Journal of Molecular Neuroscience

, Volume 39, Issue 1, pp 1–8

Neuroprotective Protein and Carboxypeptidase E


  • Hisatsugu Koshimizu
    • Section on Cellular Neurobiology, Program on Developmental Neuroscience, Eunice Kennedy Shriver National Institute of Child Health and Human DevelopmentNational Institutes of Health
  • Vladimir Senatorov
    • Section on Cellular Neurobiology, Program on Developmental Neuroscience, Eunice Kennedy Shriver National Institute of Child Health and Human DevelopmentNational Institutes of Health
    • Section on Cellular Neurobiology, Program on Developmental Neuroscience, Eunice Kennedy Shriver National Institute of Child Health and Human DevelopmentNational Institutes of Health
    • Elton Laboratory for Molecular Neuoendocrinology, Department of Human Molecular Genetics and Biochemistry, Sackler Faculty of MedicineTel Aviv University

DOI: 10.1007/s12031-008-9164-5

Cite this article as:
Koshimizu, H., Senatorov, V., Loh, Y.P. et al. J Mol Neurosci (2009) 39: 1. doi:10.1007/s12031-008-9164-5


This review outlines the neuroprotective activities and structural specificities of two distinct proteins, activity-dependent neuroprotective protein, a protein assigned transcription factor/chromatin remodeling activity, and carboxypeptidase E, a classic exopeptidase. Future studies will elucidate how these two versatile proteins converge onto a similar endpoint: neuroprotection.

Role of Carboxypeptidase E in Neuroprotection

Carboxypeptidase E (CPE) is classically known as an exopeptidase that cleaves carboxy-terminal basic amino acids from neuropeptide and peptide hormone intermediates resulting in the production of bioactive peptides in neuroendocrine cells. However, a body of work has indicated that CPE expressed in the nervous and endocrine systems plays multiple non-enzymatic roles in addition to being an exopeptidase (Fricker and Snyder 1983). The soluble form of CPE acts as a processing enzyme (Hook 1984). The membrane-bound form of CPE serves as a sorting receptor for some proneuropeptides and pro-brain-derived neurotrophic factor (proBDNF) to target them into the regulated secretory pathway (Cool et al. 1997; Lou et al. 2005). Additionally, the cytoplasmic tail of the transmembrane form of CPE was shown to be a key molecule in the anchoring of adrenocorticotropic hormone and BDNF vesicles to the microtubule-based transport system for post-Golgi delivery of the vesicles to the release site (Park et al. 2008a, b). Thus, it would be expected that lack of CPE would have a major impact on the function of the endocrine and nervous systems. Indeed, CPE knock-out (KO) mice exhibit not only endocrinological deficits such as diabetes, infertility, and obesity (Cawley et al. 2004) but also neurological deficits including poor muscle tone (Cawley et al. 2004) and diminished glutamate-transmission-mediated b-wave in their retinograms (Zhu et al. 2005). Lack of CPE also resulted in diminished activity-dependent secretion of BDNF from cortical neurons of CPE-KO mice (Lou et al. 2005). In a Caenorhabditis elegans mutant lacking CPE, a defect in acetylcholine neurotransmission at the neuromuscular junction was observed (Jacob and Kaplan 2003).

Recently, we found that absence of CPE in CPE-KO mice had a profound effect on memory, synaptic physiology, and the cytoarchitecture of the hippocampus (Woronowicz et al. 2008). Moreover, the adult CPE-KO mice displayed deficits in memory consolidation as revealed by the Morris water maze, object preference, and social transmission of food preference tests. These mice also showed no evoked long-term potentiation. Additionally, CPE-KO mice at 4 weeks of age and older, but not at 3 weeks of age, exhibited marked degeneration, specifically of the pyramidal neurons in the hippocampal CA3 region which normally expresses high levels of CPE. Immunohistochemistry revealed that the neuronal marker, NeuN, was reduced, suggesting that neurodegeneration took place in the CA3 region. The astrocyte marker, glial fibrillary acidic protein (GFAP), was increased in the CA3 area of CPE-KO mice, characteristic of gliosis. Given that the population of neurons which are generated by adult neurogenesis is ∼10% in hippocampus (Imayoshi et al. 2008), the degeneration of CA3 neurons in CPE-KO mice would be caused mainly by neuronal cell death rather than a decrease in adult neurogenesis. Additionally, immunoreactivity for calbindin, a marker molecule for mature neurons, was dramatically decreased in adult CPE−/− CA1 pyramidal neurons. These defects are similar to the neurodegeneration in the CA3 region and to the major loss of calbindin in CA1 neurons observed in rodent models for hippocampal injury, hyperexcitability, and temporal lobe epilepsy induced by kainic acid treatment (Shetty and Hattiangady 2007). Hence, the possibility that neurodegeneration in CPE−/− CA3 region and loss of immunoreactive calbindin in CA1 is due to one of these hippocampal insults cannot be ruled out. Furthermore, calbindin staining showed early termination of the mossy fibers before reaching the CA1 region in the hippocampus of CPE−/− mice. Ex vivo studies revealed that apoptosis of primary cultured rat hippocampal neurons induced by hydrogen peroxide, a reactive oxygen species (ROS), was dramatically suppressed when CPE was overexpressed in these cells (Woronowicz et al. 2008). These findings indicate that CPE is essential for survival of CA3 neurons and maintenance of hippocampal mossy fiber morphology as well as function, including consolidation of memory. Additional evidence in support of this neuroprotective role of CPE in the hippocampus comes from the observation that, after transient global ischemia, there was a greater and more sustained increase in CPE expression in the CA3 region of the hippocampus which correlated with survival of those neurons. The observation was in contrast to neurons of the CA1 region that only showed a transient increase of CPE expression and were more vulnerable (Jin et al. 2001)

To determine whether CPE has a neuroprotective role on other central nervous system (CNS) neurons, we investigated the effect of applied low potassium (LK) on cultured cerebellar granule neurons (CGN), which is a well-used in vitro model system of neuronal apoptosis (Yamagishi et al. 2001). We found that the apoptosis induced by LK in CPE+/− CGN culture was significantly increased in comparison with CPE+/+-cultured CGN, demonstrating that CPE plays a neuroprotective role in CGN as well (Fig. 1).
Fig. 1

CPE inhibits LK-induced apoptosis in cultured cerebellar granule neurons. Primary cerebellar granule neuron culture was prepared, and cell death induction was performed according to Hatanaka’s method using postnatal day 6 CPE+/+ and CPE+/− mice (Yamagishi et al. 2001). Briefly, cells were gently dissociated using a plastic pipette after digestion with papain at 37°C and seeded into chamber slides coated with polylysine. The cultures were then incubated with 26 mM potassium (High K+, HK)containing 10% fetal bovine serum (FBS) and 90% Dulbecco’s modified Eagle’s medium (DMEM) for 5 days. Apoptosis was induced by switching the medium into serum-free 5 mM potassium (Low K+, LK)-containing DMEM. One day after the cell death induction, the CGN cultures were fixed in 4% paraformaldehyde (PFA), and apoptosis was quantified by diamidino-2-phenyl-indole (DAPI) staining and terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay (Promega). One millimolar DAPI (Sigma-Aldrich) was used for DAPI staining. There was no significant difference in cell viability between CPE+/+ and CPE+/− culture before the cell death induction (data not shown). a LK-induced apoptosis was detected by DAPI staining. Arrows indicate the pattern of DAPI staining of nuclei which exhibited apoptosis. Bar represents 10 μm. b Number of cells exhibiting apoptosis was significantly increased in the CPE+/− culture. More than 600 cells from two to three animals were counted for each genotype. *P < 0.05 (Student’s t test). c LK-induced apoptosis was detected by TUNEL assay. Arrows indicate cells which exhibited apoptosis. Bar represents 10 μm. d Number of TUNEL-positive cells was significantly increased in CPE+/− culture. More than 800 cells from two to three animals were counted for each genotype. *P < 0.05 (Student’s t test)

Consistent with the findings in the CNS neurons, a recent study showed that CPE also plays a major role in endoplasmic reticulum (ER) stress and apoptosis in pancreatic islet beta-cells. Palmitate treatment of pancreatic islet beta-cells resulted in a significant decrease in the CPE protein level. The CPE was translocated to the Golgi and degraded in lysosomes. ER stress and apoptosis were up-regulated when CPE expression was suppressed, while overexpression of CPE inhibited ER stress and apoptosis (Jeffrey et al. 2008). Thus, an anti-apoptotic role of CPE seems not to be limited to neurons.

The mechanism by which CPE protect CNS neurons from cell death is currently unknown. One possibility is that CPE is required for the sorting and processing of proteins and peptides that promote cell survival such as activity-dependent neuroprotective protein (ADNP; Gozes 2007), humanin, a peptide which suppresses cell death induced by Alzheimer’s disease-related insults in vitro (Niikura et al. 2006) and BDNF which may be attenuated in some neurodegenerative diseases, such as Alzheimer’s disease, Parkinson’s disease, and Huntington’s disease (Chao et al. 2006; Zuccato and Cattaneo 2007).

To address whether the CPE exopeptidase activity is required for its neuroprotective function, we infected rat hippocampal neuronal cultures with adenovirus carrying wild type-CPE or CPE-E300Q, an enzymatically inactive form of CPE. The cultures were then challenged with hydrogen peroxide as a ROS. Overexpression of CPEE300Q significantly prevented the ROS-induced apoptosis, similar to that of wild-type CPE, suggesting that CPE neuroprotective action on hippocampal neurons is independent of its exopeptidase activity (Fig. 2). Suffice to say, this E300Q mutant CPE could still function as a sorting receptor for the regulated secretory pathway: for example, proBDNF which is sorted by CPE into secretory granules for processing and regulated secretion of mature BDNF from cerebral cortical neurons (Lou et al. 2005; Park et al. 2008a, b). BDNF has been implicated in the promotion of survival of cultured CNS neurons, although its effect is weaker than insulin-like growth factor (Yamada et al. 2001). However, hippocampal development is normal, and no neuronal degeneration is observed in the hippocampus of BDNF−/− and BDNF+/− adult mice (Korte et al. 1995). Taken together, CPE’s role in sorting of BDNF could provide contributory function but not likely to be a major factor in the neuroprotective effect of CPE on hippocampus.
Fig. 2

Neuroprotection of ROS-induced apoptosis of hippocampal neurons by CPE is independent of its enzymatic activity. Primary E18 rat hippocampal neurons (NeuroPure, Genlantis, San Diego, CA) were dissociated and seeded into chamber slides. The culture was then incubated with 10% FBS and 90% Dulbecco’s modified Eagle’s medium. After 4–5 days, the culture medium was changed to Neurobasal medium (Gibco) supplied with 2% B27 supplement (Gibco) and 10 mM cytosin-arabinosid. After 5–6 days incubation, the culture was transfected with adenovirus encoding LacZ, CPE wild type or CPE E300Q in MOI 100 for 2 days. More than 80% cells were MAP-2 positive (data not shown). Cell death was induced with 50 μM hydrogen peroxide for 1 day, and the hippocampal cultures were then fixed in 4% PFA and stained with 1 μM DAPI. Fluorescent images were captured with a confocal microscope (LSM 510, Carl Zeiss). a Arrows indicate the pattern of DAPI staining of nuclei which executed apoptosis. Bar represents 5 μm. b Relative amount of ROS-induced cell death was quantified according to the formula as follows ([relative ROS-induced cell death (%)] = 100 × {[apoptotic nuclei H2O2+]/[total nuclei H2O2+] − [apoptotic nuclei H2O2]/[total nuclei H2O2]}/{[apoptotic nuclei H2O2+]con./[total nuclei H2O2+]con. − [apoptotic nuclei H2O2]con./[total nuclei H2O2]con.}). More than 900 cells were counted for each group. There was no significant difference in cell death in ROS-untreated cultures (basal cell death, data not shown). ROS-induced cell death in CPE wild type (CPE-WT)- and CPE E300Q-expressing culture is significantly inhibited in comparison with that of Lac Z-expressing control culture (Con.). *P < 0.05 (analysis of variance)

CPE may function in a manner yet to be discovered, such as preventing ER stress, as suggested in pancreatic beta cells (Jeffrey et al. 2008). Alternatively, CPE might transmit a signal for neuroprotection by an unknown mechanism. The catalytic site of carboxypeptidases is highly conserved (Aloy et al. 2001). Interestingly, there is a structural homology between the catalytic site of d,d-carboxypeptidase and that of sonic hedgehog secreted protein (Shh-N), which is involved in embryonic tissue patterning in an enzyme activity-independent fashion (Fuse et al. 1999). It is possible that CPE may function in a similar manner in signaling to mediate neuroprotection. Future identification of molecules which interact with CPE will facilitate the discovery of the mechanism underlying CPE-mediated neuroprotection.

ADNP and Neuroprotection

Activity-dependent neuroprotective protein (Gozes 2007) was originally cloned from P19 mouse carcinoma cells (Bassan et al. 1999). The ADNP (Sigalov et al. 2000; Zamostiano et al. 2001) deduced protein structure contains zinc fingers, a proline-rich region, a nuclear bipartite localization signal, cellular export and import signals, and a homeobox domain profile, suggesting multiple functions. The hADNP gene was mapped to chromosome 20q12-13.2 (Zamostiano et al. 2001), a region potentially associated with mental function (Borozdin et al. 2007; Vulih-Shultzman et al. 2007).

Recombinant human ADNP protected against beta amyloid peptide, Alzheimer’s disease-related toxicity, and oxidative stress in pheochromocytoma cells (Steingart and Gozes 2006). ADNP knockout mice exhibited failure of cranial neural tube closure and death on E8.5-9.5 (Pinhasov et al. 2003). Furthermore, the heterozygous ADNP+/− mice showed neuronal/glial pathology and reduced cognitive functions (Vulih-Shultzman et al. 2007). In turn, the neuroprotective peptide motif of ADNP, NAP (Bassan et al. 1999; Gozes et al. 2005a, b), was shown to enhance neural tube closure which is compromised in conditions of fetal alcohol intoxication (Chen et al. 2005; Sari and Gozes 2006) and to enhance cognitive function in various animal models (Gozes et al. 2005a, b), including the heterozygous ADNP+/− mice (Vulih-Shultzman et al. 2007).

Mouse ADNP is expressed at the time of neural tube closure and is sustained throughout embryogenesis (Pinhasov et al. 2003). Mouse embryos exposed to alcohol intoxication (E8) exhibited temporal changes in ADNP expression (Poggi et al. 2002). Comparison of tissues revealed an enrichment of ADNP expression in brain-derived structures (Bassan et al. 1999; Zamostiano et al. 2001) and identified ADNP transcripts in rat astrocytes (Bassan et al. 1999).

In a mouse model of head injury (Zaltzman et al. 2004), a trend toward a reduction at 24 h in the injured cerebral hemisphere was observed for ADNP mRNA as compared to the non-injured hemisphere, contrasting with a significant increase detected in the injured cerebral hemisphere 29 days post injury. The increase in mRNA was further measured at the protein level analyzed by immunohistochemistry. ADNP was localized to glia cells in the thalamus of the injured side of the brain (Gozes et al. 2005a, b) and was suggested to be a part of an endogenous compensatory mechanism, with NAP treatment providing additional protection (Beni-Adani et al. 2001).

Gennet et al. (2008) hypothesized that ADNP may be translocated to the nucleus in stressed or injured cells (Gennet et al. 2008). Up-regulation of ADNP mRNA expression as well as ADNP-like immunoreactivity has also been suggested to occur in activated microglial cells 1 month following injury since strong glial expression was found following traumatic brain injury in mice (Gozes et al. 2005a, b). It is thus possible that ADNP expression is regulated by injury. Other potential regulators of ADNP expression include nerve growth factor (Thippeswamy et al. 2007) and the nitric oxide–cyclic guanosine monophosphate pathway (Cosgrave et al. 2008). Interestingly, in the arcuate nucleus, a region exhibiting brain plasticity in the adult, ADNP expression has shown sexual dichotomy and changes with the estrous cycle, suggesting regulation by sex hormones (see also Dangoor et al. 2005) and an involvement with neuronal plasticity (Furman et al. 2005).

A decrease in ADNP expression was observed in the cerebral cortex and astrocytes from prenatal-ethanol-exposed (PEE) rat fetuses. Furthermore, cocultures of PEE astrocytes with control neurons caused a marked decrease in neuronal growth, differentiation, and synaptic connections relative to the cocultures with control astrocytes, effects that were prevented by the addition of NAP (Pascual and Gueri 2007).

Affymetrix microarrays revealed marked differences in expression profiles between ADNP-deficient mouse embryos (E9) and ADNP-expressing embryos. Up-regulated transcripts in ADNP-deficient embryos encoded proteins enriched in the visceral endoderm such as apolipoproteins, cathepsins, and methallotionins, while down-regulated transcripts consisted of organogenesis markers, including neurogenesis (Ngfr, neurogenin1, neurod1) and heart development (Myl2). ADNP-chromatin immunoprecipitation showed direct interactions with multiple relevant gene promoters including members of the up-regulated as well as the down-regulated gene clusters in pluripotent P19 cells. A comparison between nondifferentiated and neurodifferentiated P19 cells revealed increased interaction of ADNP with chromatin from differentiated cells (Mandel et al. 2007).

To elucidate ADNP functional importance in neuronal differentiation, the P19 cell culture model was used for the determination of ADNP distribution and functional importance after neuronal differentiation. Results showed that ADNP distribution in the cytoplasm and nucleus is unique to neuronal differentiated cells in culture compared to cardiovascular and nondifferentiated pluri-potent cells where ADNP-like immunoreactivity was mostly detected in the nucleus. ADNP-like immunohistochemical localization to the neuronal cytoplasm and neurites was shown here not only in the cellular model but also in brain cerebral cortical and olfactory bulb sections. Small hairpin RNA ADNP down-regulation was used to further investigate ADNP involvement in p19 neurodifferentiation. A robust ∼80% reduction in ADNP led to a substantial reduction in embryoid body formation and to a significant reduction (∼50%) in neurite numbers. These results position ADNP in direct association with neuronal cell differentiation and maturation (Mandel et al. 2008).

Since ADNP contains both a nuclear localization signal as well as a leucine-rich nuclear export signal (Zamostiano et al. 2001), it is no surprise that it is distributed both in the nuclei as well as in the cytoplasm of neurons and astrocytes (Furman et al. 2004). Its differential distribution mostly in neuronal/glial cell-like population (both in P19 as well as in brain sections) may suggest a regulatory function associated specifically with neuronal/glial differentiation and maintenance. A similar finding was reported for the homeobox containing transcription factor Otx2. Subcellular nuclei/cytoplasm localization of Otx2 was shown to determine the fates of rod photoreceptor and bipolar cells during their generation and initial differentiation in the neonatal mouse retina. This suggested that the cell type-specific movement of transcription factors from the cytoplasm to the nucleus and vice versa may be an important mechanism for regulating cell fate determination (Baas et al. 2000). Another example similar to that shown here with ADNP is the homeoprotein Emx2, which is found in the nucleus of immature and mature olfactory sensory neurons, as well as in the axonal compartment of these neurons. Its association with high-density particles and interaction with elF4E strongly suggest that this transcription factor has new non-nuclear functions, most probably related to the local control of protein translation in the olfactory sensory neuron axons (Nedelec et al. 2004). Interestingly, both Otx2 and engrailed2 were also shown to interact with Elf4e (Nedelec et al. 2004) through a putative interaction sequence YXXXXL, where X is any amino acid and Y is any hydrophobic amino acid (Wilhelm et al. 2003). Two potential sequences of this Elf4e putative binding motif are found on the ADNP protein sequence YCNRYLP and YKPGVLL, making it a potential regulator of protein translation in neurons.

Previously, ADNP was found to differentially associate with chromatin upon P19 differentiation, and immunoprecipitation results identified ADNP to be associated with heterochromatin protein 1 alpha (Mandel et al. 2007). The HEK293 human embryonic kidney cell line, that allows efficient transfection with recombinant DNA, was also used as a model for ADNP-interacting proteins and cellular function. Recombinant green fluorescent protein (GFP)-ADNP was localized to HEK293 cell nuclei. When nuclear extracts were subjected to immunoprecipitation with specific GFP antibodies, several minor protein bands were observed in addition to GFP-ADNP. Protein digests and mass spectrometry identified BRG1, BAF250a, and BAF170, all components of the mating type switching/sucrose nonfermenting (SWI/SNF) chromatin remodeling complex as proteins that co-immunoprecipitate with ADNP. These results were verified utilizing BRG1 antibodies. ADNP shRNA down-regulation resulted in changes in cell morphology, reduction in cell process formation, and cell–cell interaction, functions that are closely associated with the SWI/SNF complex multi-functionality (Mandel and Gozes 2007). ADNP is thus suggested to have multiple interacting proteins uncovering a molecular basis for the essential function of the ADNP gene and protein.

The involvement of ADNP in neurodifferentiation may stem from its potential function associated with chromatin remodeling and transcriptional regulation, as well as its potential interaction with Elf4e, which is associated with protein translation. Additionally, ADNP function in neuronal regulation may also invoke a cytoplamic interaction with cytoskeletal elements that maintain neurite architecture. In this respect, ADNP immunoreactivity was shown before to occasionally decorate microtubules in astrocytes (Furman et al. 2004), and the neuroprotective ADNP peptide, NAP, was shown to provide neuroprotection through interaction with microtubules (Divinski et al. 2004; Gozes and Divinski 2004; Divinski et al. 2006; Matsuoka et al. 2007). ADNP+/− mice showed increased pathology of the microtubule-associated protein, tau (Vulih-Shultzman et al. 2007), and NAP treatment protected against tau hyperphosphorylation in this ADNP+/− model, as well as in the triple transgenic mouse model of Alzheimer’s disease (Matsuoka et al. 2007).

While NAP does not affect dividing cells and preferentially interacts with beta3 tubulin (Gozes et al. 2003; Divinski et al. 2006), ADNP over-expression has been suggested to be involved in tumorogenesis, which is tightly associated with enhanced cell division. ADNP knock-down in dividing cancer cells resulted in inhibition of cell division and cell death, suggesting differential effects for ADNP in different cell types (Zamostiano et al. 2001). Furthermore, the regulation of ADNP activity may be associated with alternative RNA splicing (Zamostiano et al. 2001) as well as posttranslational modification, as the apparent molecular weight of ADNP may appear higher than that calculated from the amino acid composition (Mandel et al. 2008).

In previous studies using an antibody directed to NAP, the NAPVSIPQ peptide fragment in ADNP (Furman et al. 2004), ADNP-like immunoreactivity was found in both the cytoplasmic and in the nuclear cell fractions of astrocytes. In addition, ADNP-like immunoreactivity was detected in the extracellular milieu of astrocytes, and its content was increased by ∼1.4-fold after incubation of the astrocytes with vasoactive intestinal peptide (VIP). VIP is known to activate astrocytes to secrete neuroprotective/neurotrophic factors, and it is suggested that ADNP constitutes part of this VIP-stimulated protective pathway (Bassan et al. 1999; Pinhasov et al. 2003; Zusev and Gozes 2004; Steingart and Gozes 2006).

Additionally, our studies have shown differential expression of ADNP in the male and female hypothalamus and oscillations with the estrous cycle that were suggested to be involved in synaptic plasticity (Furman et al. 2005). Recently, evidence was obtained for direct requirement for ADNP expression in the process of neurite outgrowth, suggesting effects not only during development but also affecting brain plasticity, recovery from injury, and protection against neurodegeneration (Mandel et al. 2008).

The Neuroprotection Concept

This review outlines the activity of two proteins: CPE—originally discovered as a protease, and ADNP that exhibits transcription factor structural characteristics that are important in neuroprotection. Furthermore, both proteins have been associated with learning and memory, and decreased expression resulted in decreased performance in cognitive tests. These two diverse proteins exhibiting similar functional outcomes may have converging mechanisms of action involving multiple pathways.


We would like to thank Chip Dye and Dr. Vincent Schram (Microscopy and Imaging Core, Eunice Kennedy Shriver National Institute of Child Health and Development, National Institutes of Health) for assistance with microscopic imaging and Dr. Douglas E. Brenneman for critical reading of the manuscript. This study was done in collaboration between Dr. Y. Peng Loh and Prof. Illana Gozes under the support of: The United States Israel Binational Science Foundation, the Lily and Avraham Gildor Chair for the Investigation of Growth Factors, The Adams Super Center for Brain Studies, the Dr. Diana and Zelma Elton (Elbaum) Laboratory for Molecular Neuroendocrinology, the Intramural Research Program of the Eunice Kennedy Shriver National Institute of Child Health, and Human Development and Allon Therapeutics Inc. Prof. Gozes serves as the chief Scientific Officer of Allon Therapeutics Inc.

Copyright information

© Humana Press 2009