ARTICLE
Received 12 Aug 2016
| Accepted 29 Dec 2016 | Published 17 Feb 2017
MVP-mediated exosomal sorting of miR-193a
promotes colon cancer progression
Yun Teng
1,
*, Yi Ren
2,
*, Xin Hu
3,4
, Jingyao Mu
1
, Abhilash Samykutty
1
, Xiaoying Zhuang
1
, Zhongbin Deng
1
,
Anil Kumar
1
, Lifeng Zhang
1
, Michael L. Merchant
5
, Jun Yan
1
, Donald M. Miller
1
& Huang-Ge Zhang
1,6
Exosomes are emerging mediators of intercellular communication; whether the release of
exosomes has an effect on the exosome donor cells in addition to the recipient cells has not
been investigated to any extent. Here, we examine different exosomal miRNA expression
profiles in primary mouse colon tumour, liver metastasis of colon cancer and naive
colon tissues. In more advanced disease, higher levels of tumour suppressor miRNAs are
encapsulated in the exosomes. miR-193a interacts with major vault protein (MVP). Knockout
of MVP leads to miR-193a accumulation in the exosomal donor cells instead of exosomes,
inhibiting tumour progression. Furthermore, miR-193a causes cell cycle G1 arrest and cell
proliferation repression through targeting of Caprin1, which upregulates Ccnd2 and c-Myc.
Human colon cancer patients with more advanced disease show higher levels of circulating
exosomal miR-193a. In summary, our data demonstrate that MVP-mediated selective sorting
of tumour suppressor miRNA into exosomes promotes tumour progression.
DOI: 10.1038/ncomms14448
OPEN
1
James Graham Brown Cancer Center, University of Louisville, Louisville, Kentucky 40202, USA.
2
Department of Breast and Thyroid Surgery, Huai’an First
People’s Hospital, Huai’an, Jiangsu 223001, China.
3
Program in Biostatistics, Bioinformatics and Systems Biology, The University of Texas Graduate School of
Biomedical Sciences at Houston, Houston, Texas 77030, USA.
4
Department of Genomic Medicine, The University of Texas MD Anderson Cancer Center,
Houston, Texas 77030, USA.
5
Kidney Disease Program and Clinical Proteomics Center, University of Louisville, Louisville, Kentucky, USA.
6
Robley Rex VA
Medical Center, Louisville, Kentucky 40206, USA. * These authors contributed equally to this work. Correspondence and requests for materials shouldbe
addressed to H.-G.Z. (email: H0zhan17@louisville.edu) or to Y.T. (email: yun.teng@louisville.edu).
NATURE COMMUNICATIONS | 8:14448 | DOI: 10.1038/nc omms14448 | www.nature.com/naturecommunications 1
C
ancer cells secrete extracellular vesicles (EVs) that promote
cancer progression
1–4
, and these EVs contain markers for
potential cancer diagnosis and prognosis
5,6
. A complete
working model of EV-mediated biological effects has not been
demonstrated in a fully physiological in vivo context and is
urgently needed for understanding the in vivo fate of cancer EVs.
Most studies published thus far analyse the function of EV
populations isolated from the supernatants of cultured cells
7–9
.
One of the challenges of this experimental methodology is
whether EV secretion in vitro by tumour cells is capable of
achieving the necessary result in vivo. A number of approaches
have been undertaken to address this challenge by either
interfering with in vivo EV biogenesis in cancer cells with
siRNA knockdown
10,11
or with chemical inhibitors
12
to inhibit
EV release. However, EVs, including exosomes, are released from
many different types of cells, including cancer cells as well as
non-cancer cells. Without a known cancer exosome-specific
marker, we cannot define the biological effect of cancer
cell-derived exosomes from non-cancer cell-derived exosomes
in an in vivo model.
Recently, modifications of EVs that allow tracking of their
target cells in vivo have been reported
13–15
. In the current study,
we generated a stable colon cancer cell line utilizing a vector
16
expressing a luciferase protein fused to a biotin acceptor peptide
with a transmembrane (TM) domain of platelet-derived growth
factor. The fused proteins with a TM domain localize in the
plasma membrane as well as in secreted EVs (ref. 16). Therefore,
exosomes released from this stable colon cancer cell line are
biotinylated and can be isolated from mixed EVs with
streptavidin-coated beads, allowing for their miRNA profile to
be further analysed. The tumour tissue-derived exosome miRNA
profile is detected in the exosomes from peripheral blood of
tumour-bearing mice but not naive mice. In this study, as a proof
of concept, we hypothesize that tumour exosomes selectively sort
tumour suppressor miRNA into exosomes, whereas oncogenic
miRNA is kept in the tumour cell regardless of the level of
miRNA expressed in the cell.
Results
Isolation and characterization of tumour-specific exosomes.
The presence of EV RNAs in tissues and in fluids, including
blood, together with the changes in EV RNA expression in
various pathological conditions, has implicated EV RNAs as
informative biomarkers of progression and early diagnosis for
cancer
17–22
. However, EVs are released from many different types
of cells, including tumour and non-tumour cells. The challenge is
to distinguish between EVs released from tumour cells and those
released from non-tumour cells. To achieve this overall goal,
we generated stably transduced colon cancer CT26 cells with
lentivirus vectors
16
that enable the isolation of EVs from CT26
tumour cells (Fig. 1a). The expression of this construct, as
described in Fig. 1a, provides a biotin binding moiety for isolation
of exosomes, as well as Gaussia luciferase (Gluc) and green
fluorescent protein tags for monitoring CT-26 tumours in organs
and biological fluids (Supplementary Fig. 1a). The results
generated from luciferase assays with coelenterazine (CTZ)
indicate that the EVs released from CT26 cells are stably
transfected with the lentivirus vector expressing Gluc, biotin
acceptor peptide and TM domain and that these cells exhibited a
higher luciferase activity compared with the cells expressing Gluc
without the TM domain (Supplementary Fig. 1b). Moreover,
more Gluc activity in the medium and less Gluc activity in the
whole cell lysates was detected in the cells transfected with
membrane-bound Gluc than in cells transfected with GlucB.
Lentiviral vector stable transfection had no significant effects on
the numbers of exosomes that were shed or the levels of
miR-193a, miR-18a or miR-126a in the exosomes released from
stably transfected CT26 cells compared with wild-type CT26 cells
(Supplementary Fig. 1c,d). To characterize the tumour EVs
released from primary and metastatic cancer in the liver of colon
cancer mice, CT26 cells stably expressing both EV-Gluc and
biotin ligase BirA were administered to BALB/c mice by colonic
submucosa or intrasplenic injection as described
23
. Two weeks
after injection, a tumour was evident in hematoxylin and eosin
(H&E)-stained sectioned colon and liver (Supplementary Fig. 2a)
and in confocal imaging of sectioned liver (Supplementary
Fig. 2b). Gluc and BirA expression were visualized with green
fluorescent protein-tagged Gluc and red fluorescent mCherry-
tagged BirA in metastatic lesions in the liver, but not in the
adjacent normal liver tissue (Supplementary Fig. 2b). To estimate
Gluc luciferase activity in vivo, CTZ or phosphate-buffered saline
(PBS; control) was injected via tail vein injection into BALB/c
mice-bearing CT26 tumours with Gluc expression. Five minutes
after the injection, CTZ-injected mice revealed a significant
amount of Gluc signal in the liver of GlucB-expressing mice, but
not in PBS-injected tumour-bearing mice (Supplementary
Fig. 2c). These results confirm an in vivo biological activity and
stability of Gluc imaging reporter in liver metastatic CT26
tumour cells.
To further determine if tumour-specific exosomes can be
isolated from metastatic liver, exosomes were isolated from liver
metastasis of colon cancer, followed by purification with sucrose
gradient centrifugation. Isolated exosomes were dot blotted on
nitrocellulose membranes followed by probing with anti-CD63
antibody and streptavidin-conjugated Alexa Fluor 488. Dot blot
analysis showed that exosomes from both naı
¨
ve mouse liver and
liver with metastatic colon cancer expressed exosomal marker
CD63 with the same fluorescent intensity (Supplementary
Fig. 3a). However, biotinylated EV-GlucB could be detected in
liver with metastatic colon cancer-derived exosomes, but not in
naı
¨
ve mouse liver-derived exosomes (Supplementary Fig. 3a).
Western blot analysis demonstrated the presence of exosomal
protein marker CD63 and the absence of the endoplasmic
reticulum protein Calnexin in exosomes (Supplementary Fig. 3b).
Exosomal morphology and size distribution were further
evaluated using electron microscopy (Supplementary Fig. 3c)
and Zetasizer Nano ZS analysis (Supplementary Fig. 3d),
respectively. Metastatic liver exosomes had a diameter of
48.6
±
4.6 nm (means
±
standard error of the mean (s.e.m.)) and
were smaller than naı
¨
ve mouse liver-derived exosomes, which had
a diameter of 83.6
±
7.8 nm (means
±
s.e.m.) (Supplementary
Fig. 3d, left). The number of exosomes released from liver
metastasis of colon cancer was much higher than that from the
liver of naı
¨
ve mice (Supplementary Fig. 3e).
Exosomal miRNA as an indicator for tumour progression.
Identification of a unique miRNA profile encapsulated in
tumour cell exosomes but not in non-tumour cell exosomes is
essential for clinical applications. Tumour exosomal miRNA
has recently been extensively studied for use as a
diagnostic marker
20,24–33
and as an indicator for disease
progression
3,10,34
. To further determine whether our approach
could identify an exosomal miRNA profile that mirrors disease
progression, we performed an miRNA microarray and did
comparative analysis of miRNome in exosomes isolated from
normal colon, primary colon cancer tissue and colon tumour
metastasis to the liver (Fig. 1b, Supplementary Fig. 2a). Among
these differentially expressed miRNAs, 6.9% of the miRNA was
uniquely detected in normal colon tissue-derived exosomes,
11.5% of the miRNA was detected in primary colon tumour
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14448
2 NATURE COMMUNICATIONS | 8:14448 | DOI: 10.1038/ncomms14448 | www.nature.com/naturecommunications
Lentivirus
a d
b
ef
c
GlucB reporter
Small RNA
Release
Naive colon
Primary colon cancer
Liver metastasis of colon cancer
Liver metastasis of
colon cancer at day 3
Log2
(fold change)
6.9%
16.7%12.2%
42.6%
2.8%
11.5% 7.3%
Extracellular
vesicles (EVs)
CT26/
GlucB
Biotin
ligase
Biotin
10
mmu-miR-1892
mmu-miR-504-5p
mmu-miR-764-3p
mmu-miR-654-3p
mmu-miR-875-3p
mmu-miR-711
mmu-miR-380-3p
mmu-miR-882
mmu-miR-450b-3p
mmu-miR-466f-5p
mmu-miR-154-3p
mmu-miR-200b-5p
mmu-miR-883b-3p
mmu-miR-1199-5p
mmu-miR-683
mmu-miR-672-5p
mmu-miR-147-3p
mmu-miR-196a-5p
mmu-miR-96-5p
mmu-miR-181d-5p
mmu-miR-9-5p
mmu-miR-146b-5p
mmu-miR-301b-3p
mmu-miR-182-5p
mmu-miR-33-5p
mmu-miR-423-5p
mmu-miR-301a-3p
mmu-miR-151-5p
mmu-miR-17-5p
mmu-miR-205-5p
mmu-miR-375-3p
mmu-miR-15b-5p
mmu-miR-1195
mmu-miR-196b-5p
mmu-miR-669b-5p
mmu-miR-193a-3p
mmu-miR-339-5p
mmu-miR-678
mmu-miR-10a-5p
mmu-miR-331-3p
mmu-miR-677-5p
mmu-miR-203-3p
mmu-miR-148a-3p
mmu-miR-22-3p
mmu-miR-126a-3p
mmu-miR-222-3p
mmu-miR-192-5p
mmu-miR-30b-5p
mmu-miR-714
0.0
7.0–0.6–7.0
20.188507
40.377014
Liver metastasis Exo
Liver metast 14d exo3
Liver metast 14d exo1
Liver metast 14d exo2
Liver metast 7d exo3
Liver metast 7d exo2
Liver metast 7d exo1
Liver metast 3d exo3
Liver metast 3d exo2
Liver metast 3d exo1
Primary colon Ca exo3
Primary colon Ca exo2
Primary colon Ca exo1
Naive colon exo3
Naive colon exo1
Naive colon exo2
33.322495
16.661247
0.0
Naive colon Exo
mmu-let-7i-5p
miRNAs of exosomes
miRNAs in exosomes (qPCR)miRNAs in exosomes (microarray)
600
45
30
15
10
–10
–20
0
Naïve colon
Primary colon cancer
Liver metas 3 days
Liver metas 7 days
Liver metas 14 days
Naive colon
Primary colon cance
r
Liver metas 3 days
Liver metas 14 days
#
#
#
*
*
*
*
*
#
#
#
*
*
*
#
#
#
#
#
*
*
*
*
*
*
*
*
#
*
#
*
#
*
#
*
#
*
#
*
#
*
#
*
#
*
#
*
#
*
*
#
*
#
*
*
*
*
*
*
*
#
*
#
*
#
*
#
*
#
*
#
*
#
*
#
*
#
*
#
*
#
*
#
*
#
*
#
*
*
*
#
#
#
#
*
*
*
#
*
#
*
*
#
*
#
#
*
*
*
*
*
*
*
400
200
15
–15
–300
–600
miR-10a
miR-22
miR-30b
miR-126a
miR-148a
miR-193a
miR-200b
miR-196a
miR-196b
miR-10a
miR-22
miR-30b
miR-126a
miR-148a
miR-193a
miR-200b
miR-151
miR-196a
miR-196b
miR-151
0
Fold change
g
Peripheral plasma exosomes Faeces exosomes
#
#
*
#
*
#
*
#
*
*
#
*
*
*
*
*
*
*
*
*
#
*
#
*
#
*
#
*
#
*
#
*
#
*
#
*
#
*
#
*
#
*
*
#
*
#
#
*
*
*
*
*
*
*
*
*
*
*
#
*
#
*
#
*
*
*
*
*
*
–15
–10
–5
0
5
10
15
miR-10a
miR-22
miR-30b
miR-126a
miR-148a
miR-193a
miR-200b
miR-151
miR-196a
miR-196b
miR-10a
miR-22
miR-30 b
miR-126a
miR-148a
miR-193 a
miR-200b
miR-151
miR-196a
miR-196b
Fold change
–10
0
10
20
30
Fold change
Fold change
–5
–4 –2 0 2 4 6
Primary colon cancer
Log2
(fold change)
5
0
Biotin acceptor peptide
Streptavidin
Gaussia luciferase (Gluc)
Primary colon cancer Exo
Figure 1 | Identification of exosome miRNA profile that represents primary colon cancer and metastatic colon cancer in the liver. (a) Schematic
diagram for isolation of extracellular vesicles (EVs) from colon cancer CT26 cell line with multimodal imaging report. CT26 cells stably transduced with a
lentiviral vector expressing membrane-bound Gaussia luciferase (GlucB) and biotin ligase (BirA). (b) Venn diagram summarizing unique and shared
exosomal miRNAs detected in the tissues of naı¨ve colon, primary colon cancer and metastatic mouse colon cancer in the liver using miRNA microarray data
(n ¼ 5 mice per group). (c) Microarray data visualization by scatter plot comparing exosomal miRNAs detected in primary colon cancer (x axis) and
metastatic colon cancer in the liver at day 3 (y axis) after a CT26 cell intrasplenic injection. (d) Heat map depicting changes in miRNAs with a statistically
significant (Po0.05) change in the exosomal miRNAs from normal mouse colon, primary colon cancer tissue and metastatic colon cancer in the liver at
days 3, 7 and 14 after injection of CT26 cells (n ¼ 3 mice per group). All tumour-derived exosomes were isolated with streptavidin magnetic beads.
Microarray analysis results (e) and qPCR verification (f ) of selected exosomal miRNAs from the source as described in d.(g) qPCR analysis of the
plasma- (left panel) or faeces- (right panel) derived exosomes from the source are depicted in d.*Po0.05 versus naı¨ve colon;
#
Po0.05 versus primary
colon cancer (two-tailed t-test). Data are representative of three independent experiments (error bars, s.e.m.).
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14448 ARTICLE
NATURE COMMUNICATIONS | 8:14448 | DOI: 10.1038/nc omms14448 | www.nature.com/naturecommunications 3
exosomes and 7.3% of the miRNA was detected in metastatic
tumour exosomes (Fig. 1b). Next, exosomal miRNA profiles from
liver metastasis of colon cancer were compared with the exosomal
miRNA profiles from primary colon tumour tissue. In the scatter
plot (Fig. 1c), each point represents the expression value of a
given miRNA. Compared with exosomal RNAs from primary
colon tumour tissue, exosomal RNA profiles from liver metastasis
of colon cancer displayed a different distribution (Fig. 1c). The
green and red dots represent the higher level of exosomal
miRNAs detected in the primary colon cancer and liver
metastasis, respectively, and the grey dots represent similar
levels of exosomal miRNAs detected in primary and metastatic
colon cancer. The criteria used to screen differences in miRNA
encapsulation in the exosomes between normal tissue versus
primary colon tumour tissue, and primary colon tumour tissue
versus metastatic tumour tissue were based on fold changes of
43.0 or o 3.0. Fifty of the miRNAs met these criteria and were
selected for further analysis (Table 1). A heat map of the 50
selected miRNAs demonstrated a gene cluster and sample cluster
according to the level of miRNAs in the exosomes from normal
tissue, primary colon tumour tissue and metastatic colon cancer
in the liver (Fig. 1d). To verify the microarray results, seven of the
upregulated miRNAs and three of the downregulated miRNAs in
liver metastasis of colon cancer (Fig. 1e) were randomly chosen
and confirmed by qPCR (Fig. 1f). The qPCR results indicated that
these miRNAs were encapsulated into exosomes that were
subsequently released into circulation (Fig. 1g, left panel) and
excreted into the gut (Fig. 1g, right panel) from metastatic CT26
tumour cells in the liver. The stable character of exosomal
miRNAs in biological fluids and faeces indicates that miRNAs in
exosomes can be used as potential biomarkers for clinical
diagnosis and prognosis.
Tumour suppressor miRNAs selectively sort into exosomes.To
determine whether the miRNA repertoires of exosomes differ
from those of their donor cells, the profiles of miRNAs from
exosomes and their donor cells were quantitatively analysed.
Scatter plot (Supplementary Fig. 4a,b) results demonstrated a
difference in exosomal RNA profiles from primary colon tumour
(Supplementary Fig. 4a) and naı
¨
ve colon (Supplementary Fig. 4b)
from their donor tissues. We then calculated the ratios of any
given miRNA from exosomes and their donor cells (Fig. 2a). Our
data suggest that loading miRNAs into exosomes is not a passive
process (Fig. 2a). In primary colon tumour-derived exosomes,
26.7% of miRNAs analysed are higher and 47.5% of miRNAs are
lower when compared to exosomal donor tumour cells; 25.8% of
miRNAs analysed are present in the exosomes in both types of
cells (Fig. 2a, top panel). However, when CT26 colon tumour cells
metastasize to the liver, the pattern of the CT26 tumour exosomal
miRNA profile is altered (Fig. 2a, middle panel) in comparison to
the pattern of the exosomal miRNAs expressed in the primary
colon tumour. Some of the exosomal miRNAs retain the same
patterns as indicated in yellow (Fig. 2a, bottom panel) regardless
of whether they are in the primary colon tumour or metastatic
colon cancer in the liver.
To investigate whether the level of tumour exosomal miRNA is
sorted based on the miRNA biological function, the miRNA
Table 1 | List of significant higher level or lower level miRNAs presented in the tumour exosomes (fold change).
Higher in tumour exo Primary cancer/Naı
¨
ve
(log2)
Metast/Primary
cancer (log2)
Lower in
tumour exo
Primary cancer/Naı
¨
ve
(log2)
Metast/Primary
cancer (log2)
mmu-miR-10a-5p 2.17 5.22 mmu-miR-423-5p 2.44 10.21
mmu-miR-126a-3p 6.21 4.76 mmu-miR-301a-3p 3.11 10.04
mmu-miR-22-3p 8.01 4.19 mmu-miR-33-5p 5.73 7.22
mmu-miR-192-5p 4.11 4.11 mmu-miR-9-5p 1.93 5.92
mmu-miR-339-5p 4.51 4.01 mmu-miR-151-5p 3.75 5.20
mmu-miR-148a-3p 8.59 3.94 mmu-miR-196b-5p 3.51 5.05
mmu-miR-193a-3p 5.24 3.75 mmu-miR-147-3p 3.83 5.03
mmu-miR-30b-5p 2.73 3.74 mmu-let-7i-5p 2.02 4.99
mmu-miR-200b-5p 1.94 3.34 mmu-miR-1195 2.30 4.97
mmu-miR-677-5p 2.31 3.17 mmu-miR-96-5p 3.41 4.62
mmu-miR-154-3p 1.72 2.96 mmu-miR-669b-5p 2.15 4.60
mmu-miR-678 3.02 2.79 mmu-miR-380-3p 2.36 4.17
mmu-miR-222-3p 2.59 2.57 mmu-miR-196a-5p 5.16 4.02
mmu-miR-203-3p 6.86 2.52 mmu-miR-375-3p 2.74 3.99
mmu-miR-331-3p 3.13 2.44 mmu-miR-181d-5p 3.32 3.90
mmu-miR-714 12.01 2.25 mmu-miR-672-5p 2.58 3.74
mmu-miR-883b-3p 1.84 1.94 mmu-miR-654-3p 2.43 3.74
mmu-miR-1199-5p 1.74 3.49
mmu-miR-683 1.80 3.43
mmu-miR-882 1.94 3.35
mmu-miR-875-3p 1.89 3.29
mmu-miR-711 1.87 3.20
mmu-miR-182-5p 2.45 3.15
mmu-miR-466f-5p 2.23 3.13
mmu-miR-764-3p 3.50 2.85
mmu-miR-504-5p 2.16 2.62
mmu-miR-17-5p 4.07 2.61
mmu-miR-1892 1.38 2.57
mmu-miR-205-5p 4.53 2.51
mmu-miR-146b-5p 1.79 2.46
mmu-miR-301b-3p 2.90 2.16
mmu-miR-15b-5p 4.14 1.92
mmu-miR-450b-3p 3.68 1.74
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14448
4 NATURE COMMUNICATIONS | 8:14448 | DOI: 10.1038/ncomms14448 | www.nature.com/naturecommunications
profiles were summarized (Table 2) based on miRNA-oncogenic
versus tumour-suppressive effects. We found that among the
three types of exosomes isolated from normal colon tissue,
primary colon tumour tissue and metastatic colon tumour in the
liver, metastatic CT26 colon-derived exosomes have the highest
level of tumour-suppressive miRNAs and the lowest level of
oncogenic miRNAs. We then investigated whether such a
difference is determined by the level of miRNA expressed in
the parent tissue. As summarized in Table 2, we noticed that most
of the miRNAs (miR-10a-5p, miR-193a-3p, miR-200b-5p,
miR-222-3p) that are actively sorted into exosomes have tumour
suppressive effects involving cell growth suppression, whereas
miRNAs (miR-196a/b, miR-181d-5p, miR-155-5p) that have
oncogenic effects are retained in the tumour cells even though the
levels of the oncogenic miRNAs are higher in their donor cells
than in the exosomes. This finding was further demonstrated by
reverse transcription-quantitative PCR (RT-qPCR) analysis of
tumour-suppressive miR-193a (Fig. 2b), miR-18a (Fig. 2c) and
oncogenic miR-21 (Fig. 2d), as an example. Our results indicate
that the level of miR-18a and miR-193a in the exosomes from
either primary colon tumour tissue or metastatic liver of colon
tumour is higher than in their donor tumour tissues. In contrast,
the level of oncogenic miR-21 was much higher in the primary
colon tumour tissue and metastatic colon tumour in the liver than
in their exosomes. Collectively, these data suggest that oncogenic
miRNAs are upregulated and that tumour suppressive miRNAs
are downregulated in the tumour, and this phenomenon is clearly
observed in metastatic colon tumour in the liver. Sorting
oncogenic miRNAs from exosomal donor cells into their
exosomes is suppressed, whereas sorting tumour suppressive
miRNAs into exosomes is enhanced. This may be one of the
mechanisms underlying tumour exosome-mediated promotion of
tumour progression.
Effects of microenvironment on the exosomal miRNA. The
majority of the published data show the biological activities of
tumour EVs using in vitro cultured tumour cell-derived EVs. This
may not accurately represent the case for tumour EV released
from tumour tissue because multiple factors derived from tumour
tissue have a remarkable effect on the composition of tumour
EVs, and those factors do not exist in the culture medium. As
proof of concept, we compared the levels of selected miRNAs
(Fig. 3) present in exosomes released from in vitro cultured CT26
cells (culture medium environment) from primary colon cancer,
CT26 cell subcutaneous xenograft, metastatic CT26 tumour
isolated from mouse liver and from exosomes circulating in the
peripheral blood. The results generated from qPCR show that
miR-126a, miR-148a and miR-193a are significantly higher in the
exosomes released from metastatic CT26 cells and circulating in
the peripheral blood of metastatic colon cancer in the liver, but
not from primary colon cancer or subcutaneous xenografts.
However, miR-22, miR-196a and miR-196b are decreased in the
exosomes from metastatic colon tumour in the liver (Fig. 3)
compared with exosomes from in vitro cultured CT26 cells. These
changes are specific as other miRNAs, including miR-10a,
miR-30b, miR-200b and miR-151, are not changed in amount
regardless of the origin, whether from the exosomes of cultured
tumour cells or metastatic CT26 cells, suggesting that the
microenvironment has an effect on the composition of the
exosomal miRNA profile. To further confirm that exosomes
isolated from colon tumour tissue do not contain other intra-
cellular microvesicles such as multivesicular bodies, the exosomes
were isolated from the supernatants of 12 h-ex vivo-cultured
CT26 colon cancer cells isolated from colon tumours. Exosomal
miRNAs isolated from the supernatants of ex vivo-cultured CT26
colon cancer cells and from CT26 colon cancer tissue were qPCR
analysed. The results suggest that levels of the exosomal miRNAs
in the exosomes from the supernatants of ex vivo-cultured CT26
cells are not significantly different from exosomal miRNAs in the
CT26 colon cancer tissue (Supplementary Fig. 5).
Oncosuppressor miR-193a directly targets Caprin1. We further
hypothesize that exporting tumour-suppressive miRNA such as
miR-193a from exosome donor cells into exosomes is a benefit for
colon cancer metastasis to the liver. We first searched miRNA
databases for potential miR-193a targets that may contribute or
promote tumour progression. Three public miRNA databases
(TargetScan, Pictar and MicroRNA) all predicted that cell cycle-
associated protein Caprin1 might be a target for miR-193a
(Fig. 4a), and the 3
0
-UTR of Caprin1 contains a highly conserved
binding site from position 2288 to 2309 for miR-193a (Fig. 4a,b).
To determine whether miR-193a could target Caprin1 in colon
cancer cells, we transfected the mature mouse miR-193a mimic
into CT26 cells. The CT26 cells overexpressing miR-193a (Fig. 4c,
left panel) have significantly downregulated Caprin1 mRNA
expression (Fig. 4c, right panel) as well as Caprin1 protein
expression (Fig. 4d). We found that CCND2 and c-MYC,
which are regulated by Caprin1, are also decreased as a result of
miR-193a treatment (Fig. 4c,d). The impact of miR-193a over-
expression on the inhibition of cell proliferation was further
confirmed by the Caprin1 siRNA knockdown in CT26 colon
cancer cells (Fig. 4e). To ascertain the direct effect of miR-193a on
Caprin1, a mutant construct that would disrupt the predicted
miR-193a binding site was generated from pEZX-MT01-Caprin1
containing a full length 3
0
UTR of Caprin1 mRNA (Gene
Accession: NM_001111289). We performed a luciferase reporter
assay by co-transfecting a vector containing Caprin1 3
0
UTR-fused
luciferase and miR-193a or control miRNA as a negative control.
Overexpression of miR-193a decreased the luciferase activity of
the reporter with the 3
0
UTR of Caprin1 by approximately 56% in
CT26 cells (Fig. 4f). However, the mutation that disrupted the
binding site for miR-193a entirely restored luciferase activity.
Moreover, overexpression of anti-sense miR-193a (miRNA
inhibitor) caused induction of luciferase; however, there was no
inductive effect of the anti-sense miR-193a on the activity of the
reporter with a mutant 3
0
UTR of Caprin1 (Fig. 4f). These results
demonstrate that Caprin1 is a target of miR-193a in colon cancer
cells. The tumour suppression role of miR-193a was further
supported by the fact that overexpression of miR-193a inhibited
CT26 cell proliferation and significantly prolonged survival
of colon cancer-bearing mice (Fig. 4g). Cell cycle assessment
suggested that miR-193a causes a G1 phase arrest in the cell cycle
(Fig. 4h,i).
MVP regulates the loading of miR-193a to exosomes. Although
the miRNA repertoires of exosomes differ from those of their
donor cells, the explanation or mechanism for how this occurs is
still unknown. We hypothesized that host factor(s) might play a
role in miRNA sorting from exosomal donor cells to their exo-
somes. To test our hypothesis, biotin-labelled miR-193a complex
was isolated from exosomal lysates using streptavidin beads.
A typical staining pattern of the Bio-miR-193a complex obtained
from CT26 exosomal extracts on sodium dodecyl sulfate poly-
acrylamide gel electrophoresis (SDS-PAGE) is shown in Fig. 5a,
left panel. In-gel digestion-MALDI-TOF mass spectrometry (MS)
analysis was carried out for identification of proteins that are
specifically present in the Bio-miR-193a complex sample but not
in the control bio-miRNA complex. Major vault protein (MVP)
was subsequently identified as a potential miR-193a binding
protein by MS (Fig. 5a, right panel), and this interaction was
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14448 ARTICLE
NATURE COMMUNICATIONS | 8:14448 | DOI: 10.1038/nc omms14448 | www.nature.com/naturecommunications 5