Background

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disorder in which selective loss of motor neurones in the cerebral cortex, brainstem and spinal cord leads to progressive paralysis. In the majority of cases of ALS, the cause of motor neurone degeneration is unknown, although a number of pathogenic processes, including oxidative stress, excitotoxicity, inflammation, and mitochondrial and neurofilament dysfunction, are thought to play important roles. Ten percent of cases of ALS are familial, and in 20% of these, a causative mutation is found in the gene encoding superoxide dismutase I (SOD1), a free radical scavenger. In the SOD1 rodent model of ALS, overexpression of human mutant SOD1 causes adult onset motor neurone degeneration.

Vascular endothelial growth factor (VEGF) is an endothelial cell mitogen that stimulates angiogenesis in response to hypoxia, in the developing embryo and in a number of pathological conditions, such as tumour growth. VEGF transcription is upregulated by binding of hypoxia inducible factor (HIF-1) to a hypoxia response element (HRE) in the 5' promoter region of the gene. In 2001, Oosthuyse et al deleted the HRE of the VEGF gene in mice, to generate VEGFδ/δ mice [1], which express reduced levels of Vegf in neural tissue under both baseline and hypoxic conditions. VEGFδ/δ mice show 60% mortality at or around birth, and surviving transgenic mice are slightly smaller than their wild-type littermates. At 5 months of age, they develop a motor neurodegenerative phenotype that resembles ALS, with impairment of motor behaviours and motor tests, such as the treadmill test. Electrophysiological studies show signs of denervation and compensatory reinnervation, muscle histology shows neurogenic atrophy, and peripheral nerves show loss of large myelinated motor axons. In the spinal cord and brainstem, similar numbers of motor neurones are present until 3 months of age, but by 17 months, there is a 30% reduction in motor neurone numbers, with a reactive astrocytosis, and neurofilament inclusions in surviving neurones [1]. The mechanism of neurodegeneration in VEGFδ/δ mice is unknown. Chronic hypoxia has been proposed, as although vascular structure in the sciatic nerve and spinal cord is normal, baseline neural blood flow is reduced by 44%. In addition to vascular development, however, VEGF plays a central role in the development of the nervous system, and may be required for survival of adult neurones [2]. Disruption of these functions may determine the development of neuronal degeneration in VEGFδ/δ mice.

VEGF has neurotrophic effects in vitro in a wide range of culture conditions, promoting cell survival and neurite outgrowth, via its tyrosine kinase receptor, VEGFR2 [3]. In vivo, VEGF administration prolongs survival in the SOD1 model of ALS [4]. This is likely to be a direct neuroprotective effect, as SOD1 mice crossed with mice with neuronal overexpression of VEGFR2 also exhibit delayed disease progression [5]. In the developing nervous system, VEGF released by neuroblasts and glia induces the ingrowth of capillaries, and deletion of a single copy of VEGF is lethal [6, 7]. VEGF stimulates neurogenesis directly [8], and via endothelial cell proliferation, to form vascular niches in which neurogenesis is stimulated by endothelial-derived BDNF [9, 10]. Further interaction between vascular and neuronal development is seen in the closely aligned growth of blood vessels and peripheral nerves, directed by attractive and repulsive cues on the growth cone [11]. The VEGF co-receptor, Neuropilin-1 mediates both repulsive signals of the semaphorin family, and attractant signals of VEGF [11, 12]. Thus VEGF, with its angiogenic and neurotrophic actions via VEGFR2, and shared receptor with the semaphorin axon guidance factors, may be a key player in the parallel development of vascular and nervous systems.

A recent meta-analysis of association studies of VEGF polymorphisms with ALS showed an increased risk of ALS in male patients with the -2578AA genotype, which lowers VEGF expression[13]. We have previously shown that levels of expression of VEGF and its main agonist receptor are reduced in the spinal cord of patients with ALS[14]. This study aims to clarify the molecular mechanisms of neurodegeneration in the VEGFδ/δ mouse, by determination of the transcriptional profile of isolated spinal motor neurones in the transgenic mouse, compared to its wild type littermate. Understanding the role of VEGF in the survival and death of motor neurones in this mouse model of ALS may have implications for the human disease.

We report that adult VEGFδ/δ mice show reduction in expression, from symptom onset, of genes involved in nervous system development, particularly in axonogenesis and synapse formation, and that axon outgrowth is reduced in motor neurone cultures derived from VEGFδ/δ mice. These changes suggest a role for VEGF in the maintenance of neuronal circuitry, disruption of which may result in a dying-back axonopathy.

Methods

Experimental animals

Adult female VEGFδ/δ mice and wild type littermates (Vesalius Research Institute, Leuven, Belgium) were used in this study. VEGFδ/δ mice, on a Swiss/129 background, had homozygous deletion of the HIF-1 response element (hypoxia response element) in the promoter region of the VEGF gene, and were generated as previously described [1]. All mice were housed in conventional facilities with a 12 h light/dark cycle with access to food at libitum. The local animal ethical committee, the Ethische Commissie Dierproeven (ECD) at the Catholic University Leuven, Belgium, approved the VEGFδ/δ mouse experiments.

Tissue collection

Three female VEGFδ/δ transgenic mice, and three gender-matched wild-type littermate controls were sacrificed at 3 months (pre-symptomatic), 5 months (onset of symptoms) and 14 months (late stage disease) of age, by overdose of isoflurane inhalational anaesthetic. Post mortem, animals were perfused by intracardiac injection of 15 ml sterile phosphate buffer with 30% sucrose, and the CNS was dissected and frozen in Cryo-M-Bed embedding compound (Bright, Huntingdon, UK). The procedure of sucrose perfusion and dissection of the spinal cord was conducted rapidly, within a maximum of ten minutes from terminal anaesthesia to snap freezing of tissue, to ensure optimal preservation of RNA quality. Lumbar spinal cord sections (10 μm) were fixed in 70% ethanol, washed in DEPC-treated water, and stained for 1 minute in a solution of 0.1% w/v Toluidine Blue in 0.1 M sodium phosphate. They were then washed and dehydrated through graded ethanol concentrations (70, 90 and 100%), and xylene.

Laser capture microdissection, RNA isolation and amplification

Spinal motor neurones, identified by staining, anatomical location, size and morphology, were isolated on Capsure Macro LCM caps using the Arcturus PixCell II laser capture microdissection system (Arcturus Bioscience, Mountain View, CA). Approximately 1500 motor neurones were dissected from each spinal cord, and >50 ng RNA extracted using the PicoPure™ RNA isolation kit (Arcturus), according to manufacturer's instructions. RNA amplification was carried out using a linear amplification process in 2 cycles, with the GeneChip two cycle target labelling and control kit (Affymetrix, Santa Clara, USA), and MEGAscript® T7 kit (Ambion, Austin, USA). The linear amplification technique has been shown to generate highly reproducible gene expression profiles from small starting quantities of RNA [15]. This procedure produced 50-100 μg of biotin labelled antisense RNA for each sample, the quality and quantity of which was assayed using the Agilent bioanalyser and Nanodrop™ 1000 spectrophotometer (Thermo Scientific, Wilmington, USA).

Quality control parameters

At each stage of extraction, amplification and microarray analysis, we carried out quality control (QC) measures, according to Affymetrix protocols, to ensure that RNA was of sufficient quality and was matched between samples. Where QC outliers were identified, these samples were excluded from the analysis. We used visual interpretation of RNA profiles from bioanalyser traces of extracted and amplified RNA to determine RNA quality, as the 28S:18S ratio has been shown to have only weak correlation with gene expression levels in downstream experiments[16]. The RNA profiles obtained from laser captured material in this study were comparable to profiles with modest RNA degradation, that has been shown to have little effect on the results of microarray analysis [17, 18] (Figure 1) Following amplification, the most frequent length of RNA amplicons, at 500-1000 kB, was consistent with Affymetrix protocols. Affymetrix and Bioconductor software was used to generate, from microarray data, the QC measures of average background, signal intensity, percent present calls, and RNA degradation plots, to ensure that samples were comparable in all parameters.

Figure 1
figure 1

Representative bioanalyser traces of RNA samples pre- and post-amplification at 3 months, 5 months and 14 months.

Full size image

Affymetrix GeneChip processing

15 μg amplified cRNA was fragmented by heating to 94°C for 20 minutes, and spiked hybridization controls were added. Each sample was hybridized to one mouse 430A2 GeneChip (Affymetrix) according to manufacturer's protocols. Following overnight hybridization at 42°C, GeneChips underwent stringency washes in a Fluidics Station 400, then were scanned in the GeneChip Scanner 3000 to detect fluorescent hybridization signals. These were analysed by the Affymetrix GeneChip Operating System (GCOS) to generate an overall hybridization signal for each transcript from 11 representative perfect match and mismatch probe pairs.

Microarray data analysis

CEL files generated by GCOS were imported for further analysis into Array Assist software (Stratagene, La Jolla, USA), where probe level analysis was carried out using the GC Robust Multichip Average (GC-RMA) algorithm. Following GCRMA processing, data was filtered to remove those genes whose expression was at or around the background signal level of the chip. Any probe set that returned a signal of <50 on more than 3 chips at each time point was excluded from further analysis. This signal filter did not remove genes that were only expressed in one experimental group. Differential gene expression was determined using an unpaired t-test, to generate a list of genes that were significantly differentially expressed between transgenic mice and their wild type littermates, at each time point.

Gene ontology analysis

Gene Ontology (GO) terms that reflect the function of the corresponding genes are assigned to each probe set by Affymetrix software. This GO information was used to determine which functionally related groups of genes were over-represented amongst significantly differentially expressed genes at each time point. The frequencies of GO terms represented in the significant gene lists were compared to a denominator list of genes that can be expressed by motor neurones in health or disease. GO analysis was carried out without regard for fold change, in order to limit type II errors and in order to detect more subtle changes in gene expression, for example in transcriptional regulators, which may be biologically significant. The denominator list was generated from the array data for each time point, by extracting those genes whose representative probe sets returned a signal of >50 in 3 or more chips. Statistical analysis of GO term enrichment was carried out using DAVID software (NIAID/NIH; http://david.abcc.ncifcrf.gov/summary.jsp, [19]. Literature review was also used to identify significantly differentially expressed genes with functions relating to those identified as enriched by DAVID analysis.

Verification of microarray results by Quantitative rt-PCR (QPCR)

A proportion of genes identified as significantly differentially expressed were selected for verification by QPCR, on the basis of robust microarray data confirming differential gene expression, involvement in a biological process identified as enriched by GO analysis, or a known function in neurodegeneration. Verification addresses the possibility of false positive microarray signals, due to cross-hybridization with related genes, concern about the accuracy of array probe sets, and uncertainty about the hybridization kinetics of multiple reactions occurring on the miniature scale of an array chip. RNA was extracted from 1000-1500 cells, isolated by laser capture microdissection from the lumbar spinal cord of the population of mice used in the microarray experiment and, where available, a second population of transgenic mice with wild-type littermates. RNA was extracted, quantified as previously described, and reverse transcribed to cDNA using Superscript II reverse transcriptase, according to manufacturer's protocol (Invitrogen, San Diego, CA). Primers used in for verification are shown in Table 1. QPCR was performed using 12.5 ng cDNA, 1×SYBR Green PCR master mix (Applied Biosystems, Foster City, CA), and forward and reverse primers at optimized concentrations, to a total volume of 20 μl. After an initial denaturation at 95°C for 10 mins, templates were amplified by 40 cycles of 95°C for 15 secs and 60°C for 1 minute, on an MX3000P Real-Time PCR system (Stratagene). A dissociation curve was then generated to ensure amplification of a single product, and absence of primer dimers. For each primer pair, a standard curve was generated to determine the efficiency of the PCR reaction over a range of template concentrations from 0.3 ng/μl to 25 ng/μl, using cDNA synthesized from mouse universal RNA. The efficiency for each set of primers was 100+/-10%, such that gene expression values, normalized to ß-actin expression, could be determined using the ddCt calculation (ABI PRISM 7700 Sequence Detection System protocol; Applied Biosystems). An unpaired t-test was used to determine the statistical significance of any differences in gene expression. β-actin hybridization signals determined by microarray analysis confirmed that there was no significant difference in β-actin expression between wild type and transgenic mice. To determine the effect of the choice of the normalizing gene on the verification of microarray results by QPCR, the expression of four genes was also determined using two alternative normalizing genes, HspB8 and Nme1. These were identified by microarray analysis as having the most consistent levels of expression in spinal motor neurones at each time point, with the lowest coefficient of variation of hybridization signals.

Table 1 Primer pairs used for QPCR experiments
Full size table

Neural VEGF quantitation

Ten 10 μm sections of cervical cord were taken from each mouse used in the microarray study. RNA was extracted using the PicoPure™ RNA isolation kit (Arcturus) and reverse transcribed to cDNA with Superscript II reverse transcriptase (Invitrogen), according to manufacturer's protocols. Neural VEGF expression was assessed by QPCR, as described above, normalized to the expression of ß-actin. Primer sequences and concentrations are shown in Table 1. VEGF expression was compared between VEGFδ/δ and wild-type mice using an unpaired t test.

Isolation of embryonic motor neurons and quantification of axonal outgrowth

Cultures of spinal motor neurons from E12.5 mice (VEGFwt/wt, n = 8; VEGFδ/wt, n = 14; and VEGFδ/δ, n = 4) were prepared by a panning technique using a monoclonal anti-p75NTR antibody (Millipore Bioscience Research Reagents)[20]. The ventrolateral parts of individual lumbar spinal cords were dissected and transferred to HBSS containing 10 mM 2-mercaptoethanol. After treatment with trypsin (0.05%, 10 min), single cell suspensions were generated by titration. The cells were plated on an anti-p75NTR coated culture dish (24 well; Greiner) and left at room temperature for 30 min. The individual wells were subsequently washed with HBSS (three times), and the attaching cells were then isolated from the plate with depolarizing saline (0.8% NaCl, 35 mM KCl, and 1 mM 2-mercaptoethanol) and plated at a density of 1500 cells per well on laminin-coated coverslips in Greiner four-well culture dishes. Motor neurons from single embryos were cultured for 7 days in quadruplicates in the presence of BDNF and CNTF (1 ng/ml each, 2000 cells/well initially). Surviving neurons were counted on day 0, 1 and day 7. Initial counting of plated cells was done when all cells were attached to the culture dish at 4 h after plating. At day 7, wells were fixed with fresh 4% paraformaldehyde in phosphate buffer and subjected to immunocytochemistry. Two coverslips with motor neurons of each embryo were stained with antibodies against MAP2 and P-Tau to distinguish between axons and dendrites. Axon lengths of VEGFwt/wt (n = 202), VEGFδ/wt, (n = 406) and VEGFδ/δ (n = 122) neurones were measured using the Leica Confocal Software (Leica, Germany), and compared using a t test.

Results

Neural VEGF expression

VEGFδ/δ mice used in this study showed a reduction in expression of VEGF mRNA in the cervical spinal cord of 33% compared to their wild type littermates (data not shown). A similar reduction in baseline spinal VEGF protein expression in the spinal cord, although not of mRNA expression, was described by Oosthuyse et al[1].

General features of differential gene expression between VEGFδ/δ and wild type mice

Between 15 and 18% of probe sets at each time point passed the signal filter, and their corresponding genes were considered expressed by motor neurones. Statistical analysis of these genes showed that 324 genes were significantly differentially expressed (at p < 0.05 level) between VEGFδ/δ and wild type mice at 3 months, 382 genes at 5 months, and 689 genes at late stage. The total numbers of genes which are upregulated and downregulated, with three different p values (0.05, 0.01 and 0.001) are given for each time point in Table 2. Prior to symptom onset, the majority of differentially expressed genes were upregulated. At disease onset, this pattern had reversed, and the majority of genes were significantly downregulated. At late stage disease there was marked transcriptional downregulation, shown in Figure 2. Fold changes between VEGFδ/δ mice and their littermates were, in general, smaller than those seen in parallel experimental design using the mutant SOD1 model of ALS [21].

Figure 2
figure 2

Histogram of fold change values of significantly differentially expressed genes at 3 months (green), 5 months (blue) and late stage (red). Upregulated genes are plotted on the positive axis, downregulated genes on the negative axis.

Full size image
Table 2 Numbers of significantly differentially expressed genes at each time point
Full size table

Gene ontology analysis

Gene ontology (GO) terms which were significantly enriched (at level p < 0.05) amongst genes significantly differentially expressed between transgenic and wild type mice were identified for each time point. Over-represented GO terms at each time point are given in Tables 3, 4 and 5. Four GO clusters were identified: Ontology terms concerned with mitochondrial function and energy production, with steroid metabolism, with axonogenesis and nervous system development, and with gene expression and RNA metabolism. There is overlap between GO terms enriched at 3 months and 5 months, and between 5 months and late stage disease, suggesting that the processes represented by these terms are affected in transgenic mice in a sequential manner, as disease progresses.

Table 3 GO terms significantly enriched amongst differentially expressed genes at 3 months
Full size table
Table 4 GO terms significantly enriched amongst differentially expressed genes at 5 months
Full size table
Table 5 GO terms significantly enriched amongst differentially expressed genes at 14 months
Full size table

Mitochondrial function and energy production

In presymptomatic mice, fewer transcriptional changes were seen in transgenic mice, compared to their wild-type littermates than at the other time points of disease. The most significantly over-represented GO terms related to cellular energy production (TCA cycle metabolism, generation of precursor metabolites and energy and carbohydrate metabolism). The differentially expressed genes included several genes encoding TCA cycle enzymes and components of the electron transport chain (Table 6). Three genes encoding TCA cycle enzymes, Mdh2, Ldh2 and Oxct1, are upregulated, while Ldh1, an isoform that favours glycolysis over oxidative metabolism [22], is downregulated. In the electron transport chain, genes that produce components of 4 out of the 5 complexes, and Pcdc8, which is required for mitochondrial oxidative phosphorylation [23], show upregulation (Figure 3). These gene expression changes would be consistent with a small but significant increase in oxidative metabolism in neurones of VEGFδ/δ mice in the early stages of disease. Concomitant with this increase, there was upregulation of the free radical scavenging enzymes, Prdx2 and Sod2.

Figure 3
figure 3

Diagrammatic representation of the components of the mitochondrial electron transport chain, and the constituent proteins encoded by genes significantly upregulated in VEGFδ/δ mice.

Full size image
Table 6 Differentially regulated genes involved in cellular energy production at 3 months
Full size table

Cholesterol metabolism

At disease onset, there was enrichment of genes assigned GO terms relating to steroid metabolism, all of which were downregulated (Table 7). These included five genes that catalyse reactions in the final committed pathway to cholesterol synthesis pathway (Figure 4). Sqle is a rate limiting step in this pathway [24]. Prkaa2 is a catalytic subunit of AMPK, which regulates HMG CoA reductase activity. Ldlr and Sorl1 have similar functions in binding LDL, the major cholesterol carrying lipoprotein of plasma, and transporting it into cells. Hsd17b7 catalyses the synthesis of steroid hormones from cholesterol, while Stard4 promotes transport of cholesterol across the mitochondrial membrane, and stimulates steroidogenesis [25]. Nr3c1 is the glucocorticoid receptor.

Figure 4
figure 4

Schematic representation of the cholesterol biosynthesis pathway, with genes that are differentially regulated in VEGFδ/δ mice at 5 months highlighted in red. Cytochrome B5 reductase is an electron carrier for 5-desaturase and methyl sterol oxidase [119]

Full size image
Table 7 Differentially expressed genes in the category of 'Steroid metabolism' at 5 months
Full size table

Nervous system development

At all stages of disease, there was differential expression of genes involved in nervous system development, with prominent enrichment of GO terms relating to nervous system development in the significant gene lists at both 5 months and 14 months. Genes with functions relating to nervous system development at these time points, the majority of which are downregulated, are shown in Table 8 and Figure 5. Several are identified by more than one probe set or at more than one time point.

Figure 5
figure 5

Schematic representation of genes involved in neuronal migration, neurite outgrowth and formation and maintenance of the neuromuscular junction, which are all downregulated in the VEGF transgenic mouse.

Full size image
Table 8 Differentially expressed genes in the category of 'Nervous system development'
Full size table

There is downregulation in VEGFδ/δ mice of genes that promote neurite outgrowth, such as Nrn1[26], Slitrk1[27], RasGrf1[28] and Serpine2, which is an inhibitor of thrombin [29]. Thrombin causes neurite retraction and neuronal death via its receptor F2r, which is upregulated at 3 and 5 months [30]. Tsc1, which inhibits axon growth, is also upregulated at both time points [31]. Growth factors such and NGF, BDNF, GDNF and TGFß2 promote neuronal differentiation and outgrowth [32, 33], There is downregulation in motor neurones of VEGFδ/δ mice of the BDNF receptor, Ntrk2, and its downstream mediator, Tiam1[34]; the GDNF receptor, Gfra1; several components of the TGFß2 signaling pathway; and Lmtk2, a mediator of NGF signalling [35]. Rufy3 is implicated in the formation of a single axon, to determine neuronal polarity [36]. Apbb2 interacts with amyloid precursor protein, which is required for the maintenance of dendrites and synapses [37]. KLF7 is a transcription factor with a key role in neuronal morphogenesis. Null mutations in Klf7 lead to deficits in neuronal outgrowth and axon guidance [38]. The Rho GTPase pathway controls cytoskeletal reorganization to regulate neurone outgrowth, and the maturation and maintenance of dendritic spines. The Rho GTPase, Rac, is activated by vav3[39]. Cyfip1 interacts with Rac and null mutations lead to defects in axon growth, guidance and branching, and in the organization of the neuromuscular junction [40].

The GO category 'microtubule-based process' is over-represented at 5 months, with downregulation of genes that are assigned this function. Microtubules are prominent elements of the neuronal cytoskeleton, involved in the growth and maintenance of neurites, and along which motor proteins move. Microtubule associated proteins bind to the tubulin substrate that make up microtubules, and regulate their stability [41]. Tubulin ß5 and three microtubule associated proteins, Maptau, Mtap1b and Mtap6, are downregulated in VEGFδ/δ mice, as are two motor proteins, kinesin 2 and Dnclic2, and BICD2 which binds cargoes to dynein. Disruption of the dynein-dynactin pathway through prevention of BICD2 uncoupling causes motor neurone degeneration in mice [42]. Ttl causes post-translational modifications of αtubulin that are essential to neurite extension and normal brain development [43]. Tbce is a tubulin chaperone, required for stabilization of neuronal processes, a mutation of which causes the progressive motor neuronopathy of the pmn mouse [44]. Dystonin is thought to have a role in cytoskeletal cross-linking leading to axonal stability [45].

Differentially expressed genes in the VEGFδ/δ model encode a number of proteins that mediate attractive and repulsive growth cone signals during axonal guidance, including Nrp1 and its ligand sema3A[46] and Nrp2[47]. Two downregulated genes have regulatory functions in the growth cone tip: Dpysl3 and DCAMKII[48, 49].

Cellular adhesion molecules are important in contact-dependent regulation of axonal growth, and in the control of neuronal migration. Fourteen adhesion molecules and related molecules are downregulated in the VEGFδ/δ mouse. Alcam is a member of the immunoglobulin superfamily which has a specific role in the guidance of motor axons and formation of neuromuscular junctions (NMJs) [50]. Both L1 and cd24 interact and cooperate with each other as potent stimulators of neurite outgrowth [51]. Cd24, and Numb, which mediates endocytosis of the cell-adhesion molecule L1 [52] are downregulated in VEGFδ/δ mice. The DCC (deleted in colon cancer) subgroup of the immunoglobulin superfamily are ligands for netrin, with roles in the migration and guidance of axonal growth cones. Igdcc4 is a newly recognized member of this group, which is downregulated at 5 months [53], as is the netrin Ntng1[54]. Both Pip5k1c and Astn1 are required for normal neurone migration [55, 56].

The formation of synapses involves guidance of axonal processes towards target cells, target recognition, followed by recruitment of pre- and post-synaptic elements. The synaptic connections between motor neurones and muscle exhibit both functional and anatomical plasticity after maturation, with changes in synaptic strength, and the formation and retraction of neuronal sprouts from synaptic terminals or Nodes of Ranvier. This enables the neuromuscular system to compensate for growth, changes in muscle use, and damage or disease [57]. Adhesion molecules play a central role in the formation and plasticity of synapses, several of which are downregulated in the VEGFδ/δ mouse, including two protocadherins; cadherin 10 [58, 59]; two members of the neurexin family, Nrxn3 and Cntnap2, which bind postsynaptic neuroligins [60, 61]; and SynCAM, a member of the immunoglobulin superfamily. Rnf6 binds to LIM kinase 1, which regulates actin dynamics and is important in determining synaptic structure [62]. Calcitonin-gene-related peptide, of which both polypeptide chains are downregulated in the VEGFδ/δ mouse, is released from motor neurones to stimulate acetylcholine receptor synthesis by muscle, at the NMJ [63]. Reelin is an extracellular matrix protein with a well-recognized function in neuronal migration. More recently, it has been shown to play a role in synaptic plasticity [64], and in the maturation of synaptic contacts during development, by refinement of NMDA-receptor subunit composition [65]. The activation of downstream signalling pathways of reelin is cholesterol-dependent [66]. Fos and Jun form the heterodimer, AP-1, which plays a central role in controlling development, growth, survival and plasticity of neurones. AP-1 has also been shown to positively regulate synapse strength and number, acting upstream of CREB [67]. Both Fos and CREB are downregulated in the VEGFδ/δ mouse.

Proteins of the canonical wingless signalling pathway participate in the assembly of the NMJ, with crucial components being the wingless co-receptor, arrow, Dishevelled and GSK3ß. GSK3ß functions by regulating the structure of the microtubule cytoskeleton, probably via the microtubule-associated protein, MAP1B. Mutations in this pathway cause aberrant NMJ formation, with reduction in number of synaptic boutons [68]. Dishevelled, GSK3ß and MAP1B are downregulated in the VEGFδ/δ model, as is CXXC4 which regulates the wnt-dishevelled signalling pathway [69]. TGFß also plays role in the development and functioning of synapses [70]. There is downregulation in the VEGFδ/δ mouse of TGFß2; D0H4S114, which regulates TGF signalling [71]; Smad4 and ß-spectrin at disease onset, which associate in response to TGFß, and are required for the assembly of the NMJ [72, 73].

Pre-synaptic proteins

Concomitant with the downregulation of genes involved in the formation and morphological plasticity of synapses, a reduction was seen in the expression of several genes encoding proteins that comprise the pre-synaptic machinery (Table 9), including several genes involved in the fusion of synaptic vesicles, which is mediated by the SNARE complex of VAMP/synaptobrevin, syntaxin and SNAP25. Both VAMP3 and Nap b were downregulated in VEGFδ/δ mice, as was the syntaxin-binding protein, Stxbp6; SytI1, one of a family of calcium-binding proteins that need to be bound to the SNARE complex for pore opening to occur; the SNAP associated protein, Snapap, which enhances association of synaptotagmin with the SNARE complex [74]; Unc13C which binds to and activates syntaxin [75]; gelsolin which disassembles the actin network to liberate synaptic vesicles for release [76]; piccolo, which is a scaffolding protein involved in the organization of synaptic active zones, where synaptic vesicles dock and fuse [77]; and synapsins I and II, which modulate neurotransmitter release, possibly by maintaining a pool of synaptic vesicles near to the active zone [78]. PPI controls activity of ion channels and signal transduction enzymes to determine functional synaptic plasticity, and downregulation of several of its regulatory and catalytic subunits was seen, in addition to Phactr1 and 2, which regulate PPI activity [79].

Table 9 Genes encoding presynaptic proteins, differentially expressed in VEGFδ/δ mice
Full size table

RNA processing

At late stage disease, GO terms related to gene expression and its regulation are enriched amongst significantly differentially expressed genes, including the sub-categories of mRNA metabolism and processing, and RNA splicing (Table 10). Of the 651 named genes differentially expressed at 14 months, 143 (22%) have functions relating to gene expression, and of these, 132 (92%) are downregulated. This finding is interesting in light of the massive transcriptional downregulation seen at late stage disease.

Table 10 Differentially expressed genes in the category 'Gene expression'
Full size table

Among the changes seen in this functionally category of genes, was downregulation of a number of splicing factors, and other genes involved in mRNA processing. Several ribosomal components also showed reduced expression. The cAMP responsive element binding protein, CREB1 is a stimulus-inducible transcription factor which mediates nuclear responses underlying the development, function and plasticity of the nervous system [80]. Downregulation of several members of the CREB family is seen in VEGFδ/δ mice-CREB1, CREBBP, CREBzf, its interacting partner, ATF4 (CREB2), and Cri2. Inactivation of neuronal CREB1 has been shown to cause defects in neurone migration, similar to those observed with reelin [81], and to stimulate neurodegeneration [82]. There is also downregulation of the androgen receptor, and related nuclear receptors, Nr1d2, Nr2c2 and the oestrogen receptor, Esrrg. Steroid hormones have wide ranging effects on the structure and function of the nervous system. Androgen receptor function is known to be important for motor neurone survival, and disruption of this function in Kennedy's disease contributes to the motor neurone degeneration seen in this condition [83].

Verification of microarray results by QPCR

The results of QPCR verification for the 15 representative genes chosen is shown in Table 11. Eight genes showed a significant downregulation of expression in VEGFδ/δ mice, which supported the microarray findings. A further 4 showed a trend towards downregulation which failed to reach significance, while 4 showed no change in expression despite a finding of significant expression by microarray analysis. The majority of microarray publications have indicated that arrays and QPCR analysis usually support each other qualitatively. However, it is well recognized that significant quantitative differences occur between microarray and QPCR data [84]. This may be related to gene specific variation in the hybridization kinetics associated with the two technologies, low fold changes or hybridization signals in the microarray experiment, or lack of transcript concordance between the probes used for microarray and QPCR analysis. The proportion of genes validated by this study using QPCR, with the larger sample sizes at 3 and 5 months, is comparable to that found by other studies [85, 86]. Although not all changes seen on microarray were validated by QPCR, it has been argued that where the focus of microarray analysis is the overall pattern of gene expression rather than the response of a few genes, as in this study, there is less utility in confirming the expression differences of individual genes [87].

Table 11 Comparison of QPCR and microarray data of the 15 genes selected for verification
Full size table

Reduced axon growth of VEGFδ/δ motor neurones grown in vitro

Downregulation of genes involved in nervous system development and axonogenesis was observed in motor neurones of VEGFδ/δ mice. A functional correlation of this finding was demonstrated by the growth of primary motor neurones from VEGFwt/wt, VEGFδ/wtand VEGFδ/δ mice in vitro. Survival of primary motor neurone cultures grown under basal conditions for 7 days was unaffected by deletion of the hypoxia response element of the VEGF gene (data not shown). A significant reduction in the length of axons was observed in cultures of motor neurones homozygous for the HRE deletion, compared to wild-type motor neurones (472 ± 26 μm vs 562 ± 31 μm, p = 0.047; figure 6a). The outgrowth of dendrites was unaffected (p = 0.34; figure 6b).

Figure 6
figure 6

The average length of the longest axonal process observed in primary motor neuron cultures isolated from E12.5VEGFδ/δ (n = 4), VEGFδ/wt (n = 14) and VEGFwt/wt (n = 8) mice (Figure 6a). These motor neurons were grown for 7 days on laminin with BDNF and CNTF. The average length of dendrites isolated from the same mice is also shown (Figure 6b). Asterisk indicates p < 0.05.

Full size image

Discussion

Upregulation of oxidative phosphorylation in presymptomatic mice

A small but significant upregulation of genes involved in the TCA cycle and oxidative phosphorylation was observed in VEGFδ/δ mice in presymptomatic stages. This finding is not in keeping with the 'chronic ischaemia' hypothesis of neurodegeneration in the VEGFδ/δ mouse model: were motor neurones in a state of chronic oxygen deprivation due to reduced baseline neural blood flow, oxidative phosphorylation would have been downregulated. A similar upregulation of oxidative phosphorylation was seen in motor neurones in the SOD1 mouse model of ALS [21], and this may be a non-specific adaptation to cellular stress in the early stages of disease.

Transcriptional downregulation and changes in the regulation of gene expression

One of the most marked changes observed in the VEGFδ/δ mouse model was transcriptional downregulation with disease progression, which was accompanied in late stage disease by a reduction in expression of genes involved in the regulation of gene expression, particularly mRNA processing. Similar transcriptional downregulation was observed in a study of the SOD1 mouse model of ALS [21], and in the ageing human brain [88]. The cause of this transcriptional downregulation is not known, although several mechanisms may be proposed. Firstly, it may be due to low level oxidative modification of nuclear DNA, which has been shown to accompany downregulation of gene expression in the ageing human brain [88]. Secondly, reduction in expression of VEGF could result in reduced induction of sequence-specific transcription factors, such as Fos, which was robustly downregulated in this study, with consequent downregulation of genes regulated by those factors [89, 90]. Thirdly, epigenetic modifications which cause suppression of gene expression, such as DNA methylation, increase with age and are accelerated in neurodegenerative diseases [91].

Downregulation of cholesterol metabolism at disease onset

Five genes in the final committed pathway to cholesterol synthesis, and two receptors which bind cholesterol and transport it into the cells, were downregulated in the VEGFδ/δ mouse. Neuronal cholesterol is either synthesised by neurones or produced by astrocytes, bound to apolipoprotein E (APOE), and taken up by neurones via the low density lipoprotein receptor[92]. Cholesterol-rich lipid rafts in the growth cone are required for downstream signaling of adhesion molecules and guidance receptors during axon growth and guidance. Cholesterol stimulates the formation of synapses, and has important roles in synaptic function, and the release of neurotransmitters [93–96]. A recent study showed that the neurotrophic factor, BDNF, promotes synaptic development in cortical neurones via the stimulation of cholesterol biosynthesis [97], and the findings of this study suggest that VEGF may have a similar function. Cholesterol homeostasis is already implicated in the pathogenesis of neurodegenerative disease. There is a strong association between Alzheimer's disease and the APOE4 allele, which has reduced efficacy in cholesterol delivery to cells, and in stimulating neurite outgrowth [98]. Downregulation in the expression of genes involved in the cholesterol synthesis pathway is seen in cortical tissue of patients with Huntington's disease, and in mouse and cell models of the disease [99, 100].

Neurite outgrowth and synaptogenesis

In adult mice, reduction in neural expression of VEGF, through deletion of its hypoxia-response element, causes downregulation of genes which are known to play a role in the growth of neuronal processes and formation of synapses during embryonic development. These include promoters of outgrowth of neuronal processes, components of neurotrophic signalling pathways, cell adhesion molecules, and the cholesterol synthesis pathway. The most significantly enriched functional gene categories at 5 months relate to axon extension and axon maintenance. This finding was supported by in vitro data: primary motor neurones from VEGFδ/δ mice show a significant reduction in axon outgrowth compared to their wild-type littermates. This effect of VEGF was specific to that neuronal compartment, with no effect on dendrite growth, or on cell survival. Exogenous VEGF protein stimulates axon outgrowth in cultures of primary motor neurones [20] although to a lesser degree than exogenous BDNF and CNTF. Despite addition of BDNF and CNTF to the primary neurone cultures in this study, axon outgrowth from VEGFδ/δ motor neurones was significantly reduced, indicating that the effect of VEGF on axonal outgrowth cannot be substituted for by the presence of other growth factors.

Differential expression of genes relating to the growth of neuronal processes, and formation of synapses was an unexpected finding in the adult animal, but there is growing evidence that changes involved in the plasticity of the nervous system and in the maintenance of neuronal networks in the adult animal recapitulate, to some degree, those underlying the formation of neuronal networks during development: cell adhesion molecules that mediate target recognition and synapse formation during embryogenesis also lead to changes in synaptic efficacy in the adult nervous system [101]; BMP signalling at the Drosophila NMJ is not only required for normal synaptic growth, but also for synaptic stabilization, via LIM kinase-1, and disruption of this pathway leads to synaptic disassembly and retraction [102]; the wnt signalling pathway maintains activity-dependent axon stability in adult olfactory neurones [103]; neurotrophins were first recognized as target-dependent survival factors for developing neurones during embryogenesis, but they have also been shown to promote synaptic stability and maintain neuronal processes in response to mechanical axonal injury [104].

Although the NMJ is considered a relatively stable structure, plasticity of the motor system is seen in response to changing physiological demands and to pathological conditions. Alterations in synaptic structure and function occur in the motor cortex, spinal cord and NMJ in response to exercise [105–107], while partial denervation or paralysis results in sprouting and reinnervation from adjacent nerve fibres [57]. The observation of fibre-type grouping in ageing muscle indicates that denervation and reinnervation of muscle fibres occurs with normal ageing [108]. Motor units differ in the plasticity of their synapses. Synapses formed on Fast Synapsing (FaSyn) muscle, or by fast-fatiguable (FF)-type motor units exhibit relatively little synaptic plasticity, compared to Delayed Synapsing (DeSyn) or slow (S)-type motor units. Motor units with less synaptic plasticity exhibit early susceptibility to loss in motor neurone degenerative disease, and in the SOD1 mouse model of ALS, a progressive impairment of stimulus-induced synaptic sprouting was observed over the course of the disease, suggesting that the absence of synaptic plasticity, and disease-induced synaptic loss are mechanistically linked [109, 110].

The transcriptional changes observed in VEGFδ/δ mice would be predicted to result in a reduced capacity for the morphological adaptations that occur during plasticity of the motor unit, resulting in increased vulnerability to degeneration. As neuronal processes and synapses are required continuously to retract and reform during the process of synaptic plasticity, downregulation of genes involved in neurite growth and synapse formation would be likely to result in gradual attrition of synapses and distal cellular processes. Synapse retraction causes loss of access of the neurone to trophic signals from target tissue. A recent study has shown that muscle hypermetabolism is sufficient to cause degeneration of NMJs, and subsequent loss of spinal cord motor neurones [111]. Although ALS is characterized pathologically by loss of motor neurone somata, and gliosis in the anterior horn of the spinal cord, these post mortem findings may not reflect changes occurring earlier in the disease. There is evidence in both human ALS and in mouse models, that motor neurone death occurs by the type of 'dying back' axonal degenerative pathophysiology, that would be predicted in this model, with defects of neuromuscular transmission, end-plate denervation, and 'peripheral pruning' of axons as the earliest observed events [112–114]. VEGFδ/δ mice similarly show relative preservation of spinal motor neurones, at stages of disease where there is marked axonal loss [1].

The relevance of this finding in the VEGFδ/δ mouse model of ALS to the human disease is unknown, but failure of synaptic plasticity has been proposed as a pathogenic mechanism in Alzheimer's disease [115]. A failure to maintain neuronal processes and synapses in the face of increasing demands of neuronal plasticity would explain two epidemiological observations in ALS: the associations between exposure to vigorous physical activity [116, 117] or skeletal fractures [118], and the risk of developing the disease.

An alternative hypothesis is that the pathogenic insult in VEGFδ/δ mice occurs during their development, when a lack of VEGF may lead to aberrant neuronal guidance and formation of neural networks. Subtle defects in neuronal positioning and in synaptic circuitry in VEGFδ/δ mice could result in reduced stability of synaptic connections, or a reduction in target-derived neurotrophic support, with consequent increased loss of motor neurones during ageing. This hypothesis is supported by the observation that embryonic motor neurones show a reduction in axonal outgrowth in vitro.

Conclusions

Reduction in expression of VEGF, through deletion of a regulatory promoter region of the gene, results in adult-onset motor neurone degeneration that resembles human ALS. We have presented evidence here that this phenotype is accompanied by reduction in expression, from symptom onset, of the cholesterol synthesis pathway, and genes involved in nervous system development, including axonogenesis, synapse formation, growth factor signalling pathways, cell adhesion and microtubule-based processes. These findings raise the possibility that VEGF is required for the maintenance of distal neuronal processes in the adult animal, perhaps through promotion of remodelling of distal processes and synapses in the face of the demands of neuronal plasticity. A reduction in VEGF expression in VEGFδ/δ mice may lead to failure of the maintenance of neuronal circuitry, causing axonal retraction and cell death.