Shedding of the amyloid precursor protein-like protein APLP2 by disintegrin-metalloproteinases

Retinoic acid-induced upregulation of substrate and proteinase ADAM10 during neuronal cell differentiation


F. Fahrenholz, Institute of Biochemistry, Johannes Gutenberg-University, Becherweg 30, D-55128 Mainz, Germany


Cleavage of the amyloid precursor protein (APP) within the amyloid-beta (Aβ) sequence by the α-secretase prevents the formation of toxic Aβ peptides. It has been shown that the disintegrin-metalloproteinases ADAM10 and TACE (ADAM17) act as α-secretases and stimulate the generation of a soluble neuroprotective fragment of APP, APPsα. Here we demonstrate that the related APP-like protein 2 (APLP2), which has been shown to be essential for development and survival of mice, is also a substrate for both proteinases. Overexpression of either ADAM10 or TACE in HEK293 cells increased the release of neurotrophic soluble APLP2 severalfold. The strongest inhibition of APLP2 shedding in neuroblastoma cells was observed with an ADAM10-preferring inhibitor. Transgenic mice with neuron-specific overexpression of ADAM10 showed significantly increased levels of soluble APLP2 and its C-terminal fragments. To elucidate a possible regulatory mechanism of APLP2 shedding in the neuronal context, we examined retinoic acid-induced differentiation of neuroblastoma cells. Retinoic acid treatment of two neuroblastoma cell lines upregulated the expression of both APLP2 and ADAM10, thus leading to an increased release of soluble APLP2.


a disintegrin and metalloproteinase


catalytically inactive dominant negative mutant form of ADAM10


APP-like protein 1


APP-like protein 2


cleaved soluble APLP2


amyloid precursor protein


β-site APP-cleaving enzyme


chondroitin sulfate glycosaminoglycan


protein kinase C




poly(vinylidene difluoride)


retinoic acid


tumor necrosis factor-α converting enzyme

The amyloid precursor protein (APP) is a member of a protein family in mammals that includes the APP-like proteins APLP1 and APLP2 [1]. All APP/APLP family members are type I integral membrane proteins with large extracellular ectodomains and short cytoplasmic tails. Compared with APP, both APLPs are highly homologous in their amino acid sequence (e.g. APLP2/APP 52% identical, 71% similar) [2] and are proteolytically processed in a similar way. The N-terminal ectodomains are released by a shedding enzyme [2,3], whereas the C-termini remain in the membrane [2,4,5] and can be further processed to release a cytoplasmic fragment with signaling properties [4,6,7].

Further elucidation of APLP2-processing is of relevance with regard to the outstanding function of this protein, which was derived from knockout experiments. Whereas a double knockout of APP and APLP1 did not show severe phenotypic changes in mice, the combined knockout of APLP2 with both of the other APP family members resulted in postnatal lethality [8,9]. This shows that APLP2 and/or one of its proteolytic fragments are essential for normal development and survival, and may compensate the lack of either APP or APLP1. Whereas APP orthologs have been identified in lower and higher vertebrates, a recent publication revealed the existence of the first nonmammalian APLP2 in Xenopus laevis and its overall high percentage of conserved amino acids implies an important role for this member of the APP superfamily [10]. Although BACE [11–13] and the γ-secretase [6,22] have previously been identified as proteinases involved in the proteolytic processing of the APP relatives, it remains to be shown whether APLPs are also subject to cleavage by disintegrin-metalloproteinases (ADAMs) which act as α-secretases for APP [14–16].

Shedding of APLP2 can be induced by activation of protein kinase C (PKC) in human corneal epithelial cells [17]. Moreover, a decline in the membrane-anchored C-terminal fragments of APLP1 and APLP2 by the hydroxamic acid-based inhibitors batimastat and TAPI-2 was shown recently [11]. Using deletion mutants, metalloproteinase-dependent cleavage of APLPs was shown to occur at a similar distance to the membrane as is known for APP. Thus, an α-secretase-like activity seems to release the APLP2 ectodomain, but the proteinases involved are not yet identified.

Three members of the ADAM family have been shown to act as α-secretases [14,15,18]. We restricted our investigations on APLP2 shedding to ADAM10 and tumor necrosis factor-α converting enzyme (TACE, ADAM17), because purified ADAM9 failed to cleave a synthetic APP peptide at the major α-secretase cleavage site [19], and ADAM9 knockout mice exhibit unchanged APP processing [20]. ADAM10, in contrast, was recently shown to process APP in vivo and to prevent plaque formation in an Alzheimer's disease mouse model [16].

ADAM10 and TACE, which cleave APP, have been implicated in ectodomain shedding of other substrates such as cytokines [21], growth factors and their receptors [22,23], and adhesion molecules [24]. If ADAMs have several cellular substrates, how are physiologically relevant processing events coordinated? One possibility is a common up- or downregulation of substrate and sheddase during cell-fate decisions. Differentiation of neuronal cell types through retinoic acid (RA) leads to the upregulation of both APP [25] and APLP2 [26,27]. Therefore, we investigated the effect of RA on ADAM10 and TACE expression in neuroblastoma cell lines. In this study we provide evidence for a common upregulation of ADAM10 and its newly identified substrate APLP2 by RA-induced neuronal cell differentiation which resulted in an enhanced release of neurotrophic secreted APLP2 [28].


Phorbol-12-myristate-13-acetate-induced APLP2 ectodomain shedding

To study the effect of the PKC activator phorbol-12-myristate-13-acetate (PMA) on endogenous APLP2 shedding, we stimulated HEK293, SKNMC and SH-SY5Y cells with 1 µm PMA and performed Western blot analysis of proteins from cell supernatants. It has been shown that a large fraction of APLP2 and its secreted soluble derivative is modified by the addition of chondroitin sulfate glycosaminoglycan (CS-GAG) at a single site (Ser614) in the extracellular domain. This gives rise to the secretion of molecules with an apparent molecular mass between 130 and 170 kDa (Fig. 1). Two minor sharp bands between 95 and 120 kDa probably represent, according to earlier studies, APLP2s and truncated APLP2s without CS-GAG-modification (for post-translational modification of APLP2 see Slunt et al. [2] and Thinakaran and colleagues [29,30]). PMA treatment of all tested cell lines resulted in a significant increase in secreted endogenous APLP2 indicating that shedding of APLP2, like that of APP, is stimulated by PMA in neuronal and non-neuronal cell lines (Fig. 1).

Figure 1.

Enhancement of APLP2 secretion in HEK293, SKNnc and SH-SY5Y cells by the PKC activator PMA. PMA was added at a final concentration of 1 µm for 4.5 h, proteins in the cell supernatants were then precipitated and analyzed by western blotting using the antibody D2II. A representative example of three independent experiments is shown. Arrows indicate differentially modified APLP2s.

Inhibition of APLP2 ectodomain shedding by metalloproteinase inhibitors

It is known that the shedding of various transmembrane substrates is inhibited by hydroxamic acid-based inhibitors [31,23,32]. GM6001, a broad-spectrum hydroxamic acid-based inhibitor of matrixmetalloproteinases (MMPs) and ADAMs, decreased basal APPsα and APLP2s secretion to 60 and 75%, respectively, of untreated cells (Fig. 2A,B, lanes 1 and 3) and reduced the PMA-stimulated amount of both shed ectodomains to almost the level of control cells without inhibitor (Fig. 2A,B, lanes 2 and 4). This suggested participation of either ADAMs and/or MMPs in the processing of APLP2. There was more pronounced inhibition of constitutive shedding by the inhibitor GI254023X, which has a 100-fold higher potency to inhibit recombinant ADAM10 than recombinant TACE [33,34]. When compared with solvent-treated control cells, both APPsα and APLP2s were decreased to 30% (Fig. 2A,B, lanes 5 and 6), showing that ADAM10 is strongly involved in the shedding of APLP2.

Figure 2.

Influence of metalloproteinase inhibitors on APPsα and APLP2s secretion of neuroblastoma cells. Detection of shed (A) APLP2 and (B) APPsα in SKNMC cells treated with metalloproteinase inhibitors. GI254023X, GM6001 and the inactive GM6001NK were added at a final concentration of 10 µm for overnight preincubation, proteins of the cell supernatant were then collected for 4.5 h. PMA (1 µm) was added directly during the collection period. Secreted APLP2s and APPsα were detected as described in Experimental procedures (lane 1, control; lane 2, PMA; lane 3, GM6001; lane 4, PMA/G6001; lane 5, control; lane 6, GI254023X). Full-length APLP2 (APLP2Fl) was analyzed in cell lysates (A, lower) to confirm that the inhibitors did not alter steady-state levels of the protein. Representative blots are shown. Quantitative analysis of (C) APLP2 and (D) APPsα secretion. Quantification for APLP2 was carried out taking into account all three APLP2 protein forms. Values are the mean ± SD of three independent experiments. Control cells treated with the solvent or the inactive compound GM6001NK (indicated as GM –) were set to 100% (One-way ANOVA: *P < 0.05, **P < 0.01). (E) Detection of constitutive APLP2 shedding in SH-SY5Y cells. Cells were pretreated with increasing doses of GI254023X (0.3–10 µm) for 30 min. After 4 h treatment with freshly added inhibitor, the conditioned media were harvested and the amount of secreted APLP2 was determined. Data represent the mean ± SD of three independent experiments performed in duplicate. The inhibitor dose–response curve was generated using the software graphpad prism 4.02 (GraphPad Software Inc., San Diego, CA, USA).

With both inhibitors the amount of full-length APLP2 was comparable with control cells (Fig. 2A) and did not increase upon inhibited processing. Because α-secretase cleavage of APP occurs at the surface of neuronal cells [35], only a small fraction of the total cellular APP is cleaved, which generally does not result in a decrease in the full-length protein [36,37]. Therefore, reduction of APLP2 proteolysis by hydroxamic acid-based inhibitors might also affect only minor pools of the cellular protein resulting in an unchanged steady-state level.

To compare the cell-based inhibitory effect of GI254023X on APLP2 shedding with recently published data for shedding of other ADAM substrates like the interleukin-6 receptor [38], we applied the inhibitor in concentrations ranging from 0.3 to 10 µm to SH-SY5Y cells (Fig. 2E). The IC50 value for inhibition of APLP2 shedding by GI254023X was in the micromolar range (1.7 µm) showing a reduction of potency in cellular assays as compared to its effect on recombinant ADAM10 with IC50 values in the nanomolar range [33]. In comparison, inhibition of the interleukin-6 receptor shedding in COS cells occurred with a potency of 1.8 µm[38] and therefore was in the same range as found for cellular APLP2 shedding.

Inhibition of APLP2 ectodomain shedding by a specific β-secretase inhibitor

Another proteinase suggested to be an APLP2-cleaving enzyme is BACE-1 [11,13]. To elucidate whether, in cells of neuronal origin, APLP2 is processed by β-secretase, we tested the effect of the tripeptidic β-secretase inhibitor [(N-benzyloxycarbonyl-val-leu-leu-leucinal) Z-VLL-CHO] on APLP2 shedding in the human astroglioma cells U373. These cells overexpress human wild-type APP and therefore allow detection of BACE-1-generated secreted APPsβ, which is normally found at very low concentrations in the cell supernatant [41]. As shown in Fig. 3, both ectodomains were reduced significantly by applying the β-secretase-specific inhibitor. For APPsβ we found a decreased shedding of ≈ 50% of control cells. For APLP2s shedding was inhibited to a significant but lesser extent (reduction of ∼ 30% compared with control cells).

Figure 3.

Effect of the β-secretase inhibitor II on the ectodomain shedding of APLP2 in astroglioma cells overexpressing APP. (A) Western blots of secreted APPsβ and APLP2s upon β-secretase inhibitor treatment of U373hwtAPP cells. Following preincubation for 18 h with 25 µm of the β-secretase inhibitor II, shedding of APPsβ and APLP2s was analyzed in western blots with the antibodies 192 Wt or D2II. (B) Quantitation of APPsβ and APLP2s. The amount of shed proteins was quantified in three independent experiments. For secreted APLP2 all detectable protein bands above the 98 kDa marker band were taken into account (unpaired Student's t-test: *P < 0.05).

Because the antibody available against the APLP2 extracellular region (D2II) recognizes both the BACE-1- and α-like cleavage product of APLP2, APLP2s in cell supernatants reflect the effect of both shedding processes. Probably therefore the effects on the α-like cleavage of APLP2 by metalloproteinase inhibitors (Fig. 2) or on the β-like cleavage by a BACE-1 inhibitor (Fig. 3) are probably not as strong as for the processing of APP, which is monitored by specific antibodies (α-cleavage, 6E10, Fig. 2B; β-cleavage 192 Wt, Fig. 3A).

Enhancement of APLP2 secretion by overexpression of the α-secretases ADAM10 and TACE

To identify the proteinases that participate in APLP2 shedding, we examined cells overexpressing the α-secretase ADAM10 or TACE (Fig. 4A–C). Stable overexpression of either proteinase resulted in ≈ 2.5–3.5-fold more soluble APLP2 in the culture supernatant than in control cells (Fig. 4A,B). Because expression levels of the two proteinases differed (TACE being expressed at higher levels, Fig. 4C), we are not able to determine from the data which of the two enzymes preferentially cleaves APLP2. In all cases, overexpression did not significantly alter the steady-state levels of cellular APLP2 (data not shown), therefore the observed effects are not due to enhanced expression levels of APLP2.

Figure 4.

Influence of ADAM10, TACE and dominant negative ADAM10, overexpressed in HEK293 cells, on APLP2 shedding. (A) Immunoblot of secreted APLP2 with antibody D2II in ADAM10 and TACE overexpressing cells. (B) Quantification of APLP2s in ADAM10 and TACE overexpressing cells (mean ± SD of three experiments performed in duplicate, unpaired Student's t-test: **P < 0.01). As control, HEK cells transfected with the empty vector pcDNA3 were used and set to 100%. A representative example is shown. (C) Immunoblot of overexpressed ADAM10 and TACE. The overexpressed proteinases were detected in cell lysates by an anti-HA serum. (D) Immunoblot of secreted APLP2 in ADAM10DN overexpressing cells. A longer exposure time as in (A) was chosen to demonstrate the reduction of basal secretion of APLP2 by ADAM10DN. (E) Quantification of APLP2s in ADAM10DN overexpressing cells (mean ± SD of three experiments performed in duplicates, unpaired Student's t-test: **P < 0.01). (F) Immunoblot of overexpressed dominant negative ADAM10. The overexpressed mutated form of ADAM10 was detected in cell lysates by an antibody against the fused Flag-epitope.

Effect of a dominant negative mutant of ADAM10 on APLP2 shedding

To verify the APLP2-shedding activity of endogenous ADAM10, we used a cell line with stable overexpression of a dominant negative form of ADAM10 (Fig. 4F). This mutant protein carries the E384A point mutation in the zinc-binding region of ADAM10, which is known in Drosophila melanogaster[40] and in HEK293 cells [14] to suppress endogenous ADAM10 activity. HEK ADAM10DN cells showed a decreased APLP2 secretion of ≈ 60% compared with nontransfected HEK293 cells (Fig. 4D,E), whereas expression of full-length APLP2 was not significantly affected (data not shown). Thus, dominant negative ADAM10 inhibits the endogenous APLP2 sheddase activity.

Influence of overexpressed ADAM10 on the proteolytical processing of APLP2 in transgenic mice

Cleavage of APLP2 in vivo was demonstrated by western blots comparing brain homogenates from FVB/N mice and APLP2 knockout mice (Fig. 5A). In FVB/N mice (Wt) the antibody D2II against the N-terminal part of APLP2 detected a double band (Fig. 5A, lane 1). The CS-GAG-modified protein species were almost not detectable according to the low levels of this form in the brain as described for rat neuronal tissue [41]. By using antibody CT12, two C-terminal processing products of APLP2 were identified (C-stub I and II, Fig. 5A, lane 1). These stubs were also detected in HEK cells which had been treated for 20 h with the γ-secretase inhibitor DAPT before cell lysis (results not shown).

Figure 5.

Analysis of APLP2 proteolysis in transgenic mice overexpressing ADAM10. (A) APLP2 processing products in mice. The high specificity of the antibodies D2II and CT12 (recognizing either an N-terminal epitope or an epitope at the very end of the C-terminus) is demonstrated by comparing brain homogenates of wild-type with APLP2 knockout mice in western blots. (B) Detection of overexpressed ADAM10 in transgenic mice. ADAM10 transgenic mice were 10 weeks old. As controls we used nontransgenic littermates (Wt) of the same age. (C) Detection of APLP2s and the membrane-bound C-stubs. The amounts of shed APLP2 and the C-terminal stubs were quantified by Western blotting using membrane and soluble fractions derived from brain homogenates. (D) Quantitation of APLP2 processing products in transgenic mice. The values of shed APLP2 (APLP2s) and both C-stubs (C-stub I and C-stub II) were quantified for eight animals of each group in at least two independent western blot experiments and normalized to the full-length protein form (mean ± SD, unpaired Student's t-test: *P < 0.05, **P < 0.01).

To examine the α-like cleavage of APLP2 by ADAM10 in vivo, we investigated the influence of overexpressed ADAM10 in a transgenic mouse line. These mice overexpress bovine ADAM10 under the control of a neuron-specific Thy1 promoter [16]. Expression of the HA-tagged ADAM10 protein in brains of transgenic mice was verified by immunoblotting with the anti-HA serum Y-11. Both the immature and the mature forms of ADAM10 were detectable with a dominance of the catalytically active, mature form (Fig. 5B).

To analyze APLP2 processing, soluble and membrane-bound proteins from brain homogenates were subjected to immunoblotting using either the D2II or the CT12 antibody. We detected an enhanced amount of secreted APLP2 protein fragments (170%) by comparing ADAM10 transgenic mice with wild-type littermates (Fig. 5C,D). When we examined the amount of C-terminal stubs, we noticed a roughly twofold increase in both C-stubs in ADAM10 transgenic mice (Fig. 5C,D). No fragment corresponding to an APLP2 Cβ-stub could be detected by immunoblotting with the CT12 antibody in mouse brain homogenates, and therefore both identified C-stubs probably correspond to α-secretase-like cleavage products.

To exclude the possibility that the observed effects result from an altered expression intensity due to overexpression of the proteinase, we performed real-time RT-PCR experiments with mouse brain mRNA. APLP2-mRNA levels in transgenic and in control mice were not significantly different (P > 0.4; n = 5, data not shown).

Effect of RA on APP, APLP2 and ADAM10 expression in neuroblastoma cell lines

Because APP and APLP2 expression is enhanced during neuronal differentiation [26,27], we wanted to elucidate the effect of RA-induced differentiation of neuroblastoma cell lines on ADAM10 and TACE expression and on the release of secreted APLP2 and APPsα. For neuronal (N)-type SH-SY5Y cells, differentiation by RA was accompanied by the generation of long cellular outgrowths. Under the same conditions, the more Schwann-like SKNMC cells changed their morphology only slightly but revealed strongly decreased proliferative properties (Fig. 6A); for a characterization of both cell lines during differentiation see Voigt and Zintl [42].

Figure 6.

Morphological changes and mRNA levels in neuroblastoma cell lines upon RA-treatment. Cells were treated for 4 days with 1 µm RA. (A) Microscopic image of RA-differentiated neuroblastoma cell lines. Morphological changes as cellular outgrowths and loss of adherence were determined as markers of differentiation using light microscopy. (B) Real-time RT-PCR for mRNA quantitation. Changes in mRNA for APLP2 and the proteinases ADAM10 and BACE-1 were investigated using real-time RT-PCR. Experiments were performed three times in duplicate and amounts of mRNAs were normalized to GAPDH mRNA. Values are given as mean ± SD and results obtained with control cells were set to 100% (unpaired Student's t-test: *P < 0.05, **P < 0.01).

The effect of RA-induced differentiation on either the substrate APLP2 or the proteinase ADAM10 was analyzed using real-time RT-PCR for quantification of mRNAs. At the mRNA level, APLP2 was increased significantly in both RA-differentiated cell lines SH-SY5Y and SKNMC (Fig. 6B). Also, ADAM10 mRNA was strongly increased as we have recently shown for SH-SY5Y cells [45]. Interestingly, both mRNA species were induced more strongly in the N-type neuroblastoma cell line SH-SY5Y than in the more Schwann-like SKNMC cells. As APLP2 is also known to be processed by BACE (see above), we also quantified the mRNA of BACE-1 in SH-SY5Y cells. Although ADAM10 mRNA was induced to ≈ 250% compared with undifferentiated cells, we found only a slight, but significant increase in the amount of BACE-1 mRNA (Fig. 6B and 147% of control). At the protein level, the enhancement of APLP2 and ADAM10 was confirmed for both cell lines (Fig. 7A,B). Again in the N-type SH-SY5Y the increase in both, the APLP2 and the ADAM10 protein was stronger than in SKNMC cells, where significant increase occurred only in the immature form (Fig. 7B).

Figure 7.

Expression of APLP2 (A) and ADAM10 (B) in differentiated neuroblastoma cell lines. Cell lysates of RA-differentiated SH-SY5Y and SKNMC cells were subjected to 7.5% SDS/PAGE, and the proteins were detected by immunoblotting using primary antibodies against the C-termini. Experiments were performed three times in duplicate, representative immunoblots are shown. Values are given as mean ± SD and results obtained with control cells were set to 100% (unpaired Student's t-test: *P < 0.05, **P < 0.01).

In contrast to ADAM10 expression, we could not detect increased TACE protein levels upon RA treatment in our experiments (Fig. 8). Although TACE remained unchanged in SH-SY5Y cells, the amount of the pro- and the mature form of this proteinase even decreased in SKNMC cells, revealing reduced stability compared with ADAM10. Therefore, the concerted upregulation of APLP2 and its sheddase during RA-induced neuronal differentiation appears to be specific for ADAM10.

Figure 8.

Expression of TACE in differentiated neuroblastoma cell lines. SH-SY5Y and SKNMC cells were differentiated with RA for 4 days, and the mature and the immature form of the proteinase were detected by immunoblotting. Values are given as mean ± SD of three independent experiments, and results obtained with control cells were set to 100% (unpaired Student's t-test: *P < 0.05, **P < 0.01).

In both neuroblastoma cell lines we found, upon RA treatment, an increase of APLP2 shedding. Soluble APLP2 in supernatants of differentiated cells was enhanced to 150% for SH-SY5Y and 180% for SKNMC compared with undifferentiated control cells (Fig. 9A). Also, in SH-SY5Y cells the secretion of APPsα was found to be enhanced significantly to > 200% of control cells due to increased expression of the α-secretase ADAM10. This phenomenon was also seen in SKNMC cells although to a lesser extent (Fig. 9B).

Figure 9.

Proteolytical processing of APP and APLP2 in RA-treated neuroblastoma cells. Western blots and quantification of (A) APLP2s and of (B) APPsα in RA-differentiated neuroblastoma cell lines. Cells were treated as described in Experimental procedures. Precipitated proteins of cell supernatants were subjected to 7.5% SDS/PAGE and immunoblotted. Detection was performed with the antibodies 6E10 and D2II. Values of the quantitative analysis are mean ± SD and significances were determined using paired Student's t-test (*< 0.05, **< 0.01). Experiments were performed three times in duplicate, representative immunoblots are shown.


We report the cleavage of the mammalian APP-related protein APLP2 by the disintegrin and metalloproteinases ADAM10 and TACE (ADAM17), and a common upregulation of ADAM10 and its substrate by RA.

The main criteria for the involvement of ADAMs, the enhancement of APLP2 shedding by phorbolesters and decreasing amounts of APLP2s by hydroxamic acid derivatives, were fulfilled. Overexpression of ADAM10 as well as of TACE resulted in increased secretion of APLP2s from cultured cells. Also, a dominant negative form of ADAM10 reduced the shedding of APLP2.

Because the ADAM10-preferring inhibitor GI254023X displayed the most pronounced effect by reducing APLP2s to ≈ 30% of control cells, we conclude that ADAM10, as shown for APP [14], plays an important role in the secretion of the APLP2 ectodomain. We were also able to demonstrate the influence of the α-secretase ADAM10 on APLP2 processing in vivo. Transgenic mice with neuronal overexpression of ADAM10 showed significantly increased amounts of shed APLP2 as well as C-terminal processing products.

Because soluble APLP2 was shown to induce neurogenesis in the subventricular zone of adult mouse brain [44] and enhances neurite outgrowth [28], the proteolytical processes that generate APLP2s may be important for the generation and survival of neuronal cells. The elevation of APP and APLP1 and APLP2 in differentiated SH-SY5Y [27] suggests an important function for the expression and proteolysis of APP family members especially in neuronal cell populations. In support of this hypothesis, we found enhanced secretion of the extracellular domains of APP and APLP2 upon RA treatment, which might correspond to increased expression of ADAM10 in both SH-SY5Y and SKNMC cell lines. We cannot completely rule out the possibility that the increase in secretion of soluble APLP2 following treatment with RA may also be due to the increase in the amount of APLP2 and not because of the increase in ADAM10 expression. But because the BACE-1 mRNA level was increased to a lesser extent, a major role of the nonamyloidogenic pathway and ADAM10 in differentiating neuronal cells may be supposed. Recent findings [43] demonstrate a conserved binding site for retinoid receptors in the promoter sequence of ADAM10 and an increase of promoter activity by RA. These results suggest a RA-induced regulation of this disintegrin-metalloproteinase by nuclear receptors. Because TACE was not positively affected by RA, but even degraded in SKNMC cells, we demonstrate again a higher stability of ADAM10 compared with TACE, which was also selectively degraded after PMA treatment of cultured cells [45].

In late-onset Alzheimer's disease there is genetic, metabolic and dietary evidence for defective retinoid transport and function [46–48]. In accordance with these findings, is the observation that the impairment of long-term potentiation induced by experimental vitamin A deficiency in adult mice can be reversed by direct application of RA to hippocampal slices [49]. Recently, we demonstrated that overexpression of ADAM10 in APP[V717I] transgenic mice prevented plaque formation and rescued the impairments of hippocampal long-term potentiation, thus suggesting a beneficial role of the α-secretase ADAM10 in memory and learning [16]. Because ADAM10 together with its substrates is upregulated via RA our results suggest that bioactive retinoids in the hippocampus could lead to an increased α-secretase activity and to an increased release of the neurotrophic-soluble ectodomains of APP and APLP2. Further studies are necessary to support this conclusion in vivo and to delineate the regulatory mechanism of RA-induced α-like cleavage of APLP2.

Experimental procedures


PMA and all-trans-RA were purchased from Sigma (St. Louis, MO, USA), the broad-spectrum inhibitor GM6001 (Galardin) and the corresponding inactive control compound (GM6001NK), as well as the β-secretase inhibitor II, were from Calbiochem (San Diego, CA, USA). Each was dissolved as stock in dimethylsufoxide and kept at −20 °C.

Primary antibodies

The following antibodies were used for western blot analysis: D2II, a rabbit polyclonal antibody against the N-terminus of APLP2; CT12, a rabbit polyclonal antibody against the C-terminus of APLP2 (both kindly provided by G. Thinakaran, University of Chicago, IL); 6E10 (Signet Laboratories, Dedham, MA, USA) against APPsα; 192 Wt (S. Sinha, Elan Pharmaceuticals, San Francisco, CA, USA) against APP residues 591–596, detecting only β-secretase-cleaved soluble APP (APPsβ) antibodies against the C-termini of human ADAM10 and 17 (Chemicon, Temecula, CA, USA). Overexpressed proteinases were detected with the anti-HA serum Y-11 (Santa Cruz Biotechnology, Santa Cruz, CA, USA) or anti-Flag serum M2 (Stratagene, La Jolla, CA, USA).

Constructs and mutagenesis

The cDNAs of murine TACE [50] and bovine ADAM10 [14] were fused with a DNA-sequence coding for a hemagglutinin epitope (YPYDVDDYA), and dominant negative ADAM10 was tagged with a Flag-epitope (DYKDDDDK) as described previously [14]. Expression of the tagged proteinases was performed by using the vector pcDNA3 (Invitrogen, Carlsbad, CA, USA).

Cell culture and transfections

HEK293 cells stably overexpressing either HA-tagged ADAM10, Flag-tagged dominant negative ADAM10 or HA-tagged TACE (named HEK ADAM10, HEK ADAM10DN and HEK TACE, respectively) were cultured in Dulbecco's modified Eagle's medium (DMEM; containing 10% fetal calf serum, 2 mm glutamine, 100 U·mL−1 penicillin, 100 µg·mL−1 streptomycin). SKNMC cells were cultured in DMEM complete medium supplemented with 1% sodium pyruvate, and SH-SY5Y cells were cultivated in Ham's F12 medium [containing 10% (v/v) fetal bovine serum, 2 mm glutamine, 100 U·mL−1 penicillin and 100 µg·mL−1 streptomycin]. For the astroglioma cell line U373 MEM supplemented with 10% (v/v) fetal bovine serum, 2 mm glutamine, 100 U·mL−1 penicillin, 100 µg·mL−1 streptomycin, 1% (w/v) sodium pyruvate and 1% (w/v) nonessential amino acids was used.

Stable transfections of HEK293 cells were performed by using the calcium phosphate precipitation method followed by selection of transfected cells with G418 (1 mg·mL−1). For differentiation of the neuroblastoma cell lines, cells were seeded on 10 cm culture plates after adjusting the cell number (SH-SY5Y 2.5–5 × 105, SKNMC 0.5–1.0 × 105 cells) and grown for 72 h. The medium was replaced by fresh phenol red-free medium containing 1 µm RA, the cells were incubated for 4 days, and the RA-containing medium was changed daily.

Western blot analysis of TACE and ADAM10

Cell pellets were washed with NaCl/Pi and dissolved in Laemmli buffer containing 100 mm dithiothreitol, heated to 95 °C for 10 min, separated by SDS/PAGE on 7.5% gels and transferred to poly(vinylidene difluoride) (PVDF) membranes. Bound antibodies against the endogenous or overexpressed proteinases were visualized by applying alkaline phosphatase coupled antibodies and the chemiluminescence substrate CDPstar (Tropix, Foster City, CA, USA). Emitted light was detected by using a digital camera and quantified with the software aida 3.50 (Raytest, Straubenhardt, Germany).

Western blot analysis of APP, APLP2 and their processing products

Cells were grown close to confluency, washed with serum-free culture medium and incubated for 4.5 h in serum-free culture medium containing 2 mm glutamine, 100 U·mL−1 penicillin, 100 mg·mL−1 streptomycin, 10 µg·mL−1 fatty acid-free bovine serum albumin and activators or inhibitors as indicated. PMA (1 µm) was added directly to the serum-free harvesting medium (with 2 mm glutamine, 100 U·mL−1 penicillin, 100 µg·mL−1 streptomycin and 10 µg·mL−1 fatty acid-free bovine serum albumin) for 4.5 h. The inhibitors GM6001, its negative control and GI254023X (10 µm) were added to the cells 18 h prior harvesting and also to the harvesting medium. For the dose–response curve of GI254023X SH-SY5Y cells were preincubated for 30 min with varying amounts of the inhibitor followed by a harvesting period of 4 h with freshly added inhibitor. Proteins of the culture medium were precipitated with 10% trichloroacetic acid and collected by centrifugation. The pellets were washed twice with ice-cold acetone, dried and dissolved in Laemmli buffer containing 100 mm dithiothreitol and heated to 95 °C for 10 min. Aliquots corresponding to equivalent protein contents of cells were separated by SDS/PAGE on 7.5% gels and blotted onto PVDF membranes. Soluble APLP2 was detected with antibody D2II (1 : 2500), followed by incubation with anti-rabbit serum either coupled to alkaline phosphatase (Tropix) or 35S labeled (Amersham Biosciences, Arlington Heights, IL, USA). Shed APPsα and APPsβ were detected by using the antibodies 6E10 and 192 Wt, respectively, in combination with secondary antibodies either coupled to alkaline phosphatase or 35S-labeled. Bound antibodies were visualized by using a digital camera or the BAS Reader (Fujifilm, Düsseldorf, Germany), and quantified as described above. For detection of full-length APLP2 and its membrane-bound C-stubs, cells were centrifuged for 3 min, 960 g, 4 °C. An aliquot of the cells was taken for quantification of the protein content. The residual cells were dissolved in an adequate volume of 1.5 × Nu-PAGE buffer (Invitrogen) containing 100 mm dithiothreitol, heated to 70 °C for 10 min, separated on 4–12% Nu-PAGE gels (Invitrogen) and transferred to PVDF membranes. As primary antibody we used CT12. Detection of APLP2 protein fragments was performed as described above for the soluble proteins.

Preparation of mouse brain homogenates from transgenic mice

The generation of transgenic mice with neuron-specific overexpression of bovine ADAM10 has been described previously [16]. Transgenity of mice was confirmed by PCR and by detection of the overexpressed HA-tagged ADAM10 proteins by western blotting. Mice were chosen for the experiments with a 1.3-fold increase in the amount of ADAM10 compared with their wild-type litter-mates.

Brains of 10-week-old mice (ADAM10 or wild-type nontransgenic littermates) were dissected and homogenized in 200 mm Tris/HCl (pH 8.4) in the presence of proteinase inhibitors (complete mini, Roche, Mannheim, Germany). Homogenates were centrifuged at 135 000 g for 1.75 h at 4 °C for sedimentation of cellular membranes. The supernatants containing the soluble proteins were removed and the membrane pellet was suspended in NaCl/Tris. The protein concentrations of both fractions were determined. Proteins were separated on polyacrylamide gels and blotted onto PVDF membrane as described above. As secondary antibody we used 35S-labeled secondary antibodies. For quantification the BAS Reader (Fujifilm) and the software aida 3.50 were used.

Real-time RT-PCR

Total RNA was isolated using the RNeasy Kit (Qiagen, Hilden, Germany). RNA concentration and quality was determined by spectrophotometry. Aliquots of the RNAs were dissolved in RNAse-free water (Sigma) to a concentration of 50 ng·µL−1. Real-time RT-PCR primers were designed for human GAPDH, ADAM10, BACE and APLP2 from Gene bank mRNA (cDNA) sequences utilizing the primer express 1.5 software (Applied Biosystems, Foster City, CA, USA).


Real-time RT-PCR was performed using the one-step QuantiTectSYBRGreen RT-PCR-Kit (Qiagen), the ABIPrism 7000 (Applied Biosystems), 250 ng RNA and the specific primer pairs (0.5 µm of each primer). Reverse transcription was performed at 50 °C for 30 min. The quantitative PCR was induced by heating to 95 °C, followed by 45 PCR cycles (one cycle contained the following steps: 15 s at 95 °C; 30 s at 55 °C; 30 s at 72 °C). The specificity of each primer pair was confirmed by melting curve analysis and agarose gel electrophoresis. The quantity of mRNA was calculated using either the ΔΔCt method, when PCR efficiency was close to 100%, or a standard curve (e.g. for BACE). The mRNA of the housekeeping gene GAPDH was unchanged under differentiation conditions, and all other mRNAs were normalized to it.


We thank A. Roth for excellent technical assistance; R. Black for the murine TACE cDNA; C. Prinzen for introduction of the HA-tag into the TACE cDNA, and G. Thinakaran for providing the APLP2 cDNA and the antibodies CT12 and D2II. We are grateful to Dr I. Hussain, Glaxo SmithKline (Harlow, UK) for putting the inhibitor GI254023X at our disposal. This work was supported by the DFG priority program 1085/3-Cellular mechanisms of Alzheimer's disease.