β-Amyloid precursor protein (APP) is a type 1 membrane glycoprotein found at the cell surface and in endosomal, endoplasmic reticulum, and Golgi membranes in many cell types 123. APP is expressed in the developing and mature brain at the somal cell surface, along growing axons and at the axonal growth cone. Cleavage by β- and then γ-secretase generates Aβ peptide, the apparent pathogenic agent in Alzheimer's disease (reviewed in 4). However, during brain development, a much larger fraction of APP is processed by α-secretases, which cleave within the Aβ sequence 5. This cleavage releases a soluble, amino-terminal APP fragment (referred to as APPs-α) and leaves a membrane-retained carboxy-terminal fragment (CTF) of 83 amino acids (C83). γ-Secretase can then cleave C83, yielding a free intracellular domain (AICD) and a small secreted peptide termed p3 6.

APP has two homologues in mouse, human and other mammals, APLP1 and APLP2 78, which also may be processed by α-, β, and γ-secretases in a manner similar to APP 910111213. Targeted deletions of these genes reveal functional redundancy among the three APP family members. Single genetic deletions in APP, APLP1 or APLP2 are each viable, whereas APLP2 deletion in combination with either APP or APLP1 deletion leads to early postnatal lethality 141516. The in vivo effects of disruption of these genes in mice have been difficult to pinpoint due to this functional redundancy. However, cells from these mice can be cultured and manipulated in vitro to examine the phenotypic effects of these deletions in detail.

Neuronal and glial co-cultures plated from the hippocampi of APP knock-out mice displayed shorter axonal outgrowth after 1 day in vitro (DIV) and longer axons after 3 DIV compared to neurons plated from wild-type brains 17. These results suggested that deletion of endogenous APP initially inhibits neurite outgrowth but over time stimulates neurite outgrowth, supporting the hypothesis that APP may first regulate adhesion to the substratum and then act to inhibit neurite outgrowth.

The secreted extracellular domain of APP, which is endogenously produced by α-secretase (APPs-α) or β-secretase (APPs-β), has also been shown to play functional roles in several assays. Although APPs-α can rescue some of the effects of APP deletion 18, studies of neurite outgrowth in culture have not directly tested whether APPs-α application can rescue the effects of APP deletion. Although neuronal and glial co-cultures from APP knock-out mouse hippocampi displayed longer axons after 3 DIV than neurons from wild-type brains 17, these authors found that co-culture of wild-type or APP knock-out neurons with APP knock-out astrocytes increased neurite outgrowth 17. The latter result taken by itself suggests that APPs secreted by astrocytes would normally inhibit neurite outgrowth. In contrast, other studies have shown that exogenous addition of APPs-α to cultured neurons actually promotes neurite outgrowth in a fashion similar to deleting APP in neural cultures. For example, APPs-α was found to increase neurite outgrowth in several assays, including in primary cortical cells 19, cortical explants 20, and PC12 cells 2122. Interpretation of these various results requires a more complex explanation for how full length APP and soluble APPs-α from neurons and glia interact functionally.

Several lines of evidence suggest that APP may interact functionally with integrins to mediate neurite outgrowth. Similar to APP, integrins have been shown to regulate neurite outgrowth 2324, and APP and integrins (α5, α1, and β1) co-localize in developing neurons in both growth cones and along axons 252627. However, no studies have addressed whether APP and integrin family members functionally or biochemically interact.

As reviewed above, both the absence of APP and the application of its major secreted derivative (APPs-α) can elongate neurites in culture. These seemingly paradoxical observations, which we confirm herein, led us to hypothesize that APPs-α may act to oppose effects of cell-surface APP and its homologues by competitively binding to factors that normally interact with the holoprotein to inhibit neurite outgrowth. Here, we show that deletion of each of the full-length APP family members acts in a cell-autonomous manner to promote neurite outgrowth in primary neuronal cells, and that adding the soluble derivative from each APP family member to wild-type cells similarly stimulates neurite outgrowth. Importantly, we show that APPs-α does not stimulate neurite outgrowth in the absence of APP. Our results support the hypothesis that secreted APPs-α can inhibit the activity of cell-surface APP. Further, we provide evidence that APP and APPs-α can each interact biochemically and functionally with integrins to mediate these effects.

In order to confirm and extend previous studies showing that APP plays a role in neurite outgrowth, we plated embryonic day 18 (E18) primary hippocampal or cortical neurons from APP knock out and wild-type littermate mice on CC2 coated two-chamber slides. After 3 DIV, cells were fixed and immunostained for βIII-tubulin, and the length of the longest neurite (defined here as the axon) was measured. Using these culture conditions, APP knock out cultures displayed longer axons than did neurons in wild-type cultures for both hippocampal (Figure 1a) and cortical (Figure 1f) neurons.

Next, we sought to determine the effects of APPs-α in this neurite outgrowth assay. Conditioned media (CM) from either Chinese hamster ovary (CHO) cells stably expressing human APPs-α or untransfected (control) CHO cells were added to wild-type E18 primary neurons for three days. APPs-α in the CM was present over the 3-day period after some degradation during the first 24 h (Figure 1c). In agreement with previous studies, the CM of the APP-overexpressing cells increased axon outgrowth over three days (Figure 1b). We found that the CM of CHO lines stably expressing APLP1 or APLP2 10 conferred similar axon-elongating activity (Figure 1b).

Besides APPs-α, whole CM from the CHO transfectants contains numerous additional proteins that could act in conjunction with or oppose the neurite-elongating activity of APPs-α. To assess whether APPs-α by itself can induce axonal outgrowth, His-tagged APPs-α was purified from baculovirus-transduced insect cells. Pure, recombinant APPs-α induced axonal outgrowth to a level similar to APPs-α from CM when similar concentrations of APPs-α were applied (Figures 1c–e). In contrast, purified APPs-α with a mutation of cysteine 117, which is required to form a disulfide bridge important for determining the secondary structure of the extracellular domain of APP 28, did not increase the length of neurites in this assay (Figure 1d, e).

Because both loss of APP and addition of APPs-α induced an increase in axonal outgrowth in this assay, we hypothesized that APPs-α may bind to factors that the extracellular domain of cell-surface APP also binds to, thereby acting to compete with cell-surface APP and inhibit its neurite-regulating activity. As an initial test of this hypothesis, we applied APPs-α to APP knock-out neurons to investigate whether APPs-α would still increase axonal outgrowth in the absence of cell surface APP. Although APPs-α again increased axonal outgrowth in wild-type neurons, it did not further increase outgrowth in neurons plated from APP knock-out littermates (Figure 1f). This result indicates that APP must be present for the neurite-enhancing effects of APPs-α to occur.

When APP knock-out neurons are plated and examined for effects on neurite outgrowth as above, the cultures lack both full-length APP and secreted APPs-α. To determine whether APP is required in a cell-autonomous manner for neurite regulation, we transfected APP small hairpin RNAs (shRNAs) into primary rat neurons at a low efficiency. We purposefully achieved less than 1% transfection efficiency in primary neurons using a modified lipofectamine protocol, in order to examine cell autonomous effects. Several APP shRNAs were designed to target rodent APP, and two shRNAs were selected that either efficiently knocked down APP (APP-active) or did not knock it down (APP-inactive) 29. E17 primary rat neurons were transfected with active or inactive APP shRNA plus green fluorescent protein (GFP) or else with GFP alone, and three days later, the cells were fixed and immunostained for βIII-tubulin. The transfected neurons were identified by their GFP fluorescence (Figure 2a, b), and their longest neurites were measured as before. Transfection of APP shRNA-active significantly increased the length of the longest neurite relative to transfection with either APP shRNA-inactive or GFP alone (Figure 2c), consistent with a cell-autonomous function of APP in regulating neurite length.

Next, we sought to determine whether the effects on neurites of altering APP expression that we observed in cultured neurons also occurred in vivo. To this end, control DNA (GFP alone) or APP shRNA-active plus GFP were electroporated in utero into E14 or E17 neuronal precursor cells lining the lateral ventricle of rat embryos. Data from our lab and others have shown that the rate of co-electroporation of two plasmids of equal molarity is between 90% and 95%, and that in vivo electroporation of APP shRNA and GFP results in an efficient knock down of APP in GFP-positive, electroporated cells 293031. The electroporated embryos were harvested at postnatal day 5, and their brains were fixed and sectioned coronally. With electroporation of GFP alone, neuronal precursors migrated into the cortical plate and differentiated, displaying radially aligned neurites, as expected (Figure 2d, f). Previously, we reported that co-electroporation of APP shRNA-active and GFP resulted in a strong decrease in APP expression and a corresponding defect in entry into the cortical plate for the large majority of developing neurons electroporated at E13 and harvested at E19, with greater than 80% of cells failing to migrate correctly 29. Here, we electroporated APP shRNA constructs at E14 and harvested at a postnatal time point (postnatal day 5) when migration and differentiation are complete. The APP-silenced cells continued to show a defect in cortical plate entry at this late time point (Figure 2h versus 2i), and the cells that were able to migrate into the cortical plate exhibited abnormal neurites, with radial and tangential processes showing excessive branching and often a beaded appearance in subsets of cells (Figure 2e, g).

The phenotype described above following knock down of APP in vivo occurred in only a subset of cells and was difficult to quantify accurately. Therefore, in order to quantify the qualitative phenotype we observed in vivo, regions of the cortex electroporated at E14 were dissected two days after electroporation (to allow time for APP knock down) and cultured and analyzed using the same neurite outgrowth assay described above. Similar to the in vitro Lipofectamine transfection experiments, plating of the in utero electroporated cells revealed APP knockdown in only a small subset of all of the neurons (less than 10%), and GFP fluorescence was used to identify these electroporated cells. In agreement with the Lipofectamine transfection experiments, cells electroporated in vivo with APP shRNA-active displayed longer axons than those electroporated with GFP alone (Figure 2j). In addition, these cells showed fewer primary neurites (Figure 2k), but there was no significant difference in axonal branching between control GFP electroporations and APP shRNA-active plus GFP co-electroporations (data not shown).

APP has two family members (APLP1 and APLP2) that appear to play redundant roles with APP for at least some of its functions and may compensate for a loss of APP in knock-out mice 141516. As shown in Figure 1b, APLP1-s and APLP2-s application each induced an increase in axonal outgrowth. In order to determine whether knock down of APLP1 or APLP2 resulted in a similar increase in axonal outgrowth, shRNAs were generated that efficiently knock down each of these family members (Figure 2l, m) and electroporated in utero and analyzed as described for APP. These analyses showed a small but significant increase in axonal outgrowth upon APLP1 knock down (an increase with APLP2 knock down failed to reach significance) and significant decreases in the number of primary neurites with knock down of APLP1 or APLP2, relative to control GFP electroporations (Figure 2j, k).

Using this in vivo electroporation paradigm, the analysis is limited to the specific subset of cells that are electroporated (those that undergo their terminal mitosis on or near E14, the day of electroporation). Therefore, the lack of a significant change in axonal length with APLP2 knock down may be due to high endogenous expression levels of APLP2 in the subset of cortical cells electroporated at E14. To test whether APLP2 expression could be observed in embryonic cortical cells, APLP2 expression was analyzed in E18 primary cortical cultures. In these cultures, APP and APLP2 were expressed in similar patterns in cell bodies, neurites and growth cones in a subset of cells (Figure 3a–d). To test if APLP2 contributes to axonal outgrowth in such cells, primary neurons were plated from E18 APLP2 knock-out cortices. This analysis showed that deletion of APLP2 inall cortical cells resulted in a significant increase in axonal outgrowth (Figure 3e).

Previous studies have shown that APP and integrins co-localize in primary neurons, specifically at points of substrate contact 252627. Similar to APP family members, integrins have been shown to regulate neurite outgrowth (for review, see 32. In order to determine whether APP and integrins physically interact, CHO cells were transiently transfected with FLAG-tagged murine Integrin β1 (Itgβ1-FLAG) and the 695 amino acid splice variant of APP. The lysates of these cells were subjected to immunoprecipitation for Itgβ1 with M2 anti-FLAG antibody, and the immunoprecipitates were probed for APP, revealing the co-immunoprecipitation of APP and Itgβ1 (Figure 4a). The carboxy-terminal domains of both APP and Itgβ1 contain NPXY motifs, which have been shown to bind to common interacting partners such as Fe65 and Dab1 (for review, see 3334. To determine if this domain is necessary for APP-Itgβ1 co-immunoprecipitation, APP with two point mutations in this motif (APP-NATA) and the cleaved extracellular domain of APP (APPs-α) were tested for their ability to co-immunoprecipitate with Itgβ1. Both APP-NATA and APPs-α were able to interact with Itgβ1 (Figure 4a). In contrast, APPs-α with a point mutation in cysteine 117 did not co-immunoprecipitate with Itgβ1. Further, another type I transmembrane domain receptor (angiotensin converting enzyme) that was also FLAG-tagged at its carboxyl terminus and expressed at a level similar to APP did not co-immunoprecipitated with APP (Figure 4b). Similar to APP, APLP1 and APLP2 also co-immunoprecipitated with Itgβ1 (Figure 4c). These immunoprecipitation experiments were all performed with overexpressed proteins. In order to confirm that APP interacts with Itgβ1 under endogenous conditions in neuronal cells, co-immunoprecipitation experiments were performed on these proteins from mouse and rat brain homogenates, as well as in rat primary neurons, and these all confirmed the interaction (Figure 4d, e).

To address whether APP and integrins interact functionally, E18 primary cortical neurons from wild-type or APP knock-out mice were plated and treated with an Itgβ1 blocking antibody (Ha2/5; BD Biosciences; San Jose, CA, USA) or an isotype-matched control antibody. Previous experiments have shown that blocking Itgβ1 function has been shown both to inhibit neurite outgrowth or to have no effect, based upon the assay utilized 243536. Our assay conditions did not reveal significant effects on neurite length of application of the Itgβ1 blocking antibody to wild-type neurons relative to the addition of the control antibody (Figure 5a, b). However, application of the Itgβ1 blocking antibody to APP knock-out neurons or else to APPs-α rich CM-treated cultures fully inhibited the axon-outgrowth promoting effect of each of these (Figure 5a, b).

Because APPs-α is able to biochemically interact with Itgβ1 in a fashion similar to full length APP (Figure 4a), we hypothesized that APPs-α may be able to compete with full-length APP for binding to Itgβ1. To address this possibility, APP with a carboxy-terminal FLAG tag (APP-FLAG) was co-transfected with Itgβ1 into CHO cells. After 48 h, CM from cells expressing APPs-α or from untransfected control cells was added to the co-transfected cultures for 8–12 h. Cultures were then lysed, and APP and Itgβ1 were co-immunoprecipitated with anti-FLAG agarose (Figure 5c). In a second approach, Itgβ1-FLAG was co-transfected with APP, and after 48 h, purified His-tagged APP was applied to the cultures for 8–12 h (Figure 5d). Quantification of these experiments using Licor imaging showed that in the presence of APPs-α, there was an approximately 70% decrease in the amount of APP-Itgβ1 co-immunoprecipitated relative to the amount of immunoprecipitated FLAG-tagged construct using both of these experimental approaches (Figure 5c, d). These results suggest that APPs-α can compete with full-length APP for binding to Itgβ1.

Studies from multiple labs have now established that APP plays a role in neurite outgrowth. However, the findings are sometimes inconsistent as to whether APP enhances or inhibits neurite outgrowth. It appears that APP may have differential effects on outgrowth and branching of neurites depending on the type of neuronal culture assayed (dissociated primary neurons, explants, or immortalized neuronal cell lines), the timing of analysis after plating (1 day versus 3 days versus 5 days), and the substrate upon which the neurons are plated (CC2, laminin, poly-D-lysine, fibronectin, and so on). In order to address the mechanism through which APP is acting to affect neurite outgrowth, we utilized a single primary neuronal culture system to directly compare the effects of genetic loss of APP and the application of APPs-α.

Loss of APP through either genomic deletion or shRNA-induced silencing resulted in significantly increased neurite elongation. Similarly, APPs-α application significantly stimulated outgrowth of the longest neurite in the same assay but, importantly, had no effect in cells that did not express APP. These results support our hypothesis that APPs-α acts to enhance neuronal process outgrowth by inhibiting the activity of cell surface APP, as diagrammed in Figure 5e. It should be noted that there are other possible explanations for the result; for example, the neuronal processes may not be competent to grow any longer beyond the length observed with APP deletion after 3 DIV. However, further support for the hypothesis derives from our experiments showing that APPs-α is able to block both the physical and functional interaction of full length APP and Itgβ1 (Figure 5). The model in Figure 5e incorporates these data to schematize the interactions we have shown to regulate neurite outgrowth in our assay. Under 'wild-type' conditions, APP inhibits the outgrowth-promoting activity of Itgβ1. In the absence of APP (that is, in APP knock out neurons or APP shRNA-transfected neurons), Itgβ1 enhances neurite outgrowth. Application of APPs-α to wild-type neurons can compete with cell surface APP for binding to Itgβ1 (as we demonstrate in Figure 5c, d), thus inhibiting the ability of APP to regulate Itgβ1-induced neurite outgrowth. In the absence of cell surface APP, APPs-α has no effect.

Like APP family members, integrins are single-pass, type 1 transmembrane proteins that mediate adhesion to extracellular matrix factors and regulate neurite outgrowth (reviewed in 37). Integrins exist as heterodimers, and previous studies have shown that APP co-localizes with integrin α 1β1 and α 5β1 heterodimers in growth cones 252638 and at point contacts but not focal adhesions in embryonic neurons 25. Based upon data presented here, the extracellular domain of APP is able to interact with Itgβ1, and APPs-α can compete with full length APP for binding to Itgβ1 at the cell surface. Both APP and Itgβ1 have NPXY motifs that interact with Fe65, a cytoplasmic adaptor protein that has the ability to bind two NPXY motifs simultaneously to form tripartite complexes, as it does with APP and LRP 3940. However, full length APP with mutations of its NPXY domain retains the ability to interact with Itgβ1. Thus, APP is likely to interact with Itgβ1 through both its extracellular domain and indirectly through its intracellular domain via interactions with adaptor proteins such as Fe65. In agreement with a role for Fe65 in this process, Fe65 has been found to co-localize with APP and Itgβ1 in growth cones 26, and it was recently shown that blocking the ability of Fe65 to bind to APP slowed axonal outgrowth 41. Our results showing a biochemical interaction between APP and Itgβ1 are in agreement with recent studies presented by the Rebeck lab 42.

The two other members of the mammalian APP family, APLP1 and APLP2, are also processed by α-, β-, and γ-secretases, releasing analogous cleavage products. The carboxy-terminal domains of all three family members are highly similar in amino acid sequence, but their extracellular domains are quite divergent. Surprisingly, the soluble cleavage products of APLP1 and APLP2 also stimulated neurite outgrowth in our in vitro assay. In addition, we observed that knock down or knock out of APLP1 and APLP2 in neurons resulted in an increase in neurite length, similar to the situation in APP knock-out neurons, suggesting that APLP1 and APLP2 at the cell surface can also regulate neurite outgrowth. Moreover, we found that full length APLP1 and APLP2 also interact with Itgβ1. Taken together, these results suggest that APP family members all play a role in regulating neurite outgrowth.

We recently showed that acute APP knock-down in embryonic rodent cortex dramatically retards the migration of cortical precursor cells into the cortical plate 29. In addition, we show here that APP knock-down in the cortex results in neurons that display abnormally formed and disorganized neurites. Each of these effects was due to a cell-autonomous loss of APP, as the total number of cells receiving APP shRNA was very low, and full length APP at the cell surface appears to mediate these effects. We hypothesize that the effects of loss of APP on neuronal precursor migration and neurite outgrowth are functionally linked through the ability of APP to differentially interact with factors at or near the cell surface (for example, Itgβ1), and that APPs-α may regulate each of these processes.

A recent study showed that knock-in of APPs-α to the APP genomic locus can rescue many of the defects observed in APP knock out animals, including reductions in brain and body weight, defects in long term potentiation, and certain behavioral defects in spatial learning, exploratory activity, and locomotor activities 18. These studies suggest that APPs-α can substitute for a deletion of APP to regulate these processes. However, we find that APPs-α does not rescue defects in neurite outgrowth observed upon deletion of APP. Rather, APPs-α application to wild-type neurons phenocopies APP deletion. Similarly, migration defects observed upon APP loss-of-function also are not rescued by APPs-α, and full length APP is required for cortical plate entry 29. Thus, both full length APP and its principal cleavage product APPs-α can apparently act as independent factors with multiple, sometimes distinct activities. Whereas APPs-α has evolved its own independent functions 18, our results indicate that it also has the ability to regulate the function of full length APP in some developmental processes.

APP: β-Amyloid precursor protein; CHO: Chinese hamster ovary; CM: conditioned media; DIV: days in vitro; E: embryonic day; GFP: green fluorescent protein; Itgβ1: Integrin β1; PBS: phosphate-buffered saline; shRNA: small hairpin RNAs.

The authors declare that they have no competing interests.

TLY—P and DJS conceived of the design of the study, TLY—P coordinated and contributed to the execution of all experiments performed in this study, analysis and interpretation of the data, and drafted the manuscript. AC generated purified APPs-α, RC generated neuronal cultures and aided in generation of plasmids used in the study, CM participated in the analysis of neurite outgrowth assays. In addition to study design, DJS participated in the interpretation of the data and aided in writing of the manuscript. All authors approved the final manuscript.