Abstract

   An intrauterine hyperglycemic environment has long-lasting effects on
   the offspring. Recent studies focused on fetal tissues, whereas we
   studied the development and molecular alteration of the placenta. By
   intercrossing male and female adult control (C) and first-generation
   offspring mice with gestational diabetes mellitus (F1-GDM), we obtained
   four groups of second generation (F2) offspring: 1) C♂-C♀, 2) C♂-GDM♀,
   3) GDM♂-C♀, 4) GDM♂- GDM♀. Placental weights in F1-GDM offspring were
   lower than in the control group. Placental weights in F2-offspring
   decreased through the paternal line. Placental RNA was extracted and
   analyzed using microarrays on day18.5 of pregnancy. This revealed 35
   upregulated imprinted genes and 10 down-regulated imprinted genes.
   Dlk1and Gtl2 were especially down-regulated and up-regulated,
   respectively, due to their abnormal methylation status. These findings
   suggest that intrauterine hyperglycemia decreased placental weight in
   the first generation, and this was transmitted paternally to the second
   generation in mice. They also suggest intrauterine hyperglycemia leads
   to abnormal placental Dlk1-Gtl2 expression due to DNA methylation in
   first and second generation mice.

   Keywords: epigenetic regulation, microarray analysis, intrauterine
   hyperglycemia environment, placenta, intergenerational transmission

INTRODUCTION

   Gestational diabetes mellitus (GDM) develops in 3% to 10% of pregnant
   women, and is associated with hyperglycemia during pregnancy [[34]1].
   An adverse intrauterine environment will affect multiple generations
   [[35]2, [36]3]. Offspring born from a GDM pregnancy suffer long-term
   outcomes like obesity, hypertension, type 2 diabetes mellitus (T2D)
   [[37]4–[38]6].

   The placenta conducts the maternal-fetal transport of nutrients and
   gases, and is exposed to the intrauterine conditions which may
   adversely affect placental and fetal development. This makes it a
   valuable tissue for investigating the molecular effects of GDM [[39]7,
   [40]8]. The placenta is more susceptible to maternal perturbations than
   the fetus, and complications may be transmitted to the fetus
   [[41]9–[42]11].

   Epigenetic processes modulate gene transcription, and can be inherited
   in a parent-of-origin-specific manner [[43]12, [44]13]. Methylation of
   DNA cytosine residue at the carbon 5 position is a common epigenetic
   feature and is often found in the sequence context CpG. Methylated DNA
   in gene promoters typically represses the corresponding gene [[45]14].
   Modified local and global DNA methylation patterns are found in the
   placenta of newborns with low birth weight [[46]15, [47]16]. The DNA
   methylation profile of mice placenta reflects the fetal growth and
   development with a potential impact on the offspring's health.

   We hypothesized that a GDM-inflicted environment could negatively
   affect placental function. We established a GDM mouse model of
   intrauterine hyperglycemia described by Ding et al. (2012) [[48]2] to
   study how the intrauterine environment of GDM affects the placenta in
   both first and second filial generations.

RESULTS

Placental and fetal weight in F1-offspring and F2-offspring of intrauterine
hyperglycemia

   We established a GDM mouse model by inducing hyperglycemia after
   pregnancy, and female and male F1 adults of control and GDM mice were
   intercrossed to obtain F2 offspring (Figure [49]1A). Placenta weight
   and fetal weight in F1-GDM mice was less than control offspring (0.099
   ± 0.002 g, p < 0.01; 1.058 ± 0.027 g, p < 0.01). However, in the F2-GDM
   offspring, placental weight decreased in only the GDM♂-C♀ group (0.110
   ± 0.003 g, p < 0.05) and the GDM♂-GDM♀ (0.108 ± 0.002 g, p < 0.01)
   (Figure [50]1B). Fetal weight was not different between the three
   F2-offspring (Figure [51]1C). No matter whether the phenotype changed
   or not on F1-GDM offspring or the F2-GDM offspring, there were no
   significant difference from the aspect of weight ratio (fetal birth
   weight/placental weight) (Figure [52]1D). The results suggested that
   intrauterine hyperglycemia had an intergenerational effect on mouse
   placenta.

Figure 1. Experiment design, placental weight, fetal weight, fetal/placental
weight of mice for control group (C), F1-GDM group, F2-GDM group.

   Figure 1
   [53]Open in a new tab

   (A) Experimental design. Circles designate females and squares
   designate males. Note that mating pairs were non-siblings. (B)
   Placental weight of all five groups. (C) Fetal weight of all five
   groups. (D) Ratio of fetal weight over placental weight of all five
   groups on. In (B)-(D), data are presented as mean ± SE (n>20 mice per
   group).

Effects of intrauterine hyperglycemia on global placental expression patterns
in both f1 and f2 offspring

   Microarray analysis showed distinct gene expression patterns for the
   F1-GDM and F2-GDM groups (Figure [54]2A). The F1-GDM group had 366
   genes (~2% of the total) whose expression was changed by >1.5 fold with
   P<0.05 or by <0.5 fold with P<0.05, which was comprised of 322
   up-regulated genes and 44 down-regulated genes ([55]Supplementary Table
   2). The F2-GDM group had 154 genes (~1.5% of the total) whose
   expression was changed by >1.5-fold with P<0.05, including 84
   up-regulated genes and 70 down-regulated genes placentae
   ([56]Supplementary 1). Between F1-GDM and F2-GDM placenta, there are
   only 16 up-regulated genes and 15 down-regulated genes compared with
   control group. Intrauterine hyperglycemia has different effects on gene
   expression in the placenta of F1 and F2 offspring. Real-time
   quantitative PCR analysis of 15 select genes confirmed the microarray
   results ([57]Supplementary Figure 1).

Figure 2.

   [58]Figure 2
   [59]Open in a new tab

   The basic statistic results of the genes identified in placenta both in
   3 groups according to microarray results (control group: C1, C2, C3;
   F1-GDM group: F1-1, F1-2, F1-3; F2-GDM groups (GDM♂- GDM♀): F2-1, F2-2,
   F2-3) (A) the number of identified differentiated probes including
   imprinted and non-imprinted genes; (B) the number of identified
   differentiated imprinted genes probes; (red: up-regulated genes; green:
   down-regulated genes; yellow: the feature-selected genes). (C)
   Hierarchical clustering of differentially expressed imprinted genes.
   Clustering was based on the expression levels of genes that were
   analyzed by the feature selection. Bar color represents a logarithmic
   scale from -3.0 to +3.0.

   To investigate the influence of intrauterine hyperglycemia on imprinted
   genes, we selected differentiated genes with a fold change > 1.2 fold
   and P < 0.05 or less than 0.8 with P<0.05. This revealed 35
   up-regulated genes (18 genes in the F1-GDM group and 17 genes in
   F2-GDM) and 10 down-regulated genes (4 genes in the F1-GDM group and 6
   genes in F2-GDM) (Figure [60]2B, [61]Supplementary Table 3). The heat
   map in Figure [62]2C was based on hierarchical clustering of 9 samples
   (3 control samples, 3 F1-GDM samples, 3 F2-GDM samples), which
   exhibited a direct concept of expression alteration. Five imprinted
   genes were altered in the same way in both groups, including Gtl2/Meg3,
   Asb4, Ube3a, Ascl2, Peg3 ([63]Supplementary Table 3). After multiple
   probes in microarray analysis, only Gtl2/Meg3 showed the same change in
   F1-GDM and F2-GDM placentae.

Bioinformation analysis

   The differentiated imprinted gene data was analyzed using gene ontology
   (GO) analysis to extract information relevant to involved pathways. In
   the biological process (BP) analysis, the majority of genes were
   classified into circadian rhythm, post-embryonic development,
   regulation of polysaccharide metabolic process, and regulation of
   glycogen metabolic process (Figure [64]3A). The cell component (CC)
   analysis showed that most genes affected the cytosol (Figure [65]3B).
   Molecular functional (MF) classification revealed that most genes were
   involved in ubiquitin-like protein transferase activity, insulin
   receptor binding, and ubiquitin protein ligase activity (Figure
   [66]3C).

Figure 3. Bio-information analysis for both F1-GDM and F2-GDM groups.

   Figure 3
   [67]Open in a new tab

   (A, B, C) GO annotation of total identified differentiated expressed
   imprinted genes in F1-GDM or F2-GDM (GDM♂- GDM♀) in three categories;
   (A) biological process (BP); (B) cellular component (CC); (C) molecular
   function (MF); (D) distribution of enriched KEGG pathway. (E) A
   dysfunctional regulation gene network was constructed based on the
   gene-gene (or protein-protein) interactions from KEGG, String and
   Starbase. Columns refer to related pathways, which are colored with
   gradient colors from midnight blue (smaller p-value) to lighter blue
   (bigger p-value) in the comparison between F1-GDM group and Control
   group or F2-GDM group and Control group.

   The KEGG pathway enrichment analysis of DEGs also provided insight into
   the cellular pathways associated with these DEGs. Seventeen pathways,
   including phenylalanine metabolism, histidine metabolism, and tyrosine
   metabolism, corresponded to differentially-expressed imprinted genes in
   F1-GDM placenta and F2-GDM placenta.

Common modulating network in F1-GDM and F2-GDM

   A dysfunctional regulation gene network was constructed based on the
   gene-gene (or protein-protein) interactions from KEGG, String and
   Starbase (Figure [68]3E). The imprinted gene Gtl2/Meg3 was upregulated
   in both F1 and F2 offspring of GDM groups. Meg3 is expressed
   predominantly from the maternal allele, in a reciprocal pattern from
   the nearby Dlk1 (delta-like 1 homolog) gene. Meg3 binds the PRC2
   chromatin modification complex (including gene product of Ezh2) in
   mouse embryonic stem cells, and controls Dlk1 expression at the
   Dlk1-Meg3 imprinted locus [[69]17]. Dlk1 expression is down-regulated
   in F2 GDM, supporting the dysregulation of Meg3-Dlk1 imprinted genes.

   Meg3-Dlk1 imprinted genes may downregulate downstream genes, such as
   TGFB1, Igf2, etc, which are also differentially expressed in both F1
   GDM and F2 GDM. Meg3 overexpression inhibit the TGF- β1 -stimulated
   cell proliferation and induced apoptosis [[70]18], and because high
   glucose can stimulate TGFBI expression [[71]19], decreasing Meg3 may
   stimulate TGFBI expression in an intrauterine hyperglycemic
   environment.

   Igf2 expression was down-regulated in placenta of F1-GDM and F2-GDM
   groups, which is similar to former observations that the imprinted
   genes Igf2 and H19 were down-regulated in pancreatic islets in both F1
   and F2 offspring of GDM [[72]2]. In addition, low Igf2 levels predict
   weight gain in subjects with T2D [[73]20]. The association between Meg3
   and Igf2 expression was observed in both F1 and F2 generations, and
   DLK1/Meg3 has been associated with IGF2 [[74]21]. These results suggest
   abnormal Meg3/Dlk1 expression is associated with developing diabetes.

Dlk1 and Gtl2 expression in intrauterine hyperglycemia

   According to the microarray assay, Dlk1 and Gtl2 were differentially
   expressed in both F1-GDM and F2-GDM placentae. Dlk1 and Gtl2 are
   closely associated with placental arborization and development of the
   spongiotrophoblastic and labyrinthine layers [[75]22]. We found that
   Dlk1 mRNA levels were lower in F1-GDM placenta than in the control
   group (Figure [76]4A). However, the relative Gtl2 mRNA level was higher
   (Figure [77]4B). In the F2 offspring of intrauterine hyperglycemia,
   relative Dlk1 mRNA levels decreased in the GDM♂-GDM♀ and GDM♂-C ♀
   groups (Figure [78]4C). Relative Meg3/Gtl2 mRNA levels were higher in
   C♂-GDM♀ and GDM♂-GDM♀ placentae (Figure [79]4D). These results suggest
   the imprinted gene had transmitted to next generation.

Figure 4. Dlk1, Gtl2 gene expression assessed by real-time quantitative PCR.

   [80]Figure 4
   [81]Open in a new tab

   In (A-B), Dlk1 and Gtl2 expression in placentae of F1-offspring on
   D18.5 (n=9 mice on for both F1-Control and F1-GDM offspring). In (C-D),
   Dlk1 and Gtl2 expression in placentae of F2-offspring on D18.5 (n=6
   mice for every group of F2 offspring). In (E-F), Dlk1 and Gtl2
   expression cultured in medium containing indicated concentrations of
   glucose of 24h (n=4 replicates/group) in at least three independent
   isolations. Data were analyzed with the Eq. 2-^ΔΔCT, where ΔΔCT = ΔCT
   (treatment group) -ΔCT (control group), and ΔCT = ΔCT (sample) - ΔCT
   (internal control). The values were presented as relative expression
   levels of mRNA. In all panels, data are presented as mean ± SE,
   ^*P<0.05, ^**P<0.01. Significance was determined by Student t test in
   F1 offspring and ANOVA in F2 offspring.

   In order to verify the direct effect of high-glucose on placental
   development, we collected mouse primary trophoblast cells at 14.5 dpc
   and cultured in a medium containing different concentrations of glucose
   for 24h. High glucose (25 mmol/L) decreased Dlk1 mRNA levels in
   trophoblast cells, whereas Gtl2 mRNA levels increased (Figure [82]4E,
   4F).

Intrauterine hyperglycemia induced hypermethylation of Dlk1-DMR and
hypomethylation of IG-DMR and Gtl2-DMR in placenta

   As imprinted genes, Dlk1 and Gtl2 allelic expression in mice is mainly
   regulated by allele-specific methylation at three DMRs. We collected
   placentae from 18.5dpc mice from the control, F1-GDM, and GDM♂-GDM♀
   groups. We analyzed methylation levels of 24 CpGs of the Dlk1-DMR, 32
   CpGs of the IG-DMR, and 12 CpGs of the Gtl2-DMR by bisulfate genomic
   sequencing PCR (BSP) (Figure [83]5A).

Figure 5. Methylation analysis of Dlk1/Gtl2 DMRs by bisulfate genomic
sequencing PCR.

   Figure 5
   [84]Open in a new tab

   (A) Schematic representation of mouse imprinted locus, showing the
   relative position of the Dlk1 and Gtl2 genes and indicating the
   location of the three DMRs known to contribute to Dlk1/Gtl2 imprinting.
   Exons are known as black (Dlk1 gene) and white (Gtl2 gene) rectangles
   with arrows for transcription start sites. The locations of the three
   DMRs within the Dlk1/Gtl2 imprinted locus represented by boxes are
   shaded to indicate preferential methylation of the maternal (M) or
   paternal (P) allele in each region. (B) Methylation status of
   individual DNA strands of Dlk1-DMR containing 24CpG sites and the
   average methylation ratio in each CpG site. (C) Methylation status of
   individual DNA strands of IG-DMR containing 32CpG sites and the average
   methylation ratio in each CpG site. (D) Methylation status of
   individual DNA strands of Gtl2-DMR containing 12CpG sites and the
   average methylation ratio in each CpG site. Ten clones per mouse; a
   total of 30 clones per group were sequenced. Each line represents the
   sequence of a single clone. CpG sites were shown as blank
   (unmethylated) or filled black (methylated) circles. In the histograms,
   results were expressed as methylation percentage of each CpG site. (n=3
   mice per group). ^*p<0.05 vs control (χ^2 test).

   Dlk1-DMR methylation levels were higher in the F1-GDM and GDM♂-GDM♀
   groups than the control group (Figure [85]5B). Analyzed by BSP, IG-DMR
   was highly methylated in control group (90% methylated), but not in the
   F1-GDM and GDM♂-GDM♀ groups (Figure [86]5C). Gtl2-DMR was less
   methylated in the F1-GDM and GDM♂-GDM♀ groups than in the control group
   (Figure [87]5D). These results indicate that intrauterine hyperglycemia
   could lead to Dlk1-DMR hypermethylation and IG-DMR and Gtl2-DMR
   hypomethylation in both F1 and F2 offspring.

DISCUSSION

   Hyperglycemia during pregnancy may lead to low birth weight [[88]23,
   [89]24], and low birth weight is associated with an increased risk of
   obesity, diabetes, and cardiovascular diseases during adult life
   [[90]25]. We found that intrauterine hyperglycemia induced lower fetal
   weight and placental weight in F1-offspring. Most GDM mice were
   susceptible to intrauterine hyperglycemia early in the pregnancy, which
   caused poor fetal nutrition [[91]26]. Undernutrition reduced birth
   weight, impaired glucose tolerance and obesity in both first- and
   second- generation offspring [[92]27]. In F2 offspring, placental
   weight decreased through the paternal line, but not the maternal line.
   Paternal phenotype transmission might be associated with the
   susceptibility of sperm in an intrauterine hyperglycemic environment
   [[93]2].

   According to the microarray analysis, intrauterine hyperglycemia
   affected 37 imprinted genes ([94]Supplementary 2). H19-Igf2 and
   Gtl2-Dlk1 have complementary functions in the mouse placenta
   [[95]28–[96]30]. We focused on the imprinted gene Dlk1 and Gtl2. In our
   study, down-regulated Dlk1 expression and up-regulated Gtl2 expression
   was present in F1 and F2 offspring. In the F2 offspring, decreased Dlk1
   levels were passed paternally, while increased Gtl2 levels were passed
   maternally. Dlk1 protein includes transmembrane protein and soluble
   protein, which promotes the activation of the insulin/IGF-I signaling
   pathway and adipogenesis inhibition [[97]29, [98]31]. Increased Meg3
   expression may suppress TGF-β and notch signaling pathway genes, such
   as TGFBI. That may lead to abnormal differentiation of the pancreas,
   which is an essential organ for insulin secretion.

   Genomic imprinting regulates the expression of a large group of
   monoallelically expressed genes in a parent-of-origin specific manner.
   Allele-specific DNA methylation occurs at differentially methylated
   regions (DMRs) of these genes. Allelic expression of imprinted genes
   Dlk1 and Gtl2 is regulated by allele-specific methylation at three
   DMRs: Dlk1DMR, IG DMR, and Gtl2 DMR [[99]32–[100]34]. We examined all
   three Dlk1/Gtl2 DMRs in placenta of F1-GDM and GDM♂-GDM♀ mice on
   embryonic day 18.5. Compared with control group, methylation levels
   were higher at Dlk1-DMR, and lower at IG-DMR and Gtl2-DMR in F1-GDM and
   GDM♂-GDM♀ placentae. These results suggested that in F1 and F2
   offspring after intrauterine hyperglycemia, altered Dlk1 and Gtl2
   expression in placenta is associated with altered DMR methylation.

   By establishing a GDM mouse model, we found that intrauterine
   hyperglycemia reduced the placental weight in both F1-GDM and F2-GDM
   offspring, and was paternally transmitted to the next generation.
   Intrauterine hyperglycemia alters imprinted genes, and could cause
   abnormal Dlk1 and Gtl2 expression related to dysregulation of Dlk1-Gtl2
   methylation in placenta of both F1 and F2 offspring. This epigenetic
   alteration has an intergenerational effect, which could explain
   adult-originated disease in the GDM offspring.

MATERIALS AND METHODS

Animal care and model

   The animal care and treatment protocols complied with institutional
   guidelines for laboratory animals established and approved by the
   Institutional Animal Care and Use Committee (IACUC), School of
   Medicine, Zhejiang University. Eight-week-old virgin female ICR mice
   (n=50) were mated with normal males. The mice were housed under a 12:12
   h light/dark cycle at 25 ± 0.5°C with 50-60% humidity, and were fed ad
   libitum with a standard diet and water.

   A copulation plug present after overnight mating determined pregnancy
   onset (defined as day 0 (D0) of pregnancy), and the females were
   randomly divided into a control group and an intrauterine hyperglycemia
   group with GDM (GDM group). After 12-h fast, Mice in the GDM group were
   induced with a single intraperitoneal injection of streptozotocin (STZ,
   Sigma, St. Louis, MO) in 0.1 mM citrate buffer (pH4.5) at a dose of 150
   mg/kg body weight (BW). Mice in the Control group received an equal
   volume of citrate buffer. On D3 of pregnancy, diabetes was confirmed by
   measuring blood glucose concentration through the tail vein and
   determined as a glucose level above 16.7 mM (300 mg/dl). Maternal blood
   glucose was remeasured on D7 and D20 of pregnancy to confirm the
   diabetic condition as previous described [[101]35, [102]36].

   Every litter size was controlled to 10-12 at birth to assure
   uniformity. The GDM pups were fostered by normoglycemic females until
   they were three weeks old. The female (♀) and male (♂) F1 adults of
   control (F1-Control) and GDM (F1-GDM) mice were intercrossed to
   generate F2 offspring, including 4 groups: 1) C♂-C♀, 2) C♂-GDM♀, 3)
   GDM♂-C♀, 4) GDM♂- GDM♀.

Placenta collection and in-vitro primary trophoblast cell cultivation

   The placentae were obtained and weighed at D18.5 from the pregnant mice
   in the control, F1-GDM, and F2-GDM groups. We used placentas 14.5 days
   post coitum (dpc) to isolate primary trophoblast cells as previous
   described [[103]37]. Isolated trophoblast cells were incubated in
   RPMI-1640 containing different glucose concentrations (5 mmol/L or 25
   mmol/L) for 24h (37°C).

RNA isolation, microarray analysis of placenta

   We obtained placentae of control, F1-GDM, and GDM♂-GDM♀ groups (3
   placentae per group) near term (18.5 dpc), each from a different
   litter. Placental RNA was isolated by TRIzol Reagent (Invitrogen Life
   Technologies, Carlsbad, CA, USA). After first-strand and second-strand
   cRNA preparation, placental cRNA was processed for use on the
   Affymetrix Mouse 430 2.0 array (Affymetrix, Santa Clara, CA) according
   to the manufacturer's instructions.

   Data were analyzed with Affymetrix GeneChip Operating Software
   Version1.4. Comparisons were made between control group and F1-GDM
   group, and between control group and F2-GDM group. Each placental
   sample was processed and analyzed independently. After normalization
   with dChip ([104]http://www.dchip.org/) estimation, the fold change and
   statistical significance (q-values) were calculated with SAM version
   2.23. Genes with more than 1.5-fold or less than 0.5-fold change with P
   < 0.05 were further analyzed by Molecule Annotation System and Cluster
   3.0.

Gene expression (quantitative PCR)

   Expression of select placental genes was verified by real-time
   quantitative PCR. Total placental RNA was isolated using RNAiso
   (TaKaRa, Japan), and cDNA was synthesized using oligo-dT and random
   primers (TaKaRa, Japan) for real-time quantitative PCR (ABI Prism
   7900HT; Applied Biosystems, Foster City, CA). GAPDH was the internal
   control. Full list of primer sequences is shown in [105]Supplementary
   Table 1.

Bioinformatics analysis

   The multi-omics data analysis tool, OmicsBean
   ([106]http://www.omicsbean.com:88/), was used to assign biological
   functions, subcellular locations, and molecular functions to each gene
   based on Gene Ontology (GO) categories. The overlaps between the lists
   of DEGs were detected by Fisher's exact test. P-value denotes the
   significance of a GO term enrichment in DEGs clusters and/or pathway
   correlations (P-value < 0.05 was considered significant). The Kyoto
   Encyclopedia of Genes and Genomes (KEGG) pathway analysis was performed
   in order to enrich high-level functions in the defined biological
   systems.

Gene interaction network construction

   Gene interactions from String database [[107]38] and KEGG database
   [[108]39] were used to annotate the gene interaction network between
   the differentially expressed genes in F1-GDM and F2-GDM mice. For
   protein-protein interactions (PPIs) in String database, PPIs with
   confidence score over 0.7 were taken into consideration. For non-coding
   RNAs in differentially expressed genes, the protein-RNA interactions
   were predicted using StarBase v2.0 [[109]40]. For all gene-gene
   interactions, Cytoscape was used to construct the gene-gene
   interactions.

DNA methylation (bisulfite genomic sequencing PCR)

   Genomic DNA was extracted from placenta of D18.5 mice in the control,
   F1-GDM, and GDM♂-GDM♀ groups. Bisulfite was converted using the EpiTect
   bisulfite kit (Qiagen, Valencia, CA) according to the manufacturer's
   instructions to deaminate cytosine to uracil; 5-methyl-cytosine was
   protected from deamination. The methylation status of Dlk1
   differentially methylated region (DMR), Gtl2-DMR, and intergenic
   differentially methylated region (IG-DMR) was determined by cloning and
   sequencing of bisulfite-treated DNA. The full list of primer sequences
   is shown in [110]Supplementary Table 2. The purified PCR products were
   cloned using the pMD 19-T vector system (Takara, Dalian, China). The
   cloned sequence was analyzed with 3730 DNA Analyzer Polymers (Applied
   Biosystems, Carlsbad, CA).

Statistical analysis

   Data were analyzed using SPSS 16.0, and were presented as mean ± SE.
   Statistical analysis including unpaired two-tailed Student's t-test,
   one-way analysis of variance (ANOVA), or Chi-square test were performed
   as described in the figure legends. P < 0.05 was considered
   statistically significant.

SUPPLEMENTARY MATERIALS FIGURE AND TABLES

   [111]oncotarget-09-22398-s001.pdf^ (1.2MB, pdf)
   [112]oncotarget-09-22398-s002.xlsx^ (169.9KB, xlsx)
   [113]oncotarget-09-22398-s003.xlsx^ (92KB, xlsx)

Acknowledgments