Glycosyltransferases regulate the expression of Golgi phosphoprotein 3 (GOLPH3)

15 16 Glycosphingolipid glycosyltransferases (GGTs) can organize as multienzyme complexes localized along 17 the Golgi complex. However, the influence of the relative presence of GGTs on the localization of their 18 clients is unclear. Here, we determine that expression of certain full -length GGTs increases the levels of 19 Golgi phosphoprotein 3 (GOLPH3), an adaptor oncoprotein involved in Golgi trafficking and organization. 20 Furthermore, we demonstrate that expression of the N-terminal domain of GGTs, which lacks the catalytic 21 domain, is sufficient to achieve this regulation on GOLPH3 in a cell type -dependent manner. We also 22 identify the N -terminal domain of β4GalT -VI GGT as an inhibitor of GOLPH3 expression and thus a 23 potential therapeutic application, since GOLPH3 overexpression is associated with progression and poor 24 prognosis of multiple tumor types. Our data further suggest that the cytoplasmic tail of β4GalT -VI N -25 terminal domain interferes with the ability of GOLPH3 to interact with phosphatidylinositol 4 -phosphate, 26 which consequently reduces the levels of GOLPH3, thereby impairing its function in the acquisition of 27 mesenchymal features. 28 29 30 31 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint

32 33 Glycosyltransferases play a crucial role in cell physiology, contributing to glycosylation of proteins 34 and lipids. Their abnormal expression and /or activity are associated with several human diseases and 35 disorders (Cumin et al., 2022; Lopez & Schnaar, 2009; Varki et al., 2022) . The canonical function of 36 glycosyltransferases is to catalyze the transfer of a monosaccharide from an activated donor molecule to a 37 specific acceptor molecule, forming a glycosidic bond. With few exceptions, the enzymatic activity of 38 glycosphingolipid glycosyltransferases (referred as GGTs) is mainly carried out in the lumen of the Golgi 39 cisternae. These enzymes have type II protein topology with an N -terminal domain (NTD) containing a 40 cytoplasmic tail, a transmembrane domain (TMD) and a stem region that connects the TMD to the C -41 terminal catalytic domain oriented toward the lumen of the Golgi (Vilcaes et al., 2011) (Fig.1A). Several 42 molecular features within the NTD promote the retention and localization of GGTs to specific sub -Golgi 43 compartments (Chumpen Ramirez et al., 2017; Maccioni, Quiroga, & Ferrari, 2011) . The NTD additionally 44 mediates the association and organization of GGTs into distinct multienzyme complexes (Maccioni, 45 Quiroga, & Spessott, 2011) . How the relative presence of these GGTs influence the localization of their 46 clients is n ot fully elucidated in the field . The highly conserved Golgi phosphoprotein 3 (GOLPH3) is 47 considered the first Golgi resident oncogene protein. GOLPH3 is overexpressed in many human tumors, 48 correlating with poorer patient survival (Li et al., 2012; Wang et al., 2014; Xue et al., 2014; Zeng et al., 2012). 49 Recent findings from our laboratory show that the physical association between ST3Gal -II and β3GalT-IV 50 GGTs (Fig.1B) through their NTDs is mediated by GOLPH3 (Ruggiero et al., 2022). While the retention of 51 these two enzymes at the Golgi complex does not require GOLPH3, other GGTs bind GOLPH3 via their 52 cytoplasmic tails and this interaction influences their distribution within the Golgi as well as their protein 53 levels through regulated lysosomal degradation (Rizzo et al., 2021) . The GGT ST8Sia -I forms a ternary 54 complex at the Golgi with the upstream GGTs (grey box in Fig.1B) thus facilitating the synthesis of b-series 55 gangliosides (yellow box in Fig.1B). Through this association, ST8Sia-I modulates the sub-Golgi localization 56 of the other GGTs in the complex (Uliana, Crespo, et al., 2006). These results agree with the possibility that 57 topological distribution along the Golgi complex of GGTs that form part of multienzyme complexes relies 58 on the relative levels of the partners (McCormick et al., 2000; Seko & Yamashita, 2005; Spessott et al., 2012). 59 In this scenario, given that GOLPH3 is a crucial partner in the formation of the GGT complexes, it is also 60 possible that the presence of GGTs may affect GOLPH3 localization. 61 In the present work, we demonstrate that the expression of ST8Sia-I, β3GalT -IV and ST3Gal -II 62 enzymes increase the expression of GOLPH3 in CHO -K1 cells. Moreover, the presence of NTDs from 63 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint ST8Sia-I, β3GalT-IV and ST3Gal -II GGTs, which lacks the catalytic domain, are sufficient to increase the 64 expression of GOLPH3 in a cell type -specific manner. Moreover, our results indicate that presence of the 65 NTD of β4GalT-VI GGT strongly inhibits the levels of GOLPH3 in cancer cell lines and consequently, 66 reduces their capacity to undergo epithelial -mesenchymal transition. These results uncover a novel 67 inhibitory mechanism of GOLPH3 levels with potential therapeutic applications in cancer. 68 69 70 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint

71 72 Ganglioside glycosyltransferases expression modifies GOLPH3 levels 73 74 To study the influence of glycosphingolipids metabolism on GOLPH3 expression, we took 75 advantage of the CHO -K1 cell lines already established and well -documented in our laboratory ( Fig.1B) 76 (Rodriguez-Walker et al., 2015; Uliana, Giraudo, et al., 2006; Vilcaes et al., 2011). Wild-type CHO-K1 (CHO-77 K1wt) cell line is virtually devoid of ST8Sia -I (Daniotti et al., 2000) and β4GalNAcT-I (Rosales Fritz et al., 78 1997) activities, in agreement with a recent RNA-Seq data in which the ST8Sia -I gene showed almost no 79 expression (Vishwanathan et al., 2015). Consequently, the glycosphingolipids pattern of CHO-K1wt cells is 80 dominated by glucosylceramide ( GlcCer) and GM3 ganglioside (Crespo et al., 2002) . ST8Sia-I 81 glycosyltransferase is a key regulatory enzyme controlling the synthesis of b- and c-series gangliosides. The 82 stable expression of ST8Sia -I in CHO -K1 cells (CHO -K1ST8Sia-I) generates primarily the ganglioside GD3 83 (Vilcaes et al., 2011) (Fig.1B-C). On the other hand, CHO -K1 cells stably expressing β4GalNAcT -I and 84 β3GalT-IV glycosyltransferases (CHO-K1β4GalNAcT-I/β3GalT-IV) synthesize the a -series gangliosides GM1 and 85 GD1a (Fig.1B-C). Several reports show that GOLPH3 is localized at the Golgi apparatus (Tenorio et al., 86 2016; Xing et al., 2016). In CHO-K1 cells, the spatial distribution of GOLPH3 is also confined mainly to the 87 Golgi cisternae where it is alongside specific markers of cis -Golgi GM130 and trans-Golgi network P230 88 (Fig.1D and 1E, respectively). By immunofluorescence (IF) analysis, we found that CHO-K1ST8Sia-I and CHO-89 K1β4GalNAcT-I/β3GalT-IV cells expressed significantly higher levels of GOLPH3 compared to wild -type (WT) 90 condition ( Fig.1F). The results were also confirmed by Western Blot (WB) ( Fig.1E), suggesting that the 91 presence of these GGTs increases the expression of GOLPH3 in CHO-K1 cells. 92 93 N-terminal domain of GGTs is sufficient to modify GOLPH3 levels in CHO-K1 cells 94 95 It has been described that the NTD of GGTs ( Fig.1A) is necessary and sufficient to confer Golgi 96 localization of GGTs (Maccioni, Quiroga, & Spessott, 2011) and is involved in covalent and non -covalent 97 associations in some glycosyltransferase complexes (Hassinen et al., 2011) . In addition, we previously 98 demonstrated that both NTDs of β3GalT -IV and ST3Gal-II glycosyltransferases are physically associated 99 and also that GOLPH3 interacts with both enzymes (Ruggiero et al., 2022) . Thus, this prompted us to ask 100 whether the NTDs of β3GalT -IV and ST3Gal -II enzymes influence GOLPH3 expression. To test this 101 possibility, the NTD of ST3Gal -II fused to mCherry (ST3Gal -II(1-51)-mCherry) (Ruggiero et al., 2015) and the 102 fusion protein of β3GalT-IV containing amino acids 1–52 fused to YFP (β3GalT-IV(1-52)-YFP) (Uliana, Crespo, 103 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint et al., 2006) were expressed transiently in CHO-K1wt cells. As shown in Fig.2A-B, WB analysis indicated 104 that NTD of these enzymes were sufficient to significantly increase the levels of GOLPH3 compared to 105 control condition. Since GGT NTDs lack the catalytic domain, our results also suggest that the increase in 106 GOLPH3 expression is independent of gangliosides synthesis. T o examine this possibility both CHO-107 K1ST8Sia-I and CHO-K1β4GalNAcT-I/β3GalT-IV cells were grown for four days in the presence of 1.2 μM P4, a GlcCer 108 synthase inhibitor. Under these culture conditions the synthesis of glycosphingolipids including 109 gangliosides is blocked and cell membranes are essentially devoid of them, as previously described (Vilcaes 110 et al., 2011) . As expected, P4 treatment remarkably reduced the levels of GD3 as well as GM1 and GD1a 111 expression in CHO-K1ST8Sia-I and CHO -K1β4GalNAcT-I/β3GalT-IV cells, respectively ( Fig.2C). However, the P4 112 treatment had no effect on GOLPH3 protein levels in both cell lines (Fig.2E-F). Taken together, these results 113 support the idea that the NTD of GGTs regulate GOLPH3 expression independently of the catalytic domain 114 or the presence of gangliosides. 115 116 N-terminal domain of GGTs modify GOLPH3 expression in breast cancer cells 117 118 GOLPH3 expression is elevated in various types of solid tumors and it is associated with poor 119 overall survival in patients with breast cancer (Kuna & Field, 2019; Scott et al., 2009; Zeng et al., 2012) . 120 MDA-MB-231 and MCF7 are widely used human breast adenocarcinoma cell lines. While MDA -MB-231 121 are highly aggressive, invasive and have a more mesenchymal phenotype, MCF7 retain some epithelial 122 features including hormone dependency (Theodossiou et al., 2019). Both cell lines express higher levels of 123 GOLPH3 compared to MCF10A (Tenorio et al., 2016), a non-tumorigenic epithelial cell line widely used as 124 in vitro model for studying normal breast cell function and transformation. To further explore the role of 125 NTD domain of GGTs in GOLPH3 expression, we extend ed our analysis to the human breast 126 adenocarcinomas MCF7 and MDA-MB-231 cell lines and the MCF10A human mammary epithelial cell line. 127 To do this, ST3Gal-II(1-51)-mCherry, β3GalT-IV(1-52)-YFP NTDs and the NTD of ST8Sia-I containing amino acids 1–128 57 fused to YFP (ST8Sia -I(1-57)-YFP) (Chumpen Ramirez et al., 2017) were expressed transiently in the three 129 cell lines. Then, cells were fixed, and processed for localization of GOLPH3 via immunofluorescence 130 staining. The results showed that ST8Sia-I(1-57)-YFP positive cells upregulate GOLPH3 in MCF10A and MCF7 131 lines (Fig.3A). However, the levels of GOLPH3 in MDA -MB-231 remained unchanged in cells positive for 132 ST8Sia-I(1-57)-YFP (Fig.3A). On the other hand, the expression of β3GalT -IV(1-52)-YFP (Fig.3B) or ST3Gal -II(1-51)-133 mCherry (Fig.3C) did not modify the levels of GOLPH3 in the three cell lines tested . As mentioned before, 134 GOLPH3 interacts with the cyto plasmic tails of several GGTs including β4GalT-V (Rizzo et al., 2021) . 135 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint β4GalT-V is a lactosylceramide synthase and together with β4GalT -VI are responsible for the production 136 of lactosylceramide, a key precursor in ganglioside biosynthesis (see Fig.1B). Since β4GalT-VI forms the 137 ternary complex β4GalT -VI/ST3Gal-V/ST8Sia-I through their N -terminal domains in CHO -K1 cells 138 (Maccioni, Quiroga, & Spessott, 2011) , it might also mimic the effect of ST8Sia-I on GOLPH3 expressio n. 139 We thus examined whether the NTDs of β4GalT-V and β4GalT-VI both fused to GFP (β4GalT-V(1-52)-GFP and 140 β4GalT-VI(1-52)-GFP; respectively) modif y GOLPH3 expression in breast cancer and mammalian epithelial 141 cells. Unexpectedly, the expression of β4GalT-V (1-52)-GFP did not affect the expression of GOLPH3 in the 142 three cell lines tested (Fig.4A-B). Similarly, the presence of β4GalT -VI (1-52)-GFP in MCF10A cells did not 143 change GOLPH3 levels either. Moreover, GOLPH3 was remarkably decreased in MCF7 and MDA-MB-231 144 cells positive for β4GalT -VI (1-52)-GFP. In summary, these findings indicate that different NTDs of GGTs can 145 upregulate or downregulate the expression of GOLPH3 in eukaryotic cells, in a cell type -dependent 146 manner. 147 148 The N-terminal domain of β4GalT-VI glycosyltransferase impairs the acquisition of mesenchymal-like 149 cancer cell features 150 151 GOLPH3 promotes the epithelial -mesenchymal transition (EMT) process, contributing to tumor 152 growth, metastasis, and poor prognosis in several types of cancer. Also, GOLPH3 overexpression is 153 concomitant with the upregulation of vimentin expression (Zhu et al., 2025), one of the key mesenchymal 154 markers that is upregulated during metastasis and cancer progression (Liu et al., 2015; Usman et al., 2021). 155 Since GOLPH3 levels are significantly downregulated in breast adenocarcinoma cells expressing the NTD 156 of β4GalT -VI, it is then possible that β4GalT -VI may participate in the inhibition of EMT progression 157 through the GOLPH3/Vimentin pathway. To address this, we first determined vimentin levels in cells 158 transfected with β4GalT-VI (1-52)-GFP. As shown in Fig.S1A-C, MDA-MB-231 cells expressing β4GalT -VI(1-52)-159 GFP had decreased vimentin levels, which correlated with a diminution in GOLPH3. Since MCF7 cells do 160 not express appreciable levels of vimentin (Liu et al., 2015), we decided to validate the effect of β4GalT -VI 161 (1-52)-GFP in the human skin melanoma cell line SK-MEL-28, widely used for studying metastatic progression 162 and potential therapeutic targets. Under the same experimental condition, our results confirmed a similar 163 negative regulation on the levels of vimentin and GOLPH3 by β4GalT-VI (1-52)-GFP in this cell line (Fig.S1B-164 C), indicating that the regulatory effect of β4GalT-VI may extent to multiple tumor types. Collectively, these 165 findings support the premise that the NTD of β4GalT-VI plays a role in the inhibition of EMT pathway. To 166 evaluate this hypothesis, we induced EMT in MCF10A and MDA -MB-231 cells using TGF-β1 (Deshmukh 167 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint et al., 2021) with or without the concurrent expression of β4GalT-VI (1-52)-GFP via lentiviral particles (Fig.5A). 168 As expected, TGF-β1 treatment resulted in robust increases in vimentin and GOLPH3 levels compared to 169 control condition in both cell lines (Fig.5B), indicating the acquisition of mesenchymal -like features. In 170 MCF10A cells, the presence of β4GalT -VI (1-52)-GFP partially impaired the expression of vimentin and 171 GOLPH3 in response to TGF -β1 treatment. Furthermore, β4GalT -VI (1-52)-GFP expression led to a ~ 10-fold 172 reduction in vimentin and GOLPH3 levels compared to uninfected MDA-MB-231 cells also treated with 173 TGF-β1 (Fig.5B). Finally, we tested whether expression of β4GalT -VI NTD has preventive potential, thus 174 rendering the cells impervious to EMT induction, by transducing MDA -MB-231 cells before TGF -β1 175 addition ( Fig.5C). The results show that the presence of β4GalT -VI (1-52)-GFP before exposure to TGF -β1 176 completely abolished the increase of both GOLPH3 and vimentin ( Fig.5D). Our results strongly indicate 177 that the NTD of β4GalT -VI prevents the acquisition of EMT features , uncovering a novel potential 178 therapeutic target for cancer prevention and treatment. 179 180 In silico analysis predicts different interaction surfaces between GOLPH3 and the cytoplasmic tails of 181 β4-galactosyltransferases V and VI 182 183 Biochemical assays and bioinformatic analyses proposed that GOLPH3 interact s with the short 184 cytoplasmic tails of numerous Golgi residents through its membrane-proximal polybasic stretches (Welch 185 et al., 2021) . Cytoplasmic tail (CT) of β4GalT -V and β4GalT -VI share a similarity of 60% . While the two 186 arginine-rich regions in β4GalT-V CT contribute to its net positive charge, these regions are less present in 187 β4GalT-VI CT (Fig.6A). These observations, together with the electrostatic charge distribution of GOLPH3 188 (face-down view from the membrane plane , Fig.6B), suggest that different regions in GOLPH3 may 189 mediate the interaction with the CTs of β4 -galactosyltransferases. We therefore used ClusPro server 190 (https://cluspro.org) to perform computational docking of GOLPH3/β4GalT -V and GOLPH3/β4GalT -VI 191 pairs. β4GalT-V CT was predicted to mainly interact with GOLPH3 via the electronegative region (Fig.6C), 192 consistent with previous reports (Rizzo et al., 2021; Welch et al., 2021)). On the contrary, the CT of β4GalT-193 VI adopted a different orientation, binding to GOLPH3 near the phosphatidylinositol 4-phosphate (PI4P) 194 binding pocket and the hydrophobic β-hairpin motif. In particular, β4GalT-VI R6 mediates electrostatic 195 interactions with GOLPH3 E175 which is next to R174 residue, key in binding PI4P and thus influencing 196 GOLPH3 localization to the Golgi (Dippold et al., 2009; Wood et al., 2009) (Fig.6C). For GOLPH3 R174, in 197 silico analysis predicted hydrogen bonds and van der Waals (VW) contacts with M1, S2, V3, L4 and R5 198 residues from β4GalT-VI CT. A detailed recent study showed that the interactions between residues E159 199 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint and R13; D247 and R9/R12; D258/D262 and R2/R4 formed as a key binding hub of GOLPH3 and β4GalT-V 200 complex (Theodoropoulou et al., 2025). Based on this information, we performed a ClusPro Biased Global 201 Docking of the GOLPH3/β4GalT -V and GOLPH3/β4GalT -VI pairs, guided by the key binding hub 202 identified above (Fig.6D). The resulting models were highly consistent with the blind docking condition. 203 β4GalT-V CT was predicted to contact GOLPH3 via the charged surface ( Fig.6D) while interaction of the 204 CT of β4GalT -VI occurred adjacent to the PI4P binding pocket and the hydrophobic β -hairpin motif of 205 GOLPH3. To further test the specificity of the predicted interactions between GOLPH3 and β4GalT-VI CT, 206 we carried out a new ClusPro Global Docking but this time we performed the residue changes R2S and 207 R4L in the CT of β4GalT -V to mimic those of β4GalT -VI CT. Under these conditions, the orientation of 208 β4GalT-V CT(R2S and R4L) became similar to that of β4GalT-VI CT (Fig.6E). Moreover, when the opposite switch 209 was performed (S2R), β4GalT-VI CT acquired analogous orientation to β4GalT-V CT. These in silico results 210 support the premise that β4GalT -VI CT (R2S) interferes with GOLPH3 binding to PI4P. Previous work 211 showed that hydrophilic residues placed at the tip of the hydrophobic β-hairpin of GOLPH3 (L195E/L196E) 212 interfere with GOLPH3 binding to PI4P -containing liposomes and prevented its Golgi localization 213 (Rahajeng et al., 2019). As mentioned above, docking revealed that the same hydrophobic β-hairpin (amino 214 acids 190-201) of GOLPH3 interacts with β4GalT -VI CT primarily via h ydrogen bonds and VW forces. In 215 this model, the interactions L195 with R13; L196 with R13; F197 with R9, D198 with R9; M199 with R6 and 216 R9; and T200 with R6 between GOLPH3 and β4GalT -VI CT, respectively, strongly suggest a reduction in 217 the hydrophobicity of GOLPH3 β-hairpin preventing its insertion and stabilization at the Golgi membrane. 218 219 220 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint

221 222 The Golgi complex plays a pivotal role in the processing, sorting and transport of proteins and 223 glycosphingolipids to different cell compartments (Iglesias-Artola et al., 2025; van Meer & Sprong, 2004) . 224 With the contributions of several laboratories, we have gained knowledge about how ganglioside 225 glycosyltransferases (GGTs) are organized through the Golgi cisternae and how this organization is crucial 226 for achieving the unique and precise glycosphingolipids pattern on cell surfaces. We now know that the 227 N-terminal domain of GGTs (Fig.1A) determines their localization within sub-Golgi compartment and that 228 the oncoprotein GOLPH3 interacts with and modulates the formation of glycosyltransferases complexes. 229 To the extent of our knowledge , here we present a novel role of several GGTs as regulators of GOLPH3 230 levels in a cell-type specific manner. 231 CHO-K1 cells stabl y expressing ST8Sia -I glycosyltransferase, a key enzyme controlling b-series 232 ganglioside biosynthesis ( Fig.1B), display augmented levels of GOLPH3 respect to wild type cells which 233 lack ST8Sia-I. Likewise, CHO-K1 cells expressing β4GalNAcT-I and β3GalT-IV glycosyltransferases, critical 234 for the synthesis of major complex gangliosides such as GM1 and GD1a gangliosides, also show increased 235 GOLPH3 levels. These results indicate that the presence of GGTs associated with different ganglioside 236 synthetic pathways regulates GOLPH3 expression at the Golgi complex. Furthermore, solely the NTDs of 237 ST3Gal-II and β3GalT-IV glycosyltransferases, which form an enzymatic complex modulated by GOLPH3, 238 are sufficient to increase the levels of this oncoprotein. It has also been reported that GOLPH3 over-239 expression in HeLa cells enhances the Golgi retention and residence time of β4GalT-V, preventing its 240 delivery to lysosomes. Collectively, our results reveal that GGTs and GOLPH3 are part of a bi -directional 241 regulatory mechanism where they influence each other’s levels and localization. 242 We identified the NTD of ST8Sia-I as a likely positive regulator of GOLPH3 levels in CHO-K1 cells. 243 Similarly, GOLPH3 levels are also increased in both human mammary MCF10A and breast carcinoma 244 MCF7 cells expressing the NTD of ST8Sia -I. However, the presence of this NTD in the highly aggressive 245 MDA-MB-231 cells does not change GOLPH3 expression. The transcriptional expression of the St8Sia-I 246 gene was observed in MDA -MB-231 cells but not in MCF10A and MCF7 cells (Cazet et al., 2009) . In 247 addition, based on RNA -seq analysis from Gene Expression Omnibus (GEO) repository (Messier et al., 248 2016), the gene expression of ST8SIA 1 in MCF10A and MCF7 cells also seems to be lower than in MDA -249 MB-231 cells (Fig.S2B). While GD3 ganglioside (product of St8Sia-I) is not detected in MDA -MB-231 cells 250 (Cazet et al., 2009), the expression of GT3 ganglioside (which builds up by the action of St8Sia-I on GD3) is 251 observed, indicating an active St8Sia-I. Under our experimental conditions, the GD3 ganglioside is also not 252 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint detected in MCF10A and MCF7 cells ( Fig.S2C). These data open the possibility that the expression of 253 GOLPH3 remains unaltered in MDA -MB-231 cells after the expression of NTD from St8Sia-I because of 254 endogenous levels of this enzyme being already plateaued leading to a ceiling effect on St8Sia-I-dependent 255 GOLPH3 expression. We also determined that the NTDs of β3GalT -IV and ST3Gal-II do not influence the 256 expression of GOLPH3 in MCF10A, MCF7 or MDA-MB-231 cells, contrary to the upregulation observed in 257 CHO-K1 cells. In line with a recent study (Cavdarli et al., 2020) , all three human breast cell lines have 258 similar mRNA levels of B3GalT4 gene (GSE75168, (Messier et al., 2016)) and they also express its product, 259 GM1 ganglioside (Fig.S2D and E). Similarly, we found that these cell lines have comparable mRNA levels 260 for ST3Gal-II (Fig.S2F). In addition, the ST3Gal -II expression is also evidenced by the fact that MFC7 and 261 MDA-MB-231 cells express GD1a ganglioside and SSEA4 globoside, respectively ( Fig.S2G and H). These 262 findings indicate that the bi -directional regulation between the NTD of GGTs and GOLPH3 depends on 263 cell type, specifically on the endogenous expression pattern of different GGTs across cells, uncovering a 264 new regulatory layer in the mechanisms that control GOLPH3 and glycosphingolipid metabolism in the 265 Golgi complex. In line with th ese results , a recent study suggests that other clients of yeast GOLPH3 266 (Vps74) are also involved in the recruitment of Vps74 to Golgi cisternae (Lesniak et al., 2025). 267 The β -galactosyltransferases β4GalT -V and β4GalT -VI are responsible for the synthesis of 268 lactosylceramide, the core structure of glycosphingolipids including gangliosides (Fig.1B). It was 269 previously demonstrated that the NTD of β4GalT-VI physically interacts with ST3Gal-V and ST8Sia-I GGTs 270 and that ST8Sia -I is able to promote changes in the sub-Golgi localization of β4GalT -VI. GOLPH3 also 271 interacts with the cytoplasmic tail of several GGTs, including β4GalT-V (Rizzo et al., 2021). While β4GalT-272 V is expressed in various tissues, β4GalT -VI is preferentially expressed in the brain (Lo et al., 1998; 273 Yoshihara et al., 2018). Accordingly, MCF10A, MCF7 and MDA-MB-231 cell lines have higher mRNA levels 274 of β4GalT -V compared to the mRNA levels of β4GalT -VI ( Fig.S3A). Our findings show that GOLPH3 275 expression remains unchanged in the different human breast cell lines positive for β4GalT -V NTD, in line 276 with our observations showing no effects on GOLPH3 levels for NTDs of β3GalT -IV and ST3Gal-II GGTs 277 and supporting a ceiling effect. In addition, the presence of β4GalT -VI NTD in the non-tumorigenic 278 MCF10A epithelial cell line does not modify GOLPH3 levels as well. However, the expression of β4GalT -279 VI NTD in both human breast cancer cell lines, MCF7 and MDA-MB-231, promotes a significant reduction 280 in GOLPH3 levels. Localization of GOLPH3 to Golgi membranes requires its binding to PI4P. Furthermore, 281 the PI4P-binding capacity of GOLPH3 promotes metastasis in breast and lung cancer (Tokuda et al., 2014). 282 To identify the molecular underpinnings of these different effects of β4GalT -V/VI CTs on GOLPH3 we 283 performed a computational docking analysis. For β4GalT -V CT the unbiased docking predicts binding to 284 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint the electronegative region of GOLPH3, agreeing with previous reports (Theodoropoulou et al., 2025; Welch 285 et al., 2021), and validating our in silico approach. β4GalT-VI CT, on the other hand, was predicted to bind 286 GOLPH3 near the PI4P -binding domain and the hydrophobic β -hairpin motif. Considering that both the 287 PI4P binding pocket and the hydrophobic β-hairpin motif on the surface of GOLPH3 facing the membrane 288 plane are crucial for its retention in the Golgi complex (Dippold et al., 2009; Rahajeng et al., 2019) , our 289 docking analysis thus suggests that β4GalT-VI CT may interfere with GOLPH3 ability to interact with the 290 Golgi membrane. Further supporting these results, only exchanging a few key amino acids on the CTs of 291 β4GalT-V and β4GalT -VI to match the other are sufficient to switch the preferred binding surface on 292 GOLPH3. 293 Tumors become more invasive and malignant by undergoing the EMT process. High expression of 294 GOLPH3 is observed in several types of cancer and is related to EMT progression and poor prognosis. 295 Vimentin upregulation contributes to the acquisition of invasive and metastatic properties in cancer cells, 296 in part, due to its role in regulating cell migration, cell adhesion and cytoskeletal reorganization 297 (Ostrowska-Podhorodecka et al., 2022; Vuoriluoto et al., 2011; Wu et al., 2018) . GOLPH3 stimulates EMT 298 via the upregulation of vimentin and activation of several signaling pathways (Giansanti & Piergentili, 299 2022; Gong et al., 2022; Li et al., 2022; Scott et al., 2009; Sechi et al., 2020; Tan et al., 2017; Wang et al., 2020) . 300 On the contrary, t he downregulation of GOLPH3 in aggressive endometrial carcinoma cells led to 301 inhibition of vimentin expression and also inhibited the endometrial carcinoma cell invasion in vitro and in 302 vivo by regulating EMT (Wen et al., 2019) . Punicalagin, a bioactive molecule proposed as a promising 303 chemopreventive therapy in breast cancer, also inhibits EMT by regulating GOLPH3 in breast cancer cells 304 lines (Pan et al., 2020) . Herein, we uncovered that the presence of just the NTD of β4GalT -VI promotes a 305 reduction in vimentin and GOLPH3 levels from human melanoma and breast cancer cells. While cells 306 concurrently treated with TGF -β1 and infected with lentivir al particles carrying the β4GalT -VI NTD 307 partially block the upregulation of GOLPH3 and vimentin, cells infected with β4GalT -VI NTD 24 h before 308 TGF-β1 stimulation have completely abolished increase of GOLPH3 and vimentin. Taken together, these 309 findings strongly suggest that the CT of β4GalT-VI NTD interferes with the ability of GOLPH3 to interact 310 with PI4P, and consequently it reduces the levels of GOLPH3 impairing its function in the acquisition of 311 mesenchymal features. 312 The Golgi complex, the central trafficking hub of the cell, orchestrates the post-translational 313 modification and sorting of proteins and lipids within the secretory pathway (Hellicar et al., 2022) . Our 314 current knowledge about the function of GGTs places them as responsible for ganglioside biosynthesis , 315 and imbalances in glycosphingolipids metabolism cause severe metabolic disorders (Cumin et al., 2022; 316 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint Daniotti et al., 2013; Morrison et al., 2025; Vo et al., 2025; Wennekes et al., 2009). Herein, we have determined 317 a novel role of GGTs as regulator s of the oncoprotein GOLPH3. The presence of pre-existing GGTs in the 318 cell and the amino acid sequence information within the NTD of GGTs arise as new players in the 319 mechanisms that control GOLPH3 levels in the cell. This study also opens new avenues to identify specific 320 molecules that interfere and/or inhibit GOLPH3 pathways which will not only provide more specific tools 321 to investigate glycolipid metabolism but also offer new therapeutic approaches with less off-target effects. 322 323 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint

324 325 Cell Lines 326 327 CHO-K1, MDA MB 231, MCF7, MCF10A and SK-MEL-28 cells (ATCC, Manassas, VA, USA) were 328 cultured at 37°C, in a humidified atmosphere with 5% CO2 in Dulbecco’s modified Eagle’s medium 329 (DMEM) supplemented with 10% (v/v) fetal bovine serum and antibiotics (100 µg/mL penicillin and 100 330 µg/mL streptomycin). MCF10 cells were additionally supplemented with 0.5 μg/mL hydrocortisone 331 (H0888.10G; Sigma), 0.1µg/mL cholera toxin (C8052 -1MG; Sigma) and 20ng/ml epidermal growth factor 332 (13247-051; Invitrogen). CHO-K1 cells stable expressing ST8Sia-I and β4GalNAcT-I/β3GalT-IV were 333 obtained as described in (Crespo et al., 2002). 334 335 Electroporation, Transfection and Transduction 336 337 CHO-K1 cells were transfected by electroporation, 5 x 10 5 cells were resuspended in Ingenio® Cuvettes 338 (2mm gap) with 100 µL of electroporation mix (Ruggiero et al., 2022), containing 1µg of the corresponding 339 plasmid and then pulsed in a BTX Electro Cell Manipulator 600 (voltage: 155 V, resistance: 186 W, 340 capacitance: 950 µF). After electroporation, cells were seeded in a 35-mm-diameter dish and allowed for 24 341 h of protein expression. MDA MB 231, MCF7 and MCF10A cells were transfected with Fugene 6 342 transfection reagent and allowed for 24h of protein expression. Lentiviruses for β4GalT-VI (1-52)-GFP was 343 produced in HEK293T cells by cotransfecting of pFUGW vectors and 3 packaging plasmids (pCMV -VSV-344 G, pMDLg/pRRE, pRSV-Rev) using Fugene 6 transfection reagent. Fresh, cleared supernatants containing 345 lentiviruses were used for infection of MCF10A and MDA-MB-231 cells. 346 347 Ganglioside synthesis inhibition 348 349 CHO-K1ST8Sia-I and CHO -K1β4GalNAcT-I/β3GalT-IV cells were treated with 2 μM dl-threo-phenyl-2-350 hexadecanoylamino-3-pyrrolidino-1-propanol (P4) for four days to reduce the glycolipid content , as 351 described previously (Vilcaes et al., 2011) . After treatment, cells were processed for immunofluorescence 352 or western blotting. Inhibition of glycolipid synthesis was observed by immunodetection of GD3 , GM1 or 353 GD1a gangliosides. 354 355 Plasmids 356 357 The plasmids coding for the N-terminal domains ST3Gal-II(1-51)-mCherry, β3GalT-IV(1-52)-YFP and ST8Sia-I(1-57)-YFP, 358 were constructed as described in ref. (Ruggiero et al., 2015) , (Uliana, Crespo, et al., 2006) and (Chumpen 359 Ramirez et al., 2017), respectively. The N-terminal domains of β3GalT-V and β3GalT-VI containing amino 360 acids 1–52 were synthesized by Genscript. Then, these N-terminal domains were subcloned into the FUGW 361 expression plasmid containing the GFP sequence after the C-terminus of the peptides. 362 363 Antibodies 364 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint We used the following primary antibodies: polyclonal rabbit Anti -GOLPH3/MIDAS antibody (ab98023; 365 ABCAM) (WB 1:2000), polyclonal rabbbit anti-GOLPH3 (A13121; Abclonal) (IF 1:1000), monoclonal mouse 366 anti-GM130 (610823; BD Bioscience) (IF 1:500), monoclonal mouse anti-GD1a (#GD1a-1; DSHB) (IF 1:300), 367 monoclonal mouse anti -GD3 (clone R24; gift from P.H.H. Lopez, CIQUIBIC -CONICET-UNC) (1:300), 368 monoclonal mouse anti -Vimentin (AMF17b mAb; DSHB) (IF 1:1000), monoclonal mouse anti -α-tubulin 369 (cat# T9026; Sigma-Aldrich) (WB 1:10000). 370 Secondary fluorochrome -conjugated antibodies were from Thermo Fisher: Alexa Fluor -488–conjugated 371 donkey anti mouse IgG, Alexa Fluor -568–conjugated donkey anti rabbit IgG, Alexa Fluor -488–conjugated 372 donkey anti rabbit IgG, Alexa Fluor -647-conjugated donkey anti mouse IgG. Secondary antibodies were 373 used at a dilution 1:1000. 374 375 Western Blot 376 377 Cells grown in 35-mm dishes for 24 h were harvested and lysed on ice for 20 min in 120 μL of lysis 378 buffer (50 mM Tris-HCl, pH 7.2; 1% Triton X-100; 300 mM NaCl; 1 mM PMSF; cOmplete™ Mini EDTA-free 379 protease inhibitor cocktail, Roche). Lysates were mixed with 4× Laemmli sample buffer (Bio-Rad, Hercules, 380 CA, USA) supplemented with 5% 2 -mercaptoethanol and heated at 95 °C for 10 min. Proteins were 381 separated on 12% SDS –polyacrylamide gels under reducing conditions and transferred to nitrocellulose 382 membranes at 350 mA for 60 min. Membranes were blocked with 5% (w/v) non -fat dry milk in PBS for 60 383 min and incubated overnight at 4 °C with primary antibodies. After three washes with PBS, membranes 384 were incubated with IRDye -conjugated secondary antibodies (goat anti -mouse IgG, 800CW; goat anti -385 rabbit IgG, 680CW; LI -COR) diluted 1:10,000 in PBS for 60 min at room temperature. Protein bands were 386 visualized using an Odyssey infrared imaging system (LI -COR Biotechnology, Lincoln, NE, USA). 387 Molecular weights were estimated using calibrated protein standards run in parallel. 388 389 Confocal Immunofluorescence Microscopy 390 391 For cell surface ganglioside labeling, cells were grown on Lab-Tek II chambered coverglass (Thermofisher), 392 and incubated on ice for 20 min, before adding the corresponding primary antibody and/or Alexa Fluor 393 555-conjugated cholera toxin subunit B (C34776; Thermofisher). After a 90 min incubation, cells were 394 washed three times with DMEM and fixed with 1% (w/v) paraformaldehyde in phosphate-buffered saline 395 (PBS) for 5 min. For intracellular labeling, cells were fixed with 1% (w/v) paraformaldehyde in PBS and 396 permeabilized using Saponin 0,1% in PBS. After blocking with 2% bovine serum albumin (BSA), primary 397 antibodies were diluted in 2% BSA and incubated overnight at 4°C. Secondary antibodies were incubated 398 1 h at room temperature. 399 Confocal images were acquired using an Olympus Fluoview FV -1200 or a Zeiss LSM 980 Confocal 400 Microscope (Carl Zeiss). Confocal z-stacks were collected with an oil immersion objective (63X; NA1.4) and 401 0.25 μm slices. Full z-stacks of GOLPH3, GM130, P230 and vimentin signal were taken in different regions 402 of each coverslip/sample. Fiji 3D ImageJ Suite plugin was used for 3D segmentation and quantification of 403 GOLPH3 levels (whole cell fluorescence intensity ). For vimentin signal, w hole cell fluorescence intensity 404 quantification was performed using Z -projection tool (SUM slices) on the entire stack after background 405 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint subtraction with Rolling ball algorithm. Manders' correlation coefficient was measured using the Just 406 Another Colocalization Plugin in ImageJ software. 407 408 qPCR (Quantitative Real-time PCR) 409 410 SYBR Green-based qPCR was performed as we previously described (Vilcaes et al., 2020). Primers specific 411 for human ST3Gal-II were designed and purchased from Invitrogen (Carlsbad, CA, USA). Reactions were 412 carried out in a Rotor-Gene Q thermocycler (Qiagen, Hilden, Germany). Each 15 μl reaction contained 1 μl 413 of cDNA template, 0.8 μM of each primer, and 7.5 μl of Real Mix (Biodynamics, Buenos Aires, Argentina). 414 Cycling conditions were as follows: initial polymerase activation at 95 °C for 30 s, followed by 40 cycles of 415 denaturation at 95 °C for 30 s, annealing at 60 °C for 30 s, and extension at 72 °C for 30 s. Standard curves 416 were generated in duplicate using 1:5 serial dilutions of cDNA from MCF10A, MCF7 and MDA MB 231 417 cells. All samples were analyzed in triplicate. Melt curve analysis was used to confirm product specificity, 418 and both standard curve linearity and amplification efficiency were optimized. Data were processed using 419 Rotor-Gene Q software (Qiagen, Hilden, Germany), and relative expression levels were normalized to the 420 geometric mean of three reference genes: PUM1, GAPDH and 18S rRNA. 421 422 Protein-Protein docking 423 424 The crystal structure of human GOLPH3 (Wood et al., 2009) was used for docking analyses. The structure 425 was preprocessed in UCSF Chimera v1.19 by removing water molecules and sulfate ions. The cytoplasmic 426 tails of β4GalT -V and β4GalT -VI were obtained from the AlphaFold Protein Structure Database 427 (https://alphafold.ebi.ac.uk, O43286 and AF -Q9UBX8-F1 respectively). Predicted full -length structures 428 were processed in Chimera to retain only the cytosolic regions (residues 1–14). In silico–mutated tails were 429 generated using the AlphaFold Server (https://alphafoldserver.com/) based on the corresponding modified 430 amino acid sequences (CT -β4GalT-V R2S/R4L: MSALRGLLRLPRRS and CT-β4GalT-VI S2R: 431 MRVLRRMMRVSNRS). Protein–protein docking between GOLPH3 and each cytosolic tail was performed 432 using ClusPro (https://cluspro.org/; (Jones et al., 2022)). For biased (or guided) docking runs, residues 247, 433 258, and 262 of GOLPH3 were defined as attracting. The “Electrostatic -favored” models were selected for 434 further analysis. Docking complexes were analyzed in Chimera using the Find Clashes/Contacts tool to 435 identify residue–residue interactions. Contact lists were further processed using a custom Python script 436 (compatible with Chimera’s internal Python 2.7 environment) to classify the interactions into hydrogen 437 bonds, electrostatic interactions, and van der Waals contacts. Hydrogen bonds were defined by a donor –438 acceptor distance ≤ 3.5 Å and D–H···A angle ≥ 120°. Interaction classification was based on distance, atomic 439 overlap, and residue charge properties. 440 441 Bioinformatics analysis 442 443 To explore the transcriptional expression of ganglioside glycosyltransferases in MCF10A, MCF7 and MDA-444 MB-231 cells, we incorporated a dataset from the Gene Expression Omnibus (GEO, 445 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint https://www.ncbi.nlm.nih.gov/geo/): GSE75168; (Messier et al., 2016). Dataset GSE75168 (RNA-Seq of cell 446 lines MCF10A, MCF7 and MDA-MB-231) was used for external validation. 447 448 Statistical Analysis 449 450 Data was plotted and analyzed using Prism software (GraphPad). For statistical analysis, data of individual 451 cells for each independent culture was compiled into sub -columns for each experimental group and then 452 analyzed using Nested ANOVA or Nested t test. For all experiments, when ANOVA test revealed 453 significant effects, suitable post-tests were applied to perform multiple comparisons using the wt or control 454 groups as reference. For simplicity the figures show asterisks for significance (* p<0.05, ** p<0.01, *** 455 p<0.001, **** p<0.0001). 456 457 458 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint Data availability 459 460 The data supporting the findings of this study are included in the paper and its supplemental information 461 and will be available from the corresponding author s (Natali L. Chanaday, email: 462 [email protected]; and A. Alejandro Vilcaes, email: 463 [email protected]) upon request. 464 465

466 467 We would like to thank Eve Gautreaux for her valuable feedback on the manuscript. The authors 468 acknowledge the technical and imaging assistance of Dra. Cecilia Sampedro, Dr. Carlos Mas, Dr. Pilar 469 Crespo and Dr. Gonzalo Quassollo from the Centro de Micro y Nanoscopía de Córdoba – CEMINCO – 470 CONICET – Universidad Nacional de Córdoba, Córdoba, Argentina. http://ceminco.conicet.unc.edu.ar. 471 This work was supported in part by grants from Secretaría de Ciencia y Tecnología (SECyT), Universidad 472 Nacional de Córdoba (UNC), the University of Pennsylvania and the Margaret Q. Landenberger Research 473 Foundation Award to N.L.C. We thank the CDB Microscopy Core (RRID SCR_022373) of the University of 474 Pennsylvania for the use of their instruments. 475 476 477 Author contributions: N. Martínez-Koteski, S. Rasino and F.M. Ruggiero: conceptualization, data curation, 478 formal analysis, investigation, methodology, validation, visualization, and writing—original draft, review, 479 and editing. P.H.H. Lopez and G.D. Fidelio: conceptualization, resources, and writing —review and 480 editing. A.A. Vilcaes and N.L. Chanaday: conceptualization, data curation, formal analysis, 481 investigation, funding acquisition, methodology, project administration, resources, supervision, 482 visualization, and writing—original draft, review, and editing. 483 484 Disclosures: The authors declare no competing interests. 485 486 487 488 489 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint Figure legends 490 491 Figure 1. Stable expression of Ganglioside glycosyltransferases in CHO -K1 cells increase GOLPH3 492 expression. (A) Schematic representation of topology and domain organization of type II transmembrane 493 glycosyltransferases. CT - cytoplasmic tail, TMD - Trans Membrane Domain and SR - Stem Region. (B) 494 Biosynthesis pathway for 0 -, a- and b-series gangliosides. Colored boxes indicate the main gangliosides 495 expressed in CHO -K1 (grey), CHO -K1ST8Sia-I (yellow) and CHO -K1β4GalNAcT-I/β3GalT-IV (pink) cells. (C) 496 Immunofluorescence showing GD3 (yellow) and GM1/GD1a (red/magenta) gangliosides expression in 497 CHO-K1ST8Sia-I and CHO-K1β4GalNAcT-I/β3GalT-IVcells respectively. (D-E) Representative confocal images of CHO-498 K1 cells stained with GOLPH3 (green), DAPI (blue) and GM130 or P230 ( red; D and E respectively). Bar 499 charts showing Manders’ coefficient (red overlapped with green ). (F-G) Endogenous expression of 500 GOLPH3 in CHO -K1, CHO-K1ST8Sia-I and CHO-K1β4GalNAcT-I/β3GalT-IV cells, analyzed by immunofluorescence 501 (F) and western blot (G). For immunofluorescence, statistical analysis was performed using nested one-502 way ANOVA with Tukey multiple comparisons . Bars represent the mean ± SD of three independent 503 experiments. For immunoblot, statistical significance was determined using two-tailed, unpaired t test (ns: 504 not significant). Scale bars: 10 µm. 505 506 Figure 2. N-terminal domain of GGTs modify GOLPH3 levels independently of ganglioside synthesis 507 in CHO-K1 cells. (A-B) Western Blot analysis and quantification of GOLPH3 expression in CHO-K1 cells 508 transfected with β3Galt-IV(1-52)-YFP and ST3Gal -II(1-51)mcherry. Data represent the mean ± SEM of three 509 independent experiments; statistical significance was determined using one -way ANOVA. (C) 510 representative images of endogenous expression of GD3 and GD1a/GM1 gangliosides in CHO-K1ST8Sia-I and 511 CHO-K1β4GalNAcT-I/β3GalT-IV cells (respectively) treated with P4 (+ P4) or vehicle ( -P4). (D-E) Western Blot 512 analysis and quantification of GOLPH3 expression from cells following the P4 treatment mentioned in C. 513 Data represent the ± SD of three independent experiments; statistical significance was determined using 514 two-tailed, unpaired t test (ns: not significant). Scale bars: 10 µm. 515 516 Figure 3. GOLPH3 expression is regulate d by NTDs in a cell type -dependent manner. (A-C) 517 Immunofluorescence and quantification of MCF10A, MCF7 and MDA -MB-231 cells transfected (+) with 518 NTDs of ST8Sia -I (A), β3GalT-IV (B), and ST3Gal -II (C) glycosyltransferases . Cells were stained for 519 GOLPH3 (grey) and DAPI (blue). Data represent the mean ± SD of three independent experiments, with 520 statistical significance evaluated by nested t-test (ns: not significant). Scale bars: 10 µm. 521 522 Figure 4. The N -terminal domain of β4GalT-V and β4GALT-VI exert opposite effects on GOLPH3 523 expression in breast cancer cell lines . (A-B) Immunofluorescence of MCF10A, MCF7 and MDA -MB-231 524 cells expressing the NTDs of β4GalT-V (A) and β4GALT-VI (B) glycosyltransferases stained for GOLPH3 525 (grey) and DAPI (blue). Arrowheads (green) indicate cells transfected with β4GALT-V(1-52)-GFP or β4GALT-526 VI(1-52)-GFP. Comparison of GOLPH3 levels (untransfected vs transfected cells) is presented as the mean ± SD 527 of three independent experiments, with statistical significance determined by nested t -test (ns: not 528 significant). Scale bars: 10 µm. 529 530 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint Figure 5. β4GalT-VI N -terminal domain prevents the acquisition of EMT features. (A) Schematic 531 representation of the experimental procedure performed in B . (B) TGF-β1 treatment was achieved 532 simultaneously with the lentiviral transduction of β4GALT-VI(1-52)-GFP. Representative confocal images of 533 MCF10A (top) and MDA-MB-231 (bottom) cells treated with TGF-β1, TGF-β1+lentiviral particles of 534 β4GALT-VI(1-52)-GFP or without TGF-β1 treatment (control). (C) Schematic representation of the experimental 535 procedure performed in D. (D) Lentiviral transduction of β4GALT-VI(1-52)-GFP was achieved 24 h before TGF-536 β1 treatment. Representative confocal images of MDA-MB-231 cells treated with TGF-β1, TGF-β1+lentiviral 537 particles of β4GALT-VI(1-52)-GFP or without TGF-β1 treatment (control). Arrowheads (green) indicate cells 538 infected with β4GALT-VI(1-52)-GFP. Comparison of GOLPH3 (gray) and vimentin (magenta) levels was 539 performed using nested one -way ANOVA with Tukey multiple comparisons. Bars represent the mean ± 540 SD of three independent experiments. Nuclei are stained with DAPI (blue). Scale bars: 10 µm. 541 542 Figure 6. Protein–protein docking predicts differential binding of β4GalT-V and β4GalT-VI cytoplasmic 543 tails to GOLPH3 . (A) Homology comparison between the cytoplasmic tails of β4GalT-V and β4GalT-VI. 544 Arginine-rich regions are shown in blue. (B) On the left, Scheme showing the binding of GOLPH3 to 545 membrane (Dippold et al., 2009; Rahajeng et al., 2019; Welch et al., 2021) . Arrows indicates PI4P binding 546 pocket and hydrophobic β -hairpin motif. On the right, surface charge of GOLPH3 (generated with 547 Chimera; red, negative; blue, positive) exposes a flat surface containing an electronegative zone (red, 548 negative; blue, positive) (Dippold et al., 2009; Welch et al., 2021) . (C-D) Global (C) and biased global (D) 549 docking of β4GalT-V (blue) and β4GalT-VI (black) cytoplasmic tails with GOLPH3. Biased global docking 550 was performed by defining D247, D258 and D262 GOLPH3 residues as attractive sites, as described 551 previously (Theodoropoulou et al., 2025) . GOLPH3 R174 and E175 residues are highlighted in cyan and 552 orange, respectively. Side chains of putative amino acid residues in the β4GalT-V cytoplasmic tail that 553 interact with GOLPH3 are shown. (E) mutants from β4GalT-V (R2S/R4L, blue) and β4GalT-VI (S2R, black) 554 cytoplasmic tails were docked with GOLPH3 using ClusPro Global Docking . Side chains of mutated 555 residues in β4GalT-V and β4GalT-VI cytoplasmic tails are highlighted in pink and blue, respectively. 556 557 558 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint Supplemental data 559 560 Figure S1. GOLPH3 and vimentin levels in MDA -MB-231 and SK -MEL-28 cells expressing β4GALT-561 VI(1-52)-GFP. (A-B) Immunofluorescence of MDA-MB-231 (A) and SK-MEL-28 (B) cells expressing β4GALT-562 VI(1-52)-GFP stained for vimentin (magenta), GOLPH3 (grey) and DAPI (blue). Arrowheads (green) indicate 563 cells transfected with β4GALT-VI(1-52)-GFP. (C) Comparison of GOLPH3 and vimentin levels was performed 564 using nested t-test. Bars represent the mean ± SD of three (MDA-MB-231, bars on the left ) and two (SK -565 MEL-28, bars on the right) independent experiments. Scale bars: 10 µm. 566 567 Figure S2. Ganglioside Glycosyltransferases gene expression profiling in human breast cell lines. (A) 568 Biosynthesis pathway for Globo series and; a- and b-series gangliosides. Enzymes (highlighted in pink) 569 that synthesize GD3 (yellow), GM1 (red), GD1a (green) and SSEA4 (magenta) are shown. (B-D and F) 570 Dataset GSE75168 (Messier et al., 2016) containing RNA-Seq of MCF10A, MCF7 and MDA -MB-231 cell 571 lines, was used for external validation. Relative gene expression of ST8SIA1 (B), B3GALT4 (D) and 572 ST3GAL2 (F, bottom bar charts) is shown. Each bar in the graph (averages of 3 replicates) represents the 573 expression measurement extracted from the TPM normalized expression counts . (C-E-G- and H) 574 Immunofluorescence showing GD3, GM1, GD1a and SSEA4 expression in MCF10A, MCF7 and MDA-MB-575 231 cells. (F) Top Bar charts, represent relative levels of ST3Gal-II transcripts analyzed by RT -qPCR. Total 576 RNA was purified and reverse -transcribed from three cell lines. The expression values for RT-qPCR are 577 given relative to the expression levels of 18S rRNA. Data represent mean ± SD from three biological 578 replicates, each in triplicate. Scale bars: 10 µm. 579 580 Figure S3. B4GALT5 and B4GALT5 gene expression profiling in human breast cell lines. (A) Dataset 581 GSE75168 (Messier et al., 2016) containing RNA-Seq of MCF10A, MCF7 and MDA -MB-231 cell lines, was 582 used for external validation. Each bar in the graph (averages of 3 replicates ) represents the expression 583 measurement extracted from the TPM normalized expression counts. 584 585 586 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint

587 588 Cavdarli, S., Yamakawa, N., Clarisse, C., Aoki, K., Brysbaert, G., Le Doussal, J. M.,…Groux -Degroote, S. 589 (2020). Profiling of. Int J Mol Sci, 21(1). https://doi.org/10.3390/ijms21010370 590 Cazet, A., Groux-Degroote, S., Teylaert, B., Kwon, K. M., Lehoux, S., Slomianny, C.,…Delannoy, P. (2009). 591 GD3 synthase overexpression enhances proliferation and migration of MDA-MB-231 breast cancer 592 cells. Biol Chem, 390(7), 601-609. https://doi.org/10.1515/BC.2009.054 593 Chumpen Ramirez, S., Ruggiero, F. M., Daniotti, J. L., & Valdez Taubas, J. (2017). Ganglioside 594 glycosyltransferases are S-acylated at conserved cysteine residues involved in homodimerisation. 595 Biochem J, 474(16), 2803-2816. https://doi.org/10.1042/BCJ20170124 596 Crespo, P. M., Zurita, A. R., & Daniotti, J. L. (2002). Effect of gangliosides on the distribution of a 597 glycosylphosphatidylinositol-anchored protein in plasma membrane from Chinese hamster ovary-598 K1 cells. J Biol Chem, 277(47), 44731-44739. https://doi.org/10.1074/jbc.M204604200 599 Cumin, C., Huang, Y. L., Rossdam, C., Ruoff, F., Céspedes, S. P., Liang, C. Y.,…Jacob, F. (2022). 600 Glycosphingolipids are mediators of cancer plasticity through independent signaling pathways. 601 Cell Rep, 40(7), 111181. https://doi.org/10.1016/j.celrep.2022.111181 602 Daniotti, J. L., Martina, J. A., Giraudo, C. G., Zurita, A. R., & Maccioni, H. J. (2000). GM3 alpha2,8 -603 sialyltransferase (GD3 synthase): protein characterization and sub -golgi location in CHO-K1 cells. 604 J Neurochem, 74(4), 1711-1720. 605 Daniotti, J. L., Vilcaes, A. A., Torres Demichelis, V., Ruggiero, F. M., & Rodriguez -Walker, M. (2013). 606 Glycosylation of Glycolipids in Cancer: Basis for Development of Novel Therapeutic Approaches. 607 Front Oncol, 3, 306. https://doi.org/10.3389/fonc.2013.00306 608 Deshmukh, A. P., Vasaikar, S. V., Tomczak, K., Tripathi, S., den Hollander, P., Arslan, E.,…Mani, S. A. 609 (2021). Identification of EMT signaling cross-talk and gene regulatory networks by single-cell RNA 610 sequencing. Proc Natl Acad Sci U S A, 118(19). https://doi.org/10.1073/pnas.2102050118 611 Dippold, H. C., Ng, M. M., Farber -Katz, S. E., Lee, S. K., Kerr, M. L., Peterman, M. C.,…Field, S. J. (2009). 612 GOLPH3 bridges phosphatidylinositol -4- phosphate and actomyosin to stretch and shape the 613 Golgi to promote budding. Cell, 139(2), 337-351. https://doi.org/10.1016/j.cell.2009.07.052 614 Giansanti, M. G., & Piergentili, R. (2022). Linking GOLPH3 and Extracellular Vesicles Content -a Potential 615 New Route in Cancer Physiopathology and a Promising Therapeutic Target is in Sight? Technol 616 Cancer Res Treat, 21, 15330338221135724. https://doi.org/10.1177/15330338221135724 617 Gong, L. Y., Tu, T., Zhu, J., Hu, A. P., Song, J. W., Huang, J. Q.,…Chen, Y. (2022). Golgi phosphoprotein 3 618 induces autophagy and epithelial -mesenchymal transition to promote metastasis in colon cancer. 619 Cell Death Discov, 8(1), 76. https://doi.org/10.1038/s41420-022-00864-2 620 Hassinen, A., Pujol, F. M., Kokkonen, N., Pieters, C., Kihlström, M., Korhonen, K., & Kellokumpu, S. (2011). 621 Functional organization of Golgi N - and O -glycosylation pathways involves pH -dependent 622 complex formation that is impaired in cancer cells. J Biol Chem , 286(44), 38329 -38340. 623 https://doi.org/10.1074/jbc.M111.277681 624 Hellicar, J., Stevenson, N. L., Stephens, D. J., & Lowe, M. (2022). Supply chain logistics - the role of the Golgi 625 complex in extracellular matrix production and maintenance. J Cell Sci , 135(1). 626 https://doi.org/10.1242/jcs.258879 627 Iglesias-Artola, J. M., Böhlig, K., Schuhmann, K., Cook, K. C., Lennartz, H. M., Schuhmacher, M.,…Nadler, 628 A. (2025). Quantitative imaging of lipid transport in mammalian cells. Nature, 646(8084), 474-482. 629 https://doi.org/10.1038/s41586-025-09432-x 630 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint Jones, G., Jindal, A., Ghani, U., Kotelnikov, S., Egbert, M., Hashemi, N.,…Kozakov, D. (2022). Elucidation 631 of protein function using computational docking and hotspot analysis by ClusPro and FTMap. Acta 632 Crystallogr D Struct Biol, 78(Pt 6), 690-697. https://doi.org/10.1107/S2059798322002741 633 Kuna, R. S., & Field, S. J. (2019). GOLPH3: a Golgi phosphatidylinositol(4)phosphate effector that directs 634 vesicle trafficking and drives cancer. J Lipid Res, 60(2), 269-275. https://doi.org/10.1194/jlr.R088328 635 Lesniak, A. M., Ye, Z., & Banfield, D. K. (2025). Identification of the client -binding site on the Golgi 636 membrane protein adaptor Vps74/yGOLPH3. iScience, 28(10), 113494. 637 https://doi.org/10.1016/j.isci.2025.113494 638 Li, H., Guo, L., Chen, S. W., Zhao, X. H., Zhuang, S. M., Wang, L. P.,…Song, M. (2012). GOLPH3 639 overexpression correlates with tumor progression and poor prognosis in patients with clinically 640 N0 oral tongue cancer. J Transl Med, 10, 168. https://doi.org/10.1186/1479-5876-10-168 641 Li, Y., Zhang, H., Long, W., Gao, M., Guo, W., & Yu, L. (2022). Inhibition of NLRP3 and Golph3 ameliorates 642 diabetes-induced neuroinflammation. Aging (Albany NY) , 14(21), 8745 -8762. 643 https://doi.org/10.18632/aging.204363 644 Liu, C. Y., Lin, H. H., Tang, M. J., & Wang, Y. K. (2015). Vimentin contributes to epithelial -mesenchymal 645 transition cancer cell mechanics by mediating cytoskeletal organization and focal adhesion 646 maturation. Oncotarget, 6(18), 15966-15983. https://doi.org/10.18632/oncotarget.3862 647 Lo, N. W., Shaper, J. H., Pevsner, J., & Shaper, N. L. (1998). The expanding beta 4-galactosyltransferase gene 648 family: messages from the databanks. Glycobiology, 8(5), 517 -526. 649 https://doi.org/10.1093/glycob/8.5.517 650 Lopez, P. H., & Schnaar, R. L. (2009). Gangliosides in cell recognition and membrane protein regulation. 651 Curr Opin Struct Biol, 19(5), 549-557. https://doi.org/S0959-440X(09)00092-X [pii] 652 10.1016/j.sbi.2009.06.001 653 Maccioni, H. J., Quiroga, R., & Ferrari, M. L. (2011). Cellular and molecular biology of glycosphingolipid 654 glycosylation. J Neurochem, 117(4), 589-602. https://doi.org/10.1111/j.1471-4159.2011.07232.x 655 Maccioni, H. J., Quiroga, R., & Spessott, W. (2011). Organization of the synthesis of glycolipid 656 oligosaccharides in the Golgi complex. FEBS Lett , 585(11), 1691 -1698. 657 https://doi.org/10.1016/j.febslet.2011.03.030 658 McCormick, C., Duncan, G., Goutsos, K. T., & Tufaro, F. (2000). The putative tumor suppressors EXT1 and 659 EXT2 form a stable complex that accumulates in the Golgi apparatus and catalyzes the synthesis of 660 heparan sulfate. Proc Natl Acad Sci U S A, 97(2), 668-673. https://doi.org/10.1073/pnas.97.2.668 661 Messier, T. L., Gordon, J. A., Boyd, J. R., Tye, C. E., Browne, G., Stein, J. L.,…Stein, G. S. (2016). Histone H3 662 lysine 4 acetylation and methylation dynamics define breast cancer subtypes. Oncotarget, 7(5), 5094-663 5109. https://doi.org/10.18632/oncotarget.6922 664 Morrison, T. A., Vigee, J., Tovar, K. A., Talley, T. A., Mujal, A. M., Kono, M.,…O'Shea, J. J. (2025). Selective 665 requirement of glycosphingolipid synthesis for natural killer and cytotoxic T cells. Cell, 188(13), 666 3497-3512.e3416. https://doi.org/10.1016/j.cell.2025.04.007 667 Ostrowska-Podhorodecka, Z., Ding, I., Norouzi, M., & McCulloch, C. A. (2022). Impact of Vimentin on 668 Regulation of Cell Signaling and Matrix Remodeling. Front Cell Dev Biol , 10, 869069. 669 https://doi.org/10.3389/fcell.2022.869069 670 Pan, L., Duan, Y., Ma, F., & Lou, L. (2020). Punicalagin inhibits the viability, migration, invasion, and EMT 671 by regulating GOLPH3 in breast cancer cells. J Recept Signal Transduct Res , 40(2), 173 -180. 672 https://doi.org/10.1080/10799893.2020.1719152 673 Rahajeng, J., Kuna, R. S., Makowski, S. L., Tran, T. T. T., Buschman, M. D., Li, S.,…Field, S. J. (2019). Efficient 674 Golgi Forward Trafficking Requires GOLPH3-Driven, PI4P-Dependent Membrane Curvature. Dev 675 Cell, 50(5), 573-585.e575. https://doi.org/10.1016/j.devcel.2019.05.038 676 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint Rizzo, R., Russo, D., Kurokawa, K., Sahu, P., Lombardi, B., Supino, D.,…Luini, A. (2021). Golgi maturation-677 dependent glycoenzyme recycling controls glycosphingolipid biosynthesis and cell growth via 678 GOLPH3. EMBO J, 40(8), e107238. https://doi.org/10.15252/embj.2020107238 679 Rodriguez-Walker, M., Vilcaes, A. A., Garbarino -Pico, E., & Daniotti, J. L. (2015). Role of plasma -680 membrane-bound sialidase NEU3 in clathrin -mediated endocytosis. Biochem J , 470(1), 131 -144. 681 https://doi.org/10.1042/BJ20141550 682 Rosales Fritz, V. M., Daniotti, J. L., & Maccioni, H. J. (1997). Chinese hamster ovary cells lacking GM1 and 683 GD1a synthesize gangliosides upon transfection with human GM2 synthase. Biochim Biophys Acta, 684 1354(2), 153-158. https://doi.org/10.1016/s0167-4781(97)00117-6 685 Ruggiero, F. M., Martínez -Koteski, N., Cavieres, V. A., Mardones, G. A., Fidelio, G. D., Vilcaes, A. A., & 686 Daniotti, J. L. (2022). Golgi Phosphoprotein 3 Regulates the Physical Association of Glycolipid 687 Glycosyltransferases. Int J Mol Sci, 23(18). https://doi.org/10.3390/ijms231810354 688 Ruggiero, F. M., Vilcaes, A. A., Iglesias-Bartolomé, R., & Daniotti, J. L. (2015). Critical role of evolutionarily 689 conserved glycosylation at Asn211 in the intracellular trafficking and activity of sialyltransferase 690 ST3Gal-II. Biochem J. https://doi.org/10.1042/BJ20150072 691 Scott, K. L., Kabbarah, O., Liang, M. C., Ivanova, E., Anagnostou, V., Wu, J.,…Chin, L. (2009). GOLPH3 692 modulates mTOR signalling and rapamycin sensitivity in cancer. Nature, 459(7250), 1085 -1090. 693 https://doi.org/10.1038/nature08109 694 Sechi, S., Frappaolo, A., Karimpour -Ghahnavieh, A., Piergentili, R., & Giansanti, M. G. (2020). Oncogenic 695 Roles of GOLPH3 in the Physiopathology of Cancer. Int J Mol Sci , 21(3). 696 https://doi.org/10.3390/ijms21030933 697 Seko, A., & Yamashita, K. (2005). Characterization of a novel galactose beta1,3 -N-698 acetylglucosaminyltransferase (beta3Gn-T8): the complex formation of beta3Gn -T2 and beta3Gn -699 T8 enhances enzymatic activity. Glycobiology, 15(10), 943-951. https://doi.org/10.1093/glycob/cwi082 700 Spessott, W., Crespo, P. M., Daniotti, J. L., & Maccioni, H. J. (2012). Glycosyltransferase complexes improve 701 glycolipid synthesis. FEBS Lett, 586(16), 2346-2350. https://doi.org/10.1016/j.febslet.2012.05.041 702 Tan, X., Banerjee, P., Guo, H. F., Ireland, S., Pankova, D., Ahn, Y. H.,…Kurie, J. M. (2017). Epithelial -to-703 mesenchymal transition drives a pro -metastatic Golgi compaction process through scaffolding 704 protein PAQR11. J Clin Invest, 127(1), 117-131. https://doi.org/10.1172/JCI88736 705 Tenorio, M. J., Ross, B. H., Luchsinger, C., Rivera -Dictter, A., Arriagada, C., Acuña, D.,…Mardones, G. A. 706 (2016). Distinct Biochemical Pools of Golgi Phosphoprotein 3 in the Human Breast Cancer Cell 707 Lines MCF7 and MDA -MB-231. PLoS One , 11(4), e0154719. 708 https://doi.org/10.1371/journal.pone.0154719 709 Theodoropoulou, A., Nasrallah, A., Abriata, L. A., Abrami, L., Talotta, F., Marcaida, M. J.,…D’Angelo, G. 710 (2025). Molecular Regulation and Physiological Role of GOLPH3 -mediated Golgi retention. 711 bioRxiv, 2025.2006.2026.661665. https://doi.org/10.1101/2025.06.26.661665 712 Theodossiou, T. A., Ali, M., Grigalavicius, M., Grallert, B., Dillard, P., Schink, K. O.,…Berg, K. (2019). 713 Simultaneous defeat of MCF7 and MDA-MB-231 resistances by a hypericin PDT-tamoxifen hybrid 714 therapy. NPJ Breast Cancer, 5, 13. https://doi.org/10.1038/s41523-019-0108-8 715 Tokuda, E., Itoh, T., Hasegawa, J., Ijuin, T., Takeuchi, Y., Irino, Y.,…Takenawa, T. (2014). 716 Phosphatidylinositol 4-phosphate in the Golgi apparatus regulates cell-cell adhesion and invasive 717 cell migration in human breast cancer. Cancer Res, 74(11), 3054-3066. https://doi.org/10.1158/0008-718 5472.CAN-13-2441 719 Uliana, A. S., Crespo, P. M., Martina, J. A., Daniotti, J. L., & Maccioni, H. J. (2006). Modulation of GalT1 and 720 SialT1 sub -Golgi localization by SialT2 expression reveals an organellar level of glycolipid 721 synthesis control. J Biol Chem, 281(43), 32852-32860. https://doi.org/10.1074/jbc.M605805200 722 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint Uliana, A. S., Giraudo, C. G., & Maccioni, H. J. (2006). Cytoplasmic tails of SialT2 and GalNAcT impose 723 their respective proximal and distal Golgi localization. Traffic, 7(5), 604 -612. 724 https://doi.org/10.1111/j.1600-0854.2006.00413.x 725 Usman, S., Waseem, N. H., Nguyen, T. K. N., Mohsin, S., Jamal, A., Teh, M. T., & Waseem, A. (2021). 726 Vimentin Is at the Heart of Epithelial Mesenchymal Transition (EMT) Mediated Metastasis. Cancers 727 (Basel), 13(19). https://doi.org/10.3390/cancers13194985 728 van Meer, G., & Sprong, H. (2004). Membrane lipids and vesicular traffic. Curr Opin Cell Biol , 16(4), 373-729 378. https://doi.org/10.1016/j.ceb.2004.06.004 730 Varki, A., Cummings, R. D., Esko, J. D., Stanley, P., Hart, G. W., Aebi, M.,…Seeberger, P. H. (2022). 731 Essentials of Glycobiology. In. https://doi.org/NBK579905 732 Vilcaes, A. A., Demichelis, V. T., & Daniotti, J. L. (2011). Trans -activity of plasma membrane -associated 733 ganglioside sialyltransferase in mammalian cells. J Biol Chem , 286(36), 31437 -31446. 734 https://doi.org/10.1074/jbc.M111.257196 735 Vilcaes, A. A., Garbarino-Pico, E., Torres Demichelis, V., & Daniotti, J. L. (2020). Ganglioside Synthesis by 736 Plasma Membrane -Associated Sialyltransferase in Macrophages. Int J Mol Sci , 21(3). 737 https://doi.org/10.3390/ijms21031063 738 Vishwanathan, N., Yongky, A., Johnson, K. C., Fu, H. Y., Jacob, N. M., Le, H.,…Hu, W. S. (2015). Global 739 insights into the Chinese hamster and CHO cell transcriptomes. Biotechnol Bioeng, 112(5), 965-976. 740 https://doi.org/10.1002/bit.25513 741 Vo, H. G., Gonzalez -Escamilla, G., Mirzac, D., Rotaru, L., Herz, D., Groppa, S., & Bindila, L. (2025). 742 Extended coverage of human serum glycosphingolipidome by 4D -RP-LC TIMS-PASEF unravels 743 association with Parkinson's disease. Nat Commun, 16(1), 4567. https://doi.org/10.1038/s41467-025-744 59755-6 745 Vuoriluoto, K., Haugen, H., Kiviluoto, S., Mpindi, J. P., Nevo, J., Gjerdrum, C.,…Ivaska, J. (2011). Vimentin 746 regulates EMT induction by Slug and oncogenic H-Ras and migration by governing Axl expression 747 in breast cancer. Oncogene, 30(12), 1436-1448. https://doi.org/10.1038/onc.2010.509 748 Wang, K., Jiang, S., Huang, A., Gao, Y., Peng, B., Li, Z.,…Li, W. (2020). GOLPH3 Promotes Cancer Growth 749 by Interacting With STIP1 and Regulating Telomerase Activity in Pancreatic Ductal 750 Adenocarcinoma. Front Oncol, 10, 575358. https://doi.org/10.3389/fonc.2020.575358 751 Wang, Z., Jiang, B., Chen, L., Di, J., Cui, M., Liu, M.,…Su, X. (2014). GOLPH3 predicts survival of colorectal 752 cancer patients treated with 5 -fluorouracil-based adjuvant chemotherapy. J Transl Med , 12, 15. 753 https://doi.org/10.1186/1479-5876-12-15 754 Welch, L. G., Peak-Chew, S. Y., Begum, F., Stevens, T. J., & Munro, S. (2021). GOLPH3 and GOLPH3L are 755 broad-spectrum COPI adaptors for sorting into intra -Golgi transport vesicles. J Cell Biol , 220(10). 756 https://doi.org/10.1083/jcb.202106115 757 Wen, Y., Tan, X., Wu, X., Wu, Q., Qin, Y., Liang, M.,…Xie, R. (2019). Golgi phosphoprotein 3 (GOLPH3) 758 promotes endometrial carcinoma cell invasion and migration by regulating the epithelial -759 mesenchymal transition. Cancer Biomark, 26(1), 21-30. https://doi.org/10.3233/CBM-190096 760 Wennekes, T., van den Berg, R. J., Boot, R. G., van der Marel, G. A., Overkleeft, H. S., & Aerts, J. M. (2009). 761 Glycosphingolipids--nature, function, and pharmacological modulation. Angew Chem Int Ed Engl , 762 48(47), 8848-8869. https://doi.org/10.1002/anie.200902620 763 Wood, C. S., Schmitz, K. R., Bessman, N. J., Setty, T. G., Ferguson, K. M., & Burd, C. G. (2009). PtdIns4P 764 recognition by Vps74/GOLPH3 links PtdIns 4 -kinase signaling to retrograde Golgi trafficking. J 765 Cell Biol, 187(7), 967-975. https://doi.org/10.1083/jcb.200909063 766 Wu, S., Du, Y., Beckford, J., & Alachkar, H. (2018). Upregulation of the EMT marker vimentin is associated 767 with poor clinical outcome in acute myeloid leukemia. J Transl Med , 16(1), 170. 768 https://doi.org/10.1186/s12967-018-1539-y 769 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint Xing, M., Peterman, M. C., Davis, R. L., Oegema, K., Shiau, A. K., & Field, S. J. (2016). GOLPH3 drives cell 770 migration by promoting Golgi reorientation and directional trafficking to the leading edge. Mol 771 Biol Cell, 27(24), 3828-3840. https://doi.org/10.1091/mbc.E16-01-0005 772 Xue, Y., Wu, G., Liao, Y., Xiao, G., Ma, X., Zou, X.,…Huang, R. (2014). GOLPH3 is a novel marker of poor 773 prognosis and a potential therapeutic target in human renal cell carcinoma. Br J Cancer , 110(9), 774 2250-2260. https://doi.org/10.1038/bjc.2014.124 775 Yoshihara, T., Satake, H., Nishie, T., Okino, N., Hatta, T., Otani, H.,…Asano, M. (2018). Lactosylceramide 776 synthases encoded by B4galt5 and 6 genes are pivotal for neuronal generation and myelin 777 formation in mice. PLoS Genet, 14(8), e1007545. https://doi.org/10.1371/journal.pgen.1007545 778 Zeng, Z., Lin, H., Zhao, X., Liu, G., Wang, X., Xu, R.,…Song, L. (2012). Overexpression of GOLPH3 779 promotes proliferation and tumorigenicity in breast cancer via suppression of the FOXO1 780 transcription factor. Clin Cancer Res , 18(15), 4059-4069. https://doi.org/10.1158/1078-0432.CCR-11-781 3156 782 Zhu, K., Fan, J., Cai, H., Zhou, C., Gong, Z., Li, Z., & Yu, J. (2025). The highly expressed. J Gastrointest Oncol, 783 16(2), 415-434. https://doi.org/10.21037/jgo-2025-193 784 785 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint P230/DAPI GOLPH3 merge D A B GlcCer- ceramide Gal GalNAc Neu5Ac Cer_ _ Cer_ _ _ _Cer_ _ _ _Cer_ _ _ _ _Cer_ _ _ __ GM3 GM2 GM1 GD1a Cer_ GlcCer LacCer ST3Gal-Vβ4GalT-VI _Cer_ _ _Cer_ _ _ _Cer_ _ _ __ GA2 GA1 GM1b β3GalT-IV β4GalNAcT-I ST3Gal-II “a” Cer_ _ _ _ GD3 Cer_ _ _ _ _ GD2 Cer_ _ _ _ _ _ GD1b GT1b Cer_ _ _ _ _ __ ST8Sia-I “b” “0” CHO-K1wt G F CHO-K1wt CHO-K1ST8Sia-I CHO-K1β4GalNAcT-I/β3GalT-IV GOLPH3GOLPH3 GOLPH3 0.0 0.5 1.0 1.5 2.0 wt β4GalNAcT-I/β3GalT-IV ST8Sia-I ns GOLPH3 (fold of change) merge GM130/DAPI GOLPH3 TMD SR CT Lipid Bilayer Golgi Lumen N-terminal domain (NTD) C-catalytic domain Cytoplasm CHO-K1ST8Sia-I CHO-K1β4GalNAcT-I/β3GalT-IV C Mander's coefficientP230 0.0 0.2 0.4 0.6 0.8 GM130 0.0 0.2 0.4 0.6 0.8Mander's coefficient E GD3 GD1aGM1 37 - 50 - GOLPH3 α-tubulin kDa CHO-K1 ST8Sia-I - + - β4GalNAcT-I - - + β3GalT-IV - - + /gid00192 0 1 2 3GOLPH3 (fold of change)wt ST8Sia-I β4GalNAcT-I/β3GalT-IV ✱ ✱ Figure 1 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint A B C CHO-K1ST8Sia-I CHO-K1β4GalNAcT-I/β3GalT-IV -P4 +P4 -P4 -P4+P4 +P4 GD3 GD3 GD1a GM1 GD1a GM1 D 37 - 50 - kDa GOLPH3 α-tubulin ns 0.0 0.5 1.0 1.5 ns -P4 +P4 -P4 +P4 CHO-K1ST8Sia-I CHO-K1β4GalNAcT-I/β3GalT-IV - +P4 CHO-K1ST8Sia-I CHO-K1β4GalNAcT-I/β3GalT-IV - + E GOLPH3 expression (fold change) GOLPH3 expression (fold change) control ST3Gal-II (1-51)-mCherry β3GalT-IV (1-52)-YFP 0.0 0.5 1.0 1.5 2.0 - + - - - + ST3Gal-II(1-51)-mCherry β3GalT-IV(1-52)-YFP GOLPH3 α-tubulin 37 - 50 - kDa CHO-K1 ✱ ✱ Figure 2 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint A B C GOLPH3 fluorecence Intensity (fold change) 0.0 0.5 1.0 1.5 ✱control 0.0 0.5 1.0 1.5 ✱ 0.0 0.5 1.0 1.5 ns control ST8Sia-I(1-57)-YFP control ST8Sia-I(1-57)-YFP ST8Sia-I(1-57)-YFP MCF10AMCF7MDA-MB-231 β3GalT-IV(1-52)-YFP GOLPH3GOLPH3 ( )- ( )+ β3GalT-IV(1-52)-YFP /gid00192 /gid00192 MCF10AMCF7MDA-MB-231 GOLPH3 GOLPH3 GOLPH3 GOLPH3GOLPH3 ST3GalT-II(1-51)-mCherry ( )- ( )+ ST3GalT-II(1-51)-mCherry /gid00192 /gid00192 MCF10A MCF7 MDA-MB-231 GOLPH3 fluorecence Intensity (fold change) control β3GalT-IV (1-52)-YFP 0.0 0.5 1.0 1.5 ns 0.0 0.5 1.0 1.5 ns control β3GalT-IV (1-52)-YFP 0.0 0.5 1.0 1.5 ns control β3GalT-IV (1-52)-YFP 0.0 0.5 1.0 1.5 ns ST3GalT-II (1-51) control 0.0 0.5 1.0 1.5 ns ST3GalT-II (1-51) control 0.0 0.5 1.0 1.5 ns ST3GalT-II (1-51) control GOLPH3 fluorecence Intensity (fold change) MCF10A MCF7 MDA-MB-231 MCF10AMCF7MDA-MB-231 ST8Sia-I(1-57)-YFP GOLPH3 GOLPH3 ST8Sia-I(1-57)-YFP ( )- ( )+ /gid00192 /gid00192 Figure 3 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint A MCF10A MDA-MB-231MCF7 GOLPH3 GM130 β4GalT-VI(1-52)-GFP GOLPH3 β4GalT-VI(1-52)-GFP GM130 GOLPH3 GM130β4GalT-VI(1-52)-GFP transfecteduntransfected B GOLPH3 GM130β4GalT-V(1-52)-GFP GOLPH3 GM130β4GalT-V(1-52)-GFP GOLPH3 GM130β4GalT-V(1-52)-GFP GOLPH3 (fold change) MCF10A MDA-MB-231MCF7 ns 0.0 0.5 1.0 1.5 ✱ ✱ 0.0 0.5 1.0 1.5 ✱ ✱ 0.0 0.5 1.0 1.5 transfecteduntransfected ns 0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5 ns 0.0 0.5 1.0 1.5 ns GOLPH3 (fold change) GOLPH3 (fold change) GOLPH3 (fold change)GOLPH3 (fold change) GOLPH3 (fold change) Figure 4 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint A 0 5 10 15 20 25 ✱ ✱ ✱ ✱ ✱ 0 5 10 15 ✱ ✱ ✱ ✱ ✱ ✱ GOLPH3 (fold change) vimentin (fold change) 0 5 10 15 ✱ ✱ ✱ ✱ 0 2 4 6 8 ✱ ✱ ✱ ✱ vimentin (fold change)GOLPH3 (fold change) C lentivirus (β4GalT-VI(1-52)-GFP) IF Day 0 Day 2 Day 4 B TGF-β1 control TGF-β1+β4GalT-VI(1-52)-GFP MDA-MB-231 GOLPH3 GOLPH3 GOLPH3 vimentinvimentin vimentin MCF10A GOLPH3 GOLPH3 vimentinvimentin vimentin GOLPH3 GOLPH3 GOLPH3 vimentin vimentin vimentin control lentivirus (β4GalT-VI(1-52)-GFP) IF Day 0 Day 2 Day 4Day -1 TGF-β D 0 5 10 15 20 ns ✱✱ vimentin (fold change)GOLPH3 (fold change) 0 5 10 15 ns ✱✱ ✱✱ TGF-β1 TGF-β TGF-β1 TGF-β1 control TGF-β1 β4GalT-VI(1-52)-GFP+TGF-β1 ✱✱ control TGF-β1+β4GalT-VI(1-52)-GFPTGF-β1 MDA-MB-231 TGF-β1+β4GalT-VI(1-52)-GFP GOLPH3 Figure 5 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint L4 R5 M1 R6 R9 R13 ClusPro Global Docking β4GalT-V CT β4GalT-VI CT ClusPro Biased Global Docking ClusPro Global DockingE B C D A PI4P binding pocket β-hairpin β-hairpin GOLPH3 90o PI4P binding pocket Lipid Bilayer /gid00192 negative charged surface positive charged surface R174 E175 L195 L196 F197 N-term. L4 R5 M1 R6 R9 R13 β4GalT-V CT β4GalT-VI CT R174 R175 L195 L196 F197 N-term. β4GalT-V CT(R2S/R4L) β4GalT-VI CT(S2R) R2S R4L R174 R175 L195 L196 F197 N-term. S2R cytoplasmic tail (CT) MS V L R R N MM V R S RS MRAR R G R L L L R P RS Similarity: 9/14 (60.0%) β4GalT-VI β4GalT-V N-terminus (N-term.) Figure 6 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint A C MDA-MB-231 SK-MEL-28 transfected non-transfected 0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5 MDA-MB-231 GOLPH3 SK-MEL-28 ✱ ✱ vimentin β4GalT-VI(1-52)-GFP/ GOLPH3vimentin β4GalT-VI(1-52)-GFP/ GOLPH3vimentin β4GalT-VI(1-52)-GFP/ GOLPH3vimentin β4GalT-VI(1-52)-GFP/ B vimentin (fold change)vimentin (fold change) Figure S1 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint B D GD3 GM1 MCF10A MCF7 GD3 GM1 MDA-MB-231 GD3 GM1 GM1 GM1 GM1 MCF10A MCF7 MDA-MB-231 F C E MCF10A MDA-MB-231 MCF7 0.0 0.1 0.2 0.3 0.4 0.5relative gene expression (TPM) ST8SIA1 β4GalT-VI Cer_ _ _ GM3 _Cer_ _ _ GM2 _Cer_ _ _ _ GM1 _Cer_ _ _ __ GD1a Cer_ GlcCer Cer_ _ LacCer ST3Gal-V β3GalT-IVβ4GalNAcT-I ST3Gal-II “a” Cer_ _ _ Cer_ _ _ _ Cer_ _ _ _ _ _Cer_ _ _ _ _ β3GalNAcT-IαGalT-I β3GalT-V ST3Gal-II Gb3 Gb4 SSEA3 SSEA4 ST8Sia-I Cer_ _ _ _ GD3 “b” Cer_ _ _ _ _ GD2 Cer_ _ _ _ _ _ GD1b GT1b Cer_ _ _ _ _ __ Globo series A G B3GALT4 0 2 4 6 8relative gene expression (TPM) H SSEA4 SSEA4 SSEA4 MCF10A MCF7 MDA-MB-231 GD1a GD1a GM1 GD1a GM1 GM1 MCF10A MCF7 MDA-MB-231 0.0 0.5 1.0 1.5 Relative mRNA (ST3GAL2/18S) levels ST3GAL2 GSE75168 MCF10A MDA-MB-231 MCF7 GSE75168 ST3GAL2 0 2 4 6relative gene expression (TPM) MCF10A MDA-MB-231 MCF7 GSE75168 MCF10A MDA-MB-231 MCF7 GM1 GM1 GM1 Figure S2 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint A B4GALT5B4GALT6 0 2 4 15 30 45 60relative gene expression (TPM) MCF10A MDA-MB-231 MCF7 GSE75168 Figure S3 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2025.12.31.695843doi: bioRxiv preprint