Clinical significance of exosomes as potential biomarkers in cancer

Abstract

BACKGROUND

Exosomes are microvesicles, measuring 30-100 nm in diameter. They are widely distributed in body fluids, including blood, bile, urine and saliva. Cancer-derived exosomes carry a wide variety of DNA, RNA, proteins and lipids, and may serve as novel biomarkers in cancer.

AIM

To summarize the performance of exosomal biomarkers in cancer diagnosis and prognosis.

METHODS

Relevant publications in the literature were identified by search of the “PubMed” database up to September 11, 2018. The quality of the included studies was assessed by QUADAS-2 and REMARK. For assessment of diagnostic biomarkers, 47 biomarkers and 2240 patients from 30 studies were included.

RESULTS

Our results suggested that these exosomal biomarkers had excellent diagnostic ability in various types of cancer, with good sensitivity and specificity. For assessment of prognostic markers, 50 biomarkers and 4797 patients from 42 studies were included. We observed that exosomal biomarkers had prognostic values in overall survival, disease-free survival and recurrence-free survival.

CONCLUSION

Exosomes can function as potential biomarkers in cancer diagnosis and prognosis.

Core tip: Cancer-derived exosomes carry a wide variety of DNA, RNA, proteins and lipids, which may serve as novel biomarkers in cancer. The current systematic review and meta-analysis summarized the performance of exosomal biomarkers in cancer diagnosis and prognosis. We analyzed 47 diagnostic markers and 50 prognostic markers from 56 studies with various type of cancer. We found that exosomal biomarkers had both diagnostic and prognostic power in many cancers.

INTRODUCTION

Cancer is the uncontrolled growth of cells and eventually leads to death. Cancer is the second cause of death, contributing to more than 8.8 million deaths every year[1,2]. Among various types of cancer, lung cancer, gastrointestinal cancers (GI cancer), including liver cancer, pancreatic cancer and colorectal cancer, and breast cancer are the most common cause of cancer-related death[2-4]. Although chemotherapy, targeted therapy, surgical recession and radiotherapy can effectively prolong survival of patients, the survival rate of cancer is still very low, especially in GI cancer, being less than 20%[2]. One of the major reasons is the late diagnosis of cancer, in which patients are already with advanced and metastatic tumors. As a result, no therapies can effectively kill the cancer cells. The situation is even worse in pancreatic cancers at distant stage, with 5-year survival rate of only 3%[2].

Since more than half of the patients present with locally advanced or metastatic stage, early diagnosis and early treatment are fundamentally important for better prognosis. Therefore, many tumor makers have been developed, aiming at accurately detecting various types of cancer and monitoring the disease progression. Blood test of the tumor antigens carcinoembryonic antigen, carbohydrate antigen 19-9, and carbohydrate antigen 125 (known as CEA, CA19-9 and CA125 respectively) are commonly used for detection of many cancers, such as GI cancers, ovarian cancer and breast cancer[5-8]. However, the sensitivity of these cancer biomarkers is unsatisfactory[9-12]. Also, the fecal occult blood test of colorectal cancer and the invasion endoscopic detection of gastric and colon cancer represent a great inconvenience to the patients. Therefore, highly sensitive and non-invasive diagnostic markers are urgently needed for early detection of cancer.

Exosomes are microvesicles of 30-100 nm diameter, which are secreted by both normal cells and cancer cells. They are distributed in many body fluids such as blood, saliva and urine, and carry various types of biomolecules, including RNA, proteins and lipids, for inter-cellular communication[13-15]. During cancer development, cancer cells secrete more exosomes, with significant changes in composition[16-18]. These facilitate communication within the tumor environment, acquisition of drug resistance, and metastasis to distant organs[19-21]. Although many potential non-invasive biomarkers have been developed using liquid biopsy, such as serum and urine, studies have found that these biomarkers are commonly located in the exosomes[22,23]. Enriching these exosomal biomarkers could achieve a higher diagnostic and prognostic efficiency[24-26]. Thus, exosomal biomarkers can be novel targets in cancer diagnosis and prognosis.

The objective of this systemic review and meta-analysis is to evaluate the diagnostic and prognostic potential of exosomes in patients with various types of cancer, based on current available data. This information will help in the development of novel non-invasive biomarkers for sensitive and specific diagnosis and prognosis of cancer.

MATERIALS AND METHODS

Search strategy

Electronic literature search was performed using the PubMed database, without any language restriction. Articles related to exosomes in cancer from 2010 to September 11, 2018 were identified using the following key words: “exosome” and “cancer” and ““diagnosis” or “prognosis””.

Inclusion and exclusion criteria

Articles were reviewed by their titles, key words, abstracts and full text to identify eligible studies. Eligible studies were included based on the following inclusion criteria: (1) The original article was related to exosomal diagnostic or prognostic markers in cancer; (2) At least 10 patients and 10 matched controls were enrolled in the study; (3) For diagnostic markers, enough information, such as specificity and sensitivity, was provided to construct 2 × 2 table [true positive (TP), true negative (TN), false positive (FP), false negative (FN)]; and (4) For prognostic markers, enough information was provided to estimate the hazard ratios (HRs) and confidence intervals (CIs). The exclusion criteria were as follows: (1) Duplicate articles; (2) Review articles, abstracts, comments, letters, case-report; (3) Fundamental research or animal study; (4) Diagnostic or prognostic marker that was not specific to exosome; (5) Sample size was less than 10; (6) Performance of the biomarker was not statistically significant; or (7) Incomplete information to estimate diagnostic or prognostic accuracy.

Data extraction

Two reviewers (Chi-Hin Wong and Yang-Chao Chen) independently reviewed and extracted the data from the eligible studies according to the listed criteria. Any disagreement was resolved by consensus among the authors. The following data from included studies were extracted: first author’s name, year of publication, sample size, cancer type, country of origin, source of exosome, isolation method of exosome, and detection method of biomarkers. For diagnostic studies, data for the cut-off value of tested targets, sensitivity, specificity, and area under the receiver operating characteristics curve (ROC) were also extracted. For prognostic studies, data for survival analysis, cut-off value, multivariable HR and its 95%CI were extracted. If odds ratio (OR) was reported, OR was converted to relative risk using the formula introduced by Zhang and Yu[27]. If either OR or HR was not reported, the method introduced by Tierney et al[28] was used to estimate the HR and its 95%CI from a Kaplan-Meier plot.

Quality assessment

For diagnostic studies, the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) was used to assess the quality of studies for the meta-analysis[29]. Briefly, 14 questions covering the patient selection, patient flow, index test and reference standard test were applied to each study and an answer of “Yes”, “No” or “Unclear” was given to each study. Only answers of “Yes” were given a score.

For prognostic study, the quality of studies was assessed according to reporting recommendations for tumor marker prognostic studies (REMARK)[30]. Briefly, a checklist of 20 items was generated, covering patients’ characteristics, samples’ source and storage, assay methods, statistical analysis, and data interpretation. A score was given when the study fulfilled the requirement of each item.

Statistical analysis

The statistical analysis of the diagnostic performance of biomarkers was performed using Meta-DiSc 1.4[31]. The 2 × 2 table of each study was used to assess the pooled sensitivity, specificity, positive likelihood ratio (PLR) and negative likelihood ratio (NLR). Also, the summary receiver operating characteristic (SROC) curve was plotted; the area under the curve (AUC) was calculated and Q* index was estimated to assess the overall performance in cancer diagnosis. An AUC of 0.5 suggested no diagnostic ability; 0.7-0.8 suggested acceptable diagnostic performance; 0.8-0.9 was considered excellent, and 0.9-1.0 suggested outstanding performance[32]. Q* was defined at a point in which sensitivity and specificity are equal. For statistical analysis of the prognostic performance of biomarkers, forest plots were constructed using the HR and its 95%CI of each biomarker to assess the overall prognostic performance of biomarkers on overall survival (OS), disease-free survival (DFS) and recurrence-free survival (RFS). Graphpad Prism 6 was used in constructing the forest plots. To elevate the heterogeneity between studies, Cochran-Q test and inconsistency index (I²) statistics were calculated[33,34]. P-value of < 0.05 for Cochran-Q test or I² >50% suggested the presence of heterogeneity.

RESULTS

Literature search

Initially, 1233 articles were identified based on the search strategies. Based on title and abstract screening, 705 were not related to exosome biomarkers in cancer diagnosis or prognosis, and 287 were review articles. Upon further full-text review, 56 studies were basic studies, 42 studies with sample size less than 10 in either group (test group or control group), 12 studies analyzed the performance of combined markers, 70 studies did not provide enough information for analysis, and 5 studies were without statistical significance. Finally, 56 eligible studies were included for systematic review (Figure 1). Of these, 22 candidate studies were related to diagnosis, 34 candidate studies were related to prognosis, and 8 studies were related to both diagnosis and prognosis.

Figure 1.

Literature search process to select studies which evaluated the diagnostic or prognostic performance of exosomal biomarkers in cancer.

Assessment of study quality

For diagnostic studies, the QUADAS-2 system was used to assess the study quality (Figure 2A). Most of the studies on diagnosis were with moderate-to-high quality, revealed by low risk of publication bias. However, there may be risk of bias in “patient selection” and “flow and timing”. This may due to control-based design in most of the studies. Also, time between the index test and the reference test is poorly reported. Importantly, many studies did not provide enough information on how the patients were selected and classified. Patients excluded from the 2 × 2 table were often observed in some studies.

Figure 2.

Quality assessment of the studies in this meta-analysis. A: QUADAS-2 system was used to assess the quality of diagnostic studies; B: REMARK checklist was used to assess the quality of prognostic studies.

The REMARK system was used to assess the quality of prognostic studies (Figure 2B). Most of the studies ( > 90%) clearly stated the objective, biomarkers examined, source of exosomes, and methodology of isolation and detection. Also, most of the studies clearly defined the clinical endpoints and the period of the follow-up time. However, details in patient’s characteristics during the follow-up period, such as the use of post-operative adjuvant therapy which significantly affects the OS and DFS, were lacking in most of the studies. Importantly, some studies did not clearly report the clinicpathological characteristics of the patients enrolled. Also, some studies did not show the relationship of the tested biomarkers to prognostic variables, including tumor stages and tumor differentiation. Twelve prognostic marker studies did not perform univariable or multivariable analysis. Twenty-eight of the enrolled studies reported multivariable analysis in prognostic markers, but only five studies clearly stated the adjustment factors.

Diagnostic markers

Diagnostic markers from 30 studies were included in the meta-analysis (Table 1). More than a half of these studies were related to GI cancers (4 studies were about colon cancer; 5 studies were related to liver cancer; 4 studies were about pancreatic or pancreatobiliary tract cancer; and 4 studies were related to gastric cancer). A total of 2240 patients were included in the meta-analysis, with 12 studies having enrolled < 50 patients, 16 studies having enrolled 50-100 patients, and 6 studies having enrolled > 100 patients. There were 47 diagnostic biomarkers analyzed in the meta-analysis. There were 42.6% of the biomarkers as miRNAs, followed by lncRNAs (36.2%) and proteins (19.1%). Notably, 6 studies analyzed the diagnostic performance of exosomal miR-21 in various types of cancer. Also, 61.3%, 16.1%, 12.9%, 3.2% and 3.2% of the biomarkers were detected in serum, plasma, urine, saliva and bile respectively.

Table 1.

Studies included for meta-analysis of exosomal biomarkers in cancer diagnosis

Ref.	Country	Cancer type	Stage	Control	Number of Control	Number of patients	Sample	Isolation method of exosome	Marker	Detection method	Cut-off	TP	TN	FP	FN
Sun et al[35]	China	Colorectal	All	Healthy	32	92	Plasma	UC	CPNE3	ELISA	0.143 pg/μg exosome	62	27	5	30
Ogata-Kawata et al[36]	Japan	Colorectal	All	Healthy	11	88	Serum	UC	miR-1246	qRT-PCR	1.45	84	10	4	1
									miR-23a		0.3100	81	11	7	0
									miR-21		1.08	54	10	34	1
									miR-150		0.08	49	11	39	0
									let-7a		0.9	44	10	44	1
									miR-223		1.72	41	10	47	1
									miR-1224-5p		0.5	28	11	60	0
									miR-1229		0.06	20	11	68	0
Liu et al[37]	China	Colorectal	All	Healthy and benign	320	148	Serum	ExoQuick	CRNDE-h	qRT-PCR	0.02	104	302	18	44
Uratani et al[38]	Japan	Colorectal	NR	Healthy	47	26	Serum	ExoQuick	miR-21	qRT-PCR	Youden index	18	38	9	8
Lin et al[39]	China	Gastric	All	Healthy	60	51	Plasma	UC	lncUEGC1	qRT-PCR	NR	45	50	10	6
									lncUEGC2		NR	46	34	26	17
Zhao et al[40]	China	Gastric	All	Healthy	120	126	Serum	NR	HOTTIP	qRT-PCR	1.72	88	102	18	38
Pang et al[41]	China	Gastric	All	Healthy	37	40	Serum	ExoQuick	ZFAS1	qRT-PCR	NR	32	28	9	8
Yang et al[42]	China	Gastric	All	Healthy	80	80	Serum	ExoQuick	miR-423-5p	qRT-PCR	NR	65	46	34	15
Goto et al[43]	Japan	Pancreatic	All	Healthy and advanced pancreatic cancer	22	23	Serum	ExoQuick	miR-191	qRT-PCR	Distance = (1-sensitivity)2 + (1-specificity)2 in ROC curve	18	17	5	5
									miR-21			20	18	4	3
									miR-451a			16	18	4	7
Melo et al[44]	Germany	Pancreatic	All	Healthy	100	190	Serum	UC	GPC1	Flow cytometry	Youden index	190	100	0	0
Que et al[45]	China	Pancreatic	All	Non-PDAC	27	22	Serum	UC	miR-17-5p	qRT-PCR	6.826	20	20	7	2
									miR-21		7.693	18	26	1	4
Machida et al[46]	Japan	Pancreatobiliary tract	II-IV	Healthy	13	12	Saliva	Total exosome isolation kit	miR-1246	qRT-PCR	13.77	8	13	0	4
									miR-4644		-5.205	9	10	3	3
Xu et al[47]	China	Liver	All	Chronic hepatitis B	68	88	Serum	Total exosome isolation kit	hnRNPH1	qRT-PCR	0.67	75	52	16	13
Sun et al[48]	China	Liver	All	Healthy	56	56	Serum	Total exosome isolation kit	LINC00161	qRT-PCR	NR	42	41	15	14
Xu et al[49]	China	Liver	All	Chronic hepatitis B	96	60	Serum	Total exosome isolation kit	ENSG00000258332.1	qRT-PCR	1.345	43	80	16	17
					60	55			ENSG00000258332.1		1.366	40	48	12	15
					96	60			LINC00635		1.69	46	75	21	14
					60	55			LINC00635		1.532	44	45	15	11
Goldvaser et al[50]	Israel	Pan-cancer (not include liver)		Healthy	45	98	Serum	Total exosome isolation kit	hTERT	qRT-PCR	NR	61	45	0	37
		Liver	NR	Healthy	45	35					NR	21	45	0	14
Zhang et al[51]	China	Lung	All	Healthy	30	77	Serum	ExoQuick	MALAT-1	qRT-PCR	NR	62	21	9	15
Sun et al[52]	China	Lung	All	Healthy	15	15	Plasma	UC	14-3-3ζ	ELISA		9	12	3	6
Li et al[53]	NR	Ovarian		Benign	21	50	Serum	UC	ephrinA2	ELISA	20.4 ng/L	44	17	4	6
Meng et al[54]	NR	Ovarian	All	Benign	20	163	Serum	Total exosome isolation kit	miR-200a	PCR+ qRT-PCR	Youden index	135	18	2	28
									miR-200b			86	20	0	77
									miR-200c			51	20	0	112
Pan et al[55]	Germany	Ovarian	All	Healthy	29	106	Plasma	ExoQuick	miR-21	PCR+ qRT-PCR	Youden index	65	24	5	41
									miR-100			66	21	8	40
									miR-200b			68	25	4	38
									miR-320			59	20	9	47
Bryzgunova et al[56]	Russia	Prostate	All	Healthy	20	14	Urine	UC	miR-125	qRT-PCR	NR	12	13	7	2
									miR-19b		NR	11	19	1	3
Wang et al[57]	China	Prostate	II-IV	Healthy	30	34	Plasma	Total exosome isolation kit	SAP30L-AS1	qRT-PCR	NR	21	25	5	13
									SChLAP1		NR	30	23	7	4
Øverbye et al[58]	NR	Prostate	All	Healthy	15	16	Urine	UC	ADIRF	Mass spectrometry	Youden index	12	16	0	3
									TMEM256			14	16	0	1
Işın et al[59]	NR	Prostate	All	BPH	49	30	Urine	Urine Exosome RNA Isolation Kit	LincRNA-p21	qRT-PCR	0.181	20	31	18	10
Wang et al[60]	China	Laryngeal	All	Vocal cord polyps	49	52	Serum	ExoQuick	miR-21	qRT-PCR	0.043	36	40	9	16
									HOTAIR		0.032	48	28	21	4
Alegre et al[61]	NR	Melanoma	NR	Healthy	25	53	Serum	ExoQuick	exo-MIA	ELISA	1.4 μg/L	42	20	5	11
									exo-S100B	ELISA	0.015 μg/L	42	20	5	11
Manterola et al[62]	France	GBM	NR	Healthy	30	50	Serum	ExoQuick	RNU6	qRT-PCR	0.372	33	20	10	17
Chen et al[63]	Taiwan	Bladder	All	hernia	81	140	Urine	UC	TACSTD2	ELISA	2.47 ng/mL	103	62	19	37
Ge et al[64]	China	Cholangiocarcinoma	All	Biliary obstruction	56	35	Bile	UC	ENST00000588480.1	qRT-PCR	NR	22	41	15	13

UC: Ultracentrifugation; NR: Not reported.

Since a wide range of cancers was studied by different groups, we separated the diagnostic biomarkers according to cancer types and meta-analyzed cancer types with more than three biomarkers studied. Therefore, we focused on colorectal cancer (4 studies with 11 biomarkers), gastric cancer (4 studies with 5 biomarkers), pancreatic cancer (4 studies with 8 biomarkers), liver cancer (4 studies with 7 biomarkers), and prostate cancer (4 studies with 7 biomarkers (Figures 3-7). We observed that the pooled biomarkers had a good specificity of 0.87 but poor sensitivity of 0.57 in colorectal cancer diagnosis (Figure 3A and B). The PLR and NLR were 2.02 and 0.21 respectively (Figure 3C and D). The diagnostic OR was 20.35 (Figure 3E). Importantly, the AUC of the SROC curve was 0.89 and the Q* was 0.82 (Figure 3F). In diagnosis of gastric cancer, we observed that the pooled biomarkers had a good sensitivity of 0.77 and specificity of 0.73 with PLR, NLR, AUC of the SROC curve and Q* of 2.94, 0.32, 9.88, 0.84 and 0.77 respectively (Figure 4). For diagnosis of pancreatic cancer, we also observed the pooled biomarkers had an excellent sensitivity of 0.91 and specificity of 0.90 with PLR, NLR, AUC of the SROC curve and Q* of 6.35, 0.19, 40.71, 0.94 and 0.88 respectively (Figure 5). In liver cancer, the pooled biomarkers had a good diagnostic sensitivity of 0.76 and specificity of 0.80 with PLR, NLR, AUC of the SROC curve and Q* of 3.51, 0.32, 12.45, 0.85 and 0.78 respectively (Figure 6). The pooled biomarkers also had a good sensitivity of 0.77 and specificity of 0.79 in detecting prostate cancer with PLR, NLR, AUC of the SROC curve and Q* of 3.84, 0.28, 17.88, 0.88 and 0.80 respectively (Figure 7). The high sensitivity, specificity and Q* demonstrated that the pooled biomarkers could effectively discriminate cancer patients from healthy people or non-cancer patients.

Figure 3.

Forest plot of pooled (A) sensitivity, (B) specificity, (C) positive likelihood ratio, (D) negative likelihood ratio, (E) diagnostic odds ratio and (F) SROC curve of exosomal biomarkers in diagnosis of colon cancer. SROC: Summary receiver operating characteristic.

Figure 7.

Forest plot of pooled (A) sensitivity, (B) specificity, (C) positive likelihood ratio, (D) negative likelihood ratio, (E) diagnostic odds ratio, and (F) SROC curve of exosomal biomarkers in diagnosis of prostate cancers. SROC: Summary receiver operating characteristic.

Figure 4.

Forest plot of pooled (A) sensitivity, (B) specificity, (C) positive likelihood ratio, (D) negative likelihood ratio, (E) diagnostic odds ratio, and (F) SROC curve of exosomal biomarkers in diagnosis of gastric cancer. SROC: Summary receiver operating characteristic.

Figure 5.

Forest plot of pooled (A) sensitivity, (B) specificity, (C) positive likelihood ratio, (D) negative likelihood ratio, (E) diagnostic odds ratio and (F) SROC curve of exosomal biomarkers in diagnosis of pancreatic cancer. SROC: Summary receiver operating characteristic.

Figure 6.

Forest plot of pooled (A) sensitivity, (B) specificity, (C) positive likelihood ratio, (D) negative likelihood ratio, (E) diagnostic odds ratio, and (F) SROC curve of exosomal biomarkers in diagnosis of liver cancers. SROC: Summary receiver operating characteristic.

Prognostic markers

Prognostic biomarkers from 42 studies were included in the systematic review (Table 2). In total, 4797 patients were represented among the studies, with 7 studies having enrolled < 50 patients, 15 studies having enrolled 50-100 patients, and 20 studies having enrolled > 100 patients. There were 50 prognostic biomarkers analyzed in the systematic review, with 60% of the biomarkers being miRNAs, followed by lncRNAs (18%) and proteins (16%). Also, 50%, 43%, 2.4%, 2.4% and 2.4% of the biomarkers were detected in serum, plasma, bile, ascetic fluid and cell-free effusion supernatant respectively. For the included studies, 92.9%, 26.2% and 9.5% used OS, DFS and RFS respectively as the primary endpoints. In addition, a wide range of cancers was studied by the different groups. More than one-half of the included studies were related to GI cancers (11 studies were about colorectal or colon cancer, 5 studies were related to liver cancer, 5 studies were about pancreatic cancer, and 4 studies were related to gastric cancer). In this meta-analysis, we separated studies according to clinical endpoints and focused on cancer types with more than three biomarkers studied.

Table 2.

Studies included for meta-analysis of exosomal biomarkers in cancer prognosis

Ref.

Period

Country

Sample Size

Cancer Type

Stage

Sample

Isolation method of exosome

Marker

Detection method

Cut-off value

Survival analysis

HR (95%CI)

Peng et al[65]

2008-2014

China

108

Colorectal

All

Serum

Total exosome isolation kit

miR-548c-5p

qRT-PCR

NR

OS

3.40 (1.02‐11.27)

Sun et al[35]

2012-2017

China

92

Colorectal

All

Plasma

UC

CPNE3

ELISA

≥ 0.143 pg/μg exosome

OS

3.0 (1.0-8.9)

≥ 0.143 pg/μg exosome

DFS

2.5 (1.1-5.5)

Tsukamoto et al[66]

2002-2012

Japan

326

Colorectal

II-IV

Plasma

UC

miR-21

qRT-PCR

> median

OS

2.28 (1.81-5.74)

DFS

2.34 (1.87- 4.60)

Liu et al[37]

2007-2010

China

148

Colorectal

All

Serum

ExoQuick

CRNDE-h

qRT-PCR

> 0.02

OS

2.000 (1.269-3.154)

Liu et al[67]

2006-2011

United States

84

Colorectal

II-III

Serum

ExoQuick

miR-4772-3p

qRT-PCR

≥ 27.88

OS

6.19 (1.50-25.5)

≥ 27.88

RFS

5.48 (2.49-12.1)

Liu et al[24]

2013-2014

China

158

Colorectal

All

Plasma

UC

lncRNA GAS5

qRT-PCR

NR

OS

0.265 (0.082 -0.844)

RFS

0.449 (0.194- 0.909)

miR-221

qRT-PCR

NR

OS

2.141 (1.368-3.054)

RFS

1.600 (1.162-2.007)

Gao et al[68]

2011-2014

China

108

Colorectal

All

Serum

ExoQuick

91H

qRT-PCR

≥ 0.85

RFS

7.14 (1.23-21.35)

Yan et al[69]

NR

NR

168

Colorectal

All

Serum

Total Exosome Isolation kit

miR-6803

qRT-PCR

NR

OS

2.93 (1.35-6.37)

DFS

3.26 (1.56-6.81)

Li et al[70]

2013-2015

China

85

Colorectal

III

Plasma

ExoCapTM

GPC1

Flow cytometry

> mean

OS

1.89 (1.23-2.89)

Silva et al[71]

2003-2009

Spain

91

Colorectal

All

Plasma

UC

Exosome

Flow cytometry of EpCAM

High

OS

0.87 (0.57-1.32)

Matsumura et al[72]

1992-2007

Japan

209

Colorectal

All

Serum

UC

miR-19

qRT-PCR

> mean

O

2.49 (1.12-6.61)

DFS

2.49 (1.12-6.61)

Yan et al[73]

2012-2015

China

142

Colorectal

All

Serum

Total Exosome Isolation kit

miR-6869-5p

qRT-PCR

< mean

OS

2.32 (1.08-4.99)

Santasusagna et al[25]

2009-2013

Spain

32

Colon

I-III

Plasma

UC

miR-141

qRT-PCR

High

OS

1.89 (0.93-3.83)

Zhao et al[40]

2011-2012

China

126

Gastric

All

Serum

NR

HOTTIP

qRT-PCR

> 1.72

OS

2.037 (1.085-3.823)

Liu et al[74]

2012-2017

China

76

Gastric

All

Serum

Total Exosome Isolation kit

miR-451

qRT-PCR

> median

5yr-OS

4.344 (2.853‐5.721)

Yang et al[42]

NR

China

80

Gastric

All

Serum

ExoQuick

miR-423-5p

qRT-PCR

> median

DFS

1.93 (1.25-2.99)

OS

1.42 (0.92-2.20)

Kumata et al[75]

2006-2013

Japan

232

Gastric

All

Plasma

UC

miR23b

qRT-PCR

> 0.78

OS

0.57 (0.370.78)

DFS

0.64 (0.410.91)

Zhou et al[76]

2010-2014

China

152

Pancreatic

All

Plasma

ExoQuick

miR-125b-5p

qRT-PCR

< median

OS

0.285 (0.108-0.75)

Li et al[77]

2012-2016

China

87

Pancreatic

All

Plasma

NR

circPDE8A

qRT-PCR

> median

OS

1.764 (1.064-2.925)

Goto et al[43]

2013-2015

Japan

32

Pancreatic

All

Serum

ExoQuick

miR-21

qRT-PCR

> median

OS

4.071 (1.832-11.996)

Takahasi et al[78]

2013-2017

Japan

50

Pancreatic

I-II

Plasma

UC

miR-451a

qRT-PCR

> 1.75

OS

3.20 (1.07-11.94)

DFS

2.87 (1.23-7.23)

Xu et al[49]

2012-2016

China

60

Liver

All

Serum

Total Exosome Isolation kit

ENSG00000258332.1

qRT-PCR

> 1.845

OS

2.22 (1.34-3.68)

LINC00635

qRT-PCR

> 2.100

OS

1.46 (0.88-2.43)

Shi et al[79]

2008-2011

China

126

Liver

All

Serum

Total Exosome Isolation kit

miR-638

qRT-PCR

NR

3yr-OS

3.52 (1.37-6.02)

5yr-OS

2.80 (1.24-4.31)

Liu et al[26]

2012

China

128

Liver

All

Serum

ExoQuick

miR-125b

qRT-PCR

< median

RFS

0.14 (0.07-0.29)

OS

0.36 (0.18-0.74)

Xue et al[80]

2015-2017

China

85

Liver

All

Serum

Total Exosome Isolation kit

miR-93

qRT-PCR

NR

OS

1.47 (0.96-2.25)

Liu et al[81]

2008-2013

China

32

Hepatoblastoma (children)

All

Serum

ExoQuick

miR-21

qRT-PCR

NR

EFS

1.434 (1.257-2.766)

Matsumoto et al[82]

2011-2012

Japan

66

Esophageal

All

Plasma

Total Exosome Isolation kit

exosome

AChE activity

< 600 x 108/mL

OS

2.177 (1.085-3.605)

Lu et al[83]

2007-2015

China

110

Nasopharyngeal

All

Plasma

UC

miR-9

qRT-PCR

NR

OS

1.5 (1.03-2.18)

Ye et al[84]

2011-2013

China

83

Nasopharyngeal

II-IV

Serum

UC

protein concentration

BCA assay

> 11 μg/mL

DFS

214.22 (139.27-329.49)

Huang et al[85]

NR

NR

23

Prostate

All

Plasma

ExoQuick

miR-1290

qRT-PCR

> mean

OS

1.79(1.30-2.48)

miR-375

qRT-PCR

> mean

OS

2.69(1.52-4.77)

Tang et al[86]

NR

NR

35

Ovarian

All

Ascitic fluid

UC

E-cadherin

NR

> 10 μg/mL

OS

1.82 (0.53-3.58)

Vaksman et al[87]

1998-2003

86

Ovarian

III-IV

Effusion supernatant

ExoQuick

miR-21

qRT-PCR

> median

OS

1.70 (1.1-2.59)

Kanaoka et al[88]

2012-2017

Japan

285

Lung

I-III

Plasma

UC

miR-451a

qRT-PCR

> 1.45

OS

6.06 (2.61-15.94)

DFS

2.55 (1.44-4.65)

Liu et al[89]

2012-2014

China

196

Lung

All

Plasma

ExoQuick

miR-23b-3p

qRT-PCR

High

OS

2.42 (1.45-4.04)

miR-21-5p

qRT-PCR

OS

2.12(1.28-3.49)

miR-10b-5p

qRT-PCR

OS

2.22 (1.18-4.16)

Liu et al[90]

2012-2014

China

208

Lung

All

Plasma

ExoQuick

Exosome

AChE activity

OS

1.72 (1.05-2.83)

Sandfeld-Paulsen et al[91]

2011-2014

Denmark

276

Lung

All

Plasma

/

CD171

ELISA

NR

OS

0.56 (0.41-0.79)

Flotilin1

ELISA

NR

OS

0.63 (0.46-0.86)

HER3

ELISA

NR

OS

0.63 (0.46-0.86)

GRP78

ELISA

NR

OS

0.69 (0.51-0.91)

Manier et al[92]

2006-2008

France

156

Multiple myeloma

All

Plasma

ExoQuick

let-7b

qRT-PCR

< median

OS

2.83 (1.07-7.50)

let-7b

qRT-PCR

< median

DFS

1.90 (1.22-2.94)

let-7e

qRT-PCR

< median

DFS

2.01 (1.30-3.11)

miR-106a

qRT-PCR

< median

DFS

2.34 (1.52-3.61)

miR-106b

qRT-PCR

< median

DFS

3.54 (2.21-5.68)

miR-155

qRT-PCR

< median

OS

2.41 (0.96-6.05)

miR-155

qRT-PCR

< median

DFS

1.76 (1.15-2.69)

miR-16

qRT-PCR

< median

DFS

2.21 (1.41-3.47)

miR-17

qRT-PCR

< median

DFS

2.29 (1.48-3.55)

miR-18a

qRT-PCR

< median

DFS

4.52 (1.57-12.98)

miR-18a

qRT-PCR

< median

OS

2.76 (1.79-4.26)

miR-20a

qRT-PCR

< median

DFS

2.31 (1.52-3.53)

Alegre et al[61]

NR

NR

53

Melanoma

NR

Serum

ExoQuick

MIA

ELISA

2.5 μg/L

OS

1.28 (0.65-2.51)

Lan et al[93]

2011-2012

China

60

Glioma

All

Serum

ExoQuick

miR-301a

qRT-PCR

>median

OS

4.4 (3.1-9.6)

Ge et al[64]

NR

China

35

Cholangiocarcinoma

All

Bile

UC

ENST00000588480.1

qRT-PCR

> median

OS

2.40 (1.24-4.66)

ENST00000517758.1

qRT-PCR

OS

1.55 (0.80-3.01)

Fujii et al[94]

2005-2014

Japan

108

Renal cell

I-III

Serum

Total Exosome Isolation kit

miR-224

qRT-PCR

> median

OS

9.1 (1.8-166.1)

UC: Ultracentrifugation; OS: Overall survival; DFS: Disease-free survival; RFS: Recurrence free survival; EFC: Event-free survival; NR: Not reported.

For 13 biomarkers with OS reported in colon cancer, the pooled HR was 1.833 with I² of 62.14% and P = 0.002 (Figure 8A). Also, for 5 biomarkers with DFS reported in colon cancer, the pooled HR was 3.035 with I² of 0.00% and P = 0.536 (Figure 8B). Furthermore, for 4 biomarkers with RFS reported in colon cancer, the pooled HR was 1.645 with I² of 89.61% and P = 0.000 (Figure 8C). Apart from colon cancer, for the 4 biomarkers with OS reported in gastric cancer, the pooled HR was 1.836 with I² of 96.71 and P = 0.000 (Figure 9). In addition, for the 4 biomarkers with OS reported in pancreatic cancer, the pooled HR was 1.537 with I² of 81.50 and P = 0.001 (Figure 10). For 5 biomarkers, the pooled HR was 1.828, I² of 84.48% and P = 0.000 for prognosing OS in liver cancer (Figure 11). Also, 9 biomarkers with the pooled HR of 0.895, I² of 89.50% and P = 0.000 were reported to function as prognostic biomarkers of OS in lung cancer (Figure 12). These results demonstrated that exosomes were associated with OS, DFS and RFS in various types of cancer.

Figure 8.

Forest plot evaluating the effect of exosomal markers on overall survival (A), disease-free survival (B), and (C) recurrence-free survival of patients with colon cancer.

Figure 9.

Forest plot evaluating the effect of exosomal markers on overall survival of patients with gastric cancer.

Figure 10.

Forest plot evaluating the effect of exosomal markers on overall survival of patients with pancreatic cancer.

Figure 11.

Forest plot evaluating the effect of exosomal markers on overall survival of patients with liver cancer.

Figure 12.

Forest plot evaluating the effect of exosomal markers on overall survival of patients with lung cancer.

DISCUSSION

Exosomes play important roles in cancer development via intercellular communication, promoting cell metastasis and developing drug resistance[19-21]. Importantly, exosomes are frequently secreted by the cancers and are widely distributed in many body fluids. Therefore, they can be detected in blood, saliva and urine. Exosomal biomarkers have better performance in cancer diagnosis and prognosis than liquid biopsy used alone[24-26]. However, the methods of isolating exosomes from liquid biopsy varies between studies. Ultracentrifugation or the use of commercial isolation kits are common methods in extracting exosomes. Ultracentrifugation gives highly pure exosomes but the isolation efficiency is relatively low; whereas, the use of commercial kits maximizes the efficiency with the loss of purity[95,96]. Therefore, a standardized protocol of detecting exosomal biomarkers is greatly needed.

There are some limitations of our meta-analysis. We excluded studies that utilized combined biomarkers because this cannot tell the performance of individual biomarkers[97,98]. For example, a six-microRNA panel was developed for diagnosis of lung cancer but miR-409-3p, miR-425-5p and miR-584-5p were not significantly dysregulated in patients’ exosomes[98]. This may reduce the diagnostic performance of other biomarkers in the same panel. Since many of the individual biomarkers in the panel were significantly differentially expressed in cancer exosomes, further studies may be needed to explore the correlation of these potential biomarkers with patients’ characteristics and their performances in cancer diagnosis and prognosis.

A further limitation is that we focused on exosomal markers only in cancer diagnosis and prognosis and excluded tissue-based biomarkers from this meta-analysis. In fact, many studies have reported that expression levels in exosomes and in tissues are highly associated[35,66]. This suggests that many exosomal markers can reflect the situation in cancer cells, and this notion has been developed for potential biomarkers in various cancers. Importantly, this strong association may also suggest that many tissue-based biomarkers can be developed into non-invasive exosomal biomarkers in cancer diagnosis.

Notably, most of the included studies are retrospective, having been performed on stored samples. However, the main disadvantage of the retrospective study is its lack of complete clinicpathological information[30], which lowers the quality of study. Despite the above limitations, our meta-analysis indicates that exosomes can be potential biomarkers in cancer diagnosis and prognosis. Further large prospective studies are greatly needed to clarify the performance of exosomal biomarkers in cancer diagnosis and prognosis.

ARTICLE HIGHLIGHTS

Research background

Exosomes, which are widely distributed in body fluids, including blood, bile, urine and saliva, are microvesicles of 30-100 nm diameter in size. Cancer-derived exosomes carry a wide variety of DNA, RNA, proteins and lipids, and may serve as novel biomarkers in cancer.

Research motivation

Exosomes may function as exosomal biomarkers in cancer diagnosis and prognosis.

Research objectives

To summarize the performance of exosomal biomarkers in cancer diagnosis and prognosis.

Research methods

Relevant studies in the literature were identified using the PubMed database. QUADAS-2 and REMARK were used to assess the quality of the included studies. For diagnostic biomarkers, 47 biomarkers and 2240 patients from 30 studies were included.

Research results

These exosomal biomarkers had excellent diagnostic ability in various types of cancer, with good sensitivity and specificity. A total of 50 biomarkers and 4797 patients from 42 studies were included for the prognostic markers. We observed that exosomal biomarkers had prognostic values in overall survival, disease-free survival and recurrence-free survival.

Research conclusions

Exosomes could be potential biomarkers in cancer diagnosis and prognosis.

Research perspectives

Further large prospective studies are needed to clarity the performance of exosomal biomarkers in cancer diagnosis and prognosis, through exosomes can be potential biomarkers in cancer diagnosis and prognosis.

ACKNOWLEDGEMENTS

The work described in this paper was supported by grants from the General Research Fund, Research Grants Council of Hong Kong (CUHK462713, 14102714, 14136416 and 14171217), National Natural Science Foundation of China (8142730 and 81672323) and Direct Grant from CUHK to YC.