Learn more: PMC Disclaimer | PMC Copyright Notice
Effects of N-glycan precursor length diversity on quality control of protein folding and on protein glycosylation
Abstract
Asparagine-linked glycans (N-glycans) of medically important protists have much to tell us about the evolution of N-glycosylation and of N-glycan-dependent quality control (N-glycan QC) of protein folding in the endoplasmic reticulum. While host N-glycans are built upon a dolichol-pyrophosphate-linked precursor with 14 sugars (Glc3Man9GlcNAc2), protist N-glycan precursors vary from Glc3Man9GlcNAc2 (Acanthamoeba) to Man9GlcNAc2 (Trypanosoma) to Glc3Man5GlcNAc2 (Toxoplasma) to Man5GlcNAc2 (Entamoeba, Trichomonas, and Eimeria) to GlcNAc2 (Plasmodium and Giardia) to zero (Theileria). As related organisms have differing N-glycan lengths (e.g. Toxoplasma, Eimeria, Plasmodium, and Theileria), the present N-glycan variation is based upon secondary loss of Alg genes, which encode enzymes that add sugars to the N-glycan precursor. An N-glycan precursor with Man5GlcNAc2 is necessary but not sufficient for N-glycan QC, which is predicted by the presence of the UDP-glucose:glucosyltransferase (UGGT) plus calreticulin and/or calnexin. As many parasites lack glucose in their N-glycan precursor, UGGT product may be identified by inhibition of glucosidase II. The presence of an armless calnexin in Toxoplasma suggests secondary loss of N-glycan QC from coccidia. Positive selection for N-glycan sites occurs in secreted proteins of organisms with NG-QC and is based upon an increased likelihood of threonine but not serine in the second position versus asparagine. In contrast, there appears to be selection against N-glycan length in Plasmodium and N-glycan site density in Toxoplasma. Finally, there is suggestive evidence for N-glycan-dependent ERAD in Trichomonas, which glycosylates and degrades the exogenous reporter mutant carboxypeptidase Y (CPY*).
1. Introduction
Parasitic protists cause malaria (Plasmodium falciparum), diarrhea (Giardia lamblia and Cryptosporidium parvum), dysentery (Entamoeba histolytica), neurological illness (Trypanosoma brucei and Toxoplasma gondii), heart disease (Trypanosoma cruzi), mucocutaneous lesions (Leishmania sp.), keratitis (Acanthamoeba castellanii), and sexual transmitted infections (Trichomonas vaginalis). Whole genome sequences of these medically important protists predict the “parts list” of proteins present in each organism, as well as those proteins that are not present [1–8]. The prediction of absence, which is important for analysis of secondary loss, is based upon the paucity of pseudogenes in these parasites, with the exception of Trichomonas [9]. Phylogenetic trees of small subunit rRNA genes or of housekeeping proteins allow us to identify protists with common ancestry, which correlates with their appearance (e.g. Acanthamoeba and Entamoeba) [10]. These trees also show that protists differ more from each other than do metazoans, fungi, and plants, but they fail to identify the oldest extant eukaryote. Instead they identify a set of deeply divergent eukaryotes including Giardia, Trichomonas, and Trypanosoma, which are called “excavates” [11]. In addition, subsequent to bottle-necking, the common ancestor of all eukaryotes has most certainly been lost [12].
Because of secondary loss of sets of genes (e.g. Alg genes that encode enzymes that make precursors to asparagine-linked glycans that are a focus here), it is possible to identify protists with a simpler, “ancestor-like” set of enzymes that make N-glycan precursors [13–15]. These protists with short or very short N-glycans show us that eukaryotes can grow and cause disease without N-glycan-dependent quality control of protein folding (N-glycan QC) or N-glycan-dependent ER-associated degradation (N-glycan ERAD) [16–18]. In turn, the protists without N-glycan QC demonstrate its impact on N-glycan site density in secreted proteins, by comparison to the vast majority of eukaryotes that have this system (see Section 4).
2. The present diversity of N-glycan precursors
2.1 Alg enzymes make N-glycan precursors
When we began this line of research ten years ago, we expected that protist N-glycans would resemble those of vast majority of metazoans, fungi, and plants that are composed of 14 sugars (Glc3Man9GlcNAc2) (Fig. 1). The exception to this rule, of course, was Trypanosoma cruzi, which is missing the three glucose residues [19]. This absence was exploited by Armando Parodi, who characterized the UDP-glucose:glucosyltransferase (UGGT), the key enzyme in N-glycan QC, in the absence of confounding glucose residues on the N-glycan precursor [16]. In addition, there was a great deal of confusion as to whether Plasmodium is or is not able to make N-glycans [20].
Alg enzymes predict N-glycan precursors. Secondary loss of Alg enzymes from apicomplexan parasites (top) and from fungi (bottom) predicts the present diversity of N-glycan precursors. Toxoplasma has all the Alg enzymes except those that add mannose in the ER lumen and so makes a precursor with Glc3Man5GlcNAc2. Boxes outline those Alg enzymes and the N-glycan precursors of the other parasites. Similarly, Saccharomyces has a complete set of Alg enzymes and makes Glc3Man9GlcNAc2, while Cryptococcus is missing those that add glucose. Encephalitozoon, like Theileria has no Alg enzymes and so makes no N-glycans. Drawn after ref. [29].
The vast majority of Alg genes, which encode enzymes that make the N-glycan precursor that is transferred to Asn of proteins in the ER lumen, were discovered using a “mannose suicide” experiment [21]. Briefly, mutagenized yeast were labeled with tritiated mannose and then allowed to sit in a −70°C freezer for months. Upon thawing, those yeast knockouts that survived had retained less radioactive mannose, because synthesis of their N-glycan precursors was blocked (e.g. ΔAlg1 was unable to add the first mannose to dolichol-PP-GlcNAc2). Alg genes, which are numbered based upon the order of their discovery, include those that make dolichol-PP-Man5GlcNAc2 (Fig. 1). These enzymes that use cytosolic UDP-GlcNAc and GDP-Man to add sugars to dolichol-phosphate are similar to those present in bacteria [14, 22]. In contrast, the enzymes that use dolichol-P-Man and dolichol-P-Glc to make dolichol-PP-Glc3Man9GlcNAc2 in the lumen of the ER resemble enzymes that make the GPI-anchor precursor but do not resemble bacterial enzymes [23].
The extant eukaryotes appear to descend from a common ancestor that made Glc3Man9GlcNAc2 (see argument below), and there are no alternative enzymes to catalyze the addition of a particular sugar to the N-glycan. A minor exception is the fusion of Alg13 and Alg14, which add the second GlcNAc to the N-glycan precursor, in Entamoeba and Dictyostelium. Further, there is no diversity in N-glycan precursors, as occurs for complex N-glycans made in the Golgi. Therefore, it is relatively easy to identify the Alg enzymes from the “parts list” of proteins of each parasite using BLASTP, as well as simple treeing methods to distinguish paralogs that result from gene duplication (e.g. Alg1, Alg2, and Alg11 that add the first five mannoses to the N-glycan precursor) [15].
2.2 Protists with short N-glycan precursors
Protists make all possible combinations of N-glycan precursors that are reasonable (e.g. it is not possible to add Glc to N-glycans that are missing a mannose arm) [15]. For example, coccidian parasites, which have oocyst walls and are spread by the fecal-oral route (Toxoplasma and Cryptosporidium that infect humans and Eimeria that infects chickens), are each missing enzymes that add four mannose to the N-glycan precursor (Alg3, Alg9, and Alg12). In some cases, they are also missing enzymes that add glucose (Alg6, Alg8, and Alg10), so that their N-glycan precursor then contains Glc0-3Man5GlcNAc2 (Fig. 1).
Entamoeba and Trichomonas, which are unrelated to each other, are missing all the enzymes in the ER lumen that add mannose and glucose and so make a precursor that contains Man5GlcNAc2. These N-glycans were confirmed by metabolic labeling of parasites with tritiated mannose and analyzing N-glycans released with PNGaseF on a sizing column with yeast standards [15]. Alternatively, one can analyze released N-glycans or tryptic glycopeptides by mass spectrometry. Note that the parasite Man5GlcNAc2, which contains a single arm attached to 1,3-linked mannose that is the target of UGGT (Fig. 2), differs from mammalian or yeast Man5GlcNAc2 that results from ER mannosidase digestion of Man9GlcNAc2 [24]. Indeed the ER mannosidase converts the Entamoeba and Trichomonas N-glycans to biantennery Man3 (Man3GlcNAc2), which is made by Golgi mannosidases in the human host. Interestingly, protists have the set of three cytosolic mannosyltransferases (Alg1, Alg2, and Alg11) that make Man5GlcNAc2 or have none at all [14,15]. In addition, Rft1, which has been genetically linked in yeast with transfer of Man5GlcNAc2 from the cytosol to the lumen of the ER but does not appear by itself to be the “flippase” for dolichol-PP-Man5GlcNAc2, shares the same phylogenetic profile as the cytosolic mannosyltransferases [15, 25–27].
Predicted N-glycan QC and N-glycan ERAD in higher eukaryotes (left) and Trichomonas (right). Enzymes and substrates involved in N-glycan QC are marked in red, while those involved in N-glycan ERAD are marked in blue). N-glycan sugars are as in Fig. 1. Abbreviations for this figure only are glucosidases I and II (GlsI and GlsII), calreticulin (CRT), and calnexin (CNX). Asterisks mark 1,6-linked mannose recognized by the OS-9 lectin. N-glycan QC and N-glycan ERAD are absent in Giardia and Plasmodium. Drawn after ref. [18].
2.3 Protists with very short N-glycan precursors
Plasmodium and Giardia, which are also unrelated to each other, are missing all of the Alg enzymes that add mannose or glucose (as well as Rft1) and so make a precursor that contains just GlcNAc2 [15, 28, 29]. This result explains why tritiated mannose went into GPI-anchors but not N-glycans of Plasmodium [20]. Instead N-glycans of Plasmodium and Giardia are labeled with tritiated glucosamine, which is converted in the parasite to GlcNAc. Finally, Theileria, the cause of bovine malaria as well as lymphoma, has no N-glycans [30].
Because the coccidians (Toxoplasma, Cryptococcus, and Eimeria) and related apicomplexan parasites that infect red blood cells (RBCs) (Plasmodium, Babesia, and Theileria) share recent common ancestry, the present diversity of their N-glycans is based upon secondary loss of Alg genes from an ancestor that made Glc3Man5GlcNAc2 [15, 29]. Selection against N-glycan length in parasites that infect RBCs is likely because secreted proteins targeted to the RBC cytosol must thread through a narrow channel called the PTEX [31]. Acanthamoeba, which shares common ancestry with Entamoeba that makes Man5GlcNAc2, has a complete set of Alg enzymes and so makes an N-glycan like that of the host (Glc3Man9GlcNAc2) [4, 7, 15]. Secondary loss of N-glycans also occurs in fungi, if rarely. Cryptococcus is missing Alg enzymes that add glucose in the ER lumen and so makes an N-glycan precursor with Man9GlcNAc2 (like that of Trypanosoma), while Encephalitozoon (also known as microsporidium), a fungus with a markedly reduced genome, makes no N-glycans (Fig. 1) [15, 32].
2.4 Effects on the OST, complex N-glycans, and GPI-anchors
The oligosaccharyltransferase (OST) that moves the N-glycan from the precursor is composed of a catalytic peptide (STT3), which shares common ancestry with the bacterial enzyme, as well as multiple other peptides in metazoans, fungi, and plants [33]. The OST of Giardia and Trypanosoma is composed of a single catalytic peptide, while other protists have fewer non-catalytic peptides than higher eukaryotes. The OST of a particular eukaryote often prefers the endogenous N-glycan (e.g. Glc3Man9GlcNAc2 for yeast and man or Man5GlcNAc2 for Entamoeba and Trichomonas) but this is not the case for Trypanosoma that transfer either Glc3Man9GlcNAc2 or Man9GlcNAc2 [34]. While the Giardia OST transfers endogenous GlcNAc2, it is also transfers with no particular preference exogenous N-glycans varying in size from Man3GlcNAc2to Glc3Man9GlcNAc2. Despite its short N-glycan, the occupied N-glycan sites of Giardia predominantly contain threonine (NxT) rather than serine (NxS) [28]. This result is consistent with previous observations that OSTs preferentially glycosylate NxT and that glycans are not part of the recognition site for STT3 [35–38]. An exceptional case is Trypanosoma brucei, in which a first STT3 transfers a truncated N-glycan precursor (Man5GlcNAc2) to N-glycan sites with an acidic amino acid in the -2 position, and then a second STT3 transfers the full-length N-glycan Man9GlcNAc2 to the canonical N-glycan site [39]. This transfer has important consequences, as Trypanosoma are missing the Golgi mannosidase and so cannot convert Man9GlcNAc2 to Man3GlcNAc2 (the building block for complex N-glycans).
Some parasites modify their N-glycans in the Golgi (e.g. Trypanosoma and Trichomonas add LacNAc, while Entamoeba adds Gal and Glc) [40, 41]. Many N-glycans, however, remain unmodified by ER mannosidases and Golgi glycosyltransferases and so are recognized by lectins that bind to the single 1,3-linked mannose arm of Man5GlcNAc2 (e.g. cyanovirin-N binding to Entamoeba, Trichomonas, Toxoplasma, and Cryptosporidium) [29, 42, 43]. Similarly, wheat germ agglutinin or Griffonia simplicifolia lectin II binds to unmodified GlcNAc2 of Giardia and Plasmodium [28, 29]. Lectins that bind unmodified N-glycans are useful reagents for affinity purification of parasite glycoproteins from mass spectrometry. Lectins (e.g. cyanovirin-N and griffithsin), which bind high mannose N-glycans on gp120 of HIV and are candidate compounds to prevent heterosexual spread of HIV, also bind unmodified N-glycans on the surface of Trichomonas (our unpublished data) [44].
Secondary loss of enzymes (designated GPI in yeast or PIG in metazoans) that make GPI-anchors also occurs in protists but is independent of N-glycan status [45, 46]. For example, Giardia and Entamoeba are missing the mannosyltransferases that add the 3rd mannose to the GPI-precursor, while Trichomonas is missing the entire set of GPI-synthetic enzymes [2, 8, 47]. Interestingly, Trichomonas retains the Alg5 enzyme that makes Dol-P-Glc despite the absence of glucose in its N-glycan precursor. This result, which supports the idea that an ancestor of Trichomonas once glucosylated its N-glycan precursor, led us to the discovery that Dol-P-Glc is used to make O-linked glycans [48]. Finally, Trypanosoma cruzi and some Leishmania have a predicted glucosidase I that removes the terminal Glc on Glc3Man9GlcNAc2. Since these parasites make an N-glycan precursor with Man9GlcNAc2, the presence of the glucosidase I gene is evidence for the secondary loss of these Alg6, Alg8, and Alg10 genes [14, 15]. The present function of T. cruzi glucosidase I is unclear, as no glucosidase I activity was detected in membrane extracts of the parasite [49].
In summary, secondary loss of Alg genes encoding enzymes that make the N-glycan precursor explains the present diversity of N-glycans among medically important protists. While N-glycan precursor length has profound effects on N-glycan QC, N-glycan ERAD, and N-glycan site density in secreted proteins (next sections), it appears to have no effect on the ability of these parasites to grow and cause a wide range of human illnesses. In contrast, decreased function or loss of function of the Alg enzymes that make N-glycan precursors have debilitating effects on human development, resulting in an array of phenotypes that are referred to as type 1 congenital disorders of glycosylation (CDG-I) [50].
3. Effect of N-glycan diversity on quality control of protein folding
3.1 N-glycan QC
Because the UGGT in the ER lumen adds glucose to a single arm attached to 1,3-linked mannose of the N-glycan bound to protein, protists that make N-glycan precursors with at least seven sugars (Man5GlcNAc2) (e.g. Entamoeba, Trichomonas, Toxoplasma, Cryptosporidium, Trypanosoma, and Acanthamoeba) are theoretically capable of N-glycan QC (Fig. 2) [15, 16, 18, 19]. Conversely, those organisms with very short N-glycan precursors (e.g. Plasmodium and Giardia) or no N-glycans (Theileria and Encephalitozoon) are incapable of N-glycan QC (or N-glycan ERAD, as discussed in Section 3) [17]. Consistent with these predictions, UGGT, glucosidase II that removes glucose from GlcMan5GlcNAc2 or GlcMan9GlcNAc2, and lectins that bind glucosylated N-glycans (calnexin and/or calreticulin) are present in Entamoeba, Trichomonas, Trypanosoma, and Acanthamoeba, while they are absent in Plasmodium, Giardia, and Theileria. It does not appear to matter whether organisms with N-glycan QC have calreticulin only (Entamoeba and Trypanosoma), calnexin only (Saccharomyces), or both lectins (Trichomonas) [18]. Organisms with predicted N-glycan QC also have the N-glycan-binding lectin that binds well-folded glycoproteins and directs them to the Golgi (ERGIC-53 and related VIP proteins), as well as a UDP-Glc transporter that moves the nucleotide sugar from the cytosol to the ER-lumen [51, 52]. The UDP-Glc transporter was first identified by expressing Entamoeba protein in Giardia that does not have NG-QC and does not transport UDP-Glc [53].
UGGT activity was demonstrated in Entamoeba and Trichomonas, which do not have Glc in their N-glycan precursors, by inhibiting glucosidase II with castanospermine and seeing a large increase in the amount of GlcMan5GlcNAc2 present in lysed parasites [16–18]. Alternatively, membrane extracts of Entamoeba were shown to use UDP-Glc to glucosylate denatured thyroglobulin or to add Glc to Man5GlcNAc2 attached to an iodinated NYT peptide. The latter result suggests that the Entamoeba UGGT is active even when attached to a short peptide, which is not the case for the S. pombe or Trypanosoma UGGT.
To our surprise, Toxoplasma and Cryptosporidium, which make N-glycan precursors with Glc2-3Man5GlcNAc2, do not have UGGT, ERGIC-53, or intact calnexin and/or calreticulin (Fig. 2) [18]. While Toxoplasma has a calnexin that is missing the arm that binds the protein disulfide isomerase (PDI), it does not appear to have N-glycan QC, as suggested by the absence of positive selection for N-glycan sites in its secreted proteins (see Section 3) [54]. In contrast, positive selection for N-glycan sites is found in all other eukaryotes with N-glycan QC. This result suggests that the product of Toxoplasma glucosidases I and II (GlcMan5GlcNAc2) is not bound and refolded by the armless calnexin. In contrast, the product of Saccharomyces glucosidases I and II (GlcMan9GlcNAc2) appears to be bound and refolded by calnexin, as there is positive selection for N-glycan sites in its proteins despite the fact that the Saccharomyces UGGT ortholog (Kre5) does not glucosylate N-glycans [55].
3.2 N-glycan ERAD
Whether protists might have N-glycan ERAD is more difficult to judge from the parts list of their predicted proteins, because the set of effectors is not as well-defined as those for N-glycan QC (e.g. the dislocon is not molecularly identified) [17, 18, 56]. Further numerous components of N-glycan ERAD (e.g. BIP, PDI, peptidylprolyl isomerase (PPI), HRD1, Der1, and CDC48) are not specific to this pathway and are present in organisms such as Giardia and Plasmodium that do not have N-glycan ERAD. In turn, N-glycan ERAD activity is judged by overexpression of an exogenous misfolded protein, which may be secreted or membrane bound, rather than by inhibition of glucosidase II or by a direct assay of membrane extracts, as for N-glycan QC. With these caveats, Trichomonas and Trypanosoma, which have N-glycan QC, have orthologs of Mns1 and EDEM, respectively, and each has mannosidase activity as a recombinant protein (Fig. 2) [18]. Trichomonas also has a cytosolic PNGaseF, which is active as a recombinant protein. Trichomonas adds N-glycans to wild-type carboxypeptidase Y (CPY) of Saccharomyces, as well as to mutant CPY*, which misfolds and is degraded in a proteasome-dependent manner (our unpublished data). Trichomonas has a distant homolog to OS-9, the lectin that binds 1,6-linked mannose. The 1,6-linked mannose (marked by an asterisk in Fig. 2) is revealed by host ER mannosidases prior to dislocation of the misfolded protein into the cytosol. The unprocessed N-glycan of Trichomonas (Man5GlcNAc2) contains the 1,2-linked mannose recognized by UGGT, as well as the 1,6-linked mannose recognized by OS-9, suggesting a possible “tug of war” between N-glycan QC and N-glycan ERAD.
The safest conclusion then is that some but not all protists have N-glycan QC, while the evidence for N-glycan ERAD is unclear. In contrast, all protists have proteins involved in N-glycan-independent QC of proteins folding (BIP and other chaperones, PDI, and PPI), although most parasites are missing the transmembrane kinase/RNAse (IRE1) that triggers the unfolded protein response in higher eukaryotes [57]. All protists also have proteins involved in N-glycan-independent ERAD, although studies with misfolded reporter proteins lacking N-glycan sites have not been performed [18]. Remarkably, Toxoplasma has a second set of ERAD proteins (Der1 and Cdc48) that appear to be essential for transporting nuclear-encoded proteins from the ER lumen to a chloroplast-derived organelle called the apicoplast [58].
4. Effect of N-glycan QC on protein evolution
4.1 Hypotheses tested
Addition of N-glycans may affect protein folding in four ways. First, addition of N-glycans affects the thermodynamic stability of the protein [59]. Second, by association with the Sec61 translocon, the OST affects protein folding when it binds the nascent peptide and adds N-glycans [14]. Third, N-glycans are the target for UGGT, calnexin/calreticulin, and ERGIC53 [16]. Fourth, they are part of N-glycan ERAD [17, 56]. To judge the relative importance of these processes, we determined whether there is a positive selection for N-glycan site density (NxT or NxS) in secreted proteins versus cytosolic proteins, where no N-glycans are added. We also compared N-glycans site densities in the secreted proteins of the vast majority of eukaryotes that have N-glycan QC with those of protists (Plasmodium, Giardia, Toxoplasma, Cryptosporidium, and Theileria) that do not have N-glycan QC. To do this experiment, we collected protein sequences from as many dissimilar organisms as were available in 2009, removed redundant sequences that fill the NR database at the NCBI, and separated the remaining unique sequences into secreted proteins (with an N-terminal signal peptide) and nucleocytosolic proteins with no signal peptide or transmembrane helices [60, 61]. We then compared the expected density of N-glycan sites based upon the amino acid composition of the proteins of each organism with the actual N-glycan site density (Fig. 3).
Positive selection for N-glycan sites in secreted proteins of eukaryotes with N-glycan QC. Densities of N-glycan sites with Thr (per 500 amino acids) in secreted proteins of eukaryotes with N-glycan QC (top) and those without N-glycan QC (bottom) are plotted versus the AT content of the genome. Note that for each organism there are two spots marked: the N-glycan density expected by amino acid composition of proteins (green) and the actual N-glycan density (magenta). For selected organisms, a vertical bar connects the two spots: Homo (Hs), Saccharomyces (Sc), Dictyostelium (Dd), Entamoeba (Eh), Plasmodium (Pf), and Toxoplasma (Tg). In both plots, the expected N-glycan density increases with AT content, as Asn is encoded by AAT/C. In organisms with N-glycan QC there is a difference between the actual density of N-glycan sites with Thr and the expected difference (positive selection). There is no difference between actual and expected N-glycan site density for organisms without N-glycan QC (no selection). Drawn after ref. [54].
4.2 Selection for N-glycan sites with Thr but not Ser
There were three prominent results (Fig. 3). First, independent of whether organisms have long or short N-glycans and whether they have or do not have N-glycan QC, those with AT-rich genomes (e.g. Plasmodium, Dictyostelium, and Entamoeba) have a greater density of expected N-glycan sites, as asparagine is encoded by AT-rich codons (AAT/C). Conversely, Toxoplasma with a GC-rich genome has fewer expected N-glycan sites. Humans are for the most part AT-GC neutral and have an intermediate density of expected N-glycan sites, while Saccharomyces is slightly AT-rich and has more expected N-glycan sites. To our knowledge, this is the only effect of codon usage on protein glycosylation, as Ser and Thr that are O-glycosylated are encoded by a balance of AT- and GC-rich codons [18, 62].
Second, in organisms with N-glycan QC the actual density of N-glycans sites with Thr (NxT) in their secreted proteins is markedly increased versus the expected density of sites, and this increase is independent of AT-content. In contrast and to our surprise, there was no increase in density of N-glycan sites with Ser (NxS) in organisms with N-glycan QC. For these same organisms, the expected and actual N-glycans site densities with Thr or Ser were the same in nucleocytosolic proteins (negative controls), which are not N-glycosylated. The expected N-glycans site densities were the same for secreted and nucleocytosolic proteins of organisms with N-glycan QC, showing that increase in the actual N-glycan site density is not because of an increase in the density of Asn or Thr. Instead the increase in the actual density of N-glycan sites with Thr (NxT) is based upon an increased likelihood that Thr is in the +2 position with regards to Asn. The increase in the N-glycan site density in hemagglutinin of influenza virus as it mutates over a 20 year period in the human host is also based upon an increased likelihood that Thr or Ser is in the +2 position with regards to Asn rather than an increase in the density of Asn, Thr, or Ser [18, 63]. Similarly, the very high density of N-glycan sites on gp120 of the HIV is predominantly explained by an increased likelihood that Thr or Ser is in the +2 position with regards to Asn.
Third, in organisms without N-glycan QC, there is no difference between the actual N-glycan site density with Thr or Ser and the expected N-glycan site density for its secreted proteins. This result shows that N-glycan QC, not just the presence of N-glycans that are as many as 9 to 10 sugars long in Cryptosporidium and Toxoplasma, is the important determinant for positive selection for N-glycan sites in secreted proteins.
4.3 Possible explanations for these surprising results
Despite the preference of OSTs for N-glycan sites with Thr over those with Ser, ~1/3 of occupied N-glycan sites contains Ser [37]. There is then no easy explanation for why there is positive selection for N-glycan sites with Thr but not Ser in the secreted proteins of organisms with N-glycan QC. There must be some component of the N-glycan QC system that is sensitive to the site of N-glycans and not just to the status of the N-glycan or of glycoprotein folding. Regardless of the mechanism, these evolutionary studies of N-glycan site density in secreted proteins are a strong argument for the importance of N-glycan QC (and possibly N-glycan ERAD that is present in the vast majority of these organisms). Such evidence has been relatively hard to come by experimentally, where knockout of UGGT gives a phenotype only when yeast or trypanosomes are severely stressed [64, 65]. Finally, as an exception that perhaps proves the rule, there is negative selection against the sites of N-glycosylation in Toxoplasma proteins that pass through the ER and then thread through a pore into the apicoplast, so that many of these proteins have no N-glycan sites [29].
Conclusions and future studies
Studies of medically important protists has shown the importance of secondary loss of Alg enzymes to explain the evolution of N-glycans and to show that N-glycan length, N-glycan QC, and N-glycan ERAD are not as important for viability of single cell organisms as for higher eukaryotes [50]. These studies have defined what appears to be the minimum proteins involved in N-glycan QC and have shown the mechanism whereby N-glycan QC selects for N-glycan site density in secreted proteins. Unanswered questions revolve primarily around how N-glycan precursors are flipped into the ER lumen, whether there is N-glycan ERAD in some protists, and why there is selection for N-glycan sites with Thr but not Ser in secreted proteins of eukaryotes with N-glycan QC.
Acknowledgments
We thank a terrific set of students and post-doctoral fellows, who performed this work. In alphabetical order, they are Giulia Bandini, Sulagna Banerjee, Guy Bushkin, Andrea Carpentieri, Aparajita Chatterjee, John Cipollo, Jike Cui, Kariona Grabińska, John Haserick, Paula Magnelli, Edwin Motari, and Dan Ratner. We thank collaborators including Cathy Costello, Reid Gilmore, Marc-Jan Gubbels, Carlos Hirschberg, Barry O’Keefe, and Temple Smith. This work was supported by the NIH grants GM031318, AI44070, and AI048082.
Abbreviations
| CPY | Saccharomyces carboxypeptidase Y |
| CPY* | misfolded CPY mutant |
| N-glyan | Asn-linked glycan |
| PPI | peptidylprolyl isomerase |
| PDI | protein disulfide isomerase |
| QC | quality control |
| UGGT | UDP-glucose:glucosyltransferase |
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.