A conserved MFS orchestrates a subset of O-glycosylation to facilitate macrophage dissemination and tissue invasion

Aberrant display of the truncated core1 O-glycan T-antigen is a common feature of human cancer cells that correlates with metastasis. Here we show that T-antigen in Drosophila melanogaster macrophages is involved in their developmentally programmed tissue invasion. Higher macrophage T-antigen levels require an atypical major facilitator superfamily (MFS) member that we named Minerva which enables macrophage dissemination and invasion. We characterize for the first time the T and Tn glycoform O-glycoproteome of the Drosophila melanogaster embryo, and determine that Minerva increases the presence of T-antigen on protein pathways previously linked to cancer, most strongly on the protein sulfhydryl oxidase Qsox1 which we show is required for macrophage invasion. Minerva’s vertebrate ortholog, MFSD1, rescues the minerva mutant’s migration and T-antigen glycosylation defects. We thus identify a key conserved regulator that orchestrates O-glycosylation on a protein subset to activate a program governing migration steps important for both development and cancer metastasis.


INTRODUCTION 35 36
The set of proteins expressed by a cell defines much of its potential capacities. However, 37 a diverse set of modifications can occur after the protein is produced to alter its function 38 and thus determine the cell's final behavior. One of the most frequent, voluminous and 39 variable of such alterations is glycosylation, in which sugars are added onto the oxygen 40 (O) of a serine or threonine or onto the nitrogen (N) of an asparagine (Kornfeld and 41 Kornfeld, 1985;Marshall, 1972;Ohtsubo and Marth, 2006). O-linked addition can occur 42 on cytoplasmic and nuclear proteins in eukaryotes (Comer and Hart, 2000;Hart et al., 43 2011), but the most extensive N-and O-linked glycosylation occurs during the transit of 44 a protein through the secretory pathway. A series of sugar molecules are added starting in 45 the endoplasmic reticulum (ER) or cis-Golgi and continuing to be incorporated and 46 removed until passage through the trans Golgi network is complete (Aebi, 2013;Stanley 47 et al., 2009). N-linked glycosylation is initiated in the ER at consensus NxS/T X≠P site, 48 whereas the most common GalNAc-type O-linked glycosylation is initiated in the early 49 Golgi and glycosites display no clear sequence motifs, apart from a prevalence of 50 neighboring prolines (Bennett et al., 2012;Christlet and Veluraja, 2001). Glycosylation 51 can affect protein folding, stability and localization as well as serve specific roles in fine-52 tuning protein processing and functions such as protein adhesion and signaling (Goth et 53 al., 2018;Varki, 2017). The basic process by which such glycosylation occurs has been 54 well studied. However our understanding of how specific glycan structures participate in 55 modulating particular cellular functions is still at its beginning. 56 The need to understand the regulation of O-glycosylation is particularly relevant 57 for cancer (Fu et al., 2016;Häuselmann and Borsig, 2014). The truncated O-glycans 58 called T and Tn antigen are not normally found on most mature human cells (Cao et al., 59 1996) but up to 95% of cells from many cancer types display these at high levels (Boland 60 et al., 1982;Cao et al., 1996;Howard and Taylor, 1980;Limas and Lange, 1986; Orntoft 61 the whole embryo because we were unable to obtain enough protein from FACSed 277 macrophages or to isolate CRISPR-induced full knockouts of minerva in the S2R+ 278 macrophage-like cell line. We observed that several bands detected with the anti-T mAb 279 were absent or reduced in the minerva mutant (Fig 4A), indicating an effect on a subset 280 of proteins. We wished to obtain a more comprehensive view of the proteins affected by 281 Minerva. Since there is little information about Drosophila O-glycoproteins and O-282 glycosites (Schwientek et al., 2007;Aoki and Tiemeyer, 2010), we used lectin-enriched 283 O-glycoproteomics to identify proteins displaying T and Tn glycoforms in Stage 11/12 284 embryos from wild type and mrva 3102 mutants (Fig S4A). We labeled tryptic digests of 285 embryonic protein extracts from control or mutant embryos with stable dimethyl groups 286 carrying medium (C 2 H 2 D 4 ) or light (C 2 H 6 ) isotopes respectively to allow each genotype to 287 be identified in mixed samples (Boersema et al., 2009;Schjoldager et al., 2012Schjoldager et al., , 2015. 288 The pooled extracts were passed over a Jacalin column to enrich for T and Tn O-289 glycopeptides; the eluate was analyzed by mass spectrometry to identify and quantify T 290 and Tn modified glycopeptides in the wild type and the mutant sample through a 291 comparison of the ratio of the light and medium isotope labeling channels for each 292 glycopeptide. In the wild type we identified T and Tn glycopeptides at 936 glycosites 293 derived from 270 proteins (Table S1 and Fig 4B). 62% of the identified O-glycoproteins 294 and 77% of identified glycosites contained only Tn O-glycans. 33% of the identified O-295 glycoproteins and 23% of glycosites displayed a mixture of T or Tn O-glycans, and 5% 296 of identified O-glycoproteins and 4% of glycosites had solely T O-glycans (Fig 4C). In 297 agreement with previous studies (Steentoft et al., 2013), only one glycosite was found in 298 most of the identified O-glycoproteins (44%) (Fig 4D). In 20% we found two sites, and 299 some glycoproteins had up to 27 glycosites. The identified O-glycosites were mainly on 300 threonine residues, (78.5%) with some on serines (21.2%) and very few on tyrosines 301 (0.3%) (Fig S4B). Metabolism, cuticle development, and receptors were the most 302 common functional assignments for the glycoproteins (Fig S4C). To assess the changes 303 in glycosylation in the mrva mutant we utilized two cutoffs, a three-fold and a more 304 stringent ten-fold cutoff. The majority of the quantifiable Tn and T O-glycoproteome was 305 unaltered between the wild type and the mrva 3102 mutant, with only 63 proteins (23%) 306 showing more than a three-fold change and 18 (6%) a ten-fold shift (Fig 4F). We 307 observed both increases and decreases in the levels of T and Tn modification on proteins 308 in the mutant (Fig 4F,G, Table S2), but a greater number of proteins showed decreased 309 than increased T antigen levels. 67% of the vertebrate orthologs of Drosophila proteins 310 displaying shifts in this O-glycosylation have previously been linked to cancer (Fig 4H,  311 Table S2). These proteins were affected at specific sites, with 40% of glycosites on these 312 proteins changed more than three fold and only 14% more than ten fold. The glycosite 313 shifts in T antigen occurred either without significant alterations in Tn (33% of glycosites 314 had only decreased T antigen, 17% of glycosites had only increased T antigen) or with 315 changes in T antigen occurring in the same direction as the changes in Tn (22% of 316 glycosites both Tn and T antigen increased, 22% of glycosites both Tn and T decreased) 317 (Table S2). Only 1% of glycosites displayed decreased T antigen with a significant 318 increase in Tn. Interestingly, a higher proportion of the glycoproteins with altered O-319 glycosylation in the mrva 3102 mutant had multiple glycosites than the general 320 glycoproteome (Fig 4D) (P value=0.005 for ten-fold changes). We conclude that Minerva 321 affects O-glycosylation occupancy on a small subset of O-glycoproteins, many of whose 322 vertebrate orthologs have been linked to cancer, with both T and Tn O-glycopeptides 323 being affected. 324 325

Minerva raises T antigen levels on proteins required for invasion 326
Given that the knockdown of the C1GalTA enzyme which blocks Tn to T conversion 327 produced a germband invasion defect, we examined the known functions of the 18 328 proteins with lower T antigen in the absence of Minerva to distinguish which processes 329 Minerva could influence to facilitate invasion (Fig 4H). We excluded two proteins 330 involved in eggshell and cuticle production. To spot proteins whose reduced T antigen-331 containing glycopeptides are caused directly by alterations in glycosylation rather than 332 indirectly by decreased protein expression in the mrva mutant, we checked if 333 glycosylation at other identified glycosites was unchanged or increased. We identified ten 334 such proteins, several of which were in pathways that had been previously linked to 335 invasion in vertebrates. Qsox1, a predicted sulfhydryl oxidase required for the secretion, 336 and thus potential folding of EGF repeats (Tien et al., 2008) showed the strongest 337 alterations of any protein, with a 50-fold decrease in T antigen levels in the mrva mutant.

12
The mammalian ortholog has been shown to affect disulfide bond formation, is 339 overexpressed in some cancers, promotes Matrigel invasion, and can serve as a negative 340 341 prognostic indicator in human cancer patients (Chakravarthi et al., 2007;Katchman et al., 342 2011;Lake and Faigel, 2014). Dtg, with a 13-fold reduction in T antigen (Hodar et al., 343 2014), and Put with a five-fold reduction (Letsou et al., 1995) respond to signaling by the 344 BMP-like ligand, Dpp. Gp150 shows a four fold decrease in T antigen and modulates 345 Notch signaling (Fetchko et al., 2002;Li, 2003). Notch and BMP promote invasion and 346 metastasis in mice (Bach et al., 2018;Garcia and Kandel, 2012;Owens et al., 2015;347 Pickup et al., 2015;Sahlgren et al., 2008;Sonoshita et al., 2011). Dpp signaling directs 348 histoblast invasion in the fly (Ninov et al., 2010). To test if Qsox1, the protein with the 349 strongest changes in T antigen in the minerva mutant is required for germband invasion, 350 we examined RNAi knockdown of Qsox1 in macrophages and a P element mutant in the 351 5'UTR of the Qsox1 gene. In both cases we observed reduced numbers of macrophages 352 in the germband (Fig 4I,J) (30% for RNAi and 42% for mutant) and a concomitant 353 increase of macrophages on the neighboring yolk (Fig S4D,E). There was no change in 354 total cell number in RNAi knockdown embryos (Fig S4F). For technical reasons we did 355 not examine this in the P element mutant line which only grew robustly when combined 356 with a cytoplasmic macrophage marker. We conclude that Mrva is required to increase T 357 O-glycans on a subset of the glycosites of selected glycoproteins involved in protein 358 folding, glycosylation and signaling in pathways frequently linked to promoting cancer 359 metastasis. Its strongest effect is on a predicted sulfhydryl oxidase which is required in 360 macrophages for their germband invasion, the Drosophila ortholog of the mammalian 361 cancer protein, QSOX1. 362 363

Conservation of Minerva's function in macrophage invasion and T antigen 364 modification by its mammalian ortholog MFSD1 365
To determine if our studies could ultimately be relevant for mammalian biology and 366 therefore also cancer research, we searched for a mammalian ortholog. MFSD1 from mus 367 musculus, shows strong sequence similarity with Mrva, with 50% of amino acids 368 displaying identity and 68% conservation (Fig 5A, Fig S5A). A transfected C-terminally 369 with the Golgi marker GRASP65 in murine MC-38 colon carcinoma cells (Fig 5B, Fig  371   S5C-D). mmMFSD1 expression in macrophages in mrva 3102 mutant embryos can 372 completely rescue the germband invasion defect (Fig 5C,D). This macrophage-specific 373 expression of MFSD1 also resulted in higher levels of T antigen on macrophages when 374 compared to those in mrva 3102 mutants (Fig 5E,F) identifying a key regulator of this O-glycosylation, Minerva, with an unexpected role for 387 a member of the major facilitator superfamily. As O-glycosites cannot as yet be reliably 388 predicted, our proteomic characterization in a highly genetically accessible organism will 389 permit future studies on how glycosylation affects cell behavior; we highlight T and Tn 390 O-glycosylated receptors in Table 1 to further this goal. Our demonstration that a 391 conserved protein affects invasion and the appearance of the cancer-associated core1 T 392 glycoform on a set of proteins connected to invasion may have implications for cancer. 393 394

Modifications of the O-glycoproteome by an MFS family member 395
Our identification of a MFS family member as a regulator of O-glycosylation is 396 surprising. MFS family members can serve as transporters and shuttle a wide variety of 397 substrates (Quistgaard et al., 2016;Reddy et al., 2012). Minerva is localized to the Golgi 398 and displays homology to sugar transporters; Minerva could thus affect O-glycosylation 399 through substrate availability. However, the lower and higher levels of glycosylation in 400 the mrva 3102 mutant we observe are hard to reconcile with this hypothesis. Given that the 401 changes in T antigen on individual glycosites in the mrva mutant are found either with no 402 significant change in Tn or with a change in the same direction (Table S2), regulation 403 appears to occur at the initial GalNAc addition on the protein subset as well as on further 404 T antigen elaboration. 95% of the proteins with 10-fold altered glycosylation in the mrva 405 mutant had multiple O-glycosylation sugar modifications compared to 56% of the general 406 O-glycoproteome. Greatly enhanced glycosylation of protein sequences containing an 407 existing glycan modification is observed for some GalNAc-Ts due to a lectin 408 domain (Hassan et al., 2000;Kubota et al., 2006;Revoredo et al., 2016) (Boskovski et al., 2015); the Drosophila 418 ortholog of the responsible GalNAc transferase is also essential for embryogenesis 419 (Bennett et al., 2010;Schwientek et al., 2002). Thus the changed glycosylation we 420 observe on components of the Notch and Dpp pathways could alter transcription 421 (Hamaratoglu et al., 2014;Ntziachristos et al., 2014), shifting protein levels and thereby 422 changing the ratio of some glycopeptides in the mrva mutant relative to the wild type. 423 Proteins in which glycosylation at other sites is unchanged or changed in the opposite 424 direction are those most likely to be directly affected by Minerva. Such proteins include 425 ones involved in protein folding and O-glycan addition and removal (Fig 4I) (Tien et al., 426 2008). If changes in the glycosylation of these proteins alters their specificity or activity, 427 some of the shifts we observe in our glycoproteomic analysis could be indirect in a 428 different way; an initial effect of Minerva on the glycosylation of regulators of protein 429 folding and glycosylation could change how these primary Minerva targets affect the 430 glycosylation of a second wave of proteins. 431  Hoefler G, Guertl B. 2012. miR-192, miR-194, miR-215, miR-200c Table S1 and Table S2.   Valoskova et al. Figure S2 fixed S2