Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
Research ArticleComprehensive Characterization of Androgen-ResponsivelncRNAs Mediated Regulatory Network in Hormone-Related Cancers
DanWang ,1Mingyue Li,2 Jing Li,3 XuechaoWan,1 Yan Huang,1 Chenji Wang,1 Pu Zhang,1
Yangguang Xu,1 Zhe Kong,1 Yali Lu,1 Xinmei Wang,3 Chuan Liu,4 Chaoneng Ji ,1
and Liang Li 4
1State Key Laboratory of Genetic Engineering, Shanghai Engineering Research Center of Industrial Microorganisms, School ofLife Science, Fudan University, Shanghai 200433, China2Department of Rehabilitation Medicine, The Third Affiliated Hospital, Sun Yat-Sen University, Guangzhou 510620, China3Department of Pathology, Zibo Central Hospital, Zibo 2550036, China4Department of Thyroid and Breast Surgery, Zibo Central Hospital, Zibo 255036, China
Correspondence should be addressed to Chaoneng Ji; [email protected] and Liang Li; [email protected]
Received 26 June 2020; Revised 10 August 2020; Accepted 23 August 2020; Published 6 October 2020
Academic Editor: Jian Ma
Copyright © 2020 Dan Wang et al. This is an open access article distributed under the Creative Commons Attribution License,which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
The AR signaling pathway plays an important role in initiation and progression of many hormone-related cancers includingprostate, bladder, kidney, lung, and breast cancer. However, the potential roles of androgen-responsive long noncoding RNAs(lncRNAs) in hormone-related cancers remained unclear. In the present study, we identified 469 novel androgen-responsivelncRNAs using microarray data. After validating the accuracy of the array data, we constructed a transcriptional network whichcontained more than 30 transcriptional factors using ChIP-seq data to explore upstream regulators of androgen-responsivelncRNAs. Next, we conducted bioinformatics analysis to identify lncRNA-miRNA-mRNA regulatory network. To explore thepotential roles of androgen-responsive lncRNAs in hormone-related cancers, we performed coexpression network and PPInetwork analyses using TCGA data. GO and KEGG analyses showed these lncRNAs were mainly involved in regulating signaltransduction, transcription, development, cell adhesion, immune response, cell differentiation, and MAPK signaling pathway.We also highlight the prognostic value of HPN-AS1, TPTEP1, and LINC00623 in cancer outcomes. Our results suggest thatandrogen-responsive lncRNAs played important roles in regulating hormone-related cancer progression and could be novelmolecular biomarkers.
1. Introduction
In the previous decade, one of the most interesting findingsin the field of cancer biology was the novel annotation ofnoncoding RNA transcripts [1]. Long noncoding RNAs(lncRNAs), a kind of transcripts of more than 200 nucleo-tides without protein coding function, had been a rising starin cancer research. LncRNAs could act as oncogenes ortumor suppressors via regulating downstream targets at epi-genetic [2], transcriptional [3], and posttranscriptional [4]levels. No more than 100 lncRNAs including PCAT-14 [5],NEAT1 [6], PCAT5 [7], and HOTAIR [8] were identified
to play important roles in regulating tumor progression anddysregulated in different types of cancers (including prostatecancer, breast cancer, and kidney cancer). However, accord-ing to Sahu et al.’s report in 2015, more than 60000 lncRNAslocated in the whole human genome [9]. Thus, the functionalroles of most lncRNAs in cancers were still unknown.
Androgen/androgen receptor (AR) signaling has animportant role in initiation and progression of manyhormone-related cancers including prostate, bladder, kidney,lung, breast, and liver cancer [10]. Emerging evidences havedemonstrated that androgen-responsive proteins and miR-NAs regulated a series of cancer biological processes
HindawiDisease MarkersVolume 2020, Article ID 8884450, 18 pageshttps://doi.org/10.1155/2020/8884450
https://orcid.org/0000-0002-8530-0891https://orcid.org/0000-0002-5457-5067https://orcid.org/0000-0003-1919-8048https://creativecommons.org/licenses/by/4.0/https://creativecommons.org/licenses/by/4.0/https://doi.org/10.1155/2020/8884450
including cell cycle, apoptosis, and migration [11]. Ourgroup and others have focused on exploring the molecularfunctions of androgen-responsive lncRNAs in PCa [12–14].However, whether AR-regulated lncRNAs has differentialroles in the individual cells within these tumors that containa variety of cell types remains unclear. The understanding ofthe roles of AR downstream lncRNAs in many hormone-related tumors may eventually help us to develop better ther-apeutic approaches by targeting the AR.
In this study, we first identified novel androgen-responsive lncRNAs using microarray data. To exploreupstream regulating factors, we constructed a transcriptionalnetwork using ChIP-seq data. Next, we conducted bioinfor-matics analysis to identify lncRNA-miRNA-mRNA regula-tory network in hormone-related cancers. To explore thepotential roles of androgen-responsive lncRNAs inhormone-related cancers, we performed coexpression net-work analysis and GO and KEGG analyses using TCGA data.Our results suggest that androgen-responsive lncRNAsplayed important roles in regulating hormone-related cancerprogression and could act as novel molecular biomarkers.
2. Material and Methods
2.1. Cell Culture and Androgen Treatment. LNCaP cellswere purchased from the American Type Culture Collec-tion (Manassas, USA) which was confirmed by short tan-dem repeat (STR) analysis. All experiments were carriedout by each cell line at passages below 30. LNCaP cells weremaintained in RPMI 1640 medium (Corning, USA) sup-plemented with 10% FBS (Hyclone, USA) and cultured at37°C in 5% CO2. Androgen treatment assay was performedas described previously [15].
2.2. Microarray and Expression Data Sets. Total RNA of sam-ples were isolated by using TRIzol (Invitrogen) and theRNeasy mini kit (QIAGEN). Total RNA was quantified bythe NanoDrop ND-2000 (Thermo Scientific), and the RNAintegrity was assessed using Agilent Bioanalyzer 2100 (Agi-lent Technologies).
Total RNA from LNCaP cells, treated with 10nM DHT orethanol for 8 hours, were hybridized to SBC-ceRNA (4∗180K).The sample labeling, microarray hybridization, and washingwere performed based on the manufacturer’s standard proto-cols. The raw microarray data are listed in supplementarytable 1. To begin with, the raw data was normalized with thequantile algorithm. The probes that have at least 1 conditionout of 2 conditions and have flags in “P” were chosen forfurther data analysis. Differentially expressed lncRNAs werethen identified through fold change. The threshold set for up-and downregulated genes was a fold change ≥ 2:0.
2.3. Construction of Androgen-Responsive ceRNA Networks inPCa. We predicted the interactions between differentiallyexpressed lncRNAs and their target miRNAs theoreticallyby using miRcode. Finally, TargetScan and StarBase data-bases were both used to identify miRNAs which suppressmRNAs. The networks were drawn using Cytoscape 3.0(Figures 1(a)–1(d)).
2.4. Real-Time Reverse Transcription PCR (qRT-PCR)Analysis. qRT-PCR for lncRNAs and mRNAs was performedas described previously [16]. The Ct values were normalizedusing β-actin as internal control to estimate the differentexpression of genes. Relative mRNA expression was calcu-lated using the 2-ΔΔCt method. Each sample was run in tripli-cate to ensure quantitative accuracy.
2.5. Hierarchical Clustering Analysis. To generate an over-view of lncRNA/mRNA expression profiles between thetwo groups, the hierarchical clustering analysis was per-formed based on the expression value of all targets and themost significant differentially expressed lncRNA/mRNAsusing Cluster and TreeView program.
2.6. Gene Ontology (GO) and Pathway Analysis. We con-ducted Gene Ontology (GO) analysis to construct meaning-ful annotation of genes by using MAS system provided byCapitalBio company (Molecule Annotation System, http://bioinfo.capitalbio.com/mas3/). The ontology has covereddomains of biological processes, cellular components, andmolecular functions. The enriched GO terms were presentedby enrichment scores. KEGG pathway analysis was carriedout to determine the involvement of differentially expressedmRNAs in different biological pathways. P < 0:05 was usedas the criterion for statistical significance.
2.7. Statistical Analysis. The numerical data was presented asmean ± standard deviation (SD) of at least three determina-tions. Statistical comparisons between groups of normalizeddata were performed using T-test or Mann–Whitney U-testaccording to the test condition. A P < 0:05 was consideredstatistical significance with a 95% confidence level.
3. Result
3.1. Identification of Differentially Expressed Androgen-Responsive lncRNAs in Prostate Cancer Patients. The work-flow of this study is shown in supplementary description.High-throughput microarray assay was performed to identifyandrogen-responsive lncRNAs in LNCaP cells after 8 h treat-ment of DHT. The volcano plot filtering identified changedlncRNAs with statistical significance between two groups(Figure 2(a)). And the differentially expressed lncRNAs weredisplayed through fold change filtering (Figure 2(a)). Hierar-chical clustering showed androgen-induced and reducedlncRNAs in PCa (Figure 2(b)).
As a result, 469 lncRNAs were detected to be differen-tially regulated by fold change ≥ 2:0, among which 285lncRNAs were upregulated while 184 lncRNAs were down-regulated. Among these lncRNAs, upregulated lncRNAs aremore common than downregulated lncRNAs. The distribu-tion of the lncRNAs on the human chromosomes is depicted,and we did not observe specific chromosomal preference ofandrogen-responsive lncRNAs (Figure 2(c)). Most ofdifferentially expressed lncRNAs are transcribed from theintergenic regions (Figure 2(d)).
3.2. Validation of Androgen-Responsive lncRNAs in LNCaPCells. Three lncRNAs (lnc-FAM105A-2, AC003090.1, and
2 Disease Markers
http://bioinfo.capitalbio.com/mas3/http://bioinfo.capitalbio.com/mas3/
(a)
(b)
Figure 1: Continued.
3Disease Markers
RP11-31E13.2) with significant expression changing wererandomly selected for further qRT-PCR validation. To fur-ther validate array data, we also detected lnc-FAM105A-2(Figure 3(a)), AC003090.1 (Figure 3(b)), and RP11-31E13.2(Figure 3(c)) expressions in LNCaP cells afer 8 h treatmentwith different doses of DHT (0, 0.1, 1, 10, 100, and1000 nM). We also detected the expression of these lncRNAs
after DHT treatment in a time-dependent assay. lnc-FAM105A-2 expression was also significantly elevated (25-fold in 8 h and 20-fold in 24 h) upon DHT stimulation(Figure 3(d)). AC003090.1 expression was significantlyupregulated (sixfold in 24 h and eightfold in 48h) uponDHT stimulation (Figure 3(e)). However, we observed thatRP11-31E13.2 was significantly downregulated (20 percent
(c)
(d)
Figure 1: The predicted interactions between differentially expressed lncRNAs and their target miRNAs. We constructed androgen-inducedlncRNAs-miRNA (a) or androgen-induced lncRNAs-miRNA-mRNA networks (b). (c, d) The predicted androgen-reduced lncRNAs-miRNA(c) or androgen-reduced lncRNAs-miRNA-mRNA (d) networks were constructed based on bioinformatics analysis.
4 Disease Markers
in 12 h and 24h) after DHT treatment (Figure 3(f)). Weobserved the same expression tendency under DHT stimula-tion in a dose series. These results were consistent withmicroarray data.
To test the hypothesis that these lncRNAs were AR tar-gets, we assessed the presence of AR peaks in AC003090.1,lnc-FAM105A-2, and RP11-31E13.2 genomic region usingAR-ChIP-seq data generated in VCaP cells [16]. We foundthat AR peaks in these lncRNA loci significantly increasedafter DHT treatment and also observed that these peaks weresuppressed upon treatment with AR antagonist bicalutamide(Figures 3(g)–3(i)). Next, we detected AC003090.1, lnc-FAM105A-2, and RP11-31E13.2 after AR knockdown(Figure 3(j)). Notably, we found that lnc-FAM105A-2 andAC003090.1 were significantly downregulated and the
expression of RP11-31E13.2 was induced after AR silencingwhich depends on DHT treatment (Figures 3(k)–3(m)).
3.3. Construction of Upstream Transcriptional RegulatoryNetwork of Androgen-Responsive lncRNAs. In the presentstudy, we found only 30% androgen-responsive lncRNAshad AREs, suggesting that other transcriptional factors mayalso participate in regulating lncRNAs expression in responseto DHT. To construct upstream transcriptional regulatorynetwork of androgen-responsive lncRNAs, we used ChIP-seq data from ChIP base. Thirty TFs including AR, GATA6,and NFKB were identified. Cytoscape 3.0 was used to drawthe transcriptional regulatory network. Interestingly, someTFs such as ERG, EP300, and FOXA1 had been reported asAR targets or cofactors (Figure 4(a)).
15
10
DH
Tjia
8h5
5 10DHT_8 h
15
(a)
DHT-8 h DHT + 8 h3.00
2.00
1.00
0.00
–1.00
–2.00
–3.00
(b)
1000
800
Cou
nt
600
400
200
0
chr1
chr2
chr3
chr4
chr5
chr6
chr7
chr8
chr9
chr1
0ch
r11
chr1
2ch
r13
chr1
4ch
r15
chr1
6ch
r17
chr1
8ch
r19
chr2
0ch
r21
chr2
2ch
rXch
rY
Up-regulated IncRNADown-regulated IncRNATotal
(c)
Up-regulated IncRNADown-regulated IncRNA
5000
4000
3000
2000
1000
0
Bidi
rect
iona
l
Exon
ic_a
ntise
nse
Exon
ic_s
ense
Inte
rgen
ic
Intro
nic_
antis
ense
Intro
nic_
sens
e
Cou
nt
(d)
Figure 2: The microarray assay for androgen-responsive lncRNAs in LNCaP cells after 8 h treatment of DHT. (a) The volcano plot showingdifferentially expressed lncRNAs. (b) The heat map showing differentially expressed lncRNAs. (c) The statistics of androgen-responsivelncRNAs arranged by chromosome. (d) The classification of differentially expressed lncRNAs.
5Disease Markers
40
30
20
10Re
lativ
e exp
ress
ion
ofIn
c-FA
M10
5A-2
00 0.1 1 10 100 1000
(a)
Relat
ive e
xpre
ssio
n of
AC0
0309
0.1
10
8
6
4
2
00 0.1 1 10 100 1000
(b)
Relat
ive e
xpre
ssio
n of
RP1
1.31
E13.
2
1.5
1.0
0.5
0.00 0.1 1 10 100 1000
(c)
30
20
10
0
0 h
1 h
2 hF
old
chan
ged
over
neg
ativ
e con
trol
(Inc-
FAM
105A
-2)
4 h
8 h
12 h
24 h
48 h
DHT–DHT+
(d)
10
8
6
4
2
0
Fold
chan
ged
over
neg
ativ
eco
ntro
l (AC
0030
90.1
)
0 h
1 h
2 h
4 h
8 h
12 h
24 h
48 h
DHT–DHT+
(e)
2.0
1.5
1.0
0.5
0.0
Fold
chan
ged
over
neg
ativ
e con
trol
(RP1
1-31
E13.
2)
0 h
1 h
2 h
4 h
8 h
12 h
24 h
48 h
DHT–DHT+
(f)
Control 115
25.600 kb 25.610 kb 25.620 kb 25.630 kb69 kb
25.640 kb 25.650 kb 25.660 kb
015
015
015
015
0
0
15
Control 2
DHT1
DHT2
DHT+bicaltutamide 1
DHT+bicaltutamide 2
AC003090.1
(g)
Figure 3: Continued.
6 Disease Markers
3.4. Construction of Androgen-Responsive ceRNA Networks inPCa. One of the most important functions of lncRNAs wascompetitively binding to miRNA to affect protein-codinggenes translation and posttranscriptional regulation. Thus,we constructed an androgen-responsive ceRNA network in
PCa, combined with our previous identified androgen-responsive mRNAs and miRNAs datasets. We predicted theinteractions between differentially expressed lncRNAs andtheir target miRNAs theoretically by using miRcode. Finally,TargetScan and StarBase databases were both used to identify
Control 14
78.700 kb 78.720 kb 78.740 kb 78.760 kb149 kb
78.780 kb 78.800 kb 78.820 kb 78.840 kb
04
04
04
04
0
0
4
Control 2
DHT1
DHT2
DHT+bicaltutamide 1
DHT+bicaltutamide 2
RP11-31E13.2
(h)
Control 12
14.616 kb 14.618 kb 14.620 kb 14.622 kb12 kb
14.624 kb 14.626 kb
02
02
02
02
0
0
2
Control 2
DHT1
DHT2
DHT+bicaltutamide 1
DHT+bicaltutamide 2
Lnc-FAM105A-2
(i)
3
2
1
0
Relat
ive e
xpre
ssio
n of
IncR
NA
s
AR
Inc-
FAM
105A
-2
AC00
3090
.1
RP11
-31E
13.2
NCsiAR-1
(j)
4
3
2
1
0
Relat
ive e
xpre
ssio
n of
Inc-
FAM
105A
-2
NCsiAR
DHT– DHT+
ns
⁎⁎
(k)
1.5
1.0
0.5
0.0
Relat
ive e
xpre
ssio
n of
AC0
0309
0.1
NCsiAR
DHT– DHT+
ns
⁎
(l)
1.5
2.0
1.0
0.5
0.0
Relat
ive e
xpre
ssio
n of
RP1
1.31
E13.
2
NCsiAR
DHT–
ns
DHT+
⁎⁎
(m)
Figure 3: The validation of androgen-responsive lncRNAs in LNCaP cells. (a–c) RT-PCR analyses of lnc-FAM105A-2, AC003090.1, andRP11-31E13.2’s expressions in LNCaP cells treated with DHT for 24 h in dose series of 0 nM, 0.1 nM, 1 nM, 10 nM, 100 nM, and 1000 nM.Values of expressions treated with equal volume of vehicle were used as control. (d–f) RT-PCR analyses of lnc-FAM105A-2, AC003090.1,and RP11-31E13.2’s expressions in LNCaP cells treated with 10 nM DHT in time series of 0 h, 2 h, 8 h, 24 h, and 48 h. Values ofexpressions treated with equal volume of vehicle in the same time series were used as control. (g–i) Genome browser view of theAC003090.1, RP11-31E13.2, and lnc-FAM105A-2 genomic locus for AR ChIP-seq data tracks obtained from VCaP cells treated witheither vehicle or dihydrotestosterone (DHT) alone or combinations including DHT+Bicalutamide. Significant AR binding observed ineach data track are represented as peaks. (j) The expressions of lnc-FAM105A-2 and AC003090.1 and RP11-31E13.2 after AR knockdowncompared with NC in LNCaP cells. (k–m) The expressions of lnc-FAM105A-2, AC003090.1, and RP11-31E13.2, after AR knockdown inLNCaP cells with or without DHT treatment. Results are presented as the mean ± sd of three independent experiments. Significance wasdefined as P < 0:05 (∗P < 0:05, ∗∗P < 0:01, and ∗∗∗P < 0:001).
7Disease Markers
miRNAs which suppress mRNAs. The networks were drawnusing Cytoscape 3.0 (Figures 1(a)–1(d)).
3.5. The Expression Patterns of Androgen-ResponsivelncRNAs in Human Tissues. To better understand thetissue-specific expression characteristics of androgen-responsive lncRNAs in different human tissues, we analyzedRNA sequencing profiles (RNA-seq) from 22 differenthuman tissues in NOCODE database. Interestingly, we foundandrogen-responsive lncRNAs were significantly overex-pressed in testes, thyroid, adrenal, prostate, breast, ovary,
kidney, and lung (Figure 5(a)). We also compared the expres-sion levels of the same number of lncRNAs which were ran-domly selected and found the expression of these randomlyselected lncRNAs in 22 different human tissues were thesame (Figure 5(b)). These results revealed the potentialimportant roles of androgen-responsive lncRNAs inhormone-related tumors.
3.6. The Expression of Androgen-Responsive lncRNAs IsDysregulated in Hormone-Related Cancers. To characterizetumor-associated dysregulation of androgen-responsive
Figure 4: The clusters of upstream transcriptional regulatory network of androgen-responsive lncRNAs for identifying unknown transcriptionalfactors. Diagram analyses ChIP-seq data from ChIPbase. Construction of upstream transcriptional network that regulate androgen-responsivelncRNA expression. Circle: core TFs; red diamond: androgen-reduced lncRNAs; blue diamond: androgen-induced lncRNAs.
8 Disease Markers
lncRNA expression, we analyzed RNA sequencing profiles(RNA-seq) from 2037 tumors across 6 hormone-relatedtumors, including prostate cancer (PRAD), breast cancer(BRCA), kidney renal clear cell cancer (KIRC), kidney renalpapillary cell cancer (KIRP), ovarian serious cystadenocarci-noma (OV), and testicular germ cell tumors (TGCT) inTCGA. The information of clinical samples is listed in sup-plementary table 2. However, we did not find normalsamples in OV and TGCT datasets. Compared to theirnormal samples, we found that more than 45% lncRNAswere significantly up- or downregulated in PRAD (48%),BRCA (58%), KIRC (60%), and KIRP (45%) samples,indicating the important roles of androgen-responsivelncRNAs in these cancers (Figures 6(a)–6(d)). Bycomparing the dysregulated lncRNAs in different cancer
types, we found that ~10% of these altered lncRNAs werecancer-type specific, and the rest were shared by at leasttwo cancer types. We identified only 8 lncRNAs wasupregulated and 3 lncRNAs was downregulated in all fourcancer types (Figures 6(e) and 6(f)). The lncRNAs whosedysregulated expression was shared or unique amongdifferent cancer types are listed in supplementary table 3.
3.7. Biological Functions of Androgen-Responsive lncRNAs inHormone-Related Cancers. To predict the functions of theandrogen-responsive lncRNAs in hormone-related cancers,we adopted methods as described by Guttman et al. and Shenet al. We first constructed coexpression analysis to identifythe correlation between differentially expressed mRNAsand lncRNAs. Next, we mapped the androgen-responsive
Androgen-responsive IncRNAs
Relat
ive I
ncRN
As l
evels
(fpk
m)
20000
108642
0.2
0.1
0.0
Teste
s_F
Teste
s_R
Thyr
oid
Adre
nal
Brai
n
Pros
tate
Brea
st
Ova
ry
Kidn
ey
Lung
Adip
ose
Col
on
Fore
skin
Hea
rt
Hel
a_R
HLF
_1
HLF
_2
Live
r
Live
r_R
Lym
phno
de
Skelt
alm
uscle
Whi
tebl
oodc
ell
Nocode
(a)
Random selected IncRNAs
Relat
ive I
ncRN
As l
evels
(fpk
m)
20000
108642
0.2
0.1
0.0
Teste
s_F
Teste
s_R
Thyr
oid
Adre
nal
Brai
n
Pros
tate
Brea
st
Ova
ry
Kidn
ey
Lung
Adip
ose
Col
on
Fore
skin
Hea
rt
Hel
a_R
HLF
_1
HLF
_2
Live
r
Live
r_R
Lym
phno
de
Skelt
alm
uscle
Whi
tebl
oodc
ell
Nocode
(b)
Figure 5: The expression patterns of androgen-responsive lncRNAs in human tissues. The relative expression levels plot in testes, thyroid,adrenal, prostate, breast, ovary, kidney, and lung were showed for (a) androgen-responsive lncRNAs and (b) randomly selected lncRNAs.We observed that androgen-responsive lncRNAs were overexpressed in in hormone-related tissues such as testes, thyroid, adrenal,prostate, breast, and ovary.
9Disease Markers
lncRNAs coexpressed genes to STRING with experimentallyvalidated interactions score > 0:4. Then, PPI networks wereconstructed using the Cytoscape software. Interestingly, wefound lncRNAs mediated PPI networks in different kindsof cancers were not the same; however, a PPI network formedby olfactory receptor family existed in PRAD (Figure 7(a)),BRCA (Figure 7(b)), KIRC (Figure 7(c)), and KIRP
(Figure 7(d)). The olfactory receptor proteins are membersof GPCR family and shared a 7-transmembrane domainstructure with many neurotransmitter and hormonereceptors.
Next, we performed GO and KEGG pathway analyses foreach given lncRNA using the set of coexpressed mRNAs(Figures 7(e)-7(l)). In this study, the top 50 related mRNAs
PRAD
(a)
KIRC
(b)
(c) (d)
(e) (f)
BRCA KIRP
4.53
3.02
1.51
0.00
–1.51
–3.02
–4.53
Normal
Tumor
KIRP
6(5.8%)
9(8.7%)
9(8.7%)
6(5.8%)
4(3.8%)
8(7.7%)
7(6.7%)
0(0%)
1(1%)
2(1.9%)
9(8.7%)
4(3.8%)
7(6.7%)
27(26%)
5(4.8%)
PRAD
KIRPBRCA
15(14.3%)
10(9.5%)
24(22.9%)
5(4.8%)
3(2.9%)
1(1%)
3(2.9%)
9(8.6%)
6(5.7%)
3(2.9%) 0
(0%)6
(5.7%)
4(3.8%)
6(5.7%)
10(9.5%)
KIRPPRAD
KIRPBRCA
Figure 6: The dysregulated androgen-responsive lncRNAs in hormone-related cancers. The heat map of RNA sequencing profiles (RNA-seq)in cancer hormone-related tumors including (a) prostate cancer (PRAD), (b) kidney renal clear cell cancer (KIRC), (c) breast cancer (BRCA),and (d) kidney renal papillary cell cancer (KIRP). The overlap of four hormone-related tumors for (e) upregulated and (f) down-regulatedlncRNAs.
10 Disease Markers
(a) (b)
(c)
Figure 7: Continued.
11Disease Markers
(d)
Apoptosis
PRAD IncRNAs related GO analysis
Positive regulation of cell proliferation
Lipid metabolism
RNA splicing
Nervous system development
Cell differentiation
Translational elongation
Protein amino acid phosphorylation
Cell adhesion
Oxidation reduction
Development
Ion transport
Transcription
0 50 100Number of genes
150 200
Signal transduction
Immune response
(e)
BRCA IncRNAs related GO analysis
0 50 100Number of genes
150 250200
Apoptosis
Sodium ion transport
Potassium ion transport
Transport
Negative regulation of cell proliferation
Proteolysis
Cell differentiation
Protein amino acid phosphorylation
Immune response
Oxidation reduction
Ion transport
Cell adhesion
Development
Transcription
Signal transduction
(f)
Figure 7: Continued.
12 Disease Markers
KIRC IncRNAs related GO analysis
0 50 100Number of genes
150 200
Lipid metabolism
Proteolysis
Nervous system development
Spermatogenesis
Protein amino acid phosphorylation
Cell differentiation
Visual perception
Sodium ion transport
Immune response
Cell adhesion
Development
Ion transport
Oxidation reduction
Transcription
Signal transduction
(g)
KIRP IncRNAs related GO analysis
0 50 100Number of genes
150 200
Cell-cell signaling
Nervous system development
Visual perception
Positive regulation of cell proliferation
Cell differentiation
Proteolysis
Cell adhesion
Spermatogenesis
Protein amino acid phosphorylation
Oxidation reduction
Development
Ion transport
Immune response
Transcription
Signal transduction
(h)
PRAD IncRNAs related pathway analysis
Insulin signaling pathway
ErbB signaling pathway
PPAR signaling pathway
Wnt signaling pathway
Fc epsilon RI signaling pathway
GnRH signaling pathway
B cell receptor signaling pathway
Jak-STAT signaling pathway
MAPK signaling pathway
T cell receptor signaling pathway
Calcium signaling pathway
0 10 20Number of genes
30 40
(i)
BRCA IncRNAs related pathway analysis
0 10 20Number of genes
30 5040
mTOR signaling pathway
Calcium signaling pathway
ErbB signaling pathway
Insulin signaling pathway
Wnt signaling pathway
VEGF signaling pathway
Fc epsilon RI signaling pathway
Jak-STAT signaling pathway
B cell receptor signaling pathway
GnRH signaling pathway
T cell receptor signaling pathway
PPAR signaling pathway
MAPK signaling pathway
(j)
KIRC IncRNAs related pathway analysis
0 10 20Number of genes
30
Calcium signaling pathway
PPAR signaling pathway
MAPK signaling pathway
Pentose phosphate pathway
T cell receptor signaling pathway
(k)
KIRP IncRNAs related pathway analysis
0 10 20Number of genes
30 40
T cell receptor signaling pathway
GnRH signaling pathway
Calcium signaling pathway
MAPK signaling pathway
(l)
Figure 7: The biological functions of androgen-responsive lncRNAs in hormone-related cancers. The clusters of androgen-responsivelncRNAs mediated PPI networks in hormone-related cancers including (a) prostate cancer (PRAD), (b) breast cancer (BRCA), (c) kidneyrenal clear cell cancer (KIRC), and (d) kidney renal papillary cell cancer (KIRP). The GO analysis of lncRNAs in (e) PRAD, (f) BRCA, (g)KIRC, and (h) KIRP. The KEGG pathway analysis of lncRNAs in (i) PRAD, (j) BRCA, (k) KIRC, and (l) KIRP using the set ofcoexpressed mRNAs.
13Disease Markers
100
80
60
0 500 1000
p = 0.0098
Time (days)
PRAD survival
Ove
rall
surv
ival
(%)
1500 2000
HPN-AS1-high (n = 69)HPN-AS1-low (n = 69)
(a)
100
80
60
400 50
p = 0.0026
Time (months)
BRCA survival
Ove
rall
surv
ival
(%)
100
HPN-AS1-high (n = 823)HPN-AS1-low (n = 274)
(b)
100
80
60
40
20
00 50
p = ns
Time (months)
KIPC survival
Ove
rall
surv
ival
(%)
100
HPN-AS1-high (n = 399)HPN-AS1-low (n = 134)
(c)
100
80
60
0 500 1000
p = 0.0013
Time (days)
PRAD survival
Ove
rall
surv
ival
(%)
1500 2000
TPTEP1-high (n = 69)TPTEP1-low (n = 69)
(d)
100
80
60
400 50
p = ns
Time (months)
BRCA survival
Ove
rall
surv
ival
(%)
100
TPTEP1-high (n = 823)TPTEP1-low (n = 274)
(e)
100
80
60
40
20
00 50
p = 0.047
Time (months)
KIPC survival
Ove
rall
surv
ival
(%)
100
TPTEP1-high (n = 399)TPTEP1-low (n = 134)
(f)
Figure 8: Continued.
14 Disease Markers
of each lncRNAs were classified according to GO terms. Thetop 15 generally changed biological processes were listed. GOanalysis revealed that 9 biological processes (including signaltransduction, transcription, development, cell adhesion, iontransport, oxidation reduction, immune response, proteinamino acid phosphorylation, and cell differentiation) werewidely regulated by androgen-responsive lncRNAs in PRAD(Figure 7(e)), BRCA (Figure 7(f)), KIRC (Figure 7(g)), andKIRP (Figure 7(h)). We also observed some androgen-responsive lncRNAs-regulated biological processes were can-cer specific, for example, RNA splicing and translationalelongation in PRAD, cell-cell signaling, visual perceptionand spermatogenesis in KIRP and KIRC, and negative regu-lation of cell proliferation, transport, and potassium iontransport in BRAD.
KEGG analysis showed that androgen-responsivelncRNAs in PRAD were enriched in the calcium signalingpathway, T cell receptor signaling pathway, MAPK signalingpathway, Jak-STAT signaling pathway (Figure 7(i)).Androgen-responsive lncRNAs in BRCA were enriched in
the MAPK signaling pathway, PPAR signaling pathway, Tcell receptor signaling pathway (Figure 7(j)), GnRH signalingpathway. Androgen-responsive lncRNAs in KIRC wereenriched in the T cell receptor signaling pathway, Pentosephosphate pathway, MAPK signaling pathway, PPAR signal-ing pathway (Figure 7(k)). GO analysis revealed thatandrogen-responsive lncRNAs in KIRP were enriched inthe MAPK signaling pathway and calcium signaling pathway(Figure 7(l)).
3.8. Prognostic Significance of Androgen-Responsive lncRNAsin Hormone-Related Cancers. To evaluate possible prognosticvalue of androgen-responsive lncRNAs in hormone-relatedcancers, we analyzed TCGA database and found that theselncRNAs showed significant association with cancer progres-sion. Of note, we found three lncRNAs (HPN-AS1, TPTEP1,and LINC00623) could serve as biomarkers for PRAD(Figures 8(a), 8(d), and 8(g)), BRCA (Figures 8(b), 8(e), and8(h)), and KIRC (Figures 8(c), 8(f), and 8(i)). Kaplan-Meieranalysis showed that the BCR-free survival rates were lower
100
80
60
0 500 1000
p = ns
Time (days)
PRAD survival
Ove
rall
surv
ival
(%)
1500 2000
LINC00623-high (n = 69)LINC00623-low (n = 69)
(g)
100
80
60
400 50
p = 0.038
Time (months)
BRCA survival
Ove
rall
surv
ival
(%)
100
LINC00623-high (n = 823)LINC00623-low (n = 274)
(h)
100
80
60
40
20
00 50
p = 0.0081
Time (months)
KIPC survival
Ove
rall
surv
ival
(%)
100
LINC00623-high (n = 399)LINC00623-low (n = 134)
(i)
Figure 8: Androgen-responsive lncRNAs’ pathologic features in hormone-related cancers. (a–c) Kaplan-Meier curve analysis for thecorrelation between HPN-AS1 expression and disease-free survival time in patients with PRAD, BRAC, and KIRC. (d–f) Kaplan-Meiercurve analysis for the correlation between TPTEP1 expression and disease-free survival time in patients with PRAD, BRAC, and KIRC.(g–i) Kaplan-Meier curve analysis for the correlation between LINC00623 expression and disease-free survival time in patients withPRAD, BRAC, and KIRC. Significance was defined as P < 0:05 (∗P < 0:05, ∗∗P < 0:01, and ∗∗∗P < 0:001).
15Disease Markers
in HPN-AS1-high, TPTEP1-low, and LINC00623-highgroups in PRAD and KIRC patients. Interestingly, weobserved the high levels of TPTEP1 and LINC00623 werecorrelated with a longer biochemical recurrence-free survivaltimes in BRAC.
4. Discussion
AR signaling plays important roles in regulating tumorigen-esis and metastasis in several cancers including, kidney, lung,breast, and testis cancer [17]. LncRNAs have been reportedas important regulators of cell growth, proliferation, and dif-ferentiation in many types of cancer, such as prostate, blad-der, and kidney cancer [18]. Our group and others havefocused on exploring the molecular functions of androgen-responsive lncRNAs in PCa. However, there is still a lack ofextensively identification of androgen response lncRNAs. Inthis study, we focused on androgen-responsive lncRNAs todetermine their roles in hormone-related cancers. To identifyandrogen-responsive lncRNAs, we performed high-throughput microarray assay in PCa cell line LNCaP. A totalof 285 lncRNAs were upregulated and 184 lncRNAs weredownregulated after DHT treatment. Three lncRNAs(AC003090.1, lnc-FAM105A-2, and RP11-31E13.2) with sig-nificant expression change were randomly selected for fur-ther qRT-PCR validation. Our qRT-PCR results wereconsistent with microarray data. Importantly, ChIP-seq datashowed AR peaks in these lncRNA loci significantlyincreased after DHT treatment and these peaks were sup-pressed upon treatment with AR antagonist bicalutamide,which suggest these lncRNAs were direct targets of AR.
The functional importance of lncRNA has been paidmore and more attention by researchers. Recently, a fewstudies had indicated some transcriptional factors, includingEZH2 [19], AR, MYC [20], and ERα [21], could regulatelncRNAs’ expression. However, the transcriptional regula-tion of the majority of lncRNAs remained unknown. Ourgroups had identified a series of androgen-responsivelncRNAs [12]. Of note, in our previous study, agilent humanlncRNA array was performed to simultaneously observelncRNA expressions in androgen-dependent LNCaP cellsunder DHT stimulation in time points of 0 h and 2h, respec-tively. The “0h” worked as the control representing cellularstatus before DHT stimulation. Supervised analysis of themicroarray data showed 3767 deregulated lncRNA tran-scripts (1991 transcripts upregulated and 1776 transcriptsdown-regulated) produced from 2980 genes, with an averageexpression level over 2-fold change in 2 h compared to 0 h(GSE72866). In this study, high-throughput microarray assay(SBC-ceRNA (4∗180K)). was performed to identifyandrogen-responsive lncRNAs in LNCaP cells after 8 h treat-ment of DHT. Several differences between the both studieswere observed. First, the platform used in both studies wasdifferent. Second, our previous studies mainly aimed to iden-tify early response lncRNAs. This mainly aimed to identifylate response lncRNAs. We also found that more than 50 per-cent androgen-responsive lncRNAs did not contain AREs,suggesting that other TFs may also be involved in regulatinglncRNA expression. Interestingly, Crea et al. also found that
androgen-regulated PCAT18 was not a direct target of AR[22]. To comprehensively reveal upstream transcriptionalregulatory network of androgen-responsive lncRNAs, weanalyzed ChIP-seq data from ChIPbase.
More than 30 TFs were identified in our transcriptionalnetwork. Among them, AR, GATA6, and NFKB showedthe most widely regulation of androgen-responsive lncRNAs.Interestingly, we also identified that many hormone recep-tors (such as ER and GR) and AR-cofactors (includingFOXA1, EP300, and GR) participated in regulating theselncRNAs. By analyzing our previous reports, we found that85 percent of these transcriptional factors were also regulatedby AR.
To understand the tissue-specific expression characteris-tics of androgen-responsive lncRNAs, we analyzed theirexpressions in 22 different human tissues and foundandrogen-responsive lncRNAs were significantly overex-pressed in testes, thyroid, adrenal, prostate, breast, ovary,kidney, and lung. Hormone-related cancers, such as prostatecancer and breast cancer, were the most commonly diag-nosed cancers in the worldwide. Accumulating evidencedemonstrated that AR also played key roles in regulatinghormone-related cancer progression. In the previous reports,a series of lncRNAs (such as CTBP1-AS in PCa [13],lncARSR in renal cancer [23], and NKILA in breast cancer[24] were reported to play key roles in regulating cancer pro-gression. However, there were no reports that revealedandrogen-responsive lncRNA roles in these cancers.
To characterize tumor-associated dysregulation ofandrogen-responsive lncRNAs’ expression, we analyzedRNA sequencing profiles (RNA-seq) from 2037 tumorsacross 6 cancer hormone-related tumors (PRAD, BRCA,KIRC, KIRP, OV, and TGCT) in TCGA and found that morethan 45% lncRNAs were significantly up- or downregulatedin PRAD (48%), BRCA (58%), KIRC (60%), and KIRP(45%) samples, indicating the important roles of androgen-responsive lncRNAs in these cancers. To predict the func-tions of the androgen-responsive lncRNAs in hormone-related cancers, we performed GO and KEGG pathway anal-yses using the set of coexpressed mRNAs. Our results showedthat androgen-responsive lncRNA played different roles inhormone-related cancers. Eight lncRNAs were upregulated,and 3 lncRNAs were downregulated in all four cancer types.To evaluated prognostic values of differentially expressedandrogen-responsive lncRNAs in hormone-related cancers,we then analyzed TCGA database and found that threelncRNAs (HPN-AS1, TPTEP1, and LINC00623) could serveas biomarkers for PRAD, BRCA ,and KIRC due to theirexpression levels, which were significantly correlated to over-all survival.
In conclusion, our study represents the comprehensiveanalysis of androgen-responsive lncRNAs in prostate cancer.To reveal the upstream regulating factors, we constructedtranscriptional network of lncRNAs for the first time. Wealso analyzed the expression patterns of lncRNAs in 22 dif-ferent kinds of human tissues. Of note, we found thatandrogen-responsive lncRNAs were dysregulated inhormone-related cancers such as PRAD, BRCA and KIRC.GO and KEGG pathway analyses showed that androgen-
16 Disease Markers
responsive lncRNAs played different roles in differenthormone-related cancers. We also highlight the prognosticvalue of HPN-AS1, TPTEP1, and LINC00623 in cancer out-comes. Further studies, however, will be needed to develop adeeper understanding of the mechanistic roles of androgen-responsive lncRNAs in the development and progression ofcancer.
Data Availability
The data used to support the findings of this study are avail-able from the corresponding authors upon request.
Conflicts of Interest
The authors declare no financial conflicts of interest.
Authors’ Contributions
Dan Wang, Chaoneng Ji, and Liang Li conceptualized thestudy. Mingyue Li, Jing. Li, Xuechao Wan, Yan Huang, andChenji Wang performed data curation. Pu Zhang and Yang-guang Xu made formal analysis. Zhe Kong, Yali Lu, XinmeiWang, and Chuan Liu performed the investigation.Writing-origin draft: Dan Wang and Xuechao Wan wrotethe original draft. Dan Wang, Chaoneng Ji, and Liang Lireviewed and edited the manuscript.
Acknowledgments
This work was supported by grant 81773024 from theNational Natural Science Foundation of China and ProjectSKLGE-1901 supported by the Open Research Funds of theState Key Laboratory of Genetic Engineering, Fudan Univer-sity. This work was supported by a grant 20ZR1404500 fromShanghai Science and Technology Commission.
Supplementary Materials
Supplementary 1. the raw microarray data were listed in sup-plementary table 1.
Supplementary 2. the information of clinical samples waslisted in supplementary table 2.
Supplementary 3. the lncRNAs whose dysregulated expres-sion was shared or unique among different cancer types arelisted in supplementary table 3.
References
[1] M. Huarte, “The emerging role of lncRNAs in cancer,” NatureMedicine, vol. 21, no. 11, pp. 1253–1261, 2015.
[2] H. L. Carlson, J. J. Quinn, Y. W. Yang, C. K. Thornburg, H. Y.Chang, and H. S. Stadler, “LncRNA-HIT functions as an epige-netic regulator of chondrogenesis through its recruitment ofp100/CBP complexes,” PLoS Genetics, vol. 11, no. 12, articlee1005680, 2015.
[3] C. A. McHugh, C. K. Chen, A. Chow et al., “The Xist lncRNAinteracts directly with SHARP to silence transcription throughHDAC3,” Nature, vol. 521, no. 7551, pp. 232–236, 2015.
[4] J. H. Yoon, K. Abdelmohsen, and M. Gorospe, “Posttranscrip-tional gene regulation by long noncoding RNA,” Journal ofMolecular Biology, vol. 425, no. 19, pp. 3723–3730, 2013.
[5] Y. Wang, Y. Hu, G. Wu et al., “Long noncoding RNA PCAT-14 induces proliferation and invasion by hepatocellular carci-noma cells by inducing methylation of miR-372,” Oncotarget,vol. 8, no. 21, pp. 34429–34441, 2017.
[6] M. Idogawa, T. Ohashi, Y. Sasaki, H. Nakase, and T. Tokino,“Long non-coding RNA NEAT1 is a transcriptional target ofp53 and modulates p53-induced transactivation and tumor-suppressor function,” International Journal of Cancer,vol. 140, no. 12, pp. 2785–2791, 2017.
[7] A. Ylipaa, K. Kivinummi, A. Kohvakka et al., “Transcriptomesequencing reveals PCAT5 as a novel ERG-regulated longnoncoding RNA in prostate cancer,” Cancer Research,vol. 75, no. 19, pp. 4026–4031, 2015.
[8] A. Zhang, J. C. Zhao, J. Kim et al., “LncRNA HOTAIRenhances the androgen-receptor-mediated transcriptionalprogram and drives castration-resistant prostate cancer,” CellReports, vol. 13, no. 1, pp. 209–221, 2015.
[9] A. Sahu, U. Singhal, and A. M. Chinnaiyan, “Long noncodingRNAs in cancer: from function to translation,” Trends Cancer,vol. 1, no. 2, pp. 93–109, 2015.
[10] A. D. Lamb, C. E. Massie, and D. E. Neal, “The transcriptionalprogramme of the androgen receptor (AR) in prostate cancer,”BJU International, vol. 113, no. 3, pp. 358–366, 2014.
[11] S. L. Romero-Cordoba, I. Salido-Guadarrama, M. Rodriguez-Dorantes, and A. Hidalgo-Miranda, “miRNA biogenesis: bio-logical impact in the development of cancer,” Cancer Biology& Therapy, vol. 15, no. 11, pp. 1444–1455, 2014.
[12] X. Wan, W. Huang, S. Yang et al., “Identification of androgen-responsive lncRNAs as diagnostic and prognostic markers forprostate cancer,” Oncotarget, vol. 7, no. 37, pp. 60503–60518,2016.
[13] K. Takayama, K. Horie-Inoue, S. Katayama et al., “Androgen-responsive long noncoding RNA CTBP1-AS promotes pros-tate cancer,” The EMBO Journal, vol. 32, no. 12, pp. 1665–1680, 2013.
[14] A. Misawa, K. I. Takayama, T. Fujimura, Y. Homma,Y. Suzuki, and S. Inoue, “Androgen-induced lncRNAPOTEF-AS1 regulates apoptosis-related pathway to facilitatecell survival in prostate cancer cells,” Cancer Science,vol. 108, no. 3, pp. 373–379, 2017.
[15] X. Li, X. Wan, H. Chen et al., “Identification of miR-133b andRB1CC1 as independent predictors for biochemical recurrenceand potential therapeutic targets for prostate cancer,” ClinicalCancer Research, vol. 20, no. 9, pp. 2312–2325, 2014.
[16] W. Mo, J. Zhang, X. Li et al., “Identification of Novel AR-Targeted MicroRNAs Mediating Androgen Signalling throughCritical Pathways to Regulate Cell Viability in Prostate Can-cer,” PLoS One, vol. 8, no. 2, p. e56592, 2013.
[17] C. Chang, S. O. Lee, S. Yeh, and T. M. Chang, “Androgenreceptor (AR) differential roles in hormone-related tumorsincluding prostate, bladder, kidney, lung, breast and liver,”Oncogene, vol. 33, no. 25, pp. 3225–3234, 2014.
[18] E. S. Martens-Uzunova, R. Böttcher, C. M. Croce, G. Jenster,T. Visakorpi, and G. A. Calin, “Long noncoding RNA in pros-tate, bladder, and kidney cancer,” European Urology, vol. 65,no. 6, pp. 1140–1151, 2014.
[19] M. Sun, X. H. Liu, K. H. Lu et al., “EZH2-mediated epigeneticsuppression of long noncoding RNA SPRY4-IT1 promote s
17Disease Markers
http://downloads.hindawi.com/journals/dm/2020/8884450.f1.xlsxhttp://downloads.hindawi.com/journals/dm/2020/8884450.f2.xlsxhttp://downloads.hindawi.com/journals/dm/2020/8884450.f3.xlsx
NSCLC cell proliferation and metastasis by affecting theepithelial-mesenchymal transition,” Cell Death & Disease,vol. 5, no. 6, article e1298, 2014.
[20] G. Doose, A. Haake, S. H. Bernhart et al., “MINCR is a MYC-induced lncRNA able to modulate MYC's transcriptional net-work in Burkitt lymphoma cells,” Proceedings of the NationalAcademy of Sciences of the United States of America, vol. 112,no. 38, pp. E5261–E5270, 2015.
[21] D. Chakravarty, A. Sboner, S. S. Nair et al., “The oestrogenreceptor alpha-regulated lncRNA NEAT1 is a critical modula-tor of prostate cancer,” Nature Communications, vol. 5, no. 1,p. 5383, 2014.
[22] F. Crea, A. Watahiki, L. Quagliata et al., “Identification of along non-coding RNA as a novel biomarker and potentialtherapeutic target for metastatic prostate cancer,” Oncotarget,vol. 5, no. 3, pp. 764–774, 2014.
[23] L. Qu, J. Ding, C. Chen et al., “Exosome-transmitted lncARSRpromotes sunitinib resistance in renal cancer by acting as acompeting endogenous RNA,” Cancer Cell, vol. 29, no. 5,pp. 653–668, 2016.
[24] J. M. Dijkstra and D. B. Alexander, “The “NF-ĸB interactinglong noncoding RNA” (NKILA) transcript is antisense tocancer-associated gene PMEPA1,” F1000Research, vol. 4,p. 96, 2015.
18 Disease Markers
Comprehensive Characterization of Androgen-Responsive lncRNAs Mediated Regulatory Network in Hormone-Related Cancers1. Introduction2. Material and Methods2.1. Cell Culture and Androgen Treatment2.2. Microarray and Expression Data Sets2.3. Construction of Androgen-Responsive ceRNA Networks in PCa2.4. Real-Time Reverse Transcription PCR (qRT-PCR) Analysis2.5. Hierarchical Clustering Analysis2.6. Gene Ontology (GO) and Pathway Analysis2.7. Statistical Analysis
3. Result3.1. Identification of Differentially Expressed Androgen-Responsive lncRNAs in Prostate Cancer Patients3.2. Validation of Androgen-Responsive lncRNAs in LNCaP Cells3.3. Construction of Upstream Transcriptional Regulatory Network of Androgen-Responsive lncRNAs3.4. Construction of Androgen-Responsive ceRNA Networks in PCa3.5. The Expression Patterns of Androgen-Responsive lncRNAs in Human Tissues3.6. The Expression of Androgen-Responsive lncRNAs Is Dysregulated in Hormone-Related Cancers3.7. Biological Functions of Androgen-Responsive lncRNAs in Hormone-Related Cancers3.8. Prognostic Significance of Androgen-Responsive lncRNAs in Hormone-Related Cancers
4. DiscussionData AvailabilityConflicts of InterestAuthors’ ContributionsAcknowledgmentsSupplementary Materials