From eligible studies, data were extracted from univariate Kaplan-Meier curves using WebPlotDigitizer (automeris.io) and individual patient data reconstructed using the IPDfromKM R-package (v. 0.1.10) [11]. To estimate the hazards ratio (HR) and associated standard error (SE(HR)), a univariate Cox-Proportional Hazards model was fitted. All models compared the low expression group to the high expression group of a certain mucin to evaluate the association between aberrant expression of a specific mucin type and survival outcomes, i.e., overall survival (OS) and cancer-specific survival (CSS). Final generic inverse variance meta-analysis was performed using meta R-package (v. 8.2-1). A random effects model was fitted to estimate pooled effect and it's 95% confidence interval. Forest plots were constructed to visualize HRs of individual studies as well as the pooled HR for each mucin investigated. A pooled HR > 1 indicated a worse prognosis in patients with high expression of the investigated mucin, whereas a pooled HR ≤ 1 suggested a better outcome. Heterogeneity between studies was assessed using the χ2, with alpha set at 0.1, I² statistic for estimating magnitude of heterogeneity and τ2 for between study variance estimation, while potential publication bias was examined with Begg's funnel plots and Egger's linear regression test. A P-value < 0.05 was considered to indicate significant publication bias. Meta-analysis was performed using the meta package in R (RStudio) including at least 2 eligible studies investigating survival endpoints in relation to aberrant expression of a certain mucin type.
The literature search identified 322 records that were screened against title and abstract and from this excluded 226 studies. Ninety-one studies were assessed for full-text eligibility (Fig. 1). Of these, 58 studies were retained, including 49 evaluating mucin expression in adenocarcinomas (i.e., 37 in CRC, 5 in intestinal adenocarcinomas, and 7 in rare tumours) and 16 in benign precursor lesions. Seven out of 58 investigated mucin expression in both benign and malignant lesions (Fig. 1). The majority of papers investigated mucin expression by immunohistochemical staining while a minority used RNA-based approaches (Supplementary Fig. S1).
Among the 58 articles reviewed, MUC2 was the most frequently investigated mucin as its expression was explored in 41 studies (Fig. 2a-f). MUC5AC was the second most often studied mucin, being described in 37 studies followed by MUC1 (26 studies) and MUC6 (19 studies; Fig. 2a-f). Co-expression of MUC1, MUC2 and MUC5AC was also often studied, as visualized in the upset plots and network charts (Fig. 2a-d). A detailed summary of the included studies can be found in Supplementary Tables S1-S3, whereas an overview of the number of mucins investigated per study and the total number of studies investigating each lesion type are depicted in Supplementary Figs. S2 and S3. Most studies (53%) assessed mucin expression in tissue samples from Asian populations, but North-American (12%), European (25%), North-African (5%), and Australian (4%) populations were also explored (Supplementary Tables S1-S3 and Fig. S4).
Eleven studies examined mucin expression in serrated polyps (Fig. 2e; Supplementary Table S1). There was an overall consensus that the expression of the secreted mucins MUC2 and MUC5AC was significantly increased in serrated polyps (Fig. 3b) [12,13,14,15,16,17]. Although MUC5AC is a gastric mucin type with low-level intestinal expression (Fig. 3a), this glycoprotein is aberrantly expressed in the upper segment of crypts near the luminal cell surface in serrated polyps [13, 14]. Furthermore, MUC5AC hypomethylation, which impacts the MUC5AC gene promotor and thus its expression level, was also more frequently detected in serrated polyps harbouring BRAF mutations, CIMP-H and/or MSI phenotypes [17]. By comparison, MUC2 hypomethylation was seen in polyps harbouring or lacking oncogenic mutations and correlated with the proximal colonic location in microvesicular hyperplastic polyps and sessile serrated lesions [17]. Additionally, expression of MUC6, another gastric mucin that is absent in the normal intestinal mucosa (Fig. 3a), is also increased in serrated polyps and predominantly in sessile serrated lesions and traditional serrated adenomas from the proximal colon (Fig. 3b) [14, 18,19,20]. Regarding the transmembrane mucins, expression of MUC1 seems to be increased in sessile serrated lesions, whereas a reduced expression was noted in hyperplastic polyps (Fig. 3b). This study also revealed that positive IHC staining for MUC1 and MUC6 and negative IHC staining for MUC2 in colorectal polyps associated with an increased risk of invasion in the mucosa or muscularis mucosae [18]. Besides, one study found a positive MUC13 staining in hyperplastic polyps (Fig. 3b), albeit with a sample size of two polyps [21].
Eight studies investigated mucin expression in adenomatous polyps (Fig. 2e; Supplementary Table S1). They also found significant higher expression levels of MUC2 and MUC5AC (Fig. 3b) albeit with a lower IHC staining intensity compared to what has been described in serrated polyps [15, 22, 23]. Additionally, MUC5AC hypomethylation seems to be absent in conventional adenomas [15, 16, 23]. Expression of the secreted MUC5B and MUC6 mucins was evident in only a fraction of adenomas, i.e., 27% and 10%, respectively [12], whereas the presence of the transmembrane MUC1 and MUC17 mucins was overall more pronounced in these precursor lesions [12]. Furthermore, multivariate regression analyses showed that aberrant MUC2, MUC5AC, and MUC17 signatures have the potential to discriminate between adenoma-adenocarcinoma progression and hyperplastic polyps [12], while other studies highlighted that high-level MUC1 and low-level MUC2 expression (Fig. 3b) correlates with a more severe precursor lesion and progression towards adenocarcinoma [24, 25]. More specifically, expression of MUC1 is low in the healthy intestinal mucosa (Fig. 3a) and in adenomas with mild to moderate dysplasia whereas its expression level significantly increased with the grade of dysplasia and further progression towards adenocarcinoma. The opposite was however seen for MUC2. This intestinal-type secreted mucin is abundantly secreted by intestinal goblet cells in the healthy mucosa (Fig. 3a), but its expression decreases in adenomas with mild dysplasia and is even lower in adenomas with moderate and severe dysplasia [24].
A significant loss of MUC2 expression in CRC adenocarcinomas compared to normal colorectal tissues was reported in most of the included studies (Fig. 3c; Supplementary Table S2) [25,26,27,28,29,30,31,32] and such tumours particularly originated from the distal (left-sided) colon and rectum. On the contrary, significant increased MUC2 expression was also described in CRC adenocarcinomas, specifically in the mucinous CRC subtypes (Fig. 3c) that arose in the proximal (right-sided) colon [26, 28, 33,34,35,36,37,38]. Regarding CRC outcome, low-level MUC2 significantly correlated with a poor overall survival in CRC, disease recurrence and disease progression, whereas high-level MUC2 was rather linked to a longer disease-free survival in patients with CRC of any tumour stage [31, 32, 34, 39,40,41]. However, others highlighted that the overall 3-year survival is significantly improved in mucinous adenocarcinoma with low-level MUC2 than high-level MUC2 [42]. This discrepancy in survival between high and low MUC2 expression was also reflected in the meta-analysis, which found no significant association between MUC2 expression and overall survival (Fig. 4a, Supplementary Fig. S11). A significant association between low-level MUC2 and worse outcome when investigating cancer-specific data was however, seen (Fig. 4b; Supplementary Figs. S7, S9). Of note, the pooled studies did not distinguish between mucinous and non-mucinous adenocarcinomas. Several studies also reported de novo expression of MUC5AC in CRC adenocarcinomas [12, 35, 43,44,45,46]. Interestingly, MUC5AC hypomethylation associated with high MUC5AC expression and together with aberrant MUC2 expression, are linked to the MSI CRC subtype [28]. Furthermore, high-level MUC5AC expression correlated with the proximal tumour location, an increased overall survival (as supported by the meta-analysis; Fig. 4c; Supplementary Figs. S10, S12), longer progression-free survival, improved disease free survival and a decreased risk of recurrence or metastatic disease in the post-operative period [34, 43, 47]. MUC5B, highly expressed in the lower crypts of the healthy colon, was not significantly altered in CRC tumours compared to their normal counterparts [12, 44], although its expression was associated with the presence of tumour-infiltrating lymphocytes and mucinous differentiation [35]. Similarly to MUC5AC, MUC6 is significantly upregulated in CRC (Fig. 3c) [34, 35] and correlated with the presence of tumour-infiltrating lymphocytes, proximal colonic location, mucinous tumour differentiation, and 100% progression-free survival (Supplementary Table S2) [34, 35].
Regarding transmembrane mucin expression, a significant heterogeneity in the included studies investigating MUC1 expression was noted. Whereas some stated that MUC1 expression was not detectable in the healthy mucosa [48,49,50] compared to others, all studies highlighted a significant increased MUC1 expression in CRC (Fig. 3c), albeit with a variability in MUC1 positive staining [43]. MUC1 expression also positively correlated with the development of high-grade dysplasia, tumour stage (i.e., pTNM and pM, but not pN), and the depth of tumour, lymphatic and venous invasion [25, 34, 51,52,53]. Well and moderately differentiated CRC adenocarcinomas seemed to have a significantly lower MUC1 expression than poorly differentiated CRC tumours whereas significant increased MUC1 expression can occur in both mucinous and non-mucinous tumour subtypes (Fig. 3c) [24,25,26, 30, 34, 54]. One study, investigating the difference in MUC1 expression between the tumour centre and invasion front, even highlighted that the intensity of MUC1 expression was the strongest in the tumour centre of poorly differentiated CRCs [39]. Furthermore, increased MUC1 expression significantly associated with worse overall and cancer-specific survival (as supported by the meta-analysis; Fig. 4d, e; Supplementary Figs. S5, S6, S8, S13), depending on the cellular location (i.e., cytoplasm versus apical staining) [51, 55], as well as with poorer post-operative survival in CRC tumours showing stromal-dominant MUC1 staining [55]. One study reported aberrant MUC3 expression in 84% of the investigated CRC tumours [56], whereas others [12] highlighted a significant increase in MUC4 expression in mucinous adenocarcinomas (Fig. 3c). MUC13, a predominant transmembrane mucin expressed in the normal colorectal epithelium and mainly by columnar epithelial cells (Fig. 3a), is aberrantly expressed (Fig. 3c) in the cytoplasm of CRC tumour cells [21]. Such high cytoplasmic MUC13 expression associated with poorly differentiated and late-stage tumours, specifically of the non-mucinous left-sided tumour subtype, and worse overall survival [21]. Other studies further described a reduced MUC12 and MUC15 expression in CRC tumours (Fig. 3c) [57, 58], whereas significant increased MUC20 expression also associated with poor prognosis in CRC (Supplementary Table S2; Fig. 3c) [59].
In total, 5 studies investigated mucin expression in adenocarcinomas of the middle (jejunum) and distal part (ileum) of the small intestine (Fig. 4; Supplementary Table S2) [52, 60,61,62,63]. A significant loss of MUC2 expression was described in small intestinal adenocarcinomas (Fig. 3c) and occurred more often in this type of malignancy compared to CRC adenocarcinomas [52, 60]. Furthermore, loss of this secreted glycoprotein was also linked to the tumour's behaviour as its expression level negatively associated with lymphatic invasion, tumour size and non-mucinous, poorly differentiated tumours (Supplementary Table S2) [60]. Also a significant increase in MUC5AC expression was noted in a large amount of small intestinal adenocarcinoma samples compared to non-neoplastic intestinal epithelium (Fig. 3c), albeit at a lower frequency compared to CRC tumours [52, 60]. Regarding associations with clinicopathological findings, increased MUC5AC expression positively correlated with low-grade tumours, poor prognosis, lymph node metastasis and MSI in combination with low MUC2 secretion (Supplementary Table S2) [61, 62]. Small intestinal tumours with high MUC6 expression were associated with lymph node metastasis [60], whereas other studies reported that MUC6-positive tumours were more often characterized by a lower T-classification (Fig. 3c, Supplementary Table S2) [61]. Whereas a significant increase of MUC1 expression was found in small intestinal adenocarcinomas, particularly in poorly differentiated tumours, no clear difference in MUC4 and MUC16 expression was seen compared to the normal intestinal mucosa (Fig. 3c) [52, 60, 61]. Furthermore, expression of MUC1 and MUC16 was also higher in cases with deeper invasion depth and venous invasion and their expression level positively correlated with a poor outcome (Supplementary Table S2) [60, 61, 63].
Aberrant expression of MUC1, MUC2, and MUC5AC has also been described in colonic signet-ring cell carcinoma (SRCC) (Fig. 3d) [64]. Expression of MUC1 and MUC2 was not significantly different in medullary colon carcinoma, compared to poorly differentiated colonic carcinoma (Fig. 3d) [65]. Furthermore, colorectal adenocarcinomas with enteroblastic differentiation (CAED) and high-level MUC5AC tend to have larger tumour masses and originate from the right side of the colon [66]. Finally, different types of mucinous appendiceal neoplasms, namely low-grade appendiceal mucinous neoplasm (LAMN) and mucinous appendiceal adenocarcinoma, have also shown to positively correlate with increased MUC2 and MUC5AC expression [67,68,69,70].