Cellulose binding and the timing of expression influence protein targeting to the double-layered cyst wall of Acanthamoeba

Abstract

The cyst wall of the eye pathogen Acanthamoeba castellanii contains cellulose and has ectocyst and endocyst layers connected by conical ostioles. Cyst walls contain families of lectins that localize to the ectocyst layer (Jonah) or the endocyst layer and ostioles (Luke and Leo). How lectins and an abundant laccase bind cellulose and why proteins go to locations in the wall are not known and are the focus of the studies here. Structural predictions identified β-jelly-roll folds (BJRFs) of Luke and sets of four disulfide knots (4DKs) of Leo, each of which contains linear arrays of aromatic amino acids, also present in carbohydrate-binding modules of bacterial and plant endocellulases. Ala mutations showed that these aromatics are necessary for cellulose binding and proper localization of Luke and Leo in the Acanthamoeba cyst wall. BJRFs of Luke, 4DKs of Leo, a single β-helical fold (BHF) of Jonah, and a copper oxidase domain of the laccase each bind to glycopolymers in both layers of deproteinated cyst walls. Promoter swaps showed that ectocyst localization does not just correlate with but is caused by early encystation-specific expression, while localization in the endocyst layer and ostioles is caused by later expression. Evolutionary studies showed distinct modes of assembly of duplicated domains in Luke, Leo, and Jonah lectins and suggested Jonah BHFs originated from bacteria, Luke BJRFs share common ancestry with slime molds, while 4DKs of Leo are unique to Acanthamoeba.IMPORTANCEAcanthamoebae is the only human parasite with cellulose in its cyst wall and conical ostioles that connect its inner and outer layers. Cyst walls are important virulence factors because they make Acanthamoebae resistant to surface disinfectants, hand sanitizers, contact lens sterilizers, and antibiotics applied to the eye. The goal here was to understand better how proteins are targeted to specific locations in the cyst wall. To this end, we identified three new proteins in the outer layer of the cyst wall, which may be targets for diagnostic antibodies in corneal scrapings. We used structural predictions and mutated proteins to show linear arrays of aromatic amino acids of two unrelated wall proteins are necessary for binding cellulose and proper wall localization. We showed early expression during encystation causes proteins to localize to the outer layer, while later expression causes proteins to localize to the inner layer and the ostioles.

Keywords: Acanthamoeba; AlphaFold structure; cellulose binding; cyst wall proteins; domain evolution; promoter swap.

Fig 1

A linear array of three aromatic amino acids in the β-jelly-roll folds (BJRFs) of Luke-2 bind cellulose and direct the protein to the endocyst layer and ostioles. (A) Drawings show Luke-2 and Luke-3 contain a signal peptide (purple) and two or three BJRFs (orange), respectively, which are separated by Ser-rich spacers (yellow). (B) AlphaFold structure with the confidence colored shows that Luke-2 contains two BJRFs, which are separated by a 47-aa, unstructured spacer rich in Ser (yellow). (C) Luke-3 has three BJRFs separated by 42- and 33-aa long, unstructured spacers that are also Ser-rich. (D) Foldseek shows the N-half BJRF of Luke-2 shares a similar structure with a BJRF of Dictyostelium (DDB_G0292224) wall protein (E-value = 4.99e−7 for foldseek, RMSD = 1.79, and 30.3% identity over a 121-amino acid overlap), which together suggest a relatively recent common ancestry. Importantly, the BJRFs of Luke-2 and the Dictyostelium wall protein each have three aromatic amino acids that form a linear array (pink arrows). (E) The N-half BJRF of Luke-2 also shares three aromatic acids with the CBM2 of a Cellulomonas fimi endoglucanase (blue), the structure of which has been solved (PDB 1EXH). The E-value for the N-terminal BJRF of Luke-2 and the CBM2 of C. fimi is 9.72e−5; the RMSD is 5.01; and the positional is 16% identity over a 105-aa overlap, which together suggest a more remote ancestry. (F and G) Surface views of N- and C-half BJRFs of Luke-2 highlight linear arrays of Trp (blue) and Phe (pink), which were mutated to Ala in MBP fusions made in the periplasm of E. coli or in GFP-tagged proteins expressed in transfected Ac that are encysting. (H) Western blots show that MBP alone (red arrow) fails to bind to Avicel cellulose, while full-length MBP–Luke-2 and MBP–Leo-A fusion proteins (blue arrows) each bind well to Avicel cellulose. In contrast, cellulose binding is lost when six aromatics are mutated to Ala in both Luke-2 and Leo-A. The extra lower molecular weight bands in the WT and Ala mutants of Leo-A are likely MBP only (red arrows), as only the higher molecular weight intact WT fusion protein binds to Avicel. (I) Confocal microscopy shows WT Luke-2–GFP localizes to the endocyst layer, which is labeled with CFW (pink arrowheads), and forms a flat ring around ostioles (green arrowheads). Here and in Fig. 2 to 4, the ectocyst layer (yellow arrowheads) is labeled with WGA that binds to chitin. (J) When five Tyr and a Phe are mutated to Ala, Ala-mut Luke-2–GFP no longer localizes to endocyst layer or ostioles. These experiments support the idea that the Ala-mutant of Luke-2 mislocalizes in cyst walls because of the loss of cellulose binding. Confocal micrographs here and in Fig. 2 to 5 were shot with a ×60 objective, and 3D reconstructions were made from 0.1-µm optical sections. Scale bars for panels I and J are each 5 µm.

Fig 2

Although unique sets of four disulfide knots (4DKs) of Leo lectins show no resemblance to BJRFs of Luke, linear arrays of three aromatic amino acids bind cellulose and direct the protein to the endocyst layer and ostioles (Leo-A) or ectocyst layer [Leo-S(c)]. (A) Drawings show Leo-A contains a signal peptide (purple) and two adjacent sets of four 4DKs (blue), while Leo-S(c), which was corrected by the new transcriptome, contains a Thr-rich spacer (light blue) between sets of 4DKs. (B) AlphaFold with the confidence colored shows Leo-A is composed of two adjacent sets of 4DKs (red spheres), which connect short, parallel loops. Foldseek had no hit with 4DKs of Leo, suggesting that the domain is unique to Ac or is present only in organisms not part of Foldseek searches. (C) Surface view of Leo-A shows three Tyr residues (green) form a linear array in each 4DK, which were mutated to Ala to test the effects on binding to Avicel cellulose by MBP-Leo-A (see Fig. 1H). (D) Surface view of Leo-S(c) shows a 252-aa-long unstructured, spacer rich in Thr (blue) between 4DKs, each of which contains linearly arrayed Tyr residues. (E) Confocal microscopy shows WT Leo-A localizes to the endocyst layer (pink arrowheads) and forms a flat ring around ostioles (green arrowheads). (F) When six Tyr are mutated to Ala, Ala-mut Leo-A (gray arrowheads) no longer localizes to the endocyst layer. The results here and in Fig. 1J support the idea that the Ala-mutant of Leo-A, like the Ala-mutant of Luke-2 (Fig. 1), mislocalizes in cyst walls because of the loss of cellulose binding. (G) Early in encystation (12 h), secretory vesicles (SV) containing Leo-S(c) fill the cytosol of Ac, which lack a cyst wall as shown by failure to label with WGA or CFW. (H) Later in encystation (24 h), Leo-S(c) has a somewhat patchy distribution in the single-layered wall, which now labels with WGA and CFW. (I) In mature cyst walls (72 h), Leo-S(c) has a homogeneous distribution in the ectocyst layer (yellow arrowheads).These results show that despite their shared 4DKs, Leo-S(c) and Leo-A do not localize to the same place in the cyst wall. Scale bars for panels E to I are each 5 µm.

Fig 3

Jonah lectins, which have either one or three three-sided, β-helical folds (BHFs), like those of ice-binding proteins of Antarctic bacteria, are made early in encystation and localize to the ectocyst layer. (A) Drawings show that Jonah-1 contains a signal peptide (purple), followed by a Thr-rich spacer (light blue), and a single BHF (green). Jonah-3(i), which was incorrectly predicted by AmoebaDB, has 12 TMHs (red) and three BHFs, while Jonah-3(c), corrected by the new transcriptome, has three BHFs, three Ser-rich domains (yellow), and no TMHs. (B) AlphaFold structure with confidence colored shows that the Jonah-1 lectin is composed of a single three-sided, β-helical fold (BHF), as well as two α-helices (function unknown), and a 170-aa, unstructured domain rich in Thr (light blue). (C) Jonah-3-(C) has three BHFs (red, blue, and yellow) and 132-, 267-, and 308-aa long, unstructured domains rich in Ser (yellow). (D) End-view with confidence colored shows that the Jonah-1 BHF is three-sided (BHFs may also be two-sided). (E) Foldseek shows the Jonah-1 BHF (white) resembles the BHF of Colwellia sp. (blue), the structure of which has been solved (PDB 3WP9). The E-value is 5.10e−4; the RMSD is 9.21; and the sequence identity is 12% over a 245-aa overlap, which together suggest a distant shared ancestry. While surface views reveal numerous aromatic acids, none were in linear arrays so that we declined to make Ala mutations to test cellulose-binding sites. (F) Confocal microscopy shows Jonah-3(i), which was incorrectly predicted by AmoebaDB to contain 12 TMHs, forms rings around ostioles (green arrowheads) when expressed under its own promoter with a GFP tag. Here, the endocyst layer is marked by CFW and WGA (pink arrowheads), while the ectocyst layer (yellow arrowheads) is weakly labeled. (G) In contrast, Jonah-3(c)-GFP, which contains no TMHs, localizes to the ectocyst layer (yellow arrowheads). (H) After 12 h of encystation, Ac are filled with secretory vesicles (SV) containing Jonah-1–mCherry, as well as CFW-labeled vesicles most likely containing cellulose, while Luke-2-GFP is absent. (I) After 24 h of encystation, Jonah-1–mCherry localizes to the ectocyst layer (yellow arrowheads), while CFW and Luke-2-GFP are predominantly in secretory vesicles. (J) After 48 h of encystation, the cyst wall has the appearance of mature cysts with Jonah-1–mCherry in the ectocyst layer, Luke-2-GFP in ostioles and to a lesser extent in the endocyst layer, and CFW predominantly in the endocyst layer. These double labels, which clearly show that Jonah-1 is made early in encystation, while Luke-2 is made later, justify Jonah-1 and Luke-2 in the first promoter swap in Fig. 5. Scale bars for panels F to J are each 5 µm.

Fig 4

An abundant laccase, which resembles a bacterial spore coat protein, is made early in encystation and localizes to the ectocyst layer. (A) The Ac laccase-1 has signal peptide (purple) and three copper oxidase domains, separated by short, unstructured spacers lacking either Ser or Thr. (B) AlphaFold shows CuRO-1 (blue), CuRO-2 (red), and CuRO-3 (yellow) domains of Ac laccase-1, all of which are predicted with confidence (not shown). (C) Foldseek shows that CuRo-1 of the Ac laccase-1 closely resembles the first copper oxidase domain of Bacillus subtilis spore coat protein, which has been crystallized (PDB 4YVN). Typical for an enzyme, the E-value is 1.32e−59 and the RMSD = 2.75 with a sequence identity of 39% over a 505-aa overlap for three copper oxidase domains. (D) Confocal microscopy shows that after 18 h of encystation, secretory vesicles (SV) containing laccase-1-GFP fill the cytosol of Ac, which lacks a cyst wall as shown by the failure to label with WGA or CFW. (E) After 24 h of encystation, laccase-1-GFP has a homogeneous distribution in the single-layered wall, which labels with WGA and CFW. (F) In mature cyst walls (72 h), laccase-1-GFP has a homogeneous distribution in the ectocyst layer (yellow arrowheads) with ostioles appearing as dimples (green arrowheads). Double labels were used to compare the localization of laccase-1-RFP and Leo-A-GFP during encystation. (G) After 18 h of encystation, Ac are filled with secretory vesicles (SV) containing laccase-1-RFP, while Leo-A-GFP is absent. (H) After 24 h of encystation, laccase-1-RFP localizes to the ectocyst layer (yellow arrowheads), while CFW and Luke-2-GFP are predominantly in secretory vesicles. (I) After 48− h of encystation, the cyst wall has the appearance of mature cysts with laccase-1-RFP in the ectocyst layer, Leo-A-GFP in ostioles (green arrowheads) and to a lesser extent in the endocyst layer (pink arrowheads), and CFW predominantly in the endocyst layer. The clear demonstration here that laccase-1 is made early in encystation, while Leo-A is made later, justifies the use of laccase-1 and Leo-A in the second promoter swap in Fig. 5. Scale bars for panels D to I are each 5 µm.

Fig 5

Localization of cyst wall proteins in the ectocyst layer, endocyst layer, and/or ostioles is for the most part determined by timing of expression during encystation. Deproteinated cyst walls were incubated with cyst wall MBP fusion proteins conjugated to Alexa Fluor 647. (A to D) MBP fusions for the BHF of Jonah-1, the CuRO-1 of laccase-1, and the entire Luke-2 and Leo-A minus their signal peptides all bind to both the endocyst layer (pink arrow heads), ectocyst layer (yellow arrowheads), and ostioles (green arrowheads). (E) In the first promoter swap, Jonah-1-GFP expressed under its own early promoter localizes to the ectocyst layer (see also Fig. 1B). (F) Expression under the later Luke-2 promoter causes Jonah-1-GFP to localize to the ostioles, which increase in thickness and number, while there is minimal localization of Jonah-1-GFP to the endocyst layer. (I) Conversely, Luke-2-GFP under its own later promoter localizes to the endocyst layer and forms a narrow ring around ostioles (see also Fig. 2B). (J) Expression under the early Jonah-1 promoter causes Luke-2 to relocate to the ectocyst layer. Remarkably, the narrow ring of Luke-2-GFP around the ostioles appears the same under either its own or the Jonah-1 promoter, suggesting its target glycopolymer is available early and later during encystation. (G) In the second promoter swap, laccase-1-GFP expressed under its own early promoter localizes to the ectocyst layer (see also Fig. 4D). (H) Expression under the later Leo-A promoter causes laccase-1-GFP to localize to the ostioles with minimal localization to the endocyst layer. (K) Conversely, Leo-A-GFP under its own later promoter localizes to the endocyst layer and forms a narrow ring around ostioles. (L) Expression under the early laccase-1 promoter causes Leo-A to localize to the ectocyst layer and forms a narrow ring around ostioles. These experiments show that under both early and later promoters, Jonah-1 and laccase-1 have similar localizations in the cyst wall, suggesting that each bind to the same glycopolymer(s). Luke-2 and Leo-A also have similar localizations under early and later promoters, but the localization is distinct from that of Jonah-1 and laccase-1, suggesting pairs of wall proteins are binding to different glycopolymers. Scale bars for panels A to L are each 5 µm. Manual counts carried out on transgenic cysts for GFP tagged protein localization to the outer cyst wall (red bar), inner cyst wall (green bar), mis-localized (orange bar) or no expression (blue bar). (M) Jonah-1-GFP under own promoter showing 93% of the cysts have protein localized to the outer cyst wall, and 84% of cysts under Luke-2 later promoter localizing to the inner cyst wall. (N) Laccase-1-GFP under own promoter showing 94% of the cysts have protein localized to the outer cyst wall, and under Leo-A later promoter, 86% of the cysts have protein localized to the inner cyst wall. (O) Luke-2-GFP under own promoter shows 94% of cysts have protein localized to the inner cyst wall, 96% to the outer cyst wall under Jonah-1 promoter, and 89% mis-localized in ala-mutated cysts. (P) Leo-A-GFP under own promoter shows 92% of cysts have protein localized to the inner cyst wall, 97% to the outer cyst wall under laccase-1 promoter, and 89% mis-localized in ala-mutated cysts. The counting was done three times, and a representative graph was plotted with mean ± SD.

Fig 6

A neighbor-joining tree with bootstraps marked reveals common ancestry of N-terminal BJRFs of Luke and C-terminal BJRFs of slime molds and shows Luke-2s are truncated versions of Luke-3. A phylogenetic tree contains BJRFs of Ac (outer green boarder) or slime molds (outer gray boarder). BJRFs of Luke lectins, Ac M12 proteases, and some BJRFs of slime molds contain a single disulfide knot (inner blue-green boarder), while other BJRFs of slime molds have two disulfide knots (inner red boarder). See File S3 for their sequences. BJRFs of bacterial and plant endocellulases, which have no disulfides and are much less like BJRFs of Ac and slime molds, were not included in the tree. N-terminal, middle, and C-terminal BJRFs of 10 Luke-3s form three distinct clades marked in orange, purple, and green, respectively, while N- and C-terminal BJRFs of five Luke-2s are marked with blue and red circles, respectively. C-terminal BJRFs of 19 slime mold wall proteins and enzymes form a clade marked in light gray, while N-terminal and middle BJRFs of slime mold wall proteins form a clade marked in dark gray. Although the bootstrap support is modest (0.53), clades BJRFs at N-termini of Luke lectins and C-termini of slime mold proteins, each of which have a single disulfide, appear to share common ancestry. In contrast, common ancestry of clades of middle and C-termini BJRFs of Luke has strong bootstrap support (1), and common ancestry of the clade of BJRFs of slime molds with two disulfides has strong bootstrap support (0.76). Cartoons of domains of Luke-2s show they are missing either the N-terminal BJRF (Luke-2-4 and Luke-2-5), middle BJRF (Luke-2-2 and Luke-2-3), or C-terminal BJRF (Luke-2-1).

Fig 7

A maximum likelihood tree (JTT matrix-based model) shows one copy of Leo-S with a long Ser-rich spacer between 4DKs, which are unique to Ac, has been duplicated nine times, while one Leo-A with adjacent 4DKs has been duplicated once, and two other Leo-As are single copy. A phylogenetic tree contains 4DKs of Leo lectins, which are unique to Ac. See File S3 for their sequences. One clade, which has reasonable bootstrap support (0.76), includes N-terminal 4DKs (blue) of 10 Leo-Ss and a single Leo-A, while another clade with bootstrap support includes C-terminal 4DKs (green) of 10 Leo-Ss. This result suggests that the spacer was added between 4DKs, and then the genes were duplicated nine times. In both clades, Leo-S-1 and Leo-S-2 look more like each other than Leo-S-3 to Leo-S-10, suggesting their genes were duplicated most recently. Leo-As with adjacent 4DKs (orange) appear to have been made independently and replicated once (Leo-A-1 and Leo-A-2) or not replicated (Leo-A-3). Finally, Leo-S-11 has no relationship to the other Leo-Ss or Leo-As, suggesting its origin is unique.

Fig 8

A maximum likelihood tree (JTT matrix-based model) shows BHFs of Jonah, which were derived from bacteria by HGT a long time ago, are stable as a single copy in five Jonah1s, while one Jonah-3 with three BHFs was duplicated twice, and in the other Jonah-29, three has not been duplicated. A phylogenetic tree shows that BHFs of Jonah lectins (dark green border) are distinct from those of bacteria (pink border), so that it is not clear from which bacterium the Ac genes are derived by HGT. See File S3 for their sequences. BHFs of five Jonah-1s are in the same clade as N-terminal, middle, and C-terminal BHFs of Jonah-3-4 (yellow), suggesting they share common ancestry, and each gene is stable. N-terminal, middle, and C-terminal BHFs of Jonah-3-1 to Jonah-3-3 are in the same clade (green), suggesting a single Jonah-3 with three BHFs was duplicated twice.