O-GalNAc Glycosylation Activates MBL-Mediated Complement and Coagulation Cascades to Drive Organotropic Metastasis

Abstract

Liver metastasis is prevalent among patients with neuroendocrine prostate cancer (NEPC) and other types of neuroendocrine (NE) cancers, featuring with an aggressive clinical course and a dismal prognosis. However, the cellular and molecular mechanisms underlying liver-specific metastatic tropism in NE cancers remain poorly understood. Intriguingly, it is found that NEPC liver metastatic foci are frequently associated with thrombi. NEPC cells express an aberrantly elevated level of glycosyltransferase Galnt9. Notably, the Galnt9-mediated O-GalNAc glycosylation on the cell membrane of NE cancer cells, particularly on the adhesion molecule Annexin A2, activates the mannose-binding lectin (MBL) complement signaling in the liver. This cascade stimulates platelet activation and thrombus formation, ultimately facilitating hepatic metastasis of NEPC. Inhibition of O-GalNAc glycosylation or knockdown of Galnt9 demonstrates efficacy in restraining the liver metastasis of NEPC, small cell lung cancer (SCLC), and colorectal neuroendocrine cancer. These findings identify Galnt9-mediated O-GalNAc glycosylation as a targetable mechanism driving liver metastasis through activation of MBL complement and coagulation cascades across a broad spectrum of NE cancers.

Keywords: O‐GalNAc glycosylation; complement; glant9; liver metastasis; mannose binding lectin (MBL); neuroendocrine prostate cancer.

Figure 1

Thrombi are frequently in close proximity to liver metastatic foci in NEPC tumor‐bearing mice. a) The schematic diagram of the animal model for liver metastasis of NEPC via orthotopic inoculation of rb1 ^Δ/Δ p53 ^Δ/Δ murine NEPC organoids into the prostate of WT C57BL/6 mice. b–e) H&E and IHC c–e) staining showing that the CD56⁺SYP⁺AR^‐ micro‐metastatic foci are attached to or surrounded by thrombi in the liver of NEPC‐tumor bearing mice at the early metastatic stage. Scale bar = 100 µm. f–h) H&E staining showing that the micro‐metastatic foci are attached to or surrounded by thrombi in the liver of 3 individual NEPC‐tumor bearing mice at the early metastatic stage. Scale bar = 50 µm i–l) IHC staining against CD41 and Fibrin confirms thrombi aggregations around micro‐metastatic NEPC tumor proliferating cell clusters at the early metastatic stage. Scale bar = 50 µm m,n) IHC staining against CD41 and Fibrin confirms thrombi aggregations adjacent to macro‐metastatic tumor cell clusters at the late metastatic stage (scale bar = 100 µm, upper panel). Indicative areas are enlarged to represent a detailed colocalization between thrombi and liver metastasis. (scale bar = 100 µm, lower panel). o,p) IHC staining against the neuroendocrine markers CD56 and the proliferating marker Ki67, validating the identity of NEPC cells at the late metastatic stage (scale bar = 100 µm).q,r) Quantification results showing that 8 out of 9 rb1 ^Δ/Δ p53 ^Δ/Δ NEPC tumor‐bearing mice at the early metastatic stage and 12 out of 15 rb1 ^Δ/Δ p53 ^Δ/Δ NEPC tumor‐bearing mice at the late metastatic stage exhibit evident colocalization between thrombi and liver metastasis.

Figure 2

High MBL pathway activity is enriched in PCa liver metastasis and GalNAc is a major glycosylation form on NEPC cell surface. a) The KEGG analysis showing that the “complement and coagulation cascades” is most upregulated signature in liver metastasis versus primary prostate tumors of rb1 ^Δ/Δ p53 ^Δ/Δ NEPC tumor‐bearing mice (n = 3, mice). b,c) GSEA plots reveal that “complement and coagulation cascade” (b) and “MBL pathway activity” are significantly elevated in liver metastatic lesions compared to metastasis in other anatomic sites based on analysis of the SU2C prostate cancer dataset. d) Kaplan–Meier survival analysis of the SU2C prostate cancer dataset reveal that PCa patients with high MBL activity (n = 39) exhibit significantly shorter survival compared to PCa patients with low MBL activity (n = 39). The log rank‐test was applied for statistics. e,f) GSEA plots showing that the “cell surface glycosylation” is significantly elevated in NEPC compared to PrAD based on analysis of Beltran and SU2C prostate cancer datasets. g) Lectin microarray results that the O‐GalNAc glycosylation, as revealed by the strongest signal of vicia villosa (VVA) lectin toward the membrane fraction extracted from murine rb1 ^Δ/Δ p53 ^Δ/Δ NEPC organoids.

Figure 3

NEPC cells bind to MBL and activate the complement cascade. a,b) Schematic diagrams of experimental design for MBL binding and MBL activation assays. c,d) Flow cytometric analysis and quantification results showing the MBL binding capability of WT prostate epithelial organoids, myc ^hi pten ^Δ/Δ PrAD, and rb1 ^Δ/Δ p53 ^Δ/Δ NEPC tumor organoids. e) Quantification data showing the MBL activation capability of WT prostate epithelial organoids, myc ^hi pten ^Δ/Δ PrAD, and rb1 ^Δ/Δ p53 ^Δ/Δ NEPC tumor organoids. f–h) Flow cytometric analysis and quantification results showing that the effect of O‐glycosidase and/or α‐mannosidase treatment on MBL binding (f‐g) and activation activities (h) of rb1 ^Δ/Δ p53 ^Δ/Δ NEPC organoids. i) A schematic strategy reveals that the rb1 ^Δ/Δ p53 ^Δ/Δ NEPC organoids were intravenously (i.v.) injected into WT and MBL2‐knockout (MBL2‐KO) mice, validating the essential role of MBL in mediating liver metastasis in NEPC tumor‐bearing mice. j‐l) Dissected livers, H&E staining, and quantification of the liver metastatic foci number of WT and MBL2‐KO mice intravenously inoculated with rb1 ^Δ/Δ p53 ^Δ/Δ NEPC organoids (n = 4, mice). For statistics in this figure, the two‐tail unpaired Student's‐t test was applied for, and the one‐way ANOVA test was applied for (d‐e) and (g‐h). Data were shown as means ± SD.

Figure 4

Inhibition of O‐GalNAc glycosylation or Galnt9‐KD attenuates MBL binding and activation and inhibits liver metastasis in NEPC. a) The volcano pot depicts the upregulated, downregulated, and unchanged genes, which encode the enzymes involved in the first two steps of glycosylation based on Beltran human prostate cancer dataset. b) Immunoblots showing the protein level of Galnt9 in WT prostate, myc ^hi pten ^Δ/Δ PrAD, and rb1 ^Δ/Δ p53 ^Δ/Δ NEPC organoids. c) Immunoblots showing the protein level of GALNT9 in VCaP, LAPC4, 22Rv1, Du145, PC3 and LASCPC‐01 cells. d,e) IF staining and median fluorescence intensity (MFI) quantification showing the upregulated O‐GalNAc glycosylation in rb1 ^Δ/Δ p53 ^Δ/Δ NEPC organoids compared to myc ^hi pten ^Δ/Δ PrAD counterparts (scale bar = 20 µm). f–g) IF staining and MFI quantification showing the upregulated O‐GalNAc glycosylation in rb1 ^Δ/Δ p53 ^Δ/Δ NEPC organoids compared to myc ^hi pten ^Δ/Δ PrAD counterparts (scale bar = 10 µm). h–j) O‐GalNAc inhibitor Benzyl‐α‐GalNAc (5 µm, treated for 24 h) significantly suppressed the MBL binding and activation capabilities of rb1 ^Δ/Δ p53 ^Δ/Δ NEPC organoids. k,l) IF staining and MFI quantification showing the upregulated O‐GalNAc glycosylation in rb1 ^Δ/Δ p53 ^Δ/Δ‐scramble and rb1 ^Δ/Δ p53 ^Δ/Δ‐shGalnt9 NEPC organoids (scale bar = 50 µm). m–o) Galnt9‐KD in rb1 ^Δ/Δ p53 ^Δ/Δ NEPC organoids resulted in significantly decreased MBL binding (m‐n) and activation capabilities. p–q) IF staining images and MFI quantification revealed the O‐GalNAc glycosylation status in LAPC4, VCaP, and LASCPC‐01 cells (scale bar = 10 µm). r–t) GALNT9‐KD in human NEPC LASCPC‐01 cells resulted in significantly decreased MBL binding and activation capabilities. u–w) Dissected livers, H&E staining images, and quantification of the liver metastatic foci number w) of the C57BL/6 recipients inoculated with rb1 ^Δ/Δ p53 ^Δ/Δ‐scramble (n = 7, mice) and rb1 ^Δ/Δ p53 ^Δ/Δ‐shGalnt9 organoids (n = 8, mice). For statistics in this figure, the two‐tail unpaired Student's‐t test was applied for (e), (g), (i‐j) and (w), and the one‐way ANOVA test was applied for (n‐o), (q), and (s‐t). Data were shown as means ± SD.

Figure 5

Galnt9‐OE in PrAD enhances MBL activation and promotes liver metastasis. a) Immunoblotting assay confirming the ectopic expression of Galnt9 (Galnt9‐OE) in myc ^hi pten ^Δ/Δ PrAD organoids. b–d) Galnt9‐OE in myc ^hi pten ^Δ/Δ PrAD organoids significantly increased MBL binding and activation capabilities. e) Immunoblotting assay showing that Galnt9 is overexpressed in murine RM‐1 PrAD cells. f–h) Galnt9‐OE in murine RM‐1 PrAD cells significantly increased MBL binding and activation capabilities. i) Immunoblotting assay confirming the ectopic expression of GALNT9 (GALNT9‐OE) in human LAPC4 PrAD cells. j–l) GALNT9‐OE in human LAPC4 PrAD cells significantly increased MBL binding j,k) and activation l) capabilities. m) Immunoblotting assay confirming the ectopic expression of GALNT9 (GALNT9‐OE) in human VCaP PrAD cells. n–p) GALNT9‐OE in human VCaP PrAD cells significantly increased MBL binding and activation p) capabilities. q–s) Dissected livers, H&E staining images, and quantification of the liver metastatic foci number of the C57BL/6 recipients intravenously inoculated with RM‐1‐vector and RM‐1‐Galnt9‐OE cells (n = 6, mice). For statistics in this figure, the two‐tail unpaired Student's‐t test was applied for (g‐h), (k‐l), (o‐p) and (s), and the one‐way ANOVA test was applied for (c‐d). Data were shown as means ± SD.

Figure 6

Glycosylation‐triggered MBL pathway promotes platelet activation and facilitates NEPC liver metastasis. a) The KEGG analysis showing that the “proteoglycans in cancer” and “platelet activation” were upregulated in rb1 ^Δ/Δ p53 ^Δ/Δ‐scramble organoids versus rb1 ^Δ/Δ p53 ^Δ/Δ‐shGalnt9 organoid counterparts. b) The GSEA plot reveals that the “complement and coagulation cascades” hallmark genes were significantly enriched in rb1 ^Δ/Δ p53 ^Δ/Δ‐scramble organoids in comparison to rb1 ^Δ/Δ p53 ^Δ/Δ‐shGalnt9 organoid counterparts. c) The in vivo platelet activation assay reveals that C57BL/6 recipients inoculated with rb1 ^Δ/Δ p53 ^Δ/Δ‐shGalnt9 organoids (n = 8, mice) showed significantly decreased platelet activation capability in comparison to the mice‐receiving rb1 ^Δ/Δ p53 ^Δ/Δ‐scramble organoids (n = 7, mice). d) The tail vein bleeding assay shows that C57BL/6 recipients inoculated with rb1 ^Δ/Δ p53 ^Δ/Δ‐shGalnt9 organoids (n = 8, mice) takes significantly longer time in coagulation compared to the mice‐receiving rb1 ^Δ/Δ p53 ^Δ/Δ‐scramble organoids (n = 7, mice). e) A schematic diagram of experimental design for platelet activation assay. f) Galnt9‐KD in rb1 ^Δ/Δ p53 ^Δ/Δ organoids leads to significantly deduced platelet activation capability compared to scramble control rb1 ^Δ/Δ p53 ^Δ/Δ organoids in vitro. g) GALNT9‐KD in human NEPC LASCPC‐01 cells leads to significantly reduced platelet activation capability compared to scramble control shRNA‐transfected LASCPC‐01 cells in vitro. h) Galnt9‐OE in murine myc ^hi pten ^Δ/Δ PrAD organoids significantly increased platelet activation capability compared to vector control PrAD organoids. i) Galnt9‐OE‐induced platelet activation in PrAD was diminished in MBL2‐KO recipients versus WT recipient counterparts (n = 5 mice). j) Galnt9‐OE‐mediated coagulation, as revealed by shorter bleeding time, in PrAD is reversed in MBL2‐KO recipients versus WT recipient counterparts (n = 5 mice). k–n) Depletion of platelet in vivo using anti‐CD42b antibody significantly attenuates the liver metastatic burdens in rb1 ^Δ/Δ p53 ^Δ/Δ organoid‐inoculated mice, as revealed by reduced liver metastatic foci numbers in contrast to anti‐IgG control treated tumor‐bearing mice (n = 4, mice). Scale bar = 5 mm in (m). For statistics in this figure, the two‐tail unpaired Student's‐t test was applied for (c‐d), (k) and (n), and the one‐way ANOVA test was applied for (f‐i). Data were shown as means ± SD.

Figure 7

Blockade of O‐GalNAc suppresses liver metastasis in a wide range of NE cancers. a–d) The O‐GalNAc inhibitor benzyl‐α‐GalNAc treatment leads to significantly reduced liver metastatic burdens in rb1 ^Δ/Δ p53 ^Δ/Δ organoid‐inoculated C57BL/6 mice (n = 7), as exemplified by representative H&E staining imagesscale bar = 5 mm) and quantification data. e‐h) Benzyl‐α‐GalNAc treatment results in significantly attenuated liver metastatic burdens in SCLC NCI‐H82‐inoculated nude mice (n = 6), as revealed by representative H&E staining images, scale bar = 5 mm) and quantification data. i–l) Benzyl‐α‐GalNAc treatment leads to significantly repressed liver metastatic lesions in neuroendocrine colon cancer cell COLO‐320DM inoculated nude mice (n = 4), as exemplified by representative H&E staining images, scale bar = 5 mm) and quantification results. For statistics in (d), (h), and (l) student's t‐test was applied and data were shown as mean ± SD. P value ≤ 0.01 was considered as statistically significant. For statistics in this figure, the two‐tail unpaired Student's‐t test was applied for (d), (h) and (l). Data were shown as means ± SD.

Figure 8

O‐glycosylated ANXA2 is a key protein that binds to and activates the MBL complement pathway in NEPC. a) Schematic diagrams illustrating the workflow for screening Galnt9 substrates using immunoprecipitation (IP) and mass spectrometry (MS). b) MS screening data showing a candidate list of Galnt9 substrates. c) Immunoprecipitation (IP) assay showing that Galnt9‐KD attenuates the O‐GalNAc glycosylation in murine rb1 ^Δ/Δ p53 ^Δ/Δ NEPC cells. d) Immunoprecipitation (IP) assay demonstrates that Galnt9‐KD impairs the interaction between Anxa2 and MBL2. e) Immunoprecipitation (IP) assay demonstrates the mutations of the key O‐GalNAc glycosylated sites of Anxa2 including the 18th, 22nd, and 26th Serine residues (S18A, S22A, S26A) impaired O‐Glycosylation levels as compared to the intact WT Anxa2 construct. f–h) Mutations of Anxa2 in the 22nd and 26th serine sites result in significantly reduced MBL binding and activation capabilities. i‐l) Anxa2‐KD in rb1 ^Δ/Δ p53 ^Δ/Δ NEPC organoids leads to significantly reduced MBL binding j,k) and activation l) capabilities as compared to scramble controls. m–o) Dissected livers, H&E staining images, and quantification results on liver metastatic foci number of the C57BL/6 recipients inoculated with rb1 ^Δ/Δ p53 ^Δ/Δ‐scramble organoids and rb1 ^Δ/Δ p53 ^Δ/Δ‐shAnxa2 organoids (n = 7, mice). For statistics in this figure, the two‐tail unpaired Student's‐t test was applied for (o), and the one‐way ANOVA test was applied for (g‐h) and (k‐l). Data were shown as means ± SD.