Abstract: The Galileo mission provided the first data set that allowed us to investigate large-scale spatial and temporal variability in the plasma conditions in Jupiter's magnetosphere. It was the first satellite to actually maintain an orbit around Jupiter as opposed to performing a flyby like Ulysses, Voyager 1 & 2, and Pioneer 10 & 11. We used the fluxes from the PLS instrument on Galileo to determine the physical characteristics of the heavy ions in the Io torus plasma. After developing rigorous filters to properly fit the data, we constructed two-dimensional meridional profiles of Jupiter’s inner magnetosphere. We determined no significant variation in these properties with respect to local time.
Most people are aware that Earth has a magnetic field that causes a compass needle to orient itself toward the poles. However, beyond this context, many do not give the existence of this field a second thought. After all, it does not seem to have any direct effects on daily life. Yet in ways researchers are only beginning to understand, Earth’s magnetic field appears to play a crucial role in mediating the variable conditions of the space environment known as space weather. Just as Earth’s weather can become violent and cause extensive damage, severe space weather events have cost some companies hundreds of millions of dollars in damage [Fletcher, 2003]. Without a magnetic field surrounding the planet in a protective bubble called a magnetosphere, space weather would have prevented the existence of life on Earth. First, I will review the significance of magnetospheres and what has been done to understand them. I will then focus on the primary topic of this thesis: learning about the Jovian magnetosphere through data analysis of Galileo PLS instrument observations.
1.1 What is a magnetosphere?
According to David Stern , study of the Earth’s magnetic field began with the invention of the magnetic compass in China around 1000 A.D. Over the next several centuries, observers found the field configuration of the Earth to be dipolar (like that of a bar magnet), and although the field seemed to be constant over time, it would occasionally be disturbed for a day or so. Some disturbances were simultaneously observed in vastly different parts of the world, which implied that they were occurring on a global scale. It was not apparent what kind of mechanism could affect such large regions of space until astronomers realized that there was a correlation between magnetic disturbances and the number of sunspots on the sun.
Although this correlation was not direct, Stern claims that it was enough to motivate Norwegian scientist Kristian Birkeland to conduct experiments that involved moving charged particles around a spherical dipole magnetic field. In 1896, Birkeland concluded that streams of electrons originating from the sun could potentially be deflected towards the Earth’s magnetic poles, causing magnetic disturbances and the aurora. Stern notes that no one had observed electron beams emitting from the sun toward Earth, but Birkeland’s theory peaked the interest of theorists like Sidney Chapman and Vincent Ferraro . They extended Birkeland’s theory to include interactions with clouds of ions and electrons, which are collectively called plasma. Their model predicted that the Earth’s magnetic field could be distorted to form a cavity of low-density plasma around the Earth, effectively isolating it from the sun’s incoming plasma streams. The sun’s plasma streams were thought to be rare and short-lived; however, a theory proposed by Eugene Parker  predicted that the sun was constantly emitting supersonic plasma in all directions. With the advent of the Space Age, satellite measurements were able to confirm Parker’s predictions. Therefore, the cavity of low-density plasma is not a temporary conformation, but an ever-present entity shielding the Earth from constant bombardment of high-energy solar plasma. It is now known as a “magnetosphere” [Stern, 1989].
1.2 Why do we care?
Even with some background on what a magnetosphere is, it may not be obvious why it deserves attention. After all, most people would not know that the magnetosphere exists if they could not observe its effects on magnetically responsive objects like compass needles. Interestingly, life on Earth depends as much on the planet’s magnetosphere as it does on sunlight or water, especially as technology becomes increasingly electronic. All power grids are comprised of long wires with AC current constantly flowing through them. The power grids are only stable because the magnetosphere is able to shield the wires from any external electric or magnetic fields caused by space weather. However, the magnetosphere is not infallible, and when a violent space weather event is directed toward Earth, the ‘bubble’ around Earth undulates and contorts significantly. According to the laws of electricity and magnetism, a changing magnetic field will induce an electric field. This new electric field caused by the magnetosphere’s changing shape can create currents that are too large for existing power grids to handle. For example, on March 13, 1989, six million residents across the province of Quebec lost all electrical power for over nine hours, because the Hydro-Quebec power system was rapidly overloaded as the result of an intense geomagnetic storm (as these disturbances have come to be known) [Boteler et al., 1998].
Fortunately, the effects of space weather are somewhat limited to the Earth’s surface, because our thick atmosphere is another buffer against space weather. However, it should be noted that the atmosphere would be stripped away without a magnetosphere to protect it [Vasyliūnas, 2011]. High above the atmosphere, plasma or electromagnetic fields predominantly exist, creating an environment that is much more susceptible to changes in the magnetosphere. While this may not have been a serious problem 75 years ago, societies are now heavily invested in satellite technology, which is extremely vulnerable to space weather effects. On January 20, 1994, two of Canada’s Telesat communication satellites began to spin out of control due to electrostatic discharge (ESD) within their gyroscopic circuits. These ESDs were the result of spacecraft charging, meaning that the satellites gained a high level of charge from unusually high levels of plasma bombarding them. The strong electric fields created by the charged spacecraft permanently damaged the circuit elements. The high plasma levels were caused by a geomagnetic storm. Besides having to deal with a whole country of angry customers, Telesat incurred costs of $50 to $70 million in order to repair the satellites [Bedingfield et al., 1996]. High-cost problems, radio communications blackouts, radiation poisoning to astronauts, and other negative effects will continue to threaten our technological progress unless we can gain deeper understanding of the magnetosphere and how it responds to space weather.
1.3 What has been done to understand magnetospheres?
Prior to the satellite era, methods of measuring the Earth’s magnetic field were limited, and only a narrow range could be observed. The number of magnetic field measurements has dramatically increased since Sputnik 3 was launched in 1958 [Olsen et al., 2010], and the resolution of the measurements is constantly improving. There are currently several missions in orbit around the Earth, investigating the physical conditions of the magnetosphere. The Van Allen Probes, launched in August of 2012, are studying the hostile environment within the Van Allen radiation belts, which are distinct regions of high-energy plasma encircling the Earth. The probes are investigating the possible sources of the belts, and they are collecting data on how the belts respond to geomagnetic activity. Another active area of magnetospheric research is a phenomenon known as “magnetic reconnection.” Normally, magnetic field lines exert enough intrinsic pressure to keep them distinguishable from other field lines. However, when external pressures are high enough to squeeze opposing field lines together, the lines effectively “reconnect” and release huge amounts of energy as they re-equilibrate. The Time History of Events and Macroscale Interactions during Substorms (THEMIS) mission is composed of five NASA satellites that have been orbiting the Earth since 2007. THEMIS has already confirmed an electromagnetic connection between the Earth and sun via Birkeland or “field-aligned” currents, and it was the first mission to directly prove that magnetic reconnection is the mechanism that is responsible for geomagnetic storms. To follow up on these observations, the Magnetospheric Multiscale Mission (MMS) will use four satellites flying in a constant spatial configuration to probe the physics of magnetic reconnection on the microscopic scale [Olsen et al., 2010]. All of these missions will hopefully give us better insight into how our particular magnetosphere behaves. However, our magnetosphere is not the only system available for study.
Magnetospheres appear to be a ubiquitous phenomenon among celestial objects, and this makes sense in the context of current magnetic dynamo theory. Magnetic dynamos are entities that give rise to internal magnetic fields, and in the presence of the sun’s outward plasma stream, these fields form magnetospheres [for more details about magnetic dynamos, see Finn and Ott (1988)]. This means that we should expect to see a magnetosphere around most of the planets, and we do. Therefore, we can send spacecraft to other planets and study their magnetospheres to better understand how Earth’s magnetosphere behaves. The Voyager and Pioneer missions flew by the gas giants and took the first measurements of extraterrestrial magnetospheres. The Cassini spacecraft flew by Jupiter in late 2000 and is currently orbiting Saturn, providing high-quality long-term coverage of the system and sending back invaluable data. The Galileo spacecraft remained in orbit around Jupiter from 1995 to 2003 to carry out a similar objective. The remainder of this paper will discuss the data that the Galileo spacecraft collected, how it can be analyzed, and what physical implications can be determined.
1.4 What would we expect to see?
Figure 1.1 Similar to the Earth, Jupiter’s magnetic field is bent by plasma that streams outward from the sun (solar wind). Unlike the Earth, Jupiter’s magnetosphere derives significant power from the planet’s rotation and plasma ejected from its moons. [Bagenal, 1992]
The general profile of Jupiter’s magnetosphere is given by Figure 1.1 and is described in Chapter 24 of Jupiter [Bagenal et al., 2007]. Unlike the Earth’s magnetosphere, which is primarily shaped by forces from the outward solar plasma stream (known as “solar wind”), Jupiter’s magnetosphere is “rotation driven,” meaning that the majority of magnetospheric energy originates from Jupiter’s angular momentum. In the inner magnetosphere [~ 5-30 Jupiter radii (Rj)], the plasma rotates around the planet at the same rate as the magnetic field, a condition called co-rotation. Unlike Earth, there are significant sources of plasma within the magnetosphere that come primarily from volcanically ejected material from the moon Io. This internal plasma allows Jupiter’s magnetosphere to be much larger than a magnetosphere that is formed by a dipolar magnetic field with no plasma. Knowing this, we expected to see a peak in plasma density near Io’s orbit (~5.9 Rj) and some sort of decaying behavior with increasing radial distance. These preliminary expectations helped us to understand if our data analysis was headed in the right direction. However, before we could analyze the data, we had to understand a few things about it.
Galileo was the first spacecraft to maintain an orbit in an outer planet’s magnetosphere, and it was a milestone in understanding the dynamics within Jupiter’s magnetosphere. Because no other mission to the gas giant has ever been able to take observations over timescales longer than several days, the information sent back from Galileo remains the largest available data set on the Jovian system. It is supplemented by simultaneous remote observations made from Earth-based telescopes [Reviewed by Bagenal et al., 2007]. The Plasma Science instrument (PLS) comprises seven separate anodes, each connecting to the same nested set of three quarter-spherical plate electrostatic analyzers (ESAs). An ESA is essentially two plates that create a potential difference and only allow particles with a certain energy-to-charge ratio to pass through them. Particles that do not possess this ratio are deflected onto one of the plates and do not hit the detector. The PLS counts how many particles hit the detector, and the ESA limits the amount of energy these detected particles can have. The PLS was pre-programmed to step the ESA through 64 different energy values, called energy bins, ranging from 1 eV to ~50 keV. The PLS, which was attached to the spun part of the spacecraft, was designed to “sweep” through all 64 energy bins during one spacecraft “spin,” (i.e., rotation period). The seven anodes were spread out in a fan shape on the surface of the PLS [Figure 2.1], so that it could obtain three-dimensional spatial resolution over the course of a spin.
Figure 2.1 The spatial configuration of the seven PLS anodes. Their unusual geometries are designed to prevent mechanical impurities from distorting the electric field in the ESA. Note: Galileo's spin direction was actually opposite from what is indicated in the diagram.
Unfortunately, the main antenna of Galileo was not able to open completely, so it was unable to transmit data back to Earth. Engineers managed to transmit all of the data from the back-up low-gain antenna, but this reduced the data transfer rate from 134,000 bps to about 160 bps, which is about as fast as a Morse code operator can send a message [Taylor et al., 2002]. To sufficiently compress the PLS data for uplink, counts for every fourth energy bin were measured, as opposed to sequentially stepping through each energy bin. This means that a full energy sweep took four spacecraft spins, and thus the temporal resolution was reduced by a factor of four. On top of this, although the PLS could measure ions and electrons, the electron instrument failed on the first orbit, so we only examined ion data. After the data was properly organized, we were left with a total of 114,133 so-called “merged spins,” which I will refer to as “records.” From this point, we were able to begin our data analysis.
There are several valid ways that a given data set can be analyzed. Perhaps if Galileo’s main antenna had deployed properly, the data that was sent back would have had much better resolution, allowing us to employ more sophisticated analysis techniques than what was used. However, when the data are as noisy and sparse as what was returned from the PLS instrument, fancy and rigorous methods do not yield any better results than basic methods. Therefore, we made a few assumptions that might not seem reasonable for plasma in the Io torus, but they ultimately do not harm the precision of our final results.
3.11 The Maxwellian Velocity Distribution
We decided that fitting our data to a theoretical model was the best way to proceed with our analysis. By assuming that the observed plasma was in thermodynamic equilibrium, we concluded that the plasma should follow a Maxwellian velocity distribution (vectors are denoted by boldface):
where n = ion number density, m = ion mass, T = ion temperature, = single ion velocity, = bulk flow velocity, and = ion phase space density. All ions are assumed to have a common bulk flow, , which is assumed to be near the co-rotation of Jupiter. This fundamental result from thermodynamics is essentially a Gaussian distribution of individual particle velocities, peaking at the bulk flow velocity with the thermal velocity as the standard deviation. For a given ion species, the theoretical model is uniquely determined by the plasma bulk flow velocity vector (three components), the number density, and the temperature. We can adjust these parameters until the curve that they produce is the best possible fit of the counts data from a single record. However, we must define what “best possible fit” means, and we have to develop a way to obtain it.
One can get a quantitative measure of how well a curve “fits” the data by summing up all of the deviations between the data points and the corresponding points on the curve. To prevent positive and negative deviations from cancelling each other out, we sum the square of these deviations and divide by the number of points to give us an average deviation between the curve and the data. The general formula for this average deviation, known as chi-squared, is given by:
Further details about chi-squared minimization and Equation 3.2 can be found in Bevington and Robinson , but the important point is that chi-squared gives us a quantitative measure of how good a fit is. Because the Maxwellian velocity distribution is entirely determined by bulk flow, temperature, and density, these parameters also determine the value of chi-squared. If the number of adjustable parameters is M, one could imagine that finding the best possible fit could be obtained by wandering around on an M-dimensional chi-squared hypersurface until a global minimum was reached. The parameter values corresponding to this minimum are the best representation of the plasma that produced a record’s counts data.
We can get an estimate of the uncertainties of these parameters through further chi-squared analysis. When very close to the minimum, the chi-squared surface should resemble a parabola when plotted with respect to any of the parameters. The width of the parabola gives an estimate of how uncertain a parameter is. The actual calculation of these uncertainties involve numerical approximations of the second derivatives of chi-squared with respect to each parameter, and these details can also be found in Bevington and Robinson .
3.12 Single species
A Maxwellian curve is specific for each ion species (in particular, a given mass-to-charge ratio), so we could not create a best-fit curve of the data until we made an a priori assumption about the ion species present in the Io torus. Luckily, the Cassini spacecraft flew by Jupiter between 2000 and 2001 to receive a gravity assist to get to Saturn. During its encounter, the Cassini Ultraviolet Imaging Spectrograph (UVIS) instrument took observations of the UV emissions from sulfur and oxygen ions trapped by Jupiter’s inner magnetic field, creating the so-called Io torus. These observations gave us the relative abundance of the major ion species in the torus. If the data were perfect, we could have found the individual best-fit Maxwellian curves for each species. However, the data records did not show distinguishable peaks that we could attribute to different ions; they only showed the sum of their individual curves as seen in Figure 3.1.
Figure 3.1 The individual species fits were done independently of the total flux fit. The data do not indicate which fit is more accurate.
Therefore, we calculated an average mass-to-charge ratio of 13.67 from the UVIS measurements [Delamere and Bagenal, 2005]. This value comes from taking the sum of the mass-to-charge ratios of S+, S++, S+++, O+, and O++, with each term weighted by its relative abundance, and dividing that sum by the sum of the relative abundances. From there, we fit the Maxwellian curve of this hypothetical single species for each record.
There are also some secondary assumptions we made in order to keep the complexity of our analysis on par with the quality of the data. For example, ordinarily motion that is parallel to the magnetic field produces one ion temperature, and motion that is perpendicular to the magnetic field produces another ion temperature. Due to the data quality, it is futile to distinguish between these two temperatures, so we assumed the plasma had the same temperature in all directions. In addition, a Maxwellian distribution arises only when plasma is in thermodynamic equilibrium, so when we assumed this model was valid, we also implicitly assumed that plasma conditions were not changing significantly over the course of a record.
3.2 Data Pruning
Because the Maxwellian distribution is entirely determined by number density, temperature, and three components of the bulk flow velocity vector, the curve that most closely fits the data within a record gives the best-fit values of these five parameters. However, “best fit” does not necessarily mean “good fit”; it just means that the curve was the best fit out of all the fits that were computationally possible for that record. Therefore, it is entirely plausible that the parameters derived from the fits are physically unreasonable, even if we have placed proper constraints in the fitting routine. This is a potential issue with any finite data set, but the poor quality of the PLS data made this problem particularly significant.
Figure 3.2 There is too much scatter to indicate what is going on physically. Low-resolution means that the accumulation time of a count measurement was 0.5 seconds. For mid-resolution, the accumulation time was 0.2667 s, and the high-resolution accumulation time was 0.1 s.
The reason for examining these five parameters was to spot trends that correspond to known physical processes. However, when the fit values are as scattered as they are in Figure 3.2, there is no chance of understanding what is going on. There seems to be an indication of trend lines in some of the parameters, but we could not confidently say that they exist. If the model is correct, then the trends we seek are present, but they are buried underneath a smattering of “bad” data. However, “bad” is an ambiguous term, and if we want to identify which points are “bad,” we must first define what “bad” means. In order to interpret the data accurately, we have to determine “bad” based on how well the fits represent the counts data from which they were generated.
Because chi-squared represents the total deviation of the fit from the data points, we initially assumed that this value alone would determine how good a fit was. In an ideal world, as we lower the maximum chi-squared value of points to be considered, we should eventually see the expected trends emerging from the scatter. However, the PLS data set is not infinite, and as we discard more points, we are left with fewer and fewer points from which to create trend lines. Thus, as we lowered our upper limit on acceptable chi-squared values, the number of points remaining was too small to use to deduce trends. In addition, there remained a great deal of scatter no matter how much we lowered our upper chi-squared limit. This forced us to develop more rigorous conditions that would qualify a data point as “bad.”
To find better conditions, we scrutinized the curves of best fit corresponding to these points and searched for patterns. Occasionally, we examined a record with plenty of data in all anodes and sectors that produced a plot like Figure 3.3. The x-axis is sequential energy bin steps that translate to time. The high-frequency wiggles in the fit curve account for the energy bin sweeping, and the low-frequency wiggles account for the physical spinning of the spacecraft relative to the bulk flow direction.
Figure 3.3 Here is an example of a record with a high number of counts above the background level. This scenario makes the data very smooth and easy to fit to, meaning the parameters that determine the fit curve have low uncertainties.
However, plots of this quality were far and few between because most records had fewer data points, making it more difficult for the fitting routine to generate a best-fit curve for them. As expected, the best-fit curves typically showed extreme deviation from the data points, causing the best-fit parameter values to be wildly unrealistic. However, very few points mean very few deviations, so the chi-squared values for these records were deceivingly small. This explained why lowering our acceptable chi-squared limit didn’t eliminate the scatter.
Figure 3.4 The scatter in the fit parameters is due to most of the counts plots looking like this. Notice how the fit curve peaks where there is nothing, one of the data peaks is one-sided, and most of the record is just background noise.
We also realized that the fit curve exhibits bizarre behavior when there is little data to fit to. As you can see from Figure 3.4, there are peaks that only have one side, and sometimes there are peaks that occur where there is no peak in the counts at all. This is a serious problem, considering the overwhelming majority of the records do not have many data points. However, Figure 3.4 also shows that a large portion of the data used for fitting a curve is just background noise, meaning that there is no real data there. If we want to minimize the probability that a fit curve will produce a false peak, we do not want our fitting routine to consider points that are too close to the background level. In light of this, we forced the fits to only consider the anode with the peak number of counts and the neighboring anodes on either side of it. Also, because the width of a Maxwellian peak corresponds to the plasma temperature, one-sided peaks cannot possibly represent actual physical conditions. Therefore, we had our program filter for one-sided peaks within a record, and if any were discovered, that record wasn’t considered for fitting. These conditions, however, were not adequate for filtering out enough “bad” data points to show clear trends in the plasma parameters. Up to this point, we had only imposed criteria to ensure that the fit curves were “physical,” which did not mean that the fit curves would follow trends that we believed to be present in the data. Nevertheless, we could get an idea of the accuracy of a parameter by analyzing the uncertainty produced by a chi-squared analysis. It is difficult to have confidence in a point that has an error that is greater than the value itself. Also, if a point has an error that is unrealistically small, it is probably due to a small number of points with counts above background in a record, rendering it as suspicious as a point with large error. From these principles, we decided that the error-to-value (ETV) ratio of a point is another good way to determine the “badness” of a fit. Upon reexamination of the individual fits, we decided that an upper ETV limit of 1 and a lower ETV limit of 0.01 were reasonable. After including all of these conditions, which we called the “pruning” process, we were successful in obtaining data points that showed clear trends.
Figure 3.4 Our pruning was able to filter out enough scatter to show trends more clearly.
Figure 3.4 allows us to see co-rotation behavior in the azimuthal velocity component (v-phi) and a peak in ion number density (n) near the radius of Io’s orbit. Beyond the radius, the ion number density decays as an inverse power law. These results indicated that our pruning was effective enough to begin analysis on the best-fit values rather than the counts data.
It might seem unusual that there are so few regions where the counts are high enough above background for our fitting routine to be applied, but this is due to the high background level in the region we’re studying. The high background level is caused by energetic electrons diffusing outward from Jupiter’s radiation belts and flooding the PLS detector.
3.3 Fitting Radial Profiles
Analogous to the method of fitting a model to the counts data with respect to energy-per-charge, we can fit straight lines to any one of the five plasma parameters with respect to radial distance (R). By plotting the fits on logarithmic axes, any exponential behavior can be transformed into linear behavior. We can fit the data to a power law curve, i.e. f(x) = axb, where “a” and “b” are constants that we can adjust to minimize chi-squared. However, if we take the logarithm of this type of equation, we get ln[f(x)] = ln(axb) = ln(a) + b*ln(x). Now we have an equation that is linear for the logarithms of the variables, which we can redefine as new variables. If we redefine the constants as well, we get the form y(x) = A + Bx. Using this equation, we applied simple linear regression algorithms to fit the logarithms of the plasma parameters as functions of ln(R). This method is preferable to other IDL fitting routines that use nonlinear models and typically return less accurate best-fit results and uncertainties. The difference between these methods is exhibited in Figures 3.5 and 3.6.
Although we could have fit the entire data set at once with respect to each plasma parameter and considered the determined behavior, this would have been too broad of an analysis. The data were taken over the course of nearly eight years, and there was plenty of temporal variability that could have been missed if we had not examined the data epoch by epoch. In the case of satellite data, a characteristic timescale that we used to define an epoch was an orbital period. This makes sense, because Galileo’s orbits were never shorter than a couple of months, allowing ample time for conditions to change between measurements. For these reasons, we decided to separate the data by orbit number with each orbit defined as the time when the satellite passed through apojove. Then we were able to fit straight lines to each of the plasma parameters for each orbit, giving 34 different slopes and intercepts to analyze for variation over time.
Figure 3.5 Best-fit lines made for orbit 6 using IDL CURVEFIT routine in linear space. Error bars were not included in these particular plots because they were not necessary to make the best-fit lines (we could have done this, but we were not confident enough in the uncertainties of the parameters). Significant figures were neglected in creating these plots.
Figure 3.6 Best-fit lines made for orbit 6 with IDL LINFIT routine in logarithmic space.
3.4 Two-dimensional extrapolation
We were able to develop a method for one-dimensional analysis of the fits, and they gave us a decent, if not somewhat abstract, picture of how the plasma was distributed in Jupiter’s inner magnetosphere (by its density). Since we wanted to extend this analysis to higher dimensions, we had to employ creative methods.
Unfortunately, Galileo never deviated more than a few degrees in latitude from the equator, so we have hardly any information on theta dependences. However, because the plasma tends to co-rotate with Jupiter, it experiences an associated centrifugal force that tends to fling it radially outward from the rotation axis. Additionally, a given field line in a dipolar field is farthest from the planet at the equator [Bagenal and Delamere, 2011]. Therefore, we expected the density to peak near the equator and decay above and below it. The density distribution as a function of height, z, away from the equator can be solved analytically for small z, and it is given by Hill and Michel :
n(z) = n0e-(z/H)^2 (3.3)
H = [2/3 kTi/(mpAi)]1/2 = H0[Ti/Ai]1/2 (3.4)
Here, = Jupiter’s rotation rate = 9.925 hours, mp = mass of a proton, H0 = .64 Rj, Ti is the ion temperature in eV, and Ai is the average ion mass in amu. With our best-fit lines for density, we could determine n0 with respect to R. Because we also had the temperature best-fit lines versus R, we could determine the scale height H as a function of R, assuming that the temperature didn’t vary too much with latitude. This extrapolation determined the density for all z and R, allowing us to create two-dimensional profiles of the plasma distribution, which is discussed in Section 4.5.
4.1 Demonstrating the Effects of Pruning
Figure 4.1 Fit values over entire Galileo mission, without pruning. Colors represent the local time sector in which each record was taken. Green = dawn (3 – 9 hrs), Red = noon (9-15 hrs), Violet = dusk (15-21 hrs), and Blue = midnight (21 hrs – 3 hrs).
Figure 4.2 Fit values with pruning for the entire Galileo mission.
By analyzing the fit curves in the individual counts data records, we were able to identify several recurring features that severely skewed the fits and caused them to return unrealistic plasma parameters. After we reprogrammed the fitting routine to correct for these features, the fits returned plasma parameters that were realistic enough for continued analysis. To ensure that these fits had actually improved, we analyzed a few more of the records that had survived pruning and were convinced that the fits were sufficiently close to the counts data. Even if we had not been able to obtain any significant scientific results from the data analysis, at least we were confident that our pruning process was effective enough to interpret bad data.
4.2 Factoring in Trajectory Information
It is easy to identify variability in plasma conditions over time without considering what causes those variations. Because the plasma also varies in space, we have to take into account the plasma effects at the position of the spacecraft before we could say much about the overall physical processes of the plasma. Therefore, we plotted Galileo’s trajectory information and the plasma parameters, which were obtained from the University of Iowa ephemeris website.
Figure 4.3 Plasma parameters and spacecraft trajectory with respect to time.
4.3 Other position variables contain additional information
Thus far, the data have only been analyzed with respect to radial distance, but there are several other spatial coordinates we could have used as independent variables. Since we wanted to obtain a comprehensive understanding of spatial variability, we plotted the plasma parameters with respect to each of these coordinates.
Figure 4.4 Keeping trajectory information juxtaposed with the plasma parameters helps us determine the sources of some effects.
Figure 4.5 Plasma parameters and trajectory with respect to spacecraft latitude
Figure 4.6 Plasma parameters and spacecraft trajectory with respect to local time.
4.4 Best-fit line variations with time
We calculated best-fit lines for the best-fit parameters vs. R in each orbit. Figure 4.7 is an example. This gave us 34 different slopes and intercepts for both density and temperature, allowing us to plot how these fit lines varied between orbits, which is called temporal variability and is shown in Figure 4.8.
Figure 4.7 Obtaining lines of best-fit for the temperature, density, and azimuthal speed with respect to R gave us a first-order measure of these parameters’ spatial distributions.
Figure 4.8 Temporal variations in the best-fit lines. From the top plot downward, the plots show the temperature best-fit line’s y-intercept (with units of eV), the temperature best-fit line’s slope (unitless), the density best-fit line’s y-intercept (units of m-3), and the density best-fit line’s slope (unitless). TA and TS stand for “temperature amplitude” and “temperature slope,” and NA and NS follow the same convention for density.
The y-intercepts for the best-fit lines varied dramatically because not all orbits had the same coverage. For example, as seen in Figure 4.9, Galileo only took measurements between 5 and 8 Rj during orbit 0 (orbital insertion), so the behavior of the plasma beyond 8 Rj could not be incorporated. This caused the fit lines in this orbit to be more skewed than those in other orbits with more coverage and demonstrates the need to fit the density to multiple lines with different intercepts and slopes in order to account for all behavior between 5 and 30 Rj [Bagenal and Delamere, 2011]. The slopes of the fit lines showed less temporal variation than the y-intercepts, having average values of (2 + 1.2) for the temperature and (-6 + 2.5) for the density. This is in agreement with other reports of ion temperature increasing with radial distance in the Io plasma torus [Bagenal et al., 2007]. These averages were used to construct a two-dimensional profile of the plasma density, which is shown in Figure 4.10.
Figure 4.9 Orbits with small coverage ranges can have very different best-fit lines than orbits with wider coverage.
Figure 4.10 Two-dimensional profile created from the averages of the best-fit slopes and amplitudes for both temperature and density.
4.5 Best-fit lines for each local time over whole mission
Krupp et al.  have suggested, using energetic particle data, that there should be a strong asymmetry between the azimuthal flow speeds (v-phi) in the dawn and dusk sectors on the order of ~100 km/s. If this is true, we should be able to see significant differences in the best-fit lines for v-phi, if we separate data points by local time sectors. Figures 4.11-14 show the best-fit lines for dawn, noon, dusk and midnight, respectively.
Figure 4.10 Best-fit lines for all mission data recorded in the dawn sector.
Figure 4.11 Best-fit lines for all mission data recorded in the noon sector.
Figure 4.12 Best-fit lines for all mission data recorded in the dusk sector.
Figure 4.13 Best-fit lines for all mission data recorded in the midnight sector.
In these plots, we can see slight differences in the best-fit lines for v-phi of each sector. However, as seen in previous plots, these differences are probably due to coverage differences rather than physical asymmetries. No consistent asymmetry on the order of 100 km/s was found in v-phi, and the largest uncertainties in the fits are on the order of ~10 km/s. Therefore, the data seem to refute the results of Krupp et al. . Nevertheless, we can still look for asymmetries in density or temperature. To illustrate the potential asymmetries, we again constructed a two-dimensional density profile for each local time sector. Figure 4.15 shows the density profiles corresponding to each local time sector.
Figure 4.15 Two-dimensional density profiles constructed from the best-fit lines for each local time sector over the entire Galileo mission. From left to right, the top two plots correspond to dawn and noon, and the bottom two plots correspond to dusk and midnight.
From these profiles, there appears to be no significant variation in the density or temperature (which affects the width of the distribution) with respect to local time.
Data analysis is a complex endeavor. One has to align the measurements with expected results without discarding information that could potentially contradict expectations. After all, most scientific breakthroughs begin with measurements that contradict popular belief. To make matters worse for the Galileo PLS data, most of the records had little to no counts above background, so there were few data points to fit. Eliminating bad points would have left too few points to indicate trends. Therefore, we had to tailor our pruning to sift out the bad parts of individual records instead of eliminating whole records. Once we fit trend lines to the points that survived pruning, we began to see theoretically predicted behavior.
We did not detect any significant variations in the plasma conditions over the time of the Galileo mission, but this conclusion could have been different if each orbit had had equal coverage. The spatial variations in the plasma parameters followed trends that agreed with [Delamere and Bagenal, 2005] and [Bagenal et al., 2007], although we did not find any local time asymmetries comparable to what was found by Krupp et al. .
Bagenal, F. (1992), Giant planet magnetospheres, Annual Review of Earth and Planetary Sciences, 20, 289-328.
Bagenal, F. and P.A. Delamere (2011), Flow of mass and energy in the magnetospheres of Jupiter and Saturn, Journal of Geophysical Research, 116, A05209, doi:10.1029/2010JA016294
Bagenal, F., T. E. Dowling, and W. B. McKinnon (Eds.) (2007), Jupiter: The Planet, Satellites and Magnetosphere, Vol. 1, Cambridge University Press, Cambridge, United Kingdom.
Bedingfield, Keith L., Richard D. Leach, and Margaret B. Alexander (1996), Spacecraft System Failures and Anomalies Attributed to the Natural Space Environment, National Aeronautics and Space Administration, Marshall Space Flight Center, Huntsville, Alabama, USA.
Bevington, Philip R., and D.K. Robinson (2003), Data Reduction and Error Analysis for the Physical Sciences, 3rd ed., McGraw-Hill, New York, New York, USA.
Boteler, D. H., R. J. Pirjola, and H. Nevanlinna (1998), The effects of geomagnetic disturbances on electrical systems at the Earth's surface, Advances in Space Research,22, 17-27.
Chapman, S., and V.C.A. Ferraro (1933), A new theory of magnetic storms, Journal of Geophysical Research, 38, 79-96, doi:10.1029/TE038i002p00079
Delamere, P. A., and F. Bagenal (2005), Radial variations in the Io plasma torus during the Cassini era, Journal of Geophysical Research, 110, A12223, doi:10.1029/2005JA011251
Finn, J.M. and E. Ott (1988), Chaotic flows and magnetic dynamos, Phys. Rev. Lett.;(United States), 60, doi:10.1103/PhysRevLett.60.760
Fletcher, L. (2003), The costs and benefits of space weather, Contemporary Physics, 44, 451-453, doi:10.1080/0010751031000135968
Hill, T.W. and F.C. Michel (1976), Heavy ions from the Galilean satellites and the centrifugal distortion of the Jovian magnetosphere, Journal of Geophysical Research, 81, 4561-4565, doi:10.1029/JA081i025p04561.
Krupp, N., A. Lagg, S. Livi, B. Wilken, J. Woch, E.C. Roelof, and D. J. Williams (2001), Global flows of energetic ions in Jupiter’s equatorial plane: First-order approximation,Journal of Geophysical Research, 106, 26017-26032, doi:10.1029/2000JA900138
Olsen, Nils, G. Hulot, and T. J. Sabaka (2010), Measuring the Earth’s magnetic field from space: concepts of past, present and future missions, Space science reviews, 155, 65-93.
Parker, E.N. (1958), Dynamics of the Interplanetary Gas and Magnetic Fields,Astrophysical Journal, 128, 664-676.
Stern, David P. (1989), A brief history of magnetospheric physics before the spaceflight era, Reviews of Geophysics, 27.1, 103-114.
Taylor, J., Cheung, K. M., and Seo, D. (2002), Galileo telecommunications, DESCANSO Design and Performance Summary Series, National Aeronautics and Space Administration, Marshall Space Flight Center, Huntsville, Alabama, USA.
Vasyliūnas, Vytenis M. (2011), Physics of magnetospheric variability, Space Science Reviews, 158, 91-118.
Localization of Extracellular Regulated and Calcium-Dependent Calmodulin Protein Kinases […], Dora Panyi
Historically, stress habituation research has largely focused on the reduction of the hypothalamic-pituitary-adrenal axis response to repeated stress exposure. The habituation of the HPA axis to prolonged stress is caused, in part, by negative feedback mechanisms, such as the activation of extensive stress-related neural circuitries and more complex enzymatic cascades underlying associative memory mechanisms. The results of previous investigations have implicated the MAP kinase and CaM kinase pathways as potential signal transduction pathways that are activated in stress adaptation. In addition, the amygdala and the hippocampus are areas where learning and memory occur in the brain. However, a recent study has shown that the posterior hypothalamus may also be a critical site for simple forms of learning like habituation. Building on existing research, this project seeks to determine if microtubule associated protein kinases 1/2 (ERK) and calcium-dependent calmodulin kinase II alpha (CaMKIIα) are present in cells characterized as active by stress-induced c-Fos protein expression in the posterior hypothalamic (PH) nucleus. In the following series of studies, we demonstrate with immunohistochemistry that ERK and CaMKIIα but not Elk-1 can be localized in the stress-responsive neurons of the posterior hypothalamus. Immunohistochemistry was used to measure the functional activation of ERK by assessing the level of phosphorylation (pERK) and the presence of CaMKII in the posterior hypothalamus of acutely stressed rats. The phosphorylation of ERK is time-dependent; basal expression was observed in control animals, but a significant increase was observed in rats exposed to 15-30 minutes of stress. It has been shown that Fos protein is expressed in active pathways in the brain, and we found that pERK but not CaMKII is significantly co-expressed in the same cells with Fos. These findings suggest that ERK is present in neurons that are thought to be signaling during stressful exposures. ERK could also be involved in long-term plastic changes underlying habituation to repeated stress.
I would like to express my greatest gratitude to the people who have supported me throughout this project beginning with the members of the Campeau Lab who have taught me the basics of neuroscientific research, lifted the monotony of work in the dungeon, and who are happy to see me succeed now. It was an honor working with them.
This work could not have been accomplished without the social support of my family and closest friends. I am lucky to have a wonderful family who will support me no matter what. Their accomplishments and inexhaustible encouragement have shaped my strong-willed personality that was absolutely necessary in the completion of this project. As for my friends, they have been there for me at every step along the way. I will cherish the experiences we have shared and will share together in the future.
I would like to extend my gratitude to the members of my thesis committee for their constructive criticism, love of science and academia, and time devoted to hearing this finished project.
Most of all, I would like to thank my mentor and principal investigator of the Campeau Lab, Dr. Serge Campeau, who gave me the opportunity to become an undergraduate researcher as a sophomore and has helped me receive three research scholarships over the years. Without his tremendous guidance, insight, and invaluable motivation in times of distress due to fruitless research, this project would not have been possible.
1.1 The Phenomenon of Stress Habituation
Habituation is a widespread phenomenon observed in most species and is defined as a reduction in response amplitude to a stimulus upon repeated exposure to the same stimulus. While habituation is considered the simplest form of learning and is exhibited ubiquitously in vertebrates and invertebrates, the underlying cellular mechanisms mediating this process have yet to be fully understood (Esdin, Pearce & Glanzman, 2010). Habituation is established from most classes of stimuli, including stressors like noise exposure or restraint. These stressors also elicit habituation, leading to reduced stress-elicited responses following repeated stress exposure (Grissom & Bhatnagar, 2008).
There are several measures of stress-elicited responses in rats that can be observed following habituation to stress. Among these measures are the release of the hypothalamo-pituitary-adrenocortical axis (HPA) hormones, which are adrenocorticotropin hormone (ACTH) and cortisol (or corticosterone in rodents). Other measures include the adrenomedullary hormones, which are epinephrine and norepinephrine, and various autonomic responses such as tachycardia, blood pressure, hyperthermic core body responses, and several other behavioral responses.
The neuronal circuitry associated with psychological stress responses involves the limbic system. Psychological stressors first activate the limbic system, which projects to several subcortical areas such as the Preoptic Area (POA), the dorsomedial hypothalamus (DMH), and the bed nucleus of the stria terminalis (BNST). Theses subcortical areas then relay the sensory information to the medial parvocellular paraventricular nucleus of the hypothalamus (mpPVN) to initiate HPA axis response to stress. The limbic system is made up of interconnected nuclei such as the amygdala and hypothalamus and hippocampus, which are critical for motivated behaviors, emotions, and long-term memory. Of these limbic-associated structures, the hypothalamus is an important effector node for responses regulated by the limbic system (see Figure 1). Responses to psychological stressors can be innate, like a rat’s fear of predators (cats, ferrets, etc.), or alternatively they can be conditioned, such as the classical conditioning of weak visual stimuli and acoustic stimuli paired with electrical shocks. One brain region that is responsible for the formation of contextual memory in response to both innate and conditioned stressors is the amygdala. When activated by an environmental stressor, the amygdala will send signals to several subcortical relay nuclei.
Fig.1 The ventral subiculum and medial prefrontal cortex excite the BNST/DMH/POA through glutamatergic projections upon GABAergic neurons of the BNST, which in turn inhibit the mpPVN through activation of GABAergic neurons. The basolateral amygdala excites the medial and central amygdala, which excite the mpPVN through disinhibition of the BST, DMH and POA (Jankord & Herman, 2008).
1.2 The posterior hypothalamus as a region important in stress habituation
It has been shown that inactivation of the posterior hypothalamic nucleus/dorsomedial hypothalamus region with the GABAA receptor agonist, muscimol, reduces many acute stress-elicited responses when injected immediately prior to stress (Nyhuis, Day & Campeau, 2012). Muscimol also disrupts stress habituation when injected prior to repeated stressor exposure. In addition, vehicle-injected animals repeatedly exposed to loud noise showed corticosterone habituation, which was attenuated following repeated injections of muscimol prior to each stress exposure (Nyhuis, Day &Campeau, 2012).
The rostral Raphe Pallidus (rRPa) and the paraventricular hypothalamic nucleus (PVN) have previously been shown to be active during acute stress. The posterior hypothalamic region of acutely stressed rats that are injected with retrograde tracers in the rRPa and/or the PVN displayed the highest levels of co-localization with the immediate early gene protein product Fos. This indicates that the posterior hypothalamic nucleus projects to these two areas, further supporting its involvement in acute responses to stress. The posterior hypothalamus also receives sensory information from the auditory thalamus, creating a connection between auditory processing and a region that could be critical in mediating stress responses (Campeau & Watson, 2000). Combined with additional anatomical evidence indicating that the posterior hypothalamic nucleus is innervated by several sensory regions and projects to many brain regions that regulate neuroendocrine and autonomic responses elicited by stress, this region is ideally situated to modify its input-output relationship through synaptic modifications and provide the basis for habituation to stress (Nyhuis, Day & Campeau, 2011).
1.3 Stress Habituation and Synaptic Plasticity – LTP & LTD as possible mechanisms
Adaptation to stress at the cellular level involves the process of synaptic plasticity, which is the strengthening or weakening of a synapse between two neurons due to use or disuse. The flexibility of synapses to undergo morphological and functional changes is central to understanding learning and memory. Synaptic strengthening occurs via an extension of dendritic spines, the insertion of excitatory receptors, and/or activation of silent synapses of the post-synaptic cell, which magnifies signaling between connected neurons. The weakening of the synapse occurs through the internalization of receptors from the membrane and the retraction of dendritic spines.
The leading candidate mechanisms of synaptic modifications associated with a variety of learning and memory processes include long-term potentiation (LTP) and long-term depression (LTD) (Malenka & Bear, 2004). Preliminary results indicate that posterior hypothalamic nucleus neurons projecting to the autonomic nucleus that controls tachycardia and hyperthermia (rRPa) contain mostly glutamate as their neurotransmitters (H. Day & S. Campeau, unpublished results). However, several additional neurons of the posterior hypothalamus contain the neurotransmitter GABA. Conceivably, LTP- or LTD-like mechanisms could mediate the development of response reductions upon glutamate or GABA-containing neurons at the level of the posterior hypothalamic nucleus.
Several intracellular signaling cascades and their associated enzymes have been reported to mediate synaptic plasticity via LTP or LTD-like mechanisms. One of the effects of LTP in several brain regions is phosphorylation, increased insertion, and increased average current that is mediated through α-amino-3-hydroxy-5-methyl4-isoxazole propionic acid (AMPA) subtype of excitatory amino acid receptors. Once the post-synaptic membrane is depolarized through AMPA receptors in response to a stimulus, magnesium-gated NMDA receptors can be activated to further depolarize post-synaptic elements and allow the influx of calcium, which serves as a second messenger associated with the strengthening or weakening of synapses. This effect is mediated through a signaling cascade involving the phosphorylation/activation of several enzymes, including calcium/calmodulin-dependent protein kinase II (CaMKII), protein kinase A (PKA), or protein kinase C (PKC). The signaling cascade eventually leads to the phosphorylation of the transcription factor, cAMP-response element binding protein (CREB), which is necessary for the effects observed on post-synaptic AMPA receptors (Kennedy, Beale, Carlisle, & Washburn, 2005; Malenka & Bear, 2004). Interestingly, LTD has been associated with several dephosphorylating processes, including AMPA receptor subunits and CREB, in addition to the phosphorylation of the transcription factor Elk-1 (Thiels, Kanterewicz, Norman, Trzaskos, & Klann, 2002). The removal of AMPA receptors from the membrane and the consequent reduction in the amplitude of post-synaptic depolarization leading to decreased signal transduction provides a probable mechanism for explaining the response reduction associated with adaptation to stress. However, this current explanation has not been proven.
As mentioned above, there are several protein kinases and transcription factors that play a part in the phosphorylation and dephosphorylation of AMPA receptors enhancing plastic changes. If these kinases play a role in habituation, they would be expected to be present in posterior hypothalamic neurons, to display some increases in functional activity following a stressor, and to decrease in functional activity with repeated exposure to a stimulus. In an effort to further understand the plastic processes underlying the development of stress habituation, the localization of kinase enzymes that are associated with a variety of plastic processes at the level of the posterior hypothalamic nucleus was investigated.
1.4 Role of CREB, CAMKII, pERK and Elk-1 in Synaptic Plasticity
Multiple signal transduction pathways have been implicated in mediating habituation, learning, and memory formation, including calcium/calmodulin-dependent kinases, mitogen activated protein kinases (MAPK), protein kinase A, and protein kinase C, among others. These pathways converge on the transcription factor CREB, which is conducive to plastic changes when active (Barco, Jancic & Kandel, 2008). There is an extraordinary level of interaction and convergence between these transduction pathways, increasing the complexity of synaptic plasticity. Due to this overlap, events in one pathway can activate enzymes in another, leading to the simultaneous induction and propagation of these signaling cascades. A fine integration of these pathways is and should be necessary for functional alterations of neurons. Due to this complexity, we are focused on the first two pathways, the CaMK and MAPK signaling pathways (Sweatt, 2004).
In order to understand how cAMP-dependent synaptic plasticity by the MAPK and CaMK pathways occurs, it is important to illustrate the role of CREB in this process. There are several signal transduction pathways that converge on the phosphorylation of CREB, but the two of interest are the CaMK and MAPK pathways (see Figure 2). While the interaction between CREB and CaMKII remains largely unresolved, evidence shows that increased levels of activated CREB (phosphorylated or phospho-CREB) and CaMKII are the causes of the generation of silent synapses in plasticity. Through the regulation of CaMKII, CREB can augment AMPA receptor density in the synapse (Marie, Morishita, Yu, Calakos, & Malenka, 2005). Aplysia phosphoCREB (ApCREB) controls the transcription of several immediate, early-response genes that play a role in habituation, such as the expression of ubiquitin hydrolase in aplysia, which regulates the regulatory domain of PKA. PKA constrains long-term facilitation. CREB also controls the production and translocation of mRNA that is critical for the generation of new synaptic connections branching from the synapse that initially received stimulation. Therefore, gene expression regulated by phospho-CREB contributes to the growth of new synapses and the stabilization of existing synaptic connections (Barco, Jancic & Kandel, 2008).
Fig 2: External stimuli, such as stress, regulate the phosphorylation of CREB via converging protein-kinase pathways.
1.4.2 CaM Kinase
CaMK is part of a GαQ-G protein-coupled receptor-signaling pathway. When a ligand binds to a metabotropic receptor, a conformational change occurs in the associated Gq protein, causing a guanosine diphosphate to be exchanged for a guanosine triphosphate and the alpha subunit (GTP coupled) to dissociate from the βγ subunits. The GTP-Gq complex then binds to an inactive membrane-bound phospholipase C, activating it and causing the cleavage of phosphatidylinositol diphosphate (PIP2) into inositol triphosphate (IP3) and diacylglycerol (DAG). IP3 is released into the cytosol, where it triggers Ca2+release from the endoplasmic reticulum. DAG remains in the membrane, where it activates protein kinase C (PKC). After intracellular calcium levels rise, calmodulin binds four calcium ions and changes its conformation to form an active calcium/calmodulin complex. The two globular ends of the complex attach to the inhibitory domain of a target protein, in this case CaMKII (see Figure 2). Upon calmodulin binding, the kinase engages in autophoshorylation that perpetuates its activity even after calcium levels drop. Once phosphorylated/activated, CaM Kinase will catalyze the phosphorylation of substrate proteins by transferring phosphate groups from adenosine triphosphate (ATP) to the hydroxyl (OH) groups of serine and threonine residues in the substrate proteins. The addition of an electrically charged phosphate group induces a conformational change and activates the biological activity of the target proteins (Nestler, Hyman & Malenka, 2009).
Fig 3: Ca2+-calmodulin-dependent protein kinase signaling pathway (Soderling, 1999)
The CaM kinase pathway is thought to underlie neuronal plasticity associated with learning and memory through the activation and deactivation of CREB by CaMKII and calcineurin (protein phosphotase 2B), respectively (Barco, Jancic & Kandel, 2008), and through interaction with AMPA receptors. Tetrameric AMPA receptors are comprised of four different subunits; the phosphorylation or dephosphorylation of GluR1 and GluR2 has been linked to LTP and LTD. Upon LTP induction, CaMKII attached to membrane-bound NMDA receptors is known to recruit the protein, SAP97, a member of the membrane-associated guanylate kinase (MAGUK) protein family. SAP97 has been shown to interact with the C-terminal tails of the GluR1 subunit, facilitating the insertion of cytosolic AMPA receptors into the postsynaptic density. As mentioned before, the addition of AMPA receptors into the synapse contributes to the strengthening of synaptic communication and dendritic arborization. This effect indicates that LTP/LTD is dependent on CaM kinase (Lee, 2006). Here we aim to investigate whether CaMKII is expressed in posterior hypothalamic neurons that also display stress-induced increases in the immediate early gene c-Fos. Investigation of CaMKII could illuminate its putative role in the plasticity underlying habituation to repeated stress exposure. In order to do this, immunohistochemical detection of CaMKII was performed following an acute exposure to loud noise stressors.
1.4.3 MAP Kinase
The mitogen-activated protein kinase (MAPK) pathway has polymodal activation. Calcium can directly trigger the activation of protein kinase C, which will activate rat sarcoma protein (Ras) or indirectly trigger protein kinase A to activate Ras-related protein 1 (Rap 1). Ras can also be activated through neurotropic factors binding to receptor tyrosine kinases. Once these second and third messengers have been activated, Ras will phosphorylate Raf-1 and Rap 1, which will phosphorylate B-Raf and eventually lead to the phosphorylation and activation of MEK and ERK (Figure 3). Finally, the kinase cascade activates the activation protein 1 (AP-1) family of transcription factors. The most well known AP-1 transcription factors are Jun and Fos, which heterodimerize and move to the nucleus to affect transcriptional changes (Nestler, Hyman & Malenka, 2009).
Fig 4: Mitogen activated protein kinase pathway (Thiels & Klann, 2001).
Learning and memory deficits from Ras/ERK hyper activation in a mouse model of neurofibromatosis demonstrate that healthy functioning of ERK is necessary for synaptic plasticity associated with learning (Sweatt, 2004). The activation of the ERK signaling pathway via acetylcholine-activated hippocampal muscarinic receptors shows that without sufficient ERK phosphorylation, the induction of LTP does not happen. Along with evidence that ERK blockers impede the progression of LTP, it can be concluded that ERK is critical to the induction of LTP (Giovannini, 2006). Since LTP is one of the candidate mechanisms underlying adaptation to stress, ERK expression in the PH could mediate long-term synaptic plasticity.
A rise in post-synaptic levels, the translocation to the nucleus, and phosphorylation of apCREB-2 by pERK implicate ERK in long-term facilitation (LTF). These changes were observed in the gill-withdrawal reflex in aplysia and in cell cultures of LTF-inducing sensory cells that were subjected to repeated 5-HT applications (Michael, Martin, Seger, Ning, Baston & Kandel, 1998). Long-term facilitation (usually used in reference to aplysia) and long-term potentiation (used in reference to mammals) are phenomenologically the same in that they both increase synaptic efficacy. The fact that transcriptional regulators and a cell adhesion molecule (apCAM) both contain consensus phosphorylation sites for ERK supports the role of ERK in synaptic plasticity in habituation. These experiments, combined with the finding that LTF is blocked by a MEK antagonist directly responsible for the activation of ERK, further support the possible causal relationship between ERK and plasticity (Thiels & Klann, 2001). Observed increases in ERK2 mRNA levels after the establishment of NMDA receptor-dependent LTD and the subsequent CREB-dependent transcription seem to indicate a link between ERK cascades and CREB-dependent transcription. Similarly, elevated expressions of B-Raf and Raf-1 (found upstream in the ERK activation cascade) after LTP induction in dentate gyrus cells consistently demonstrate that ERK is mediating post-LTP gene expression. Therefore, based on these findings, our second aim was to detect the presence of the active form of ERK, known as phospho-ERK (pERK) in the posterior hypothalamus, especially in neurons also expressing stress-induced Fos protein elevations.
1.4.4 p-ELK-1 upregulated in LTD
Elk-1 is a transcriptional activator that is responsible for the transcription of c-Fos. Elk-1 seems to be a downstream target of ERK in the MAPK pathway (Tian & Karin, 1999). Elk-1 has a transcriptional binding domain on the C terminal tail, which has a conserved consensus sequence for MAP kinases (Cruzalegui, Cano & Treisman, 1999). Findings that demonstrate the occurrence of synaptic plasticity via Elk-1/CREB interaction following traumatic brain injury suggest that ERK may be responsible for transcriptional changes induced by CREB through Elk-1 (Hu, Liu, Bramlett, Sick, Alonso, Chen & Dietrich, 2004). From these findings, we hypothesize that stress habituation is associated with Elk-1 activation and propose to look for the expression of the activated form of this protein, phospho-Elk-1 (pElk-1), in posterior hypothalamic neurons also expressing stress-evoked Fos induction.
1.5 Fos as a marker of activated neurons
Responses to acute stress exposure at the cellular level can be measured by looking at levels of the transcription factor Fos. C-Fos is an immediate early gene (IEG) that implicates Fos as a marker for activated neurons. An IEG is a gene that responds to a wide variety of stimuli before the synthesis of new proteins. An IEG has a low level of expression under basal conditions and rapid induction in response to extracellular signals. C-Fos mRNA is detectable a few minutes after transcriptional stimulation, peaks within 30-60 minutes, and diminishes after two to three hours. C-Fos gene expression results from the rapid phosphorylation of CREB (cAMP response element binding protein) via protein kinase A and CaM kinase. Due to the low level of c-Fos expression under basal conditions and its sizable induction by a variety of stimuli, including various stressors, c-Fos has been extensively employed as a tool for specifying neuronal pathways displaying increased activity in the behavioral and neuroendocrine responses induced by stress. Brain regions that are shown to express c-Fos mRNA in response to stimuli include the lateral septum and anterior BNST (Kovacs, 1998). Since ERK and CaMKII have been closely associated with a variety of plastic processes and because posterior hypothalamic neurons appear to be necessary in the mediation of habituation to stress, the current study was designed to determine if ERK and CaMKII are possible transduction systems in the cells of the PH that are known to be active through Fos due to loud noise stressors.
Behavioral Procedure: Male Sprague–Dawley rats (n=14) were housed in groups of four in colony facility cages and were provided with food and water. In the afternoon before the experiment, the animals were transferred to individual cages and taken to the testing room, where they were placed in noise boxes to adjust to the new environment. This was necessary so that molecular, hormonal, and cardiovascular changes in response to stress would be attributable to the stressor and not to the novel environment or transportation. Each acoustic chamber was double wooden, ventilated, and contained a red fluorescent light. The intensity level was 100 decibels (A scale) and the exposure time was 30 minutes for the experimental group on the morning of the experiment. After the loud noise exposure, the animals remained in the chambers for an additional 15 minutes. The control group stayed in the chamber for the same duration of time. The duration of exposure and the post-stress interval was chosen to maximize the detection of loud noise-induced ERK phosphorylation, which is relatively quick in response to stress exposure, and the detection of loud noise-induced Fos, which requires a longer elapsed time for expression (Masini, Babb, Nyhuis, Day & Campeau, 2012).
Tissue processing and sectioning: For the first experiment, the animals (n=4) were anesthetized with sodium pentobarbital (Fatal Plus) and transcardially perfused. For the second (n=6) and third experiment (n=4), following the noise or control exposures, the animals were decapitated and their brains were quickly removed. The brains were blocked on ice and placed in vials to be soaked in fixative solutions. We used this alternate fixation procedure for the second and third experiments to prepare the tissue instead of the traditional aortic perfusions, because our first experiment yielded strong expression of phospho-ERK (pERK) in the control rats, which could have been caused by the lengthy perfusion manipulations. In order to obtain more consistent results and lower basal values, a faster tissue isolation method was used in subsequent experiments. The brains were immersed in sodium acetate-buffered paraformaldehyde (4% PFA/0.1M sodium acetate, pH=6.5) and chilled for ~6-8 hours at 4º C on a rotating platform. A pH of 6.5 was chosen because it provides an ideal permeation rate of the PFA into brain tissue (Khan & Watts, 2004). After eight hours, the brains were removed from the initial solution, blotted dry, and transferred to a solution of sodium borate buffered PFA (8% PFA/0.2 M sodium borate, pH=9.5) and soaked for four days at 4º C while shaking. The brains were transferred to a sodium phosphate buffered glycerol solution (0.1 M sodium phosphate/20% glycerol solution, pH=7.4) and incubated for two days at 4o C while shaking until they were completely saturated and had no buoyancy. The brains were flash frozen in -30o C isopentanes, wrapped in foil, and stored in -80º C freezer until sectioning. Sectioning was performed using a cryostat (Leica 1450). The brain blocks were mounted using M1 and were cut into 30 μm coronal sections with no roll plate at -28º C. The sections were deposited into 6-well plates filled with cryoprotectant solution.
Immunohistochemistry: In order to determine the titration of antibody that should be used to label cells, immunohistochemistry was performed on sections of the brains of acutely stressed rats (n=4). The brain sections were predominantly chosen from the hypothalamic paraventricular nucleus (PVN), the bed nucleus of the stria terminalis (BNST), and lateral septum, since these areas had been shown to display pERK, CaMKII, and Fos induction in response to stress. On Day 1, the tissue was washed in 6x5’ in 1xPBS (0.1 M (10x) PBS: 80 g NaCl, 2 g KCl, 2 g KH2PO4, 11.5 g Na2HPO4x7H2O dissolved in 1 L of Milli-Q H2O, pH=7.4). The tissue was incubated in hydrogen peroxide (0.3%) for 15 minutes at room temperature while shaking on a rotating plate. The tissue was then washed in 6x5’ in 1xPBS. Next, it was incubated in Avidin blocking reagent for 20 minutes at room temperature, shaking, followed by 1x5’ in 1xPBS wash. The tissue was incubated in Biotin blocking reagent (Avidin/Biotin blocking reagent kit; Vector Laboratories Inc., Burlingame, CA) for 20 minutes in a similar manner to the Avidin, followed by 1x5’ in 1xPBS wash. Then the tissue was pre-incubated in a buffer (1% bovine serum albumin, 0.5% Triton-X 100, 0.25% Carrageenan-Lambda) and dissolved in PBS (0.01 M immune buffer) for one hour at room temperature. The sections were transferred into the primary antibody that had been appropriately diluted in immune buffer and allowed to react for 48 hours at 4º C while shaking. On the last processing day, the sections were washed 6x5’ in 1xPBS and incubated for two hours in the secondary antibody, which was diluted 1:200 in immune buffer. The tissue was washed in 6x5’ in 1xPBS and incubated for two hours in an Avidin-Biotin Complex (ABC kit VECTASTAIN Elite Kit, Vector Labs) for HRP conjugated complex, which was diluted 1:1000 in immune buffer and allowed to cross-link for at least 30 minutes before usage. Sections were then washed in 6x5’ in 1xPBS and prepared for the diaminobenzidine (DAB) reaction (5.0 mL Milli-Q H2O, 2 drops buffer, 2 drops H2O2, 4 drops DAB), which would produce a dark brown product that was visible under light microscopy.
Double fluorescent immunohistochemistry: To test for co-expression of Fos/CaMKIIα and Fos/pERK, immunohistochemistry was performed. On Day 1, brain sections of the PVN, lateral septum, bed nucleus of the stria terminalis, and posterior hypothalamus were selected and washed 6x5’ in 1x phosphate buffered saline, followed by a one hour pre-incubation in block buffer (5% normal donkey serum, 0.5% TritonX-100 dissolved in 1xPBS). The sections were washed again 6x5’ in 1xPBS and incubated in the first primary antibody (diluted in immune buffer, 2.5% normal donkey serum, 0.5% TritonX-100 dissolved in 1xPBS) for 48 hours at 4º C while shaking on a platform. Rabbit anti-Fos (antibody #SC-42; Santa Cruz, CA) was used at a 1:6000 dilution for the most optimal results. On Day 2, the sections were removed from the antibody, washed 6x5’ in 1xPBS, and incubated in the first secondary antibody (1:200 dilution) for two hours at 4º C while shaking. A donkey anti-rabbit antibody conjugated to AlexaFluor 549 (Jackson ImmunoResearch Labs Inc., WestGrove, PA) was used to detect Fos. Sections were hidden from light with an aluminum foil cover after application of the first secondary antibody, since this antibody was fluorescent and therefore sensitive to light. Immediately following the first secondary antibody incubation, sections were washed 6x5’ in 1xPBS, soaked in 4% PFA for 10 minutes, washed again 6x5’ in 1xPBS, and incubated in the second primary antibody for 48 hours at 4º C. CaMKIIalpha (Calcium/Calmodulin dependent protein kinase – ThermoFisher; #MA1048) in a 1:2000 dilution or an anti-pERK antibody (Cell Signaling; #9101S) in a 1:4000 dilution was used for optimal results. On Day 3, the sections were washed 6x5’ in 1xPBS and incubated in the second secondary antibody for two hours (1:200 dilution) while shaking. The sections were washed again 6x5’ in 1xPBS and floated under weak lighting conditions in a PBS /gelatin mounting medium (0.5% gelatin, 0.05% chromium potassium sulfate in Milli-Q water) on superfrost plus slides. The sections were coverslipped in mounting medium (Vector Laboratories; Vectashield Hard Set Mounting Medium with DAPI #H-1500) and stored in the refrigerator at 4º C until microscopic analysis.
Fluorescent microscopic analysis: A ZEISS fluorescent microscope was used to conduct cell counts with cell imaging by AxioVision. Four circular areas around the third ventricle in the posterior hypothalamus were imaged at 20x magnification (Figure 4). The secondary antibodies had different fluorescent characteristics (AlexaFluor 488 vs. 549 or 594) that were detected with the combination of dichroic filters on the Zeiss epifluorescent microscope.
Fig 5: The coronal slice depicting the anatomical region around the anterodorsal posterior hypothalamus. The red box shows where cell counts were made in four different fields surrounding the dorsal third ventricle. The area (Bregma = -3.12) at this level is the posterior hypothalamus.
3.1 Experiment 1: Finding titres for pELK-1, pERK and CaMKIIα while looking at expression in the PH
When antibody dilutions allow for the maximum amount of antibodies to be associated with antigen and when those concentrations can be accurately detected, the measured values are called titres. The optimal titres for the antibodies tested were determined by IHC (Table 1).
CAM Kinase II alpha
Table 1: Shown above are the optimal concentrations of antibodies used for the detection of Fos, p-ELK, pERK and CaM kinase.
IHC was performed on tissue to test for pElk-1, which has been implicated to activate the transcription factor CREB. However, pElk expression could not be robustly or consistently observed. The induction of IEG Fos was minimal in controls and highest at 30 minutes. Strong pERK and CaMKII expression was observed at the level of the posterior hypothalamus. Further experiments focused on these last two protein kinases.
3.2 Experiment 2: Activation of pERK, but not CaMKII is time-dependent
To determine whether the activation of ERK is time-dependent, we measured pERK concentrations in animals that were exposed to 15 minutes of noise and 30 minutes of noise as well as a control group (n = 2/group). For pERK cell counts (Table 2), a repeated measures ANOVA was run with the number of sections counted as a within-subjects variable (2) and group counted as a between-subjects variable (control, 15 min, 30 min noise). As expected, no differences were observed between sections within animals or an interaction with group. However, there was a significant effect of group (F(1,3)=9.72 p<0.05) as displayed in Figure 5. A comparison test on the pERK cell counts across groups revealed that the 15 minute noise groups were different from both control and 30 minute noise groups (p < 0.05). The control and 30 minute noise groups were not different from each other (p > 0.05). There was no effect from any factors on CaM kinase cell counts. Due to problems with tissue availability, the 15 minute time point for CaMKII was not investigated.
15 min noise
30 min noise
Table 2. Protein kinase levels measured by immunohistochemical DAB reactions. PERK levels rise at 15 minutes and return back to baseline levels at 30 minutes of noise exposure. There is no significant difference between the control and 30 minute noise group CaMKII levels.
CONTROL 15 MIN NOISE 30 MIN NOISE
Fig. 5. The pictures depict the area around the third ventricle. PERK levels are at baseline levels in control animals, peak at 15 minutes of noise exposure, and regress at 30 minutes.
3.3 Experiment 3: Co-expression of Fos/pERK was significant but Fos/CaM Kinase was insignificant
An analysis of variance (ANOVA) with group as a between-subject factor (control, 30 minute noise) showed a significant induction of Fos protein in the 30 minute noise group (F(1,3)=38.84, p=0.025). PERK and Fos displayed a high level of co-expression (Table 3 and left panel of Figure 6). However, co-localization of CaMKII in Fos cells was very low (Table 3 and right panel of Figure 6).
30 min noise
Table 3. The table shows the mean and standard error for the control and 30 minute stress group Fos expression and co-expression of pERK and CaMKII in Fos positive cells.
Fig. 6. Comparative pictures of pERK/Fos co-expression and CaMKII/Fos co-expression in the posterior hypothalamus of stressed rats. Magnification for the left panel is 400X, while magnification in the right panel is 200X. The color red represents Fos, while green represents pERK or CaMKII.
The present study explored the localization of pELK-1, pERK, and CaMKII in the posterior hypothalamus of rats following an acute episode of stress. Elk has been shown to be activated by ERK during plasticity and to induce transcriptional changes downstream via CREB (Tian & Karin, 1999). However, the antibody used to detect pElk-1 in the current study did not consistently detect a protein product. Since pElk-1was tested in tissue perfused in the traditional manner, the inconclusive results could have been affected by a dark or partially visible background. Further investigations should measure pElk-1 expression using the new perfusion method. Because of these technical limitations, no conclusions can be made about the localization/activation of Elk-1 by stress in the posterior hypothalamus.
The discovery of increased phospho-ERK levels in stressed animals suggests that pERK could be part of a signaling pathway that mediates the plasticity associated with noise habituation. Since noise was the last sensory stimulation the animals experienced before they were sacrificed and pERK levels increased in the PH, it can be concluded that the noise exposure directly facilitated this change.
The next steps were to measure the expression of pERK at different time intervals of stress and to look for it in cells characterized as active with Fos. We showed that phosho-ERK levels were highest at 15 minutes of noise exposure and declined at 30 minutes. The time-dependent activation of ERK is congruent with the theory of habituation; ERK signaling takes place shortly after the stimulus (15 minutes) and is turned off with prolonged stress (30 minutes). The PH sends projections to the mpPVN, a site where releasing hormones are secreted; therefore, the intracellular plastic changes of pERK could contribute to attenuated HPA axis response in subsequent stress exposure.
The third experiment showed more Fos expression in stressed animals than the controls, but the ratio of co-expression was similar for the controls and stress conditions. This effect was caused by the high expression of pERK in the controls and the stressed animals, which indicates that the activation of ERK was caused by factors beyond stress. The novel environment of the acoustic chambers was not a likely source of stress for the control animals, because they were allowed one night of adaptation before testing day. Similarly, the duration of sacrifice was too short (~1 minute) to have activated signal transduction systems. PERK presence in the control groups could mean that ERK was basally phosphorylated, providing other functions than plasticity. This means that stress-induced phosphorylation of ERK had a rapid onset and termination, causing stress-induced extra phosphorylation that was only detectable at a specific time.
A limitation of the study was that extremely high levels of pERK in control and stressed subjects prevented the exact number of cells expressing pERK from being manually quantified. A better procedure to determine if there was a difference in pERK levels between control and stressed subjects would be a western blot, which provides a more accurate quantitative analysis of protein expression.
PERK was widely expressed in active PH neurons, which supports the high expression levels we observed with the 15 minute DAB reaction. Further investigation could indicate whether a larger ratio of Fos/pERK co-expression occurs at the 15 minute time point. The next step would be to conduct a between-subjects study to observe the phosphorylation patterns of ERK at additional time intervals from 0 to 45 minutes. Stress exposure should be extended to 45 minutes, because pERK levels at 30 minutes were still higher than pERK in baseline controls. Most habituation occurs within the first periods of stress exposure; therefore, phosphorylation of ERK would not be expected in the later stages of repeated exposures.
While the highest pERK levels occurred at 15 minutes, Fos levels peaked at ~30 minutes, meaning that only the Fos mRNA levels would have been detectable at 15 minutes. To better correlate pERK/Fos co-expression, Fos mRNA, not Fos protein, should be measured with in situ hybridization, and immunohistochemical analysis should be performed for pERK. If Fos/pERK co-expression is highest around 15 minutes and declines by 45 minutes, we can infer that the transcriptional changes leading to altered gene expression and reduction of neuroendocrine and autonomic responses to stress may be mediated in part by ERK.
In order to further understand the relationship between pERK and HPA axis habituation, a within-subjects study could be conducted to observe pERK and Fos expression in the posterior hypothalamus. Rodents would be exposed to 15 minutes of noise for several consecutive days. A subset of the sample size would be sacrificed and used for research each day after the stress exposure. The subjects would be stressed on consecutive days, because HPA axis corticosterone habituation only takes place after the second stress exposure, meaning that some learning and memory formation has to occur. HPA axis hormones, such as ACTH and corticosterone, would be measured following every stress exposure. In situ hybridization would be used to measure Fos protein, and immunohistochemistry would be used to measure pERK. Fos levels, pERK levels, and HPA axis hormones would be expected to decrease between sessions. This study would help correlate pERK habituation and its possible role in regulating the steady decrease of the HPA axis response following repeated exposures to stress.
By using DAB IHC in the first experiment, CaM kinase was shown to be present in the PH of the control groups and stressed animals. The presence of CaM kinase in the PH of both groups was expected, since the kinase is constitutively expressed in cells. We were also not looking for the activated/phosphorylated state of CaMKII.
Protein levels in the second experiment showed no significant differences between the control groups and stressed animals. A limitation of this experiment was that we did not have enough tissue from the 15-minute stress group to analyze CaMKII due to problems with blocking and sectioning. However, the level of expression was not expected to change, and we were not looking for the phosphorylated form of the enzyme.
Double fluorescent immuno studies showed low or nonexistent co-expression with Fos in the control animals vs. stressed animals. This result does not support our hypothesis that cells that are active during stress contain the CaMK signal transduction pathway at the level of the PH. Low expression of CaMK could be attributed to the type of secondary antibody used in DAB IHC vs. fluorescent IHC. The DAB reaction showed much better expression of CaMKII as opposed to the double fluorescent immuno assays, suggesting that the fluorescent antibody was not as sensitive to the protein as the antibody used in the DAB reaction.
Currently, the molecular processes explaining the plastic changes that occur in response to acute stress remain largely unresolved. Whether the plastic changes are a LTD or LTP process mainly depends on the types of cells, excitatory or inhibitory, that become activated in response to stress; however, this is not conclusive. The limitations of this study provide basis and direction for future investigations to find a direct relationship between the effects of pERK and CaM kinase on stress habituation. It is difficult to study intermediate signaling molecules due to their pleitropic nature; further experimentation will have to shed light on their unexplained behavior.
In order to gauge the role of ERK in habituation and LTP/LTD processes, it would be interesting to test whether ERK inhibition would attenuate stress-elicited habituation. Detection of phospho-CREB would help to distinguish the ligand binding of the stress adaptation pathway from the nuclear activity of that pathway. Measurement of NMDA/AMPA receptor levels before and after repeated stress exposure could provide further evidence as to whether this process is LTP or LTD-related. Using microdialysis to measure the neurotransmitter signals from the posterior hypothalamus to the PVN and raphe pallidus could provide insight into the participating neurotransmitter systems and elucidate LTP or LTD-like mechanisms.
It is important to study the neural mechanisms involved in habituation to stress, because the negative feedback mechanisms of the HPA axis do not completely account for habituation. It has been shown in adrenalectomized rats that habituation to an unanticipated noise still occurs despite the lack of glucocorticoids that are responsible for the negative feedback loop (Davis & Zolovick, 1974).
Habituation to stressful stimuli is a beneficial evolutionary mechanism. Hans Selye’s General Adaptation Syndrome proposes that excessive activation of the HPA axis and the autonomic nervous system in the exhaustion stage can be detrimental to an organism. This is supported by numerous conditions like anxiety disorder, post-traumatic stress disorder, and depression, which can be caused by an overactive HPA axis. An overactive HPA axis can bring about a feeling of helplessness when stress does come. Repeated exposure to stress is the catalyst for a plethora of psychiatric diseases; therefore, understanding the neural mechanisms underlying stress adaptation could illuminate the etiology of these diseases and aid in the development of novel therapeutics.
Barco, A., Jancic, D., & Kandel, E. (2008). Creb-dependent transcription and synaptic plasticity.Transcriptional Regulation by Neuronal Activity, 127-154. Retrieved fromhttp://download.springer.com/static/pdf/500/chp%3A10.1007%2F978-0-387-73609-9_7.pdf?auth66=1363119278_e9a9cbd61d0a103de5c0f7f7ddaf9e0f&ext=.pdf
Campeau, S, & Watson, S.J.Jr. (2000). Connections of some auditory-responsive posteior thalamic nuclei putatively involved in activation of the hypothalamo-pituitary-adrenocortical axis in respons to audiogenic stress: An anterograde and retrograde tract tracing study combined with Fos expression. Journal of Comparative Neurology, 423: 474-491.
Cruzalegui, F., Cano, E., & Treisman, R. (1999). Erk activation induces phosphorylation of elk-1 at multiple s/t-p motifs to high stoichiometry.Oncogene, 18(56), 7948-7957.
Davis, M., & Zolovick, A. (1974). Habituation of the startle response in adrenalectomized rats. Physiology & Behavior, 8(4), 579-584.
Erondu, N., & Kennedy, M. (1985). Regional distribution of type ii ca2 /calmodulin-dependent protein kinase in rat brain. The Journal of Neuroscience, 5(12), 3270-3277.
Esdin, J., Pearce, K., & Glanzman, D. (2010). Long-term habituation of the gill-withdrawal reflex in aplysia requires gene transcription, calcineurin and l-type voltage-gated calcium channels. Frontiers in Behavioral Neuroscience, 4(181), doi: 10.3389/fnbeh.2010.00181
Giovannini, M. (2006). The role of the extracellular signal-regulated kinase pathway in memory encoding.Reviews in the Neurosciences, 17(6), 619-634.
Grissom, N., & Bhatnagar, S. (2008). Habituation to repeated stress: Get used to it. Neurobiol. Learn. Mem, 92(2), 215-224.
Hu, B., Liu, C., Bramlett, H., Sick, T., Alonso, O., Chen, S., & Dietrich, D. (2004). Changes in trkb-erk1/2-creb/elk-1 pathways in hippocampal mossy fiber organization after traumatic brain injury. Journal of Cerebral Blood Flow & Metabolism, 24, 934-943. doi: 10.1097/01.WCB.0000125888.56462.A1
Jankord, R., & Herman, J. (2008). Limbic regulation of hypothalamo-pituitary-adrenocortical function during acute and chronic stress. Stress, Neurotransmitters, and Hormones, 1148, 64-73. doi: 10.1196/annals.1410.012
Kennedy, M. B., Beale, H. C., Carlisle, H. J., & Washburn, L. R. (2005). Integration of biochemical signalling in spines. Nat. Rev. Neurosci, 6(6), 423-434.
Khan, A., & Watts, A. (2004). Intravenous 2-deoxy-d-glucose injection rapidly elevates levels of the phosphorylated forms of p44/42 mitogen-activated protein kinases (extracellularly regulated kinases 1/2) in rat hypothalamic parvicellular paraventricular neurons. Neuroendocrinology, 145(1), 351-359. Retrieved fromhttp://endo.endojournals.org/content/145/1/351.full
Kovacs, K. (1998). c-fos as a transcription factor: a stressful (re)view from a functional map. Neurochemistry International, 33, 287-297.
Lee HK. AMPA Receptor Phosphorylation in Synaptic Plasticity: Insights from Knockin Mice. In: Kittler JT, Moss SJ, editors. The Dynamic Synapse: Molecular Methods in Ionotropic Receptor Biology. Boca Raton (FL): CRC Press; 2006. Chapter 14. Available from: http://www.ncbi.nlm.nih.gov/books/NBK2549/
Malenka, R. C., & Bear, M. F. (2004). LTP and LTD: an embarrassment of riches.Neuron, 44(1), 5-21.
Marie H., Morishita W., Yu X., Calakos N., Malenka R.C. Generation of silent synapses by acute in vivo expression of CaMKIV and CREB (2005) Neuron, 45 (5) , pp. 741-752.
Masini, C., Babb, J., Nyhuis, T., Day, H., & Campeau, S. (2012). Auditory cortex lesions do not disrupt habituation of hpa axis responses to repeated noise stress. Brain Research, 1443, 18-26.
Masini, C., Day, H., & Campeau, S. (2009). Reversible inactivation of the auditory thalamus disrupts hpa axis habituation to repeated loud noise stress exposures. 1276, 123-130. doi: 10.1016/j.brainres.2009.04.023
Michael, D., Martin, K., Seger, R., Ning, M., Baston, R., & Kandel, E. (1998). Repeated pulses of serotonin required for long-term facilitation activate mitogen-activated protein kinase in sensory neurons of aplysia. Neurobiology, 95, 1864–1869.
Nestler, E., Hyman, S., & Malenka, R. (2009). Molecular neuropharmacology: A foundation for clinical neuroscience. (2nd ed. ed.). China: McGraw-Hill Companies, Inc.
Nyhuis, T., Day, H., Campeau, S. Neurons in the posterior hypothalamus express Fos in response to audiogenic stress and project to both the raphe pallidus and paraventricular hypothalamic nucleus. Program No. 928.05. 2011 Neuroscience Meeting Planner. Washington, DC: Society for Neuroscience, 2011. Online.
Nyhuis, T., Day, H., Campeau, S. Reversible inactivation of a subregion of the posterior hypothalamus disrupts HPA axis habituation to repeated audiogenic stress exposures in rats. Program No. 191.01. Neuroscience Meeting Planner. Louisiana: Society for Neuroscience, 2012. Online.
Popoli, M., Gennarelli, M., & Racagni, G. (2002). Modulation of synaptic plasticity by stress and antidepressants. Bipolar Disorders, 4(3), 166-182. doi: 10.1034/j.1399-5618.2002.01159.x
Soderling, T. (1999). The ca21 –calmodulin-dependent protein kinase cascade. Trends in Biochemical Sciences, 24(6), 232-236. doi: 10.1016/S0968-0004(99)01383-3
Sweatt, D. (2004). Mitogen-activated protein kinases in synaptic plasticity and memory. Current Opinion in Neurobiology, 14, 311-317. doi: 10.1016/j.conb.2004.04.001
Thiels, E., & Klann, E. (2001). Extracellular signal-regulated kinase, synaptic plasticity, and memory. Reviews in the Neurosciences, 12(4), 327-345.
Thiels, E., Kanterewicz, B. I., Norman, E. D., Trzaskos, J. M., & Klann, E. (2002). Long-term depression in the adult hippocampus in vivo involves activation of extracellular signal-regulated kinase and phosphorylation of Elk-1. J. Neurosci, 22(6), 2054-2062.
Tian, J., & Karin, M. (1999). Stimulation of elk1 transcriptional activity by mitogen-activated protein kinases is negatively regulated by protein phosphatase 2b (calcineurin). The Journal of Biological Chemistry, 274, 15173-15180. doi: 10.1074/jbc.274.21.15173