Abstract
STATs (Signal Transducers and Activators of Transcription) 5A and 5B
are induced during adipocyte differentiation and are primarily
activated by growth hormone (GH) and prolactin in fat cells. Previous
studies in mice lacking adipocyte GH receptor or STAT5 support their
roles in lipolysis-mediated reduction of adipose tissue mass. Male and
female mice harboring adipocyte-specific deletion of both STAT5 genes
(STAT5^AKO) exhibit increased subcutaneous or inguinal adipose tissue
mass, but no changes in visceral or gonadal fat mass. Both depots
display substantial increases in adipocyte size with no changes in
lipolysis in adipose tissue explants. RNA sequencing analysis of
subcutaneous adipose tissue and indirect calorimetry experiments reveal
sex-dependent differences in adipose gene expression and whole-body
energy expenditure, respectively, resulting from the loss of adipocyte
STAT5.
Keywords: adipose, STATs, sexual dimorphism, growth hormone,
transcription
Introduction
Adipocytes, the defining cells within adipose tissue, regulate
whole-body energy homeostasis. In the obese state, which is
characterized by excessive adipose tissue accumulation, the inability
to safely store lipids in adipose tissue contributes to elevated lipid
levels in the circulation (dyslipidemia) and/or liver (hepatic
steatosis). Obesity has become a global epidemic and is a major risk
factor for type diabetes mellitus (T2DM) and other chronic diseases.
However, some people with obesity are metabolically healthy ([37]1),
and subcutaneous fat can protect against metabolic dysfunction ([38]2).
Mouse models and humans with reduced growth hormone (GH) signaling
exhibit increased subcutaneous adiposity but improved metabolic health
([39]3–[40]7).
The Janus kinase/signal transducer and activator of transcription
(JAK/STAT) pathway mediates cellular signaling of many hormones and
cytokines ([41]8). In mice, STAT5A and 5B proteins have 96% similarity
despite being encoded by separate genes ([42]9); they each have some
essential and redundant signaling roles ([43]10, [44]11). Although
STAT5 proteins are activated by numerous cytokines/hormones, reported
phenotypes of STAT5-null mice support the requirement for STAT5 only in
proper GH and prolactin functions ([45]10).
STAT5 polymorphisms associate with altered cholesterol metabolism
([46]12), body weight ([47]9), and lipid metabolism ([48]13). In
addition, STAT5 proteins regulate adipocyte development [reviewed in
([49]14)], fat mass ([50]10), and lipid metabolism ([51]13), through
mechanisms that remain enigmatic. JAK2-STAT5 signaling in various
tissues contributes to obesity and T2DM ([52]15), and disruption of the
adipocyte JAK2-STAT5 pathway ([53]16) improves systemic metabolism and
liver function ([54]17–[55]19).
To understand the metabolic functions of adipocyte STAT5, we generated
mice lacking both Stat5 genes in adipocytes (STAT5^AKO). Compared to
littermate controls, STAT5^AKO mice fed chow diet had pronounced
increased subcutaneous fat mass, similar to mammalian models with
deficient GH signaling ([56]3–[57]7), and female STAT5^AKO mice had
decreased insulin levels and improved insulin sensitivity as indicated
by homeostasis model assessment of insulin resistance (HOMA-IR).
Although both sexes of STAT5^AKO mice had increased adiposity,
assessment of energy expenditure and whole adipose tissue gene
expression revealed that loss of adipocyte STAT5 confers sexually
dimorphic responses in mice.
Materials and Methods
Animals
Mice with Stat5a and Stat5b genes flanked by loxP sites (floxed)
([58]20) were bred to adiponectin-Cre (AdipoQ-Cre) mice to produce
offspring heterozygous for floxed STAT5 alleles and hemizygous for
AdipoQ-Cre (STAT5^fl/+:AdipoQ-Cre/+), which were then crossed with
STAT5^fl/fl:+/+ mice to create STAT5^AKO mice
(STAT5^fl/fl:AdipoQ-Cre/+) and control littermates (STAT5^fl/fl:+/+).
All mice were on a C57BL/6J background. Unless otherwise stated, mice
were housed in a temperature (22 ± 2°C)- and humidity-controlled
(45–55%) room under a 12-h light/dark cycle with free access to food
and water. Mice used for thermoneutrality experiments were housed at
28°C. All mice were 6 weeks to 11 months of age and fed standard chow
(13% kcal from fat), breeder chow (26% kcal from fat), low-fat diet, no
sucrose (LFD; 10% kcal from fat), or high-fat diet (HFD; 60% kcal from
fat), as specified in the figure legends. Mice were humanely euthanized
by carbon dioxide inhalation followed by cervical dislocation. All
regulations of the Institutional Animal Care and Use Committee at
Pennington Biomedical Research Center were strictly followed, and
experiments were performed under approved protocols 977 and 985.
Body Composition Measurements (NMR)
Body composition was measured by NMR, and adiposity was calculated as
total fat mass divided by total BW x 100.
Blood Glucose and Serum Analyses
Whole-blood and serum samples were collected following a 4-hour fast.
Whole-blood was collected via a tail prick, and glucose levels were
measured using a Breeze 2 glucometer. For serum analyses blood was
collected via cardiac puncture at the time of euthanasia. Serum GH,
insulin-like growth factor-1 (IGF-1), and insulin were measured by
ELISA according to manufacturer instructions. Serum glycerol,
non-esterified fatty acid (NEFA), triglyceride (TG) levels were
measured using colorimetric assays and a microplate reader. For TG
measurements, blood was collected from overnight-fasted mice using
capillary blood collection tubes. HOMA-IR was calculated from glucose
and insulin concentrations as follows: fasting glucose (mg/dl) ×
fasting insulin (µU/ml)/405 ([59]21, [60]22).
RNA Isolation and RT-qPCR Quantification
Tissues were flash frozen in liquid nitrogen and stored at –80°C. RNA
was extracted by homogenization in TRIzol reagent, followed by
chloroform extraction according to the TRIzol reagent manufacturer’s
protocol and further column purification using the RNeasy mini kit. RNA
was quantified using a NanoDrop spectrophotometer. cDNAs were
synthesized and expression levels determined as previously described
([61]23). Primer sequences are listed in [62]Table S1 .
Immunoblotting
Frozen tissues were homogenized in immunoprecipitation (IP) buffer
containing protease and phosphatase inhibitors as previously described
([63]24) with the addition of 100µM sodium fluoride, 2X protease, and
4X phosphatase inhibitors. Lysates were clarified and protein
concentrations were determined by bicinchoninic acid (BCA) protein
assay. Protein extracts (50 µg total protein per lane) were separated
on 7.5 or 10% sodium dodecyl sulfate (SDS)-polyacrylamide gels,
transferred to nitrocellulose membranes, probed with primary antibodies
([64] Table S2 ), and visualized as previously described ([65]24,
[66]25).
Adipose Tissue Fractionation
Adipose fractionation was performed using a protocol adapted from
([67]26). Briefly, excised adipose tissue was weighed, immediately
placed on ice in DMEM + 5% heat-inactivated fetal bovine serum (FBS),
minced, then incubated with Type 1 collagenase at 37°C in a shaking
water bath until fully digested. Following centrifugation, the floating
adipocyte layer was recovered and diluted 1:4 in 5X IP buffer; the
pellet was resuspended in ACK lysing buffer for red blood cell lysis,
filtered by successive passages through cell strainers (pore sizes of
100µm and 40µm), washed with PBS, and centrifuged at 10,000 x g to
yield the stromal vascular fraction (SVF) pellet, which was
subsequently resuspended in 1X IP buffer. Following a single
freeze/thaw cycle at -80°C, the resuspended adipocyte fraction and SVF
were each passed through a 20-gauge needle 10 times and then
centrifuged at 10,000 x g for 10 min. The aqueous layer between the
floating lipid and pellet of the adipocyte fraction, and the
supernatant of the SVF were used for immunoblotting analyses. All
centrifugations were performed at 4°C.
Adipose Histology and Adipocyte Size Analysis
Excised adipose tissue was fixed in 10% neutral buffered formalin for a
minimum of 24 hours. Tissues were processed, embedded, stained, and
imaged, and adipocyte area was calculated from > 3000 cells per mouse
per depot as previously described ([68]27).
Ex Vivo Lipolysis
Ex vivo lipolysis assays were performed as described in ([69]28). Mice
were fasted for 4 hours prior to euthanasia and white adipose tissue
(WAT) excision. Briefly, gonadal WAT (gWAT) explants (10 – 20 mg, in
triplicate) were incubated in 200 µl DMEM containing 2% fatty acid
(FA)-free BSA, with or without 10 µM isoproterenol in 96 well plates at
37°C in a 5% CO[2] and 95% humidified incubator for two hours. Released
glycerol and NEFA were quantified from the conditioned medium and
normalized to tissue weight.
Indirect Calorimetry/Energy Expenditure Analyses
Mice were weaned onto standard chow diet at 3 WOA. Beginning at 7 WOA,
mice were either switched to a characterized low-fat diet or remained
on chow diet for the duration of indirect calorimetry assessment in the
metabolic cage system starting at 10-11 WOA. Oxygen consumption and
carbon dioxide production were continuously monitored for 3 days to
calculate estimate of energy expenditure. Cages were maintained with
12-hr light/dark cycles. Locomotor activity (pedometers) was determined
by calculating the sum of all detectable motion determined using an
infrared photocell beam interruption technique (> 1 cm/s along X-, Y-,
or Z-axis) over the continuous monitoring period.
RNA-Sequencing and Analysis
For RNA sequencing analysis, RNA was isolated from subcutaneous iWAT of
6-week-old mice as described for RT-qPCR quantification. RNA integrity
numbers >7 were confirmed using a Bioanalyzer RNA 6000 chip (Agilent).
Sequencing libraries were constructed using a Quant-Seq 3’ mRNA-Seq
Library Prep kit. Each sample was prepared with a unique sample index.
Completed libraries were analyzed on the Bioanalyzer High Sensitivity
DNA chip (Agilent) to verify correct library size. All libraries were
pooled in equimolar amounts and sequenced on the NextSeq 500 sequencer
at 75bp forward read and 6bp forward index read. Primary data analysis
was performed using the Lexogen QuantSeq pipeline V1.8.8 on the Bluebee
platform for quality control, mapping, and read count tables. Raw and
processed data are deposited in the GEO database (accession number GEO:
[70]GSE113939).
Raw count matrices of RNA sequencing data were obtained via the
Rsubread ([71]29) package in R, and further processed for gene
quantification and identification of differentially expressed genes
using the limma package ([72]30). Sequencing data were aligned to the
GRCm38.84 mouse genome (mm10). Genes with at least one count per
million (CPM) reads in 6 or more samples were retained for further
analysis, resulting in 15118 genes. Gene counts were log2 transformed
and normalized for sequencing depth via the TMM method ([73]31). The
mean-variance relationship of gene-wise standard deviation to average
logCPM gene signal was assessed via the ‘voom’ method ([74]32) to
generate precision weights for each individual observation. The logCPM
values and associated precision weights were subsequently utilized to
generate empirical Bayes moderated t-statistics estimates for
identification of differentially expressed genes. To control for
multiple testing, false discovery rates were computed and genes with an
adjusted p-value ≤ 0.05 were deemed differentially expressed. Venn
analysis of gene lists were conducted via Venny
([75]https://bioinfogp.cnb.csic.es/tools/venny/index.html).
Pathway enrichment analysis on differentially expressed genes was
conducted via gene set enrichment analysis or GSEA ([76]33, [77]34),
using a custom pathway database consisting of KEGG pathways (obtained
from MSigDB) plus some user-defined custom gene sets. Due to the
genomic scale of the study, we used all genes annotated in the genome
as the background set for the GSEA (as recommended in the DAVID
tutorial ([78]https://david.ncifcrf.gov/helps/FAQs.html#16).
Quantification and Statistical Analysis
All data are expressed as mean ± SEM. GraphPad Prism 8.0 and JMP
version 14 were used for statistical analyses. For comparisons between
two independent groups, a Student’s t test was used and p < 0.05 was
considered statistically significant. For comparisons between three or
more groups, two-way ANOVA with Tukey’s multiple comparisons testing
was performed. Analysis of covariance (ANCOVA) was used to determine
differences between groups in energy expenditure using adjustments for
fat and fat-free mass. All sample sizes, statistical test methods, and
p-values are listed in the figure legends.
Data and Resource Availability
STAT5^AKO mice generated in this study are available from the
corresponding author with a completed Material Transfer Agreement, if
mice are still being bred at the time. If the STAT5^AKO mice are
unavailable at the time of the request, they can be easily re-created
by breeding STAT5^fl/fl mice, which we can provide with permission from
Dr. Lothar Hennighausen, with Adipoq-Cre mice
(B6.FVB-Tg(Adiopq-cre)1Evdr/J), which are available from Jackson
Laboratory (Stock No. 028020). We are glad to share requested mice with
reasonable compensation for shipping. Further information and requests
for resources and reagents should be directed to and will be fulfilled
by the corresponding author. Commercially available resources utilized
in these studies are described in [79]Table S2 . RNA-Seq data are
available at in the GEO repository under the accession number GEO:
[80]GSE113939.
Results
Confirmation of Adipocyte-Specific STAT5A and 5B Knockout Status in STAT5^AKO
Mice
STAT5A and 5B proteins ([81] Figures 1A, B ) and mRNA ([82] Figures 1C,
D ) were reduced, but not absent, in whole fat from STAT5^AKO mice. We
examined gene expression in several fat pads as well as liver and
skeletal muscle ([83] Figures 1C, D ). Stat5a mRNA was 50% lower in
iWAT and gWAT of knockout mice, and Stats 5a and 5b mRNA in brown
adipose tissue (BAT) were both reduced by >60%. There were no changes
in Stat5 gene expression in liver or skeletal muscle of female mice.
Although Stat5b gene expression was not reduced in iWAT ([84] Figure 1D
), STAT5B protein levels were significantly reduced ([85] Figures 1A, B
), and we observed decreased Stat5b mRNA levels in iWAT from subsequent
cohorts of STAT5^AKO mice (data not shown).
Figure 1.
[86]Figure 1
[87]Open in a new tab
Expression of STAT5A and 5B is knocked down in adipose tissue and
adipocytes of STAT5^AKO female mice. Female STAT5^AKO (AKO) mice and
their floxed (FL) littermate controls were euthanized at 3 (A–D), 4
(E), or 11 (F) months of age, and tissues were immediately collected
for protein or gene expression analyses (A–D, F) or cellular
fractionation (E). (A) Immunoblot of proteins resolved from iWAT
samples for 5 mice of each genotype. (B) Quantification of band
intensities from A (n = 5 per group). Band intensities were normalized
to the loading control ERK1/2 and represented as fold change relative
to FL mice. (C, D) Stat5a and Stat5b gene expression measured by
RT-qPCR for the indicated tissues (n = 5 – 8 mice per group). (E)
Gonadal white adipose tissue was removed from female STAT5^AKO,
heterozygous STAT5^AKO (het AKO), and homozygous floxed control (FL)
mice, fractionated into adipocytes and stromal vascular fraction (SVF),
and 100 µg (SVF) or 200 µg (adipocyte) protein subjected to western
blotting. Also shown are ERK1/2 as a loading control and adiponectin
(ADPN) as an adipocyte marker. For each protein, the SVF and adipocyte
samples were run on the same gel and the images were from the same
exposure of the same blot. (F) Eleven-month-old female mice were
injected with 1.5mg/kg mGH or vehicle (V; 0.1% BSA/PBS) for 30 minutes
prior to euthanasia and tissue collection. Cish gene expression
measured by RT-qPCR is shown (n = 2 – 3 mice per group). Significance
was determined by t-test for FL versus AKO comparisons in A-D) and is
denoted as *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. For F,
a 2-way ANOVA was used to assess treatment/genotype and tissue
variables with Tukey’s post-hoc multiple comparisons test to compare
all treatment/genotype groups for each tissue; ^# denotes p < 0.05 for
GH versus V comparisons. See also [88]Figure S1 .
Since STAT5 is also present in the stromal vascular fraction (SVF) of
adipose tissue (AT), we fractionated fat and observed the expected
decreases in STAT5 proteins in the adipocyte fractions from AKO
heterozygotes (het) and STAT5^AKO mice ([89] Figure 1E ). We also
consistently observed elevated STAT5A protein levels in the SVF of
STAT5^AKO mice. There were no changes in STAT5 protein levels in liver
or skeletal muscle in STAT5^AKO female mice (data not shown). Growth
hormone (GH) rapidly induces Cish gene expression in a STAT5-dependent
manner ([90]35). Acute GH injection increased Cish mRNA expression in
the livers of both floxed and STAT5^AKO mice, but only in iWAT from
floxed mice ([91] Figure 1F ). We made similar observations in male
mice ([92] Figure S1 ). Collectively, our results demonstrate
Cre-mediated excision of Stat5a/b only in mature adipocytes of male and
female mice.
Increased Adiposity of STAT5^AKO Mice Is Depot-Specific, but Not Correlated
With Depot-Specific Changes in Adipocyte Size or BAT Activity
As shown in [93]Figures 2A, B and [94]S2A, B , female and male
STAT5^AKO mice had significantly higher iWAT weights, without
discernable changes in gWAT or other fat depots, except for small
statistically significant increases in retroperitoneal and brown fat.
Notably, only the subcutaneous iWAT depot of STAT5^AKO mice was
consistently larger in multiple cohorts of mice at different ages (data
not shown). Histological analysis revealed hypertrophic adipocytes in
both iWAT and gWAT of female ([95] Figures 2D–F ) and male STAT5^AKO
mice ([96] Figures S2D–F ).
Figure 2.
[97]Figure 2
[98]Open in a new tab
Female STAT5^AKO mice have increased subcutaneous adiposity when fed
chow diet. Female STAT5^AKO (AKO) and floxed (FL) littermate control
mice were weaned onto regular chow diet (13% kcal from fat). (A)
Representative images of inguinal and gonadal white adipose tissue
depots (iWAT and gWAT) from five-month-old mice. (B) Weights of white
adipose tissue depots (iWAT, gWAT, retroperitoneal - rWAT, mesenteric -
mWAT), brown adipose tissue (BAT), gastrocnemius skeletal muscle
(Gastroc), and liver collected from 10-week-old mice (n = 11-13). (C)
Mice (n = 5-6 per group) were housed at different temperatures
beginning at weaning (3 weeks of age). Body composition was measured at
9 weeks of age, adiposity was calculated as fat mass divided by total
body weight for each animal. (D) Representative images of H&E-stained
inguinal and gonadal white adipose tissue depots (iWAT and gWAT) from
five-month-old mice. Quantification of total mean adipocyte area (E)
and fat cell size distribution (F) from H&E-stained images shown in D
(n = 4 mice per genotype). Significance was determined by t-test and is
denoted as *p < 0.05, **p < 0.01, or ****p < 0.0001 for AKO versus FL
comparisons. See also [99]Figures S2, S3 .
To assess the robustness of the adiposity phenotype in STAT5^AKO mice
and the potential role of BAT, we also examined body composition at
thermoneutrality. As shown in [100]Figures 2C and [101]S2C , female and
male mice maintained increased adiposity following six weeks of
thermoneutral housing conditions. While females demonstrated no
differences in BW or lean mass at either housing temperature, male
STAT5^AKO mice had higher fat mass at both temperatures, but higher
lean mass and BW at 28°C only ([102] Figure S3 ). Taken together, these
data suggest that STAT5 signaling in BAT does not substantially
contribute to development of the adiposity phenotype.
Glucose and Lipid Metabolism in STAT5^AKO Mice
The phenotype of animal models with altered GH signaling is often
confounded by differences in IGF-1 levels. Although there was a trend
towards decreased GH levels, serum IGF-1 levels were unaltered in
STAT5^AKO females ([103] Figures 3A, B ). Notably, these mice had
significantly lower fasting insulin levels, with no difference in
glucose levels ([104] Figures 3C, D ). Therefore, STAT5^AKO females
were more insulin sensitive, as indicated by HOMA-IR ([105] Figure 3E
), despite having higher adiposity. Like females, STAT5^AKO males had
higher subcutaneous fat mass and improved HOMA-IR, which was not,
however, driven by lower insulin levels as it was in females ([106]
Figures 3H–J ). Assessment of HOMA-IR for the male STAT5^AKO mice was
confounded by high blood glucose levels ([107] Figure 3I ) of the
floxed mice since serum insulin values were not significantly different
([108] Figure 3H ). When glucose levels of chow-fed mice were measured,
we observed no difference between male control and STAT5^AKO mice (data
not shown). However, it is notable that while the control male mice had
higher than expected blood glucose levels ([109] Figure 3I ), possibly
due to acute high-fat-diet feeding for 4 days immediately prior to
euthanasia, the STAT5^AKO males maintained lower glucose levels. In
males, circulating IGF-1 and GH levels did not significantly differ
between genotypes ([110] Figures 3F, G ).
Figure 3.
[111]Figure 3
[112]Open in a new tab
STAT5^AKO mice have improved homeostasis model assessment of insulin
resistance (HOMA-IR). Female (A–E) and male (F–J) STAT5^AKO (AKO) and
floxed (FL) littermate control mice were weaned onto regular chow diet
(13% kcal from fat) and switched to a defined-composition LFD (10% kcal
from fat) at 6 weeks of age (WOA), followed by HFD (60% kcal from fat)
at 11 WOA for 4 days prior to euthanasia. Serum insulin-like growth
factor 1 (IGF-1; A, F), growth hormone (GH; B, G), insulin (C, H), and
blood glucose (D, I) levels in 3-month-old mice (n = 7 – 8). (E, J)
HOMA-IR was calculated from insulin and glucose levels, respectively in
(C, D) or (H, I). Significance was determined by t-test and is denoted
as ***p < 0.001, for AKO versus FL comparisons in females and ^##p <
0.01 in males.
It is largely accepted that GH promotes fat loss via lipolysis, and
animal models with diminished adipose tissue GH signaling have altered
lipolytic responses ([113]16–[114]19, [115]36, [116]37). We examined
gene and protein expression of several known lipolytic mediators
including adipose triglyceride lipase (ATGL), CGI-58 (comparative gene
identification-58), and β3 adrenergic receptor (ADRB3). As shown in
[117]Figure 4 , male STAT5^AKO mice have decreased gene expression of
Atgl/Pnpla2, Cgi-58/Abhd5 and Adrb3, but diminished protein expression
levels for CGI-58 only. Notably, levels of phosphoactivated
extracellular signal-regulated kinase 1/2 (ERK1/2) were increased in
STAT5^AKO mice of both sexes. While the data shown in [118]Figure 4
were generated from a non-fasted cohort of mice, gene and protein
expression analyses of iWAT from fasted mice yielded the same results
(data not shown). Lipolysis of triglycerides (TGs) results in the
release of glycerol and NEFA from adipocytes into the bloodstream.
Circulating levels of glycerol, NEFA, and TGs were not significantly
different in STAT5^AKO mice of either sex versus floxed controls ([119]
Figures 5A–F ). The serum glycerol and NEFA levels showed no genotype
differences in either sex, independent of diet (LFD/HFD –
[120]Figures 5A–D versus chow - [121]Figures S4A–D ), housing
temperature ([122] Figures S4A–D versus [123]Figures S4E–H ), or
fasting period ([124] Figures S4A–D versus [125]Figures S4I–L ).
Figure 4.
[126]Figure 4
[127]Open in a new tab
Effects of loss of adipocyte STAT5 on expression of lipolytic proteins.
Chow-fed female (F) and male (M) STAT5^AKO (AKO) and floxed (FL)
littermate control mice were euthanized at 11 weeks of age
(non-fasted), and subcutaneous iWAT was collected for gene (A) and
protein (B, C) expression analyses. (A) Gene expression of Atgl/Pnpla2,
Cgi-58/Abhd5, and Adrb3 was measured by RT-qPCR and normalized against
the reference gene Nono. (B) Protein expression was examined by
immunoblotting, and three representative samples per group are shown.
(C) Band intensities were quantified and normalized against total
ERK1/2 expression. Fold change was calculated as the relative gene or
protein expression values divided by the mean value of the floxed
control group of the same sex for each gene/protein (n = 7 per group).
Significance was determined by t-test for FL versus AKO comparisons and
is denoted as *p < 0.05, **p < 0.01, or ****p < 0.0001 for female
comparisons, while significance for male comparisons is denoted as ^#p
< 0.05 or ^##p < 0.01.
Figure 5.
[128]Figure 5
[129]Open in a new tab
STAT5^AKO mice do not show any differences in adipose tissue lipid
metabolism relative to floxed control mice. Female (A, C, E, G, I) and
male (B, D, F, H, J) STAT5^AKO (AKO) and floxed (FL) littermate control
mice were fed a defined-composition low-fat diet (LFD - 10% kcal from
fat; A–D) or regular chow (13% kcal from fat; E–J). (A–D) After 1 month
on LFD diet, at 3 months of age, serum glycerol and non-esterified
fatty acid (NEFA) levels were measured in 4 h-fasted mice. E and F)
Serum triglyceride (TG) levels were measured in 8-week-old,
overnight-fasted mice. Ex vivo lipolysis (G, H) and de novo lipogenesis
(I, J) assays were performed using gWAT and iWAT explants,
respectively, from chow-fed mice (5 months old). (G, H) Glycerol
release from gWAT explants (~20mg) into media following a 2
h-incubation period was measured under both basal and isoproterenol
(ISO)-stimulated (10µM) conditions. I and J) Incorporation of
^14C-glucose into total triglycerides was measured by incubating iWAT
explants (~50mg) with 4µCi/ml of [^14C]-U-glucose for 4.5 hours. The
triglyceride (neutral lipid) fraction was purified and [^14C] counts
were measured by scintillation counting. (A–F) t-tests were used to
test for significance between means (n = 8-12 mice per genotype). (G–J)
Two-way ANOVA with Tukey’s post-hoc multiple comparison analysis was
used to test for significance between genotypes and treatments (n = 3-6
mice per group). Significance is denoted as *p < 0.05 or ****p < 0.0001
for basal versus ISO or insulin comparisons and ^##p < 0.01 or ^####p <
0.0001 for AKO versus FL comparisons.
Increased adiposity of STAT5^AKO mice could be due to either decreased
lipolysis or increased lipogenesis. To assess these processes more
directly we performed ex vivo lipolysis and de novo lipogenesis assays,
respectively in the gWAT and iWAT of the same cohort of mice. Basal or
adrenergic-induced lipolysis altered in gWAT explants from female mice
([130] Figure 5G ) were not significantly different as determined by
levels of glycerol released into the medium. Male STAT5^AKO mice had a
modest, non-significant decrease in adrenergic-stimulated lipolysis
([131] Figure 5H ). Ex vivo ^14C-glucose incorporation into iWAT TGs, a
measure of lipogenesis, was not different, ([132] Figure 5I ) either in
basal or insulin-stimulated conditions, in STAT5^AKO females compared
to controls. In males, insulin-stimulated ^14C-glucose incorporation
into iWAT TGs was greater in controls than in STAT5^AKO mice ([133]
Figure 5J ).
The convention for these assays is to normalize measurements by tissue
weight as done in [134]Figure 5 . However, since STAT5^AKO mice have
larger adipocytes in both gWAT and iWAT (see [135]Figures 2 and [136]S2
), normalizing by tissue weight might be confounding. Normalizing by
protein or DNA content also can confound the results because adipocytes
only constitute 30 – 60% of adipose tissue cellular content even though
they make up approximately 80% of the volume ([137]38–[138]41). Since
the number of adipocytes per mg of tissue would be expected to
significantly differ between STAT5^AKO and control mice due to the
significant difference in adipocyte size within adipose tissue depots,
we also normalized the ex vivo lipolysis and lipogenesis data by
adipocyte number using an estimate of adipocyte number per mg tissue,
which was calculated using a formula to approximate adipocyte mass
(m[ad] ) that is described in ([139]41). This formula assumes that each
adipocyte is a perfect sphere with its volume equal to
[MATH:
43πr
3 :MATH]
, and the same density as triolein (0.915 g/ml):
[MATH:
mad(in μg)=0.915106×π6×d3=0.4791106×d3 :MATH]
. A statistical transformation,
[MATH: 3σ×d¯+d¯3
:MATH]
replaces d ^3 to account for skew in calculation due to the
distribution of adipocyte sizes;
[MATH: d¯ :MATH]
is the mean diameter calculated from the cross-sectional adipocyte
areas used to create [140]Figures 2 and [141]S2 , and σ is the
respective standard deviation. Adipocyte number per mg, shown in
[142]Figures S5A, B , was calculated by inversion of the adipocyte
mass. When normalized by adipocyte number per explant, glycerol release
was higher, not lower, from STAT5^AKO AT explants compared to floxed
controls, demonstrating that AT lipolysis was not decreased in
STAT5^AKO mice using either method of normalization ([143] Figures S5C,
D ). Notably, STAT5^AKO mice of both sexes had higher levels of
^14C-glucose lipid incorporation when the data were normalized by
adipocyte number ([144] Figures S5E, F ) versus tissue weight,
indicating that elevated de novo lipogenesis might confer or play a
role increased adiposity in STAT5^AKO mice.
For the room temperature (RT)- and thermoneutrality (TN)-housed cohorts
of mice that were fasted for 4h or 18h prior to tissue collection,
glycerol release, normalized by tissue weight, from iWAT and gWAT
explants was not significantly different between genotypes of both
sexes under basal conditions or in the presence of adrenergic stimuli,
isoproterenol or CL 316,243 ([145] Figures S6, S7 ). There were a few
exceptions where STAT5^AKO mice had significantly less glycerol release
than floxed controls under adrenergic-induced lipolytic conditions (see
[146]Figures S6A, B , [147]S7E ), but these instances were not
consistently observed as a function of housing temperature, fasting
period, sex, or AT depot. Moreover, when the glycerol levels were
normalized by adipocyte number, AT explants from STAT5^AKO mice did not
release less glycerol than control mice under any condition (data not
shown). Since insulin is an important anti-lipolytic hormone, we also
assessed ex vivo lipolysis for each condition in the presence of
insulin. As shown in [148]Figures S6, S7 , insulin suppressed glycerol
release under adrenergic-induced, but not basal conditions, and no
significant genotype differences were observed under any condition,
indicating that the loss of adipocyte STAT5 does not amplify the
antilipolytic effect of insulin. Taken together, these data support
increased lipogenesis, but not reduced lipolysis, as a principal driver
of the adiposity phenotype in STAT5^AKO mice.
Sex-Specific Differences in Energy Expenditure and Adipose Tissue Gene
Expression in STAT5^AKO Mice
As shown in [149]Figures 6A, C, D , female STAT5^AKO mice fed LFD had
reduced daily energy expenditure when adjusted by ANCOVA for fat-free
mass (see [150]Figure S8 for body weight and fat-free mass of mice
measured in metabolic chambers) with no differences in RER between
genotypes. As in all our cohorts of STAT5^AKO mice, there were no
differences in food intake versus controls ([151] Figure 6G ). Mice
were more active in the dark cycle, but no differences were observed
between genotypes ([152] Figure 6H ). In a separate experiment,
chow-fed male mice were also assessed in the metabolic cages. Although
there were no differences in total energy expenditure between the
genotypes ([153] Figures 6B, E ), RER was significantly increased in
STAT5^AKO males during the light cycle ([154] Figure 6F ), indicating a
shift toward carbohydrate utilization. Like the female mice, there were
no significant differences in total daily food intake or activity
between the genotypes ([155] Figures 6I , [156]J ).
Figure 6.
[157]Figure 6
[158]Open in a new tab
Female, but not male STAT5^AKO mice have reduced energy expenditure.
Floxed (FL) and STAT5^AKO (AKO) female (A, C, D, G, H) and male (B, E,
F, I, J) mice (10 weeks of age) were individually housed in metabolic
cages where oxygen and carbon dioxide concentration were continuously
monitored for 3 days. (A, B, C, E) Total energy expenditure (EE) was
calculated by multiplying the daily average rate of energy expenditure
by 24 and by the number of experiment days. As an estimate of substrate
oxidation, (D, F) respiratory exchange ratio (RER, VCO2/VO2), (G, I)
food intake and (H, J) total activity in walking meters (pedometers)
were measured during the light and dark cycles. Significance was
determined by ANCOVA with fat and fat-free masses as covariates. The *
denotes p < 0.05 and **p < 0.01 (n = 8-12/group) between genotypes.
To investigate molecular mechanisms regulated by loss of STAT5
transcriptional activity in inguinal fat, we performed mRNA sequencing
(RNA-Seq) on whole iWAT. The expression changes resulting from loss of
adipocyte STAT5 were poorly correlated between females and males (R^2 =
0.17) ([159] Figure 7A ). At an adjusted p-value ≤0.05 threshold, we
identified 387 differentially regulated transcripts in males (289
upregulated and 98 downregulated), but only 42 in females (37
upregulated and 5 downregulated) ([160] Table S3 ). Approximately 60%
(24/42) of the transcripts differentially expressed in females were
also regulated in males, resulting in a statistically significant
overlap (Fisher exact test p < 2.2E-16) between the two lists ([161]
Figure 7A , inset). Unbiased principal component analysis (PCA) shows
segregation of all four groups ([162] Figure S9 ) and indicates that
sex-specific differences persist in both STAT5^AKO mice and floxed
controls. Heatmaps for the top differentially expressed genes between
males and females appear in [163]Figure 7B . Overall, these RNA-Seq
data revealed largely sex-dependent gene expression changes in whole
iWAT from STAT5^AKO mice, with a greater transcriptomic response
observed in males but also some sex-independent changes in gene
expression.
Figure 7.
[164]Figure 7
[165]Open in a new tab
Loss of adipocyte STAT5 results in sexually dimorphic gene expression
changes within subcutaneous WAT. Inguinal WAT was collected from male
and emale STAT5^AKO and floxed (FL) control mice at 6 weeks of age, and
isolated mRNA was subjected to RNA-sequencing analysis. (A) Scatterplot
showing correlation between female and male gene expression changes
resulting from adipocyte-specific loss of STAT5 (LogFC = Log fold
change). The black dashed line is the trend line, and the correlation
coefficient (R2) is shown. Inset: Venn diagram of 405 differentially
regulated genes in male and female mice (AKO/FL) at threshold adjusted
p-value equal to 0.05, expressed as numbers of genes and percentages of
total regulated genes. (B) Heatmaps of expression levels of top
differentially regulated genes in female (F) and male (M) samples. For
females, genes with an adjusted p-value ≤ 0.1 and ≥ 2 or ≤ -2-fold
differential gene expression, and for males, genes with adjusted
p-value ≤ 0.01 and ≥ 2 or ≤ -2-fold differential gene expression are
plotted. Heatmaps are row-normalized with lower gene expression shown
in blue and higher gene expression in red. KO labels refer to STAT5^AKO
mice and WT to STAT5 floxed mice. (C) Enrichment maps of significant
pathways identified from GSEA (FDR ≤ 0.05) that share gene members.
Upper and lower panels show significant pathways in females and males,
respectively. Pathways upregulated in knockouts are colored in red, and
downregulated pathways in blue. Thickness of lines connecting pathways
is proportional to the extent of gene sharing between them. Pathways
with no shared genes are not shown. (D) Enrichment scores of selected
pathways significantly altered in both females and males, as determined
via GSEA. The upper panels for females (F) or males (M) show enrichment
plots for the ‘oxidative phosphorylation’ pathway, which was
downregulated in STAT5^AKO mice of both sexes; bottom panel for each
sex shows enrichment plot for the ‘chemokine signaling’ pathway which
was upregulated in both males and females. Pathway enrichment p-values
(FDR) are noted in each plot. (E) Pathways uniquely enriched in females
(F) or males (M). For each pathway, the relative significance of
enrichment (negative logarithm of the FDR) in females and males (gray
and white bars, respectively) are plotted side by side, with the
vertical dashed line representing an FDR of 0.01. Pathways upregulated
in knockout samples are indicated by a red border and pathways
downregulated in knockout samples are indicated by a blue border.
To understand the mechanistic implications of the observed differences
in gene expression, we performed gene-set enrichment analysis (GSEA)
([166] Table S4 ). Among KEGG pathways with a false discovery rate
(FDR) < 0.05 we identified 8 and 21 downregulated pathways in female
and male AKO samples, respectively. Of these, all 8 female-regulated
pathways were also downregulated in males, and 13 were unique to males.
Upregulated pathways numbered 43 and 33 in AKO female and male samples,
respectively, with 21 pathways common to both sexes. The enrichment
maps in [167]Figure 7C show a subset of regulated pathways in female
and male samples, in the form of a network based on the extent of gene
sharing between pathways. The ‘oxidative phosphorylation’ pathway was
downregulated in both male and female KO samples, pointing to possible
disturbances in ATP-generating processes in animals lacking adipocyte
STAT5. Several of the pathways upregulated in male and female KO
samples were related to immune-inflammatory processes including
‘chemokine signaling’ and ‘allograft rejection’ (FDR < 1E-05 in females
and 3.67E-03 in males), etc., with considerable gene sharing among the
pathways ([168] Figure 7D ). Many lipid metabolism and fuel utilization
pathways were uniquely downregulated in male KO animals, including
fatty acid metabolism, PPAR signaling, citrate cycle, propanoate
metabolism, pyruvate metabolism, etc. Interestingly, a subset of
immune-inflammatory pathways was exclusively upregulated in STAT5^AKO
females, including ‘cytokine-cytokine receptor interaction’ (FDR <1E-05
in females and 0.10 in males), ‘JAK-STAT signaling’ (FDR 6.18E-04 in
females, 0.06 in males), and ‘VEGF signaling’ (FDR 9.51E-03 in females,
0.15 in males) ([169] Figure 7E ). This sexual dimorphism may reflect
different aspects of the inflammatory response, some common to both
sexes and some unique to females.
A potential limitation of the RNA-Seq analysis is that over half of the
cells in whole iWAT are not adipocytes ([170]38, [171]39), and that a
compensatory increase in SVF STAT5 levels was observed in our STAT5^AKO
mice ([172] Figure 1E ). Therefore, although sexual dimorphism in
immune-inflammatory pathways is intriguing, it is unclear whether this
observation reflects changes in adipocyte gene expression or in the
transcriptomes and/or proportions of co-resident immune cells.
Discussion
STAT5 is a primary mediator of GH signaling, thus, understanding its
contributions to in vivo adipocyte function related to metabolic
health, is important. To that end, we have performed physiological
studies in mice lacking both STAT5 genes in adiponectin-expressing
cells. Another group has also generated this mouse model, but examined
only male mice, which exhibited higher adiposity, increased adipocyte
size, and evidence of improved insulin sensitivity ([173]16). We have
examined both male and female STAT5^AKO mice and observed increased
subcutaneous adiposity in both sexes, along with indicators of improved
insulin sensitivity on chow diets ([174] Figure 3 ). Further, the
following unpredicted phenotypic observations in our STAT5^AKO mice
offer additional insight: 1) increased adiposity is not accompanied by
changes in lipolysis and 2) significant sex-specific differences in
energy expenditure and gene expression exist in STAT5^AKO mice. These
observations reveal fundamental gaps in our understanding of adipocyte
STAT5 and its role in metabolism.
GH regulates several metabolic pathways including lipolysis,
lipogenesis, glucose uptake, and protein synthesis ([175]42). Excess
adiposity in GH-deficient patients can be improved by exogenous GH
administration ([176]43), an effect primarily attributed to GH’s
lipolytic action. Male mice with adipocyte-specific deletion of GHR
([177]36), JAK2 ([178]17–[179]19, [180]37) or STAT5 ([181]16) have
increased adiposity also attributed to decreased lipolytic rates, and
impaired lipolysis in STAT5^AKO mice has been attributed to decreased
ATGL and CGI-58 levels in subcutaneous fat ([182]16).
In contrast to published data, our STAT5^AKO mice have no significant
differences in circulating levels of glycerol or NEFAs, the products of
lipolysis, or in ex vivo lipolytic activity of AT explants versus
floxed controls ([183] Figures 5 and [184]S4–S7 ). Therefore, decreases
in lipolytic gene expression and in CGI-58 protein levels do not
correlate with reduced lipolytic rates in STAT5^AKO AT ([185] Figures 4
, [186]5 ). A variety of factors could account for the phenotypic
discrepancies between the two STAT5^AKO lines, including genetic drift,
or differences in housing conditions or gut microbiota between mouse
colonies reared at different institutions. Also, our fasting conditions
are limited to 4 hours or overnight, while increased lipolytic products
in serum of mice lacking adipocyte GHR, JAK2, or STAT5 were observed
after prolonged fasts ranging up to 48 hours ([187]16–[188]19, [189]36,
[190]37). We specifically chose a 4h fasting period because when food
intake was monitored throughout the day while the mice were in the
metabolic chambers and provided food ad libitum, their longest period
of fasting was approximately 4h. Given that the STAT5^AKO mice accrue
more subcutaneous fat mass than the floxed control mice under ab
libitum feeding conditions without eating more food, we believe that a
4h fasting period to examine the potential role of lipolysis in
development of the adiposity phenotype is more physiologically relevant
than an overnight fasting period, which would be more akin to
starvation. Notably, mice lacking adipocyte-specific expression of the
lipolytic mediator, ATGL, are phenotypically different from STAT5^AKO
mice, consistent with our findings and suggesting that obesity in
STAT5^AKO mice is not the result of reduced lipolysis.
Another consideration is that GH promotes beige fat formation via
GHR-induced STAT activation ([191]44). Impaired cold tolerance has been
reported in STAT5^AKO mice ([192]45), and the iWAT in both sexes of our
STAT5^AKO mice appears whiter ([193] Figures 2A and [194]S2A ),
suggesting lower expression of beige adipocyte-related genes. However,
we observed no alterations in expression of iWAT UCP-1 or other beiging
markers (data not shown). Also, both sexes maintained higher adiposity
under thermoneutral conditions ([195] Figures 2C and [196]S2C ),
suggesting that the increased fat mass of STAT5^AKO mice is not due to
altered BAT thermogenesis.
Based on other animal models with disrupted adipose tissue GH or STAT5
signaling ([197]16, [198]36, [199]46), we expected the increased
adiposity of STAT5^AKO mice to be dependent on decreased lipolysis, but
this was not evident in our model ([200] Figures 5 and [201]S4–S7 ),
despite clear evidence that adipocyte STAT5 contributes to AT
deposition ([202] Figures 2 and [203]S2 ). Mediators other than STAT5
have been implicated in GH-induced lipolysis, including the MEK/ERK and
PLC/PKC pathways ([204]47–[205]49) and our STAT5^AKO mice exhibited
increased levels of phosphoactivated ERK1/2 in iWAT ([206] Figure 4C ).
Although we did not observe an overall decrease in AT lipolysis in
STAT5^AKO mice, it is possible that the increased phosphorylation of
ERK1/2 elevated lipolysis ([207]50) and may have masked any lipolytic
defect that might have been the direct result of loss of adipocyte
STAT5. Additional experiments interrogating other molecular players
within the lipolysis pathway, such as hormone sensitive lipase,
perilipins, and phosphodiesterases, as well as using MEK inhibitors
will be necessary to test this hypothesis.
The specificity of the subcutaneous adiposity was also intriguing and
likely due to compensation within the visceral depots during
development. Notably, BCL6^AKO mice have a very similar adiposity
phenotype ([208]51) to STAT5^AKO mice. Both mouse models exhibit
increased subcutaneous adiposity when either BCL6 or STAT5 is knocked
out of adipocytes in a congenital manner. Transcriptomics analysis
revealed that both visceral (peri-gonadal) and subcutaneous WAT
underwent similar and highly correlated gene expression changes with
ablation of BCL6, even though only the subcutaneous depot displayed
expansion in adipocyte size and fat mass. Moreover, when BCL6 was
ablated only in adult mice using a doxycycline-inducible mouse model,
all WAT depots (visceral and subcutaneous) exhibited increased
adipocyte size and fat depot mass, supporting the hypothesis that
developmental compensation drives the difference in fat mass between
the depots in the BCL6^AKO model ([209]51). Interestingly, BCL6 is a
transcriptional target of STAT5 and these transcription factors
demonstrate reciprocal occupancy within promoters to influence
activation or repression of GH target genes ([210]52). Clearly, these
observations suggest some compensation among GH-induced and
developmental signaling pathways and warrant further study to elucidate
the specific mechanism(s) and complex crosstalk between signaling
pathways and transcription factors that ultimately give rise to
sex-specific gene expression and energy expenditure differences and
drive the increased adiposity in our STAT5^AKO mice. These studies
highlight STAT5^AKO mice as a useful tool to investigate sex-specific
differences in adiposity, GH signaling, and adipose tissue
depot-specific responses.
In summary, our STAT5^AKO mice revealed several intriguing
observations. Despite the sex-independent differences in subcutaneous
fat mass and adipocyte size, STAT5^AKO mice exhibit prominent
sex-specific differences. Male and female STAT5^AKO mice both have
higher adiposity than floxed controls, which is maintained at
thermoneutrality ([211] Figures 2 and [212]S2 ), indicating that BAT
thermogenesis is not driving the phenotype. However, this increased fat
mass is associated with decreased energy expenditure in females only
([213] Figure 5 ). Transcriptional effects in iWAT were also
sex-specific, as we observed more differentially regulated genes in
males than in females. Notably, downregulation of fatty acid
metabolism, pyruvate metabolism, peroxisome, and PPAR signaling
specifically in male STAT5^AKO animals could indicate greater
impairment of de novo adipogenesis than in female mice. Although STAT5
transcriptionally promotes adipogenesis in precursor cells
([214]53–[215]62), the use of mice with Cre expression driven by the
Adiponectin promoter restricts STAT5 ablation to fully differentiated
adipocytes. Thus, any effect of adipocyte STAT5 ablation on
adipogenesis is likely indirect, and additional experiments utilizing
animal models capable of tracking fat cell kinetics in vivo
([216]63–[217]66) are necessary to test this hypothesis. Taken
together, our data suggest that STAT5^AKO mice develop increased
adiposity through distinct sex-dependent mechanisms that are not
consistent with defects in adipose tissue lipolysis.
Data Availability Statement
The datasets presented in this study can be found in online
repositories or [218]Supplementary Material . The names of the
repository/repositories and accession number(s) can be found in the
article/[219] Supplementary Material .
Ethics Statement
The animal study was reviewed and approved by Institutional Animal Care
and Use Committee at Pennington Biomedical Research Center.
Author Contributions
Conceptualization, AR, CE, and JS. Methodology, AR, CE, HH, and PZ.
Investigation, AR, CE, HH, PZ, and TM. Formal Analysis, AR, HH, TA, and
SG. Writing – Original Draft, JS and AR. Writing – Review and Editing,
all authors. Visualization, AR, HH, TA, and SG. Supervision, JS, AR,
and CE. Funding Acquisition, JS and CE. All authors contributed to the
article and approved the submitted version.
Funding
This research project utilized facilities of the Genomics Core, the
Cell Biology and Bioimaging Core, and the Animal Metabolism and
Behavior Core at Pennington Biomedical that are supported in part by
COBRE (1P30GM118430) and NORC (NIH 2P30DK072476) center grants from the
National Institutes of Health. The Promethion Metabolic cage system was
purchased using funds from NIH shared instrumentation grant
S10OD023703. CE is supported by R03 DK122121 from NIH. This work was
supported by NIH grant R01DK052968 (JS) and pilot funding to CE from a
NORC center grant P30DK072476.
Conflict of Interest
The authors declare that the research was conducted in the absence of
any commercial or financial relationships that could be construed as a
potential conflict of interest.
Publisher’s Note
All claims expressed in this article are solely those of the authors
and do not necessarily represent those of their affiliated
organizations, or those of the publisher, the editors and the
reviewers. Any product that may be evaluated in this article, or claim
that may be made by its manufacturer, is not guaranteed or endorsed by
the publisher.
Acknowledgments