Axons are sufficient and necessary for transcriptional activation of astroglial GLT1
A microfluidic culture platform(Park et al., 2006
) (MCP, ) was used to develop a novel neuron and astrocyte co-culture system in which only axons penetrate and establish contact with astrocytes, thereby providing a system to study the role of neuronal processes on GLT1 activation at a single axon/astrocyte resolution. High-density neurons (1–3 ×106
/ml) were first plated on the right side of the MCP. GDNF (10ng/ml) was added to the left side of the MCP to stimulate axon outgrowth 24h after seeding of neurons. Extensive neuronal processes grew through channels connecting both sides of MCP 3–5 days after adding GDNF (). A representative magnified image of axon and astrocyte is shown in . More than 90% of neuronal processes growing through the channels are axons, as confirmed by unique growth cone morphology, immunoreactivity for the growth cone marker 2G13 (), and for synapsin-I (Supplementary Figure 1A
) and immunonegative for MAP2 (data not shown). The solution exchange between both sides of the MCP is minimal(Park et al., 2006
), as no DiI and DiO double-labeled astrocytes were observed in each side of the MCP (Supplementary Figure 1B
) nor was there leakage of 3
H-glutamate between chambers (data not shown). To establish neuron astrocyte co-culture in the MCP, GDNF was removed when neuronal processes were still inside the connecting channel, and astrocytes were plated into the left chamber. As shown in , axon bundles, visualized by βIII-tubulin staining, passed through the channel, entering the astrocyte side of MCP. Intensity of GLT1 immunostaining (10990±791, arbitrary units) in astrocytes with the presence of axons was significantly increased 230% () compared to that in astrocytes alone chamber (4781±385), indicating that axons are sufficient to induce GLT1 activation. Axonal activation of GLT1 expression is apparently axon number/length dependent, as the intensity of GLT1 staining (12350±732) in platforms with more axons (indicated by the total length, 7500–18889μm/mm2
) is significantly higher than that (7409±872) in platforms with fewer axons (3700–7500μm/mm2
) (). In this model paradigm, the vast majority of GLT1 activation was only seen with astrocytes contacted by axons (white arrows in ). Highest GLT1 intensity was found in all astrocytes with axon contact. Only occasional astrocytes, with no obvious direct axon/terminal contacts, were also GLT1 activated (yellow arrow in ). Subsequent treatment of cultured astrocytes with differential neuronal components, i.e., neuronal supernatant (NS), neuronal membrane (NM), or neuron conditional medium (NCM) were all sufficient to increase GLT1 expression, to a similar level (Supplementary Figure 1C, 1D
). These results would suggest that presynaptic interactions with astroglia via direct contact might activate transcriptional pathways in astroglia.
Axon-dependent transcriptional activation of GLT1 promoter and GLT1 protein expression on a microfluidic culture platform (MCP)
To study molecular modulation of astroglial synaptic-relevant pathways, we utilized BAC GLT1 eGFP promoter reporter mice(Regan et al., 2007
), which allow spatial and temporal in situ monitoring of single astrocyte GLT1 promoter activity by fluorescence reporter intensity. The expression of the reporter correlates with endogenous GLT1 promoter activation, protein expression and functional activity(Regan et al., 2007
). To monitor dynamic changes of the GLT1 promoter to axon stimulation, astrocytes derived from this mouse were added on the left side of MCP after culturing neurons for 3–5d, prior to entry of axons into the astroglial chamber. Time-lapse images were collected from 24h to 112h after co-culture (). Intensity of eGFP fluorescence, the indicator of GLT1 genomic promoter activity, in single astrocytes, was detected and represented by pseudocolor (). In the absence of axons, no GLT1 appreciable gene activation was observed. Dramatic increase of eGFP intensity was observed (from 2846±682 to 10860±941) within 48 hrs after axons approach and/or make contact with the astroglia (24h–68hr, ). The dependence of the axon-astrocyte interaction was validated by the decrease in astrocytic eGFP expression (68–112hr), following kainate induced (200μM; added to neuronal side) neural injury and subsequent axonal degeneration (). In contrast, eGFP intensity from MCP without application of kainite only decreased slightly from 68h to 112h (). Axon-dependent eGFP intensity changes, quantified from these time-lapse images, clearly indicate that axons induced transcriptional activation of GLT1. GLT1 mRNA levels were also increased after directly adding neurons to astrocyte cultures (Supplementary Figure 1E
), providing additional evidence that synaptic interaction with astroglia induces transcription activation of GLT1.
We further investigated whether glutamate receptor (GluR)-mediated neuron-astroglial signaling is involved in neuron-dependent GLT1 activation. P8 rat spinal cord organotypic slice cultures, a system faithfully mimicking the in vivo
environment (Rothstein et al., 1993
), were used to investigate altered neuron-astrocyte communication on GLT1 activation. As spinal cord slices are usually 350μm in thickness, relatively higher concentrations of pharmacological inhibitors were used in experiments. GLT1 expression levels in slice cultures were reduced in a dose-dependent manner following tetrodotoxin (TTX) treatment (), suggesting that Na+
channel mediated synaptic transmission facilitate GLT1 activation in astrocytes. In addition, treatment of slice cultures with ionotropic or metabotropic GluR antagonist cocktails also resulted in a dose-dependent decrease of GLT1 expression (). Inhibition of astroglial iGluRs in astrocytes in MCP also leads to decrease of axon-induced GLT1 expression (Supplementary Figure 2A, 2B
). Subsequent treatment of slices with individual glutamate receptor antagonists revealed that AMPA (antagonist CNQX) and mGluR 1/5 (antagonist m196) are involved in the neuron-dependent GLT1 activation (). NMDA antagonist MK801 had little effect, although high dose APV was inhibitory (). Overall, mGluR1/5 antagonists alone or in combination appeared to have the greatest effect on GLT1 expression in the slice culture paradigm (). GluRs are expressed on pre- and post-synaptic neurons as well as astrocytes, thus, inhibition of GluRs results in inhibition of synaptic transmission from neuron-to-neuron and neuron-to-astrocyte. The block of GLT1 activation by inhibition of neurotransmitter release (TTX) and by the GluR receptor antagonists suggests that synaptic activity is a strong component of pre-synaptic-based activation of astroglial GLT1.
Glutamate receptors are involved in neuron-dependent GLT1 expression
Previous in vitro preparations from mixed neuron-glial cultures had suggested that neuron conditioned medium can directly activate the expression of GLT1(Gegelashvili et al., 1997
) (Schlag et al., 1998
; Swanson et al., 1997
), and that soluble factors may be secreted from neurons to stimulate the expression of astroglial glutamate transporter. Glutamate has long been speculated as one of the soluble factors that increase GLT1 expression, but direct exposure of high concentration of glutamate in vitro to astrocyte cultures fails to increase GLT1 expression, but does alter its cytoplasmic clustering (Nakagawa et al., 2008
; Zhou and Sutherland, 2004
). In fact, elevated levels of glutamate are closely associated with loss of EAAT2 in chronic neurological injuries including ALS, Huntington’s disease, and multiple sclerosis, suggesting that other released factors or altered membrane contact may play more important roles in mediating axon-induced GLT1 activation. For example, axon membrane contact can be a potent regulator of developmental astroglial Notch signaling(Eiraku et al., 2005
). By using this novel isolated neuron astrocyte co-culture system, our results prove that astroglial activation is neuron dependant and that presynaptic interaction with astroglia including both secretion and direct membrane contact, transcriptionally activate astroglial GLT1.
Recruitment of kappa-B motif-binding phosphoprotein (KBBP) to the promoter is required for GLT1 transcriptional activation
To evaluate how neurons regulate astroglial GLT1/EAAT2, we cloned a 2.5kb upstream promoter of human EAAT2 that shares high sequence homology (>70%) with GLT1 promoter from multiple species (Rothstein et al., 2005
). To test whether this 2.5kb promoter behaves similarly as full-length (15kb) GLT1 genomic promoter in rodent astrocytes, a DsRed EAAT2 promoter reporter (2.5kb) () was transfected into astrocyte cultures derived from BAC GLT1 eGFP transgenic mice (Regan et al., 2007
). After transfection, newly prepared cortical neurons were plated on transfected astrocytes. As shown in , the presence of neurons markedly induced the expression of both DsRed and eGFP reporters in the same astrocyte, indicating that the 2.5kb promoter is functional in a rodent astrocyte environment and contains conserved cis-elements sufficient for neuron-derived activation signals.
Recruitment of Kappa B-motif binding phosphoprotein (KBBP) to GLT1 promoter is required for its activation
To identify the major promoter regions responsible for neuron-dependent activation, luciferase promoter reporters were generated by serial deletion of the 2.5kb sequence and electroporated into P2 mouse primary astrocyte cultures. Sequence deletion from −921 to −279 (pGL958 vs. pGL316 in ) sharply reduced both basal and neuron-dependent promoter activity (), with the most significant drop of activity from pGL958 to pGL557, indicating that the sequence between −921 to −520 contains primary elements for neuron-dependent GLT1 promoter activation. Multiple highly conserved transcription factor binding sites are present in the region from −921 to −520 by sequence analysis. Site-directed mutagenesis of pGL958 generated multiple mutant luciferase promoter reporters, which were examined in astrocyte cultures (). The 958bp-mutant 2 exhibited the greatest reduction in luciferase activity without altering co-transfected β-galactosidase activity, suggesting that the sequence mutated from −688 to −679 GGGTGGGTGT is essential for GLT1 promoter activity. Sequence alignment across ten species characterized GGGTGGGTGT as an evolutionally conserved site among mammals (supplementary Figure 3A
). Notably, mutations of putative NF- B binding sites within this promoter region had no dramatic influence on neuron stimulated promoter activation (not shown).
The requirement of nucleotides −688 to −679 for GLT1 promoter activity was further tested in vivo
. GLT1 mRNA transcription is strongly induced during early postnatal development in rodent brain (Schmitt et al., 1996
; Sutherland et al., 1996
), providing an in vivo
model to test the GLT1 promoter activation. A DsRed reporter driven by the wild type 958bp or 958bp-mutant 2 EAAT2 promoters was delivered to brain of young mice using an adeno-associated virus (AAV) vector to examine in vivo reporter activation. AAV particles carrying the EAAT2 promoter-DsRed expression cassette were injected into the lateral cerebral ventricle of P0 pups of BAC GLT1 eGFP transgenic mice (Broekman et al., 2006
). As shown in , confocal analysis of coronal brain sections of mice, 3 weeks after injection with 958bp-WT-DsRed containing AAV revealed expression of the DsRed reporter in astrocytes that also express eGFP reporter. However, no expression of DsRed was found from brain sections of mice injected with 958bp-mutant 2-DsRed containing AAV. Expression of the DsRed was also seen in nearby neurons for both AAV viruses (data not shown), suggesting that both viruses are effective in driving the DsRed reporter expression in vivo. The inability of 958bp-mutant 2 promoter to drive DsRed reporter expression in astrocytes in vivo confirmed the in vitro results that the sequence from −688 to −679, GGGTGGGTGT, is essential for neuron-dependant EAAT2 promoter activity.
The nuclear factors recruited to the GGGTGGGTGT sequence, were identified by performing a gel shift assay with nuclear extracts prepared from mouse cortex at developmental time points prior to GLT1 promoter activation (P2) and after strong promoter activation (P21). Wild type and mutant oligos (45mer) that only differ in the GGGTGGGTGT were synthesized and labeled with biotin. As shown in , increased specific binding with labeled WT oligo was found from P2 to P21 mice when the GLT1 promoter was strongly activated. The addition of extra-unlabeled WT oligo completely abolished the binding, suggesting that the binding is specific to this oligo. On the other hand, no specific binding with labeled mutant (MT) oligo was found, indicating the change of core sequence from GGGTGGGTGT to AAATGCCACT abolished the recruitment of specific nuclear proteins to this cis-element during GLT1 promoter activation. In addition, no other binding with labeled MT oligo was observed, suggesting that the replacement of GGGTGGGTGT with AAATGCCACT does not introduce non-specific binding with other nuclear proteins. This direct biochemical evidence of gel shift result, along with the genetic analysis in , demonstrates that the recruitment of nuclear factors to GGGTGGGTGT is essential for neuron-stimulated GLT1 promoter activity.
To identify the actual transcription factor binding to GLT1 regulatory site GGGTGGGTGT, nuclear extracts were affinity purified with the same oligos used in gel shift analysis. After purification, proteins bound to the WT or MT oligos were resolved on 4–12% PAGE gel and visualized by silver staining. A specific band with molecular weight between 50 to 75KDa was only found with labeled WT oligo but not with MT oligo (), confirming the results from gel shift analysis. Subsequent trypsin digestion and LC/MS/MS analysis of digested peptides () revealed a nuclear protein, kappa-B motif-binding phosphoprotein (KBBP, gi|1083569), specifically binds to WT oligo.
KBBP was first identified in a T lymphoma cell line that binds to kappa B enhancer element (GGGGACTTTCC)(Ostrowski et al., 1994
). It is almost 100% homologous with human heterogeneous nuclear ribonucleoprotein K (hnRNP K) protein (Ostrowski et al., 1994
). HnRNP K protein plays diverse roles in transcription, RNA splicing, and signal transduction through its interaction with many protein partners, including protein kinases, transcriptional factors and DNA/RNA(Bomsztyk et al., 2004
). Notably, although the original kappa B enhancer element GGGGACTTTCC share partial homology with the KBBP binding sequence GGGTGGGTGT on GLT1 promoter, it is sufficient to compete out GGGTGGGTGT induced specific binding with KBBP in a dose dependent manner (supplementary Figure 3B
). In addition, further mutation of GGG
to AAA on KBBP binding sequence GGG
TGGGTGT also abolished GLT1 promoter activity (supplementary Figure 3C
), suggesting that these nucleotides are critical for KBBP binding and GLT1 promoter activity. As the mouse homolog of human hnRNP K, KBBP protein may also participate in transcription or RNA splicing. Expression of KBBP in astrocytes in vivo was first examined by using an antibody against human hnRNP K. Cortical immunoreactivity for KBBP in both P2 and P21 BAC GLT1 eGFP mice shows neuronal and astroglial immunolocalization. Notably, KBBP expression was barely detectable in P2 astrocytes and dramatically increased in astroglia from P2 to P21, which is identical to the very large increase in astroglial GLT1 promoter activity (indicated by eGFP reporter intensity) in the same astroglia (). The significant correlation between increased expression of astroglial KBBP and GLT1 promoter/protein () during early postnatal development suggests that KBBP may play an important role in transcriptional regulation of GLT1. Interestingly, KBBP expression in cultured astrocytes was also induced by neurons, coupled to the increase of GLT1 (); the coupled increases of KBBP and GLT1 are consistent with the same changes in astrocytes observed in vivo.
KBBP is essential and sufficient for neuron-dependent transactivation of GLT1
Small interference RNA (siRNA) specifically against KBBP was developed that effectively reduced KBBP expression level to 10% of that in control () (Moumen et al., 2005
), which results in significantly reduced GLT1 expression in astrocytes on neuron-astrocyte co-culture MCP (). We further tested whether a decrease of KBBP expression in astrocytes also reduces GLT1 promoter activity in vivo. To selectively silence astroglial KBBP, AAV particles that carried an expression cassette with the KBBP gene in the antisense orientation driven by the GFAP promoter, along with AAV particles that carried a DsRed reporter expression cassette driven by the same promoter, were intraventricularly injected into new born (P0) BAC GLT1 eGFP transgenic mice. AAV particles carrying luciferase in the antisense orientation served as a control. KBBP antisense was highly effective at silencing KBBP expression in cultured astrocytes (Supplementary Figure 4C
). As shown in , KBBP antisense, but not luciferase antisense, effectively silenced KBBP expression in vivo. GLT1 promoter activity in KBBP antisense expressing astrocytes, identified by DsRed, was significantly reduced as indicated by reduced eGFP intensity (, shown by yellow arrow) compared to that in untransduced astrocytes (shown by white arrow) while GLT1 promoter activity in luciferase expressing astrocytes was unaltered (, shown in yellow arrow). In aggregate, the loss-of-function analysis of KBBP by RNAi and antisense in astrocytes, along with the promoter mutagenesis analysis, indicate that recruitment of KBBP to the GLT1 promoter is required for GLT1 transcriptional activation in vivo.
We next determined if over-expression of KBBP could concomitantly alter GLT1 expression in astrocyte cultures. KBBP and luciferase overexpression constructs were prepared and electroporated into cultured astrocytes. As shown in , overexpression of KBBP, but not luciferase, was sufficient to increase GLT1 expression in the presence of neurons, though overexpression of KBBP itself was not sufficient to induce the GLT1 expression. In addition, direct treatment of cultured astrocytes with DHPG (50μM), the selective agonist for mGluR 1/5, increases both GLT1 and KBBP mRNA levels (supplementary Figure 4A, 4B
), indicating that mGluR 1/5 may play more dominant role in neuron-depednent GLT1 up-regulation. Although direct treatment of physiological GluR agonist glutamate failed to increase the GLT1 expression, this could be due to the quick uptake of extracellular glutamate by GLAST in cultured astorcytes. These results are consistent with our observations that astroglial KBBP and GLT1 are expressed in parallel in vivo during early post-natal and synaptic maturation. This hints that the maturation of synapse might influence astroglial gene activation and expression of GLT1. Conversely, an alteration of this pathway due to the loss of pre-synaptic influence of neurons on astrocytes could be an important event linking neural injury to astroglial synaptic function.
Corticospinal Tract Transection
Corticospinal tracts originating from glutamatergic upper motor neurons possess excitatory presynaptic terminals on spinal cord interneurons and lower motor neurons that are modulated by perisynaptic astroglial GluTs. This pathway allows direct in vivo assessment of astroglial transporter expression in response to the loss of axonal signals induced by spinal cord transection. Thoracic cord (segment 9) transection was performed three weeks after mice received bilateral injection of Fluororuby (FR) into motor cortex to label corticospinal tracts by anterograde FR transport. One week after transection, thin horizontal sections of spinal cord were prepared and FR signals were examined to assess axon degeneration. Above the surgical lesion, abundant FR labeled axons, typical for corticospinal tract innervation, were found bilaterally in spinal cord white matter, projecting to grey matter neurons at each segment (). FR labeled neurons were also observed in grey matter (). In contrast, FR signals (including labeled axons and grey matter neurons) were reduced in the cord well below the lesion site in lumbar spinal cord (). Quantitative analysis revealed that more than 70% of FR labeled corticospinal axons in lumbar cord were depleted as early as 7 days following spinal transection (). GLT1 protein expression in lumbar cord of lesioned mice was markedly reduced compared to that of control mice (). A pronounced loss of GLT1 immunostaining was also observed in coronal sections of lesioned lumbar cords from BAC GLT1 eGFP mice. GLT1 promoter activity, as reflected by eGFP fluorescence intensity in coronal sections of lumbar cord of lesioned BAC GLT1 eGFP mice, was focally reduced () in single astrocytes. These sharp astroglial changes were localized to regions with diminished synaptic terminals, as identified by the loss of pre-synaptic terminal marker vesicle glutamate transporter 1 (VGluT1), VGluT2 (supplementary Figure 5A
) and preservation of post-synaptic marker PSD95.
Corticospinal tract presynaptic degeneration induced by spinal cord transection results in loss of GLT1 protein and promoter activity
Overall, the loss of KBBP in individual astroglia (, yellow arrow) correlated with a decrease in GLT1 promoter-reporter eGFP intensity in same cells in lesioned lumbar cord (), along with loss of GLT1 protein (). The astroglial changes all followed the loss of presynaptic terminals. These in vivo studies demonstrate that presynaptic terminals act to maintain astroglial KBBP expression, which in turn activates GLT1 expression. Furthermore, these data provide a pathway by which abnormal axon/pre-synaptic signaling can alter astroglial synaptic integrity.
G93A SOD1 ALS Mouse Neurodegeneration
Severe loss of GLT1 protein has been observed in SOD1 G93A transgenic rodents (Howland et al., 2002
), a model of familial ALS, and in the R6/2 mice model of Huntington’s disease (Lievens et al., 2001
). The mechanism for this loss in these animal models remains unexplained. Total spinal cord EAAT2 mRNA levels were not found abnormal in prior ALS studies(Bristol and Rothstein, 1996
). However, these past studies were technically limited due to post mortem tissue artifacts and the use of low-resolution, bulk tissue homogenization and non-in situ approaches-all which limit the ability to detect focal astroglial abnormalities known to occur in human and rodent ALS (Howland et al., 2002
; Rothstein et al., 1994a
). Similar to the acute corticospinal lesion experiments and ricin toxin lesion, the loss of GLT1 protein in rodent models of ALS is focal to the region of degenerating motor neurons and spreads along the lumbar spinal cord. To better detect focal GLT1 mRNA alterations in astroglia, in situ hybridization of GLT1a and GLT1b, the two major GLT1 transcripts was performed in lumbar spinal cord of SOD1 G93A transgenic rat (). GLT1a mRNA is far more abundant than GLT1b, which is consistent with previous studies(Berger et al., 2005
). There was a pronounced focal loss of both GLT1a (70% loss compared to WT, ) and GLT1b (50% loss compared to WT, ) mRNA transcripts in the ventral horn of the SOD1 G93A mouse spinal cord, similar to reports of GLT1 protein(Howland et al., 2002
). In addition, quantitative real time PCR analysis of GLT1 mRNA from lumbar spinal cord of SOD1 G93A mice at different stage of disease progression (60, 90, 120d) also showed that GLT1 mRNA was reduced to 65% of control at disease onset and further decreased to 40% of control as disease progressed to end stage (). In contrast, GFAP mRNA levels increased gradually as disease progressed to end-stage (). To better evaluate alterations of GLT1 regulation at the single cell level, the SOD1 G93A mice were mated to the GLT1 BAC eGFP reporter mice. The presence of the BAC-eGFP promoter reporter did not change the course of disease, with mice developing disease at approximately 90 days of age and reaching end stage at ~125 days of age (n=20). BAC GLT1 eGFPxSOD1 G93A mice were sacrificed at various time points and eGFP expression, as a reflection of GLT1 promoter activity(Regan et al., 2007
), was examined and quantified by confocal microscopy of serial thin sections from lumbar cord (). A profound decrease of astroglial GLT1 promoter activity, indicated by a more than 60% loss of eGFP fluorescence intensity, was found in ventral gray matter of BAC GLT1 eGFPxSOD1 G93A mice, especially in ventral gray matter-localized astroglia near motor neurons from end stage (127d) animals (). The loss of GLT1 promoter activity/eGFP fluorescence was the result of transcriptional dysfunction in these spinal cord astroglia, rather than cell death as increased GFAP immunostaining signal was observed from astrocytes with decreased eGFP fluorescence (), as well as increased GFAP mRNA levels in lumbar spinal cord from end-stage SOD1 G93A mice (). Notably, there is a profound loss of pre-synaptic terminals in this transgenic model as a result of the loss of pre-synaptic corticospinal tract terminals, spinal interneurons and recurrent collaterals from motor neurons, as reflected by a decrease in synaptophysin expression (Supplementary Figure 4A
)(Morrison et al., 1998
GLT1 transcriptional dysfunction contributes to the loss of GLT1 protein in SOD1 G93A rodents
To determine if there was a relationship between neural injury/synaptic loss and the alteration in regulation of astroglial synaptic control in this disease model, the expression level of astroglial KBBP in ventral lumbar cord of BAC GLT1 eGFPxSOD1 G93A mice (n=5) was also examined. As shown in , KBBP immunostaining and GLT1 promoter activity/eGFP reporter fluorescence was diminished in astrocytes from ventral lumbar cord of SOD1 G93A mice. The intensity of eGFP fluorescence and KBBP immunostaining in individual astrocytes (n≥25) were highly correlated (). Decreased KBBP expression levels (<20000, arbitrary units) were correlated with lower levels of eGFP fluorescence intensity (<25000, arbitrary units), which was also found in the corticospinal transection and ricin models described above. GLT1 transcriptional activation, KBBP immunoreactivity and eGFP expression levels were unaltered in dorsal horn astrocytes from the BAC GLT1 eGFPxSOD1 G93A mice (data not shown).
In summary, KBBP is a neuron-dependant downstream nuclear factor that regulates astroglial GLT1 transcription as determined by: 1) in vitro and in vivo genetic analysis of KBBP modulating GLT1 expression, 2) pharmacologic and synaptic activity regulated expression of GLT1 transcription, 3) correlation of developmental KBBP and GLT1 expression levels in astroglia, and 4) correlation of altered in vivo pre-synaptic inputs to astroglial KBBP and GLT1 promoter activation by employing models of synaptic denervation including corticospinal lesion, single neuron toxin lesion and an animal model of ALS. These studies provide a molecular mechanism for abnormal astroglia in neurodegeneration: alteration/loss of synaptic terminals decreases the astroglial expression of KBBP, which ultimately negatively regulates astroglial expression of the peri-synaptic GLT1.