Role of FUT8 expression in clinicopathology and patient survival for various malignant tumor types: a systematic review and meta-analysis

Dysregulation of α(1,6)-fucosyltransferase (FUT8) plays significant roles in development of a variety of malignant tumor types. We collected as many relevant articles and microarray datasets as possible to assess the prognostic value of FUT8 expression in malignant tumors. For this purpose, we systematically searched PubMed, Embase, Web of Science, Springer, Chinese National Knowledge Infrastructure (CNKI), and Wan Fang, and eventually identified 7 articles and 35 microarray datasets (involving 6124 patients and 10 tumor types) for inclusion in meta-analysis. In each tumor type, FUT8 expression showed significant (p< 0.05) correlation with one or more clinicopathological parameters; these included patient gender, molecular subgroup, histological grade, TNM stage, estrogen receptor, progesterone receptor, and recurrence status. In regard to survival prognosis, FUT8 expression level was associated with overall survival in non-small cell lung cancer (NSCLC), breast cancer, diffuse large B cell lymphoma, gastric cancer, and glioma. FUT8 expression was also correlated with disease-free survival in NSCLC, breast cancer, and colorectal cancer, and with relapse-free survival in pancreatic ductal adenocarcinoma. For most tumor types, survival prognosis of patients with high FUT8 expression was related primarily to clinical features such as gender, tumor stage, age, and pathological category. Our systematic review and meta-analysis confirmed the association of FUT8 with clinicopathological features and patient survival rates for numerous malignant tumor types. Verification of prognostic value of FUT8 in these tumor types will require a large-scale study using standardized methods of detection and analysis.

Glycosylation (attachment of glycans to proteins or other organic molecules) is a common posttranslational modification in all organisms. Aberrant glycosylation is a characteristic phenomenon in carcinogenesis, plays essential roles in specific steps of tumor development [10], and directly promotes tumor progression and metastasis [11][12][13][14][15][16][17]. Glycosylation is mediated by enzymatic activities of glycosyltransferases and glycosidases in glycoproteins and/or lipids. Alterations of glycosyltransferase expression are associated with both pro-metastatic and metastasis-suppressing functions [12].
Fucosyltransferase 8 (FUT8) is an α(1,6)fucosyltransferase responsible for addition of fucose to asparagine-linked N-acetylglucosamine (GlcNAc) moieties, a common feature of N-linked glycan core structures [18]. Aberrant fucosylated glycan structures and associated glycosyltransferases have been observed in development of various cancers [19]; e.g., increased core fucosylation in BC [20], non-small cell lung cancer (NSCLC) [21], ovarian cancer [22], colon cancer [23], prostate cancer [24], and melanoma [25]. A positive feedback mechanism of FUT8-mediated receptor core fucosylation was recently shown to enhance TGF-β signaling and epithelial-mesenchymal transition (EMT), thus promoting BC cell invasion and metastasis; such fucosylation in BC patients is a potential diagnostic/ prognostic biomarker, or therapeutic target [26]. In BC patients, high FUT8 protein expression is correlated with lymphatic metastasis and stage status, whereas reduced FUT8 expression is correlated with disease-free survival and overall survival [20].
FUT8 expression level has been linked to tumor clinical features and outcomes in numerous studies. However, few attempts have been made to systematically evaluate such associations. We performed a systematic review and meta-analysis based on collection of references and Gene Expression Omnibus (GEO) microarray data, to clarify the correlations between FUT8 expression and clinical pathology and patient survival in various common types of cancer.

Characteristics of studies included in the metaanalysis
A total of 366 articles were found through database mining and manual searches (see M&M); these comprised 152 articles directly related to FUT8, and 214 articles that included microarray data and were indirectly related to FUT8. After screening out duplicate titles and abstracts, 27 full-text articles and 78 datasets remained; of these, 20 articles and 43 datasets were excluded on the basis of criteria described in M&M, finally leaving 7 articles and 35 microarray datasets (involving 6124 patients) for inclusion in the metaanalysis. Characteristics and quality scores of the included studies are summarized in Table 1. Among these, the 35 datasets involved 7 types of malignant tumors and descriptions of 29 types of clinicopathological features related to FUT8, and the 7 articles involved 5 types of malignant tumors and described correlations between FUT8 expression and patient survival.  Figure 2).   [54], [55], [56], [57], [58], [59], [60], [61], R2 platform Glioma

Prognostic value of FUT8 expression in survival of various types of malignant tumors
FUT8 expression levels in the other 4 tumor types (glioma, DLBCL, GC, BC) were also correlated with OS ( Figure 4). Among these, high FUT8 expression was associated with shorter OS in glioma (RR=

DISCUSSION
FUT8 is clearly involved in tumor initiation and progression, and in various biological behaviors of cancer, including cell proliferation, apoptosis, migration, and metastasis [27]. Prognostic values of glycosylation or fucosylated antigens in many types of cancer have been documented in several previous reviews and meta-analyses [14,15,[28][29][30]. However, very little meta-analysis of FUT8 has been performed, because of lack of sufficient studies and references. Many reports are based on use of microarrays to study cancer genomes; accordingly, we performed a systematic review after collecting GEO microarray data, and data from a large number of studies, focused on the prognostic value of FUT8 in cancer. This review / metaanalysis is, to our knowledge, the first to comprehensively clarify the association of FUT8 expression with specific clinicopathological features and survival data for various types of cancer.
Our analyses demonstrated that FUT8 expression levels were most often associated with tumor stage (n=4: BC, CRC, MBL, NSCLC), molecular classification (n=2: BC, MBL), age (n=2: glioma, ependymoma), and gender (n=2: MBL, NSCLC), but less associated with histological grade (n=1: BC) and pathological typing (n=1: glioma). FUT8 expression showed no association with disease induction, location, or family history for any of the above cancer types. FUT8 expression is therefore presumably enhanced in luminal A and B BC patients, and may be a useful survival prognostic predictor in such patients. A 2019 study indicated that FUT8 is highly expressed in ER+ BC patients, is associated with metastasis, and is a potential therapeutic target in these patients [35].
Microsatellite instability (MSI) is a molecular "fingerprint" of defective mismatch repair systems, and methods to detect MSI are well established and routinely incorporated into clinical practice. Prognosis for MSI tumors is better than that for microsatellite stable CRC [36]. In regard to CRC, we found that high FUT8 expression was associated with MSI, female gender, and TNM stages I-III. E-cadherin, a Ca 2+ -dependent cell adhesion molecule, was found to be significantly  enhanced in dense culture of FUT8-transfected colorectal adenocarcinoma cells, resulting in increased cell-cell adhesion [23]. In addition, E-cadherin truncation was significantly higher in low MSI as compared with high MSI tumors [37]. The inferred mechanism is that the low expression of FUT8 in TNM stages IV or microsatellite stable colorectal cancer may cause E-cadherin to decrease or be truncated, resulting in decreased tumor cell-cell adhesion and increased metastasis.
Medulloblastoma, a common type of malignant brain cancer that accounts for 8-10% of childhood brain tumors [38], comprises four molecular subgroups (WNT, SHH, group 3, group 4) that have differing prognoses [39]. In general, subgroup WNT has very good prognosis and rare metastasis; subgroup SHH has good prognosis in infants and intermediate prognosis and uncommon metastasis in older children; subgroup G4 has intermediate prognosis and frequent metastasis; subgroup G3 has poor prognosis and very frequent metastasis [40][41][42][43]. Our analysis revealed association of high FUT8 expression with subgroups WNT and SHH, and inverse correlation with metastasis and male gender. High FUT8 expression in medulloblastoma therefore appears to be a positive prognostic indicator, particularly for female patients. In regard to NB, we did not find any notable relationship of high FUT8 expression with clinicopathological features. On the other hand, another group reported aberrant expression of N-glycans and short O-glycans in NB cells, and regulation of their expression levels by associated glycosyltransferases [44]. GnT-V expression in NB patients was correlated with favorable prognosis and treatment outcome [45]. GALNT9 was expressed in neuroblasts derived from primary tumor [46] but not in those derived from metastatic bone marrow, and may be a useful prognostic marker for positive clinical outcome in NB patients [47].
Overall, our results are comprehensive and seemingly reliable in view of the high quality of included articles and microarray data. On the other hand, there are inherent limitations in our analysis. (i) Heterogeneity is present among studies of a given tumor type, and is difficult to address because of methodological differences (e.g., sample selection, detection method, determination of cutoff value, statistical analysis). (ii) Nearly all our included studies report a statistically significant result. Although a Begg's funnel plot indicates absence of publication bias (Supplementary Figure 3), our experience suggests that selective reporting bias is common in the literature regarding FUT8 and tumor prognosis. (iii) Roughly half of our included studies had small sample size (<100), which is often associated with inflated estimates of effect size and with high heterogeneity.
Results of our analyseswith due consideration of the above caveatshighlight the prognostic value of FUT8 expression in a variety of malignant tumors, and the important biological function of FUT8 in tumor progression. FUT8 may exert its effects in such tumors by regulating functional protein core fucosylations involved in tumor development and metastasis (e.g., L1CAM, P53, TGF-β, EGFR) [25,34,48]. Thus, dysregulation of FUT8 would have effects on tumor development, and consequently on clinical prognosis of patients. The molecular mechanisms remain to be clarified, and are being addressed in ongoing studies.
In conclusion, our systematic review and meta-analysis confirmed the association of FUT8 with clinicopathological features and patient survival rates for certain malignant tumor types. FUT8 expression is a significant predictor of prognosis for these tumors. The relative weight of FUT8 correlations with particular clinical features is currently difficult to evaluate because of the presence of many uncontrollable factors. Reliable verification of prognostic value of FUT8 in these tumor types will require a large-scale study using standardized methods of detection, analysis, and reporting.

MATERIALS AND METHODS
We performed this review based on PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) criteria.

Search strategy
The PubMed, Embase, Web of Science, Springer, Chinese National Knowledge Infrastructure (CNKI), and Wan Fang databases were searched systematically to identify and retrieve all pertinent publications through June 24, 2020. Keywords and search terms used were: FUT8 or Fut8 or Fucosyltransferase 8 or α1-6 Fucosyltransferase or Core fucosylation or Fucosyltransferase; Malignant tumor or Cancer or Carcinoma or Neoplasm; and clinical or clinicopathological or clinicopathology or prognosis predictor or survival or Odds Ratio (OR) or Hazard Ratio (HR). References in retrieved articles were screened manually. Languages of retrieved articles were restricted to English and Chinese.

Data extraction and analysis of GEO datasets
Our first step was to read and screen articles related to tumor clinicopathological features and prognosis, and search for microarray datasets through the references. In this way, 96 datasets from 214 articles were extracted for preliminary screening. The second step was to rescreen the 96 datasets in regard to presence of FUT8 AGING data, an appropriate clinical pathology track, and sufficient number of samples. In this way, 35 malignant tumor microarray datasets were selected and downloaded from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/gds) and R2 platform (http://r2.amc.nl). These datasets included a total of 4918 samples, from 7 types of malignant tumor: breast cancer (BC), colorectal cancer (CRC), ependymoma, glioma, non-small cell lung cancer (NSCLC), medulloblastoma (MBL), and neuroblastoma (NB). The datasets all used the Affymetrix Human Genome U133a or U133 Plus 2.0 expression arrays to detect expression value signals. Data were normalized using Microarray Suite (MAS) V. 5.0. In many cases, more than one probeset has been reported for FUT8 gene. We selected a probeset with highest average AGING presence signal (APS) by default. mRNA expression of FUT8 was assigned to "low" or "high" category based on average expression value of each dataset.

Inclusion and exclusion criteria
Studies were included if they met the following criteria: (1) original study focused on human subjects; (2) presented FUT8 expression data and malignant tumor clinicopathological features and/or survival data; (3) reported an OR or HR with 95% confidence interval (CI), or sufficient data were presented so that we could calculate them; (4) full text was available. Exclusion criteria were as follows: (1) study lacked key information (e.g., clinical parameters, survival curves), or usable data; (2) APS of probeset was too low, or too many data were lost; (3) HRs applied to a combination of multiple FUTs; (4) article was a review, letter, single case report, or conference abstracts. In cases of multiple articles from the same group, reporting overlapping data, only the most complete one was included. The article selection process is shown in flow diagram form in Figure 5.

Quality assessment and data extraction
Two authors (MXM and GXH) reviewed potentially eligible articles independently. Quality of each study was assessed using the Newcastle-Ottawa Scale. The following information was extracted from each included study: (1) basic information: first author's or uploaded dataset author's name, publication year, country of origin, names of malignant tumors, sample size, FUT8 expression levels, detection methods, sample type, outcome measurements, follow-up duration, cutoff value, survival analysis method; (2) p values for correlation between FUT8 expression and clinicopathological features of malignant tumors, and original data used for calculation of ORs and their 95% CIs; (3) HRs and their 95% CIs for survival analysis. If HRs were not directly accessible in the text, Kaplan-Meier survival curves were read using Engauge Digitizer (V. 4.1) to obtain data. Different datasets for a particular malignant tumor were considered as separate studies, and respective HRs were extracted. In cases of possible discrepancy, a consensus was reached by discussion among all authors.

Statistical analysis
ORs and their 95% CIs were used to estimate associations of FUT8 with clinical features of malignant tumors. For purposes of comparison, patients were divided into paired categories (e.g., male vs. female; TNM stages I, II and III vs. IV; ER/PR status positive vs. negative). For survival rates, HRs with corresponding 95% CIs were used. All ORs and HRs were calculated for high FUT8 expression. When a given FUT8 was investigated in two or more different studies, a meta-analysis was performed to combine the effect size. Z test was used to determine significance of ORs or HRs. Heterogeneity between studies was tested using Q statistic and I 2 test. When I 2 value was >50% (indicating significant heterogeneity) a random-effects model was used. For I 2 value ≤50%, a fixed-effects model was used. Statistical analyses were performed with software program Review Manager V. 5.3 (Cochrane Collaboration; London, UK). Differences with p< 0.05 were considered statistically significant.

AUTHOR CONTRIBUTIONS
MXM and XL conceived and designed the study. MXM and GXH searched and reviewed literature. MXM, GXH, ZYZ and YW contributed to data collection, analysis, interpretation and prepared tables and figures. MXM drafted the manuscript. XL and FG revised the manuscript. All authors approved the final manuscript.