We have previously shown that following traumatic brain injury (TBI) the presence of the amyloid precursor protein (APP) may be neuroprotective. APP knockout mice have increased neuronal death and worse cognitive and motor outcomes following TBI, which is rescued by treatment with exogenous sAPPα (the secreted ectodomain of APP generated by α-secretase cleavage). Two neuroprotective regions were identified in sAPPα, the N and C-terminal domains D1 and D6a/E2 respectively. As both D1 and D6a/E2 contain heparin binding activity it was hypothesized that this is responsible for the neuroprotective activity. In this study, we focused on the heparin binding site, encompassed by residues 96-110 in D1, which has previously been shown to have neurotrophic properties. We found that treatment with APP96-110 rescued motor and cognitive deficits in APP−/− mice following focal TBI. APP96-110 also provided neuroprotection in Sprague–Dawley rats following diffuse TBI. Treatment with APP96-110 significantly improved functional outcome as well as preserve histological cellular morphology in APP−/− mice following focal controlled cortical impact injury. Furthermore, following administration of APP96-110 in rats after diffuse impact acceleration TBI, motor and cognitive outcomes were significantly improved and axonal injury reduced. These data define the heparin binding site in the D1 domain of sAPPα, represented by the sequence APP96-110, as the neuroprotective site to confer neuroprotection following TBI.
The product of α-secretase cleavage of the amyloid precursor protein, sAPPα is neuroprotective following traumatic brain injury (TBI). Of interest was whether this neuroprotective activity could be further delineated to a heparin binding region within sAPPα, corresponding to the region APP96-110 (see diagram demonstrating the domain structure of sAPPα). Indeed treatment with APP96-110 improved functional outcome following TBI, an effect that was not seen with a mutated version of the peptide that had reduced heparin binding affinity.
Traumatic brain injury (TBI) is a significant health problem with an estimated 10 million people worldwide affected every year by a new TBI event (Langlois et al. 2006), with projections indicating that TBI will become the third largest cause of global disease burden by 2020 (Hyder et al. 2007). Following TBI cell death is ongoing, because of the initiation of a number of secondary injury factors such as glutamate excitotoxicity, oxidative stress and inflammation (Maas et al. 2008). Thus, a window of opportunity exists whereby neuronal damage can be limited with appropriate treatment.
It has been proposed that candidate therapeutic agents should aim to emulate the body's own neuroprotective pathways (Faden and Stoica 2007). One such agent is the amyloid precursor protein (APP), which we have previously shown to be neuroprotective in the immediate period following TBI (Corrigan et al. 2012a). These protective properties of APP may relate to the product of α-secretase pathway in sAPPα, as APP knockout mice treated with sAPPα were comparable to their wildtype counterparts (Corrigan et al. 2012c). The neuroprotective regions within sAPPα have been narrowed down to the D1 and D6a domains (Corrigan et al. 2011). The predominant isoform of sAPPα within the brain, APP695, consists of four domains: a growth-factor like domain (D1), a copper binding region (D2), an acidic region (D3) and a carbohydrate domain (D6), which can be further divided into an E2 domain (D6a) and a juxtamembrane region (D6b) (Storey and Cappai 1999; Reinhard et al. 2005). As the D1 and D6a domains both contain heparin binding sites, it was hypothesized that the neuroprotective properties of sAPPα may relate to its ability to bind heparin sulphate proteoglycans (HSPGs) (Corrigan et al. 2011).
The heparin binding site within the D1 domain (APP96-110), has neurotrophic properties as binding of this region to HSPGs promotes neurite outgrowth (Small et al. 1994). Whilst competitive binding to this region prevents functional synapse formation (Morimoto et al. 1998) and depolarisation induced neurite outgrowth (Gakhar-Koppole et al. 2008). Utilization of the heparin binding site within the D1 domain has a number of functional advantages over that located within the D6a domain. It has a sequential sequence, as well as a disulphide bridge between cysteines 98 and 105 (Rossjohn et al. 1999; Small et al. 1994), increasing the likelihood that it will maintain the correct conformational shape when assembled independently of the D1 domain. To determine the efficacy of the heparin binding site within the D1 domain of APP as a potential neuroprotective agent two studies were conducted. Firstly, APP−/− were treated with the APP96-110 peptide following focal controlled cortical impact (CCI) injury, to determine its ability to rescue deficits in these mice. Secondly, the ability of the peptide to improve outcome in non-transgenic animals was evaluated by treating Sprague–Dawley rats following diffuse impact-acceleration TBI.
All studies were performed within the guidelines established by the National Health and Medical Research Committee of Australia and were approved by the Animal Ethics Committees of the Institute of Medical and Veterinary Sciences and the University of Adelaide. Experiments were performed in compliance with the ARRIVE guidelines.
Study 1: Evaluation of the efficacy of APP96-110 in rescuing deficits in APP−/− mice
All experiments were performed on male APP−/− mice (C57BLbj x 129sv background) aged 10–12 weeks and housed in a controlled temperature environment under a 12 h light/dark cycle, with uninterrupted access to food and water. Mice were randomly assigned into each of the four experimental groups: sham, vehicle control, D1 domain treated and APP96-110 treated.
Animals were anaesthetized with an intraperitoneal injection of 2,2,2 Tribomoethanol (250 mg/kg) prior to conduction of the CCI injury, as previously described (Corrigan et al. 2012c). Briefly, a portable drill was used to create a 3 mm craniotomy over the right parieto-temporal cortex. The bone flap was removed with care taken to leave the dura intact. Mice were then placed within a stereotaxic frame and a 2 mm tip lined up so that it was directly touching the surface of the dura. Impact was then delivered at a velocity of 5 m/s via an electromagnetic piston, with a dwell time of 100 ms and a depth of 1.2 mm. Sham animals underwent a craniotomy without impact.
At 30 mins post-injury mice received treatment via an intracerebroventricular injection of 2 μL of either the D1 domain (APP28-123, n = 9), APP96-110 (n = 13) or aCSF (n = 15). The D1 domain was expressed as a secreted protein in the methylotrophic yeast, Pichia pastoris as previously described (Henry et al. 1997). APP96-110 was produced through custom peptide synthesis (Auspep, Tullamarine, Vic., Australia) using the following sequence, NWCKRGRKQCKTHPH, and was N-terminally acetylated and C- terminally amidated. Following this, 150 μL of each peptide (25 μM) was mixed with 75 μL of artificial CSF vehicle (Roch et al. 1994; Thornton et al. 2006) prior to administration. A 0.3 mm craniotomy was performed on the uninjured left side, at the stereotaxic coordinates relative to bregma: posterior 0.5 mm, lateral 1 mm (Paxinos and Franklin 2007). The needle was then lowered 2.5 mm and retracted 0.3 mm to allow injection of 2 μL of the appropriate treatment, which was administered at a rate of 0.5 μL/min. The scalp incision was then sutured and the mice returned to their cage to recover.
Motor deficits were assessed on the ledged beam, as previously described (Corrigan et al. 2012c). Briefly, following training 3 days prior to injury, mice were tested for 7 days following injury, with the number of times the underhanging ledge was used (foot faults) by the fore and hindlimb on the left side counted and averaged across two trials.
Cognitive deficits were assessed using the Barnes Maze, which has been described in detail elsewhere (Koopmans et al. 2003). In short the Barnes Maze consists of a white circular platform with 40 holes spaced at equal intervals around the perimeter. An escape box is connected to one of these holes and its location is randomized between animals. Animals were pre-trained for 5 days prior to injury, with their best time taken as their pre-injury baseline level. Assessment was conducted on days 2, 4 and 6 post-injury, with escape latency (time in seconds) for the mice to find and enter the escape box with front paws and trunk recorded.
Mice were terminally anaesthetized with pentobarbital and perfused transcardially with 10% formalin. Twelve sections per brain, 5 μm thick, were collected at 200 μm intervals to represent the region Bregma 0 to −4 and were stained with haematoxylin and eosin (H&E). These slides were then scanned (Nanozoomer, Hamamatsu City, Japan) and viewed using the associated software (NDP View Version 22.214.171.124; Hamamatsu, Hamamatsu City, Japan) by a blinded observer. As previously described lesion volume was determined by outlining the unaffected area of the cortex of each hemisphere and then calculating the hemispheric volume by summing area of each section and multiplying it by 0.2 μm (Corrigan et al. 2012c). Hippocampal injury was determined by assessing the CA region from three sections representing Bregma −1.2 to −2.1. Sequential images were taken at 40× magnification and the cell count software associated with Image J used to manually count neurons as previously described (Corrigan et al. 2012c). Furthermore, the effect of injury on the dentate gyrus was assessed by determining the area of the granular layer of the dentage gyrus in five sections located 200 μm apart (Bregma −1.5 to −3), with the volume calculated by summing the area of each section and multiplying it by 0.2.
Study 2: Examining the efficacy of APP96-110 as a potential neuroprotective agent following TBI
A total of 52 adult male Sprague–Dawley rats weighing between 360 and 400 g were randomly assigned into four groups: sham, vehicle control, APP96-110 and mAPP96-110 treatment groups (n = 13 per group). The APP96-110 was manufactured as detailed above, with the mAPP96-110 peptide also synthesised according to the following sequence: NWCNQGGKQCKTHPH (Auspep). Thus, mAPP96-110 is mutated as K99N, R100Q, R102G, to reduce the ability of this peptide to bind heparin. Animals were injured using the impact-acceleration model of diffuse TBI as described previously (Marmarou et al. 1994). Animals were anaesthetized with isoflurane and a midline incision performed. A stainless steel disc (10 mm in diameter, 3 mm in depth) was then fixed using polyacrylamide adhesive to the animal's skull so that it was located centrally between the lambda and bregma sutures. Brain injury was then induced by placing the rats on a 12 cm foam bed and releasing a 450 g weight from 2 m directly onto the stainless disc. Sham operated animals were surgically prepared but were not injured. At 30 min following injury rats received treatment via an intracerebroventricular injection of either APP96-110, mAPP96-110 or aCSF. A 0.7 mm craniotomy was performed at the stereotaxic coordinates relative to bregma: posterior 0.6 mm, lateral 1.5 mm, with the needle then lowered to 4.0 mm and retracted 0.5 mm before drug delivery at a rate of 0.5 μL/min. Animals were then removed from the device and returned to their home cage to recover.
Motor deficits produced by TBI were assessed using the rotarod. The rotarod is a motorized rotating assembly of eighteen 1 mm metal rods, with speed of the device increased from 6 to 36 revolutions per minute in intervals of 3 rpm every 10 min. Animals were pre-trained daily on the device for 5 days prior to injury and assessed daily for 7 days after injury. The length of time the animals were able to remain actively walking on the device was recorded up to a maximum of 120 s, as previously described (Corrigan et al. 2011).
Cognitive deficits produced by TBI were assessed using the Y maze, which tests spatial and recognition memory (Conrad et al. 1996). As previously described, the maze consists of three arms, given the arbitrary designation of start, other and novel arms, with a different object placed at the end of each arm, to provide a stimulus for the rats to continue to explore the maze (Corrigan et al. 2011). Animals are initially introduced into the maze with one of the arms (novel) closed off, before being re-introduced an hour later with all the arms available for exploration. All trials were recorded on video, with the number of times each arm was visited and the time spent in each arm analysed. Animals that did not enter the novel arm during the first part of the experiment were not included.
Rats were killed by perfuse fixation of 10% formalin transcardially at either 24 h or 7 days post-injury. Fixed brains were cut into six 4 mm coronal slices and embedded in paraffin. Three serial 5 μm sections were taken 300 μm apart starting from the region −3.5 from bregma and examined for axonal injury. Sections were incubated overnight with a monoclonal antibody specific to APP (Novocastra, 1 : 1000), followed by the appropriate biotintylated secondary antibody (1 : 250, Vector). They were then incubated with strepavidin conjugate, with bound antibody then detected with 3,3 diaminobenzidine tetrahydrochloride following counter-staining with haematoxylin. All slides were scanned (Nanozoomer, Hamamatsu) and viewed with the associated software (NDP view) to allow the number of APP immunopositive lengths along the corpus callosum to be counted.
Examination of the heparin binding ability of the APP96-110 and mAPP96-110 peptides
The peptides were dissolved in phosphate-buffered saline to a final concentration of 1 mM. 10 μL of peptide was injected onto a 1 ml HiTrap Heparin column (GE Healthcare, Sydney, NSW, Australia) equilibrated in Buffer A (20 mM Tris 7.5) and connected to an AKTA Purifier (GE Healthcare). The bound peptide was eluted from the column in a linear gradient of 0–1.2 M NaCl at a flow rate of 1 mL/min. Eluted peptide was detected by UV absorbance at 220 nm.
Cognitive and motor outcome data were assessed using a repeated measures two-way anova, followed by Bonferonni tests for multiple comparisons. Lesion volume, hippocampal cell counts and dendritic volume data were assessed using a one-way anova followed by Newman–Keuls tests for multiple comparisons. A p-value of less than 0.05 was considered significant. All graphical data are presented as mean ± SEM.
Study 1: Evaluation of the efficacy of APP96-110 in rescuing deficits in APP−/− mice
Effects of treatment with APP96-110 on functional outcome in APP−/− mice
As previously reported injured APP−/− mice had a significant increase in the number of foot faults as compared to sham animals and remained impaired over the entire 7 day testing period (Fig. 1a). Treatment with either the D1 protein or APP96-110 led to a reduction in the number of foot faults in APP−/− mice, which persisted over the 7 days tested. This reached significance from days 1–4 for the D1 treatment group and days 2 and 3 for the APP96-110 treatment groups (p <0.05). Similarly, cognitive deficits noted in injured APP−/− mice were rescued by treatment with either D1 or APP96-110, with significantly decreased escape latency on the Barnes Maze on days 2 and 4 post-injury (p <0.05)(Fig. 2a). When tested for their ability to learn a new location for the escape hole on day 7 post-injury, treatment with the D1 domain of sAPPα or APP96-110 significantly reduced escape latency on trial 2 (p <0.05) when compared to vehicle treated APP−/− mice (Fig. 1b). There were no significant differences between mice treated with D1 or APP96-110, in either motor or cognitive outcome.
Effect of treatment with APP96-110 on histological outcome in APP−/− mice
Following injury, mice treated with either the D1 or APP96-110 demonstrated greater histological preservation of cortical and hippocampal tissue than vehicle treated APP−/− mice (Fig. 2). Indeed, both D1 and APP96-110 led to an increase in the number of remaining CA neurons as counted within three sections when compared to vehicle treated APP−/− mice (839 ± 244 in APP−/− vs. 1479 ± 193 in D1 treated mice and 1654 ± 233 in APP96-110 treated mice). There was also an improvement in survival of dentate gyrus neurons. A calculation of lesion volume, as determined as a %cortical damage, found a reduction from 25.1 ± 1.8% in vehicle treated APP−/− mice to 19.1 ± 1.4% in mice treated with the APP96-110 and 18.2 ± 1.2% in D1 treated mice.
The heparin binding activity of APP96-110 and mAPP96-110
The finding that APP96-110 had similar neuroprotective activity to D1 supported our hypothesis this protective activity against TBI is mediated via its heparin binding activity. To test this hypothesis we mutated the heparin binding residues, K99N, R100Q, R102G, in APP96-110 (mAPP96-110). The heparin binding activity of APP96-110 and mAPP96-110 were measured on a Heparin-Sepharose column. APP96-110 eluted at 540 mM NaCl, while mAPP96-110 eluted at the significantly lower concentration of 230 mM NaCl (Fig. 3). This demonstrates that K99N, R100Q, R102G are necessary for heparin binding and that mAPP96-110 has significantly weaker heparin binding activity then APP96-110.
The heparin binding activity of APP96-110 directly correlates with its neuroprotective activity against TBI
To directly test that heparin binding activity correlates with its neuroprotective activity against TBI, we compared mAPP96-110 and APP96-110 in a diffuse injury model in Sprague–Dawley rats. Treatment with APP96-110 significantly improved motor outcome, as tested by thee rotarod, when compared with vehicle treated controls (Fig. 4a). The APP96-110 treated rats had returned to a sham level by day 6 following injury. In contrast, the mAPP96-110 treated rats displayed significantly worse performance on the rotarod from days 1 to 4 when compared to the APP96-110 treated rats. Similarly on the Y Maze, the APP96-110 treated rats, like the sham rats, spent significantly more time in the novel arm compared to the other two arms (Fig. 4b). In contrast, the mAPP96-110 treated rats had a significantly different profile to the APP96-110 treated rats with no difference in time spent between the novel arm and the other arm. There were no significant differences in exploration between the groups (data not shown).
The effects of APP96-110 on axonal injury following diffuse injury in Sprague–Dawley rats
Levels of axonal injury were determined by counting the number of APP immunopositive lengths within three serial sections of the corpus callosum (Fig. 5). A significant increase compared to shams was noted in vehicle and mAPP96-110 treated rats. In contrast treatment with APP96-110 treatment reduced axonal injury following TBI, with a significant decrease noted at day 7 compared to both vehicle and mAPP96-110 treated rats (41 ± 19 versus 126 ± 66 and 99 ± 14 APP immunopositive lengths respectively).
This study showed that residues 96-110 in D1, via their heparin binding activity, are responsible for the neuroprotective activity against TBI. These results are in line with studies showing APP96-110 has neurotrophic properties. The APP96-110 region contains a β hairpin loop formed by a disulphide bridge between cysteines 98 and 105 (Rossjohn et al. 1999; Small et al. 1994), with the presence of this bridge previously shown to be critical for promoting neurite outgrowth (Young-Pearse et al. 2008) and activation of MAP kinase (Greenberg and Kosik 1995). This most likely relates to maintenance of the correct secondary structure to facilitate binding to heparin, with binding of this region to HSPGs promoting neurite outgrowth from central and peripheral neurons (Small et al. 1994). An antibody that binds to this region can inhibit functional synapse formation (Morimoto et al. 1998), completely abolishes depolarisation induced neurite outgrowth (Gakhar-Koppole et al. 2008) and prevents the effects of sAPPα in promoting the migration and differentiation of human neural stem cells (Kwak et al. 2006).
We employed APP−/− mice in the CCI injury model as our previous work found a substantial treatment effect with sAPPα (Corrigan et al. 2012c). In contrast we could only achieve a moderate treatment effect in wild-type mice treated with sAPPα following CCI injury (Corrigan et al. 2012b). It was determined that the greatest likelihood of gaining a significant treatment effect with APP96-110 in non-transgenic animals would be to utilize APP−/− mice in the CCI injury model. We included the rat weight-drop model since treatment with D1, as well as sAPPα, resulted in a significant improvement in functional outcomes in injured rats (Corrigan et al. 2011; Thornton et al. 2006).
Treatment with the APP96-110 rescued functional deficits in these two rodent TBI models. In APP−/− mice, APP96-110 treatment caused a reduction in the number of foot faults on the ledged beam and a decrease in escape latency on the Barnes Maze. This was accompanied by a significant reduction in the amount of cortical damage and significantly greater preservation of both the CA and dentate gyrus regions of the hippocampus. APP96-110 was as effective as the entire D1 domain of sAPPα, suggesting APP96-110 is responsible for the neuroprotective activity. In rats, following diffuse impact-acceleration TBI, APP96-110 significantly improved motor and cognitive outcome, with an associated decrease in axonal injury.
Literature proposes that coulombic forces between basic amino acids (positively charged) and anionic (negatively charged) groups on the polysaccharide are of major importance for the interaction between heparin and proteins (Fromm et al. 1995). To reduce the interaction of APP96-110 peptide to heparin, we have mutated the positively charged residues K99N, R100Q, R102G in mAPP96-110. Our data strongly suggest the neuroprotective actions of APP96-110 relate to its heparin binding activity since mAPP96-110 had a significantly reduced affinity for heparin, and failed to provide any neuroprotection. Heparin sulphate chains are typically attached to core proteins to form HSPGs, which are located at the cell surface and the extracellular matrix (ECM) (Turnbull et al. 2001). Members include the syndecan family of transmembrane proteins, the glypican family attached to the cell membrane and ECM proteins such as perlecan (Raman et al. 2005). Specific sequences in the heparin sulphate chains mediate interactions with specific proteins and regulate a number of functions such as cell signalling, cell adhesion and neurite outgrowth (Spillman and Lindahl 1994), as well as responding to neuronal injury, through modulation of growth factors and axon growth (Rapraeger 2002; Lin et al. 1999). Indeed, cell surface HSPGs act as co-receptors for growth factors such as fibroblast growth factor (FGF), Wnt, and transforming growth factor-β can regulate cell proliferation and differentiation (Ford-Perriss et al. 2002; Song et al. 2005; Gallagher 2007). The majority of the binding sites for APP in the ECM and cell surface are via these HSPGs and restriction of APP binding heparin prevents its neuroprotective (Furukawa et al. 1996). This supports the findings of this study, whereby abolition of APP96-110's heparin binding activity in mAPP96-110 led to a loss of its neuroprotective actions.
Heparin is known to facilitate FGF signalling by facilitating the dimerisation of FGFRs and regulating downstream signalling pathways (Beenken and Mohammadi 2009; Itoh 2007). In vitro experiments have demonstrated that FGF is dependent on HSPGs for efficient signal transduction (Olwin and Rapraeger 1992). Heparin is also thought to promote the dimerisation of APP, as well as heterodimerisation of APP with its family members APLP1 and APLP2 (Soba et al. 2005). Of note, mutant APP and amyloid precursor like protein constructs deficient in the D1 domain showed a reduction in dimerisation (Soba et al. 2005). The APP96-110 region is essential, as the addition of a peptide encompassing APP96-110 reduced APP dimerisation (Kaden et al. 2008). It is yet to be determined the functional effects mediated by homo- and heterodimerisation of the APP family members, although it has been hypothesized that cis-dimerisation of membrane bound APP with sAPPα may allow activation of intracellular signalling pathways (Gralle et al. 2009; Reinhard et al. 2005). Alternatively cis dimerisation may simply influence the processing of APP, possibly by changing its conformational state, thereby decreasing its amyloidogenic processing by preventing access of β-secretase to its cleavage site (Kaden et al. 2008). It is possible that sAPPα may act in a similar manner to known growth factors, whereby heparin is required to facilitate binding of APP to a co-receptor, through the APP96-110 region.
The finding that APP96-110 is protective in APP−/− mice, which obviously precludes an interaction between APP96-110 and endogenous APP, indicates that sAPPα or APP96-110 are not binding to endogenous APP as a potential neuroprotective mechanism. However, APLP2 may take over some of the functions of APP in these mice. Indeed, as APP−/− mice age, levels of both APLP1 and APLP2 are known to increase in the brain relative to wild-type mice, suggesting they may compensate for some of the physiological activities of APP (Soba et al. 2005). Indeed, the functions of this family of proteins appear to be partially redundant, with only minor phenotypic effects of single knockout of any of these genes, or of APP/APLP1 knockout, whilst APP/APLP2 and APLP2/APLP1 knockout are perinatally lethal (von Koch et al. 1997; Heber et al. 2000). Of note APLP2, but not APLP1 has a heparin binding site within the D1 domain (Anliker and Muller 2006). Given the importance of APLP2, as highlighted by the combined knockout studies, as well as its ability to interact with the heparin binding site of the D1 domain, it is possible that APLP2 could be an alternate receptor for APP96-110 signalling in APP−/− mice, either dependent or independent of heparin binding.
This study has delineated the TBI neuroprotective site in the APP D1 region to the heparin binding residues, and this activity can be replicated by a 15 amino acid peptide. This activity would appear to be mediated by its ability to bind heparin, although further research will be needed to clarify this issue by identifying the receptor of APP96-110 and whether it is indeed a HSPG. Nonetheless, the availability of the crystal structure of this region (Rossjohn et al. 1999) allows the application peptidomimetics and in silico modelling to devise analogues as therapeutics for TBI.
This study was funded in part by a grant from the National Health and Medical Research Council of Australia. We thank Prof. Hui Zheng for providing the APP knockout mice and Dr Mark Habgood (University of Melbourne), Dr Jenna Zeibell and Jim Manavis for expert technical assistance. We thank Dr Chi Pham (The University of Melbourne) for purifying the D1 domain. RC is a National Health and Medical Research Council Senior Research Fellow. No authors have a conflict of interest to declare.