|Home | About | Journals | Submit | Contact Us | Français|
Amyloid β-protein precursor (AβPP) is cleaved by β- and γ-secretases to liberate amyloid beta (Aβ), the predominant protein found in the senile plaques associated with Alzheimer’s disease (AD) and Down syndrome (Masters et al., 1985). Intense investigation by the scientific community has centered on understanding the molecular pathways that underlie the production and accumulation of Aβ Therapeutics that reduce the levels of this tenacious, plaque-promoting peptide may reduce the ongoing neural dysfunction and neuronal degeneration that occurs so profoundly in AD. AβPP and Aβ production are highly complex and involve still to be elucidated combinations of transcriptional, post-transcriptional, translational and post-translational events that mediate the production, processing and clearance of these proteins. Research in our laboratory for the past two decades has focused on the role of RNA binding proteins (RBPs) in mediating the post-transcriptional as well as translational regulation of APP messenger RNA (mRNA). This review article summarizes our findings, as well as those from other laboratories, describing the identification of regulatory RBPs, where and under what conditions they interact with APP mRNA and how those interactions control AβPP and Aβ synthesis.
The APP gene is located on human chromosome 21 and codes for AβPP, a ubiquitously expressed, transmembrane protein that localizes to post-synaptic densities, axons, dendrites and neuromuscular junctions (NMJ) (Akaaboune et al., 2000; Shigematsu et al., 1992), consistent with its multiple roles in cell adhesion (Soba et al., 2005), synapse formation (Torroja et al., 1999; Yang et al., 2005) and synapse maturation (Akaaboune et al., 2000). Differential processing of AβPP by non-amyloidogenic (α- and γ-) or amyloidogenic (β- and γ-) secretases produces soluble AβPP (sAβPPα and sAβPPβ), truncated or full-length amyloid-beta (Aβ) and variable length carboxy-terminal fragments (CTFs). The over-production and ultimate aggregation of Aβ is associated with AD and Down syndrome pathology with current evidence suggesting a more pathogenic role for soluble, pre-aggregated Aβ. Despite decades of investigation, much remains to be learned about the normal physiological function of Aβ and other AβPP catabolites as well as the cellular and molecular pathways that regulate their generation. Clearly, a better understanding of AβPP biology and gene expression will provide new avenues for therapeutic intervention in a number of highly prevalent and currently untreatable neurological disorders.
Our laboratory has studied the mechanisms that underlie the stability and translation of APP mRNA. These studies were triggered by the realization that the 3’ untranslated regions (3’ UTR) of mouse, rat, human and other species of APP mRNA contained islands of substantial homology, consistent with the presence of cis-acting elements. We hypothesized that APP mRNA was subject to activation-dependent changes in stability and/or translation. Such events could account for both increased mRNA under steady state levels as well as increased translation. Indeed, we and other groups have successfully defined multiple cis-regulatory elements within the 5’-UTR, coding region (CR) and 3’-UTR of APP mRNA that mediate the post-transcriptional stability and/or translation of APP mRNA (Zaidi and Malter, 1994; Amara et al., 1999; Rogers et al., 2002; Westmark et al., 2006; Westmark and Malter, 2007) as well as identified a number of RBPs that bind to these cis-elements (Zaidi et al., 1994; Zaidi and Malter, 1995; Westmark and Malter, 2007; Broytman et al., 2008; Lee et al., 2010; Cho et al., 2010). These mRNA/RBP complexes control the degradation, stability and/or translation of APP mRNA, which directly determines how much AβPP is synthesized and thus available for processing towards or away from amyloidogenic endpoints.
This review article summarizes our findings, as well as those from the Rogers, Amara and Gorospe laboratories, describing the mRNA/RBP complexes and signaling pathways that regulate AβPP expression. A detailed understanding of the molecular interactions that regulate AβPP synthesis may provide novel avenues for therapeutic intervention in the treatment of AD and other Aβ-related disorders.
mRNA stability is an important control point in gene expression. The mRNAs of certain classes of regulatory proteins, such as oncogenes, cytokines, lymphokines and transcriptional activators are extremely labile (Brewer, 1991; Shaw and Kamen, 1986). Many of these mRNAs contain a common AUUUA pentamer in their 3’-UTR, which confers message instability through interactions with RBPs, such as AU-rich binding factor (AUBF) (Malter, 1989; Gillis and Malter, 1991), AUF1 (Wang et al., 1998) or TTP (Lai et al., 1999). In the early 1990s, we were interested in defining the contribution of mRNA stability to steady-state levels of APP mRNA in neurons. A large number of AU-rich mRNAs are expressed in the central nervous system and we hypothesized that APP would also be a labile mRNA due to the presence of multiple 3’-UTR AUUUA motifs in its 3’-UTR. The first of its four AUUUA elements in the context of an AU-rich region is located approximately 65 bases downstream from the stop codon in both human and murine APP mRNA. The AUUUA motif is frequently identified in a similar location in other labile mRNAs (Caput et al., 1986; Shaw and Kamen, 1986). Somewhat unexpectedly, we found that APP mRNA has a half-life of 4–5 hr in primary cells, which is intermediate between labile and stable housekeeping genes (Zaidi and Malter, 1994; (Westmark and Malter, 2001a). These data suggested, and were later confirmed, that the aggregate content of AUUUA motifs did not correlate well with mRNA stability. The AUUUA motifs in APP mRNA are individually arranged and separated by approximately 50 bases whereas these motifs are tandemly stacked (AUUUAUUUA etc.) in labile mRNAs such as IL-2 or GM-CSF (Lagnado et al., 1994; Zubiaga et al., 1995). Thus, the moderate instability of APP mRNA in resting cells is consistent with an alternative cis-element presumably distinct from the AUUUA motifs.
Indeed, deletion mapping of the 3’-UTR indicated that a pyrimidine-rich 29-base element (29-be) approximately 200 bases downstream from the stop codon (bases 2285–2313) mediated APP mRNA decay (Zaidi et al., 1994). This element is highly conserved in human and murine sequences and is 69% AU-rich although it lacks the canonical AUUUA motifs (Zaidi et al., 1994). Mutant APP mRNA containing a randomized 29-be was far more stable than WT with a half-life >10 hr (Zaidi and Malter, 1994). Later, we showed that heterologous mRNAs containing the 29-be in their 3’-UTR showed similar destabilization (Westmark et al., 2006). Interestingly, when repositioned into the CR by deletion of the stop codon, the 29-be failed to function (Westmark et al., 2006). These results suggested a topologic constraint on the element. We hypothesized that the transiting ribosome might displace a protein required for function and/or the tertiary conformation of the mRNA. Thus, the 29-be clearly functions in cis to destabilize APP mRNA and must be located within the 3’-UTR, not the CR.
The majority of studies regarding APP mRNA stability have focused on the 3’-UTR. As APP mRNAs are produced by alternative splicing of a single copy gene and all share an identical 3’-UTR, all resultant APP mRNAs would be co-regulated. However, whether this is actually true has not been rigorously tested in cells. In addition to the 29-be, two additional, and opposing cis-regulatory elements in the 3’-UTR of APP mRNA have been identified. Amara and colleagues found an 81-be (positions 2730–2810) that destabilized APP mRNA (Amara et al., 1999), and our laboratory identified a 52-be immediately downstream from the translation stop codon that stabilized APP mRNA (Westmark et al., 2006). The proximal 200 bases of the 3’-UTR of APP mRNA thus contain both the 29-be and 52-be. Sequence conservation over the entirety of this region suggested maintenance of an important regulatory element (Westmark et al., 2006). Computer modeling of this 250 base segment predicts a highly stable structure with the terminal 29-be and 52-be in apposition to each other (Westmark et al., 2006). To test if the aggregate structure and/or the individual elements controlled APP mRNA decay, we deleted the entire 200 base intervening sequence between the stop codon and the 29-be. These mRNAs showed similar half-lives to those simply lacking the 52-be (Westmark et al., 2006). To ensure this was due to the loss of specific regulatory sequence rather than spacing and/or changes in conformation, we deleted the 54 bases immediately upstream from the 29-be. These mutant mRNAs had near normal stability (Westmark et al., 2006). Thus, the 52-be is most likely a bona fide stabilizer of APP mRNA that acts independently but in opposition to the 29-be.
In addition to the 3’-UTR elements, there are also cis-regulatory elements in the 5’-UTR and CR of APP mRNA. As for other mRNAs, element location often (but not always) has functional repercussions. Thus not surprisingly, CR and 5’-UTR locations have been associated with translational control rather than mRNA stability. For example, Rogers and coworkers identified a type II iron response element (IRE) in the 5’-UTR of APP mRNA (+51 to +144 from the 5’ cap site) (Rogers et al., 2002). This stem loop RNA IRE was highly homologous to translational control elements in the 5’-UTR of the light (L) and heavy (H) ferritin genes and was located immediately upstream from an interleukin-1 (IL1) responsive acute box domain (+101 to +146) (Rogers et al., 1999; Rogers et al., 2002). Recently, we identified a guanine-rich area in the CR of App mRNA that bound to fragile X mental retardation protein (FMRP) [APP refers to the human and App to the mouse transcript]. Binding was maintained under basal conditions but was lost within minutes of metabotropic glutamate receptor 5 (mGluR5)-mediated signaling. As App mRNA translation was increased in the absence of FMRP binding (after mGluR5 agonists), these data suggest the CR element mediates translational repression (Westmark and Malter, 2007). These results were later confirmed by the Gorospe laboratory that further demonstrated that this element differentially interacts with FMRP and hnRNP C to modulate translation (Lee et al., 2010).
Given the numerous cis-elements described above, it is hardly surprising that multiple RBPs have been identified that interact with APP mRNA. Many of these proteins are known to bind to multiple mRNAs and have diverse functions including nuclear/cytoplasmic shuttling, RNA helicase activity, iron homeostasis and translational repression. Interestingly, with the exception of the 5’-UTR IRE-like domain, the cis-elements in APP mRNA do not show obvious or substantial homology to other mRNAs. As it is unlikely that RBPs have evolved to specifically regulate APP mRNA, these results suggest that the tertiary structure assumed by the element likely plays an important role in RBP recognition and binding.
The search for RBPs began by probing whole cell lysates from neuronal tumor lines with radioactive fragments of the 29-be. After relatively short incubations under conditions of physiologic salt and high concentrations of irrelevant, competitor RNA (typically tRNAs), reactions were treated with RNase to digest nonspecific interactions prior to nondenaturing gel electrophoresis. RBP/RNA interactions were identified by the presence of a slowly migrating band above that of free radioactive RNA in an electrophoretic gel mobility shift assay (EMSA). Specificity was established by the competition of the slowly migrating complex with increasing amounts of unlabeled probe RNA but not with similar concentrations of irrelevant RNA. The molecular mass of the complex was determined by treating reactions with UV light prior to SDS-PAGE. These methods detected multiple RBPs in cytosolic lysates prepared from normal and transformed human cells (Zaidi et al., 1994).
Obviously, identification of these RBPs was a critical next step prior to characterization of their role in APP mRNA stability. Through classical, biochemical, chromatographic methods, we fractionated cytosolic lysates prepared from human leukemia T-cells and followed binding activity by RNA EMSAs. Ultimately, the 29-be interacted in vitro with five cytoplasmic RBPs of 70, 48, 47, 39 and 38 kDa (Zaidi et al., 1994). Northwestern blotting with radiolabeled RNA confirmed that the purified proteins interacted specifically with the 29-be. After SDS-PAGE isolation and mass spectrometry analysis, the p47, p48 and p70 RBPs were identified as fragments of 110-kDa nucleolin and the p38 and p39 were identified as hnRNP C1 and hnRNP C2, respectively (Zaidi and Malter, 1995). The 70 kDa fragment of nucleolin and the hnRNP C proteins bound to the 5’-half of the 29-be while the 47 and 48 kD fragments of nucleolin likely bound to the 3’-end (Zaidi and Malter, 1995).
hnRNP C1 and C2 are members of a family of RBPs involved in the packaging of heterogeneous nuclear RNA (hnRNA) and its conversion to mRNA in the nucleus (Choi and Dreyfuss, 1984), and they showed some specificity for uridine-rich sequences (Wilusz and Shenk, 1990). hnRNP A1 and C were previously identified as RBPs that shuttle between the nucleus and cytoplasm and were capable of specifically binding to reiterated AUUUA sequences found in the 3’-UTRs of labile mRNAs (Hamilton et al., 1993). The hnRNP A and C families have been implicated in chromatin remodeling (Mahajan et al., 2005), replicator-binding protein (Huang et al., 2011), mammalian telomerase (Ford et al., 2000) and internal ribosome entry site complexes (Holcik et al., 2003). They have established roles in the prevention of transcriptional silencing (Huang et al., 2011), splicing (Choi et al., 1986) and translation (Millard et al., 2000; Holcik et al., 2003).
Nucleolin is a 110 kDa multifunctional phosphoprotein that plays roles in chromatin condensation (Erard et al., 1988), the transcription and processing of rRNA (Jordan, 1987; Egyhazi et al., 1988; Belenguer et al., 1989; Bouvet et al., 1998; Ginisty et al., 1998), transcriptional regulation (Yang et al., 1994), cell proliferation (Derenzini et al., 1995), differentiation and maintenance of neural tissue (Kibbey et al., 1995), apoptosis (Brockstedt et al., 1998), nuclear/cytoplasmic shuttling (Borer et al., 1989), mRNA stability (Zaidi et al., 1994) and mRNP assembly and masking (Yurkova and Murray, 1997). The carboxy-terminus of nucleolin has intrinsic RNA helicase activity capable of unwinding mRNA and thereby increasing its susceptibility to RNase attack (Tuteja et al., 1995). Nucleolin has been implicated in the post-transcriptional regulation of several mRNAs (Sengupta et al., 2004; Singh et al., 2004; Fahling et al., 2005). We discuss the opposing stabilizing (hnRNP C) and destabilizing (nucleolin) activities of these proteins in regard to APP mRNA post-transcriptional stability in the next section.
The 52-be adjacent to the stop codon in APP mRNA bound to several RBPs based on EMSAs when interrogated with whole cell lysates from K562 cells (Broytman et al., 2008). We purified interacting proteins by RNA affinity chromatography using a biotinylated 52-be tethered to a streptavidin affinity matrix. Protein fractions were analyzed by northwestern blotting and active bands were excised and identified by mass spectrometry. We identified 6 proteins including RCK/p54, nucleolin, plasminogen activator inhibitor-RNA binding protein 1 (PAI-RBP1), autoantigen La, elongation factor 1α (EF1α) and Y box binding protein (YB1). We confirmed these RBPs as true 52-be targets by performing EMSA with K562 lysates and radiolabeled 52-be RNA followed by excising the protein/RNA complexes, running them on SDS-PAGE and immunoblotting against the putative targets (Broytman et al., 2008). With the exception of nucleolin, which binds to the 29-be, these RBPs were not known to bind to APP mRNA. EMSA with radiolabeled RNA of the first 250 bases of the APP 3’-UTR produced 3 complexes. All six 52-be binding proteins were found in the heaviest migrating complex. The intermediate migrating complex contained nucleolin, autoantigen La and YB1, and the fastest migrating complex only contained YB1 and La (Broytman et al., 2008).
Co-immunoprecipitation studies indicated that nucleolin and EF1α both co-immunoprecipitated with all of the other 5 proteins (Broytman et al., 2008). However, the co-immunoprecipitations of nucleolin with La or YB1 was RNA-dependent, whereas its co-immunoprecipitation with RCK/p54, PAI-RBP1 and EF1α was RNA-independent. Likewise, EF1α co-immunoprecipitations with La and YB1 were RNA-dependent and co-immunoprecipitations with nucleolin, RCK//54 and PAI-RBP1 were RNA independent (Broytman et al., 2008). These data suggest that the RBP complex at the 52-be in the 3’-UTR of APP mRNA contains both protein/protein interactions (nucleolin, EF1α, RCK/p54 and PAI-RBP1) and protein/mRNA interactions (YB1, La) (Broytman et al., 2008).
To confirm that these in vitro interactions mirror those seen in intact cells, we immunoprecipitated all six RBPs from the cytosol fraction of SH-SY5Y human neuroblastoma cell line and demonstrated by RT/qPCR analysis that APP mRNA was specifically present in the co-immunoprecipitate (Broytman et al., 2008). The RBPs involved in protein/52-be RNA interactions are YB1 and La. As there is little sequence homology between their known mRNA targets and APP mRNA, YB1 and La binding at the 52-be are likely mediated by the secondary or tertiary structure of the RNA. Nucleolin, on the other hand, while a well-established RBP, is tethered to the 52-be/protein complex via protein/protein interactions. YB1 is a member of the highly conserved cold shock domain family of proteins and is a multifunctional nucleic acid binding protein implicated in transcriptional activation and repression, cellular and viral mRNA binding, mRNA stabilization and translational control (Ansari et al., 1999; Bouvet et al., 1995; Capowski et al., 2001; Chen et al., 2000; Didier et al., 1988; Ranjan et al., 1993; Stenina et al., 2000). The autoantigen La is a well-characterized RBP that has been shown to stabilize histone mRNA (McLaren et al., 1997).
The other RBPs that form protein/protein interactions at the 52-be are EF1α, RCK/p54 and PAI-RBP1. EF1α regulates the transfer of aminoacyl-tRNAs to the ribosome and is involved in the nuclear export of proteins (Khacho et al., 2008). Broytman and colleagues were the first to show that it bound to an mRNA and possibly functioned in mRNA stabilization (Broytman et al., 2008). Its role in the posttranscriptional regulation of APP mRNA could involve utilization of energy from GTP hydrolysis to facilitate the binding of cofactors, to catalyze a change in secondary/tertiary structure of the 52-be or the energy-dependent transport of the RBP/APP mRNA complexes out of the nucleus (Broytman et al., 2008). RCK/p54 is a member of the DEAD-box helicase family of proteins, which have known roles in splicing, ribosome biogenesis, RNA transport, degradation and translation (Akao et al., 1995). RCK/p54 is a component of both processing (P)-body and cytoplasmic polyadenylation element-binding protein (CPEB) translation repressor complexes (Minshall et al., 2007; Minshall et al., 2009). PAI-RBP1 is an RBP that interacts with the adenosine-rich region of the cyclic nucleotide-responsive sequence in 3’-UTR of type-1 plasminogen activator inhibitor mRNA (Heaton et al., 2001) to regulate post-transcriptional stability (Heaton et al., 2003). The 52-be does not contain an adenosine-rich stretch consistent with its RNA independent binding to PAI-RBP1.
Iron homeostasis is regulated post-transcriptionally in mammalian cells through iron regulatory proteins (IRP). IRP1 and IRP2 bind with high affinity to iron response element (IRE) stem-loop structures in the transcripts of iron storage proteins such as ferritin (Kim et al., 1995; Thomson et al., 2000). IRP are released from L- and H-ferritin transcripts during iron influx resulting in the latter’s increased translation (Cox et al., 1991; Clarke et al., 2006). IRP1 modulates AβPP translation by binding to an IRE stem-loop structure in the 5’-UTR of the APP mRNA (Cho et al., 2010). Intracellular iron chelation increases the binding of IRP1 to the IRE thereby decreasing AβPP expression. Conversely, shRNA knockdown of IRP1 increases AβPP translation (Cho et al., 2010). Interestingly, the light and heavy ferritin messages bind to both IRP1 and IRP2 whereas only IRP1 binds to the IRE of APP mRNA (Cho et al., 2010).
FMRP is a multi-functional mRNA binding protein involved in the transport, localization and translational regulation of a subset of dendritic mRNAs and is required for normal dendrite development (Santoro et al., 2011). The protein has 2 hnRNP K homology domains and 1 RGG box as well as nuclear localization and export signals. The RGG box of FMRP binds to intramolecular G-quartet sequences in target mRNAs (Darnell et al., 2001) while the KH2 domain has been proposed to bind to kissing complex RNAs based on in vitro selection assays (Darnell et al., 2005). In dendrites, FMRP is predominantly found in the post-synaptic density (PSD) associated with polysomes or nontranslating ribonucleoprotein (RNP) particles. The RNPs contain the FMRP homologs FXR1 and 2, YB1, nucleolin, the guide RNA BC1 and mRNA targets (Bagni and Greenough, 2005). Hundreds of mRNA ligands have been identified (Brown et al., 2001; Chen et al., 2003; Dolzhanskaya et al., 2003; Miyashiro et al., 2003) with many having the potential to influence synapse formation and synaptic plasticity. FMRP has been implicated in translational repression (Brown et al., 2001; Laggerbauer et al., 2001; Li et al., 2001; Mazroui et al., 2002; Miyashiro et al., 2003; Zalfa et al., 2003), and in brain, co-sediments with both translating polyribosomes (Stefani et al., 2004) and with mRNPs (Zalfa et al., 2003). In addition to G-quartets and kissing complex structures, FMRP binds to uridine-rich mRNAs (Chen et al., 2003; Dolzhanskaya et al., 2003). It has been proposed that FMRP is an “immediate early protein” at the synapse orchestrating synaptic development and plasticity (Irwin et al., 2000). FMRP binds to a guanine-rich sequence in the CR of App mRNA (Westmark and Malter, 2007) and competes for binding to this sequence with hnRNP C (Lee et al., 2010). When FMRP is bound, translation is repressed. When hnRNP C is bound, translation is active. FMRP also protects the 29-be in the 3’-UTR of App mRNA from ribonuclease digestion (Westmark and Malter, 2007), but it is not known at this time if FMRP binds directly to the 3’-UTR or if protection occurs through protein/protein interactions.
Post-transcriptional gene regulation is a complex process involving interaction of cis-regulatory elements in the 3’-UTR of mRNAs with trans factors or RBPs. These interactions are regulated by cell signaling events and likely result in alterations in higher order structures of the mRNAs. A summary of the identified cis-regulatory elements and trans factors involved in the pos-transcriptional gene regulation of APP mRNA is depicted in Figure 1. The role of microRNAs (miRNAs) is also evolving as an important control mechanism in post-transcriptional gene regulation. The effects of these cis-elements and trans-factors on message stability and translation can be studied by supplementation of a rabbit reticulocyte lysates (RRL) translation system with in vitro synthesized mRNAs harboring mutated cis-elements, transduction of bacterially expressed TAT-conjugated RBPs or transfection of cDNA constructs into cells of choice.
All APP mRNA isoforms share a common 1,134 base 3’-UTR, suggesting that the above described elements and their cognate RBPs can coordinately regulate all forms simultaneously. We directly examined the role of these trans factors in the turnover and translation of APP mRNA in RRL. A mutant APP mRNA containing a randomized 29-be (MT) was 3–4-fold more stable and synthesized 2–4-fold more AβPP than WT APP mRNA in RRL (Rajagopalan et al., 1998). Therefore, the 29-be functioned as an APP mRNA destabilizer in this system. RNA EMSAs with the RRL suggested the presence of endogenous nucleolin, but failed to show hnRNP C binding activity. However, WT APP mRNA was stabilized and coded for 6-fold more AβPP when translated in RRL system supplemented with exogenous hnRNP C. Translation of the MT APP mRNA was not affected by supplementation of the RRL with hnRNP C (Rajagopalan et al., 1998). Therefore, interactions with hnRNP C stabilized APP mRNA when those events occurred in the 3’-UTR. As described later, when hnRNP C bound to the guanine-rich CR domain, the resulting phenotype showed enhanced translation without a change in stability (Lee et al., 2010). Thus, depending on the site of interaction, RBPs can affect distinct components of the post-transcriptional regulatory system.
The 29-be also functions as a destabilizer in intact, proliferating human umbilical vein endothelial cells (HUVEC) (Rajagopalan and Malter, 2000). Optimal proliferation of HUVEC requires medium containing epidermal growth factor (EGF), basic fibroblast growth factor (bFGF), insulin-like growth factor (IGF-1) and vascular endothelial growth factor (VEGF). We transfected HUVECs with in vitro transcribed chimeric mRNAs containing the 5’-UTR and CR of beta-globin attached to the entire WT or MT 3’-UTR of APP mRNA. The WT and MT globin-APP mRNAs decayed with identical half-lives. Removal of all supplemental growth factors from the culture medium significantly accelerated the decay of the transfected WT globin-APP mRNA, but only elicited a moderate decrease in the half-life of the MT mRNA (Rajagopalan and Malter, 2000). Thus, the destabilizing function of the 29-be in the 3’-UTR of APP mRNA can be modulated through alterations in growth factor-mediated signaling.
The situation with the 52-be is more complex. As described above, nucleolin, PAI-RBP1, RCK/p54, autoantigen La, EF1α and YB1 complex with this domain (Westmark et al., 2006; Broytman et al., 2008). Since RCK/p54 belongs to a well-conserved family of RNA helicases with known roles in RNA degradation (Akao et al., 1995; Coller et al., 2001; Tanner and Linder, 2001; Chu and Rana, 2006; Weston and Sommerville, 2006), we first investigated the effects of altering its levels on APP mRNA and protein levels in human cells (Broytman et al., 2008). Augmentation of intracellular RCK/p54 protein levels, either through TAT-protein transduction or through cDNA transfection of RCK/54 constructs resulted in increased APP mRNA stability and AβPP production (Broytman et al., 2008).
In addition to the 29-be and 52-be, there is a more 3’ cis-element that also mediates mRNA decay. The half-life of APP mRNA is increased by binding of an unidentified protein to this 81 nucleotide fragment in the 3’-UTR to form a 68 kDa RNA-protein complex, which is responsive to TGFβ Amara et al., 1999). The 3’-UTR of App mRNA also contains a cytoplasmic polyadenylation element (CPE) at nucleotide 2853 and 2-AAUAAA elements at bases 3098 and 3107 (mouse sequence, accession #NM_007471). Both polyadenylation sites in App mRNA are utilized with the longer mRNA showing more efficient ribosomal mobilization and translation than the shorter (de Sauvage et al., 1992).
The first 247 bases of the 3’-UTR of APP mRNA, which include the 29-be and 52-be, are highly conserved among species and computer modeling suggests that these two cis-elements could associate in apposition to each other to form a stable structure (Westmark et al., 2006). Nucleolin and/or RCK/p54, which have RNA helicase activity (Akao et al., 1995; Tuteja et al., 1995), could mediate this RNA/RNA interaction. The hypothesis that the 2 elements interact suggests functional consequences. Consistent with this, chimeric mRNAs composed of the globin 5’-UTR and CR fused to the 3’-UTR of APP mRNA were unstable when the 52-be was removed (Westmark et al., 2006). Transfection of SH-SY5Y cells with chimeric mRNAs containing the GFP CR and APP 3’-UTR resulted in down-regulation of endogenous AβPP, but only when both the 52-be and 29-be were present in the chimera (Westmark et al., 2006). These data suggest that the 52-be acts in opposition to the 29-be to control APP mRNA stability. Since mRNA produced from transfected DNA containing the 3’-UTR of APP reduced endogenous AβPP levels (Westmark et al., 2006), our data suggest that the overexpression of exogenous APP 3’-UTR sequestered RBPs that normally functioned to stabilize endogenous APP mRNA. With reduced APP mRNA stability, less AβPP was translated. These data may also explain the observation that AβPP levels are commonly reduced in transgenic animals expressing human APP mRNAs (Rockenstein et al., 1995). From a therapeutic perspective, decoy APP 3’-UTR DNA constructs could be targeted to neuronal cells to reduce endogenous APP mRNA and protein expression in diseases such as AD and Down syndrome (Westmark et al., 2006).
Growth factors activate a number of cell surface receptors and downstream signaling pathways. We assessed the effect of agonists and antagonists of several well-studied receptors and second messenger systems on APP mRNA levels (Westmark and Malter, 2001a). Drugs that affected protein kinase A, protein kinase C and phospholipase C signal transduction pathways did not alter APP mRNA levels significantly; however, activation of extracellular-regulated kinase (ERK), a member of the mitogen-activated protein kinase (MAPK) family, as well as calcium signaling modulated APP mRNA levels. We found a rapid 35–40% reduction in steady state levels of APP mRNA upon stimulation of peripheral blood mononuclear cells (PBMC) with phorbol ester or calcium ionophore, which caused a rapid activation of ERK that could be blocked with the specific inhibitor U0126. APP mRNA underwent biphasic decay upon ERK activation with a rapid, initial decrease in message quantity followed by prolonged stability. We have proposed that the initial unstable phase was due to nucleolin, and the later stable phase was a consequence of hnRNP C binding to the 29-be. Thus, depending on which RBP occupied the 29-be, APP mRNA levels were up- or down-regulated.
Conversely, ERK activation up-regulated nucleolin mRNA levels in PBMC after stimulation with phorbol ester, which was accompanied by accumulation and decreased processing of nucleolin protein (Westmark and Malter, 2001b). Higher levels of nucleolin were accompanied by increased binding of the 70-kDa-nucleolin fragment to the 29-be. In RRL, supplementation with nucleolin decreased APP mRNA stability and protein production. The initial drop in the biphasic decay of APP mRNA corresponds temporally with increased nucleolin production. Thus, our data suggests that nucleolin is upregulated in response to ERK activation and then unwinds APP mRNA allowing RNase attack. The destabilizing effect of nucleolin on App mRNA appears independent of the 29-be cis-element as the MT APP mRNA also decays rapidly.
Other signaling molecules affect the post-transcriptional stability and translation of APP mRNA through the 81 nucleotide instability element in the 3’-UTR, the IRE in the 5’-UTR or the guanine-rich sequence in the CR. TGFβ increases the half-life and steady state levels of APP mRNA through interaction of an unidentified protein with the 81 nucleotide instability element in the 3’-UTR (Amara et al., 1999). IL-1α and IL-β up-regulate AβPP synthesis through the 5’-UTR IRE without changing steady state levels of APP mRNA (Rogers et al., 1999). Activation of group 1 mGluR signaling increases the translation of AβPP by modulating the binding of FMRP to a guanine-rich sequence in the CR of APP mRNA (Westmark and Malter, 2007).
In addition to RBPs, miRNAs can also bind to cis-regulatory elements in mRNAs to post-transcriptionally alter protein production. Several miRNAs including hsa-miR-106a, hsa-miR-520c, miR-101 and miR-20 bind to the 3’-UTR of APP mRNA and negatively regulate reporter gene expression (Patel et al., 2008; Ruberti et al., 2010; Long and Lahiri, 2011).
App mRNAs are highly associated with polyribosomes in rat brain (Denman et al., 1991) suggesting that translational regulation could play an important role in AβPP production. At least two identified cis-elements regulate translation of APP mRNA including an IRE in the 5’-UTR (Rogers et al., 2002) and a guanine-rich element in the CR (Westmark and Malter, 2007; Lee et al., 2010). RBPs, which modulate ribosomal access and initiation, often mediate translational control at the synapse. Two of the best-known examples are IRP1 and FMRP.
The 5’-UTR of APP mRNA contains an IRE that binds to IRP1, which modulates AβPP translation in response to intracellular iron levels (Cho et al., 2010). Upon intracellular iron chelation, IRP1 binding to the IRE is increased and translation is decreased.
FMRP represses the translation of a subset of dendritic mRNAs containing G-quartets. We have demonstrated through crosslinking immunprecipitation (CLIP) assays that FMRP binds directly to a guanine-rich cis-element in the CR of App mRNA (Westmark and Malter, 2007). Specific activation of group 1 mGluRs (mGluR1 and mGluR5) with dihydroxyphenylglycine (DHPG) releases FMRP from App mRNA resulting in elevated translation, which can be blocked with the mGluR5-specific inhibitor MPEP. In Fmr1KO neurons, AβPP expression is constitutively elevated and non-responsive to DHPG treatment (Westmark and Malter, 2007).
The Gorospe laboratory confirmed our finding that FMRP binds to the guanine-rich region in the CR of App mRNA and represses translation (Lee et al., 2010). They were working with a human neuroblastoma cell line, BE2-M17, and we were using brain lysates from mice. Thus, FMRP is a bona fide target for the guanine-rich cis-element in both human and rodent mRNAs as well as in both established and primary cells. The Gorospe laboratory further demonstrated through CLIP analyses that hnRNP C interacts directly with the guanine-rich CR cis-element of APP mRNA and, through a series of elegant siRNA silencing and RBP transfection experiments, that FMRP and hnRNP C modulate translation in opposing directions (Lee et al., 2010). Their data suggest that FMRP and hnRNP C competitively bind to APP mRNA to modulate protein synthesis. When hnRNP C is bound, translation is enhanced, and when FMRP is bound, translation is repressed. The interaction of hnRNP C with App mRNA was significantly more abundant in Fmr1KO mice, which lack FMRP and exhibit elevated protein translation (Lee et al., 2010).
Interestingly, the Gorospe laboratory also demonstrated that APP mRNA associates with Ago1 and Ago2 proteins through RNA-independent protein/protein interactions (Lee et al., 2010). Ago1, Ago2 and RCK/p54 reside in P-bodies, which are sites where nontranslating mRNAs accumulate before sorting for transient storage or degradation (Barbee et al., 2006; Eulalio et al., 2007; Parker and Sheth, 2007; Balagopal and Parker, 2009). They further demonstrated that FMRP interacts with RCK/p54 and does not repress translation in the absence of RCK/p54, Ago1 or Ago2 (Lee et al., 2010). When FMRP is overexpressed, APP mRNA is more abundant in RCK/p54 immunoprecipitates suggesting that FMRP represses translation by enhancing the association of APP mRNA with P-bodies. Co-transfection of a pMS2-APP reporter construct expressing a chimeric RNA with 24 tandem MS2 RNA hairpins and the APP CR cis-element with plasmid pMS2-YFP, which has a NLS and expresses the chimeric fluorescent protein MS2-YFP, allowed tracking of the subcellular localization of the MS2-APP RNA by confocal microscopy. Some RNA colocalized in the cytoplasm with RCK/p54, but was lost upon overexpression of hnRNP C or silencing of FMRP. Thus, FMRP appears to repress the translation of APP mRNA by recruiting the transcript to P-bodies, and hnRNP C promotes translation by competing with FMRP for interaction with the CR cis-element thus preventing localization of APP mRNA to P-bodies (Lee et al., 2010).
Thus, there appears to be competition between stabilizing and destabilizing RBPs at multiple cis-regulatory elements of APP mRNA. At the 29-be, hnRNP C (stabilizing) and nucleolin (destabilizing) differentially modulate message stability. At the guanine-rich area in the CR, FMRP represses and hnRNP C activates protein synthesis. When we were characterizing the FMRP binding site in App mRNA, we found that the 29-be in the 3’-UTR was also protected from T1 ribonuclease digestion of anti-FMRP immunoprecipitates (Westmark and Malter, 2007). As FMRP is a known binding partner of both nucleolin and YB1 (Ceman et al., 1999), this data suggests that a large ribonucleoprotein complex forms to bring multiple cis-regulatory elements of APP mRNA in close proximity. Overall, APP mRNA contains several identified cis-elements that bind to numerous RBPs, either directly or through protein/protein interactions, to mediate mRNA stability, decay, polyadenylation and translation.
Fragile X Syndrome (FXS) is characterized by excessive mGluR signaling (Bear et al., 2004). Slices from Fmr1KO mice reveal excess LTD in response to mGluR5 agonists (Huber et al., 2002). At a molecular level, basal mGluR5 signaling is normally attenuated by FMRP through the latter’s interactions with and translational suppression of dendritic mRNAs. Upon mGluR5 activation, translation of post-synaptic mRNAs is transiently enhanced through loss of FMRP-mediated suppression (Bear et al., 2004). We have demonstrated that activation of group 1 mGluR signaling increases translation of AβPP, which can be blocked with the specific mGluR5 antagonist MPEP (Westmark and Malter, 2007). Therefore, mGluR5 and FMRP appear to play important roles in the translation and hence accumulation of AβPP at synapses. Consistent with these data, Fmr1KO mice have constitutively elevated AβPP levels and older mice exhibit increased Aβ1–40 and Aβ1–42 cortical loads compared to WT controls (Westmark and Malter, 2007). Thus, it is very tempting to speculate that mGluR5 antagonists could be used to reduce APP mRNA translation in Alzheimer’s disease as well as FXS. As discussed below, we have recently provided evidence that the pathobiology of FXS may indeed reflect altered AβPP and Aβ levels.
As Aβ is overexpressed in the brain of Fmr1KO mice, we explored the possibility that genetic reduction of AβPP and Aβ in these mice would rescue characteristic FXS phenotypes (Westmark et al., 2011). We found partial or complete rescue of audiogenic seizures, anxiety, the ratio of mature versus immature dendritic spines and mGluR-LTD after removal of one App allele (Fmr1KO/AppHET mice) (Westmark et al., 2011). Furthermore, plasma Aβ1–42 levels were significantly reduced in full-mutation FXS males compared to age-matched controls suggesting that Aβ is sequestered in the brain and could serve as a plasma-based biomarker to facilitate disease diagnosis or assess therapeutic efficacy (Westmark et al., 2011).
The molecular basis for defective synapse formation in FXS is unknown, but likely reflects the loss of FMRP-regulated translation of synaptic mRNA. Dendrites contain all of the translational machinery (mRNAs, ribosomes, tRNA and translation initiation, elongation and release factors) necessary for protein biosynthesis (Tiedge and Brosius, 1996; Inamura et al., 2003). Our research demonstrates that App mRNA is a previously unappreciated target for FMRP-mediated translational repression at the synapse.
The ultimate reason to understand the cellular and molecular pathways underlying AβPP and Aβ production is to facilitate the rationale development of reliable and safe therapeutics. All of the currently approved drugs for the treatment of AD act on healthy neurons to (1) compensate for reduced acetylcholine in the case of cholinesterase inhibitors, or (2) modulate NMDA receptor activity in the case of memantine. They improve cognitive ability for a year or less, but do not reduce amyloid plaque or neurofibrillary tangle accumulation, neuronal degeneration or subsequent disease progression.
Aβ immunotherapy has proved very effective in reducing soluble Aβ, amyloid plaque and soluble tau as well as associated cognitive decline; however, there are questions about its safety (Broytman and Malter, 2004; Morgan, 2005). Human clinical trials testing active immunization against Aβ were suspended because 6% of the patients developed meningoencephalitis (Check, 2002). In AD mouse models characterized by prominent vascular Aβ deposition (APP23 and PDAPP mice), immunization resulted in a two-fold increase in hemorrhages (Pfeifer et al., 2002; Racke et al., 2005). The alternative, passive immunization (anti-Aβ administration), may prove just as effective in reducing Aβ without the side effects. In AD mouse models, passive immunization with an antibody directed against Aβ induced clearance of preexisting amyloid in the brain (Bard et al., 2000) and reverted Aβ-associated memory deficits (Kotilinek et al., 2002). There are many on-going trials with such agents at present.
β- and γ-secretase inhibitors could provide viable therapeutics that reduce Aβ generation. Specificity remains an issue with the use of γ-secretase inhibitors as these drugs also inhibit proteolytic processing of other proteins such as Notch that are critical for cellular function. Targeting other components of the γ-secretase complex may avoid these issues (Panza et al., 2011).
Studies described herein suggest several possible alternative avenues for therapeutic intervention. Drugs that target the metallobiology of AD, such as desferrioxamine, clioquinol, tetrathiolmolybdate, dimercaptopropanol and VK-28 as well as natural antioxidants, such as curcumin and ginko biloba, could modulate IRE/IRP1 cis/trans interactions in the 5’-UTR and thereby regulate translation of AβPP (Bandyopadhyay et al., 2010). Decoy RNAs or miRNAs that target the 3’-UTR cis-regulatory elements could modulate RBP binding and thereby regulate APP message stability. Drugs influencing mGluR5/FMRP signaling could repress translation of AβPP and reduce amyloidogenic processing. Chronic dosing with the mGluR5 inhibitor fenobam reduced Aβ levels in mouse models of AD and FXS (Malter et al., 2010). This drug has never been tested as an AD therapeutic, but it has been approved for phase I & II clinical trials (Pecknold et al., 1980; Friedmann et al., 1980; Lapierre and Oyewumi, 1982; Pecknold et al., 1982). It was also designated as an orphan drug for the treatment of FXS in November of 2006. Despite these advances and on-going studies, continued research in the field of APP mRNA/RBP interactions will likely provide novel therapeutic insights to target Aβ production and plaque formation.
In conclusion, post-transcriptional gene regulation of APP mRNA is mediated by several trans factors (RBPs) that bind to cis-regulatory elements in the UTRs or CR in response to cell signaling events. Our laboratory has identified two cis-regulatory elements in the proximal 3’-UTR and one in the CR of APP mRNA while other laboratories have identified a destabilizing element in the 3’-UTR and a translational element in the 5’-UTR. Numerous RBPs interact with these cis-elements to modulate post-transcriptional message stability and translation. It is highly probable that the sophistication of the RBP complexes that regulate the stability and translation of APP mRNA resembles the complexity of the better-studied transcriptional and translational machinery. Remaining components of the APP mRNA/RBP complex(es) remain to be identified as do the post-translational modifications and signaling pathways that mediate the mRNA/protein interactions. Our work the past two decades as well as that of our colleagues has just scraped the surface of the APP mRNA/RBPs iceberg.
This work was supported by the National Institutes of Health grants R01-AG10675, RO1-DA026067, P30-HD03352, T32-AG00213, the Alzheimer’s Drug Discovery Foundation, FRAXA Research Foundation and the Wisconsin Comprehensive Memory Program. The sponsors had no role in the writing of this review article or the decision to publish. The authors thank Dr. Pamela Westmark for critical reading of the manuscript.
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.