It has also been proposed that modulation of the ratio between Glucose-6-phosphatase (G6pc) and Glucokinase (Gck) toggles hepatocyte glucose flux between lipogenesis and glucose production, and that this operates before transcriptional changes in lipogenic transcription factors [35]. However, although an increase in G6pc:Gck liver transcript levels was seen on fasting of wild-type mice, the ratio was suppressed in Pik3r1 WT/Y657 * mice, as reported in mice on Western-type diet [35] ( Figure S6DeF). We conclude that the consistent hypolipidemia seen in Pik3r1 WT/Y657 * mice does not correlate with reduced insulin-induced lipogenesis in the liver, suggesting that the reduced blood lipid concentrations are not explained by attenuation of insulin's lipogenic actions. Given the lack of transcriptional perturbation of known lipogenic factors, global liver transcriptomic analysis was performed in fed and 16 h fasted states. Genotypes were compared in each state, and fasting gene expression was compared to fed gene expression for both genotypes. Gene-based analysis confirmed prior findings, but provided no obvious alternative explanation for hypolipidemia. Pathway analysis of predicted upstream regulators showed modest but significant differences in the fed state, with a transcriptional signature of reduced Endoplasmic Reticulum (ER) stress (upstream regulators Xbp1, its processing enzyme Ern1 (Ire1), and Atf6), and altered signatures of several cytokines including IL6 (reduced), IL21, interferons gamma and alpha 2 (increased) ( Figure 5, Table S2). More significant differences were seen after 16-h fasting, when a transcriptional profile consistent with the reduced activity of Rictor, the defining component of mTORC2, was seen with high significance, while patterns associated with Myc and Mycn were upregulated. The large preponderance of changes observed between fed and fasting states in wild-type mice, including strong upregulation of PPARa-responsive genes, was however conserved in heterozygous animals ( Figure 5, Table S2).
In view of the transcriptomic evidence of reduced Rictor activity, plasma and hepatic amino acid and b-hydroxybutyrate concentrations were determined. This revealed increased b-hydroxybutyrate in the liver from fed Pik3r1 WT/Y657 * mice with a trend towards an increase in plasma, but no difference between genotypes when fasted ( Figure S7A). The fall in plasma amino acid concentrations seen in fasting wild-type mice was lost in Pik3r1 WT/Y657 * mice, with a trend towards a generalised increase in liver amino acids ( Figure S7B,C). These findings add to the evidence for abnormal transitions between fed and fasted states in Pik3r1 WT/Y657 * mice, and suggest that reduced intra-hepatocyte amino acid concentrations on fasting do not explain the pronounced reduction in Rictor signalling suggested by transcriptomics.

DISCUSSION
Most monogenic defects causing severe IR primarily affect adipose tissue development, and metabolically phenocopy common IR [1]. In contrast, humans with severe IR due to proximal insulin signalling Figure 5: Pathway analysis of liver transcriptomes of fed and fasted mice. Volcano plots are shown for predicted upstream regulators derived from IngenuityÒ Pathway Analysis (IPAÒ) of transcriptomes of fed and fasted male Pik3r1 WT/Y657* (n ¼ 6,6) mice and wild-type littermates (n ¼ 6,6). Outside plots show regulators showing differential activity in the fasting state compared to the fed state for wild-type (WT; left) and Pik3r1 WT/Y657* (right) mice. Central plots show regulators with differential activity in Pik3r1 WT/Y657* vs WT mice in the fed state (top) and fasting state (bottom). The green dots represent data point with an activation z score <À2 or >þ2 and a p-value <1 Â 10 À5 . Regulators showing differential activity in either genotype-based comparison are coloured red (activation z score > þ2 in Pik3r1 WT/Y657* vs WT mice) or blue (activation z score < À2 in Pik3r1 WT/Y657* vs WT mice) in the outside plots showing differences based on nutritional state. Statistical analysis was performed using a general linear model with Bonferroni correction. defects do not have the dyslipidaemia, fatty liver, nor decreased plasma adiponectin concentration usual in common IR [6,7,13,36]. This was first noted in people with insulin receptor mutations [13], but more recently, has been described in SHORT syndrome [6,7], in which the lack of fatty liver and dyslipidaemia is particularly striking given concomitant lipodystrophy. Fatty liver and atherogenic dyslipidaemia are major sources of IR-related morbidity, so investigating how these are dissociated from IR in the face of proximal insulin signalling defects will potentially reveal novel approaches to mitigate IR-associated disease.
The novel mice generated, heterozygous for the Pik3r1 Y657* allele, show reduced growth, severe IR, and severely reduced adipose accretion on high-fat feeding. Although IR in Pik3r1 R649W/WT mice described previously was reported to spare skeletal muscle [31], we found skeletal muscle to be insulin resistant both in clamps, and ex vivo, in keeping with the ubiquitous expression of the mutant allele. Despite severe IR, Pik3r1 WT/Y657 * mice were hypolipidemic on chow or HFD, and in fed and fasted states, while the increase in liver triglyceride usual in lipodystrophic IR was absent, validating it as a model of the human phenomenon we aimed to study. Knockout evidence that PI3K signalling mediates insulin's stimulation of liver de novo lipogenesis led us to expect reduced transcript levels of Srebp and its target genes (e.g. [16,17]), however we found no evidence supporting this or other candidate transcriptional mechanisms including reduced lipogenic activity of Xbp1 [34]. RNA sequencing showed most insulin-responsive genes to respond normally to fasting and refeeding. Given elevated plasma insulin concentrations, this could be construed as showing compensated hepatic IR, as expected from a proximal insulin signalling defect. Fasting-responsive programmes, such as those orchestrated by PPARa, were intact in Pik3r1 WT/Y657 * mice, however pathway analysis suggested reduced ER stress signalling and showed altered inflammatory signatures in the fed state, with a strikingly reduced Rictor signature in the fasted state. These transcriptional clues may be pertinent to the protection from dyslipidaemia as ER stress signalling promotes fatty liver and dyslipidaemia [37], while Rictor drives hepatic lipogenesis [38], however further study is needed. Impaired induction of Rictor activity in this case may reflect Pik3r1's role in amino acid activation of Rictor [39]. Pik3r1 R649W/WT mice have been shown to have less subcutaneous adipose tissue [31] and reduced adipose expansion in obesogenic conditions [40], as in the current report, which was interpreted as evidence of lipodystrophy. However, constrained adipose expansion produces inflammation of overloaded, hypertrophic adipose tissue during chronic positive energy balance [1], and the normal adipocyte appearance, lack of adipose inflammation, and normal differentiation of preadipocytes ex vivo in both studies argue against this [31]. Reduced adipose mass can reflect altered energy balance rather than lipodystrophy. Reduction of PI3K pathway activity by overexpression of the lipid phosphatase Pten in mice was previously reported to enhance energy expenditure via brown adipose tissue activation [41], albeit without IR, while pharmacological or genetic inhibition of the p110a catalytic subunit of PI3K has also been associated with increased energy expenditure [40e43]. Moreover, a similar energetic phenotype on brown adipose-specific knockout of Pik3r1 has recently been described [44]. On HFD we found that food intake and energy expenditure were both increased in Pik3r1 WT/Y657 * mice. Given unaltered locomotor activity, intestinal lipid absorption, and lean tissue growth, reduced adipose accumulation shows that increased energy expenditure outweighed increased food intake, and is thus the primary abnormality. Reduced thermal insulation in smaller animals is an unlikely explanation [43]; hence, we conclude that increased energy expenditure explains reduced adipose accretion. Increased brown adipose tissue activity is the most likely explanation, as in models with p110a inhibition [40e43], but this requires further study. Increasing brown adipose tissue function reduces plasma triglyceride and cholesterol concentrations [45,46], and this reduces atherogenesis when liver scavenging of triglyceride-depleted lipoproteins is intact [45]. This accords with the hypolipidemia of Pik3r1 WT/Y657 * mice, and humans with SHORT syndrome. As brown fat is now known to occur in humans, it would be of interest to establish whether it is hyperactive in people with SHORT syndrome. In addition, given the normal lipid profile despite severe IR in people with insulin receptor mutations [13], and the reduced adipose tissue in mice harbouring a common dominant negative insulin receptor allele on an obesogenic diet [47], it will be of interest to assess whether the bioenergetic consequences of reduced PI3K signalling are also seen in the face of insulin receptor dysfunction.

CONCLUSIONS
We report a novel mouse model of SHORT syndrome recapitulating the severe IR without dyslipidaemia seen in humans. Hepatic lipogenesis and lipogenic gene expression were unimpaired, and transcriptomic pathway analysis suggested only mild nutritional state-specific abnormalities in liver gene expression. White adipose tissue was reduced on HFD, but did not show inflammation of hypertrophy, while energy expenditure was increased, raising the testable possibility that the reduced adipose mass in SHORT syndrome does not represent true "lipodystrophy". The most likely mechanism explaining uncoupling of IR from dyslipidaemia in SHORT syndrome is increased brown adipose activity, consistent with other murine evidence that inhibition of PI3K results in beneficial metabolic alterations as well as IR.