Hepatology, VOL. 0, NO.

0, 2019

HDAC6 Suppresses Let-7i-5p to Elicit

TSP1/CD47-mediated Anti-tumorigenesis
and Phagocytosis of Hepatocellular
Carcinoma
Hee Doo Yang,1-3 Hyung Seok Kim,1-3 Sang Yean Kim,1-3 Min Jeong Na,1-3 Gyeongdeok Yang,1,2 Jung Woo Eun,4 Hee Jung Wang,5
Jae Youn Cheong,4 Won Sang Park,1,2 and Suk Woo Nam1-3

Histone deacetylase 6 (HDAC6) uniquely serves as a tumor suppressor in hepatocellular carcinogenesis, but the un-
derlying mechanisms leading to tumor suppression are not fully understood. To identify comprehensive microRNAs
(miRNAs) regulated by HDAC6 in hepatocellular carcinogenesis, differential miRNA expression analysis of HDAC6-
transfected Hep3B cells was performed. Using integrative analyses of publicly available transcriptome data and miRNA
target prediction, we selected five candidate miRNAs and, through in vitro functional validation, showed that let-7i-5p
specifically suppressed thrombospondin-1 (TSP1) in hepatocellular carcinoma (HCC). Ectopic expression of antisense
let-7i-5p (AS-let-7i-5p) inhibited in vitro tumorigenesis of HCC cells. In addition, treatments of partially purified
TSP1 from culture cell media (ppTSP1) and recombinant TSP1 (rTSP1) exhibited similar effects with AS-let-7i-5p
treatment on the same HCC cells, whereas TSP1 neutralizing antibody treatment significantly attenuated these effects.
Notably, treatments of HDAC6 plasmid, AS-let-7i-5p, ppTSP1, and rTSP1 significantly suppressed in vitro angiogen-
esis and metastatic potential of HCC cells, but the co-treatment of TSP1 antibody specific to cluster of differentiation
47 (CD47) binding domain successfully blocked these effects in the same cells. Furthermore, we demonstrated that re-
covery of HDAC6 elicited let-7i-5p suppression to de-repress TSP1 expression; therefore, it occupied the CD47 recep-
tor to block CD47-SIRPα-mediated anti-phagocytosis of macrophage in HCC. We also observed that HCC-derived
exosomal let-7i-5p suppressed TSP1 of recipient hepatocyte cells. Treatments of HDAC6 plasmid, AS-let-7i-5p, and
rTSP1 suppressed tumor incidence as well as tumor growth rates in a spontaneous mouse HCC model. Conclusion:
Our findings suggest that the HDAC6–let-7i-5p–TSP1 regulatory pathway suppresses neoplastic and antiphagocytic be-
haviors of HCC by interacting with cell surface receptor CD47 in HCC and neighboring cells of tumor microenviron-
ment, providing a therapeutic target for the treatment of liver malignancy and metastasis. (Hepatology 2019;0:1-18).

H
epatocellular carcinoma (HCC) is an aggres- death worldwide.(1) However, the overall survival of
sive form of cancer, the fifth most common patients with HCC has not improved significantly
cancer and the third leading cause of cancer over the past two decades, and the mechanisms

Abbreviations: AS-let-7i-5p, antisense let-7i-5p; CD47, cluster of differentiation 47; CFSE, carboxyfluorescein succinimidyl ester; ChIP, chromatin
immunoprecipitation; EMT, the epithelial–mesenchymal transition; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; GEO, gene expression
omnibus; HCC, hepatocellular carcinoma; HDAC6, histone deacetylase 6; HUVEC, human umbilical vascular endothelial cell; IgG, immunoglobulin
G; let-7, lethal-7; miRNAs, microRNAs; N.C., negative control; NCBI, National Center for Biotechnology Information; oncomiR, oncogenic miRNA;
PARP, poly(adenosine diphosphate ribose) polymerase; ppTSP1, partially purified TSP1 from culture cell media; Ras-Tg, H-ras12V transgenic; rTSP1,
recombinant TSP1; siRNA, small interfering RNA; SIRPα, signal regulatory protein α; TCGA, The Cancer Genome Atlas; THBS1, thrombospondin-1;
TSP1, thrombospondin-1.
Received November 22, 2018; accepted March 30, 2019.
Additional Supporting Information may be found at onlinelibrary.wiley.com/doi/10.1002/hep.30657/suppinfo.
Supported by the National Research Foundation of Korea (2017R1A2B3002989).
© 2019 by the American Association for the Study of Liver Diseases.
View this article online at wileyonlinelibrary.com.
DOI 10.1002/hep.30657
Potential conflict of interest: Nothing to report.

1
YANG ET AL. Hepatology, Month 2019

responsible for the development and progression of also negatively regulates miR-27b through Rel
HCC remain poorly understood.(2) To date, molec- A/p65 to modulate the MET/PI3K/AKT pathway
ular targeted therapy has shown promise for the in diffusing large B-cell lymphoma.(9) Based on
treatment of advanced HCC,(3) but it is unclear how these reports, we hypothesized that loss or suppres-
these genetic changes cause the clinical characteristics sion of HDAC6, as a potent histone modifier, causes
observed in individual HCC patients. re-activation of comprehensive functional RNAs,
Histone deacetylase 6 (HDAC6), originally known especially oncogenic microRNAs (oncomiRs), and
as a microtubule-associated deacetylase, interacts with thereby suppresses tumor suppressors contributing
α-tubulin, and microtubule-associated HDAC6 is an to hepatocellular carcinogenesis. To this end, Venn
essential component of the lysosomal protein deg- Diagram analysis of HDAC6- and HCC-specific
radation mechanism and plays an important role in microRNAs (miRNAs) resulted in 18 miRNAs as
aggresome clearance, which is functionally implicated potentially regulated by HDAC6 in HCC. Through
in the autophagic signaling and degradation of ubiq- analysis of publicly available data sets, let-7i-5p was
uitinated proteins.(4) In several human cancers, the found to be the most significant. Target predic-
expression of HDAC6 is associated with oncogenic tion and various computational analysis suggested
transformation and tumor formation, but we have thrombospondin-1 gene (THBS1) as a direct target
recently demonstrated that HDAC6 functioned as a of let-7i-5p in HCC. The thrombospondin-1 pro-
tumor suppressor by playing a pivotal role in auto- tein (TSP1)–cluster of differentiation 47 (CD47)
phagic cell death processing in hepatocellular carcino- pathway has an important role in several fundamen-
genesis.(5,6) It was also reported that suppression of tal cellular functions, including proliferation, apop-
HDAC6 promotes angiogenesis in HCC, supporting tosis, inflammation, and atherosclerotic response.(10)
HDAC6 as tumor suppressor in hepatocellular car- Partially purified TSP1 from nontransformed
cinogenesis.(7) Thus, HDAC6 is uniquely endowed as hepatocyte culture media (ppTSP1) induced cellular
tumor suppressor among the HDAC family in hepa- death processing and simultaneously inhibited tumor
tocellular carcinogenesis. angiogenesis and metastatic behaviors of HCC cells.
In eukaryotes, transcriptional regulation occurs Treatments of HDAC6 plasmid, antisense inhibitor
through chromatin remodeling, primarily through of let-7i-5p (AS-let-7i-5p), ppTSP1, and recombi-
posttranslational modifications of histones that nant TSP1 (rTSP1) significantly enhanced macro-
package DNA into structural units. HDACs are phage phagocytosis by disrupting the CD47-signal
enzymes that play an important role in various regulatory protein α (SIRPα) interaction “don’t eat
biological processes by repressing or de-repressing me signal” between macrophage and HCC cells,
coding and noncoding gene expressions. A recent but the co-treatment of TSP1 antibody blocking
study has shown that activated transcription factor of C-terminal domain to CD47 receptor (3F352)
3 in association with HDAC6 acts to decrease the significantly attenuated these effects on HCC cells.
transcription of miR-199a-2/miR-214.(8) HDAC6 Here, we suggest that loss or suppression of HDAC6

ARTICLE INFORMATION:
From the 1 Department of Pathology, College of Medicine, the Catholic University of Korea, Seoul, Republic of Korea;
2
Functional RNomics Research Center, the Catholic University of Korea, Seoul, Republic of Korea; 3 Department of Biomedicine
& Health Sciences, Graduate School of Medicine, the Catholic University of Korea, Seoul, Republic of Korea; 4 Department
of Gastroenterology, Ajou University School of Medicine, Suwon, Republic of Korea; 5 Department of Surgery, Ajou University
School of Medicine, Suwon, Republic of Korea.

ADDRESS CORRESPONDENCE AND REPRINT REQUESTS TO:

Suk Woo Nam, Ph.D. Seoul 06591, Republic of Korea
Department of Pathology E-mail: swnam@catholic.ac.kr
College of Medicine, the Catholic University of Korea Tel.: +82-2-2258-7314
222 Banpo-daero, Seocho-gu Fax: +82-2-537-6586

2
Hepatology, Vol. 0, No. 0, 2019 YANG ET AL.

causes increase of oncomiR let-7i-5p expression, and regulator cistrome.(11) Cross-linked DNA was pre-
thereby directly represses tumor suppressor TSP1, cipitated using HDAC6 (anti-HDAC6), acetylated-
highlighting the potential therapeutic values of the histone H3 (Ac-H3), and acetylated-histone H4
signaling HDAC6–let-7i-5p–TSP1 pathway in the (Ac-H4) antibodies (Millipore, Burlington, MA).
treatment of liver cancer. The reaction was conducted in triplicate, and means
were normalized versus the normal immunoglobulin
G (IgG) control.
Materials and Methods
LUCIFERASE REPORTER
TISSUE SAMPLES PLASMIDS AND ASSAYS
A total of 88 matched pairs of HCC tissues with The 3′-UTR of THBS1 mRNA was PCR-
their corresponding noncancerous liver tissues were amplified from complementary DNAs of the SNU-387
obtained from National Biobank of Korea. Written and SNU-423 cell line and cloned into the Xho I/
informed consent was obtained from each subject Not I sites of a psiCHECK-2 vector (Promega,
according to the Declaration of Helsinki, and the Madison, WI) according to the manufacturer’s instruc-
study was approved by the institutional review of board tions. Luciferase activity was assayed using the Dual-
(IRB) of the Songeui Campus, College of Medicine, Luciferase Reporter Assay System (Promega) according
the Catholic University of Korea (IRB approval num- to the manufacturer’s instructions. The firefly lucifer-
ber: MC12EISI0106). ase activity was normalized with co-expressed Renilla
luciferase activity.
PUBLICLY AVAILABLE GENOMIC
DATA ANALYSIS IN VITRO MICROTUBULE
To recapitulate the expression of let-7-5p family FORMATION ASSAY
miRNAs and THBS1 mRNA in HCC, data were Matrigel Basement Membrane Matrix (BD
obtained from The Cancer Genome Atlas liver Biosciences, San Jose, CA) was coated in 24-well plates
hepatocellular carcinoma project (TCGA_LIHC), at 37°C for 1 hour. Then, Ras-transformed mouse
International Cancer Genome Consortium liver cancer fibroblast cell (Ras-NIH3T3) and human umbilical
RIKEN Japan (ICGC_LIRI-JP), and gene expression vascular endothelial cell (HUVEC) were seeded alone
omnibus (GEO) databases of the National Center for or co-cultured with an equivalent number of trans-
Biotechnology Information (NCBI) (GSE100017, fected SNU-423 cells. After seeding, indicated cells
GSE10694, GSE31384, GSE36915, GSE39678, were incubated overnight in the presence or absence
GSE40744, GSE76903, GSE77276, GSE77314, and of 3F352 antibody (TSP1 antibody specific to
GSE89377 [Catholic_LIHC]). Level 3 mRNA expres- CD47 binding domain; US Biological, Salem, MA).
sion data of TCGA_LIHC RNA-seq V2 were log2- The number of nodes of the tubular structures was
transformed (log2 [RSEM+1]) and used to assess the quantified.
gene expression levels.
IN VITRO PHAGOCYTOSIS ASSAY
CHROMATIN
HCC cells were digested into single cell suspen-
IMMUNOPRECIPITATION ASSAY
sions, and then labeled with carboxyfluorescein suc-
Chromatin immunoprecipitation (ChIP) assay cinimidyl ester (CFSE) (Abcam, Cambridge, United
was conducted according to the manufacturer’s Kingdom) following the manufacturer’s protocol.
instructions (Pierce Agarose ChIP kit; Thermo CFSE-labeled HCC cells were co-cultured with
Fisher Scientific, Waltham, MA). DNA was ampli- macrophage cells for 2 hours. Next, macrophages
fied by real-time quantitative PCR using primers were repeatedly washed and imaged with an IX71
against binding motif of HDAC6 on the let-7i pro- photomicrograph (Olympus, Tokyo, Japan). Macro­
moter region that were predicted using chromatin phages were then harvested and stained with Alexa

3
YANG ET AL. Hepatology, Month 2019

fluor-conjugated F4/80 (Abcam). After labeling, flow Statistical differences between groups were evaluated
cytometry was performed. The phagocytic index was by unpaired two-tailed Student t test using GraphPad
calculated as the percentage of F4/80+CFSE+ cells 7.0 software (GraphPad Software Inc., La Jolla, CA).
(Q  2) among F4/80+ cells (Q  1 + Q  2): phagocytic P values less than 0.05 were considered statistically
index (%) = [Q  2/(Q  1 + Q  2)] × 100%. significant.

MOUSE LIVER CANCER MODEL

The H-ras12V transgenic (Ras-Tg) mice were
Results
kindly provided by Dr. Dae-Yeoul Yu (Laboratory RECOVERY OF HDAC6
of Human Genomics, Korea Research Institute of DEREPRESSES LET-7I-5P
Bioscience and Biotechnology, Daejeon, Korea).(12)
IN HEPATOCELLULAR
Male mice spontaneously developed HCC beginning
at approximately 10-15 weeks of age. Mice livers were
CARCINOGENESIS
then harvested at 25 weeks of age, and processed for We previously demonstrated that HDAC6 func-
the experiments. All animal experiments were under- tions as a tumor suppressor by way of activation of
taken in accordance with the National Institutes of caspase-independent autophagic cell death through the
Health’s Guide for the Care and Use of Laboratory JNK/Beclin-1 pathway in HCC.(5) As the HDAC6 is
Animal, with approval of the Animal Experiment one of the potent epigenetic modifiers, more oncogenic
Ethics Committee of the Catholic University of signaling pathways might be governed by HDAC6 in
Korea College of Medicine. hepatocellular carcinogenesis. Thus, to identify com-
prehensive miRNAs regulated by HDAC6, HDAC6-
transfected Hep3B cells were subjected to miRNA
IN VIVO STUDIES
microarrays and compared with that of empty
Ras-Tg mice were intravenously injected with vector (Mock)-transfected Hep3B (Supporting
50 ug of pcDNA3.1_HDAC6 by using Turbofect Fig. S1A). A total of 129 miRNAs were significantly
in vivo Transfection Reagent (Invitrogen, Carlsbad, down-regulated in HDAC6-transfected Hep3B cells
CA), according to the manufacturer’s protocol.(13) In (GSE100017; <−1.5-fold change, P < 0.05). Then,
addition, Ras-Tg mice were intravenously injected these miRNAs were subjected to Venn Diagram
with Invivofectamine 3.0 (Invitrogen) containing analysis, and compared with that of aberrantly over-
0.25 mg/kg of AS-let-7i-5p according to the man- expressed miRNAs in HCC that we have previously
ufacturer’s instructions.(14) The 0.5 ug of TSP1 published (GSE39678; >1.5-fold change, P < 0.05).(15)
recombinant protein (R&D Systems, Minneapolis, From this, 18 miRNAs were identified as miRNAs
MN) was injected through the lateral tail vein. The potentially regulated by HDAC6 in HCC (Supporting
ultrasonography images were taken at 17, 19, 21, Fig. S1B). Among the most suppressed top 5 miRNAs,
and 23 weeks of age with an ultrasound machine the miR-335-3p, miR-17-5p, and miR-338-3p are not
(Philips, Amsterdam, Netherlands) by the same consistently overexpressed in the three different stud-
medical imaging expert. ies of HCC patients available from the NCBI GEO
Other assays used in this study are described in the database (GSE36915, GSE31384, and GSE76903).
Supporting Information. Kaplan-Meier survival analysis of the TCGA_LIHC
data set showed that the 5-year overall survival rate of
HCC patients with high expressions of both let-7i-5p
STATISTICAL ANALYSIS
and miR-182-5p were significantly lower than that of
Survival curves were plotted using the Kaplan- patients with low levels of let-7i-5p and miR-182-5p
Meier product limit method, and significant differ- (Supporting Fig. S2A). Then, quantitative RT-PCR
ences between survival curves were determined using analysis of these two miRNAs in HCC cell lines
the log-rank test. All experiments were performed (SNU-182, SNU-449, and SNU-475) demonstrated
at least three times, and all samples were analyzed that only the let-7i-5p was consistently suppressed by
in triplicate. Results are presented as mean ± SEM. ectopic expression of HDAC6 in all tested three HCC

4
Hepatology, Vol. 0, No. 0, 2019 YANG ET AL.

cell lines (Supporting Fig. S2B). Furthermore, among cells suppressed HDAC6, and simultaneously elic-
the eight let-7-5p family miRNAs, only the let-7i-5p ited let-7i-5p expression (Fig. 1D). Furthermore,
exhibited overexpression in HCC of seven different we observed that c-Jun knockdown and JNK inhi-
HCC studies, whereas the rest of the seven let-7-5p bition by SP600125, a JNK-specific inhibitor, treat-
family miRNAs showed no significant changes in the ment successfully restored HDAC6 expression, and
same studies available from the NCBI GEO data- thereby suppressed let-7i-5p in HCC cells, whereas
base (Supporting Table S1). Next, to validate aberrant HDAC6 knockdown rescued let-7i-5p expression
expression of let-7i-5p in HCC, we performed quan- in the same cells (Fig. 1E,F). These results suggest
titative RT-PCR analyses for both 88 matched pairs that HDAC6 directly suppresses let-7i-5p expres-
(HCC with corresponding noncancerous surrounding sion, but loss or suppression of HDAC6 elicits his-
liver tissue) of human HCCs and 11 HCC cell lines. tone acetylation to activate let-7i-5p transcription in
Fifty-four of 88 HCC patients (61.4%) exhibited HCC cells (Supporting Fig. S3B).
significant overexpression of let-7i-5p in HCC tis-
sues (Fig. 1A, left panel). Consistently, 9 of 11 tested LET-7I-5P ONCOGENIC
HCC cell lines exhibited significant overexpression POTENTIAL AND ITS
of let-7i-5p compared with MIHA, an immortalized DIRECT TARGET THBS1
non-transformed hepatocyte cell line (Fig. 1A, right IN HEPATOCELLULAR
panel).
CARCINOGENESIS
Next, to investigate whether let-7i-5p expression
is directly regulated by HDAC6 in HCC, we To verify the oncogenic functions of let-7i-5p in
searched the binding motif of HDAC6 on the hepatocellular carcinogenesis, we performed in vitro
promoter region of let-7i-5p by using Chromatin tumorigenesis experiments. Ectopic expression of
Regulator Cistrome (http://cistr​ ome.org/cr/index. HDAC6 significantly suppressed both tumor cell
php). From this, a total of 18 motifs were identified growth and proliferation in HCC cells, whereas
as putative HDAC6 binding motif, and five motifs co-transfection of let-7i-5p mimic rescued the
were located within the 1,000 base pair upstream of tumor-suppressing effects of HDAC6 (Fig. 2A,B).
the let-7i-5p promoter region (Fig. 1B, left panel). Similarly, ectopic expression of AS-let-7i-5p sup-
We then designed three set of primers covering pressed tumor cell growth and proliferation in the
these five motifs, and performed ChIP assay with same cells (Supporting Fig. S4A,B). Flow cytome-
HDAC6 antibody (anti-HDAC6). Treatment of try of annexin V-stained cells showed that ectopic
anti-HDAC6 showed significant enriched bind- expression of HDAC6 induced cellular apoptosis
ing of HDAC6 on these three let-7i-5p promoter of HCC cells, whereas co-transfection of let-7i-5p
regions compared with the control, IgG antibody, mimic rescued the pro-apoptotic effects of HDAC6.
treatment in HCC cells (Fig. 1B, middle panel). This result was confirmed by induced cleaved
We then performed ChIP assays with Ac-H3 caspase-3 and PARP (poly[adenosine diphosphate
and Ac-H4 in the presence or absence of ectopic ribose] polymerase) cleavage in the same cells
HDAC6 expression in HCC cells. Ectopic expres- (Fig. 2C,D). In keeping with these results, AS-let-7i-5p
sion of HDAC6 (pcDNA3.1_HDAC6) elicited transfection also significantly induced apoptosis of
histone deacetylations on these three let-7i-5p pro- HCC cells (Supporting Fig. S4C,D). Epithelial–
moter regions in HCC cells (Fig. 1B, right panel; mesenchymal transition (EMT) has been suggested
Supporting Fig. S3A). In addition, ectopic expres- as a key process in cancer progression. To explain
sion of HDAC6 significantly suppressed let-7i-5p the role of let-7i-5p in the malignant behavior of
expression, whereas Tubastatin A, a HDAC6 spe- HCC cells, we performed a scratch wound healing
cific inhibitor, treatment rescued let-7i-5p expres- assay. Ectopic expression of HDAC6 significantly
sion in HCC cells (Fig. 1C). Previously, we have reduced wound healing efficacy and suppressed
demonstrated that JNK/c-Jun-dependent miR- chemoattractant-stimulated migratory and invasive
221-3p directly suppresses HDAC6 in HCC.(6) As responses of HCC cells, whereas co-transfection of
expected, ectopic expression of miR-221-3p mimic let-7i-5p mimic significantly rescued these effects
in nontransformed hepatocyte MIHA and L-02 (Fig. 2E,F). Consistently, AS-let-7i-5p transfection

5
YANG ET AL. Hepatology, Month 2019

FIG. 1. HDAC6 directly regulates let-7i-5p expression in HCC. (A) Expression of let-7i-5p by quantitative RT-PCR. Left panel:
88 matched pairs of human HCC tissues with corresponding noncancerous adjacent liver tissues. Right panel: Eleven HCC cell lines
with immortalized nontransformed hepatocyte, MIHA. (B) ChIP–quantitative RT-PCR assay to assess HDAC6 binding on the
let-7i-5p promoter. Left panel: Schematic of let-7i-5p promoter region covering five putative HDAC6 binding motifs. Middle panel:
ChIP–quantitative RT-PCR assay to assess HDAC6 association with let-7i-5p promoter. The amount of DNA precipitated by either
anti-HDAC6 or control IgG was shown as fold enrichment. Right panel: ChIP–quantitative RT-PCR assay to assess enrichment
of histone deacetylation by HDAC6 on the let-7i-5p promoter region. (C) Cells were transfected with pcDNA3.1_HDAC6 in the
presence or absence of tubastatin A (HDAC6-specific inhibitor, Tub A). Left panel: Indicated protein expression analysis by western
blot. Right panel: Let-7i-5p expression analysis by quantitative RT-PCR. (D) Cells were transfected with miR-221-3p mimic. Left
panel: Indicated protein expression analysis by western blot. Right panel: Let-7i-5p expression analysis by quantitative RT-PCR.
(E) Cells were transfected with c-Jun small interfering RNA (siRNA) and then co-transfected with HDAC6 siRNA. Left panel:
Indicated protein expression analysis by western blot. Right panel: Let-7i-5p expression analysis by quantitative RT-PCR. (F) Cells
were treated with SP600125 and then co-transfected with si-HDAC6. Left panel: Indicated protein expression analysis by western
blot. Right panel: Let-7i-5p expression analysis by quantitative RT-PCR. All data are shown as the mean ± SEM. *P < 0.05, **P < 0.01,
***P < 0.001 by unpaired Student t test. Abbreviations: GAPDH, glyceraldehyde 3-phosphate dehydrogenase; N.C., negative control;
si-HDAC6, HDAC6 siRNA; si-c-Jun, c-Jun siRNA.

6
Hepatology, Vol. 0, No. 0, 2019 YANG ET AL.

7
YANG ET AL. Hepatology, Month 2019

FIG. 2. Functional characterization of oncomiR let-7i-5p in HCC. (A) Cell growth assays: MTT assay (upper) and cell viability
assay (lower). (B) Cell proliferation assays: BrdU assay (upper) and clonogenic assay (lower). (C) Flow cytometry of annexin-V-FITC/
PI-positive HCC cells. Bar graph indicates the percentage of cells displaying annexin-V single-positive and annexin-V, PI double-
positive (n = 3). (D) Western blot analysis of caspase-3, cleaved caspase-3, PARP, and cleaved PARP. GAPDH was used as a loading
control. (E) Representative cell image (left) and migration percentage of cells transfected with pcDNA3.1_HDAC6 or co-transfected
with let-7i-5p mimics in scratch wound healing assay. (F) Representative cell image (left) and number (right) of migrated and invaded
cells transfected with pcDNA3.1_HDAC6 or co-transfected with let-7i-5p mimics in modified Boyden chamber motility assay and
Transwell invasion assay. All data are shown as the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001 by unpaired Student t test.
Abbreviations: BrdU, bromodeoxyuridine; MTT, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide.

exhibited similar results (Supporting Fig. S4E,F). Furthermore, let-7i-5p exhibited significant negative
PLC/PRF/5 and Huh7 HCC cell lines are rela- correlation with both HDAC6 (r = −0.32, P < 0.01)
tively high in HDAC6 expression, whereas let-7i-5p and THBS1 (r = −0.47, P < 0.001) in the same subset,
expression is low in HCC cell lines (Fig. 1A right whereas THBS1 exhibited significant positive correla-
panel; Supporting Fig. S5A). Treatment of let-7i-5p tion with HDAC6 (r = 0.42, P < 0.01) (Supporting
mimic significantly enhanced tumor cell growth Fig. S7D). Another data set of 33 matched pairs HCCs
and proliferation, and elevated chemoattractant- (GSE77276; Tsinghua_LIHC) showed similar results
stimulated migratory and invasive responses of PLC/ with the 88 matched pairs HCC subset (Supporting
PRF/5 and Huh7 cells (Supporting Fig. S5B-D). Fig. S7E).
These results demonstrate that de-repression of let- TSP1 is a matricellular glycoprotein that can
7i-5p by loss or suppression of HDAC6 contributes be secreted by many cell types.(17) As expected, we
to hepatocellular carcinogenesis. were able to detect a relatively high amount of sec-
Next, to identify let-7i-5p-targeting molecules, retary TSP1 in the conditioned medium of MIHA
we used the target prediction program TargetScan compared with the SNU-387 and SNU-423 HCC
(http://www.targe​tscan.org/vert_72/).(16) From this, cell lines (Fig. 3A). To determine whether endoge-
1,207 genes were predicted as let-7i-5p-targeting nous THBS1 expression was selectively regulated by
molecules (data not shown). We then selected mol- let-7i-5p by direct interaction with the 3′-UTR of
ecules involved in cellular apoptosis process by using the transcripts, we cloned the 3′-UTR of THBS1
the DAVID Functional Annotation Tool 6.7 (https​:// into a reporter vector to generate the wild-type
david.ncifc​rf.gov/) as let-7i-5p inhibition-induced psiCHECK-2-THBS1_3′-UTR vector (psiCHECK-
apoptosis (Supporting Fig. S6). Of the 54 gene ele- 2-THBS1_3′-UTR_WT), and cloned the 3′-UTR
ments that involve in cell death processes (GO: of THBS1 harboring the mutated let-7i-5p
0006915, GO: 0012501), we recapitulated expres- response sequence into a reporter vector to generate
sions of most 10 down-regulated genes in our HCC mutant psiCHECK-2-THBS1_3′-UTR vectors
data set (Catholic_LIHC; GSE89377) and compared (psiCHECK-2-THBS1_3′-UTR_MTs). Then, dual
them with three additional large HCC cohort studies luciferase reporter assays of psiCHECK-2-THBS1-
(TCGA_LIHC, ICGC_LIRI-JP, and GSE77314). 3′-UTR vectors with AS-let-7i-5p were performed
To select the most stringent let-7i-5p targeting gene and compared with that of negative control treat-
in HCC, we counted the significance score in fold ment. As expected, AS-let-7i-5p treatment signifi-
changes (<−1.5 fold, P < 0.001) of four studies against cantly augmented relative luciferase activity in the
10 selected gene elements (Supporting Table S2). psiCHECK-2-THBS1_3′-UTR_WT vector group,
This analysis suggested that THBS1 is the most sig- but not in the psiCHECK-2_THBS1_3′-UTR_MTs
nificantly down-regulated gene in HCC. As expected, groups (Fig. 3B). Consistently, Dicer knockdown
THBS1 was significantly down-regulated in HCC in caused induction of TSP1, whereas treatment of
both the 50 matched pairs and the 371 HCC data sets let-7i-5p mimics attenuated the Dicer knockdown
of the TCGA_LIHC (Supporting Fig. S7A). In addi- effect on the same cells (Fig. 3C, upper panel). In
tion, quantitative RT-PCR analysis of the 88 matched addition, we were able to observe that AS-let-7i-5p
pairs of HCC subset showed that THBS1 (62 of 88, transfection augmented secretary TSP1 in the con-
70.4%) and HDAC6 (59 of 88, 67%) were significantly ditioned media of HCC cells (Fig. 3C, lower panel).
down-regulated in HCC (Supporting Fig. S7B,C). TSP1 is widely studied molecule and functions as

8
Hepatology, Vol. 0, No. 0, 2019 YANG ET AL.

9
YANG ET AL. Hepatology, Month 2019

FIG. 3. Let-7i-5p suppresses TSP1 in contributing neoplastic and malignant behaviors of HCC. (A) Detection of secretary TSP1 in
the conditioned media of MIHA, SNU-387, and SNU-423 by western blot analysis. Coomassie blue staining was used as a loading
control. (B) The target site of let-7i-5p in 3′-UTR of THBS1 is shown as a schematic representation (upper). Wild-type (psiCHECK-
2_THBS1_WT) and mutant THBS1 3′-UTR (psiCHECK-2_THBS1_MT1 or psiCHECK-2_THBS1_MT2) constructs were inserted
into psiCHECK-2 plasmid and transfected with AS-let-7i-5p (lower). (C) Western blot analysis performed in Dicer siRNA and
AS-let-7i-5p-transfected cells and their conditioned media. Upper: GAPDH was used as the loading control. Lower: Coomassie blue
staining was used as a loading control. (D) Cells transfected with AS-let-7i-5p or treated with MIHA_ppTSP1 in scratch wound
healing assay. (E) Representative cell image (left) and number (right) of migrated and invaded cells transfected with AS-let-7i-5p or
treated with MIHA_ppTSP1 in modified Boyden chamber motility assay and Transwell invasion assay, respectively. (F) Western blot
analysis of E-cadherin, N-cadherin, Fibronectin, Snail, and Slug in cells transfected with AS-let-7i-5p or treated with MIHA_ppTSP1.
GAPDH was used as a loading control. All data are shown as the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001 by unpaired Student
t test. Abbreviation: si-Dicer, Dicer siRNA.

a negative modulator of various tumor cell behav- showed that treatments of rTSP1, 293T_ppTSP1, and
iors such as cell–cell adhesion, migration and MIHA_ppTSP1 significantly induced cellular apop-
invasion, proliferation, and apoptosis. Notably, we tosis, and cleaved caspase-3 and PARP in the same
observed that AS-let-7i-5p transfection inhibited cells (Fig. 4C,D). We also observed that treatments of
chemoattractant-stimulated wound-healing efficacy rTSP1, 293T_ppTSP1, and MIHA_ppTSP1 signifi-
of HCC cells. In addition, treatment of partially cantly suppressed the chemoattractant-stimulated inva-
purified TSP1 from the MIHA cell-conditioned sive response of HCC cells, whereas the co-treatment
media (MIHA_ppTSP1) exhibited similar effect of A6.1, a TSP1-specific neutralizing antibody, atten-
on the same cells (Fig. 3D). Furthermore, modified uated anti-invasion effects of these cells (Fig. 4E).
Boyden chamber and Transwell assays showed that Again, we observed that each treatment of HDAC6
both AS-let-7i-5p and MIHA_ppTSP1 treatments expression, let-7i-5p inhibition, and cellular and
significantly inhibited chemoattractant-stimulated recombinant TSP1 significantly suppressed the inva-
migratory and invasive responses of the HCC cells sion responses of HCC cells, but was rescued by the
(Fig. 3E). To gain further insight into the regula- co-treatment of A6.1 in the same cells (Fig. 4F).
tory effect of let-7i-5p and MIHA_ppTSP1 on
EMT processing, western blot analysis was per- HDAC6–LET-7I-5P–TSP1
formed for the EMT regulatory proteins. Notably,
REGULATORY AXIS GOVERNS
N-cadherin, Fibronectin, Snail, and Slug were mark-
edly decreased, but E-cadherin, an epithelial marker,
TUMOR ANGIOGENESIS,
was simultaneously increased in both AS-let-7i-5p METASTASIS, AND
and MIHA_ppTSP1-treated HCC cells (Fig. 3F). PHAGOCYTOSIS THROUGH CD47
RECEPTOR
LET-7I-5P TARGETING Accumulating evidence has demonstrated that
TSP1 FUNCTIONS AS A secretary TSP1 mediates angiogenesis through the
TUMOR SUPPRESSOR cell surface receptor CD47.(18-20) Therefore, to con-
IN HEPATOCELLULAR firm the anti-angiogenic activity of TSP1, we per-
formed in vitro microtubule formation assay with
CARCINOGENESIS
rTSP1 and MIHA_ppTSP1 in both Ras-NIH3T3
Our data demonstrated that let-7i-5p directly reg- and HUVEC cells. Treatments of MIHA_ppTSP1
ulated TSP1. Thus, to clarify the tumor-suppressive and rTSP1 significantly suppressed in vitro microtu-
role of TSP1 in hepatocellular carcinogenesis, we per- bule formation of both Ras-NIH3T3 and HUVEC
formed in vitro tumorigenesis assays. In cell growth cells, whereas co-treatment of 3F352 (TSP1 anti-
and proliferation assays, treatments of rTSP1, partially body blocking of C-terminal domain to CD47)
purified TSP1 from the 293T cell-conditioned media significantly attenuated anti-microtubule formation
(293T_ppTSP1), and MIHA_ppTSP1 significantly activity of TSP1 (Supporting Fig. S8A,B). Next,
inhibited growth and proliferation rates of HCC cells to verify that the HDAC6–let-7i-5p–TSP1 axis
(Fig. 4A,B). Flow cytometry of annexin V–stained cells regulates tumor angiogenesis through the CD47

10
Hepatology, Vol. 0, No. 0, 2019 YANG ET AL.

11
YANG ET AL. Hepatology, Month 2019

FIG. 4. TSP1 functions as a tumor suppressor in hepatocellular carcinogenesis. (A) Cells were treated with rTSP1, 293T_ppTSP1
and MIHA_ppTSP1, respectively, and then MTT assay was performed. (B) Cells were treated in parallel with (A), and then BrdU
assay was performed. (C) Flow cytometry of annexin-V-FITC/PI positive HCC cells after treatment in parallel with (A). Bar graph
indicates the percentage of cells displaying annexin-V single-positive and annexin-V, PI double-positive (n = 3). (D) Western blot
analysis of caspase-3, cleaved caspase-3, PARP, and cleaved PARP in cells parallel with (A). GAPDH was used as a loading control.
(E) Representative cell image (upper) and number (lower) of cells treated with A6.1 after rTSP1, 293T_ppTSP1, and MIHA_ppTSP1
treatment in Transwell invasion assay. (F) Representative cell image (left) and number (right) of cells treated with A6.1 after being
transfected with pcDNA3.1_HDAC6 and AS-let-7i-5p or treated with MIHA_ppTSP1 and rTSP1 in Transwell invasion assay. All data
are shown as the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001 by unpaired Student t test. Abbreviations: BrdU, bromodeoxyuridine;
MTT, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide.

receptor, we performed in vitro microtubule forma- mechanisms of calreticulin-mediated TSP1 secre-

tion assay by using the HCC cell co-culture system. tion.(21) This led us to hypothesize that the meta-
Treatments of pcDNA3.1_HDAC6, AS-let-7i-5p, static potential of HCC is also mediated by autocrine
MIHA_ppTSP1, and rTSP1 to SNU-423 HCC loop of TSP1 and CD47 interaction. As expected, we
cells significantly suppressed microtubule forma- observed that each treatment of pcDNA3.1_HDAC6,
tion activities of both Ras-NIH3T3 and HUVEC AS-let-7i-5p, MIHA_ppTSP1, and rTSP1 signifi-
cells, whereas co-treatment of 3F352 successfully cantly suppressed migratory and invasive responses of
rescued these effects (Fig. 5A,B). Note that SNU- HCC cells. All of these effects were significantly res-
423 cells were not able to form an in vitro capil- cued by the co-treatment of 3F352 in the same cells
lary-like tube in Matrigel microtubule formation implicating the autocrine and/or paracrine mechanism
assay (Supporting Fig. S8C). These results indicate of TSP1-CD47 in HCC cells (Fig. 5C,D; Supporting
that the HDAC6–let-7i-5p–TSP1 regulatory axis Fig. S9A-D).
exerts anti-angiogenic activities by interacting with We then noted that TSP1 could compete with
the CD47 cell surface receptor. SIRPα in the CD47 interaction of the macrophage
In a previous study, TSP1-CD47 interaction con- phagocytosis regulatory network.(22) Therefore, we
trolled T cell motility and migration by autocrine hypothesized that elevated TSP1 occupies the CD47

FIG. 5. HDAC6–let-7i-5p–TSP1 regulatory axis mediates anti-angiogenesis and antimetastatic potentials through CD47 in HCC
cells. (A,B) Matrigel microtubule formation assay. HCC cells were co-cultured with Ras-NIH3T3 cells (A) or HUVECs (B) and
treated pcDNA3.1_HDAC6, AS-let-7i-5p, MIHA_ppTSP1 and rTSP1, and then rescued by 3F352 treatment. Migration (C) and
invasion (D) assay. HCC cells were transfected or treated in parallel with (A) and (B). Bar graph is presented as numbers of migrated
and invaded cells. All data are shown as the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001 by unpaired Student t test.

12
Hepatology, Vol. 0, No. 0, 2019 YANG ET AL.

receptor to disrupt CD47–SIRPα interaction, thereby nontransformed hepatocyte cells may occur to facili-
ceasing the “don’t eat me” signaling between HCC and tate malignant transformation and growth of hepato-
macrophage cells. To this end, mouse peritoneal macro- cytes in hepatocellular carcinogenesis.
phage cells were harvested and co-cultured with HCC Next, to validate our observation in vivo, Ras-Tg
cells, then compared with that of HCC cells treated mice that spontaneously develop HCC from approx-
with pcDNA3.1_HDAC6, AS-let-7i-5p, MIHA_ imately 15 weeks of age were prepared. To investi-
ppTSP1 and rTSP1, respectively. We additionally gate the cancer-preventive effect of re-activating the
performed flow cytometry–based assay, labeling HCC HDAC6–let-7i-5p–TSP1 regulatory axis in vivo,
cells with CFSE to determine the phagocytic activity mice were intravenously injected with liver-specific
of macrophage. Macrophagic cells stained with F4/80 delivery reagents Turbofect (pcDNA3.1_HDAC6),
antibody were then measured and presented as phago- Invivofectamine (AS-let-7i-5p), and rTSP1 from
cytic index. As expected, all treatments of pcDNA3.1_ 13 weeks of age weekly (Fig. 7A, upper panel).
HDAC6, AS-let-7i-5p, MIHA_ppTSP1, and rTSP1 Mouse liver HCC was detected using ultrasonog-
to the HCC cells significantly increased the phagocy- raphy from 17 weeks of age until the time of sacri-
tosis activity of macrophage, whereas co-treatment of fice at 25 weeks of age. HCC was detectable at 17
3F352 significantly attenuated the macrophage phago- weeks of age in the negative control group, in which
cytosis of HCC cells (Fig. 6A; Supporting Fig. S10). 4 of 4 mice developed large and multiple HCCs,
This implies that HDAC6-let-7i-p-TSP1 regulatory whereas HCC was detectable at 21 weeks of age in
axis re-activates “eat me” signaling by switching CD47- the pcDNA3.1_HDAC6 (except 1 mouse), AS-let-
SIRPα to CD47-TSP1 interaction between HCC and 7i-5p and rTSP1 groups, in which approximately
macrophage cells. 1-4 of 4 mice developed relatively small HCCs
Recently, accumulating evidence has shown that (Fig. 7A, lower panel; Supporting Fig. S11A-E).
tumor-derived exosomal elements, especially vari- Western blot analysis confirmed elevated expression
ous miRNAs, facilitate malignant behaviors of HCC of HDAC6 and TSP1 in the mouse liver tissues of
cells by communicating with microenvironmental the pcDNA3.1_HDAC6 and AS-let-7i-5p groups
components.(23) Following this idea, we investigated (Fig. 7B). Consistently, let-7i-5p down-regulation
the potential role of tumor cell–derived nanovesi- was observed in the same tissues of the pcDNA3.1_
cles in the regulation of TSP1 in hepatocytes. First, HDAC6 group (Fig. 7C). To show that the decrease
we isolated and purified exosomes from conditioned in tumor load is possibly due to phagocytosis and
culture media of HCC cells, then further veri- anti-angiogenesis of the HDAC6–let-7i-5p–TSP1
fied HCC-derived exosomes by detecting CD81, effects in vivo, we performed western blot analyses
CD9 and TSG101, characteristic exosome markers with F4/80 (mouse macrophage marker), CD31, and
(Fig. 6B,C). A comparison of the Ct (threshold CD34 (endothelial markers) antibodies. Elevated
cycle) value of let-7i-5p quantitative RT-PCR anal- expression of F4/80, and simultaneously reduced
ysis between exosomes and donor cell fractions expression of CD31 and CD34, were observed in
showed that let-7i-5p existed primarily in exosomes mouse liver tissues from all of the pcDNA3.1_
as opposed to donor cells (Fig. 6D). Next, we investi- HDAC6, AS-let-7i-5p, and rTSP1 groups com-
gated whether nonmalignant transformed hepatocyte pared with the corresponding control (Fig. 7D).
cells could take HCC-derived exosomes, and PKH67 Furthermore, immunohistochemical staining anal-
dye (green fluorescence)–labeled exosomes were incu- yses of the formalin-fixed, paraffin-embedded tis-
bated with MIHA. Green fluorescence in the recipi- sues from same specimens exhibited similar results
ent cells (MIHA) were successfully observed (Fig. 6E). (Fig. 7E).
We then measured the relative expression of let-7i-5p Finally, to further validate the regulatory func-
in MIHA cells incubated with HCC-derived exo- tion of the HDAC6–let-7i-5p–TSP1 axis in HCC
somes and compared that with the nontreated MIHA. metastasis in vivo, invasion and motility assays were
Treatment of HCC-derived exosomes significantly performed with the Ras-NIH3T3 cell. Each treat-
enhanced let-7i-5p expression, thereby suppressing ment of pcDNA3.1_HDAC6, AS-let-7i-5p, and
TSP1 in MIHA cells (Fig. 6F). These results imply pcDNA3.1_THBS1 significantly suppressed chemo-
that intercellular crosstalk between HCC cells and attractant-stimulated migratory and invasive responses

13
YANG ET AL. Hepatology, Month 2019

FIG. 6. Effect of CD47–TSP1 interaction in macrophage phagocytosis and exosome communication of HCC cells. (A) SNU-423 HCC
cells were labeled with CFSE and cultured with peritoneal-derived macrophages from C57BL/6 mice. The cells were incubated with
control IgG or 3F352 antibody. Left: Representative images and flow cytometry results. Right: The extent of macrophage phagocytosis
in each experiment was quantified by phagocytic index. (B) Exosomes released by HCC cells were detected by electron microscopy
(scale bar, 100 nm). (C) Western blot analysis for HSP70, HSP90, CD81, CD9, and TSG101 from donor cell lysates and isolated
exosomes. (D) The cycle threshold (Ct) value of let-7i-5p in donor cell and exosomes from SNU-387 and SNU-423 HCC cells were
analyzed by quantitative RT-PCR. (E) Microscopy images show exosome internalization. MIHA cells labeled with 4′,6-diamidino-
2-phenylindole (blue) in culture were incubated with SNU-387 or SNU-423-derived exosomes that were labeled with PKH67 (green).
(F) let-7i-5p expression analysis by quantitative RT-PCR (left). MIHA cells were incubated with HCC-derived exosomes (5 ug/mL
or 10 ug/mL). TSP1 expression analysis by western blot (right). GAPDH was used as a loading control. All data are shown as
mean ± SEM. **P < 0.01, ***P < 0.001 by unpaired Student t test. Abbreviation: DAPI, 4′,6-diamidino-2-phenylindole.

14
Hepatology, Vol. 0, No. 0, 2019 YANG ET AL.

FIG. 7. The HDAC6–let-7i-5p–TSP1 axis regulates HCC tumorigenesis in vivo. (A) In vivo transfection of pcDNA3.1_HDAC6
and AS-let-7i-5p, and the treatment of rTSP1. Timeline of injection with H-ras-transgenic mouse model (upper). Representative
ultrasonography images of Ras-Tg mice injected with pcDNA3.1_HDAC6, AS-let-7i-5p, and rTSP-1 (lower) at 21 and 23 weeks of
age, respectively, and liver image at 25 weeks of age. (B) Western blot analysis of mouse Hdac6, ac-α-tubulin, and Tsp1 expression
in liver tissues from Ras-Tg mouse model mentioned previously. GAPDH was used for loading control. (C) Quantitative RT-PCR
analysis for let-7i-5p expression. The U6 was used for loading control. (D) Western blot analysis of F4/80, CD31 and CD34 expression
in liver tissues from Ras-Tg mouse model mentioned above. GAPDH was used for loading control. (E) Immunohistochemical staining
analysis for F4/80, CD31, and CD34 in liver tissues from Ras-Tg mouse model mentioned previously. Representative images (upper)
and quantification of positivity (lower). All data are shown as the mean ± SEM. *P < 0.05, **P < 0.01 by unpaired Student t test.

of Ras-NIH3T3 cells (Fig. 8A). In vivo experimental

metastasis assay also showed that all of the same treat-
Discussion
ments significantly attenuated the metastatic potential The results of our study provide insights into the
of Ras-NIH3T3 cells observed by lung nodule quan- pivotal role of the tumor suppressor HDAC6 by
tification (Fig. 8B-D). governing oncogenic miRNA let-7i-5p to maintain

15
YANG ET AL. Hepatology, Month 2019

FIG. 8. The HDAC6–let-7i-5p–TSP1 axis regulates HCC metastasis in vivo. (A) Migration and invasion assay. Representative cell
images (left) and the number of cells (right). (B) Images of mouse lung metastasis obtained from in vivo experimental metastasis
of Ras-NIH3T3 transfected with pcDNA3.1_HDAC6, AS-let-7i-5p, and pcDNA3.1_THBS1, respectively. (C) Lung weight/body
weight ratio of each group is shown (n = 5). (D) The number of lung nodules is shown by bar graph (n = 5). All data are shown as
the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001 by unpaired Student t test. (E) Proposed schematic of HDAC6–let-7i-5p–TSP1
regulatory mechanisms in HCC and microenvironment.

tumor suppressor TSP1 and exert the antineoplas- The lethal-7 (let-7) gene was first found in the
tic effect on HCC and pro-phagocytic behavior on nematode as a key developmental regulator and
macrophage. became one of the first of two known miRNAs (the
Our previous investigations have suggested that other one is lin-4), before being identified as the first
HDAC6 functions as a tumor suppressor by medi- known human miRNA.(24,25) In general, let-7 miRNA
ating caspase-independent autophagic cell death is known to act as a tumor suppressor by targeting var-
through the JNK-activated Beclin 1-dependent path- ious oncogenic molecules including RAS, HMGA2,
way in human liver cancer.(5) Later, we demonstrated cell cycle, and apoptosis regulators in cancers.(25-27)
that loss or suppression of HDAC6 could be partly However, we observed that among the let-7 families,
due to aberrant up-regulation of oncomiR-221.(6) only the let-7i-5p was significantly overexpressed in the
Recent studies have shown that HDAC6 negatively large cohorts of HCC patients (Supporting Table S1),
regulated certain miRNAs to inhibit the pathologic and that let-7i-5p rescued the tumor-suppressing
condition. Thus, we hypothesized that loss or sup- effects of HDAC6 in HCC cells, suggesting pivotal
pression of HDAC6 could re-activate oncogenic roles as oncomiR in hepatocellular carcinogenesis
miRNAs, thereby suppressing tumor suppressors that (Fig. 2; Supporting Figs. S4 and S5). We then iden-
contribute to malignant transformation and growth tified secretary TSP1, a potent endogenous negative
of hepatocytes during hepatocellular carcinogenesis. angiogenesis regulator, as a direct target of let-7i-5p,
Of the candidate miRNAs that were specifically reg- and demonstrated the role of the HDAC6–let-7i-5p–
ulated by HDAC6, we selected let-7i-5p, as it could TSP1 regulatory axis in governing the mitogenic and
be re-depressed by HDAC6 recovery in HCC cells, metastatic potential of HCC cells (Figs. 3 and 4).
and was overexpressed in the large cohort of HCC Antitumor effects of the secreted TSP1 have been
patients (Supporting Figs. S1 and S2). Likewise, we reported to be mediated by its receptors, CD36(28) and
were able to demonstrate that HDAC6 directly regu- CD47.(20) CD47 is a ubiquitously expressed trans-
lates and suppresses let-7i-5p in HCC (Fig. 1). membrane protein that serves as a ligand for SIRPα

16
Hepatology, Vol. 0, No. 0, 2019 YANG ET AL.

and is a signaling receptor of TSP1. The TSP1-CD47 signaling axis was further validated with the spon-
pathway has an important role in several fundamen- taneous mouse HCC model (Fig. 7A). These results
tal cellular functions, including proliferation, apop- suggest the potential of targeting the HDAC6–
tosis, inflammation, and atherosclerotic response.(29) let-7i-5p–TSP1 axis as a strategy for immunotherapy
Ligation of CD47 by TSP1 has been shown to inhibit of human HCC.
nitric oxide/cGMP signaling in vascular cells, leading to The collective data presented here demonstrate
suppression of angiogenic responses.(18) Our investiga- that tumor-suppressive HDAC6–let-7i-5p–TSP1
tion demonstrated that the HDAC6–let-7i-5p–TSP1 signaling plays a pivotal role in the development and
regulatory axis mediated anti-angiogenesis and anti- progression of HCC. Normal activity and expression
metastasis of HCC cells by interacting with the CD47 of HDAC6 appear to be important in maintaining
cell surface receptor (Fig. 5). the balance of mitogenic and angiogenic signals in
Tumor cells evade immune surveillance through hepatocyte. Loss or suppression of HDAC6 induces
direct or indirect interactions with various types oncomiR let-7i-5p to suppress translation of the
of immune cell, with much recent attention being THBS1 gene, thereby interfering in the governance
focused on modifying immune cell responses as the of TSP1 in the growth, proliferation, angiogenesis,
basis for the development of new cancer treatments. and metastasis of HCC. HDAC6 inactivation also
However, recent work has demonstrated that mac- suppresses host macrophage immune surveillance
rophages can be induced to phagocytose tumor cells by inducing the CD47–SIRPα-mediated “don’t eat
through the blockade of the interaction between me” signal. HCC cells may educate the nontrans-
SIRPα and CD47,(22) and this therapeutic strategy formed hepatocyte surrounding microenvironment to
is currently the subject of multiple clinical trials in be HCC-associated hepatocyte, by communicating
cancer.(30) Moreover, the CD47 blockade has been with exosomal let-7i-5p (Fig. 8E). Altogether, these
shown to promote the stimulation of tumor-specific findings define a central role of the HDAC6–let-
cytotoxic T cells by macrophages or dendritic cells. 7i-5p–TSP1 axis in hepatocellular carcinogenesis and
However, the mechanisms underlying the initiation suggest its potential therapeutic value for the treat-
and consolidation of immune responses to tumor cells ment of HCC.
by CD47 or SIRPα blockade, as well as the possible
adverse effects of such blockade in vivo, remain to be REFERENCES
fully understood. In this regard, our results demon-
1) Torre LA, Bray F, Siegel RL, Ferlay J, Lortet-Tieulent J, Jemal A.
strate that recovery of HDAC6 suppressed let-7i-5p Global cancer statistics, 2012. CA Cancer J Clin 2015;65:87-108.
to elicit TSP1 expression; therefore, it occupied the 2) Bruix J, Boix L, Sala M, Llovet JM. Focus on hepatocellular car-
CD47 receptor to block the CD47-SIRPα-mediated cinoma. Cancer Cell 2004;5:215-219.
3) Shen YC, Hsu C, Cheng AL. Molecular targeted therapy for
anti-phagocytosis of the macrophage (Fig. 6A; advanced hepatocellular carcinoma: current status and future
Supporting Fig. S10). In addition, emerging evidence perspectives. J Gastroenterol 2010;45:794-807.
suggests that extracellular vesicles (EVs) have cru- 4) Lee JY, Koga H, Kawaguchi Y, Tang W, Wong E, Gao YS,
et al. HDAC6 controls autophagosome maturation essential
cial roles in cancer development, including premeta- for ubiquitin-selective quality-control autophagy. EMBO J
static niche formation and metastasis.(31) These EVs 2010;29:969-980.
carry molecules such as oncoproteins and oncopep- 5) Jung KH, Noh JH, Kim JK, Eun JW, Bae HJ, Chang YG, et al.
Histone deacetylase 6 functions as a tumor suppressor by acti-
tides, RNA species (e.g., miRNAs, mRNAs, and long vating c-Jun NH2-terminal kinase-mediated beclin 1-dependent
non-coding RNAs), lipids, and DNA fragments from autophagic cell death in liver cancer. Hepatology 2012;56:
644-657.
donor to recipient cells, initiating profound pheno-
6) Bae HJ, Jung KH, Eun JW, Shen Q , Kim HS, Park SJ, et al.
typic changes in the tumor microenvironment. We MicroRNA-221 governs tumor suppressor HDAC6 to potentiate
showed that HCC cells may suppress TSP1-CD47- malignant progression of liver cancer. J Hepatol 2015;63:408-419.
7) Lv Z, Weng X, Du C, Zhang C, Xiao H, Cai X, et al.
dependent antitumor behaviors of neighboring or Downregulation of HDAC6 promotes angiogenesis in hepa-
distant nontransformed hepatocyte by way of HCC- tocellular carcinoma cells and predicts poor prognosis in liver
derived exosomal let-7i-5p, possibly by facilitating transplantation patients. Mol Carcinog 2016;55:1024-1033.
8) Li C, Zhou Y, Loberg A, Tahara SM, Malik P, Kalra VK.
aberrant growth and proliferation, motility, invasion, Activated transcription factor 3 in association with histone
and angiogenesis (Fig. 6E,F). Therapeutic efficacy deacetylase 6 negatively regulates microRNA 199a2 transcription
of the tumor-suppressive HDAC6–let-7i-5p–TSP1

17
YANG ET AL. Hepatology, Month 2019

by chromatin remodeling and reduces endothelin-1 expression. 22) Chao MP, Alizadeh AA, Tang C, Myklebust JH, Varghese B,
Mol Cell Biol 2016;36:2838-2854. Gill S, et al. Anti-CD47 antibody synergizes with rituximab to
9) Jia YJ, Liu ZB, Wang WG, Sun CB, Wei P, Yang YL, et al. promote phagocytosis and eradicate non-Hodgkin lymphoma.
HDAC6 regulates microRNA-27b that suppresses proliferation, Cell 2010;142:699-713.
promotes apoptosis and target MET in diffuse large B-cell lym- 23) Fang T, Lv H, Lv G, Li T, Wang C, Han Q , et al. Tumor-
phoma. Leukemia 2018;32:703-711. derived exosomal miR-1247-3p induces cancer-associated
10) Murata Y, Saito Y, Kotani T, Matozaki T. CD47-signal regula- fibroblast activation to foster lung metastasis of liver cancer. Nat
tory protein alpha signaling system and its application to cancer Commun 2018;9:191.
immunotherapy. Cancer Sci 2018;109:2349-2357. 24) Reinhart BJ, Slack FJ, Basson M, Pasquinelli AE, Bettinger
11) Wang Q , Huang J, Sun H, Liu J, Wang J, Wang Q , et al. CR JC, Rougvie AE, et al. The 21-nucleotide let-7 RNA regu-
Cistrome: a ChIP-Seq database for chromatin regulators and lates developmental timing in Caenorhabditis elegans. Nature
histone modification linkages in human and mouse. Nucleic 2000;403:901-906.
Acids Res 2014;42:D450-D458. 25) Johnson CD, Esquela-Kerscher A, Stefani G, Byrom M,
12) Wang AG, Moon HB, Lee MR, Hwang CY, Kwon KS, Yu SL, Kelnar K, Ovcharenko D, et al. The let-7 microRNA represses
et al. Gender-dependent hepatic alterations in H-ras12V trans- cell proliferation pathways in human cells. Cancer Res 2007;67:
genic mice. J Hepatol 2005;43:836-844. 7713-7722.
13) Zhang L, Yang Z, Trottier J, Barbier O, Wang L. Long non- 26) Johnson SM, Grosshans H, Shingara J, Byrom M, Jarvis R,
coding RNA MEG3 induces cholestatic liver injury by interac- Cheng A, et al. RAS is regulated by the let-7 microRNA
tion with PTBP1 to facilitate shp mRNA decay. Hepatology family. Cell 2005;120:635-647.
2017;65:604-615. 27) Kumar MS, Erkeland SJ, Pester RE, Chen CY, Ebert MS,
14) Eguchi A, De Mollerat Du, Jeu X, Johnson CD, Nektaria A, Sharp PA, et al. Suppression of non-small cell lung tumor devel-
Feldstein AE. Liver Bid suppression for treatment of fibrosis opment by the let-7 microRNA family. Proc Natl Acad Sci U S A
associated with non-alcoholic steatohepatitis. J Hepatol 2016; 2008;105:3903-3908.
64:699-707. 28) Jimenez B, Volpert OV, Reiher F, Chang L, Munoz A, Karin
15) Kim JK, Noh JH, Jung KH, Eun JW, Bae HJ, Kim MG, et al. M, et al. c-Jun N-terminal kinase activation is required for the
Sirtuin7 oncogenic potential in human hepatocellular carcinoma inhibition of neovascularization by thrombospondin-1. Oncogene
and its regulation by the tumor suppressors MiR-125a-5p and 2001;20:3443-3448.
MiR-125b. Hepatology 2013;57:1055-1067. 29) Sick E, Jeanne A, Schneider C, Dedieu S, Takeda K, Martiny
16) Agarwal V, Bell GW, Nam JW, Bartel DP. Predicting effective mi- L. CD47 update: a multifaceted actor in the tumour microen-
croRNA target sites in mammalian mRNAs. Elife 2015;4:e05005. vironment of potential therapeutic interest. Br J Pharmacol
17) Li Y, Turpin CP, Wang S. Role of thrombospondin 1 in liver 2012;167:1415-1430.
diseases. Hepatol Res 2017;47:186-193. 30) Gordon SR, Maute RL, Dulken BW, Hutter G, George BM,
18) Isenberg JS, Ridnour LA, Dimitry J, Frazier WA, Wink DA, McCracken MN, et al. PD-1 expression by tumour-associ-
Roberts DD. CD47 is necessary for inhibition of nitric oxide- ated macrophages inhibits phagocytosis and tumour immunity.
stimulated vascular cell responses by thrombospondin-1. J Biol Nature 2017;545:495-499.
Chem 2006;281:26069-26080. 31) Xu R, Rai A, Chen M, Suwakulsiri W, Greening DW, Simpson
19) Zaslavsky A, Baek KH, Lynch RC, Short S, Grillo J, Folkman RJ. Extracellular vesicles in cancer—implications for future
J, et al. Platelet-derived thrombospondin-1 is a critical nega- improvements in cancer care. Nat Rev Clin Oncol 2018;15:
tive regulator and potential biomarker of angiogenesis. Blood 617-638.
2010;115:4605-4613.
Author names in bold designate shared co-first authorship.
20) Gao Q , Chen K, Gao L, Zheng Y, Yang YG. Thrombospondin-1
signaling through CD47 inhibits cell cycle progression and
induces senescence in endothelial cells. Cell Death Dis 2016;

21)
7:e2368.
Li SS, Forslow A, Sundqvist KG. Autocrine regulation of T cell Supporting Information
motility by calreticulin-thrombospondin-1 interaction. J Immunol
2005;174:654-661. Additional Supporting Information may be found at
onlinelibrary.wiley.com/doi/10.1002/hep.30657/suppinfo.

18