Welcome international healthcare professionals

This site is no longer supported and will not be updated with new content. You are welcome to browse and download all content already included in the site. Please note you will have to register your email address to access the site.

You are here

Tumour vasculogenic mimicry is associated with poor prognosis of human cancer patients: A systemic review and meta-analysis

European Journal of Cancer, 18, 49, pages 3914 - 3923



Vasculogenic mimicry (VM) has been reported in various malignant tumours and is known to play an important role in cancer progression and metastasis. However, the impact of VM on the overall survival of human cancer patients remains controversial. The goal of this study was to evaluate whether VM is associated with 5-year survival of human cancer patients.


Twenty-two eligible clinical studies with data on both tumour cell-dominant VM and the 5-year survival of 3062 patients involved in 15 types of cancers were pooled in the meta-analysis.


The 5-year overall survival of VM-positive and -negative cancer patients was 31% and 56%, respectively. The relative risk (RR) of the 5-year survival of VM-positive patients was significantly higher than that of VM-negative cases (RR = 1.531; 95% confidence interval (CI): 1.357–1.726; P < 0.001). Notably, metastatic melanoma patients demonstrated a higher VM rate (45.3%) than patients with primary melanoma (23.1%) and showed worse 5-year survival, suggesting that VM contributes to tumour metastasis and poor prognosis in cancer patients. Subgroup analysis indicated that a poor 5-year survival was significantly associated with eight types of VM-positive malignant tumours, such as lung, colon, liver cancers, sarcomas and melanoma; but was not associated with the seven other types of cancers, such as prostate cancer. Heterogeneity and publication biases were found among the 22 studies, mainly due to the divergent characteristics of cancers and extremely low survival rate in six types of malignant tumours.


VM-positive cancer patients show a poor 5-year overall survival compared with VM-negative malignant tumour cases, particularly in metastatic cancer.

Keywords: Vasculogenic mimicry, Angiogenesis, Neovascularisation, Malignant tumour, Metastasis, Survival, Meta-analysis.

1. Introduction

Tumour angiogenesis plays an important role in tumour growth and metastasis and has been regarded as a hallmark of cancer [1], [2], [3], and [4]. Malignant tumours can generate their vasculature in six distinct ways, namely through sprouting angiogenesis, vasculogenesis, intussusception, vessel co-option, vasculogenic mimicry (VM) and trans-differentiation of cancer stem-like cells into tumour endothelial cells [4], [5], and [6]. Aggressive tumour cells adopt the ability of embryonic vasculogenesis to produce primitive tube-like structures and networks [7] ; particularly, various malignant tumours can directly form tumour blood vessels through VM, independent of vascular endothelial cells, and tumour cell-lined blood vessels vitally support tumour oxygen and nutrition supplies, promoting cancer progression [8], [9], [10], and [11].

Tumour VM, which refers to tumour cells directly lined up to form blood vessels, was first reported in melanoma by Hendrix and colleagues in 1999 [10] . Accumulated evidence has shown that VM exists in various malignant tumours [9], [12], and [13], including melanoma [14] and [15], ovarian cancer [11] , breast cancer [16] , prostate cancer [17] , osteosarcoma [18] , bladder cancer [19] , colorectal cancer [20] , hepatocellular cancer [21] , gastric cancer [22] and [23] and lung cancer [24] . VM is intimately linked with the ability of the tumour to establish an adequate vascular supply to enhance tumour growth and metastasis, and was linked to the poor prognosis of cancer patients [10], [20], [21], [24], and [25]; however, several studies have shown that VM was not significantly associated with tumour prognosis statistically, although these studies have indicated that VM-positive cancer patients displayed a shorter survival compared with VM-negative cancer patients [25], [26], [27], and [28]. Hence, the impact of VM on the overall survival of cancer patients remains controversial.

To clarify the correlation between VM and the prognosis of cancer patients, we conducted a meta-analysis to evaluate the influence of VM on the 5-year survival of 3062 cancer cases involved in 15 types of malignant tumours. The results show that VM is significantly associated with a poor 5-year overall survival of cancer patients.

2. Methods

2.1. Literature search and study selection

The current meta-analysis was limited to studies that evaluated the prognostic implication of VM in human cancers. A systematic literature search for “vasculogenic mimicry”-related papers in Pubmed, Web of Science, EMBASE and Cochrane Library databases from 1999 to 22nd August 2012 was conducted independently by two investigators (Zhifei Cao and Meimei Bao; Supplemental Table 1 ). Using the key words “vasculogenic mimicry or periodic acid-Schiff (PAS) staining, or tumour cell-lined vessels” and “prognosis or survival”, relevant research papers regarding VM in human cancers were retrieved and screened by the two investigators separately.

2.2. Data extraction and quality assessment

Data from eligible studies were extracted and read by the two investigators independently, and any disagreement was resolved by discussion. All data using a standardised data-collection form were outlined in Table 1 and recorded in detail in Supplemental Table 2 , including the first author, language, publication year, population, tumour type, VM assay methods, total cases, VM-positive or -negative rate and 5-year survival of cancer patients.

Table 1 The relative risk of 5-year survival of vasculogenic mimicry positive human cancer patients.

Study contents Studies (n) Cases (n) VM + (n) VM+ (%) Relative risk (95% confidence interval (CI)) P I2 (%)
Total studies 22 3062 744 24.3 1.531 (1.357–1.726) <0.001∗∗ 75.2
Single cancer type
 BDMT # 1 158 35 22.2 1.193 (1.091–1.305) <0.001∗∗ NA
 Breast cancer 1 331 26 7.9 1.508 (0.856–2.659) 0.155 NA
 Colorectal cancer 2 320 62 19.4 1.789 (1.007–3.179) 0.047 89.3
 Gallbladder carcinoma 1 71 18 25.4 1.182 (1.023–1.364) 0.023 NA
 Gastric cancer 2 257 61 23.7 1.645 (0.781–3.463) 0.19 86.4
 Gliomas 1 101 13 12.9 1.231 (0.877–1.727) 0.23 NA
 HCC # 2 250 43 17.2 1.417 (1.176–1.707) <0.001∗∗ 0.0
 Medulloblastoma 1 41 9 22.0 1.422 (0.998–2.027) 0.051 NA
 Melanoma 4 606 192 31.7 2.383 (1.084–5.238) 0.031 93.4
 NSCLC # 1 160 59 36.9 1.883 (1.528–2.320) <0.001∗∗ NA
 OLSCC # 2 315 85 27.0 1.404 (1.024–1.926) 0.035 0.0
 Ovarian tumour 1 84 36 42.9 1.200 (0.545–2.644) 0.651 NA
 Prostate cancer 1 96 24 25.0 2.000 (0.355–11.265) 0.432 NA
 Sarcomas 4 232 59 25.4 1.482 (1.269–1.730) <0.001∗∗ 0.0
 TGCMT # 1 40 22 55.0 3.000 (0.984–9.142) 0.053 NA
Metastatic/primary tumours
 Metastatic tumour 1 234 106 45.3 6.210 (2.882–13.381) <0.001∗∗ NA
 Primary tumour 21 2828 638 22.6 1.479 (1.333–1.641) <0.001∗∗ 66.3
VM assay methods
 PAS staining 5 1121 278 24.8 2.001 (1.218–3.288) 0.006∗∗ 94.2
 CD31/PAS staining 10 1088 227 20.9 1.360 (1.218–1.517) <0.001∗∗ 30.7
 CD34/PAS staining 7 741 198 26.7 1.493 (1.291–1.728) <0.001∗∗ 30.2
 Pan-cytokeratin/CD34 1 112 41 36.6 1.475 (0.987–2.205) 0.058 NA
 Asian 19 2653 593 22.4 1.411 (1.291–1.543) 0.005∗∗ 81.5
 Non-Asian ## 3 409 151 36.9 3.403 (1.435–8.069) <0.001∗∗ 51.4

# BDMT: bi-directional differentiated malignant tumour; HCC: hepatocellular carcinomas; NSCLC: non-small cell lung cancer; OLSCC: oral/laryngeal squamous cell carcinoma; TGCMT: testicular germ cell malignant tumour.

## Non-Asian: including one study in the United States and two studies in the Netherlands. NA: not available due to only one study.

P < 0.05, ∗∗P < 0.01: the relative risk of 5-year survival of VM-positive cancer patients compared with VM-negative cancer patients.

2.3. Statistical analyses

The meta-analysis of the impact of VM on the 5-year survival of cancer patients was performed using Stata 12.0 software. The individual hazard ratio (HR) estimates in each type of cancer were combined into an overall relative risk (RR) using the Peto’s method, which consisted of a fixed-effects model and was tested by performing Cochrane’s Q-tests for heterogeneity, and a P value greater than 0.10 for the Cochrane’s Q-test indicated a lack of heterogeneity among studies using I2 and respective 95% confidence intervals (CIs). I2 values ⩾50% or ⩾75% indicated large or very large heterogeneity, respectively. Otherwise, the random-effects model was used in the meta-analysis. By convention, an observed RR > 1 implies a worse survival of cancer patients. The overall pooled RR was estimated by calculating the weighted average of an individual study of log RR, with weights proportional to the inverse of the variances of the study-specific log RR estimation. Survival rates on the graphical representation of the survival curves were read by Engauge Digitizer version 2.5.

In addition, we analysed the influence of VM on the 5-year survival of individual cancer types, different VM detection methods, different populations and metastasis, which were treated as the corresponding subgroups. Study estimates, along with pooled estimates, are presented as forest plots. The effect of publication bias on the reported outcomes was assessed graphically using funnel plots and empirically using regression tests according to the method reported by Egger et al. [29] and [30] and P < 0.05 was deemed a statistically significant publication bias.

3. Results

3.1. Literature search in databases

We initially identified 609 potential papers using the key words “vasculogenic mimicry” in the four databases mentioned above; however, 480 of the 609 papers were excluded because they were irrelevant to VM in human cancer patients. Among the remaining 129 articles regarding VM in human cancers, 92 were excluded due to the absence of clinical cancer prognostic data. The remaining 37 studies pertinent to VM in human cancers and the 5-year survival of the patients were evaluated using the strict standard of meta-analysis as aforementioned. Thus, 15 of the 37 studies were excluded for the following reasons: eight studies [31], [32], [33], [34], [35], [36], [37], and [38] published in the Chinese literature or meeting abstract used cancer cases and data similar to six studies announced in the English literature [23], [24], [39], [40], [41], and [42], indicating duplicate reports in bi-languages; two studies [27] and [43] examined endothelial cell-formed blood vessels, which did not meet the definition of VM, indicating tumour cell-lined vessels; three studies evaluated the correlation between morphological patterns of VM and the prognosis of cancer patients [44], [45], and [46]; however, they lacked data regarding VM and 5-year overall survival of cancer patients; additionally, two other studies [47] and [48] only demonstrated the median survival time of cancer patients and lacked 5-year survival data. Finally, 22 eligible studies [10], [20], [21], [23], [24], [25], [37], [39], [40], [41], [42], [49], [50], [51], [52], [53], [54], [55], [56], [57], [58], and [59] containing 25 datasets (one of the studies evaluated four types of cancer) were qualified to the standard of the meta-analysis. The selection steps and the reasons for exclusion are summarised in Fig. 1 .


Fig. 1 Selection of the studies with vasculogenic mimicry and 5-year survival of human cancer patients.

3.2. Characteristics of VM in cancer patients

Three thousand sixty-two cancer patients were included in the meta-analysis, comprising 122.5 cases per dataset on average among 22 studies, ranging from 37 to 331 patients. The characteristics of the 25 datasets in the 22 studies are outlined in Table 1 and are shown in detail in Supplemental Table 2 as well as in Supplemental Figs. 1 and 2 . Briefly, 21 of the 22 studies investigated VM in 2828 primary cancer cases, and one study examined VM in 234 metastatic melanoma patients. The 22 studies covered 15 types of malignant tumours, including osteosarcoma, melanoma, hepatocellular carcinoma, gastric cancer, oral/laryngeal squamous cell carcinoma, colorectal cancer, breast cancer, ovarian tumours, testicular germ cell malignant tumour, glioma, gallbladder carcinoma, prostate cancer, medulloblastoma and non-small cell lung cancer. In the subgroups of ethnicity among the 22 studies, 19 studies were performed in Asian descendants, and three were performed in non-Asian descendants, including one study in the United States [10] , and two studies in the Netherlands [20] and [50]. PAS staining, a commonly used method for identification of VM was used to detect tumour VM in paraffin-embedded tumour tissue specimens (24 of 25 datasets). In addition, PAS staining was also combined with either CD31-staining (10 datasets) or CD34 staining (seven datasets) to distinguish tumour cell-lined blood vessels (PAS-positive, CD31- and/or CD34-negative) from vascular endothelial cell-formed vessels (CD31-positive, PAS-negative). In one study [52] , the authors used pan-cytokeratin as a tumour marker plus CD34 staining to assess tumour VM.

3.3. Outcome of the meta-analysis

The meta-analysis of 3062 cancer patients showed that the rate ratio of the overall 5-year survival was 31% [95% confidence interval (CI): 26–36%] in VM-positive cancer patients ( Supplemental Fig. 1 ) and 56% (95% CI: 45–67%) in VM-negative cancer patients ( Supplemental Fig. 2 ), respectively. Notably, VM-positive cancer patients showed a very significant high RR of 5-year survival in comparison with VM-negative patients (RR = 1.531; 95% CI: 1.357–1.726; P < 0.001; Table 1 and Fig. 2 ).


Fig. 2 The impact of vasculogenic mimicry on the prognosis of cancer patients. The relative risk (RR) of 5-year survival of VM-positive cancer patients was compared with VM-negative cancer patients. Individual study is shown in the square with grey colour, and the pooled 25 datasets of 22 studies are shown in the diamond, representing the 95% confidence interval (CI) of all 22 studies. RR > 1 implied a worse survival of the cancer patients. The size of each investigation represents the weighting factor (1/SE2) assigned to the study.

Next, we analysed the heterogeneity and publication biases in the current investigation. Subgroup analysis showed that metastatic melanoma patients showed a 45.3% VM-positive rate ( Table 1 and Supplemental Fig. 5 ), whereas primary melanoma patients showed a 23.1% VM-positive rate ( Table 2 and Supplemental Fig. 5 ). Subgroup analysis also showed that the relative risk of 5-year survival of metastatic melanoma patients (RR = 6.210; 95% CI: 2.882–13.381; P < 0.001) was higher than that of primary melanoma patients (RR = 1.620; 95% CI: 1.103–2.380; P = 0.014). Apparently, metastatic melanoma showed a higher VM rate and poorer prognosis compared with primary melanoma.

Table 2 The relative risk of 5-year survival in the subgroups of melanoma patients.

Study contents Studies (n) Tissues (n) VM+ (n) VM+ (%) RR (95% confidence interval (CI)) P I2 (%)
Total studies 4 606 192 31.7% 2.383 (1.084–5.238) 0.031 93.4
 Metastatic melanoma 1 234 106 45.3% 6.210 (2.882–13.381) <0.001∗∗ NA
 Primary melanoma 3 372 86 23.1% 1.620 (1.103–2.380) 0.014 70.6
VM detection
 PAS 3 416 182 43.8% 2.987 (1.136–7.853) 0.026 84.2
 CD31 and PAS 1 190 10 5.3% 1.377 (1.172–1.618) <0.001∗∗ NA
 Asian 2 314 64 0.204 1.467 (1.139–1.891) 0.003∗∗ 53.6
 Non-Asian 2 292 128 0.438 4.516 (2.160–9.444) <0.001∗∗ 32.4

NA: Not available due to single study; relative risk (RR) > 1 implies a poor prognosis of the melanoma patients.

P < 0.05, ∗∗P < 0.01, the VM-positive cancer patients compared to the VM-negative melanoma patients.

Subgroup analysis demonstrated that not every VM-positive cancer type was related to the 5-year overall survival. A significantly poor 5-year survival was observed in eight types of VM-positive cancer ( Table 1 , Supplemental Table 2 , and Supplemental Figs. 1 and 2 ), including sarcomas (P < 0.001), melanoma (P = 0.031), hepatocellular carcinoma (P < 0.001), oral/laryngeal squamous cell carcinoma (P = 0.035), colorectal cancer (P = 0.047), gallbladder carcinoma (P = 0.023), non-small cell lung cancer (P < 0.001) and bi-directional differentiated malignant tumours (P < 0.001); however, no statistically significant difference was found in the 5-year overall survival between VM-negative and -positive cancer patients of the other seven types of cancer, including breast cancer (P = 0.155), gastric cancer (P = 0.19), ovarian tumours (P = 0.651), testicular germ cell malignant tumours (P = 0.053), gliomas (P = 0.23), prostate cancer (P = 0.432) and medulloblastoma (P = 0.051), although the survival rate in VM-positive patients of these seven cancers was lower than that of VM-negative patients ( Table 1 , Supplemental Table 2 and Supplemental Figs. 1 and 2 ). As shown the data did approach statistical significance in two of the seven tumours such as testicular germ cell malignant tumours (P = 0.053) and medulloblastoma (P = 0.051).

The relative risk analysis of different populations demonstrated a very significant association of VM with a high risk of 5-year survival in both Asian populations (2653 cases; RR = 1.411; 95% CI: 1.291–1.543; P = 0.005) and non-Asian populations [409 cases from the United States and the Netherlands; RR = 3.403; 95% CI: 1.435–8.069; P < 0.001 ( Supplemental Fig. 3 and Supplemental Table 2 )], indicating that all the populations of VM-positive cancer patients demonstrate a poor prognosis.

In addition, analysis of the subgroups using different VM detection methods showed a poor 5-year survival in the PAS-positive staining subgroup (RR = 2.001; 95% CI: 1.218–3.288; P = 0.006), CD31-negative/PAS-positive staining subgroup (RR = 1.360; 95% CI: 1.218–1.517; P < 0.001) and CD34-negative/PAS-positive staining subgroup (RR = 1.493; 95% CI: 1.291–1.728; P < 0.001; Supplemental Fig. 4 and Supplemental Table 2 ) compared with the poor 5-year survival in VM-negative cancer patients but not in the pan-cytokeratin-positive/CD34-negative staining subgroup (RR = 1.475; 95% CI: 0.987–2.205; P = 0.058; Supplemental Fig. 4 and Supplemental Table 2 ). These data confirm that PAS-positive staining is a golden standard for the detection of VM in cancers.

The publication bias examination by Egger’s test showed that a publication bias (t = −5.66, P < 0.001) existed among the 22 studies ( Fig. 3 A). However, no publication bias was observed in 18 of 22 studies if the five studies [20], [24], [25], [40], and [41] with an extremely low 5-year survival rate were excluded (t = −3.48; P = 0.161; Fig. 3 B). The main factor related to publication bias was the extraordinary low 5-year survival rate (0–6.8%) in the five studies, including that for sarcoma (0%) [25] , bi-directional differentiated malignant carcinoma (0%) [40] , colorectal cancer (0%) [20] , gallbladder cancer (0%) [41] and non-small cell lung cancer (6.8%) [24] ( Table 1 and Supplemental Table 2 ), findings that are much lower compared with the mean 5-year survival rate of 24.3% in the 22 studies as a whole ( Table 1 and Supplemental Table 2 ), implying that these five types of VM-positive cancer patients have a very poor prognosis.


Fig. 3 The heterogeneity and publication biases in the Meta-analysis. The heterogeneity and publication biases were analysed using Funnel plots and empirically utilising regression tests according to the method reported by Egger et al. The standard error of log relative risk (S.E. of log RR) for 5-year survival was plotted against log RR in every individual study as shown in a circle. The 25 datasets of 22 studies were plotted in the funnel plots ( Fig. 3 A). After the seven studies with an extremely low 5-year survival rate were excluded, there was no publication bias among the remaining 15 studies as shown in Fig. 3 B (t = −3.48, P = 0.161).

4. Discussion

The tumour vasculature is heterogeneous and has distinct morphological characteristics from normal blood vessels. Vasculogenic tumour cells adopt a way of embryonic vasculogenesis and directly form primitive, immature blood vessels consisting of various capillary-like structures, tubes and networks [60] and [61]; notably, the inner layer of tumour cell-lined blood vessels comprises extracellular matrix (ECM) proteins, such as laminin, collagens IV and VI and heparin sulphate proteoglycans, which are absent in endothelial cells [60] and [61]. When tumour cell-formed vessels are stained using PAS- and CD31-staining methods, the vessels show pinch colour (PAS-positive), while CD31-staining is negative; thus, PAS-positive and CD31-negative characteristics become the golden standard for tumour cell-dominant VM [23], [25], [39], [42], [49], [51], and [55]. Alternatively, PAS-positive and CD34-negative staining were also used to indicate tumour cell-mediated VM [21], [24], [53], [54], [56], [57], and [58]. Since Hendrix and colleagues raised a new concept of tumour VM in 1999 [10] , VM has been found in various types of malignant tumours [20], [25], [28], [39], [40], [49], [50], [52], [55], and [59]. However, scientists have divergent views regarding whether VM affects the prognosis of cancer patients or not. Several studies have indicated that VM was significantly associated with a poor 5-year survival of cancer patients [10], [21], [23], [40], [41], [42], [51], and [58]. By contrast, other studies have shown no significant correlation between VM and the 5-year survival of cancer patients [25], [26], [27], and [28]. Our meta-analysis results showed that VM was very significantly associated with a high RR of a 5-year overall survival in 22 studies with 3062 cancer patients, supporting the notion that VM plays important roles in tumour growth, progression, metastasis and has a negative impact on the long-term survival of cancer patients.

Our meta-analysis showed that metastatic melanoma patients had a higher VM rate (45.3%) than primary melanoma cases (23.1%) [10] , and metastatic melanoma possesses the high risk of a 5-year survival compared with primary melanoma ( Tables 1 , 2 and Supplemental Fig. 5 ), implying that VM contributes to tumour metastasis and poor prognosis. It is well known that metastasis is closely related to the high poor survival of cancer patients, and tumour cell-lined blood vessels can provide oxygen and nutrition supplies to tumour cells and offer channels to facilitate cancer cell migration and tumour metastasis [10] . Thus far, only one study with 234 metastatic melanoma cases investigated VM and cancer prognosis. Hence, the impact of VM on the prognosis of various other metastatic cancers is warranted. These studies will likely provide valuable information for the diagnosis and treatment of metastatic cancers.

Analysis of tumour subtypes showed that VM is significantly associated with a poor 5-year survival in eight out of 15 types of cancer as mentioned above. We noted that only a single study relevant to VM and 5-year survival was conducted in the seven types of cancer, including breast cancer, gastric cancer, ovarian tumours, testicular germ cell malignant tumours, gliomas, prostate cancer and medulloblastoma. Although the 5-year overall survival rate in VM-positive cancer patients was less than that in VM-negative cancer cases, the difference was statistically insignificant. In addition, small patient samples with a large variation in the 5-year survival in patients with ovarian tumours, testicular germ cell malignant tumours, gliomas, prostate cancer and medulloblastoma may also account for failure to reach statistical significance in five out of seven types of malignant tumours, suggesting more studies are needed to verify the impact of VM on the prognosis of these cancer patients. Notably, all the patients in the seven studies have primary tumours. Considering that metastatic melanoma has a high VM rate and poor prognosis ( Table 1 and Supplemental Fig. 5 ), the influence of VM on the prognosis of the corresponding seven types of metastatic cancers needs to be further investigated.

VM is significantly related with a poor 5-year survival in eight types of high malignant tumours, including lung cancer, colorectal cancer, hepatic cancer, melanoma, sarcomas, oral/laryngeal squamous cell carcinoma, gallbladder carcinoma, and bi-directional differentiated malignant tumours. However, VM is not significantly associated with the 5-year overall survival of the other seven types of cancer with relative moderate malignancy, including breast cancer, ovarian tumours, prostate cancer, gastric cancer, testicular germ cell malignant tumours, gliomas and medulloblastoma. The reasons for the divergent effect of VM on the prognosis of cancer patients are unclear. Malignant tumours are heterogeneous in morphology, varying in location, differing in malignancy and contrasting in metastasis. Notably, cancer metastasis accounts for approximately 90% death of all the cancer patients and is closely connected with the poor 5-year overall survival of cancer patients [2], [42], [51], [62], and [63].

Accordingly, we speculate several possibilities underlying the dissimilarity of VM relevant to the prognosis of cancers. First, tumour cell-dominant VM in high malignant tumours, such as lung cancer, colorectal cancer, hepatic cancer, melanoma and sarcomas, is more robust than that in relative moderate malignant tumours, such as breast cancer, ovarian tumours, prostate cancer and gastric cancer. Second, various oncogenes and VM master genes, including c-Myc [64] and [65], Twist1 [66] , mutant von Hippel-Lindau tumour suppressor gene [67] , mutant fibronectin ED-B [68] and [69], VE-cadherin (CDH5) [70], [71], [72], and [73], Nodal [74], [75], [76], and [77], hypoxia-inducible factor (HIF)-1alpha [78] and several cancer stem cell markers [79] , are overexpressed in the high malignant tumours and are responsible for the initiation of a vigorous VM in the cancers. Importantly, tumour cell-lined blood vessels not only supply oxygen and nutrients for a rapid tumour growth, but also provide a channel for tumour cell dissemination; in particular, the vasculogenic tumour cells in tumour cell-lined blood vessels directly contact with circular blood and can easily leave the vascular wall and migrate to distant tissues along with the blood flow, resulting in cancer metastasis. Third, in the high malignant tumours, the key VM driver genes mentioned above activate several key signalling pathways and trigger a sturdy VM [70], [71], [72], and [73]. For example, c-Myc, Twist1 and HIF-1alpha induce overexpression of cell surface adhesive molecule VE-cadherin, a master protein of VM. VE-cadherin not only acts as a bridge for tumour cells to connect and form blood vessels, but also directly binds and forms complexes with vascular endothelial growth factor receptor 2 (VEGFR2), EphA2, β-catenin and several other proteins, and enhances VEGF–VEGFR2 signalling, phosphoinositide 3-kinase (PI3K)-Akt signalling and β-catenin-mediated gene transcription, leading to a strong tumour cell-mediated vascularity, tumour growth and cancer metastasis [70], [71], [72], and [73]. For another instance, an embryonic protein nodal is overexpressed in melanoma and many other high malignant tumours and participates in Notch and mitogen-activated protein kinase (MAPK)-dependent signalling pathways. Nodal overexpression and activation of Nodal signalling pathways promote tumour VM, growth and dissemination [74], [75], [76], and [77]. Collectively, in high malignant tumours, these VM drivers and triggers initiate a robust VM, which promotes tumour growth, progression and metastasis, resulting in the poor prognosis of the cancer patients.

Although publication bias existed in the meta-analysis, the cause may be due to an extremely low 5-year survival rate in non-small cell lung cancer, sarcomas, melanoma, bi-directional differentiated malignant carcinoma, colorectal cancer and gallbladder cancer [20], [24], [25], [39], and [40]. In addition, three studies [20], [25], and [39] with small sample size (less than 100 cases per study) may also contribute to the publication bias; the weight of these three studies on the quality of the meta-analysis is light because they only account for 5.8% of the cancer patients analysed in the investigation. Nonetheless, the publication bias analysis indicates a profound effect of VM on the 5-year survival in the six types of malignant tumours mentioned above, implying that VM is a useful indicator of these malignant tumours.

Recent and rapid progresses [1] and [4] in tumour cell-mediated angiogenesis or neovascularisation, particularly VM and trans-differentiation of cancer stem-like cells into tumour endothelial cells have broadened the landscape of tumour angiogenesis and brought new opportunities for novel anti-angiogenic drug discovery and anti-cancer therapy. Since Folkman proposed the new concept of anti-tumour angiogenesis in 1971, five anti-angiogenic drugs have been used clinically and are promising for anti-cancer therapy [1] and [4]; however, these anti-angiogenic drugs have recently demonstrated a short-lived anti-cancer effect in several clinical trials and an absence of long-term survival benefits [1] and [4]. A recent meta-analysis of the anti-angiogenic drug Avastin indicated that the drug had a significant survival gain only in a subtype of colon cancer patients, but not in several other subtypes of colon cancer patients [29] . One of the fundamental reasons for the moderate anti-tumour effect might be related to current anti-angiogenic therapies not effectively inhibiting tumour cell-mediated VM and other forms of neovascularisation. Accumulated data have shown that vasculogenic tumour cells play critical roles in tumour neovascularisation, growth and metastasis, and contribute to the poor prognosis of cancer patients [10] . Hence, vasculogenic tumour cells should be targeted for novel anti-cancer drug discovery. Targeting vasculogenic tumour cells may have the effect like “one stone hits two birds” by impeding tumour cell-dominant VM and hindering tumour cell-stimulated angiogenesis because tumour cells produce various growth factors to promote ingrowth of endothelial cells in tumour tissues to form blood vessels. Because molecular and cellular mechanisms of tumour cell-dominant VM are distinct from endothelial cell-mediated sprouting angiogenesis, the combination of anti-tumour VM agents with anti-angiogenic drugs may produce a synergistic effect, and thus may improve the efficacy of anti-cancer therapy.

Conflict of interest statement

None declared.


This work was supported by grants from National Natural Science Foundation of China (Grants No. 30971138, 81172087 and 81071306), Suzhou City Scientific Research Funds (No. SS201004 and SS201138), a project funded by the priority academic program development of Jiangsu Higher Education Institutions (PAPD), Cultivation base of State Key Laboratory of Stem Cell and Biomaterials built together by Ministry of Science and Technology, Research and Innovation Project for College Graduates of Jiangsu Province (CXZZ13_0824) and Jiangsu Province’s Key Discipline of Medicine (XK201118).

Appendix A. Supplementary data


Download file

Supplemental Figures This PDF file contains Supplemental Figs. 1–5.

Download file

Supplemental Table 1 Computer-assisted search strategy and database.

Download file

Supplemental Table 2 Main characteristics and datasets from 22 eligible studies.


  • [1] P. Carmeliet, R.K. Jain. Molecular mechanisms and clinical applications of angiogenesis. Nature. 2011;473:298-307 Crossref
  • [2] D. Hanahan, R.A. Weinberg. Hallmarks of cancer: the next generation. Cell. 2011;144:646-674 Crossref
  • [3] D. Hanahan, R.A. Weinberg. The hallmarks of cancer. Cell. 2000;100:57-70 Crossref
  • [4] M. Potente, H. Gerhardt, P. Carmeliet. Basic and therapeutic aspects of angiogenesis. Cell. 2011;146:873-887 Crossref
  • [5] L. Ricci-Vitiani, R. Pallini, M. Biffoni, et al. Tumour vascularization via endothelial differentiation of glioblastoma stem-like cells. Nature. 2010;468:824-828 Crossref
  • [6] R. Wang, K. Chadalavada, J. Wilshire, et al. Glioblastoma stem-like cells give rise to tumour endothelium. Nature. 2010;468:829-833 Crossref
  • [7] B. Shang, Z. Cao, Q. Zhou. Progress in tumor vascular normalization for anticancer therapy: challenges and perspectives. Front Med. 2012;6:67-78 Crossref
  • [8] M.J. Hendrix, E.A. Seftor, P.S. Meltzer, et al. Expression and functional significance of VE-cadherin in aggressive human melanoma cells: role in vasculogenic mimicry. Proc Natl Acad Sci USA. 2001;98:8018-8023 Crossref
  • [9] A.R. Hess, N.V. Margaryan, E.A. Seftor, M.J. Hendrix. Deciphering the signaling events that promote melanoma tumor cell vasculogenic mimicry and their link to embryonic vasculogenesis: role of the Eph receptors. Dev Dyn. 2007;236:3283-3296 Crossref
  • [10] A.J. Maniotis, R. Folberg, A. Hess, et al. Vascular channel formation by human melanoma cells in vivo and in vitro: vasculogenic mimicry. Am J Pathol. 1999;155:739-752 Crossref
  • [11] A.K. Sood, E.A. Seftor, M.S. Fletcher, et al. Molecular determinants of ovarian cancer plasticity. Am J Pathol. 2001;158:1279-1288 Crossref
  • [12] J. Mei, Y.W. Jia, X.S. Cai. Vascular channel formation by osteosarcoma cells in vitro: vasculogenic mimicry. Chin-German J Clin Oncol. 2003;2:237-239
  • [13] D.W. van der Schaft, R.E. Seftor, E.A. Seftor, et al. Effects of angiogenesis inhibitors on vascular network formation by human endothelial and melanoma cells. J Natl Cancer Inst. 2004;96:1473-1477 Crossref
  • [14] Y.W. Paulis, P.M. Soetekouw, H.M. Verheul, V.C. Tjan-Heijnen, A.W. Griffioen. Signalling pathways in vasculogenic mimicry. Biochim Biophys Acta. 2010;1806:18-28 Crossref
  • [15] Z. Cao, D. Yu, S. Fu, et al. Lycorine hydrochloride selectively inhibits human ovarian cancer cell proliferation and tumor neovascularization with very low toxicity. Toxicol Lett. 2013;218:174-185 Crossref
  • [16] K. Shirakawa, H. Kobayashi, Y. Heike, et al. Hemodynamics in vasculogenic mimicry and angiogenesis of inflammatory breast cancer xenograft. Cancer Res. 2002;62:560-566
  • [17] N. Sharma, R.E. Seftor, E.A. Seftor, et al. Prostatic tumor cell plasticity involves cooperative interactions of distinct phenotypic subpopulations: role in vasculogenic mimicry. Prostate. 2002;50:189-201 Crossref
  • [18] X.S. Cai, Y.W. Jia, J. Mei, R.Y. Tang. Tumor blood vessels formation in osteosarcoma: vasculogenesis mimicry. Chin Med J (Engl). 2004;117:94-98
  • [19] A. Fujimoto, H. Onodera, A. Mori, S. Nagayama, Y. Yonenaga, T. Tachibana. Tumour plasticity and extravascular circulation in ECV304 human bladder carcinoma cells. Anticancer Res. 2006;26:59-69
  • [20] C.I. Baeten, F. Hillen, P. Pauwels, A.P. de Bruine, C.G. Baeten. Prognostic role of vasculogenic mimicry in colorectal cancer. Dis Colon Rectum. 2009;52:2028-2035 Crossref
  • [21] W.B. Liu, G.L. Xu, W.D. Jia, et al. Prognostic significance and mechanisms of patterned matrix vasculogenic mimicry in hepatocellular carcinoma. Med Oncol. 2011;28(Suppl. 1):S228-S238
  • [22] J. Jiang, W. Liu, X. Guo, et al. IRX1 influences peritoneal spreading and metastasis via inhibiting BDKRB2-dependent neovascularization on gastric cancer. Oncogene. 2011;30:4498-4508 Crossref
  • [23] M. Li, Y. Gu, Z. Zhang, et al. Vasculogenic mimicry: a new prognostic sign of gastric adenocarcinoma. Pathol Oncol Res. 2010;16:259-266 Crossref
  • [24] S. Wu, L. Yu, Z. Cheng, W. Song, L. Zhou, Y. Tao. Expression of maspin in non-small cell lung cancer and its relationship to vasculogenic mimicry. J Huazhong Univ Sci Technolog Med Sci. 2012;32:346-352 Crossref
  • [25] B. Sun, S. Zhang, X. Zhao, W. Zhang, X. Hao. Vasculogenic mimicry is associated with poor survival in patients with mesothelial sarcomas and alveolar rhabdomyosarcomas. Int J Oncol. 2004;25:1609-1614
  • [26] D. Massi, A. Franchi, M. Paglierani, et al. Vasculogenic mimicry has no prognostic significance in pT3 and pT4 cutaneous melanoma. Hum Pathol. 2004;35:496-502 Crossref
  • [27] L. Pignataro, N. Carboni, V. Midolo, et al. Clinical relevance of microvessel density in laryngeal squamous cell carcinomas. Int J Cancer. 2001;92:666-670 Crossref
  • [28] K. Shirakawa, H. Wakasugi, Y. Heike, et al. Vasculogenic mimicry and pseudo-comedo formation in breast cancer. Int J Cancer. 2002;99:821-828 Crossref
  • [29] L.T. Macedo, A.B. da Costa, A.D. Sasse. Addition of bevacizumab to first-line chemotherapy in advanced colorectal cancer: a systematic review and meta-analysis, with emphasis on chemotherapy subgroups. BMC Cancer. 2012;12:89 Crossref
  • [30] J.T. Qu, M. Wang, H.L. He, Y. Tang, X.J. Ye. The prognostic value of elevated vascular endothelial growth factor in patients with osteosarcoma: a meta-analysis and systemic review. J Cancer Res Clin Oncol. 2012;138:819-825 Crossref
  • [31] Y.Z. Fan, W. Sun, W.Z. Zhang, C.Y. Ge. Vasculogenic mimicry in human primary gallbladder carcinoma and clinical significance thereof. Zhonghua Yi Xue Za Zhi. 2007;87:145-149
  • [32] X. Hao, B. Sun, S. Zhang, X. Zhao. Microarray study of vasculogenic mimicry in bi-directional differentiation malignant tumor. Zhonghua Yi Xue Za Zhi. 2002;82:1298-1302
  • [33] X.S. Hao, B.C. Sun, S.W. Zhang, X.L. Zhao. Correlation between the expression of collgen IV, VEGF and vasculogenic mimicry. Zhonghua Zhong Liu Za Zhi. 2003;25:524-526 Crossref
  • [34] M. Li, Y. Gu, D. Zhang, B. Sun, Z. Zhang, X. Zhao. Mechanism of vasculogenic mimicry and its clinicopathologic significance in gastric adenocarcinoma. Chin J Clin Oncol. 2010;37:372-376
  • [35] W. Wang, P. Lin, B.C. Sun, et al. Role of vasculogenic mimicry and endothelium-dependent vessel in metastasis of laryngeal cancer. Zhonghua Er Bi Yan Hou Tou Jing Wai Ke Za Zhi. 2012;47:400-405
  • [36] S. Wu, Z. Cheng, L. Yu, W. Song, Y. Tao. Expression of CD82/KAI1 and HIF-1alpha in non-small cell lung cancer and their relationship to vasculogenic mimicry. Zhongguo Fei Ai Za Zhi. 2011;14:918-925
  • [37] X.L. Zhao, J. Du, S.W. Zhang, Y.X. Liu, X. Wang, B.C. Sun. A study on vasculogenic mimicry in hepatocellular carcinoma. Zhonghua Gan Zang Bing Za Zhi. 2006;14:41-44
  • [38] B.C. Sun, S.W. Zhang, X.L. Zhao, X.S. Hao. Vasculogenic mimicry is associated with shorter survival in hepatocellular carcinomas. Lab Invest. 2006;86:1302
  • [39] B. Sun, S. Zhang, D. Zhang, et al. Vasculogenic mimicry is associated with high tumor grade, invasion and metastasis, and short survival in patients with hepatocellular carcinoma. Oncol Rep. 2006;16:693-698
  • [40] B. Sun, S. Zhang, X. Zhao, X. Hao, D. Zhang. Pilot study of molecular mechanism on vasculogenic mimicry in bi-directional differentiated malignant tumors. Chin-German J Clinic Oncol. 2005;4:50-52 Crossref
  • [41] W. Sun, Z.Y. Shen, H. Zhang, et al. Overexpression of HIF-1alpha in primary gallbladder carcinoma and its relation to vasculogenic mimicry and unfavourable prognosis. Oncol Rep. 2012;27:1990-2002
  • [42] W. Wang, P. Lin, C. Han, W. Cai, X. Zhao, B. Sun. Vasculogenic mimicry contributes to lymph node metastasis of laryngeal squamous cell carcinoma. J Exp Clin Cancer Res. 2010;29:60 Crossref
  • [43] A.K. Sood, M.S. Fletcher, C.M. Zahn, et al. The clinical significance of tumor cell-lined vasculature in ovarian carcinoma: implications for anti-vasculogenic therapy. Cancer Biol Ther. 2002;1:661-664
  • [44] J.T. Buijs, A.M. Cleton, V.T. Smit, C.W. Lowik, S. Papapoulos, G.v. Pluijm. Prognostic significance of periodic acid-Schiff-positive patterns in primary breast cancer and its lymph node metastases. Breast Cancer Res Treat. 2004;84:117-130 Crossref
  • [45] A.A. Vartanian, E.V. Stepanova, S.L. Gutorov, et al. Prognostic significance of periodic acid-Schiff-positive patterns in clear cell renal cell carcinoma. Can J Urol. 2009;16:4726-4732
  • [46] M.A. Warso, A.J. Maniotis, X. Chen, et al. Prognostic significance of periodic acid-Schiff-positive patterns in primary cutaneous melanoma. Clin Cancer Res. 2001;7:473-477
  • [47] C. Han, B. Sun, W. Wang, et al. Overexpression of microtubule-associated protein-1 light chain 3 is associated with melanoma metastasis and vasculogenic mimicry. Tohoku J Exp Med. 2011;223:243-251 Crossref
  • [48] Z. Liu, Y. Li, W. Zhao, Y. Ma, X. Yang. Demonstration of vasculogenic mimicry in astrocytomas and effects of Endostar on U251 cells. Pathol Res Pract. 2011;207:645-651 Crossref
  • [49] Y. Gao, X.L. Zhao, Q. Gu, et al. Correlation of vasculogenic mimicry with clinicopathologic features and prognosis of ovarian carcinoma. Zhonghua Bing Li Xue Za Zhi. 2009;38:585-589
  • [50] F. Hillen, C.I. Baeten, A. van de Winkel, et al. Leukocyte infiltration and tumor cell plasticity are parameters of aggressiveness in primary cutaneous melanoma. Cancer Immunol Immunother. 2008;57:97-106 Crossref
  • [51] R. Liu, K. Yang, C. Meng, Z. Zhang, Y. Xu. Vasculogenic mimicry is a marker of poor prognosis in prostate cancer. Cancer Biol Ther. 2012;13:527-533 Crossref
  • [52] S.Y. Liu, L.C. Chang, L.F. Pan, Y.J. Hung, C.H. Lee, Y.S. Shieh. Clinicopathologic significance of tumor cell-lined vessel and microenvironment in oral squamous cell carcinoma. Oral Oncol. 2008;44:277-285 Crossref
  • [53] X.M. Liu, Q.P. Zhang, Y.G. Mu, et al. Clinical significance of vasculogenic mimicry in human gliomas. J Neurooncol. 2011;105:173-179 Crossref
  • [54] Z. Liu, B. Sun, L. Qi, H. Li, J. Gao, X. Leng. Zinc finger E-box binding homeobox 1 promotes vasculogenic mimicry in colorectal cancer through induction of epithelial-to-mesenchymal transition. Cancer Sci. 2012;103:813-820 Crossref
  • [55] B. Sun, S. Qie, S. Zhang, et al. Role and mechanism of vasculogenic mimicry in gastrointestinal stromal tumors. Hum Pathol. 2008;39:444-451 Crossref
  • [56] L. Wang, Y. Gu, S. Zhang, B. Sun. Pilot study and clinical prognosis significance of vasculogenic mimicry in leiomyosarcoma. Tianjin Med J. 2009;37:161-165
  • [57] S.Y. Wang, L. Yu, G.Q. Ling, et al. Vasculogenic mimicry and its clinical significance in medulloblastoma. Cancer Biol Ther. 2012;13:341-348 Crossref
  • [58] X.Y. Wang, Y. Gu, S. Zhang, B.S. Zhao. Vasculogenic mimicry in testicular germ cell malignant tumor and its significance for prognosis. Chin Tumor Clinic. 2009;36:78-83
  • [59] S. Zhang, M. Li, D. Zhang, et al. Hypoxia influences linearly patterned programmed cell necrosis and tumor blood supply patterns formation in melanoma. Lab Invest. 2009;89:575-586 Crossref
  • [60] D.A. Kirschmann, E.A. Seftor, K.M. Hardy, R.E. Seftor, M.J. Hendrix. Molecular pathways: vasculogenic mimicry in tumor cells: diagnostic and therapeutic implications. Clin Cancer Res. 2012;18:2726-2732 Crossref
  • [61] R.E. Seftor, A.R. Hess, E.A. Seftor, et al. Tumor cell vasculogenic mimicry: from controversy to therapeutic promise. Am J Pathol. 2012;181:1115-1125 Crossref
  • [62] F. van Zijl, G. Krupitza, W. Mikulits. Initial steps of metastasis: cell invasion and endothelial transmigration. Mutat Res. 2011;728:23-34 Crossref
  • [63] M.J. Hendrix, E.A. Seftor, A.R. Hess, R.E. Seftor. Vasculogenic mimicry and tumour-cell plasticity: lessons from melanoma. Nat Rev Cancer. 2003;3:411-421 Crossref
  • [64] C. Chen, S. Cai, G. Wang, et al. C-Myc enhances colon cancer cell-mediated angiogenesis through the regulation of HIF-1alpha. Biochem Biophys Res Commun. 2013;430:505-511 Crossref
  • [65] N.M. Sodir, L.B. Swigart, A.N. Karnezis, D. Hanahan, G.I. Evan, L. Soucek. Endogenous Myc maintains the tumor microenvironment. Genes Dev. 2011;25:907-916 Crossref
  • [66] T. Sun, N. Zhao, X.L. Zhao, et al. Expression and functional significance of Twist1 in hepatocellular carcinoma: its role in vasculogenic mimicry. Hepatology. 2010;51:545-556 Crossref
  • [67] M.T. Chou, J. Anthony, J.D. Bjorge, D.J. Fujita. The von Hippel-Lindau tumor suppressor protein is destabilized by Src: implications for tumor angiogenesis and progression. Genes Cancer. 2010;1:225-238 Crossref
  • [68] M. Midulla, R. Verma, M. Pignatelli, M.A. Ritter, N.S. Courtenay-Luck, A.J. George. Source of oncofetal ED-B-containing fibronectin: implications of production by both tumor and endothelial cells. Cancer Res. 2000;60:164-169
  • [69] E. Ventura, F. Sassi, A. Parodi, et al. Alternative splicing of the angiogenesis associated extra-domain B of fibronectin regulates the accessibility of the B-C loop of the type III repeat 8. PLoS One. 2010;5:e9145 Crossref
  • [70] L.Z. Zhang, J. Mei, Z.K. Qian, X.S. Cai, Y. Jiang, W.D. Huang. The role of VE-cadherin in osteosarcoma cells. Pathol Oncol Res. 2010;16:111-117 Crossref
  • [71] K.H. Ryu, K.N. Shim, S.A. Jung, K. Yoo, Y.H. Joo, J.H. Lee. Significance of preoperative tissue levels of vascular-endothelial cadherin, liver-intestine cadherin and vascular endothelial growth factor in gastric cancer. Korean J Gastroenterol. 2012;60:229-241 Crossref
  • [72] R. Liu, Z. Cao, J. Tu, et al. Lycorine hydrochloride inhibits metastatic melanoma cell-dominant vasculogenic mimicry. Pigment Cell Melanoma Res. 2012;25:630-638 Crossref
  • [73] R. Liu, Z. Cao, Y. Pan, et al. Jatrorrhizine hydrochloride inhibits the proliferation and neovascularization of C8161 metastatic melanoma cells. Anticancer Drugs. 2013;24:667-676 Crossref
  • [74] L.M. Postovit, N.V. Margaryan, E.A. Seftor, M.J. Hendrix. Role of nodal signaling and the microenvironment underlying melanoma plasticity. Pigment Cell Melanoma Res. 2008;21:348-357 Crossref
  • [75] J.C. McAllister, Q. Zhan, C. Weishaupt, M.Y. Hsu, G.F. Murphy. The embryonic morphogen, Nodal, is associated with channel-like structures in human malignant melanoma xenografts. J Cutan Pathol. 2010;37(Suppl. 1):19-25 Crossref
  • [76] K.M. Hardy, D.A. Kirschmann, E.A. Seftor, et al. Regulation of the embryonic morphogen Nodal by Notch4 facilitates manifestation of the aggressive melanoma phenotype. Cancer Res. 2010;70:10340-10350 Crossref
  • [77] D.F. Quail, G. Zhang, S.D. Findlay, D.A. Hess, L.M. Postovit. Nodal promotes invasive phenotypes via a mitogen-activated protein kinase-dependent pathway. Oncogene. 2013;10.1038/onc.2012.608
  • [78] G. Comito, M. Calvani, E. Giannoni, et al. HIF-1alpha stabilization by mitochondrial ROS promotes Met-dependent invasive growth and vasculogenic mimicry in melanoma cells. Free Radic Biol Med. 2011;51:893-904 Crossref
  • [79] K. Lirdprapamongkol, K. Chiablaem, M. Sila-Asna, R. Surarit, A. Bunyaratvej, J. Svasti. Exploring stemness gene expression and vasculogenic mimicry capacity in well- and poorly-differentiated hepatocellular carcinoma cell lines. Biochem Biophys Res Commun. 2012;422:429-435 Crossref


a Cyrus Tang Hematology Center, Jiangsu Institute of Hematology, First Affiliated Hospital of Soochow University, Key Laboratory of Thrombosis and Hemostasis, Ministry of Health, Suzhou, Jiangsu 215123, PR China

b University of Mississippi Cancer Institute, Jackson, MS 39216, USA

c Department of Pathology and Oncology, Karmanos Cancer Institute, Wayne State University, Detroit, MI 48201, USA

d Department of Pathology, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA 02215, USA

lowast Corresponding authors: Addresses: Cyrus Tang Hematology Center, Soochow University, Room 703-3505, 199 Ren Ai Road, Suzhou Industrial Park, Suzhou, Jiangsu 215123, PR China. Tel.: +86 512 65882116; fax: +86 512 65880929 (Q. Zhou). Department of Pathology, Beth Israel Deaconess Medical Center, Harvard Medical School, 330 Brookline Ave., Boston, MA 02215, USA. Tel.: +1 617 735 2474; fax: +1 617 735 2480 (Z. Wang).

1 These authors contributed equally to this work.