|Home | About | Journals | Submit | Contact Us | Français|
Retroposition is an important mechanism for gene origination. However, studies to elucidate the functions of new genes originated through retroposition, especially the functions related to diseases, are limited. We recently identified a mouse gene, Rps23 retroposed gene 1 (Rps23rg1), that regulates β-amyloid (Aβ) level and tau phosphorylation, two major pathological hallmarks of Alzheimer's disease (AD), and found that Rps23rg1 originated through retroposition of the mouse ribosomal protein S23 (Rps23) mRNA. Here we show that retroposition of Rps23 mRNA occurred multiple times in different species but only generated another functionally expressed Rps23rg1-homologous gene, Rps23rg2, in mice, whereas humans may not possess functional Rps23rg homologs. Both Rps23rg1 and Rps23rg2 are reversely transcribed relative to the parental Rps23 gene, expressed in various tissues and encode proteins that interact with adenylate cyclases. Similar to the RPS23RG1 protein, RPS23RG2 can upregulate protein kinase A activity to reduce the activity of glycogen synthase kinase-3, Aβ level and tau phosphorylation. However, the effects of RPS23RG2 are weaker than those of RPS23RG1 and such a difference could be attributed to the extra carboxyl-terminal region of RPS23RG2, which may have an inhibitory effect. In addition, we show that the transmembrane domain of RPS23RG1 is important for its function. Together, our results present a new gene family, whose products and associated signaling pathways might prevent mice from developing AD-like pathologies.
Alzheimer's disease (AD), the most common neurodegenerative disease associated with aging in humans, is pathologically characterized by the abnormal accumulation of extracellular neuritic plaques and intracellular neurofibrillary tangles (NFTs) in vulnerable brain regions. Neuritic plaques consist of deposits of variously sized small β-amyloid (Aβ) peptides (1,2), which are derived from the precursor protein (APP) through sequential proteolytic processing by β-secretase and γ-secretase (3). Multiple lines of evidence suggest that overproduction/aggregation of Aβ in the brain is the primary cause of AD pathogenesis (4). NFTs are composed largely of hyperphosphorylated twisted filaments of a microtubule-associated protein, tau (5,6). Although tau hyperphosphorylation and aggregation is believed to be downstream of Aβ pathology in AD, tau mutations causing hyperphosphorylation have been associated with frontotemporal dementia with Parkinsonism-17 and tau pathology has been found in a number of neurodegenerative diseases collectively termed tauopathy (6,7).
Genetically engineered mice have become a useful tool for AD research. However, so far, none of these mouse models can faithfully recapitulate all of the pathological features of AD. Additionally, wild-type mice do not develop age-associated AD-like pathologies (8–10). Several hypotheses have been proposed to explain the resistance of mice to AD-like pathologies, including the sequence disparity between human and mouse Aβ (and possibly tau) that underlie different aggregation capabilities (8–10), the short lifespan of mice relative to humans (10,11) and the differences in processing of human and mouse APP by BACE1 (12). However, it is possible that other genetic novelties within the mice genome may also contribute to such a resistance. Therefore, identification of new genes/proteins involved in the modulation of Aβ generation and/or tau hyperphosphorylation and the comparative study of these genes/proteins between different species may not only be important for developing AD therapeutics but also provide clues for understanding the discrepancy between humans and mice, which will be useful in developing better AD animal models.
The origination of genes with new functions is an important mechanism for generating genetic novelties within a species during its evolution. New genes can originate through different mechanisms such as exon shuffling, gene duplication, retroposition, mobile element integration, lateral gene transfer and gene fusion/fission (13). Retroposition is a process where a parental mRNA is reverse-transcribed and inserted into the organism's genome, creating duplicate genes in new genomic positions. Although a retroposed gene copy usually does not contain the promoter region from the parental gene and will die out as a processed pseudogene in most cases, it can occasionally recruit new regulatory sequences near the incorporation site and become a functionally expressed gene (13). Nevertheless, studies to elucidate the functions of these newly originated genes, especially the functions related to diseases, are limited (14,15).
We recently identified a mouse gene Rps23rg1 (Rps23 retroposed gene 1. Gene ID 546049, C330021F23Rik. Referred to as Rps23r1 in previous publication but changed here, by suggestion of the Mouse Genomic Nomenclature Committee) that originated through retroposition of the mouse ribosomal protein S23 (Rps23) mRNA (16). In the present study, we show that retroposition of Rps23 mRNA occurred multiple times in different species but only generated two functionally expressed genes, Rps23rg1 and Rps23rg2, both reversely transcribed relative to Rps23, in mice. In addition, we show that the RPS23RG protein family members function similarly to reduce AD-like pathologies (Aβ levels and tau phosphorylation) by interacting with adenylate cyclases to upregulate protein kinase A (PKA) activity and downregulate glycogen synthase kinase-3 (GSK-3) activity.
To identify RPS23RG1 homolog(s), we used the 141 amino acid-long mouse RPS23RG1 protein sequence to blast GenBank protein database and found another three mouse proteins with very high homology to RPS23RG1: EG381438 (Gene ID 381438, identities = 116/117, 99%), LOC100040998 (Gene ID 100040998, identities = 137/139, 98%) and LOC100039346 (Gene ID 100039346, identities = 126/134, 94%), of which the latter two are predicted by automated computational analysis. Hence, we designated their encoding genes as Rps23rg2 (protein RPS23RG2 for EG381438), Rps23rg3 (RPS23RG3 for LOC100040998) and Rps23rg4 (RPS23RG4 for LOC100039346), respectively (Fig. 1A). RPS23RG1 has a predicted transmembrane domain (Fig. 1A) and previously we found that RPS23RG1 is a type Ib transmembrane protein that has a normal type I transmembrane protein orientation but no cleavable signal peptides. Here we found that RPS23RG2–RPS23RG4 also have the predicted transmembrane domain (Fig. 1A), suggesting that they are also type Ib transmembrane proteins. Blast with the mouse RPS23RG1 protein sequence also identified a predicted hypothetical protein LOC738579 (Gene ID 738579, identities = 50/74, 67%) with a relatively high homology, but lacking the predicted transmembrane domain from chimpanzees (Pan troglodytes) (Fig. 1A), which we designated Rps23rg5 (and the protein as RPS23RG5). Moreover, the blast search showed that RPS23RG1 shares some homology with the CRA_d isoform of human ATG10 [ATG10 autophagy related 10 homolog (S. cerevisiae). Gene ID 83734, identities = 33/40, 82%]. However, further analyses revealed that the ATG10 CRA_d isoform originated through a different mechanism from that of Rps23rg1–Rps23rg5 (see below).
To study whether Rps23rg family genes are expressed in mice, we carried out reverse transcription–PCR (RT–PCR) and found that Rps23rg1 was expressed in various mouse tissues (Fig. 1B). In addition, we found that Rps23rg2 was also expressed (Fig. 1B). However, we did not detect the expression of Rps23rg3 or Rps23rg4 in any tissue by RT–PCR with several pairs of primers (data not shown), implying that the two genes predicted by automatic computational analysis are either not real functional genes (i.e. they are processed pseudogenes) or they are expressed at very low levels that cannot be detected by RT–PCR. For Rps23rg5, we did not study whether it is expressed in chimpanzees due to limited resources. However, since Rps23rg5 was also predicted by automated computational analysis and we failed to detect its expression in humans by RT–PCR (data not shown), even though Rps23rg5 originated before the divergence between humans and chimpanzees (see below), it is possible that Rps23rg5 may also be a pseudogene without functional expression.
In our previous study, we found that the Rps23rg1 gene originated through retroposition of the mouse Rps23 mRNA (16). Therefore, we also compared gene sequences of other Rps23rg members to those of Rps23. Consistent with the results from the Rps23rg1 comparison, sequence alignments and analyses showed that Rps23rg2–Rps23rg4 and Rps23rg5 gene sequences were highly homologous to the reverse and complementary sequences of mouse and chimpanzee Rps23 mRNA, respectively, and were all intronless within the homologous region (Fig. 2A and B). These results clearly demonstrate that all of the Rps23rg gene family members originated through retroposition of Rps23 mRNA, but are reversely transcribed, relative to their parental genes.
The Rps23 gene is highly conserved among species and belongs to the ribosomal protein family that is crucial for ribosome function (17). It has been reported that human ribosomal protein genes have generated a large number of processed pseudogenes through retroposition (18). Therefore, to further understand the origination of Rps23rg family genes, we screened for all Rps23-like sequences in human, mouse and rat genomes and constructed their phylogenic relationships. The results showed that retroposition of Rps23 occurred in all three species and occurred more frequently in rodents than in humans (Fig. 3). The results also demonstrated that Rps23rg1–Rps23rg4 originated in mice after the divergence between mice and rats, whereas Rps23rg5 originated after the divergence between rodents and primates but before the divergence between humans and chimpanzees (Fig. 3). Furthermore, to identify potential functional Rps23rg homologs in humans, we selected fragments covering 100 Kb of the 5′ and the 3′ region adjacent to each of the identified Rps23-like sequences in the human genome for gene prediction and found no Rps23rg-like genes (data not shown). We also carried out RT–PCR with primers binding regions right next to these identified human Rps23 retroposition sites and failed to obtain any positive amplification (data not shown).
We focused our functional studies on Rps23rg1 and Rps23rg2 since only these two genes were found to be expressed. RPS23RG2 is 24 amino acids shorter at the amino-terminus and 37 amino acids longer at the carboxyl-terminus than RPS23RG1. Otherwise, RPS23RG1 and RPS23RG2 have only one amino acid difference within the predicted transmembrane domain (Fig. 1A). Previously we showed that RPS23RG1 is a type Ib transmembrane protein and can be delivered to the cell surface (16). Here we found that RPS23RG2 and its truncated form lacking the 37 amino acids at the carboxyl-terminus (RPS23RG2CΔ37) can also be delivered to the cell surface (Fig. 4B), implying that these RPS23RG family members have same subcellular localizations. Since overexpressed RPS23RG1 has been found to interact with adenylate cyclases and upregulate the cAMP level, activating PKA activity and thereby inhibiting GSK-3 activity, tau phosphorylation and Aβ generation (16), we investigated whether RPS23RG2 has a similar function. We found that overexpressed RPS23RG2 also interacted with overexpressed adenylate cyclase 8 (Fig. 4A). In addition, overexpression of RPS23RG2 in human HeLa cells stably expressing human APP Swedish mutation upregulated PKA activity (with more CREB phosphorylation), decreased GSK-3α/β activities (with more GSK-3α/β phosphorylation), Aβ level and tau phosphorylation, and increased sAPPα release and APP carboxyl-terminal fragments (CTFs) levels (Fig. 4B). However, the effects of RPS23RG2 were much weaker than those of RPS23RG1 (Fig. 4B). In contrast, the truncated RPS23RG2CΔ37 form not only interacted with adenylate cyclase 8 (Fig. 4A), but also had effects on modulating PKA and GSK-3 phosphorylation/activity, APP processing/Aβ level and tau phosphorylation that were much stronger than those of RPS23RG2 and comparable to those of RPS23RG1 (Fig. 4B). These results imply that the extra carboxyl-terminus of RPS23RG2 may inhibit its effects on downstream signaling. On the other hand, the amino-terminal region of RPS23RG1 may not be functionally important since RPS23RG2CΔ37 lacks this region and has comparable effects to those of RPS23RG1.
Since RPS23RG1 is a type Ib transmembrane protein and interacts with transmembranous adenylate cyclases (16), we speculated that the transmembrane domain of the RPS23RG1 protein is important for its interaction with adenylate cyclases and for its activity. Therefore, we substituted the transmembrane domain of RPS23RG1 with that of APP or nicastrin, both of which are type I transmembrane proteins. As expected, co-immunoprecipitation studies showed that there is no interaction between adenylate cyclase 8 and the RPS23RG1 with an APP transmembrane domain substitution (Fig. 5A) or a nicastrin transmembrane domain substitution (data not shown). In addition, substitution of the RPS23RG1 transmembrane domain with that of APP or nicastrin abolished the effects of RPS23RG1 on upregulating phopshorylation of CREB and GSK-3α/β and reducing the level of Aβ (Fig. 5B).
Blast of the GenBank database with the RPS23RG1 protein sequence revealed that the amino-terminus of RPS23RG1 shares homology with the carboxyl-terminus of the rare isoform CRA_d of the human ATG10 protein, whereas RPS23RG1 does not share any homology with the common ATG10 protein (Fig. 6A). However, detailed sequence analyses revealed that the human ATG10 CRA_d isoform did not originate through retroposition of the human Rps23 mRNA but rather acquired the homology by ‘hijacking’ an exonal domain of the human Rps23 gene: the human ATG10 gene is localized on chromosome 5 next to, but reversely transcribed relative to, the human RPS23 gene (Fig. 6A). Therefore, a part of exon 4 of human RPS23 is recruited by the ATG10 CRA_d isoform during its transcription and translation, encoding the ATG10 CRA_d isoform's carboxyl-terminus. Since Rps23 is highly conserved among different species and mouse Rps23rg1 is reversely transcribed relative to mouse Rps23, the amino-terminus of RPS23RG1 becomes homologous to the carboxyl-terminus of ATG10 CRA_d isoform (Fig. 6A). The differences in origination suggest that the ATG10 CRA_d isoform should not be considered a member of the Rps23rg family. Although ATG10 is thought to be an E2-like enzyme and involved in two ubiquitin-like modifications essential for autophagosome formation (19), the exact function of ATG10 CRA_d isoform remains unknown. However, it is unlikely that ATG10 CRA_d isoform functions similarly to RPS23RG1 through the homologous domain, since the amino-terminus of RPS23RG1 is dispensable for its function (Fig. 4). Indeed, overexpression of ATG10 CRA_d isoform, as well as ATG10, had no effects on promoting phosphorylation of CREB and GSK-3α/β or on reducing the level of Aβ as RPS23RG1 did (Fig. 6B).
Here we have identified a new gene family that originated through retroposition of the Rps23 mRNA. RPS23 belongs to the ribosomal protein family which is crucial for ribosome function (17). Mammals have 79 ribosomal proteins. Interestingly, it has been found that for most if not all human ribosomal proteins, the single, functional gene has generated a large number of processed pseudogenes (18), suggesting that transcripts of these ribosomal protein genes have a tendency to retropose. Indeed, we have found multiple Rps23 retroposition events in humans and many more in mice and rats (Fig. 3). In addition, we found retroposition of Rps23 in rhesus monkeys and dogs (data not shown). Although most of the retroposed gene copies probably have become processed pseudogenes, in the case of the Rps23rg gene family, at least for Rps23rg1 and Rps23rg2, the retroposed gene copies have recruited host chromosome sequences adjacent to the incorporation sites as new regulatory elements and/or exon/intron units to form functional genes. However, the Rps23rg gene family members are reversely transcribed relative to the parental Rps23 mRNA, thereby encoding proteins with no sequence homology to those of RPS23 proteins.
Since RPS23RG family members show beneficial functions for inhibiting AD-like pathologies (Aβ production and tau phosphorylation), one priority is to identify any human homologs for potential therapeutic targeting. However, our database search found that only the human ATG10 CRA_d isoform shares homology to a dispensable domain of RPS23RG1. In addition, ATG10 CRA_d isoform did not have the same effect as RPS23RG1 and ATG10 CRA_d isoform did not originate in the same way as the Rps23rg family members (Fig. 6). Furthermore, although we found that the predicted Rps23rg5 gene originated before the divergence between humans and chimpanzees, our RT–PCR amplification with primers specific to the predicted Rps23rg5 cDNA failed to generate any positive result from various human tissue/cell samples (data not shown), suggesting that there is no functionally expressed Rps23rg5 in humans. On the other hand, we have demonstrated that the transmembrane domain of RPS23RG1 is important for its function (Fig. 5), whereas the predicted RPS23RG5 lacks this transmembrane domain (Fig. 1A). Therefore, even if RPS23RG5 is present in humans, it might not have the same function as that of RPS23RG1. Moreover, we predicted genes within the regions covering all the identified Rps23-like sequences in the human genome and found no Rps23rg-like genes. RT–PCR with primers binding regions right next to these human Rps23 retroposition sites also failed to obtain positive amplification (data not shown). Hence, although we cannot completely exclude the possibility of existence of functional Rps23rg homologous gene in humans, we believe that the chance will be low.
During aging, humans are susceptible to the pathogenesis of AD which is typically characterized by Aβ overproduction/aggregation and tau hyperphosphorylation. In contrast, wild-type mice rarely develop AD-like pathologies (8–10). The differences in AD susceptibility between humans and mice have been attributed to the sequence disparity between human and mouse Aβ (and possibly tau) that underlie different aggregation capabilities (8–10), the short lifespan of mice relative to humans (10,11) and the differences in processing of human and mouse APP by BACE1 (12). The presence of the functional Rps23rg family members in mice may provide additional explanation for why mice are resistant to AD pathogenesis. On the other hand, although it is unlikely that humans possess Rps23rg family homologs, RPS23RG family members can reduce Aβ level and tau phosphorylation not only in mouse cells (data not shown), but also in human cells (Fig. 4B), suggesting that RPS23RG protein family-mediated signaling pathways are active in humans and these pathways may be potential AD therapeutic targets.
We blasted the GenBank database with the RPS23RG1 protein sequence to identify its homologs. Rps23 cDNA sequences of humans, mice and rats were used to screen for Rps23-like sequences in the respective species' genome. Homologous sequences were aligned manually. For most identified sequences, we selected their homologous regions covering the complete RPS23 protein-encoding sequence to construct their phylogenic relationships, using the maximum parsimony method embedded in MEGA (20) and the dog Rps23 sequence as a root. We generated a consensus tree with a 50% cut-off of a 1000 bootstrap analysis. For other sequences without a homologous region covering the complete RPS23 protein-encoding sequence, we used whatever homologous region they share with other sequences to construct consensus trees. All generated consensus trees were finally combined into one consensus tree. The potential transmembrane region in the RPS23RG1 protein was predicted using PredictProtein (21). For identified Rps23-like sequences in the human genome, we selected the fragment containing 100 Kb each of the 5′ and the 3′ region adjacent to these sequences for gene prediction, using Genscan (22,23).
Maintenance of mouse neuroblastoma N2a cells, N2a cells stably expressing human APP Swedish mutation and human HeLa cells stably expressing human APP Swedish mutation has been previously described (16).
Total RNAs were extracted from tissues of a 2-month-old C57BL6 mouse using TRIzol reagent (Invitrogen, Carlsbad, CA, USA). After treatment with RNase-free DNase, equal amounts of RNAs were reverse-transcribed into first-strand cDNA using SuperScript First-Strand kit (Invitrogen). Synthesized cDNAs were used as templates for PCR with pairs of primers amplifying the Rps23rg1 cDNA (5′-primer: TGGGTGAGGGCGTGCAGCTGT; 3′-primer: ACCTTGCCCACTTGTTTGCTTTTC), the Rps23rg2 cDNA (5′-primer: GGCCCAAGAGAAAGCATTC; 3′-primer: TTGCACATTTAGCCTTAGAAC) and a conserved region of the human and the mouse Gapdh cDNAs (5′-primer: CCCTTCATTGACCTCAACTA; 3′-primer: CCTTCTCCATGGTGGTGAA).
Vectors expressing RPS23RG1, RPS23RG2 or RPS23RG2CΔ37 (RPS23RG2 lacking 37 amino acids at its carboxyl-terminus) were constructed in the pCMV-Myc vector (Clontech, Mountain View, CA, USA) with a Myc tag at the amino-terminus and a His6 tag (for RPS23RG1) or a HA tag (for RPS23RG2 and RPS23RG2CΔ37) at the carboxyl-terminus. Vectors expressing ATG10 and ATG10 CRA_d were constructed in pcDNA3.1 with a Myc-His tag at the carboxyl-terminus. A vector expressing RPS23RG1 was also constructed using the same strategy as a control. For transmembrane domain replacement studies, the transmembrane domain of RPS23RG1 (amino acids 96–118) in the pcDNA3.1 vector was substituted with that of APP (amino acids 625–648, based on APP695) or with that of nicastrin (amino acids 670–690). Cells were transiently transfected with these vectors using Lipofectmine reagent (Invitrogen). Aβ in conditioned media was either precipitated with trichloroacetic acid or immunoprecipitated with the antibody 6E10 (Covance, Emeryville, CA, USA) and then immunoblotted with 6E10. Soluble APPα (sAPPα) in conditioned media was also immunoblotted with 6E10. Cell lysates were subjected to SDS–PAGE and immunoblot with antibodies against phosphorylated CREB (Ser133) indicative of PKA activity, total CREB, phosphorylated GSK-3α/β (Ser21/9, inactive form) and total GSK-3α/β (Cell Signaling, Danvers, MA, USA), the 369 antibody recognizing APP/APP CTFs (24) and the Myc antibody 9E10 (Santa Cruz Biotechnology, Santa Cruz, CA, USA). In some experiments, cells were first transfected with a human tau vector (25) and cell lysates were additionally immunoblotted with antibodies against total tau (Abcam, Cambridge, MA, USA) and hyperphosphorylated PHF-1 tau (kindly provided by P. Davies at Albert Einstein College of Medicine). Protein levels were quantified by densitometry and normalized to those of controls for comparison (set as one arbitrary unit), using two-tailed Student's t-test. In some experiments, transfected cells were subjected to biotinylation and affinity precipitation with streptavidin beads (26) before SDS–PAGE to detect cell surface levels of RPS23RG proteins.
N2a cells were transfected with RPS23RG1, RPS23RG2, RPS23RG2CΔ37 or RPS23RG1 with an APP transmembrane domain or nicastrin transmembrane domain substitution, together with adenylate cyclase 8 (AC8) plasmid (kindly provided by D. R. Storm and G. Chan at University of Washington). Cells were then lysed in either CHAPSO buffer (1% CHAPSO, 25 mm HEPES, pH 7.4, 150 mm NaCl and 2 mm EDTA supplemented with protease inhibitors) or in NP40 buffer (1% NP40 in phosphate buffered saline, supplemented with protease inhibitors). Lysates were immunoprecipitated using mouse IgG, rabbit IgG and antibodies against Myc or adenylate cyclases (Santa Cruz Biotechnology) and Trueblot™ IP beads (eBioscience, San Diego, CA, USA), followed by immunoblot with the Myc antibody.
This work was supported in part by National Institutes of Health grants (R01 NS046673, R01 AG030197 and R01 AG021173 to H.X., and R01 NS054880 to F.-F.L.), and grants from the Alzheimer's Association (to H.X. and F.-F.L.), the American Health Assistance Foundation (to H.X.), National Natural Science Foundation of China (30973150 to Y.-w.Z.), National S&T Major Project (2009ZX09103-731 to Y.-w.Z.), the 973 Prophase Project (2010CB535004 to Y.-w.Z.) and Natural Science Funds for Distinguished Young Scholar of Fujian Province (2009J06022 to Y.-w.Z.). Y.-w.Z. is supported by the Program for New Century Excellent Talents in Universities (NCET), the Program for New Century Excellent Talents in Fujian Province Universities (NCETFJ) and Fok Ying Tung Education Foundation.
We thank P. Davies for providing the PHF-1 tau antibody and D.R. Storm and G. Chan for providing the adenylate cyclase 8 vector.
Conflict of Interest statement. None declared.