Abstract
Long non-coding RNA (LncRNAs) are newly highlighted key factors
controlling brown adipogenesis and development, but their regulatory
effect to white adipocyte is still merely understood. Deciphering their
underlying mechanism could be a novel way to discovering potential
targets of obesity. Therefore, we conducted a whole transcriptome
analysis in white adipose tissue from obese patients for the first
time. Six obese patients and five control subjects were selected for
microarray assay. Differentially expressed coding genes (DEGs), targets
of lncRNAs, and alternatively spliced genes in obesity group were
systematically compared in a functional framework based on a global
gene regulatory network. It was demonstrated that all the three kinds
of transcripts were enriched in pathways related to glucose metabolism
while only DEGs showed closer proximity to neuro-endocrine-immune
system. Thus, a lncRNA-regulated core network was constructed by a
stepwise strategy using DEGs as seed nodes. From the core network, we
identified a decreased lncRNA, uc001kfc.1, as potential cis-regulator
for phosphatase and tensin homolog (PTEN) to enhance insulin
sensitivity of white adipocytes in obese patients. We further validated
the down-regulation of uc001kfc.1 and PTEN in an independent testing
sample set enrolling 22 subjects via qRT-PCR. Although whether the
decreased uc001kfc.1 correlated with low risk of diabetes deserved to
be examined in an expanded cohort with long-term follow-up visit, the
present study highlighted the potential of lncRNA regulating glucose
homeostasis in human adipose tissue from a global perspective. With
further improvement, such network-based analyzing protocol proposed in
this study could be applied to interpreting function of more lncRNAs
from other whole transcriptome data.
Keywords: obesity, long non-coding ribonucleic acid, network, glucose
homeostasis, adipose tissue
Introduction
Over the past decade, obesity has become a global health-threatening
disease resulting from unbalanced energy intake and expenditure. Except
for the uncontrolled body weight, obesity often accompanied by
complications such as type 2 diabetes, fatty liver, and cardiovascular
diseases in a long-term and chronic manner ([33]Boles et al., 2017;
[34]Vecchie et al., 2018). Although several adipokines, cytokines and
related pathways have been implicated in obesity, its exact mechanism
still has not been elucidated yet.
Long non-coding RNA (lncRNA) is defined as transcript ≥ 200 nt with low
protein-coding potential. It has been shown to drive numerous
biological processes such as cell development, differentiation, and
metabolism since it was discovered. Recently, it was linked to
adipogenesis and cell differentiation in white and brown adipocyte
([35]Xiao et al., 2015; [36]Xiong et al., 2018). Based on mouse model,
it was found that lncRNA could dynamically regulate energy homeostasis
([37]Bai et al., 2017), control brown adipocyte differentiation
([38]Zhao et al., 2014), and maintain its morphology
([39]Alvarez-Dominguez et al., 2015). Besides, several lncRNAs
recognized from human adipose tissue were characterized to be related
to adipogenesis ([40]Ding et al., 2018). For example, Smith et al.
examined ectopic expression of HOTAIR and found that it could promote
differentiation of human gluteal preadipocytes ([41]Divoux et al.,
2014). Zhou et al. observed down regulation of MEG3 during adipogenesis
and confirmed its role in human adipogenic differentiation via
knockdown assay ([42]Li et al., 2017).
Although the above insightful studies have shown the essential role of
lncRNAs in fat biology, to interpret their mechanisms in obesity, a
systematical analysis of all transcripts is still urgently needed
because they are the base for lncRNAs’ function implementation. The
widespread effects of lncRNA on gene transcription include but are not
limited to shaping chromosome conformation, recruiting transcription
factors ([43]Marchese et al., 2017), and even indirectly targeting
downstream genes by interacting with other non-coding RNAs ([44]Cesana
et al., 2011). As a reflection of this delicate progress, whole
transcriptome provides a reliable and effective way to uncovering
latent mechanism of lncRNA in obesity.
In the present study, the whole transcriptome of adipose tissues
collected from obese individuals and control subjects were analyzed. A
stepwise network reconstruction strategy was proposed to explore
lncRNA-related functional modules altered in obesity. A decreased
lncRNA named uc001kfc.1, which was recognized from the core network,
was suggested to be related with enhanced insulin sensitivity of white
adipocytes in obese patients by cis-regulating phosphatase and tensin
homolog (PTEN). Our study highlighted the potential of lncRNA
regulating glucose homeostasis in human adipose tissue from a global
perspective, which would facilitate deeply understanding pathology of
obesity as well as developing novel therapeutic targets.
Methods
Subject Selection and Grouping
Our study was approved by the Institutional Review Board of Hebei
General Hospital and conducted according to the principles expressed in
the Declaration of Helsinki. As invasive surgery was the only way to
obtain adipose tissues, 33 asymptomatic cholecystolithiasis patients
receiving cholecystectomy were recruited as donors. All participates
had no acute symptoms induced by cholecystolithiasis such as infection,
fever, and jaundice. All volunteers had not taken any
anti-cholecystitis drugs such as magnesium sulfate, dehydrocholic acid,
ursodeoxycholic acid, and chenodeoxycholic acid before surgery.
Besides, we also excluded volunteers with medication history of weight
loss drugs such as orlistat, lorcaserin, naltrexone-bupropion,
phentermine-topiramate, and liraglutide. According to the
recommendations of the Working Group on Obesity in China, obesity was
defined as a body mass index (BMI) of at least 28 kg/m^2 ([45]Zhou,
2002). Thus, a total of 11 obese (BMI ≥ 28 kg/m^2) and 22 age-matched
control (BMI < 28 kg/m^2) subjects were enrolled in this study.
Visceral fat tissues from donors were collected during cholecystectomy
and immediately stored at −80°C until microarray assay. The use of
these subjects was approved by the hospital’s Ethics Committee and all
participants provided their written informed consents.
To evaluate the reliability of potential findings based on whole
transcriptome analysis, a training, and testing strategy was used for
sample grouping. We assigned the collected subjects into two
independent sample sets: discovery set and validation set. Totally, six
obese patients and five control subjects were randomly assigned to the
discovery set, which was used for microarray to high-throughput screen
potential transcripts critical to obese adipose functions. The
remaining samples were assigned to the validation set which was
designed to independently validate the identified transcripts.
Microarray Assay
The whole transcriptome profiles of subjects in microarray set were
detected through GeneChip^® Human Transcriptome Array 2.0 (HTA2.0)
according to the manufacturer’s instructions. The microarray assay was
performed by Shanghai Biotechnology Cooperation and the process were
detailed in the attachment file (see [46]Supplementary Method ). The
obtained microarray data was released to the National Center for
Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database
with accession number [47]GSE133786.
Principle Component Analysis
Principle component analysis of microarray data was performed through
Transcriptome Analysis Console 4.0 (TAC4.0) software to assess the
overall variances of selected subjects.
Differentially Expression Analysis
To determine a robust average unaffected by outliers, we use Tukey’s
biweight algorithm to compute average signal in each group
([48]Affymetrix, 2002). Subsequently, foldchanges were calculated as
formula (1).
[MATH: logFC= log2Biweight_SignalObesity¯− log2Biweight_SignalControl¯ :MATH]
(1)
One-way between-subject ANOVA was used to determine the significance of
difference between obesity and control group in term of P-value. Genes
with |logFC| ≥ 1.5 and P-value < 0.05 were regarded to be
differentially expressed between groups.
Assessment of Alternative Splicing Events
Here, we assessed alternative spliced genes based on transcripts from
exons, which were provided by HTA2.0 microarray. Splicing index (SI)
was used to compare the relative intensity of each exon between two
groups. SI for exon i in gene j was calculated as
[MATH: SIExon i= IntensityExon i, Obesity/IntensityGene j, ObesityIntensityExon i, Control/IntensityGene j, Control :MATH]
(2)
One-way between-subject ANOVA was used to determine each exon’s
significance of difference between obesity and control group in term of
P-value. Genes with |SI| ≥ 2 and P-value < 0.05 were regarded to be
alternatively spliced in obesity group. This process was performed by
TAC 4.0 software. Additionally, TAC provides annotation for
alternatively spliced genes based on “how well the data fits into
pre-defined splicing pattern,” which is measured by “Exon Event
Estimation Score” (EEES). Accordingly, alternatively spliced genes with
EEES > 0.2 were labeled by exon events such as intron retention,
cassette exon, alternative 3’ acceptor site, alternative 5’ donor site,
and mutually exclusive exons.
Prediction of cis- and trans-Targets
Differentially expressed lncRNAs were selected for prediction of
cis-/trans-targets. Cis-targets of lncRNAs were recognized by searching
genes intersecting the region which stretched from 10 kb upstream of
transcriptional start site to 10 kb downstream of termination site of
the interested lncRNA. To classify lncRNA trans-target genes, the BLAST
software was used to assess the impact of lncRNA binding on complete
mRNA molecules. The RNAplex program was then used to identify possible
trans-target genes of lncRNAs (e < 1E−20) ([49]Tafer and Hofacker,
2008).
Pathway Enrichment Analysis
In this paper, p-value is defined by the hypergeometric cumulative
distribution function (see [50]Supplementary Method ) to measure the
significance of candidate genes co-existing in the same Kyoto
Encyclopedia of Genes and Genomes (KEGG) pathway. A cutoff of P < 0.05
means that a pathway is significantly enriched by differential or
alternatively spliced genes. As there were few lncRNAs annotated in
pathway database, their predicted cis- and trans-targets were used
instead.
Reconstruction of Human Gene Regulatory Network
Here we constructed a human gene regulatory network based on KEGG
pathway database as our background network ([51]Kanehisa et al., 2017).
We extracted all the genes and their interactions from the *.kgml file
of each human pathway in KEGG. The obtained genes were afterwards
connected into an undirected network which contained 5,607 nodes and
51,748 edges.
Proximity Parameters
Connectivity and distance are two main parameters commonly used to
measure the proximity between two genes in a network. The former is
defined as whether there is at least one path bridging two genes, which
could measure how closely they are connected. And the latter is defined
as the length of the shortest path between two genes, denoting how far
they are away from each other. In this paper, the proximity between two
groups of genes was calculated by the average connectivity/distance of
each gene pair across two groups as previously described ([52]Yang et
al., 2012).
Generating Core Network of Obese Adipose Tissue
To deeply exploring mechanism hidden behind the adipose transcriptome,
a core network of obesity was generated. Candidate genes were
introduced in the core network step by step. Firstly, differentially
expressed genes were regarded as seed nodes. The relations between seed
nodes and their neighbors were extracted from the background network.
Then the extracted subnet was simplified by Steiner minimal tree
algorism, which helped to cut unnecessary branches mainly composed of
non-differential genes and keep important nodes bridging seed genes
([53]Klein and Ravi, 1995). Subsequently, differential lncRNAs were
linked to their targets in the simplified subnet. Finally, as the
function of a network was mainly carried out by the most connected
component, the largest component of the minimized network was extracted
as the core network of obese adipose tissue. To better interpret the
obtained core net, it was decomposed by simulated annealing algorithm
which helped to divide the network into several modules representing
different biological functions ([54]Guimerà and Amaral, 2005a;
[55]Guimerà and Amaral, 2005b).
Quantitative Real Time Polymerase Chain Reaction
Total RNA was isolated from adipose tissues by using TRIzol reagents
(Thermo Fisher, Life Technologies, USA) and following the
manufacturer’s instructions. Total RNA was quantified by NanoDrop 2000
Spectrophotometer (Thermo Fisher Scientific Inc., USA). Complementary
DNA (cDNA) was synthesized by Fast Quant cDNA Synthesis Kit (TIANGEN,
China). Quantitative PCR analysis were performed on ABI7500 Real-Time
PCR Detection System (Thermo Fisher, USA) with SuperMix Real PreMix
Plus (SYBR green) (TIANGEN, China). Sequences of primers were listed in
[56]Supplementary Table 1 . Expression data was normalized to the
geometric mean of the housekeeping gene GAPDH to control the
variability in expression levels and calculated as 2 − [(CT of
indicated genes) − (CT of GAPDH)], where CT represented the threshold
cycle for each transcript. PCR mixtures were initially heated to 95°C
for 15 min, followed by 40 cycles of 95°C for 10 s, 60°C for 20 s, and
72°C for 32 s.
Results
Overview of Samples and Altered Transcripts
Totally, 11 obese and 22 control subjects were enrolled in this study.
The basic information of all the participants was recorded in [57]Table
1 . There was no significant difference of age and gender between
obesity and control group. Obese subjects showed significantly higher
BMI, waistline, and hipline than control group, which was consistent
with the inclusion criteria. No significant differences of total
cholesterol (TC), total triglyceride (TG), high density lipoprotein
cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C)
were observed between obesity and control group (P > 0.05, Student’s t
test). It was unexpected that obesity group showed significantly lower
fasting blood glucose (FBG) than control one (P = 0.019,
Kolmogorov-Smirnov test).
Table 1.
Demographic information of the enrolled subjects.
Control group (n = 22)^§ Obesity group (n = 11)^§
Age (years) 54.82 ± 12.74 52.27 ± 13.39
Gender (male, %) 8 (36.4%) 3 (27.3%)
Waistline (cm) 85.56 ± 9.24 111.38 ± 13.61*
Hipline (cm) 95.47 ± 4.30 111.19 ± 9.79*
Height (cm) 162.59 ± 6.46 162.36 ± 6.83
Weight (kg) 63.14 ± 7.61 84.73 ± 8.52*
BMI (kg/m^2) 23.87 ± 2.24 32.21 ± 3.54*
TC (mmol/L) 4.84 ± 1.05 4.59 ± 0.58
TG (mmol/L) 1.39 ± 1.14 1.69 ± 1.03
HDL-C (mmol/L) 1.16 ± 0.31 1.32 ± 0.70
LDL-C (mmol/L) 3.31 ± 1.01 2.84 ± 0.80
FBG (mmol/L) 5.31 (4.27–6.78) 5.11 (4.75∼10.65)^#
[58]Open in a new tab
^§Information is separately presented as n (%) for dichotomous data,
median (min∼max) for non-normal data, and mean ± sd for normal data.
*A significant difference between obesity and control group exists if
p-value < 0.01 (two tailed Student’s t test).
^#A significant difference between obesity and control group exists if
p-value < 0.01 (two tailed Kolmogorov-Smirnov test).
Separately, six and five subjects were randomly selected from the
obesity and control group for microarray assay. Although the reduced
FBG in obesity group of microarray set was not significant (P = 0.079,
Kolmogorov-Smirnov test), the statistical characteristic of demographic
information about these subjects remained almost the same as the
original cohort (see [59]Supplementary Table 2 ). Whole transcriptome
profiles of adipose tissues from the selected samples were obtained
through HTA2.0 array. As quality control, PCA was performed based on
the microarray data and the top three principle component were
displayed as [60]Figure 1A , in which obese and control subjects were
assigned into distinct regions. Based on the cutoff of |logFC| > 1.5
and P-value < 0.05, 105 coding genes and 209 non-coding ones were
identified as differentially expressed transcripts (DETs). A total of
954 genes were predicted as cis- or trans-targets of the non-coding
DETs. Besides, 3,421 transcripts with |SI| > 2 and P-value < 0.05 were
filtered as alternatively spliced genes (ASGs), 679 of which were
annotated to pre-defined exon events. Overlaps of DETs, ASGs, and
predicted targets were displayed in [61]Figure 1B . As it can be seen
in [62]Figure 1B , about 1.4% differentially expressed lncRNAs (DELRs)
and 40% differentially expressed coding genes (DEGs) were alternatively
spliced, indicating different transcriptional regulation mechanisms
between coding and noncoding genes. [63]Figure 1C showed a global view
of all transcriptional alterations, including chromosome location and
fold change of DETs and ASGs. Interactions between DELRs and their
predicted targets were exhibited in center of [64]Figure 1C as well. As
it could be observed in [65]Figure 1C , none chromosome was preferred
by any type of the above transcripts in obese adipose tissue.
Figure 1.
[66]Figure 1
[67]Open in a new tab
Global view of altered transcripts in obese adipose tissue. (A)
Principle component analysis of microarray samples. (B) Overlapping of
differentially expressed transcripts. (C) Chromosome distribution of
altered transcripts. The heatmap of outer ring denotes fold changes of
all alternatively spliced genes (ASGs); the histogram of middle ring
denotes fold changes of differentially expressed coding genes (DEGs),
and the links in the center denotes interactions between differential
expressed long non-coding RNAs [differentially expressed lncRNAs
(DELRs)] and their targets. The symbols of connected DELRs, DEGs, and
annotated ASGs are respectively labeled in red, blue, and green color.
Links between DELRs and these transcripts are painted by the same color
as targets’ labels.
Disturbed Pathways Suggest Aberrant Glucose Metabolism in Obesity
Aiming to understand the disturbed primary biological process reflected
by whole transcriptome of obesity, we respectively performed enrichment
analysis for DEGs, ASGs, and targets of DELRs in KEGG pathways by using
hypergeometric cumulative distribution function. [68]Figure 2A
displayed all the pathways mapped by any class of these transcripts.
Totally, there were 11 pathways simultaneously enriched by two or more
kinds of transcripts and their corresponding names were tagged by
labels in [69]Figure 2A . Unexpectedly, there was no specific metabolic
pathway commonly enriched by the three kinds of transcripts. However,
it was noteworthy that, three (27.3%) out of the 11 commonly enriched
pathways, forkhead box protein O (FoxO), adenosine 5’-monophosphate
activated protein kinase (AMPK), and insulin signaling pathways were
typical ones regulating glucose metabolism, highlighting its
significance in obese adipose tissues.
Figure 2.
[70]Figure 2
[71]Open in a new tab
Enriched pathways and proximities to neuro-endocrine-immune (NEI)
system. (A) Enriched Kyoto Encyclopedia of Genes and Genomes pathways
of differentially expressed transcripts. (B) Average distances between
NEI genes and altered transcripts. (C) Average distances between NEI
genes and altered transcripts. *The distance/connectivity from NEI
genes to DETs is significantly different from those to non-DETs (P <
0.001, Kolmogorov-Smirnov test).
Topological Features Indicate Closer Proximity of Obese Differentially
Expressed Coding Genes to Neuro-Endocrine-Immune System
It was known that obesity was an endocrine disorder which would be
systematically regulated by the neuro-endocrine-immune (NEI) system
([72]Dixit, 2008). Thus, whether and how the altered transcripts
correlate with NEI system was examined by calculating their proximity
in the global gene regulatory network. According to dbNEI2.0, there
were a total of 1,736 genes involved in NEI system ([73]Zhang et al.,
2008). Nine hundred forty-five of them could be mapped into our
background network. Distance and connectivity between NEI genes and
each kind of altered transcripts were summarized in [74]Figure 2 , C.
Intriguingly, only DEGs showed significant shorter distance and
corresponding higher connectivity to NEI genes than those non-DEGs did
in the background network (P < 0.001, Kolmogorov-Smirnov test).
Moreover, the average distance of DEG-NEI was significantly shorter
than that of NEI-NEI (P < 0.001, Kolmogorov-Smirnov test). On the
contrary, ASGs and targets of DELRs presented significant longer
distance and lower connectivity to NEI genes (P < 0.001,
Kolmogorov-Smirnov test). The above different topological features
suggested that DEGs of obese adipose tissue pressed much closer to
physiologic perturbations in NEI system than DELRs and ASGs did.
Core Network of Obesity Reveals uc001kfc.1 as Regulator to PI3K Pathway in
Adipose Tissue
Aiming to characterize the core mechanism presented by transcriptome in
obesity, a minimized network was constructed based on the altered
transcripts in adipose tissue. The largest component of this network
was subsequently extracted as the core network of obese adipose tissue
because it undertook the main function of a network. To improve its
interpretability, the core net was decomposed and exhibited in
[75]Figure 3A . As it was shown in [76]Figure 3A , five out of seven
modules in the core network were regulated by DELRs, suggesting the
significant role of lncRNAs in obesity. The remaining two modules, M3
and M4, possessed almost half (48.8%) of inter-connections across
different modules (see [77]Supplementary Table 3 ) and hence could be
regarded as hub of the core network. The functions of DEGs in M3 and M4
could be mainly characterized as FoxO and insulin signaling pathways,
both of which were well known glucose metabolism regulatory pathways.
Figure 3.
[78]Figure 3
[79]Open in a new tab
Core network of obese adipose tissue. (A) The most connected component
in core network of obesity. Biological functions of connections within
each module are classified according to the hierarchical sort of
pathways in Kyoto Encyclopedia of Genes and Genomes pathway database.
(B) The TOP20 genes with highest degree in the core network. The bars
highlighted by red color denote genes related to PI3K pathway.
A total of 37 DEGs were contained in the core network. To measure their
contribution to the connection in the core net of obesity, degrees of
genes were measured by calculating how many other nodes that a gene
might connect with. [80]Figure 3B showed the top 20 of DEGs with
highest degree in the core network. As it was highlighted in [81]Figure
3B , the top six most connected DEGs were dominated by three genes
related to phosphatidylinositol 3-kinase (PI3K) pathway, which was
revolved around by FoxO and insulin signaling pathways that were
observed in hub modules. Additionally, there were 17 DELRs in the core
network and 12 of them were predicted to be able to regulate DEGs (see
[82]Supplementary Table 4 ). Among these DELRs, a lncRNA named
uc001kfc.1 was predicted to be able to cis-regulate a gene directly
participating in the PI3K pathway, PTEN. Notably, of all the seven
target DEGs, PTEN also possessed the most connections in the core
network (see [83]Supplementary Table 4 ). It was indicated that
uc001kfc.1 might be a vital regulator to glucose metabolism in obese
adipose tissue by potentially targeting PTEN in PI3K pathway.
Quantitative Real Time Polymerase Chain Reaction Validated Down-Regulation of
uc001kfc.1 in Independent Testing Samples
In order to validate the candidate genes important to glucose
metabolism in obesity, expression levels of altered transcripts were
further examined by quantitative real time PCR assay in the independent
testing sample set including 5 obese subjects and 17 controls. The
relative quantity and receiver operating characteristic (ROC) curves of
uc001kfc.1, PTEN, and FOXO1 were exhibited in [84]Figure 4 . Obesity
group showed significantly lower expression level of uc001kfc.1 and
PTEN than control group (P = 0.016 and 0.028, Student’s t test), which
was consistent with results from microarray (FC = −4.74 and −3.66, P =
0.021 and 0.030, one-way ANOVA). Their downstream protein, FOXO1 (FC =
−1.51, P = 0.015, one-way ANOVA), was validated to be down regulated in
obesity group as well although the difference did not reach to
significant level (P = 0.079, Student’s t test). But it was worth
noting that FOXO1 showed the best performance (AUC = 0.8) for
discrimination of obese and control subjects in testing sample set.
Figure 4.
[85]Figure 4
[86]Open in a new tab
Quantitative real-time polymerase chain reaction validation for
candidate genes. (A) Relative quantity of candidate genes. *A
significant difference between obese and control group exists if
p-value 0.05 (Student’s t test). (B) Receiver operating characteristic
curve of candidate genes.
Discussion
LncRNAs are newly proposed key factors controlling brown adipogenesis
and development, but their regulatory effect to white adipocyte is
still merely understood. Deciphering the underlying mechanism could be
a novel way to discovering potential targets of obesity. Hence, we
present the first comprehensive analysis for whole transcriptome of
visceral fat tissue from obese patients.
From network’s point of view, we systematically compared three main
kinds of transcripts, ASGs, targets of DELRs, and DEGs, in a functional
framework, which filled the gap in adipose transcriptome analysis.
Unexpectedly, only DEGs showed closer proximity to NEI genes, which was
the physiological basis of obesity. It was suggested that coding genes
might approach NEI system in a more precise manner than alternative
splicing genes and lncRNAs did. This result may be limited to the
inherent ability of the prediction process of ASGs and lncRNA’s targets
and could be complemented by abundant accumulation of these molecules
from wet experiments in the near future. Accordingly, to precisely
deciphering mechanism hidden behind transcriptome, we only use DEGs as
seed genes to start obesity’s core network construction.
Abnormal glucose homeostasis is one of the common complicating
disorders for obesity ([87]Chen et al., 2017). Other than specific
metabolic pathways, regulation of glucose metabolism were highlighted
in this study. Two typical regulatory pathways of glucose metabolism,
FoxO and insulin signaling pathways, were found to be simultaneously
enriched by different kind of transcripts. What’s more, they even
dominated main function of hub modules in the core network, indicating
significant contribution of them to the disturbed glycometabolism. PI3K
pathway is another worth noting pathway due to its high proportion
(3/6) in TOP6 genes with highest degree in the core net. It was not
only one of the most typical path for regulating glucose metabolism
([88]Huang et al., 2018), but also revolved around by FoxO and insulin
signaling pathways as basic functional part of them.
When searching for potential lncRNAs accounting for the above
perturbated processes in the core network, we found that lncRNA
participated most (5/7) functional modules. It was suggested that
lncRNA might play an extensive role in pathology of obese adipose
tissue. Enthusiastically, among the 17 DELRs in the core network of
obesity, only uc001kfc.1 was predicted as a potential regulatory lncRNA
to PTEN, the diphosphatase of phosphatidylinositol 3,4,5-trisphosphate
(PIP3). Therefore, we proposed a hypothesis as [89]Figure 3A showed: in
obesity group, decreased uc001kfc.1 might be linked to down-regulation
of PTEN, then the subsequently increased PIP3 could phosphorylate and
activate serine/threonine protein kinase (AKT). The activated AKT not
only can promote glucose uptake by helping glucose transporter type 4
(GLUT4) translocate to plasma membrane ([90]Beg et al., 2017), but also
depress gluconeogenesis via phosphorylating FOXO1 which would be
immediately exported from nucleus ([91]Pan et al., 2017).
To date, uc001kfc.1 (n381526, NONHSAT015462.2) has been only reported
in two literatures about cholesteatoma ([92]Gao et al., 2018) and
purpura nephritis ([93]Pang et al., 2019). According to NONCODE, this
lncRNA is highly expressed in human adipose tissue and no homologues in
other non-human animal models has been reported to date ([94]Fang et
al., 2018). We identified decreased uc001kfc.1 in obesity group by
whole transcriptome analysis and confirmed it in an independent testing
sample set via qPCR assay. The cis-target of this lncRNA, PTEN, was
demonstrated to be down-regulated in both microarray and testing
samples, which was consistent with previous observation in liver of
high fat diet induced obese mice ([95]Ji et al., 2015). Interestingly,
mice with adipocyte-specific elimination of PTEN have been observed to
gain more weight than wild type but retain improved insulin sensitivity
([96]Morley et al., 2015). Besides, PTEN haploinsufficiency has been
proved to be obesogenic but decrease the risk of type 2 diabetes (T2D)
owing to enhanced insulin sensitivity ([97]Pal et al., 2012). The
contrary effect of PTEN to obesity and T2D may be an explanation to our
observation of decreased FBG in obese subjects. Parallelly, as
downstream protein of PI3K pathway and transcription factor of G6PC
([98]Valenti et al., 2008), FOXO1 was detected to be decreased in
obesity group and confirmed via qPCR with AUC = 0.8. This was in
accordance with previous observation of down-regulated FOXO1 in adipose
tissue of T2D patients ([99]Rajan et al., 2016). The above results
together suggested enhanced insulin sensitivity in obese adipose
tissue, which was conflict with previous findings about co-occurrence
of obesity and impaired glucose regulation (IGR) in populations
([100]Das et al., 2014; [101]Akter et al., 2017). It may be related to
the pathological stage and the metabolic status of selected subjects.
Due to the invasive sampling means and cooperation of participants, we
only collected 11 obese and 22 control subjects in this study. Although
there were more control subjects than obese ones, the obesity rate was
still higher than the prevalence of obesity in China (15.7%) and the
reported metabolic syndrome rate (24%) in Chinese with gallstone
([102]Zhu et al., 2016; [103]Lu et al., 2017). To obtain reliable
results from microarray based on limited samples, we used Tukey’s
biweight algorithm which was robust to outliers for differential
expression analysis. Meanwhile, a training and testing strategy was
designed for independent validation of identified transcripts, which
further guaranteed the reliability of our study. As we can look forward
to, the identified lncRNA and its mechanism would shed new light on
target development of obesity once it is examined in larger cohorts and
gain-/loss-of function experiments.
In summary, the present study concentrated on searching for critical
transcripts and potential mechanisms in obese adipose tissue from a
systematical perspective. By network analysis of whole transcriptome,
our work identified decreased uc001kfc.1 as a potential lncRNA
cis-regulating PTEN to enhance insulin sensitivity of white adipocytes
in obese patients. Although abnormal glycometabolism have been revealed
in animal models and cohort studies of obesity, our observations
contributed new importance to the changes of glucose hemostasis in
obese adipose tissue. The mechanism proposed in this study may
ultimately help to develop new therapeutic interventions for the
treatment of obesity.
Conclusion
It was recently highlighted that lncRNA played essential role in fat
biology. Here, to explore the underlying mechanism of lncRNAs in
obesity, we conducted a whole transcriptome analysis in white adipose
tissues from obese patients for the first time. All the three kinds of
transcripts, DEGs, targets of DELRs, and ASGs were compared in a
functional framework based on a global gene regulatory network. They
all indicated abnormal glucose metabolism in obesity group while only
DEGs showed closer proximity to NEI system. Hence, a lncRNA-regulated
core network was constructed by using DEGs as seed nodes. From the core
network, we identified a decreased lncRNA, uc001kfc.1, as potential
cis-regulator for PTEN to enhance insulin sensitivity of white
adipocytes in obese patients.
Although whether uc001kfc.1 correlates with risk of T2D deserves to be
examined in an expanded cohort with long-term follow-up visit, the
present study provides novel insights into understanding the glucose
homeostasis in obese white adipocyte. With further improvement, the
network-based analyzing protocol proposed in this study would pave an
alternative but interesting way to deciphering function of more lncRNAs
from other whole transcriptome data.
Data Availability Statement
The microarray files are deposited at the Gene Expression Omnibus (GEO)
database and are available under the accession number: [104]GSE133786.
Ethics Statement
The present study involving human participants was reviewed and
approved by Institutional Review Board of Hebei General Hospital. All
participants provided their written informed consent to participate in
this study.
Author Contributions
LY designed the studies, carried out the research, interpreted the
results, and wrote the manuscript. XW assisted in data analysis and
revised the manuscript. HG and WZ performed the sample preparation and
data acquisition. WW assisted in data analysis. HM designed the study,
analyzed the data, reviewed and revised the manuscript, and is
responsible for the integrity of this work. All authors approved the
final version of the manuscript.
Funding
This work was supported by grants from National Natural Science
Foundation of China (81200638), National Natural Science Foundation of
Hebei, China (C2019307081), Medical Scientific Research Project of
Hebei, China (20190222, 20180019), and Government Funded Program for
Clinical Medicine Talent Training and Basic Research Project of Hebei,
China (2017).
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.
Supplementary Material
The Supplementary Material for this article can be found online at:
[105]https://www.frontiersin.org/articles/10.3389/fgene.2019.01133/full
#supplementary-material
[106]Click here for additional data file.^ (667.3KB, docx)
References