Circ-HuR suppresses HuR expression and gastric cancer progression by inhibiting CNBP transactivation
Circular RNAs (circRNAs), a subclass of non-coding RNAs, play essential roles in tumorigenesis and aggressiveness. Our previous study has identified that circAGO2 drives gastric cancer progression through activating human antigen R (HuR), a protein stabilizing AU-rich element-containing mRNAs. However, the functions and underlying mechanisms of circRNAs derived from HuR in gastric cancer progression remain elusive.
CircRNAs derived from HuR were detected by real-time quantitative RT-PCR and validated by Sanger sequencing. Biotin-labeled RNA pull-down, mass spectrometry, RNA immunoprecipitation, RNA electrophoretic mobility shift, and in vitro binding assays were applied to identify proteins interacting with circRNA. Gene expression regulation was observed by chromatin immunoprecipitation, dual-luciferase assay, real-time quantitative RT-PCR, and western blot assays. Gain- and loss-of-function studies were performed to observe the impacts of circRNA and its protein partner on the growth, invasion, and metastasis of gastric cancer cells in vitro and in vivo.
Circ-HuR (hsa_circ_0049027) was predominantly detected in the nucleus, and was down-regulated in gastric cancer tissues and cell lines. Ectopic expression of circ-HuR suppressed the growth, invasion, and metastasis of gastric cancer cells in vitro and in vivo. Mechanistically, circ-HuR interacted with CCHC-type zinc finger nucleic acid binding protein (CNBP), and subsequently restrained its binding to HuR promoter, resulting in down-regulation of HuR and repression of tumor progression.
Circ-HuR serves as a tumor suppressor to inhibit CNBP-facilitated HuR expression and gastric cancer progression, indicating a potential therapeutic target for gastric cancer.
KeywordsCircular RNAs Human antigen R Gastric cancer CCHC-type zinc finger nucleic acid binding protein
Translational start codon
CCHC-type zinc finger nucleic acid binding protein
Clustered regularly interspaced short palindromic repeats
Catenin beta 1
Dead mutant of Cas9 endonuclease
Elongation factor-1α promoter
Heparin binding growth factor
human U6 promoter
Human antigen R
Matrix metallopeptidase 14
Nucleosome assembly protein 1 like 1
Non-POU domain containing octamer binding
Quantitative real-time PCR
RNA fluorescence in situ hybridization
Reverse transcription PCR
Splicing factor proline and glutamine rich
Small guide RNA
Short hairpin RNA
Transcriptional start site
Gastric cancer is one of the leading causes of cancer-related death, mainly due to high rate of recurrence and distant metastasis in advanced cases . Thus, it is urgent to elucidate the mechanisms underlying the progression of gastric cancer. Human antigen R (HuR) is a member of the embryonic lethal abnormal visual protein (ELAV) family that is involved in nervous system development, cellular proliferation, and migration . As a RNA binding protein (RBP), HuR increases mRNA stability of target genes through binding to poly-U elements or AU-rich elements (AREs) in 3′-untranslated region (3′-UTR) , and plays an important role in tumor development, recurrence, invasion, and metastasis . Studies have shown that HuR is highly expressed in breast cancer, colorectal cancer, gastric cancer, and prostate cancer, and is closely related to clinicopathological features, lymph node metastasis, low survival rate, and poor prognosis of cancer patients [5, 6, 7, 8, 9]. However, the mechanisms regulating HuR expression during gastric cancer progression still remain to be elucidated.
Circular RNAs (circRNAs), a group of transcripts characterized by closed continuous loops stably existing in tissues and cells, are essential regulators of gene expression, while dysregulated circRNAs have been identified in almost all types of cancers . Previous studies show that circRNAs play multiple important roles in cellular physiology via acting as microRNA (miRNA) sponges, RBP-binding molecules, transcriptional regulators, or templates for protein translation. The inspiring miRNA sponging activity has been proven in CDR1as and ciRS-7 [11, 12], and ciRS-7 contains 70 target sites of miRNA-7 to act as a competing endogenous RNA. However, the majority of circRNAs harbor few binding sites for a single miRNA, and high-throughput sequencing analysis reveals that miRNA sponging mechanism cannot be widely applied across the properties of circRNAs . Recent studies have shown the emerging roles of circRNAs in control of gene expression via physical interaction with proteins. For example, circ-Amotl1 derived from angiomotin-like 1 (AMOTL1) promotes tumorigenesis by binding to and retaining c-Myc within the nuclei in breast cancer . Circ-CTNNB1 generated by β-catenin (CTNNB1) activates Wnt signaling pathway by interacting with DEAD-box polypeptide 3 . In addition, circ-EIF3J generated by eukaryotic translation elongation factor 3 J (EIF3J) interacts with U1 snRNP and RNA polymerase II to promote parental gene transcription . In our previous work, we have demonstrated that HuR is highly expressed in gastric cancer tissues and cells, while circRNA derived from Argonaute 2 (circAGO2) binds and promotes the recruitment of HuR protein to the 3′-UTR of target genes, and facilitates the proliferation, invasion, and metastasis of gastric cancer cells, suggesting that circRNA can regulate HuR activity in gastric cancer . Meanwhile, the roles of circRNAs in regulating HuR expression in gastric cancer still remain largely elusive.
In this study, we identify a circRNA consisting of exons 3, 4, and 5 of HuR (circ-HuR) as a novel tumor suppressor in gastric cancer. We discover that circ-HuR is significantly down-regulated in gastric cancer, and effectively inhibits the growth, invasion, and metastasis of gastric cancer cells. Mechanistically, circ-HuR directly interacted with CCHC-type zinc finger nucleic acid binding protein (CNBP), and acted as an inhibitor to restrain the binding of CNBP to HuR promoter, resulting in repression of HuR expression and tumor progression, which indicates the essential roles of circ-HuR and CNBP in gastric cancer progression.
Materials and methods
Patient tissues and cell culture
Tumor and adjacent normal (peritumor) tissues of 81 gastric cancer cases were obtained at surgery in Union Hospital of Tongji Medical College, with demographic and clinicopathological details indicated in Additional file 1: Table S1. The Institutional Review Board of Tongji Medical College approved human tissue study (approval number: 2011-S085). All procedures were carried out in accordance with guidelines set forth by Declaration of Helsinki. Written informed consent was obtained from all patients. Fresh tumor tissues were validated by pathological diagnosis, frozen in liquid nitrogen, and stored at − 80 °C. Human embryonic kidney HEK293T (CRL-11268) cells and gastric cancer cell lines AGS (CRL-1739), MKN-45 (JCRB0254), MKN-74 (JCRB0255) and NCI-N87 (CRL-5822) were obtained from American Type Culture Collection (Rockville, MD) and Japanese Collection of Research Bioresources Cell Bank (Osaka, Japan), authenticated by short tandem repeat (STR) profiling, and applied for study within 6 months following resuscitation of frozen aliquots. Cell lines were cultured in RPMI 1640 medium (Gibco, Carlsbad, CA, USA) supplied with 10% fetal bovine serum (Gibco) at 37 °C in a humidified atmosphere of 5% CO2.
RT-PCR and real-time quantitative RT-PCR (qRT-PCR)
Total RNA was isolated according to the instructions of RNeasy Mini Kit (QIAGEN, Stockach, Germany). For circRNA detection, treatment with RNase R (3 U/mg, Epicenter, Madison, WI, USA) was undertaken at 37 °C for 15 min, and cDNA was synthesized by using reverse transcription kit (Takara, Dalian, China). Genomic DNA (gDNA) was isolated with DNA Mini Kit (QIAGEN). Quantification of mRNA, circular RNA and gDNA was performed by using a SYBR Green PCR Kit (Takara), primers (Additional file 1: Table S2) and Real-Time PCR System (Applied Biosystems, Carlsbad, CA, USA). The levels of circRNA and mRNA were normalized to those of β-actin and determined by 2-△△Ct method.
Western blot assay
Cellular proteins were extracted with RIPA lysis buffer (Thermo Fisher Scientific, Inc., Waltham, MA, USA). Western blot assay was performed as previously described [18, 19], with antibodies specific for HuR (ab200342), cyclin D2 (CCND2, ab207604), CTNNB1 (ab32572), CNBP (ab48027), enolase 1 (ENO1, ab155102), nucleosome assembly protein 1 like 1 (NAP1L1, ab33076), matrix metalloproteinases 14 (MMP-14, ab51074), c-Myc (ab32072), Flag (ab45766), β-actin (ab125402, Abcam lnc., Cambridge, MA, USA), heparin binding growth factor (HDGF, A10435), non-POU domain containing octamer binding (NONO, A5282), splicing factor proline and glutamine rich (SFPQ, A0958, ABclonal, Wuhan, China), glutathione S-transferase (GST, sc-33,614), histone H3 (sc-24,516), or glyceraldehyde 3-phosphate dehydrogenase (GAPDH, sc-47,724, Santa Cruz Biotechnology, Santa Cruz, CA, USA).
Plasmid construction and stable transfection
Linear circ-HuR and hsa_circ_23897 were synthesized by TSINGKE (Wuhan, China) and inserted into pLCDH-ciR (Geenseed Biotech Co., Guangzhou, China). Mutation of back-splicing elements of circ-HuR or hsa_circ_23897 vector was prepared with GeneTailor™ Site-Directed Mutagenesis System (Invitrogen, Carlsbad, CA, USA) and primers (Additional file 1: Table S3). Human CNBP cDNA (540 bp) was subcloned into CV186 (Genechem Co., Ltd., Shanghai, China), while its truncations were obtained by PCR amplification (Additional file 1: Table S3) and inserted into pCMV-3Tag-1A or pGEX-6P-1 (Addgene, Cambridge, MA, USA), respectively. Two independent single guide RNAs (sgRNAs) targeting downstream region of CNBP transcription start site (Additional file 1: Table S4) were inserted into dCas9-BFP-KRAB (Addgene). Oligonucleotides specific for short hairpin RNAs (shRNAs) against circ-HuR (Additional file 1: Table S4) were inserted into GV298 (Genechem Co., Ltd). Lentiviral vectors were co-transfected with packaging plasmids psPAX2 and pMD2G into HEK293T cells. Infectious lentiviruses were harvested at 36 and 60 h after transfection, followed with concentration by ultracentrifugation (2 h at 120,000 g). Stable cell lines were obtained by selection with puromycin.
RNA fluorescence in situ hybridization (RNA-FISH)
Digoxin (DIG)-labeled antisense or sense probe for circ-HuR junction sequence (Additional file 1: Table S4) was synthesized. The probes for GAPDH and U1 were generated by in vitro transcription of PCR products (Additional file 1: Table S2) using DIG Labeling Kit (MyLab Corporation, Beijing, China). Hybridization was undertaken using Fluorescent In Situ Hybridization kit (RiboBio, Guangzhou, China) following the manufacturer’s instructions, while the nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI). The images were analyzed via a Nikon A1Si Laser Scanning Confocal Microscope (Nikon Instruments Inc., Japan).
Dual-luciferase reporter assay
Human HuR promoter (1341 bp) or c-Myc promoter (1363 bp) was amplified from gDNA using primer sets (Additional file 1: Table S3) and subcloned into pGL3-Basic (Promega, Madison, WI, USA). Human MMP-14 promoter reporter was previously described . Mutation of CNBP binding site was established using GeneTailor™ Site-Directed Mutagenesis System (Invitrogen) and primers (Additional file 1: Table S3). Human CNBP activity luciferase reporter was established by inserting oligonucleotides containing four canonical CNBP binding sites (Additional file 1: Table S3) into pGL3-Basic (Promega). Dual-luciferase assay was performed according to the manufacturer’s instructions (Promega).
Biotin-labeled RNA pull-down and mass spectrometry
Biotin-labeled oligonucleotide probes (Additional file 1: Table S4) targeting junction sites of circRNAs were synthesized (Invitrogen). Using Biotin RNA Labeling Mix kit (Roche, Indianapolis, IN, USA) and T7 RNA polymerase, the biotin-labeled RNA probes for circRNAs were in vitro transcribed as previously described [15, 17]. RNA pull-down assay was performed at room temperature, and retrieved proteins were detected by mass spectrometry analysis at Wuhan Institute of Biotechnology (Wuhan, China).
Fluorescence immunocytochemical staining
Cells were plated on coverslip and incubated with antibody specific for CNBP (ab48027, Abcam lnc.) at 4 °C overnight, and incubated with Alexa Fluor 594 goat anti-rabbit IgG and DAPI. The images were photographed under a Nikon A1Si Laser Scanning Confocal Microscope (Nikon Instruments Inc., Japan).
Chromatin immunoprecipitation (ChIP)
ChIP assay was undertaken in accordance with the manual of EZ-ChIP kit (Upstate Biotechnology, Temacula, CA, USA), with antibodies specific for CNBP (ab48027, Abcam lnc.). Primer sets were listed in Additional file 1: Table S2.
Cross-linking RNA immunoprecipitation (RIP) assay
Cells were cross-linked at 254 nm by ultraviolet light, and RIP assay was performed according to the instructions of Magna RIPTM Kit (Millipore, Bedford, MA, USA), with antibodies specific for CNBP (ab48027), ENO1 (ab155102), NAP1L1 (ab33076), Flag (ab45766, Abcam lnc.), HDGF (A10435), NONO (A5282), SFPQ (A0958, ABclonal). Co-precipitated RNAs were detected by RT-PCR or real-time qRT-PCR with specific primers (Additional file 1: Table S2).
In vitro binding assay
Four truncations of CNBP were amplified with primers (Additional file 1: Table S3), and subcloned into pCMV-3Tag-1A or pGEX-6P-1 (Addgene). GST-tagged CNBP protein was produced from E. coli as previously described [18, 21]. The Flag-CNBP, GST-CNBP, and circ-HuR complexes were pulled down using GST or Flag beads (Sigma, St. Louis, MO, USA). Circ-HuR was measured by RT-PCR with divergent primers (Additional file 1: Table S2), whereas protein was detected by SDS-PAGE and western blot.
RNA electrophoretic mobility shift assay (EMSA)
Biotin-labeled circular probe of circ-HuR was prepared as described above. RNA EMSA was conducted according to the instructions of LightShift Chemiluminescent RNA EMSA Kit (Thermo Fisher Scientific, Inc.)
In vitro cell viability, growth, and invasion assays
Xenografts in nude mice
All animal experiments were carried out in accordance with NIH Guidelines for the Care and Use of Laboratory Animals, and approved by the Animal Care Committee of Tongji Medical College (approval number: Y20080290). In vivo tumor growth and experimental metastasis studies were performed with blindly randomized four-week-old male BALB/c nude mice (n = 5 per group). For tumor growth studies, AGS cells (1 × 106) stably transfected with vectors were injected into the upper back of nude mice (n = 5 per group). For metastasis studies, AGS cells (0.4 × 106) stably transfected with vectors were injected from tail vein of nude mice (n = 5 per group). The tumor volume and survival time of each mouse were monitored and recorded, while xenografts were imaged using In-Vivo Xtreme II small animal imaging system (Bruker Corporation, Billerica, MA) and detected by hematoxylin and eosin (H&E) and immunohistochemical staining.
Immunohistochemical staining and quantitative evaluation was performed as previously described [15, 21], with antibodies specific for Ki-67 (sc-23,900, Santa Cruz Biotechnology; 1:100 dilution) or CD31 (ab28364, Abcam Inc.; 1:100 dilution). The degree of positivity was measured according to percentage of positive cancer cells.
All data were shown as mean ± standard error of the mean (SEM) processed by GraphPad Prism 5.0 (La Jolla, USA). Student’s t-test, analysis of variance (ANOVA), and chi-square analysis were used to evaluate the difference. Pearson’s correlation coefficient assay was used to analyze expression correlation. Log-rank test and Cox regression models were used to assess survival difference and hazard ratio. All statistical tests were two-sided and considered statistically significant when P values less than 0.05.
Circ-HuR is down-regulated and decreases HuR expression in gastric cancer
Over-expression of circ-HuR suppresses the growth and aggressiveness of gastric cancer
Circ-HuR interacts with CNBP protein in gastric cancer cells
CNBP promotes HuR expression, growth, and aggressiveness of gastric cancer
Circ-HuR suppresses HuR expression, growth, and invasion of gastric cancer cells via repressing CNBP transactivation
We further investigated the interplay effects between circ-HuR and CNBP in regulating HuR expression and gastric cancer progression. In a cohort of 81 gastric cancer cases, lower circ-HuR expression (P = 3.0 × 10− 3) and higher expression of CNBP (P = 7.0 × 10− 4) or HuR (P = 3.5 × 10− 2) was associated with lower survival probability of patients (Additional file 1: Figure S4 g). In addition, the circ-HuR (R = -0.602, P < 1.0 × 10− 4) or CNBP (R = 0.683, P < 1.0 × 10− 4) levels were negatively and positively correlated with those of HuR in these gastric cancer tissues, respectively (Additional file 1: Figure S4h). Cox regression analysis revealed that distant metastasis [hazard ratio (HR) = 2.809, P = 0.002], tumor-node-metastasis (TNM) stage (HR = 2.519, P = 0.015), circ-HuR levels (HR = 0.616, P = 0.012), and CNBP expression (HR = 2.643, P = 0.003) were prognostic factors for gastric cancer patients (Additional file 1: Table S1).
Circ-HuR suppresses gastric cancer progression by inhibiting CNBP transactivation in vivo
With the development of genomic analysis platform, a large number of abundant and conserved circRNAs have been identified in human tissues and cells [33, 34, 35]. Recent studies suggest that circRNAs are able to function as tumor drivers or tumor suppressors in multiply ways [10, 14, 16, 36, 37, 38, 39]. Besides action mode of binding miRNA, circRNAs participate in parental gene expression at transcriptional level [16, 40]. In this study, we discover that circ-HuR (hsa_circ_0049027), a novel nuclear circRNA, is down-regulated in gastric cancer. Over-expression of circ-HuR suppresses the growth and aggressiveness of gastric cancer in vitro and in vivo. Mechanistically, circ-HuR inhibits the enrichment of CNBP on the promoter of HuR, resulting in reduction of HuR expression (Fig. 6f). These findings highlight a novel tumor suppressive circRNA in regulating tumor growth and aggressiveness, presenting a promising therapeutic target for gastric cancer.
CNBP is a highly conserved zinc finger protein consisting of seven tandem repeats of CCHC zinc finger and an arginine/glycine-rich region, and contributes to embryonic development, organogenesis, and tumorigenesis [26, 31, 41, 42]. As a transcription factor, CNBP binds to the promoters of target genes such as macrophage colony-stimulating factor (CSF1) , MMP-14 , and c-Myc . CNBP expression is elevated in human tumor tissues, and is associated with proliferation, invasion, and migration of tumor cells in vitro and in vivo [31, 44]. In the current study, mining of public gastric cancer datasets reveals that higher expression of CNBP is associated with poor outcome of patients. Gain- and loss-of-function studies show that CNBP facilitates HuR expression, growth, and aggressiveness of cancer cells, suggesting the oncogenic roles of CNBP in gastric cancer progression. Additionally, our results indicate that RGG box domain is essential for the transcriptional activity of CNBP. We identify that circ-HuR is able to suppress the transcriptional activity of CNBP, and functions as an endogenous inhibitor for repressing the binding of CNBP to HuR promoter, resulting in down-regulation of HuR and its target genes involved in cancer progression.
HuR protein is able to regulate the expression of labile mRNAs containing AU-rich elements through increasing their half-life . As a RBP shuttling between the nucleus and cytoplasm, HuR exerts a fundamental role in tumor progression, and its cytoplasmic presence is intimately linked to mRNA stabilizing function [5, 46]. Many HuR target genes are necessary for cell growth and proliferation, such as cyclin A, cyclin B1, cyclin D2 [22, 47], hypoxia-inducible factor 1α (HIF-1α), vascular endothelial growth factor (VEGF), cyclooxygenase-2 (COX-2), and β-catenin [8, 23, 48]. Thus, it has been a promising tumor therapeutic approach via modulating the expression or activity of HuR. For example, treatment with MS-444, an inhibitor interfering the RNA binding and trafficking of HuR, results in loss of viability and induction of apoptosis in malignant glioma cells . Additionally, CMLD-2, a disruptor of interaction between HuR and mRNA targets, exerts antitumor effects in thyroid cancer cells by decreasing cell viability and increasing apoptosis , highlighting HuR as a promising therapeutic target for cancers. In this study, our evidence shows that circ-HuR is able to inhibit HuR expression and suppress the growth and aggressiveness of gastric cancer in vitro and in vivo, suggesting a potential therapeutic approach for cancers.
This study indicates that a novel nuclear circRNA, termed as circ-HuR, is down-expressed in gastric cancer tissues and cells. Circ-HuR acts as an inhibitor of CNBP transactivation by directly interacting with its RGG domain. Interestingly, CNBP promotes the expression of HuR at transcription level via binding to HuR promoter in gastric cancer cells. Meanwhile, circ-HuR restrains the transcription of HuR by inhibiting CNBP transactivation, and suppresses the growth and aggressiveness of gastric cancer in vitro and in vivo. This study extends our knowledge about the regulation of HuR expression and gastric cancer progression by circRNA, and provides a potential target for treatment of gastric cancer.
We appreciate Dr. Jouko Lohi for providing MMP-14 promoter vector.
FY and AH conceived and performed most of the experiments; DL and JW accomplished some of in vitro experiments; YG, YL, HL, YC, and XW accomplished in vivo studies; FY and AH undertook the mining of publicly available datasets; KH critically reviewed the manuscript; QT and LZ wrote the manuscript. All authors read and approved the final manuscript.
This work was granted by the National Natural Science Foundation of China (81472363, 81572423, 81672500, 81773094, 81772967, 81874085, 81874066, 81802925, 81903011, 81903008), and Fundamental Research Funds for the Central Universities (2019kfyRCPY032).
Ethics approval and consent to participate
All animal experiments were approved by the Animal Care Committee of Tongji Medical College. The Institutional Review Board of Tongji Medical College approved the human tissue study. All procedures were carried out in accordance with guidelines set forth by Declaration of Helsinki.
Consent for publication
Written informed consent was obtained from all patients.
The authors declare that they have no competing interests.
- 3.Doller A, el-S A, Huwiler A, Muller R, Radeke HH, Pfeilschifter J, Eberhardt W. Posttranslational modification of the AU-rich element binding protein HuR by protein kinase Cdelta elicits angiotensin II-induced stabilization and nuclear export of cyclooxygenase 2 mRNA. Mol Cell Biol. 2008;28:2608–25.PubMedPubMedCentralCrossRefGoogle Scholar
- 4.Vazquez-Chantada M, Fernandez-Ramos D, Embade N, Martinez-Lopez N, Varela-Rey M, Woodhoo A, Luka Z, Wagner C, Anglim PP, Finnell RH, et al. HuR/methyl-HuR and AUF1 regulate the MAT expressed during liver proliferation, differentiation, and carcinogenesis. Gastroenterology. 2010;138:1943–53.PubMedPubMedCentralCrossRefGoogle Scholar
- 8.Mrena J, Wiksten JP, Thiel A, Kokkola A, Pohjola L, Lundin J, Nordling S, Ristimaki A, Haglund C. Cyclooxygenase-2 is an independent prognostic factor in gastric cancer and its expression is regulated by the messenger RNA stability factor HuR. Clin Cancer Res. 2005;11:7362–8.PubMedCrossRefGoogle Scholar
- 22.Sun Q, Tripathi V, Yoon JH, Singh DK, Hao Q, Min KW, Davila S, Zealy RW, Li XL, Polycarpou-Schwarz M, et al. MIR100 host gene-encoded lncRNAs regulate cell cycle by modulating the interaction between HuR and its target mRNAs. Nucleic Acids Res. 2018;46:10405–16.PubMedPubMedCentralCrossRefGoogle Scholar
- 39.Wei Y, Chen X, Liang C, Ling Y, Yang X, Ye X, Zhang H, Yang P, Cui X, Ren Y, et al. A noncoding regulatory RNAs network driven by circ-CDYL acts specifically in the early stages hepatocellular carcinoma. Hepatology. 2019. https://doi.org/10.1002/hep.30795.
- 44.D'Amico D, Antonucci L, Di Magno L, Coni S, Sdruscia G, Macone A, Miele E, Infante P, Di Marcotullio L, De Smaele E, et al. Non-canonical hedgehog/AMPK-mediated control of polyamine metabolism supports neuronal and medulloblastoma cell growth. Dev Cell. 2015;35:21–35.PubMedPubMedCentralCrossRefGoogle Scholar
- 48.Pan H, Strickland A, Madhu V, Johnson ZI, Chand SN, Brody JR, Fertala A, Zheng Z, Shapiro IM, Risbud MV. RNA binding protein HuR regulates extracellular matrix gene expression and pH homeostasis independent of controlling HIF-1α signaling in nucleus pulposus cells. Matrix Biol. 2019;77:23–40.PubMedCrossRefGoogle Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.