5 December 2007. Despite a 20-year courtship, researchers have gotten to know rather little about amyloid-β precursor protein (APP). Yes, it undergoes sequential cleavage via amyloidogenic and non-amyloidogenic pathways, and it is a key player in the pathology of Alzheimer disease, but the normal function of the precursor is still up for grabs. Is it a signaling receptor, a transcription factor, a transporter of intracellular cargo? What role does it play in embryogenesis, neuronal development, apoptosis, tumorigenesis? Does it have pathological roles outside of being the source of Aβ? Progress is being made, albeit slowly. At this year’s Society for Neuroscience annual meeting, held last month in San Diego, California, an ancillary symposium explored the normal and pathological roles of APP. Titled “Function of APP Gene Family Members and Clues to AD Pathogenesis: Studies from Worms to Mammals,” and organized by Sanjay Pimplikar, Case Western Reserve University, Cleveland, Ohio, and Suzanne Guenette, Massachusetts General Hospital, Boston, the symposium was intended to provide an informal platform where researchers would be free to discuss, debate, and speculate on the normal and pathological roles of APP.

The first few talks dealt with the role of APP in lower organisms. Chris Li, City College of the City University of New York, kicked things off by outlining her work on the role of APL-1, a homolog of APP in the worm Caenorhabditis elegans. APL-1 is essential for survival in these worms. Without it, larvae cannot molt properly and never transition to their second stage. This point of death provides a platform to study individual domains of APL-1, said Li. Her group has made a series of APL-1 constructs and tested their ability to rescue lethality in APL-1 knockouts. With this approach they found that neither the transmembrane domain nor the cytoplasmic end of APL-1, which harbors a putative G protein binding domain, is essential for rescue. Rather, the extracellular part of the protein seems to be the business end. Expressing either the E1 or E2 extracellular domains was sufficient to rescue the APL-1 knockouts (see Hornsten et al., 2007). This finding is starkly reminiscent of recent work from Ulrike Muller and colleagues at Germany’s University of Heidelberg. These investigators showed that sAPPα, the α-secretase cleaved extracellular domain of APP is sufficient to rescue APP knockout phenotypes in mice (see below).

What is the function of the E1 and E2 APL-1 domains? This question is currently under investigation, said Li. What does appear certain is that the worms die because neuronal APP function fails, not for other reasons. That’s because only pan-neuronal expression of constructs led to rescue of APL-1 knockouts.

Sanjay Pimplikar described his studies on APP function in Danio rerio, better known to non-ichthyologists as the zebrafish. These tropical minnows have two APP homologs, APPa and APPb, both highly similar to human APP, especially at the C-terminal end. The human and fish AICD is 100 percent conserved, said Pimplikar, and in the complete cytoplasmic end of the protein there are only three conserved substitutions in human versus fish APP.

Pimplikar has used antisense oligonucleotides to knock down APP expression in fish embryos. He reported that silencing APPb alone was sufficient to induce severe morphological changes and to disrupt embryogenesis as early as 9-12 hours post-fertilization. The reason for these developmental problems is not known, but Pimplikar reported that human APP (hAPP) was capable of partially rescuing the APPb phenotype. Expression of hAPP reduced the number of deformed embryos and increased the number of normal embryos, though not to levels seen in wild-type zebrafish. Interestingly, hAPP with the Swedish mutation failed to rescue, which suggests that although the cytoplasmic end of hAPP is most similar to APPb, the extracellular domain is important for function, too. In this regard, Pimplikar suggested that APP processing and function may be intimately linked and inseparable.

The main advantages of working with zebrafish and worm models are the ease and speed of addressing specific questions. “The beauty of C. elegans is that we know exactly where all the neurons are and genetics experiments can be done so readily,” Guenette told ARF. But she also stressed that because of the greater biological complexity of mammals, it is important to replicate findings in mammalian models.

Guenette outlined some of her group’s work deciphering the role of APPs and their binding partners Fe65 and Fe65L1 in mouse brain development. Last year, Guenette reported that knocking out both Fe65 proteins gave a phenotype that resembled triple APP/APLP1/APLP2 knockouts (see Guenette et al., 2006). In the Fe65 double knockout (KO) mice, neurons appeared in the wrong places, particularly in the cortex. These so-called marginal zone heterotopias also appear in Muller’s APP triple KOs. Guenette found that in the Fe65 double KOs, levels and pattern of expression of APP were normal, as were levels of the intracellular domain AICD and APP C-terminal fragments. The data suggested that if Fe65 ensures proper brain development, then it does not do it by modulating APP levels or processing. But could Fe65 have some other effect on APP?

In San Diego, Guenette reported a subtle alteration in APP biology in Fe65 double KOs. She reported that treating wild-type cells with the glutamate analog NMDA elicits production of a high-molecular-weight form of APP that is substantially reduced in Fe65-negative cells. Guenette showed that this APP is not due to alternative splicing, but to post-translational modification by glycosylation. It is not yet clear if this high-molecular-weight form of APP is related to the developmental effects seen in Fe65 knockout mice. “Characterization of modified forms of APP can be pretty tricky,” said Guenette, but she noted that glycosylation can affect binding of Notch and its ligands. “The question for us is, Does glycosylation interfere with any signal that is essential for development?” she said. Guenette suggested that reduced glycosylation could be restricted to cells in a specific region of the brain, as that would explain why she does not see a difference in glycosylated APP in total brain lysates of the Fe65 knockout mice.

Tracy Young, from Dennis Selkoe’s lab at Harvard Medical School, also addressed the role of APP in development. Some of Young’s data was presented at a subsequent slide session. Young uses in utero electroporation of short hairpin RNAs to knock down APP to create mouse mosaics, with some cells expressing APP while others do not. This may be a crucially important point, because some of the phenotypes observed are not seen in APP knockout mice.

Young traced the APP-deficient cells by virtue of a green fluorescent protein (GFP) construct that is coexpressed with the shRNA. She showed how precursor cells that lack APP fail to migrate past the cortical plate. This does not seem to be a problem of motility, since the cells appear to migrate laterally. In fact, when Young focused on embryonic day-16 brains, 3 days after electroporation, she was able to see that GFP cells migrate out of the intermediate zone, before getting trapped at the cortical plate (ARF related conference story. Interestingly, the trapped cells begin to express the neuronal marker MAP2, suggesting that they begin to differentiate into neurons despite having their journey cut short.

By what mechanism does APP contribute to neuronal migration? Young has started to perform rescue experiments with various APP constructs to tease out factors that retard precursors at the cortical plate. She found that the migration defect does not appear when she electroporates in human APP constructs that are not recognized by the shRNA and are therefore expressed. Both the 751 and the 695 amino acid isoforms of APP rescue migration, as do APLP1, APLP2, and even APP with the Swedish mutation, Young reported. However, she found that the complete holoprotein is required; the extracellular or intracellular domains alone failed to rescue. Young also reported that rescue with full-length APP depended on an intact NPTY motif, found in the cytoplasmic tail. This motif binds a variety of proteins, including phosphotyrosine-binding proteins such as Dab1. In fact, Young reported that electroporation of Dab1 shRNAs lead to a similar phenotype as APP shRNA, with cells getting trapped at the cortical plate. Interestingly, Dab1 expression could partially rescue APP knockdown but not the other way around. Young suggested that this means that Dab1 works downstream of APP. It is worth noting that Dab1 has been shown to increase cell surface expression and processing of APP (see Parisiadou and Efthimiopoulos, 2006).

Both Hongmei Li from Tom Sudhof’s lab at the University of Texas Southwest Medical Center, Dallas, and Muller also addressed the role of various parts of the APP molecule. Li noted that although plenty of molecules have been discovered to bind to the cytoplasmic tail of APP, 80 percent of APP is in the extracellular space. Which is more important, she asked, the head or the tail of APP? To address this, Li attempted to rescue APP/APLP2 double KO mice.

Despite having what appears to be normal brain morphology and normal expression of synaptic protein, about 80 percent of these animals die early. Li tried to keep them alive with a flag-tagged construct that expresses only sAPPβ. Knocked in to the APP KO mice, this construct did not extend survival past postnatal day 21. It did bind to the molecular chaperone GRP78, otherwise known as BiP, however. Li said that presently it is unclear whether this binding is physiologically relevant or is merely related to misfolding of the flagged-APP protein. It is also unclear if the sAPPβ gets retained in the endoplasmic reticulum or is released from the cells.

For her part, Muller has had better success with rescue experiments using the extracytoplasmic end of APP. In San Diego, she summarized findings published last summer that sAPPα is sufficient to rescue APP KO mice (see Ring et al., 2007 and commentary).

Muller introduced a stop codon into the APP gene such that knock-in mice only expressed secreted sAPPα, the extracellular part of APP that is normally cleaved off by α-secretase. Muller found that this knock-in was sufficient to rescue the prominent phenotypes of APP-deficient mice, including their small body weight, weak grip, learning deficit in the Morris water maze, and loss of hippocampal synaptic plasticity. When crossed with APP/APLP2 double knockouts, the knock-in also survived longer. Muller noted that sAPPα levels are lowered in Alzheimer disease, which may partially explain learning and memory deficits in patients.

Overall, one important theme that emerged from the symposium was the role of the various parts of the APP molecule, said Guenette. In both C. elegans and in mice, the extracellular piece of APP, E1 or E2 in the case of worms or sAPPα in rodents, can rescue APP deficiency. In worms this seems particularly noteworthy, since these animals only have one APP homolog, APL-1, while in mice, APLP1 and APLP2 may partially compensate for APP loss. The ability of sAPPα to rescue at least some of the lethality seen in APP/APLP2 knockouts indicates the importance of this part of the molecule in mammals, as well. But the fact that the full-length molecule seems necessary to rescue developmental defects in mice suggests that different parts of the APP molecule may have different targets.—Tom Fagan.

This is Part 1 of a two-part story. See Part 2.