A rhomboid protease, EhROM1, regulates the submembrane distribution and size of the Gal/GalNAc lectin subunits in the human protozoan parasite, Entamoeba histolytica

Entamoeba histolytica is a food- and waterborne parasite that is the causative agent of amebic dysentery and amoebic liver abscesses. Adhesion is one of the most important virulence functions as it facilitates motility, colonization of host, destruction of host tissue, and uptake of nutrients by the parasite. One well-characterized parasite cell surface adhesin is the Gal/GalNAc lectin, which binds to galactose or N-acetylgalactosamine residues on host components and is composed of heavy (Hgl), intermediate (Igl), and light (Lgl) subunits. Igl has been shown to be constitutively localized to lipid rafts (cholesterol-rich membrane domains), whereas Hgl and Lgl transiently associate with rafts. When all three subunits are localized to rafts there is an increase in galactose-sensitive adhesion. Thus, submembrane location may regulate the function of this adhesion. Rhomboid proteases are a conserved family of intramembrane proteases that also participate in the regulation of parasite-host interactions. In E. histolytica, one rhomboid protease, EhROM1, cleaves Hgl as a substrate, and knockdown of its expression inhibits parasite-host interactions. Since rhomboid proteases are found within membranes, it is not surprising that lipid composition regulates their activity and enzyme-substrate binding. Given the importance of the lipid environment for both rhomboid proteases and the Gal/GalNAc lectin, we sought to gain insight into the relationship between rhomboid proteases and submembrane location of the lectin in E. histolytica. We demonstrated that EhROM1, itself, is enriched in rafts. Reducing rhomboid protease activity, either pharmacologically or genetically, correlated with an enrichment of Hgl and Lgl in rafts. In a mutant cell line with reduced EhROM1 expression, there was also a significant augmentation of the level of all three Gal/GalNAc subunits on the cell surface and an increase in the molecular weight of Hgl and Lgl. Overall, the study provides insight into the molecular mechanisms governing parasite-host adhesion for this pathogen.

that were captured by avidin affinity chromatography were resolved by SDS-PAGE and analyzed for Hgl, 169 Lgl, Igl and actin by western blotting and densitometry as described above. 170 171 Immunoprecipitation 173 Triton-insoluble membrane was isolated from wildtype and T-EhROM1-s cells as described above and 174 fractionated on sucrose gradients. Fractions 11-12 (lipid rafts), 18-19 (actin-rich membrane) and TSS were 175 used for immunoprecipitation assays. Combined fractions 11 and 12, 18 and 19 or TSS were pre-cleared 176 by incubation with 1 X 10 7 Dynabeads magnetic beads conjugated to sheep α-mouse IgG (Invitrogen Dynal 177 AS, Oslo Norway) at 4°C for 2 h. The beads were collected in a microfuge tube magnetic separation stand 178 (Promega, Madison, WI) and discarded. Protease inhibitors (as described above) and a mixture of 179 monoclonal α-Hgl antibodies (1G7:3F4; ratio 3:1 to give a final antibody concentration of 0.01 μg/μl) were 180 added to the pre-cleared fractions.

Statistical analyses
189 All values are given as means ± standard deviations (SD). Statistical analyses were performed using 190 GraphPad Prism V.6.05 with a one-way analysis of variance (ANOVA) and a Tukey-Kramer multiple 191 comparison test. P values of less than 0.05 were considered statistically significant. P values less than 0.01 192 were considered highly statistically significant. To determine if rhomboid protease activity influences the submembrane distribution of the subunits of the 204 Gal/GalNAc lectin, we characterized the lipid rafts, as previously described [18], in trophozoites exposed 205 to DCI, a compound which inhibits rhomboid proteases by alkylating an active-site histidine [34]. The lipid 206 composition of rafts confers detergent resistance. Therefore, purification of lipid rafts was initiated by 207 extraction with cold Triton X-100. This resulted in the isolation of detergent-resistant membrane (DRM), 208 which consists of both lipid raft and actin-rich membrane. Since the buoyant density of lipid rafts is less 209 than that of actin-rich membrane, these two membrane domains were further separated by sucrose density 210 gradient centrifugation. 211 212 Western blot analysis of sucrose gradient fractions revealed that the majority of Igl was found in a low-213 density region (fractions 9 to 14) ( Fig. 1). Previously, these fractions were shown to possess the highest 214 levels of cholesterol, and thus are considered lipid rafts [18]. The localization of Igl to these low-density 215 rafts is consistent with previous reports [4,18,19]. In control cells, the majority of Hgl and Lgl was 216 associated with less buoyant, actin-rich fractions (fractions 17 to 20) (Fig. 1). However, after exposure to 217 DCI, there was an increase in the proportion of Hgl and Lgl that was localized to lipid raft fractions 218 (fractions 9 to 14), whereas the submembrane distribution of Igl remained unchanged (Fig. 1). Thus, 219 inhibition of rhomboid protease activity correlates with an enrichment of Hgl and Lgl in lipid rafts in E. histolytica. Although, Lgl has not been identified as a EhROM1 substrate, the DCI-induced altered 221 distribution of this subunit may be the result of its covalent connection to Hgl. 222

223
Since chemical inhibitors may have off-target effects, we also used a genetic approach to corroborate the 224 findings made after DCI treatment. Specifically, we isolated lipid rafts and characterized the sub-membrane 225 distribution of the lectin subunits in a cell line (T-EhROM1-s) with reduced expression of EhROM1 [33]. 226 Like DCI-treated cells, T-EhROM1-s cells exhibited augmented levels of Hgl and Lgl in lipid rafts (Fig.  227 2). These data support the notion that there is an inverse relationship between the level of rhomboid consensus sequence is the "cholesterol recognition amino acid consensus" (CRAC) sequence, which 240 consists of a leucine or valine residue, followed by 1 to 5 other amino acids, a central tyrosine, 1-5 other 241 amino acids, and finally an arginine or lysine residue (L/V-X(1-5)-Y-X(1-5)-R/K). The second consensus 242 sequence is a "mirror image" of the first consensus sequence and thus, is referred to as CARC (R/K-X(1-5)-243 There is a central conserved tyrosine in CRAC sequences and a central conserved 244 tyrosine or phenylalanine in CARC sequences. In either case, these residues seems to be essential for Cell surface levels of the Gal/GalNAc subunits are higher in T-EhROM1-s cells 251 We have previously shown that enrichment of Gal/GalNAc lectin subunits in rafts correlates with increased 252 galactose-sensitive adhesion [19]. It would follow that T-EhROM1-s cells exhibit increased adhesion to 253 host cells. Contrary to this prediction, it was reported that inhibition of expression of EhROM1 correlates 254 with defects related to reduced adhesion to host cells [32,33]. The purification of rafts by detergent 255 extraction and sucrose gradient centrifugation is carried out with whole cells. Consequently, the protocol 256 does not distinguish between the adhesins in cell surface lipid rafts versus those in intracellular lipid rafts. 257

Despite raft-enrichment, one reason that T-EhROM1-s cells may have impaired adhesion is because less 258
Gal/GalNAc lectin is in cell surface rafts and more Gal/GalNac lectin is trapped in intracellular rafts. 259 Therefore, we quantified the surface levels of Hgl, Lgl and Igl in T-EhROM1-s cells using biotinylation as 260 previously described [19]. Whole cells were exposed to sulfo-NHS-SS-biotin to label surface proteins. Cell 261 lysates were subjected to avidin-agarose affinity chromatography. Equivalent fractions of the starting cell 262 lysate and affinity purified protein were analyzed for the lectin subunits by western blotting. Surprisingly, 263 surface biotinylation and affinity chromatography revealed that there were significantly higher levels of all 264 three subunits on the surface of mutant cells compared to control cells (Fig. 4) reflection of the different techniques used to quantify cell-surface proteins. We also observed an increase 268 in the level of intracellular lectin subunits. These would be found in the flow-through fractions after 269 biotinylation (Fig. 4). These increases were statistically significant for Igl (*P<0.05), but not for Hgl and 270 Lgl. These data are consistent with that of Baxt et al., [32], who reported an increase, albeit not statistically 271 significant, in the level of Hgl as detected by the ELISA method. Together, with the increases seen for the cell surface-localized subunits in this study, these observations suggest that there may be more Hgl, Lgl, 273 and Igl, overall in the mutant cell line. In the mutant cell line, Hgl was larger than its counterpart in control cells (Fig. 5). This was true for this 283 subunit in all membrane domains (i.e., rafts, actin-rich membrane, and detergent-sensitive membrane) (Fig.  284   5). However, we were surprised to see that the size of Igl was also increased in all membrane domains in 285 the mutant cell line (Fig. 5). These data support the conclusion that Hgl is a substrate of EhROM1 and 286 suggest that Igl is also a substrate of this protease. There is evidence that rhomboid proteases can act on 287 hydrophilic sequences outside of transmembrane domains [44][45][46]. Thus, it is conceivable that GPI-linked 288 Igl could be a substrate of EhROM1 even though it is predicted to reside completely outside of the 289 membrane. To test this, we examined interaction among the subunits using an immunoprecipitation approach. 295 Detergent-sensitive membrane and various fractions of detergent-resistant membrane, resolved by sucrose 296 gradient centrifugation, were mixed with anti-Hgl monoclonal antibody. Following magnetic purification, 297 the immune complexes were resolved by SDS-PAGE and analyzed by western blotting using α-Hgl, α-Lgl, 298 or α-Igl antibodies. In all membrane types in the mutant cells, there was no alteration in the interaction 299 profiles of the Gal/GalNAc lectin subunits (Fig. 6). Therefore, the increased size of Hgl and Igl in the T-300 EhROM1-s cell line did not seem to hinder the assembly of trimers, and thus, cannot explain the adhesion 301 defect. Perhaps improper cleavage of the subunits masks crucial binding sites and/or alters the three-302 dimensional structure of the trimer, in a manner that inhibits its function. 303 304

305
The major finding of this study is that rhomboid protease activity may regulate the subunit size and the sub- this conserved sequence. Accordingly, other mechanisms for rhomboid-substrate recognition must exist 367 [50]. For instance, cleavage of thrombomodulin by vertebrate rhomboid-2, RHBDL2, is directed by its 368 cytoplasmic tail rather than by sequences in its transmembrane region [51]. Furthermore, a recent study of 369 the activity of ten diverse rhomboid proteases (1 prokaryotic and 9 eukaryotic), in an in vitro reconstituted 370 membrane system, demonstrates that the requirement for specific sequences in substrates is much less 371 stringent than previously thought and instead hydrolysis is likely driven, in part, by substrate concentration 372 (rate-driven) [52]. Thus, co-localization of Igl and EhROM1 in rafts may be sufficient for Igl to be 373 hydrolyzed by EhROM1. In fact, the environment of the lipid raft, itself, may contribute to the ability of EhROM1 to target Igl or other substrates that lack the canonical cleavage signal. Two previous studies 375 [53,54] demonstrated that it was possible to transmute non-substrates into substrates simply by changing 376 membrane composition with membrane-disrupting agents, including the cholesterol-binding agent, methyl-377 β-cyclodextrin.   densitometric scans (n ≥ 3), reported as a percentage of total Hgl, Igl, or Lgl in WT cells, which was 612 arbitrarily set to 100%. There is a statistically significant increase of all three subunits on the surface of the