The role of astrocytic GFAP-‐isoforms in the cellular
response following mechanical injury
Research report
T.A.O. Muntslag
21/05/2014
Supervised by:
O.M.J.A. Stassen
Abstract
Our aim was to determine the role of astrocytic GFAP-‐isoforms in the cellular
response following traumatic brain injury (TBI). In vitro modelling of TBI was done
with a cell injury controller, which uses a short air-‐induced strain resulting in
mechanical injury. A human glioblastoma cell line (U251) was transduced to
selective overexpress GFAP-‐α, GFAP-‐δ or a mCherry control. Subsequent genetic
changes were assessed with qPCR, allowing us to study the molecular changes in
astrocytes after mechanical injury. We hypothesized that the isoforms had a
differential cellular response. Our initial stretch severity resulted in only minor
cellular changes; therefore a maximization test was performed in order to increase
the cellular effects. Although our model did not show any TBI-‐characteristic gliosis
and other subsequent changes, in both medium and severe stretch protocol two
genes were consistently upregulated in our 30-‐minutes stretch condition (FOS,
EGR1). The upregulation of FOS and EGR1 corresponds to previous experiments in
both rodent and human tissue; however, no isoform dependent cellular response
differences were found in our study. Future calcium imaging studies, and
alternative model cell lines, might provide more insights into the role of GFAP-‐
Introduction
Traumatic brain injury (TBI) is one of the most common causes of injury-‐related brain
dysfunction with a range of psychological and societal problems (Maas et al., 2008; Rapoport et al., 2012). Although the largest proportion of patients has only minor injuries like
temporary mild-‐cognitive impairment, some require considerable specialist care in order to deal with their lasting disability.
It is estimated that in the US, in the year 2000, $80 billion was spend for medical treatment with an estimated $326 billion for lost productivity (Corso et al., 2006). Even with these numbers there is an underestimation of the scope of problem, as untreated/undiagnosed individuals are not included in the study. Furthermore, TBI survivors have a higher risk of Alzheimer’s disease, stroke and epileptic seizures, which are each considerable problems on their own (Sivanandam et al., 2012; James et al., 2013; Christensen et al., 2009). Knowledge of TBI pathology might help prevent the development of further neurodegeneration and subsequent loss of function. Central to any therapeutical intervention of the sequela of TBI is knowledge about the cellular mechanisms in response to the initial injury.
On a cellular level TBI is characterized by cell death, synaptic degeneration, mitochondrial dysfunction, inflammation and ischemia by a reduction in the small vasculature (Floyd et al., 2007; Werner and Engelhardt, 2007). However, all of these detrimental effects are
secondary to the initial mechanical injury caused by the fall, blow, or other external force mediators. The physical distortion of the cell’s cytoskeleton is the initial responder to injury, which, according to the mechanochemical control hypothesis, is a determining factor for the cell’s function (Ingber et al., 2008). The cytoskeleton consists of three separate, although interrelated, categories of filament systems: microfilaments, microtubules and intermediate filaments (IFs).
Although most research has gone into neuronal alterations following TBI, astrocytes are now regarded as a heterozygous cell group important for energy metabolism, neurotransmitter clearance, defense against oxidative stress and modulation of synaptic strength (Matyash and Kettenmann, 2010; Allaman et al. 2011). These cells respond to TBI by hypertrophy, proliferation and an underlying change in gene expression (Myer et al., 2006). Moreover, recent studies have implicated astrocytic IFs in TBI pathology.
Glial fibrillary acidic protein (GFAP) –astrocyte type III IF-‐ mediates reactive gliosis in response to TBI (Middeldorp and Hol, 2011; Pekny and Pekna, 2004; Chiu and Goldman, 1985). Gliosis is believed to have an early protective role by allowing a compartmentalization of pathology after injury, although it can prove detrimental when gliosis becomes chronic and post-‐insult functional reorganization is prevented (Laird et al., 2008). Moreover, it has been suggested that presence of GFAP increases resistance to mechanical injury (Pekny and Pekna, 2004; Myer et al., 2006).
Apart from the role of GFAP in gliosis following TBI, there is limited knowledge in regard to the responses of the isoforms. The canonical isoform of GFAP is GFAP-‐ α, as it is the most abundant and widespread isoform among the 7 known in humans (Middeldorp and Hol, 2011). Most studies that report on GFAP do not distinguish among the isoforms, which suggests that any reported changes most often relate to GFAP-‐α (Reeves et al., 1989; Middeldorp et al., 2011). Another isoform of interest is GFAP-‐δ. GFAP-‐δ differs from GFAP-‐α by having interchanged exons 8 and 9 with exon 7+/7a, thereby shortening the transcript and giving it different assembly characteristics (Mamber et al., 2012; Singh et al., 2005). A relative increase of this shortened isoform above 10% impaired GFAP-‐filament formation, causing a complete collapse of the IF-‐network with GFAP in a juxtanuclear position (Perng et al., 2008). Considering their putative role in injury resistance and the opposing assembly features, it is expected that a cell with an overexpression of either GFAP-‐α or GFAP-‐δ has a differential mechanical injury response. In addition, GFAP-‐δ is a neuronal stem cell marker in humans, whereas its expression may be central to all astrocytes in mice (Mamber et al., 2012; Roelofs et al., 2005).
Mechanical stimulation/stress/injury can be modeled in vitro via multiple approaches. Application of fluid-‐shear stress, mechanical or vacuum induced radial/uniaxial stretch or magnetic microbead manipulation of integrins are all examples (Ellis et al., 1995; Cui et al. 2004; Park et al. 2004; Ingber, 2008). Moreover, studies differentiate from each other by applying either a cyclic, chronic or transient induction of injury burst (Ellis et al., 2004; Cui et al., 2004; Park et al., 2004).
In order to study the isoform specific role in the cellular response to mechanical injury, we use an endogenously low GFAP-‐expressing astrocyte cell line (U251 human glioblastoma) to selectively overexpress the α-‐ and δ-‐isoforms, or mCherry – a red monomer for control transduction (mCh). These cells are cultured on a laminin peptide coated, elastic membrane-‐ and subjected to a strain stress by a short burst of air, as described by Ellis and colleagues (1995). Although each model has its merits, by choosing a model with a transient radial injury we have chosen a model with similarities to the effects of blast induced TBI (Chen et al., 2009).
An initial microarray analysis of stretched astrocytes resulted in only minor cellular effects, and was without any differences between the cellular response of GFAP-‐α-‐ and δ; suggestive of an injury induction lacking in and/or a weak response of these cells to injury. A Gene Ontology (GO)-‐analysis of the array data did indicate that the genes, that were affected, are associated with apoptosis, autophagy and mitochondrial dysfunction. A subsequent
Ingenuity analysis of these genes located them to the mitochondrial respiratory chain, most predominantly in complexes I, III and IV. The transfer of electrons through this chain, and the subsequent movement of protons across the intermembrane of the mitochondria, is the main catalyser for cellular ATP production. Therefore, any problems in this respiratory chain could cause mitochondrial dysfunction, and a decreased ability to produce ATP in response to energy demands (Brand et al., 2011). Mitochondrial dysfunction is concomitant with an increased production of reactive oxygen species (ROS), which in turn is able to initiate further mitochondrial decay in a downward, self-‐propagating pathway, leading up to apoptosis (Lin et at al., 2006; Roberts et al., 2010; Cheng et al., 2012). Furthermore, lasting cognitive deficits in TBI-‐patients are suggested to have their origin in the dysfunction of these organelles (Lifshitz et al., 2004). The correlation of mitochondria with apoptosis, and the early time-‐point in which they are affected, suggests a primary role for these organelles in the cellular response to mechanical injury.
In sum, we hypothesized that GFAP-‐α and GFAP-‐δ affect the cellular response differentially. More specifically, we expect that mitochondria are the central mediators in response to mechanical injury.
Here we report on the current state of our study. We were unable to verify most targets of our microarray analysis due to the minor cellular changes in our initial stretches. The minimal cellular changes and possible sensitivity limitations of the acquisition method (qPCR) prevented a further verification. An increase of stretch severity did result in a larger expression level of FOS and EGR1, but no significant genetic changes were found in response of our stretch protocol. Subsequent protein measures and mitochondrial function assays are lacking.
Materials and methods
Cell culture
U251 cells were kept in 10cm dishes in a humidified incubator at 37°C with 5% CO2. Medium
was composed of 2% Fetal Calf Serum (FCS), supplemented with 1% P/S (100 U/ml penicillin, 100 ug/ml streptomycin) in a 1:1 F10/DMEM+Glutamax mixture.
Passage and preparation for stretch was done as follows. Cells in 10cm plates were washed with Versene (0.53 mM EDTA in PBS) and trypsinized for 5 minutes in the humidified incubator (i.e. 37°C with 5% CO2), using a mixture of Trypsin:Versene (1:3). Trypsinization
was stopped with FCS-‐containing medium, after which cells were spinned down for 3 minutes at 300 g and room temperature (RT). Cells were resuspended with medium and supplemented to 5 mL total volume. A haemocytometer was used for counting cells, enabling an appropriate estimation of cell density throughout experiments. Remaining cells were seeded and grown for later stretches.
Stretch protocol
Cells were plated (5 *10^5, U251) were plated on a YIGSR laminin peptide coated bioflex 6-‐ well plate, as supplied by Dunn Labortechnik Germany and kept in the humidified incubator. The cells were allowed to attach for at least 6 hours, after which the medium was changed from 2% to 0% FCS, and kept overnight (O/N) in the incubator (fig. 1). Serum starvation stops proliferation of the plated cells (Cooper et al., 2003).
Figure 1; Schematic representation of our stretch protocol. 500.000 cells were plated and kept in the incubator for 6 hours
with FCS-‐containing medium (2%), after which they were changed to and kept O/N in medium lacking FCS. The first stretch (i.e. 24-‐hours stretch) occurred the following day. The second (i.e. 30-‐minutes stretch), and subsequent RNA isolation was performed on the third day. Total duration of our stretch protocol was ~48 hours.
After our O/N incubation, two wells were subjected to our 24-‐hours stretch condition with a pressure of ~67 psi, in order to obtain a response peak pressure of ~5.7 psi (i.e. 0.39 bar). Previous (medium) stretches had a peak pressure of ~4.0 psi (i.e. 0.27 bar). 30 minutes prior to RNA-‐isolation another two other wells were subjected to a strain stress, using similar force as in the previous stretch (fig. 1). The remaining two wells served as unstretched control. Mechanical injury was induced with the ’Cell-‐Injury Controller II’ (fig. 2) (Custom Design & Fabrication, Inc., USA).
Figure 2; Cell-‐Injury Controller II (Custom Design & Fabrication, Inc., USA)
RNA-‐isolation
For RNA isolation medium was taken of the wells, and 0.5ml-‐1 ml Trisure was applied to the wells. Samples were placed in a fresh eppendorf tube and kept on ice. Separation of RNA from all other cellular components was done by addition of chloroform, incubating for 5 minutes and spinning for 15 minutes at 12.000 g and 7°C.
Spinning resulted in a 3-‐layered liquid (i.e. phenol phase, interphase, and aqueous phase). The aqueous phase was taken off the underlying inter-‐ and phenol phase (i.e. ~ 200 µL), and placed in a new eppendorf tube. This was supplemented with isopropanol (i.e. 200 µL) and 1 µL of glycogen, and stored at least O/N at -‐20°C. The phenol phase was kept and is usable for future protein isolation procedures, although stability of protein/yield is expected to be lower in this indirect way, as opposed to direct protein isolation using lysis buffer containing protease inhibitors.
After O/N storage, RNA-‐isolation was continued with a spinning step for 45 minutes at max speed (ca. 20.000 g) and 4°C). Supernatant was removed from the RNA-‐pellet and washed twice with 75% ethanol. After each, RNA pellet was spun down for 1 minute at maximum speed and 4°C. After removal of the last wash, pellet was left to air-‐dry for approximately 15 minutes. RNA was dissolved in a minimum of 15 µL MilliQ (MQ).
cDNA creation
cDNA creation was done with the Quantitect Reverse Transcription Kit by Qiagen, which uses a combination of oligo-‐dT and random primers. Yields of the U251 have readily enabled a MQ solution with RNA of 500 ng/6µL.
cDNA-‐synthesis was started by with an incubation step of our RNA-‐solution (6 µL), and an additional 1 µL of 7x gDNA Wipeout Buffer. During this step Wipeout Buffer removed excess contamination of DNA. Reverse transcription was started with the addition of 2.0 µL 5x Quantiscript RT Buffer, 0.5 µL RT Primer Mix, and 0.5 µL Quantiscript Reverse Transcriptase (RT). Thus, bringing the total reaction mixture volume to 10 µL.
qPCR
Real time qPCR was based on the monitoring of fluorescent SYBR Green I by a sequence detection system (Prism 5700; Applied Biosystems Inc, Nieuwekerk a/d IJssel, The
Netherlands). SYBR binds double stranded (ds)-‐DNA and can thus be regarded as a measure of the amount of double stranded amplicon present in that specific sample. Ideally each cycle doubles the signal (i.e. amount of ds-‐DNA), up to a set threshold. Threshold was identical for all genes, thereby providing a means of comparing RNA-‐template levels throughout samples.
cDNA samples were diluted 1:20 for real-‐time quantitative PCR assays with SYBRH Green PCR Master Mix (ABI) (2X mixture of SYBR Green 1 Dye, AmpliTaq Gold® DNA Polymerase, dNTPs with dUTP, Passive Reference 1 (ROX), and optimized buffer components). Measures were taken with a passive reference dye in the PCR mixture. Variability in concentration or volume of sample were accounted for using its fluorescence. Total mixture had a
composition as follows: 5 μL SYBR Green PCR Master Mix, 3.5 μL MQ, 0.5 μL of primer mix (2μM) and 1 μL of diluted cDNA.
The sequence detection system had several steps of measurement. First, a 2-‐minute
incubation step at 50°C, followed by a polymerase activation at 95°C for 10 minutes. Second, a cycling melting step at 95°C for 15 seconds and annealing-‐elongation at 60°C for 1 minute for 40 cycles. Continuous measure of fluorescence enabled the construction of a dissociation curve (i.e. melting profile). Melting profiles of the used primers were checked after every microplate and only single melting peaks were found; suggesting that there were no primer dimers in our RNA-‐measures (data not shown). Further technical details can be found in Kamphuis et al., 2012.
GeNORM and statistical analysis
Although our microarray gave us an insight as to which genes were the most stable in our data sets, translating findings in our medium stretch directly to the severe stretch are not recommended. Therefore, we chose to perform a GeNORM analysis in order to validate stability data of our microarray, as well as checking which genes could be considered as reference genes in the severe stretch condition. GeNORM analaysis is a pairwise variation analysis of the genes in the dataset, providing a measure for internal expression stability (Vandesompele et al., 2002). In our medium stretch protocol GAPDH, PPP3CB and RPS6 matched requirements of variation, whereas CNOT10, PPP3CB and RPS6 did in our severe stretch protocol.
Analysis for statistical significance was done with Graphpad using a non-‐parametric Kruskal-‐ Wallis test with post-‐hoc Dunn’s correction.
Results
GeNORM analysis
Using microarray and GeNORM-‐analysis we found that RPS6, PPP3CB and GAPDH are the most stable genes in our medium stretch (MS) protocol, whereas RPS6, PPP3CB and CNOT10 are the most stable in severe stretch (SS) (fig. 3 and 4). Throughout stretch and isoform conditions these genes have similar threshold cycles without any statistical differences.
Figure 3;Threshold cycles (Ct’s) for referen2es genes in medium stretch protocol. Ct’s for RPS6 (a), PPP3CB (b), and GAPDH (c),
as measured with qPCR. X-‐axis is labelled according to each cell line and the their measures in time: mCh (Ch), α, δ; and 0 (control), 30 (minutes post-‐stretch), 24 (hours post stretch), respectively. No statistical significant differences were found between stretch and isoform conditions (n=4).
Figure 4;Threshold cycles (Ct’s) for referenes genes in severe stretch protocol. Ct’s for RPS6 (a), PPP3CB (b), and CNOT10 (c),
as measured with qPCR. X-‐axis is labelled according to each cell line and the their measures in time: mCh (Ch), α, δ; and 0 (control), 30 (minutes post-‐stretch), 24 (hours post stretch), respectively. No statistical significant differences were found between stretch and isoform conditions (mCh, α, n=3; δ, n=2)
Ch.0 Ch.30 Ch.24 α.0 α.30 α.24 δ.0 δ.30 δ.24 0 1 2 3 4 RPS6
Stretch and isoform conditions
Thr e s hold c y c le (C t) (a) Ch.0 Ch.30 Ch.24 α.0 α.30 α.24 δ.0 δ.30 δ.24 0.00 0.01 0.02 0.03 0.04 0.05 PPP3CB
Stretch and isoform conditions
Thr e s hold c y c le (C t) (b) Ch.0 Ch.30 Ch.24 α.0 α.30 α.24 δ.0 δ.30 δ.24 0 5 10 15 GAPDH
Stretch and isoform conditions
Thr e s hold c y c le (C t) (c) Ch.0 Ch.30 Ch.24 α.0 α.30 α.24 δ.0 δ.30 δ.24 0 5 10 15 20 RPS6
Stretch and isoform conditions
Thr e s hold c y c le (C t) (a) Ch.0 Ch.30 Ch.24 α.0 α.30 α.24 δ.0 δ.30 δ.24 0.0 0.5 1.0 1.5 PPP3CB
Stretch and isoform conditions
Thr e s hold c y c le (C t) (b) Ch.0 Ch.30 Ch.24 α.0 α.30 α.24 δ.0 δ.30 δ.24 0.00 0.02 0.04 0.06 0.08 0.10 CNOT10
Stretch and isoform conditions
Thr e s hold c y c le (C t) (c)
Medium stretch
Initial stretching of cells was done with a pressure wave of ~46 psi, corresponding to a medium/moderate deflection of the silicon membrane and a response peak pressure of ~0.29 bar (~4.0 psi). Several putative targets from our microarray analysis were tested using qPCR. A statistical significant upregulation of FOS and EGR1 RNA-‐levels was found in our 30-‐ minutes stretch condition in all 3 cell line subtypes (i.e. mCh, α, δ) without a difference in expression between isoforms (fig. 5). However, no other changes were discovered,
suggesting that the severity of the mechanical injury could be insufficient for a clear cellular response or that the cells were inherently unresponsiveness to stretch.
Figure 5; RNA expression levels of target genes in medium stretch (MS). RNA-‐expression profiles of FOS (a/b) and EGR1(c/d),
as measured with qPCR and corrected for GAPDH, PPP3CB and RPS6. X-‐axis is labelled according to each cell line and the their measures in time: mCh (red), α (green) and δ (blue); and 0 (control), 30 (minutes post-‐stretch), 24 (hours post stretch), respectively. No statistical differences were found between the different isoforms, whereas 30-‐minutes post stretch RNA expression was statistically different from both other time-‐points (P<0.0001, ****; P<0.01, **). Non-‐parametric Kruskal-‐Wallis test with post-‐hoc Dunn’s correction was performed for statistical analysis (0 30, 24; n=12).
Ch.0 Ch.30.2 Ch .24 α.0 α.30 α.24 δ.0 δ.30 δ.24 0.000 0.005 0.010 0.015 FOS Stretch3and3isoform3condi<ons No rm al iz ed 3R NA Be xp re ss io n a) 0 30 24 0.000 0.005 0.010 0.015 FOS Stretch0condi5on No rm al iz ed 0R NA =e xp re ss io n **** ** b) Ch.0 Ch.30.2 Ch .24 α.0 α.30 α.24 δ.0 δ.30 δ.24 0.0 0.1 0.2 0.3 0.4 0.5 EGR1 Stretch4and4isoform4condi=ons No rm al iz ed 4R NA Be xp re ss io n c) 0 30 24 0.0 0.1 0.2 0.3 0.4 0.5 EGR1 Stretch1and1isoform1condi:ons No rm al iz ed 1R NA ?e xp re ss io n **** ** d)
Injury maximization
Medium stretch severity did not result in a large cellular response. Therefore, we aimed to maximize the mechanical injury by amplifying the pulse pressure to ~67 psi, corresponding to a large deflection of the silicon membrane and a response peak pressure of ~5.7 psi (0.39 bar). This pulse pressure of ~67 psi approached the maximum of our setup, thereby limiting a further increase in severity (Ellis et al., 1995). Injury severity was assessed using RNA-‐ expression levels of genes from our medium stretch protocol. Both FOS and EGR1 had an increase in their relative fold change in our 30-‐minutes stretch condition (fig. 6). Our 24-‐ hours stretch condition did not differ in their expression levels (data not shown).
Figure 6; Cell response maximization. Relative fold increase of FOS (a) and EGR1 (b) RNA-‐expression in 30-‐minutes post-‐stretch
condition, as compared to unstretched (control). Measures were obtained with qPCR and corrected for PPP3CB, RPS6, and either GAPDH (MS) or CNOT10 (SS). X-‐axis is labelled with cell line (i.e. mCh (Ch), α, or δ) and their 30-‐minutes post stretch measure. Medium stretch (MS) has a lower fold change than severe stretch (SS) for both FOS and EGR1.
Severe stretch
The severe stretch samples were gathered using our new maximized peak response pulse of ~0.39 bar (~5.7 psi). Transductions are generally quite stable, however, after prolonged use it is possible that introduced genes are not expressed as much as they were in the beginning. During our severe stretches we discovered a reduction in expression of our GFAP-‐isoforms. Therefore, to prevent a further loss of our proteins of interest, a new transduction was performed, and cells were FACS-‐sorted resulting in a more consistent and higher expression of our isoforms of interest (data not shown). No comparative quantitative measure was performed for protein level expression of both transductions. The severe stretch dataset discussed here contains both transductions, henceforth referred to as the first and second
MS#Ch SS#C h MS#α SS#α MS#δ SS#δ 0 10 20 30 40 50 FOS Stretch#and#isoform#condi<ons No rm al iz ed #R NA Be xp re ss io n a) MS#Ch SS#C h MS#α SS#α MS#δ SS#δ 0 5 10 15 20 EGR1 Stretch#and#isoform#condi;ons No rm al iz ed #R NA @e xp re ss io n b)
batch of cells. As stated, our severe stretch protocol induced an increase in FOS and EGR1 RNA-‐expression levels of our 30-‐minutes stretch condition throughout isoforms.
However, despite a larger expression of FOS and EGR1 in our severe stretch protocol, no difference between the isoforms was found (fig. 6 and 7). Of our microarray targets, no other genes were significantly affected by our stretch protocol (data not shown). Preliminary findings did indicate that the second transduction responded with greater severity,
suggesting that there was a cellular response difference between both transductions (fig. 7). Moreover, possibly obscuring smaller genetic changes.
Figure 7; RNA expression levels of target genes in severe stretch (SS). Mean (+ standard deviation) RNA-‐expression profiles of FOS (a/b) and EGR1 (c/d), as measured with qPCR and corrected for RPS6, PPP3CB and CNOT10. X-‐axis is labelled according to each cell line and the their measures in time: mCh (circle), α (triangle), δ (square); and 0 (control/unstretched), 30 (minutes post-‐stretch), 24 (hours post stretch), respectively. Dots depicted in colour are measures from the new transduction, suggesting that there is a difference between the cellular response in the first and second batch. No statistical differences were found between the different isoforms, whereas 30-‐minutes post stretch RNA expression was statistically different from both other time-‐points (P<0.01, **). Non-‐parametric Kruskal-‐Wallis test with post-‐hoc Dunn’s correction was performed for statistical analysis (0, 30, 24; n=8). Ch.0 Ch.30Ch.24 α.0 α.30 α.24 δ.0 δ..30 δ.24 *0.5 0.0 0.5 1.0 1.5 2.0 2.5 FOS Stretch4and4isoform4condi=ons No rm al ize d4 RNA *e xp re ss io n a) 0 30 24 0.0 0.5 1.0 1.5 2.0 2.5 FOS Stretch0condi5on No rm al ize d0 RNA =e xp re ss io n b) *** ** Ch.0 Ch.30 Ch.24 α.0 α.30 α.24 δ.0 δ..3 0 δ.24 0 10 20 30 40 50 EGR1 Stretch'and'isoform'condi0ons No rm al ize d' RNA 6e xp re ss io n c) 0 30 24 0 10 20 30 40 EGR1 Stretch/condi4on No rm al ize d/ RNA ;e xp re ss io n d) ** **
GFAP is non-‐reactive throughout our stretch conditions and stretch severities (fig. 8).
Looking in more detail to our GFAP-‐isoforms in our second batch of cells does suggest an upregulation of RNA (fig. 9); however, after accounting for the corresponding correction factor, no reactivity was found (suppl. fig. 1). GFAP-‐expression level differences between the 1st and 2nd batch of cells do further support a differential cellular response between both
batches (fig. 9).
Figure 8; RNA expression levels of new target genes in severe stretch (SS). Expression profiles of GFAP-‐α (a) and
GFAP-‐δ (b), as measured with qPCR and corrected for RPS6, PPP3CB and CNOT10. X-‐axis is labelled according to each cell line and the their measures in time: mCh (Ch), α, δ; and 0 (control), 30 (minutes post-‐stretch), 24 (hours post stretch), respectively. No differences in expression level were found between stretch conditions.
Figure 9; RNA expression levels of GFAP-‐α and-‐δ in severe stretch (2nd batch). Expression profiles of GFAP-‐α and
GFAP-‐δ, as measured with qPCR and corrected for RPS6, PPP3CB and CNOT10. X-‐axis is labelled according to each cell line and the their measures in time: mCh (Ch), α, δ; and 0 (control), 30 (minutes post-‐stretch), 24 (hours post stretch), respectively. No statistical analysis was performed (n=1).
ch.0 ch.30 ch.24 a.0 a.30 a.24 d.0 d.30 d.24 0.00 0.05 0.10 0.15 0.20 0.25 GFAP0α Stretch5and5isoform5condi<ons No rm al iz ed 5R NA 0e xp re ss io n a) Ch.0 Ch.30 Ch.24 α.0 α.30 α.24 δ.0 δ.30 δ.24 0.00 0.02 0.04 0.06 10 20 30 40 Stretch'and'isoform'condi0ons No rm al iz ed 'R NA 6e xp re ss io n GFAP-δ b) Ch.0Ch.30Ch.24 α.0 α.30 α.24 δ.0 δ.30 δ.24 0.00 0.05 0.10 0.15 0.20 0.25 GFAP0α" Stretch6and6isoform6condi?ons No rm al iz ed 6R NA 0e xp re ss io n 1st 2nd a) Ch.0Ch.30Ch.24 α.0 α.30 α.24 δ.0 δ.30 δ.24 0.00 0.02 0.04 0.06 0.08 0.10 10 20 30 40 Stretch2and2isoform2condi;ons No rm al iz ed 2R NA Ae xp re ss io n GFAPAδ b)
Interbatch differences in the cellular response suggested reanalysing of our initial dataset. Separating both transductions gave rise to several genes of interest in our severe stretch protocol in our second batch of cells (table 1). These changes remained after accounting for fluctuations in our correction factor (suppl. fig 1, table 1). Remarkably, most changes in our α-‐line are reductions in gene expression, whereas our δ-‐line has mostly increases in
expression.
Ch
α
δ
30/0 24/0 30/24 30/0 24/0 30/24 30/0 24/0 30/24
CKLF
COX11
CTNNB1
GFAPd
GSK3B
MAFF
PARK2
PLXNA4
SEMA4A
SMPD1
TXNRD2
UTP14C-‐1
UTP14C-‐2
Table 1; Genes affected in our second batch using a severe stretch. On the Y-‐axis genes are depicted, whereas on the
X-‐axis the comparison is shown. For example, 30/0 is the RNA-‐expression level ratio of our 30-‐minutes condition to our unstretched. Similarly, 30/24 is the ratio of our 30-‐minutes stretch condition versus the 24-‐hours condition. Red shows a reduction by more than half, whereas green depicts an increase of more than 50%. For example, the GFAP-‐δ gene has a higher expression in our 24-‐hours stretch in comparison to unstretched in our δ-‐overexpressing line.
Discussion
Microarray and GeNORM-‐analysis showed that RPS6, PPP3CB and GAPDH are the most stable reference genes in moderate stretch, whereas in the severe stretch CNOT10, RPS6, PPP3CB are the most stable genes. GAPDH is commonly regarded as a stable reference gene (Li et al., 2012), however, several studies have found the gene to be variable across species, and experimental and disease conditions (Mahoney et al., 2004; Nelissen et al., 2010). Although the gene is one of the most stable in our initial medium stretch, with an increased stretch severity the expression stability of GAPDH becomes more variable, and CNOT10 is found to be a more stable gene. This suggests that initial microarray findings –using a moderate stretch-‐ may not translate to our severe stretch, possibly overlooking genetic changes with increased mechanical injury. Moreover, differences in the cellular response were not restricted to our prospective reference genes and considerable differences in gene expression levels were found between both stretch conditions (fig. 6).
FOS is part of a family of transcription factors that associate with the Jun family in a heterodimeric manner, and is known to be upregulated in astrocytes in a wide range of conditions, of both disease and normal functioning (Chiang et al., 2012). In turn, EGR1 has been associated with damage causing conditions, and seems an integral part of the cellular response (Fan et al., 2011; Beck et al., 2008). More importantly, FOS and EGR1 are found to be upregulated after clinical and experimental TBI (Silverman et al., 1998; Michael et al., 2005; Lu et al., 2000), and are therefore in support of our findings.
Cultured astrocytes respond to trauma by increasing their intracellular calcium levels in a direct correlation with severity (Lange et al., 2012; Rzigalinski et al., 2002), thereby enabling Ca2+-‐dependent FOS and EGR1 expression (Mellström and Naranjo, 2001; Stula et al., 2000).
Furthermore, intracellular calcium levels play a role in GFAP-‐phosphorylation, thereby correlating this primary stress responder to the stability and protective capacity of GFAP (Takemura et al., 2002; Schlaepfer and Zimmerman, 2006).
Fan and colleagues (2011) showed that GFAP expression is regulated by p300, as specific mutations within p300’s EGR1-‐transactivating promoter abolishes this upregulation (Fan et al., 2011). Absence of FOS does not impair GFAP transcription, as GFAP content of astrocytes accumulated in the absence of this gene (Dragunow et al., 1990). Therefore, despite FOS
being an excellent marker for neuronal activation (Kovács et al., 2008), it is not always associated with GFAP-‐expression and EGR1 might prove a better measure for optimization success when regarding astrocytic reactive gliosis in response to our stretching protocol. The correlation of Ca2+, EGR1 and GFAP underlies their importance in the astrocytic cellular
response after mechanical injury (fig. 10). FOS might therefore only be involved in smaller cellular mechanisms.
Figure 10; Model for the correlation of Ca2+, EGR1 and GFAP in the cellular response to mechanical injury.
Using qPCR we attemped to check several microarray targets. Although some overlap in pattern could be seen in genes, e lack of overlap suggests that the initial small expression level differences can’t be delineated further using qPCR (data not shown). Furthermore, an array uses certain nucleotide probes in order to recognize the expressed RNAs. Primers are based on known refseq sequences and it is thus possible the array probes recognize
unaccounted regions of the gene, thus despite having made search-‐parameter constrictions towards exon spanning and accounting for alternative splicing, it is possible that two separate regions of the transcript are recognized. Multiple studies have pointed out that such an alignment is desirable for validation of the array (Dallas et al., 2005; Wang et al., 2006). Therefore, it is possible that the already minor effects found in the array samples, and the lack of sequence complementarity between real time PCR and microarray, prevented us from validating other target genes. When possible, results of future primers may be
improved if the microarray probe region is taken into account in the primer design.
Alternatively, it is possible that there were no other differences to detect. Despite finding an upregulation of FOS and EGR1, an upregulation of GFAP is absent in our study. Epigenetic modulators regulate genetic expression. The methylation state of the promoter of a gene determines the ‘accessibility’ for the polymerases, where less methylation equals more transcription and vice versa. It has been suggested that there is a role of epigenetic silencing of the GFAP-‐promoter with increasing malignancy (Restrepo et al., 2010). Therefore, it is possible that absence of gliosis is a consequence of epigenetic inheritance of our cell line. Subsequent genetic changes correlated with GFAP are therefore underestimated/absent.
Intermediate filaments have a possible protective role in mechanical injury, as an absence of these filaments makes cells more susceptible to that injury (Fuchs and Cleveland, 1998; Pekny and Pekna, 2004; Lane and Pekny, 2004; Silverman et al. 1998). Apart from this mechanical property, there is a hypothesis suggesting that IFs can add a non-‐mechanical dimension to IF-‐mediated signalling (Hyder et al., 2008). This phosphate ‘sponge’ hypothesis suggests that keratin – another intermediate filament – is capable of absorbing stress-‐ activated kinases, thereby reducing non-‐keratin pro apoptotic substrate phosphorylation and reducing overall apoptosis (Ku and Omary, 2006). Similarly, 14-‐3-‐3 proteins –IF solubility factors-‐ were found to be centres of kinase activity and involved in cell cycle progression (phenotypical development) (Gardino and Yaffe, 2011). In this way, IF’s are generally important for both the mediation of external forces, and the intracellular localization of primary responders. A similar role could thus be postulated for other filaments like GFAP, which could explain putative genetic changes in our second batch of cells. Whereas there is hardly any change in expression in our control line, both the α-‐ and δ-‐line might have a considerable altered response. Remarkably, it could even be suggested that our α-‐ and δ-‐ line have an apparent opposed response, with the α-‐line having an initial reduction of expression and our δ-‐line a more immediate increase in RNA-‐expression. Therefore, it is possible that an intact network – as in our α-‐line – is capable of sequestering direct cellular change by capturing necessary protein kinases and Ca2+. This non-‐mechanical function of
intermediate filaments is could thus responsible for the putative genetic changes in our overexpressing lines within our second batch (table 1). Furthermore, it may be able to explain why a proper GFAP-‐network is protective from mechanical injury. Temporal and spatial imaging of Ca2+-‐might thus provide greater insight into the cellular changes (e.g.
apoptosis) following mechanical injury.
In sum, our study found a consistent RNA upregulation of both FOS and EGR1 in our 30-‐ minutes stretch condition throughout experimental protocols. Both in vitro and in vivo studies with rodent and human models have previously shown that FOS and EGR1
expression after clinical/experimental TBI (Silverman et al., 1998; Michael et al., 2005; Lu et al., 2000; Awashti et al., 2003; Dutcher et al., 1999). We have established that U251
astrocytoma cells respond in a similar fashion, suggesting that these immediate early genes can be regarded as the central mediators in the cellular response following mechanical injury. EGR1 is of particular importance to reactive gliosis, although GFAP RNA-‐expression was non-‐responsive throughout our stretch protocols and conditions.
Moreover, no differential transcription was found between control and either overexpression of GFAP-‐α or GFAP–δ; suggesting that there is no differential cellular response between cells with either isoform overexpressed. However, preliminary results of our second batch of cells, using the severe stretch protocol, might suggest that there is a temporal and directional discrimination in their respective responses (table 1).
Future perspectives
Our study has shown the importance of FOS and EGR1 in the mechanical injury response of U251 astrocytoma cells. EGR1 is of particular interest in reactive gliosis via its possible role in GFAP expression (Fan et al, 2011). Furthermore, EGR1 regulates several tumor suppressor genes such as TGFβ1, PTEN, p53 and fibronectin (Baron et al., 2006; Liu et al., 2001). Considering that most tumors have a dysfunctional PTEN or p53, the U251’s might have a differential cellular response than a healthy astrocyte, preventing the apoptotic initiation that is characteristic of TBI (Liu et al., 2001; Brázdová et al., 2009). If we hypothesize that the GFAP isoforms have a differential apoptotic response in mechanical injury, it would serve our goal to have a model that is able to show this cascade leading up to cell death. The donor-‐derived astrocytes isolated by Van Strien (10/103, Netherlands Institute for Neuroscience) could be such a model.
An integral part of synaptic plasticity is the cellular regulation of calcium. However, during trauma excessive calcium pathology occurs, which increases ROS production and induces mitochondrial dysfunction in TBI patients (Lifshitz et al., 2004; Brookes et al., 2004). In addition, both FOS-‐ and EGR1-‐transcription are calcium dependent, although no measures of Ca2+ concentration were performed. Several dyes will be suitable for calcium imaging in our
stretching procedures (e.g. fura-‐2) (Rzigalinski et al.,1998). Alternatively, inducing mechanical injury in the presence of calcium chelators can help discriminate the physical distortion effects from transcriptional regulations.
Hartings and colleagues (2011) report on a negative correlation between spreading neuronal depolarizations and TBI outcome. However – in light of the supportive and active role of glia in neuronal function-‐ glial cells are expected to play a role in the modulation of the TBI outcome. McCall and colleagues (1996) found that GFAP-‐null mice displayed enhanced neuronal long-‐term potentiation (LTP) of both population spike amplitude and excitatory post-‐synaptic potential slope compared to control mice, without an concurrent
compensatory upregulation of either vimentin nor nestin. Astrocytes lacking GFAP could thus be restricted in their ability to limit LTP, resulting in spreading depolarizations and synaptic dysfunction. Therefore, upregulation of GFAP during gliosis could be responsible for restricting these detrimental neuronal depolarizations and subsequent damage.
It would thus be interesting to see how mechanical injury affects synaptic plasticity in a co-‐ culture of neurons and our isoform-‐specific astrocytes. In addition, Pablo and colleagues (2013) recently used an in vitro model ischemia model to describe the ROS-‐production in the presence and absence (i.e. GFAP-‐/-‐, Vim-‐/-‐) of a IF-‐network. They showed that the absence of
a network resulted in both an increased ROS-‐production, as well as a reduced metabolic function in a co-‐culture of astrocytes and neurons. This already increased our understanding of the correlation between mechanical injury, ROS, mitochondrial dysfunction, and the role intermediate filaments play in this process. We could take a step further by looking into the isoform-‐specific role in ROS production and synaptic plasticity.
References
Allaman, I., Bélanger, M., and Magistretti, P.J. (2011). Astrocyte–neuron metabolic relationships: for better and for worse. Trends in Neurosciences 34, 76–87.
Awasthi, D., Kutz, S.C., Beuerman, R., Nguyen, D., Carey, M.E., Zeiller, S., 2003. Early gene expression in the rat cortex after experimental traumatic brain injury and hypotension. Neuroscience Letters 345, 29–32. doi:10.1016/S0304-‐3940(03)00497-‐X
Baron, V., Adamson, E.D., Calogero, A., Ragona, G., and Mercola, D. (2006). The transcription factor Egr1 is a direct regulator of multiple tumor suppressors including TGFβ1, PTEN, p53, and fibronectin. Cancer Gene Therapy 13, 115–124.
Beck, H., Semisch, M., Culmsee, C., Plesnila, N., and Hatzopoulos, A.K. (2008). Egr-‐1 Regulates Expression of the Glial Scar Component Phosphacan in Astrocytes after Experimental Stroke. Am J Pathol 173, 77–92.
Brand, M.D., and Nicholls, D.G. (2011). Assessing mitochondrial dysfunction in cells. Biochemical Journal 435, 297–312.
Brazdova, M., Quante, T., Togel, L., Walter, K., Loscher, C., Tichy, V., Cincarova, L., Deppert, W., and Tolstonog, G.V. (2009). Modulation of gene expression in U251 glioblastoma cells by binding of mutant p53 R273H to intronic and intergenic sequences. Nucleic Acids Res 37, 1486–1500.
Brookes, P.S., Yoon, Y., Robotham, J.L., Anders, M.W., and Sheu, S.-‐S. (2004). Calcium, ATP, and ROS: a mitochondrial love-‐hate triangle. American Journal of Physiology -‐ Cell Physiology 287, C817–C833.
Chen, Y.C., Smith, D.H., Meaney, D.F., 2009. In-‐Vitro Approaches for Studying Blast-‐Induced Traumatic Brain Injury. J Neurotrauma 26, 861–876. doi:10.1089/neu.2008.0645
