PennScience Fall 2015
Volume 14 Issue 1
PennScience is a peer-reviewed journal of undergraduate research published by the Science and Technology Wing at the University of Pennsylvania. PennScience is an undergraduate journal that is advised by a board of faculty members. PennScience presents relevant science features, interviews, and research articles from many disciplines including biological sciences, chemistry, physics, mathematics, geological sciences, and computer sciences. PennScience is a SAC funded organization. For additional information about the journal including submission guidelines, visit www.pennscience.org or email us at pennscience@gmail.com.
EDITORIAL STAFF EDITORS-IN-CHIEF DESIGN MANAGERS
Claudia Cheung Carolyn Lye Emily Chen Courtney Connolly
WRITING MANAGERS
Grace Ragi Samip Sheth
EDITING MANAGERS
Kartik Bhamidipati Karan Pahil
BUSINESS MANAGER
Tina Huang
FACULTY ADVISORS
WRITING
Ritwik Bhatia Richard Diurba Mia Fatuzzo Andy Guo Krisna Maddy Abhinav Suri Shelly Teng Andrew Wang Ben Wang
DESIGN
Suzanne Knop Chigoziri Konkwo Alison Weiss
2 PENNSCIENCE JOURNAL | FALL 2015
Dr. M. Krimo Bokreta Dr. Jorge Santiago-Aviles EDITING
Angela Chang Cindy Chen Jane Chuprin Tiffany Huang Ila Kumar Ryan Leone Rachel Levinson Joan Lim Karishma Nanwani Jim Tse Andrew Wang Alex Wong Joyce Xu Evan Zou
CONTENTS FEATURES The Future of Alzheimer’s Neuroplasticity
06
09
Can Our Brains be Trained to Learn More Efficiently?
Memristors
12
Oliver Sacks: Contributions to Science and Writing
18
The Scientific Future of Immortality Free Will
20
15
INTERVIEW 22 with David F. Meaney, Ph.D. Solomon R. Pollack Professor and Chair of Bioengineering
RESEARCH
24
Social Stress Induces Dendritic Growth in Layer V Pyramidal Neurons of the PFC Eric Geng, Children’s Hospital of Philadelphia Department of Anesthesiology and Critical Care Medicine
30
Fabrication of Novel Nanomaterial Tungsten Disulfide (WS2) Nanopores for Solid-State DNA Sequencing Laura Beth Fulton, Drndić Condensed Matter Physics Lab, University of Pennsylvania Nano/Bio Interface Center
FALL 2015 | PENNSCIENCE JOURNAL
3
LETTER FROM THE EDITORS Dear Readers, We are delighted to bring you the first issue of the 14th volume of PennScience. The theme for this issue is dedicated to neuroscience, an inherently complex and interdisciplinary area of scientific research. Through this issue, we hope to share with you some exciting aspects to understanding the nervous system on multiple levels. We are incredibly fortunate to have such amazing writers, and we are excited to share their work in exploring the many facets of neuroscience through a series of feature articles. Shelly Teng opens with an article on Alzheimer’s disease and whether an imbalance of neurotransmitters contributes to degeneration. Mia Fatuzzo presents a thorough examination on neuroplasticity and the effectiveness, if any, of cognitive training. Ben Wang explores the value of Artificial Neural Networks in the artificial intelligence landscape. Andy Guo describes the contributions of Oliver Sacks to the medical writing community. In an article focused on the concept of whole brain emulation, Ritwik Bhatia sheds some light on the allure that immortality might someday become a reality. Richard Diurba provides a discussion on the extent to which physics and neuroscience support the idea of free will. We are also proud to publish an interview with Dr. David Meaney, whose research in the Department of Bioengineering at Penn focuses on the response of the central nervous system to traumatic injury. Moreover, we are honored to showcase the excellent research of two fellow undergraduates. Eric Geng examines the effects of repeated stress on certain neurons in the prefrontal cortex, an area of the brain that regulates executive function in humans, using the rat as a model organism. Laura Beth Fulton presents an interesting study on fabricating nanopore membranes, particularly those that contribute to DNA sequencing and techniques that reveal molecular changes and abnormalities for biomedical purposes. We also continued our series of coffee chats this semester to expand and foster scientific discussion on campus. These coffee chats offer undergraduates the opportunity to meet faculty members and learn about their research and career paths, and we plan to continue this initiative into next semester. We are especially grateful to Dr. Rahul Kohli for participating and leading a discussion on the physician-scientist profession and his research on DNA modifying enzymes and pathways that contribute to genome plasticity. Finally, we would like to thank the many groups and individuals who made this publication possible. We would first like to thank all of our managers and committee members for their hard work and dedication for the journal. The Student Activities Council and the Science and Technology Wing at Kings Court English College House generously provide our funding, without which we could not publish such a first-rate journal. We would also like to acknowledge our faculty advisors, Dr. Krimo Bokreta and Dr. Jorge Santiago-Aviles, for their constant support and advice. Lastly, we would like to thank the various members of the Penn faculty who offered us insight into their fields of research. Thank you for reading PennScience and joining us in furthering the scientific discourse and visibility of undergraduate research at Penn! Sincerely, Claudia Cheung and Carolyn Lye Co-Editors-in-Chief
4
PENNSCIENCE JOURNAL | FALL 2015
CALL FOR SUBMISSIONS Looking for a chance to publish your research? PennScience is accepting submissions for our upcoming Spring 2016 issue! Submit your independent study projects, senior design projects, reviews, and other original research articles to share your work with fellow undergraduates at Penn and beyond. Email submissions and any questions to pennscience@gmail.com.
Research in any scientific field will be considered, including but not limited to:Â
Biochemistry, Biological Sciences, Biotechnology, Chemistry, Computer Science, Engineering, Geology, Mathematics, Medicine, Physics, and Psychology
FALL 2015 | PENNSCIENCE JOURNAL
5
The Future of
FEATURES
Introduction
A
By Shelley Teng
Alzheimer ’s
lzheimer’s disease (AD) has been a part of global history since its first diagnosis in 1901 in a patient named Auguste Deter. Hospitalized for mental instability, Auguste was one of many to experience the deleterious symptoms of AD. However, we now understand that AD is not characterized by mental instability, but rather by a gradual memory loss and dementia (Francis et al., 1998). Since then, the ubiquity of the disease has also grown extensively, now affecting more than 40 million people worldwide and over 50% of people over than the age of 85 (Alz.org, 2015). As the incidence of ADhas grown, substantial research toward understanding the disease has also been progressing with an upward trend. Many researchers have pointed towards the build up of senile plaques in various areas of the brain as a cause of AD. However, ongoing and future research shows that a potential reason for AD onset may also be an imbalance of neurotransmitters in the brain. Current Theories As aforementioned, a popularly cited cause of ADis the
6
PENNSCIENCE JOURNAL | FALL 2015
production and build up of neural lesions in the brain. Neurofibrillary tangles made up of a protein called tau aggregate inside neurons. Moreover, , plaques of misfolded beta-amyloid (Aβ), another protein, tend to aggregate around neurons. This buildup can lead to dysfunction in neuron synapses and eventual neuronal death. This loss of neural activity, along with disrupted neural connectivity in parts of the brain that control memory, are possible factors contributing to the symptoms involved in AD (Murphy & Levine, 2010). In addition to this popular theory, scientists have discovered that AD onset is potentially caused by an imbalance of neurotransmitters in the brain. Specifically, these studies have found that the downregulation of a neurotransmitter, acetylcholine (ACh), in certain memoryrelated pathways of the brain plays a significant role. How does the ACh system work? Normally, ACh is essential to many cholinergic pathways, which are pathways in neurons that propagate signals of choline neurotransmittersthroughoutvariouspartsofthebrain.Generally, it acts as a neuromodulator for many functions, meaning that it helps to regulate the strengthening and firing of certain pathways
FEATURES
that trigger neural responses. Furthermore, ACh has been found to enhance memory encoding by strengthening the feedback mechanisms in neurons that store memory (Hasselmo, 2006). On a molecular level, ACh synthesis is strictly regulated by two major classes of receptors: nicotinic and muscarinic. Nicotinic receptors are ligand-gated ion channels, while muscarinic ones are G-protein coupled receptors. Although these receptors differ in the type of molecules that activate the subsequent signaling pathways, it has been shown that both are effective in binding with ACh(Akk & Auerbach, 1999). Therefore, ACh may bind to them and trigger the post-synaptic signal that allows for a certain neural action to be performed. After propagation of this pathway, the termination of the ACh signal is regulated by a protein called acetylcholinesterase (AChE). This enzyme serves to hydrolyze ACh, thus maintaining a basal level of ACh in the brain and working to keep the feedback loop steady. Oftentimes, these pathways are located in areas of the brain associated with higher cognitive function such as the neocortex, cerebral cortex, and hippocampus. Relationship to Alzheimer’s As a result, if normal ACh levels are responsible for maintaining memory encoding in the brain, then decreased levels of acetylcholine may lead to AD symptoms. Research has shown that this decline may be caused by a possible reduction in the amount of nicotinic and muscarinic receptors in the brain’s cholinergic pathways (Kihara & Shinohama, 2004). A lack of receptors to receive the ACh signal may potentially disrupt the strength of neural connections that build memory in the brain. An important pathway in the ACh system is the cholinergic pathway from the nucleus basalis of Meynert, located in the basal forebrain, to the neocortex (Levey, 1996). This pathway has a high concentration of muscarinic receptors, so the decline in their function often contributes to the onset of memory loss in the neocortex (Tsang et al., 2006). Furthermore, it has been found that loss of these cholinergic projections in the cerebral cortex also significantly contributes to cognitive deterioration in older patients (Francis et al., 1998). In addition to the decline in ACh levels, studies have also found that AChE enzymes have the ability to interact with Aβ proteins. Specifically, these interactions allow for increased levels of plaque buildup in cholinergic neurons, and thus, subsequent death. With substantial amounts of neuronal death, the level of ACh is significantly reduced, thereby reducing the brain’s ability to store memories and make synaptic connections (Rees & Brimijoin, 2003). Drug Therapies There has been extensive research for therapies that may increase the levels of acetylcholine in order to alleviate
symptoms of AD. Currently, there are two major classes of drugs that are clinically used to treat AD. One of these includes cholinesterase inhibitors, which inhibit the enzyme AChE, and thereby prevent the breakdown of ACh in the brain. The use of of these acetylcholinesterase inhibitors (AChEIs) not only prevents the breakdown of ACh, but also has been found to protect cells from free radical toxicity and beta-amyloid induced neuronal death (Tabet, 2006). The other class of drugs falls under N-methyl-d-aspartate (NMDA) receptor inhibitors. This receptor plays a significant role in the propagation of various signaling pathways in nerve cells. Generally, it is activated by the neurotransmitter glutamate. In relation to AD, it has been shown that over-activation of NMDA receptors leads to neuronal death (Lipton, 2004). Therefore, the use of NMDA receptor antagonists in certain drugs can quell the negative effects of this in various neurological disorders. Future Research While significant progress has been made in regards to AD treatment in recent history, a substantial amount of current research is taking place in order to further enhance our understanding of the disease. Here at the University of Pennsylvania, a university at the forefront of research in medicine, many notable professors are working with various new hypotheses and ideas regarding drug therapies for AD patients. Dr. Jason Karlawish, M.D., of the Perelman School of Medicine is currently involved in many prominent research projects that deal with the clinical aspects of AD. One clinical study that he is a part of is the “Anti-Amyloid in Asymptomatic Alzheimer’s” (A4) study. The project involves the use of an antiamyloid antibody and tests whether this antibody treatment can slow the effects of memory loss associated with amyloid plaque buildup in neurons (A4 Study, 2015). The study examines 1000 subjects between the ages of 65 and 80 who have no signs of cognitive deterioration, but may be at risk for the disease due to elevated levels of amyloid in their brains. By administering this drug to these subjects, researchers may analyze whether the antibody has the ability to slow neuronal death and prevent the symptoms associated with memory loss in AD. This project is especially unique because there are currently no drugs available for prescription in clinical practice that target the build of of amyloid plaques. As a result, the efforts put forth by Dr. Karlawish and his team contribute significantly to the ongoing research initiatives in this domain. In addition to his work on the A4 study, Dr. Karlawish is also conducting his own independent research regarding AD; he is working with the same patients from the A4 study, but he is focusing on the psychological effects of learning about their particular results from the study. In his project, the Study of Knowledge and Reactions to Amyloid Testing (SOKRATES), Dr. Karlawish
FALL 2015 | PENNSCIENCE JOURNAL
7
FEATURES studies how learning the results from this amyloid testing can impact an older patient’s psychological well-being and mood. In an interview with Dr. Karlawish, he stated that he hopes this project will shed light on how physicians in clinical practices should interact with patients after sharing their diagnoses with them. Conclusion AD is growing in both the number of people it is affecting, and the amount of research taking place to combat it. While recent research has shed light on the physical symptoms and the physiological processes of AD, more knowledge about the disease has yet to be discovered. The research being done at Penn and many other top institutions has placed the scientific community one step closer to finding a definitive solution to AD.
References A4 Study,. (2015). A4 Study: Now is the Time. Retrieved 27 October 2015, from http://a4study.org Akk, G., & Auerbach, A. (1999). Activation of muscle nicotinic acetylcholine receptor channels by nicotinic and muscarinic agonists. British Journal Of Pharmacology, 128(7), 1467-1476. http://dx.doi. org/10.1038/sj.bjp.0702941 Francis, P., Palmer, A., Snape, M., & Wilcock, G. (1999). The cholinergic hypothesis of Alzheimer’s disease: a review of progress. Journal Of Neurology, Neurosurgery & Psychiatry, 66(2), 137-147. http://dx.doi.org/10.1136/ jnnp.66.2.137 Hasselmo, M. (2006). The role of acetylcholine in learning and memory. Current Opinion In Neurobiology, 16(6), 710-715. http://dx.doi.org/10.1016/j. conb.2006.09.002 Kihara, T. & Shimohama, S. (2004). Alzheimer’s disease and acetylcholine receptors. Acta Neurobiol Exp (Wars), (64)1, 99-105. Latest Facts & Figures Report | Alzheimer’s Association,. (2013). Latest Alzheimer’s Facts and Figures. Retrieved 11 October 2015, from http://www. alz.org/facts/ Levey, A. (1996). Muscarinic acetylcholine receptor expression in memory circuits: Implications for treatment of Alzheimer disease. Proceedings Of The National Academy Of Sciences, 93(24), 13541-13546. http://dx.doi. org/10.1073/pnas.93.24.13541 Lipton, S. (2004). Failures and successes of NMDA receptor antagonists: Molecular basis for the use of open-channel blockers like memantine in the treatment of acute and chronic neurologic insults.Neurotherapeutics, 1(1), 101-110. http://dx.doi.org/10.1602/neurorx.1.1.101 Murphy, M. P. & Levine, H. (2010). Alzheimer’s Disease and the β-Amyloid Peptide. J Alzheimers Dis, 19(1), 311. http://dx.doi.org/10.3233/JAD-20101221 Nba.uth.tmc.edu,. (2015). Acetylcholine Neurotransmission (Section 1, Chapter 11) Neuroscience Online: An Electronic Textbook for the Neurosciences | Department of Neurobiology and Anatomy - The University of Texas Medical School at Houston. Retrieved 26 September 2015, from http://nba.uth.tmc.edu/neuroscience/s1/chapter11.html Rees, T., & Brimijoin, S. (2003). The role of acetylcholinesterase in the pathogenesis of Alzheimer’s disease. Drugs Today, 39(1), 75. http://dx.doi. org/10.1358/dot.2003.39.1.740206 Tabet, N. (2006). Acetylcholinesterase inhibitors for Alzheimer’s disease: antiinflammatories in acetylcholine clothing!. Age And Ageing, 35(4), 336-338. http://dx.doi.org/10.1093/ageing/afl027 Tsang, S., Lai, M., Kirvell, S., Francis, P., Esiri, M., & Hope, T. (2006). Impaired coupling of muscarinic M1 receptors to G-proteins in the neocortex is associated with severity of dementia in Alzheimer’s disease. Neurobiology Of Aging, 27(9), 1216-1223. http://dx.doi.org/10.1016/j. neurobiolaging.2005.07.010
8
PENNSCIENCE JOURNAL | FALL 2015
FEATURES
Neuroplasticity Can Our Brains Be Trained to Learn More Efficiently? By Mia Fatuzzo
L
umosity’s advertisement, in the span of thirty seconds, promises that, for 15 dollars a month, you too can become “quicker, sharper, more focused, and a faster learner (YouTube, 2015).” In short, Lumosity promises a better brain. Can their cognitive training — simple “brain-training” games which rely on the concept of neuroplasticity – actually improve concentration or memory? Companies which market cognitive training software, such as Lumosity and Posit Science, argue, crucially, that their participants improve at not only at the specific tasks taught by the program or game, but also untrained tasks of significance. sults from their programs are generalizable. In other words, a Lumosity user should see improvement at both the game “Speed Match,” which requires you to match symbols, and general memory assessments. Some studies affirm this claim. Other studies, including a 2013 literature review, argue that, while the games do produce improvements at the tasks at hand, they do not yield any generalizable results. These studies contend that while participants improve at “Speed Match,” they are ultimately no better at, say, locating their car in the parking lot. Therefore, the literature is mixed; the claim
FALL 2015 | PENNSCIENCE JOURNAL
9
FEATURES that cognitive training yields improvement at untrained tasks of significance is somewhat controversial. Neuroplasticity — the ability of neurons to change from experience — is the concept that inspired the development of cognitive training programs. In 2000, researchers from University College London provided a concrete example of this phenomenon. They imaged the brains of taxi drivers, required by occupation to have a vast amount of spatial knowledge, and bus drivers, who drive prescribed routes. The researchers noticed differences in grey matter — unmyelinated axons and cell bodies. The taxi drivers had more grey matter in their mid-posterior hippocampi and less grey matter in their anterior hippocampi, when compared to the bus drivers. They also noticed a correlation between years of navigation experience and hippocampal grey matter volume in the taxi drivers; right posterior grey matter volume increased and anterior volume decreased with more navigation experience. These findings suggest that spatial knowledge is associated with the pattern of hippocampal grey matter volume in taxi drivers. More importantly, they emphasize the plasticity of the hippocampus – the change in volume of grey matter – in response to environmental demands (Maguire et al, 2000). Driving a taxi, as well as juggling, speaking another language, or playing an instrument, can alter the way our brains work Musicians, for example, not only improve at playing an instrument over time, but also demonstrate neuroplasticity. Brain scans showed changed anatomical
10 PENNSCIENCE JOURNAL | FALL 2015
features in areas involved in playing a musical instrument (Munte, 2002). Scientists have identified three basic processes by which these structural changes can occur. Presynaptic facilitation occurs when the neurons release more neurotransmitter into the synapses during learning than at rest (Castellucci and Kandel, 1967). Long-term potentiation is produced when one neuron activates another neuron repeatedly (Bliss and Lomo, 1973). Finally, entirely new synapses can be created as dendrites grow new dendritic spines (Engert and Bonhoeffer, 1999). Cognitive training games hope to capitalize on the plasticity of the brain by inducing structural brain changes through carefully-designed games. These anatomical changes, ideally, will elicit improvement in not only the games, but also in generalizable memory tasks, such as remembering where you parked your car. Some research confirms the claim that cognitive training produces generalizable results. A 2009 study run by scientists from the Mayo Clinic and the University of Southern California, but funded by Posit Science, suggested that this is the case. The IMPACT study randomly assigned participants to receive either a computerized cognitive training program or general cognitive stimulation (written quizzes regarding education videos). After eight weeks of training, the group that received the cognitive training program showed significantly greater improvements in several measures of memory and attention, including a standard neuropsychological assessment, a word list recall test, a word list delayed recall test, and a letter-number
FEATURES sequencing test (Smith, 2015). Note that these areas of improvement are untrained, generalizable skills, rather than tasks taught in the program. This result lends credence to commercially available cognitive training software. However, the crucial assumption that benefits gained from specific tasks within cognitive training will transfer to untrained tasks of significance concerns many neuroscientists. Noted psychologist Edward Thorndike noticed that, when he trained young adults to estimate the area of a rectangle, they improved only at estimation of rectangles. They showed no improvement when asked to estimate, for example, the areas of other shapes or other measures such as weight (Thorndike & Woodworth, 1901a, 1901b). More recent studies also support the hypothesis that, although practice can lead to improved performance, it does not guarantee better performance on other tasks. For example, master chess players quickly memorized the locations of chess pieces only if the pieces were arranged in such a way that they could have come from an actual game. If the pieces were randomly placed, the master players were no better than novice players at the task (Chase & Simon, 1973). These studies cast doubt on the association between mastery of specific tasks and improvement in untrained tasks. Recent research more directly related to cognitive training also raises concerns about the generalizable effects of brain training games. A literature review conducted in 2013 appeared to confirm the opinion that the training does not yield improvement in untrained tasks. Researchers reviewed 23 different studies, all of which evaluated cognitive training. Using meta-analysis, they found no significant evidence of the generalization of cognitive training to other skills (Melby-Lervag, 2013). In other words, the researchers suggested that, while cognitive training certainly improves your ability to, for example, play the requisite computer game, it does not improve your ability to do arithmetic. There is clearly discord in the literature and among scientists regarding the effectiveness of cognitive training. Scientists agree that the training is not harmful, but is it worth the fifteen dollar per month* fee? The next step will likely be to determine neuronal activity during cognitive training; this will allow scientists to understand the effects of cognitive training not only externally, but also internally. Several recent studies have looked at neuronal activity during cognitive training, but it remains difficult to draw firm conclusions about the underlying neural mechanisms of the training. These studies make use of new imaging techniques such as functional magnetic resonance imaging, which uses blood flow to infer brain activity, and magnetoencephalography, which records the magnetic fields generated by brain activity. Lumosity alone boasts over fifty million members. The science may be ambiguous, but the popular vote is not (Lumosity.com, 2015).
*Lumosity, $14.95 monthly subscription fee Resources References Bliss & Lomo. (1973). Long-lasting potentiation of synaptic transmission in the dentate area of the anaesthetized rabbit following stimulation of the perforant path. The Journal of Physiology, 232, 331-356. Castellucci, V., & Kandel, E. R. (1976). Presynaptic Facilitation as a Mechanism for Behavioral Sensitization in Aplysia. Science, 194, 1176-1178. Chase, W. G., & Simon, H. A. (1973). Perception in chess. Cognitive Psychology, 4, 55–81. Engert, F., & Bonhoeffer, T. (1999). Dendritic spine changes associated with hippocampal long-term synaptic plasticity. Nature, 399, 66-70. Lumosity.com,. (2015). Lumosity Press Release Lumosity. Retrieved 9 October 2015, from http://www. lumosity.com/press/releases/lumosity-reaches-50million-members-and-1-1-billion-game-plays Munte, Thomas F. et al. (2002). The musician’s brain as a model of neuroplasticity. Nature Reviews Neuroscience, 3, 473-478. Maguire, E. A, et. al. Navigation-related structural change in the hippocampi of taxi drivers. (2000). Proceedings of the National Academy of Sciences of the United States of America, 97, 4398-4403. Melby-Lervag, M., & Hulme, C. (2012, May 21). Is Working Memory Training Effective? A Meta-Analytic Review. Developmental Psychology. Advance online publication. Smith, G. E. et. al. (2009). A cognitive training program based on principles of brain plasticity: results from the Improvement in Memory with Plasticity-based Adaptive Cognitive Training (IMPACT) study. Journal of the American Geriatrics Society, 57, 594-603. Thorndike, E. L., & Woodworth, R. S. (1901a). The influence of improvement in one mental function upon the efficiency of other functions I. Psychological Review, 8, 247–261. Thorndike, E. L., & Woodworth, R. S. (1901b). The influence of improvement in one mental function upon the efficiency of of other functions, II, Psychological Review, 8, 384-395. YouTube,. (2015). Emily Greco - Lumosity Commercial. Retrieved 9 October 2015, from https://www.youtube. com/watch?v=WfIbIsVRDcM
FALL 2015 | PENNSCIENCE JOURNAL
11
FEATURES
Memristors By Ben Wang
W
hen watching a video produced by Google’s DeepDream program, one could be forgiven for thinking that it is an artistic work produced by an avid user of hallucinogenic drugs. Once fed with a pedestrian video of a supermarket, the program spits out a rainbow-colored, psychedelic scene with floating dog heads and shifting patterns. DeepDream is Google’s way of visualizing the thought process within an Artificial Neural Network (ANN), a computer program that “learns” in a way that is similar to the human brain. By instructing the program to create a video output based on pattern recognition, computer scientists can learn more about the way ANNs process information, as well as demonstrate their immense power. While Google’s DeepDream program does not currently have any practical function, ANNs, however, are being employed to complete a variety of important tasks, including handwriting and license plate recognition. ANNs are computational models that “learn” to solve problems in a way that is similar to human processing. Each network consists of multiple layers of nodes (the “neurons”) that connect to each other via artificial synapses. These synapses hold different weights, determining the extent to which any one node’s output contributes to the input of any one node in the following layer. The nodes themselves also undergo an internal computation, though the specifics vary by model. The output of traditional artificial neurons is calculated along a sinusoidal curve, limiting the intensity of any one input. Spiking ANNs, which mimic the mechanisms of biological neurons, require the intensity of input to reach a threshold before releasing a pulse of intense output, just as neurons only fire after reaching a threshold of charge. In
12 PENNSCIENCE JOURNAL | FALL 2015
FEATURES
order for an ANNto compute or process information, it must first be “trained.” The ANN will be given a myriad of different inputs, all with attached feedback affirming or discouraging certain outputs. The crux of an ANN’s effective and accurate computation is therefore a way of learning from incorrect answers and remembering correct ones. Just as in the case of human learning, ANNs achieve this by modifying the weight of synaptic connections between nodes. Connections that result in correct answers are used repeatedly, and this repetition is reflected in increasing weight on the connection. Conversely, connections that are never required for getting a correct answer are ignored, reducing the weight of these connections.
rather an analog memory capable of remembering values along a continuous curve. This memory was due to the unique relation of flux and charge in memristors, which mathematically led to a conservation of memristance or memductance even after the voltage or current turned off. However, Chua’s memristor was a self-admitted “black box,” with no schematics for a practical model, let alone a near-ideal memristor that could be used for its analog memory capabilities. Thus, the full potential of memristor technology remained largely unrealized until 2008. When a team from Hewlett-Packard (HP) created a practical memristor device, they hoped that
However, ANNs have not come close to meeting their potential utility. Google’s DeepDream program consists of 11.2 billion parameters, or “connections” between artificial synapses, a figure that pales in comparison to the trillions of equivalent connections in biological brains. It stands to reason that larger artificial networkswith more connections may be the key to the development of high-level speech and pattern recognition, learning, and even intelligence in artificial machines. Yet computational power stands firmly in the way of such dreams; powerful neural networks can take thousands of separate central processing units to run. The problem lies in the incompatibility of traditional computer architecture with neural networks; each synapse consists of multiple circuits. Recently, a relatively new device called a memristor may be able to overcome the challenges of traditional computer architecture. Memristors were initially conceptualized by circuit engineer Leon Chua in 1971. For many years, the three fundamental circuit elements, resistors, capacitors, and inductors, had served to relate voltage, charge, current, and magnetic flux in circuits. The memristor, Chua proposed, was the “missing” fourth element, a device that could link magnetic flux and charge. In addition, his work suggested that this theoretical device had properties unique among circuit elements. The most significant of these properties was the memristor’s ability to “remember” — hence the name — the effects of past input, even after the removal of a power source. Furthermore, Chua argued, this memory was not a binary memory of two different equilibrium states, but
FALL 2015 | PENNSCIENCE JOURNAL
13
FEATURES
they could use practical memristors and their unique properties to expand the functionality of electrical circuits. By wedging together two resistors with variable resistance and separating them using a semiconducting nanofilm, the team was not only able to create a device that produced the pinched hysteresis loops, a signature pattern formed only by memristors when plotting the output and input of a device on a voltage-current graph. The device was also able to produce output consistent with that of an ideal memristor within a certain set of device states. The application of this newly created device to the construction of ANNs has been evident since the memristor’s conceptual origins. In the brain, special channels that function using sodium and potassium ions are responsible for transmitting signals. Through a circuit theoretic method, Chua demonstrated that these sodium and potassium ion channels were themselves biological memristors, displaying the characteristic pinched hysteresis loops. This fact, combined with the memristor’s ability to store analog memory, meant that memristors could very closely mimic the effect of long term potentiation in neurons. Just as biological memristors play a crucial role in remembering the weights of synaptic connections between neurons, artificial memristors can play a similar role in function of neural network-specific circuits. A widely supported design proposes a weighted synapse bridge composed of multiple memristors in series. Voltage pulses enter the bridge, and are amplified or dampened by the memristive weight of the synaptic bridge. They lead to a corresponding output of voltage that is dependent on the memristance stored in the synapse components. In addition, the voltage pulse also changes the memristance of the individual components, causing these alterations to be stored as non-volatile memory in preparation for the next pulse. Processing hardware created using memristors could run ANNs much more efficiently than traditional computer architecture, with a single memristor able to compute the output of a synapse. The industry has taken notice of the potential for the incorporation of memristors into computer hardware. “The Machine,” HP’s vision for the next generation of supercomputers, incorporates memristors heavily in its memory component. Despite
14 PENNSCIENCE JOURNAL | FALL 2015
this, the projected timeframe for the development of full memristor-incorporating hardware does not suggest the technology will be commonplace for a number of years. Meanwhile, however, ANNs will continue to grow in size and complexity on Silicon Valley servers to further push the boundaries of artificial intelligence. References Chua, Leon. “If It’s Pinched It’s a Memristor.” Memristors and Memristive Systems (2013): 17-90. Web. 30 Oct. 2015. Adhikari, S. P., Changju Yang, Hyongsuk Kim, and L. O. Chua. “Memristor Bridge Synapse-Based Neural Network and Its Learning.” IEEE Trans. Neural Netw. Learning Syst. IEEE Transactions on Neural Networks and Learning Systems 23.9 (2012): 1426-435. Web. Strukov, Dmitri, Gregory Snider, Duncan Stewart, and R.Stanley Williams. “The Missing Memristor Found.” Nature.com. Nature Publishing Group, n.d. Web. 30 Oct. 2015. Almási, Adela-Diana, Stanisław Woźniak, Valentin Cristea, Yusuf Leblebici, and Ton Engbersen. “Review of Advances in Neural Networks: Neural Design Technology Stack.” Neurocomputing (2015): n. pag. Web. 30 Oct. 2015. Chua, Leon. “Memristor, Hodgkin-Huxley, and Edge of Chaos.” Memristor Networks (2014): 67-94. Web. 30 Oct. 2015. Simonite, Tom. “HP’s Audacious Idea for Reinventing Computers | MIT Technology Review.” MIT Technology Review. MIT, 21 Apr. 2015. Web. 30 Oct. 2015. Hsu, Jeremy. “Biggest Neural Network Ever Pushes AI Deep Learning.” IEEE Spectrum. IEEE, n.d. Web. 30 Oct. 2015. Clark, Liat. “Google’s Artificial Brain Learns to Find Cat Videos.” Wired.com. Conde Nast Digital, n.d. Web. 30 Oct. 2015.
FEATURES
Oliver Sacks
Contributions to Science and Writing By Andy Guo
FALL 2015 | PENNSCIENCE JOURNAL
15
FEATURES
A
man propelled by joyful curiosity, Dr. Oliver Sacks led a life full of fascination with ways of exploring the experiences of living. As a medical doctor, Sacks had a gift for investigating the brain’s most peculiar pathways, often using his patients’ disorders as starting points for medications on the human condition. As a writer, he achieved a level of popularity unparalleled among scientists, with over a million copies of his books in print in the United States and several adapted for film (1). His “neurological novels” about his patients illuminated their characters as much as their conditions, employing a unique blend of clinical observation, and creating a deeply humanistic empathy rarely found in medical writing. For this reason, Sacks leaves behind a legacy of engagement in the scientific community while also establishing a humanistic connection with the general public. Born in 1933 in London, Oliver Sacks had an early penchant for pushing the limits of scientific endeavor. His interests in neurology stemmed from his brother Michael, who suffered from life-long psychosis and schizophrenia. Sacks wrote that his family felt “a sense of shame, of stigma, of secrecy” toward Michael, and he wanted to learn what made his brother that way (2, 3). During World War II, Sacks was sent to boarding school before attending Queen’s College, Oxford, where he received his medical degree. Sacks moved to America for an internship at the UCSF Medical Center at Mount Zion, and did his residency in neurology at the University of California, Los Angeles. Sacks moved to New York in 1965 for a fellowship at the Albert Einstein College of Medicine, and one year later, he began the clinical work that led to his best-known book, Awakenings (4). Awakenings tells the story of a group of patients in Beth Abraham Hospital, where Sacks worked as a consultant neurologist. These patients were survivors of encephalitis lethargica, better known as sleeping sickness. There has been no consensus on etiology, but several opinions have argued that it stemmed from the 1919 influenza epidemic as either an acute, or post-viral syndrome (5). The “sleep” that gave this disorder its popular name was rather unique as the body presented all signs of slumber while the sleeper remained aware of their surroundings. After a period of mild fever and general uneasiness, localized neurodegeneration would last anywhere from a few days to 30 years (6, 7). Post-
16 PENNSCIENCE JOURNAL | FALL 2015
encephalitic Parkinsonism emerged following this period of degeneration, as patients would present themselves in a comatose state Doctors had been baffled at the root causes of the disease until 1969, when Sacks tried the new Parkinson’s disease drug levodopa, L-DOPA, which had an astonishing “awakening” effect. L-DOPA is a precursor to the neurotransmitters dopamine, norepinephrine, and epinephrine. It functions in Parkinson’s patients by increasing levels of dopamine, thus reducing rigidity and increasing balance (8). In a 90day double blind trial with patients with encephalitis Sacks found varied success with L-DOPA treatment, with some patients showing no signs of improvement, while others demonstrated greater signs of vitality (9). In Awakenings, Sacks accounts that his successful patients became “much more optimistic…for there has been a significant number of patients who, following the vicissitudes of their first year on L-DOPA, came to do – and still do – extremely well.” Sacks’ patients had undergone an “enduring awakening,” able to “enjoy possibilities of life which had been impossible, unthinkable, before the coming of L-DOPA.” It is these types of historical accounts that drew the attention of both the medical community and the general public (10). According to Sacks, what made his book so intriguing to non-medical readers was his ability to “interfuse narration with medication, embedding each, so to speak, in the other…this interfusion was of case history and essay, was done at length and at leisure in Awakenings; and could never have been done within the format of any conventional article or book (9).” In this sense, this unique mutualism allowed Sacks to write about his patients in a personable and humanistic manner while still upholding the stringent traditions of medical writing. Sacks pushed the boundaries of innovation and tradition in medical writing, resulting in piece that has defined medical writing. He claims that he “Almost unconsciously, I became a storyteller at a time when medical narrative was almost extinct (11).” Awakenings was later adapted into a film starting Robert De Niro and Robin Williams in 1990. Sacks’ influence continues to reverberate with physicians and writers today. Dr. Norman Swan, a renowned Australian medical doctor and journalist, has said that his own writing
FEATURES
has been highly influenced by Sacks. Dr. Swan claims that Sacks flourished in a time “when medicine was as much of an art as it was a science,” complementing his ability to synthesize the two disciplines. Additionally, in Dr. Swan’s estimation, “humanistically, [Sacks’] contribution was immense…he really understood his patients in depth… he exposed the experience of being a patient as well as the wonderment of the natural world, and I think that’s what [Sacks’] gave people (12).” Ultimately, Sacks was able to pioneer a field of medical literature that continues to grow today. He has served as a model for aspiring physicians to pursue medical writing (13, 14). References 1. G. Cowles, Oliver Sacks, Neurologist Who Wrote About the Brain’s Quirks, Dies at 82. The New York Times. August 30, 2015. Accessed at http://www. nytimes.com/2015/08/31/science/oliver-sacks-dies-at82-neurologist-and-author-explored-the-brains-quirks. html?_r=0 2. J. Linder, After studying other people’s minds, Oliver Sacks looks at his own. The New York Post. May 3, 2015. Accessed at http://nypost.com/2015/05/03/after-studyingother-peoples-minds-oliver-sacks-looks-at-his-own/ 3. M. Roth, Why Oliver Sacks Always Goes Too Far. The Atlantic. May 16, 2015. Accessed at http://www.theatlantic. com/health/archive/2015/05/oliver-sacks-knows-what-itreally-means-to-live/393410/ 4. The Telegraph, Oliver Sacks, neurologist - obituary Neurologist who chronicled the dignified struggle of his patients in books such as Awakenings. The Telegraph. Accessed at http://www.telegraph.co.uk/news/ obituaries/11833684/Oliver-Sacks-neurologist-obituary. html 5. A. Easton, Encephalitis Lethargica. The Encephalitis Society. Accessed at http://www.encephalitis.info/ information/types-of-encephalitis/encephalitis-lethargica/ 6. The Conversation, A viral infection of the mind? The curious case of encephalitis lethargica. The Conversation.
Accessed at http://theconversation.com/a-viralinfection-of-the-mind-the-curious-case-of-encephalitislethargica-660 7. BBC News, Mystery of the forgotten plague. BBC News. Accessed at http://news.bbc.co.uk/2/hi/health/3930727.stm 8. H. Simon, D. Zieve, Parkinson Disease. The New York Times. Accessed at http://www.nytimes.com/health/ guides/disease/parkinsons-disease/levadopa-(l-dopa).html 9. O. Sacks, The origin of “Awakenings.” British Medical Journal. 287, 1968-1969 (1983). http://www.bmj.com/content/bmj/287/6409/1968.full.pdf 10. Sacks, Oliver. Awakenings. New York: HarperPerennial, 1990. Print. Awakenings, the book. ^“I have become much more optimistic than I was when I […] wrote Awakenings, for there has been a significant number of patients who, following the vicissitudes of their first years on L-DOPA, came to do – and still do – extremely well. Such patients have undergone anenduring awakening, and enjoy possibilities of life which had been impossible, unthinkable, before the coming of L-DOPA. [3]” 11. O. Sacks, Oliver Sacks: Sabbath. The New York Times. Accessed at http://www.nytimes.com/2015/08/16/opinion/ sunday/oliver-sacks-sabbath.html 12. ABC News, What was neurologist Oliver Sacks’ influence on science and humanity? ABC News Australia. Accessed at http://www.abc.net.au/news/2015-08-31/whatwas-neurologist-oliver-sacks-influence-on/6736966 13. R. Rosenbaum, Why Oliver Sacks is One of the Great Modern Adventurers. Smithsonian. Accessed at http://www.smithsonianmag.com/science-nature/whyoliver-sacks-is-one-of-the-great-modern-adventurers134428056/?no-ist 14. J. Linshi, Scientists And Writers Pay Tribute To Oliver Sacks. TIME. Accessed at http://time.com/4016218/oliversacks-death-twitter/
FALL 2015 | PENNSCIENCE JOURNAL
17
FEATURES
THE SCIENTIFIC FUTURE OF
I
mmortality is defined as the ability to live forever. As curious human beings, who are largely influenced by classical and science fictional tales, we imagine immortality as quixotic and simply unfeasible. We wonder how it would be possible to live in our bodies for hundreds, even thousands, of years. From a scientific standpoint, immortality revolves around allowing one’s brain to continue functioning after the body ceases to exist. With the exploits of scientists like Randal Koene and Winfried Denk, as well as innovations in neuroscience technology, immortality might one day develop into a controversial reality. Randal Koene, a Dutch neuroscientist and engineer, often captivates entire audiences by simply stating his lifelong goal: to upload his brain to a computer. Influenced by his father, a particle physicist, and spurred by an early interest in the human brain, Koene set out to study physics while pursuing a method to map the intricacies of the human brain. In the 1990s, Koene came across a group that shared similar interests, and together, they coined the term “whole brain emulation.” This idea, which Koene and his associates have pioneered, determines how to successfully keep a brain active, thinking, and living, through a computer. Understanding immortality requires the consideration of neuroscience and an understanding of the intricate relationship between basic science and human life. In theory, whole brain emulation attempts to “re-implement functions of a mind in another computational substrate... by copying the connectivity between those components” (1). Working toward this goal, scientists must first focus on the basic principles. For example, prior to copying the entirety of the brain, which involves complex pathways, it is critical to map out the individual particles which the brain comprises, such as the
18 PENNSCIENCE JOURNAL | FALL 2015
FEATURES
IMMORTALITY neurons. Even with a vast knowledge of how neurons work, scientists are currently unable to identify where one neuron begins and where the other ends. Well before a brain gets mapped, however, it must be preserved after death via several techniques. The most common method for preservation of any tissue involves cryogenics, which combine the chemistry and physics behind very cold temperatures. But drawbacks have been encountered. For example, at low temperatures, the formation of ice crystals can lead to the cutting and breaking of neurons and synapses which would render preservation useless. To address this, scientists are developing materials called cryoprotectants to prevent ice crystals from forming. Such antifreeze materials can also lead to dehydration, which would cause the inner circuits of the brain to shrink and ultimately tear. Thus, with every new breakthrough in technology, scientists must overcome various hurdles to achieve the next step. In addition, research over the last decade has demonstrated the metal osmium to aid in the preservation of biofilms in plants and other membranes, which has led scientists to believe that osmium may play an integral role in neuronal preservation (2). Wilfried Denk, a German physicist at the Max Planck Institute for Medical Research, has specialized in developing microscopy techniques, and has received numerous accolades for his achievements, recently being appointed as a foreign member of the National Academy of Sciences. Similar to Koene, Denk is fascinated with the inner machinations of the brain, and specifically, what exactly happens in the brain at every moment of our lives — from when we hear good news to when we attempt to remember something from ages ago. Novel breakthroughs continue to pave the path for brain immortality. Denk, along with researchers at the Max Planck Institute for Medical Research, has created a method that will allow for the mapping of a mouse brain. The experiment is currently ongoing, and through a special microscopy process, the brain tissue of a mouse can be examined at high resolutions such that each and every neuron is visible (3). In 2004, Denk created a new method that would allow scientists to analyze the entire lengths of axons, the “serial block-face” microscopy method (4). The principle of scanning electron microscopy is that the surface of a tissue is bombarded with electron beams. Previously,
by Ritwik Bhatia
through the fixation and staining of tissue, it was indeed possible to analyze smaller chunks of tissue, but never the size of a mouse brain. Now, it is possible to examine multiple layers of a tissue by just scoping the outermost. Combined with preservation techniques, the use of electroscopy paves the way for scientists to delve deeper into the conundrums of immortality. The previously mentioned techniques are all in application to the brains of mice and small animals. As such, it is highly likely that Koene’s lifelong goal to upload his brain onto a computer will not be accomplished in his lifetime. Preserving the brain and creating a map are only the two most fundamental steps. After this, scientists will be required to create further advancements in computational and robotic technology to host the vast information of the brain they wish to conquer. Even then, will Koene continue to exist after his death? At what point do we draw the line between ourselves and our brains? When we receive good news, as Denk posits, what role do our intrinsic qualities play in this reception? One who has an optimistic demeanor may view good news in one manner, while a pessimist may view it in another. Preserving a brain to function may help one remember good news, but in order to realize how they remember it, scientists will be missing a very integral part of human nature. The name of this part can be argued— the soul, the inner being—but it is something very real indeed. References 1. R. Koene, Experimental research in whole brain emulation: The need for innovative in vivo measurement techniques. Int. J. Mach. Conscious. 04, 35 (2012). 2. S. Mikula, A circuit diagram of the mouse brown. Max-Planck-Gesellschaft, 22 October 2012. 3. S. Mikula, J. Binding, W. Denk, Staining and embedding the whole mouse brain for electron microscopy. Nat Methods 12, 1198-1201 (2012). 4. W. Denk, H. Horstmann, Serial block-face scanning electron microscopy to reconstruct three-dimensional tissue nanostructure. PLoS Biology 02, 1900-1909 (2004).
FALL 2015 | PENNSCIENCE JOURNAL
19
FEATURES
Free Will By Richard Diurba
A
Penn student on average formulates millions of decisions and choices. They might skip a lecture to abscond to New Jersey, attend office hours, or put in two more hours of studying to keep medical school dreams alive. Making these decisions allows them to determine their own destinies. But, unbeknownst to students, they may not make these decisions. Free will may just be a myth. Thankfully, a large body of scientists from physics and neuroscience has dedicated themselves to the idea of free will. They offer a wide variety of answers from a wide variety of viewpoints. The idea of empiricism proving free will started with physics during the quantum mechanics revolution and remains a highly divided topic for neuroscientists. Physicists, such as Bell, argued on the topic of free will with the quantum mechanical theory of entanglement. The idea of entanglement is that two separate particles can be indirectly connected with each other. The key defining characteristic of this is that the analysis of one particle is predictive of the other particle. To describe this phenomenon, metaphorically, entanglement would be like running tests on a Penn student to learn about his twin brother at the University of California [1]. However, physicist John Bell opposed this idea. He stated that if a measurement was not a clone of the entangled particle, then the correlation could not truly explain the characteristics of the entangled particles [2]. The entangled particles could correlate with some accuracy, but ultimately could not correlate precisely with its other entangled particle. Consider again the Penn student who might study finance. Via Bell’s logic, physicists would not be able to determine if his twin brother was also studying finance because entanglement can only determine some of their shared characteristics. Bell’s theorem then became a mathematical inequality that states that all information taken from an entangled particle can only be accurate up to the point of the inequality. It essentially sets a limit to the amount entanglement can predict [1]. Thereafter, a large debate arose in physics to understand what is and is not determined by entanglement. Since the presentation of Bell’s inequality, physicists have presented findings that have confirmed and denied his work. For example, in a 2009 paper published in Nature, quantum bits (qubits) were connected with a waveform resonator and then measured [3]. If Bell’s inequality holds true, then the experimental data should show a discrepancy between the data of the entangled particle and the original particle. This was not observed.
20 PENNSCIENCE JOURNAL | FALL 2015
Instead, the discrepancy fell below the threshold for Bell’s inequality to hold [3]. By falling below the threshold, the experiment predicted more than what Bell predicted. Bell’s inequality failed and the errors in entanglement were not present in this experiment. For physicists, free will could account for these errors in Bell’s inequality. If there were perfect entanglement, then the physical constants that control neurons would dictate human behavior [4]. In this case, neurons are like entangled particles. They would experience a perfect relationship allowing nearly exact measurements of their entangled behavior. But if Bell’s inequality exists, the neurons would have some error that would need to be altered by another system to make up for the discrepancy. This system would be an individual’s free will acting to change the synaptic behavior. Like physicists, neuroscientists have similar divisions on free will. University of California Santa Barbara’s Michael Gazzaniga suggests that something analogous to perfect entanglement exists in the brain. In neuroscience, if perfect entanglement is upheld, then free will does not exist. If there is a difference between predicted and observed brain activity related to decision-making, then free will can be factored in as the difference [5]. Gazzaniga stated in an interview with Scientific American that the idea that a “little guy” representing free will controls a system of complex biological interactions appears unlikely, considering the fact that “neuroscientific experiments indicate that human decisions for action are made before the individual is consciously aware of them” [5]. Gazzaniga proved this in an experiment in which test subjects were subjected, ironically, to the stimulus of the word “bell” and then asked to respond
FEATURES with a general word [6]. Every subject responded with the word “music.” The test subjects ostensibly replied with the word “music” because of the stimulus from the experiment [6]. In this case, Bell’s inequality is “violated” in the neuroscience world, and initial conditions rule the entangled decisions. The initial stimulus of the word “bell” had a perfect reciprocal effect in producing the word “music.” This suggests that the brain follows a perfect entangled state, whereby when one part of the brain is given a stimulus the other portion precisely undergoes similar stimuli. Therefore, Bell’s inequality may not exist since it perfectly predicted and induced the word “music” as the response from its subjects.
extent to which nature determines everything. The large number of papers, debates, and experiments has presented a small percentage of the intricacies and possibilities free will can or cannot have in nature. One has to either accept the idea that religion or fate determines everything or that free will can determine an individual’s actions. The conclusion for the average undergraduate should remain clear, though; explaining to authority figures why daytrips to New Jersey takes precedence over organic chemistry will require decades of research papers on decision-making, interdisciplinary research into free will, and an abundance of questions on why New Jersey was chosen over New York.
Again, as with physics, opposition exists. An example of such opposition comes from Daniel Kahneman, a 2002 winner of the Nobel Memorial Prize in Economic Sciences. He states that within decision-making there is a wide variety of perceptions and options for free will. For instance, in his work on crises within business structures, he found that in certain crises free will could have been considered present or absent [7]. He ultimately expanded this idea into a dichotomy [6]. One subset of decisionmaking has free will and the other does not. The subset lacking free will includes daily operations, actions in which initial conditions determine reactions. However, when experiencing major issues, the brain therefore turns on free will and rationally thinks about the options provided [6]. When a crisis is determined to be important, the free will aspect of the brain is triggered. In this case, free will is determined by certain factors that can only be determined by the individual. It would, therefore, be unlikely that science could directly or indirectly prove free will because of the multiple intricacies in activating the free will aspect of the brain. Regardless, physics and neuroscience continue to support the idea of free will.
Bell’s inequality image: outreach/bellcurve.png
http://www.qolah.org/
References 1. Bell J. Speakable and uspeakable in quantum mechanics. Second Edition ed. ; 2004. 2. Bell J. On the einstein podolosky rosen paradox. Physics 1964 4 November 1964;1(3):195-290. 3. Ansmann M, Wang H, Bialczak Radoslaw, et. al. Violation of bell’s inequality in josephson phase qubits. Nature Letters 2009 24 September;461. 4. Musser G. The quantum physics of free will. Scientific American 2012 6 February. 5. Cook G. Neurosceince challenges old ideas about free will. Scientific American 2011 15 November. 6. Debiec J. Neuroscience: Capturing free will. Nature 2011 October;478(7369):322-3. 7. Kahneman D, Tversky A. Prospect theory: An analysis of decision under risk. Econometric Society 1979 March;47(2):263-92.
Neuroscience and physics both want to know the
FALL 2015 | PENNSCIENCE JOURNAL
21
INTERVIEW
CONDUCTED BY ABHINAV SURI, KRISNA MADDY AND ANDREW WANG PennScience: Dr. Meaney, your work has provided insight into traumatic brain injury. How have you seen research approaches in TBI change in the past few years and the past decade? Dr. Meaney: So the first way TBI research has changed is awareness. People are much more aware about it. The second part that has changed is that because we can do so many things scientifically, we are really starting to drill into what happens in the brain. Not just at the moment of injury, but as the brain recovers as it heals itself, [we’re understanding] what is going on for weeks, days, and months afterwards, and of course, what we see now in the popular press are diseases like chronic TBIs, Chronic Traumatic Encephalopathy (CTE), and how it’s linked to multiple concussions. That’s one aspect where we didn’t know the disease existed 15 years ago. There’s been this great interest now inrepetitive trauma andhow it affects the brain in diseases such as CTE, so we’ve become much more aware because we have the tools now to look at the brain over time about what happens after these injuries. Dr. David Meaney serves as the Solomon Pollack Professor and Chair of Bioengineering at the University of Pennsylvania. Dr. Meaney completed his PhD and his postdoctoral fellowship in bioengineering at Penn in 1991 and 1993, respectively, and joined Penn’s School of Engineering and Applied Science as a faculty member. His laboratory investigates the prevention and protection of traumatic brain injury (TBI), using mathematical models to simulate cell and nerve tissues in the central nervous system. In the past decade, Dr. Meaney’s work has been published in The Journal of Neuroscience Methods, Journal of Neurotrauma, and The Journal of Neuroscience. Recently, Dr. Meaney has led the development of injury models to further understand TBI. This work shows promising clinical applications of synthetic headgear in automotive restraint systems and athletic helmets.
22 PENNSCIENCE JOURNAL | FALL 2015
PennScience: You helped in the creation of the FluoroSNNAP, Fluorescence Single Neuron and Network Analysis Package for microcircuit imaging. What is the function of this program? Meaney: One of the nice features of these dyes is by looking at the fluctuations in the fluorescence, which literally go up, and in fractions of a second, we see an index into the activity of a neuron itself. So when the fluorescence goes up and comes down, many refer to this as a burst in action potential. And it’s nice, because what you do is you label individual neurons. If you can imagine in a field of view, if this neuron goes up in one instantof time, and at the same instant of time this neuron fluorescence goes up, then what we then do is we draw an association and infer that these are functionally connected to each other, even though we don’t have a way of directly probing that. What we would like to do is develop a set of tools that anybody could download for free, and it would be a robust set of tools to analyze the effects of TBI on a microscopic level in a fluorescence neuron network, facilitating hundreds of researchers in this field. PennScience: What are stretchable microelectrode arrays,
INTERVIEW
SMEA’s? How have they been employed in discovering the functional mechanisms and consequences of TBI? Meaney: We typically use electronics to record electrode signals with gold plated electrodes. But when you stretch or pull on gold, it breaks pretty easily. However, if you could plate the gold on a very flexible substrate, you could stretch that substrate and you wouldn’t break the connections in the gold. So if you had a circuit that was embedded and made from this gold, by stretching the clear membrane, you wouldn’t break the circuit. That was important because it mimicked a traumatic brain injury. It takes a slice of brain tissue from the hippocampus, puts it on this very flexible membrane, and freely stretches the membrane, which the slice of tissue is adhered to. This mimicks the kind of mechanical motion deformation that occurs in the brain during a concussion. It’s essentially a “concussion in a dish.” A stretchable microelectrode array allows you to get a reading immediately prior to and immediately following the stretch event, so you can infer what is going on in the brain during an actual TBI. Before SMEAs, we would get nothing. PennScience: With these new technologies, TBI research has increasingly come to the forefront of the neuroscience field. In the next decade, where do you see research in the field going? Are there any particular biomedical treatments, cures, or applications that seem on the verge? Dr. Meaney: It’s a big field now. I’ve always liked that part of it where there’s a single injury and you can kind of look at many different phases of this kind of injury. Protection, prevention, helmets, there is going to continue to be a big push about making that next generation helmet and going beyond that. Not only are they making helmets that are protecting a bit better, but they are also embedding sensors in these helmets when a football player, let’s say, tackles another player. You actually have a measure of the acceleration their head experiences when they tackle another person, so that has led to this notion that not only will we design helmets that will be safer, we will also have a way of monitoring exposure in players. But, we have yet to figure out how to interpret this data correctly. I also think there will continue to be a push for finding ways to identify patients that have had concussions and which of them, through whatever tests or indicators, will have persistent problems from concussions. If there’s something we can do to identify if they are experiencing very early signs of CTE, if there’s something we can do with those patients at that time, maybe we can change the course of the disease before it really gets bad. So I think we’re going to look early at the disease and treat it as it happens. PennScience: You also teach biomechanics and you’re the chair of the Department of Bioengineering. How does
your role as an as an educator fit into your research? Dr. Meaney: One of the things that you do as a researcher is you really drill into something, and then you’re forced to come back up a few levels. To present it at a level where people understand it as much as you do and that they’re interested in what you did, teaching is the same way. I find a similarity in teaching and educating and doing the research. It’s often said that a good researcher is a good educator, and they are correlated to each other. I think education is a good part of what we do and it’s embedded in the right people. We just have to find the mix of right people. Good educators become better researchers because of this process of taking something really complicated and detailed, and to come up a few levels and say I want to talk about this in a more conceptual way. PennScience: Why did you choose TBI? Dr. Meaney: So I did my graduate work in this area, and one of the classes I had was an independent study. We had rotations where we went to the neurointensive unit at Penn and talked to the nurses that were taking care of the severely injured patients. So you actually tour the ICU and you find out and see how devastating these brain injuries are. You realize how prevalent they are. Fifty thousand people die, and over two hundred thousand people have these severe head injuries every year where they’re never the same for the rest of their lives. And I looked at that, and I said, “Well, if I’m going to make a difference, this is something that I want to sink my teeth into.” And I’ve never looked back after that. PennScience: What advice do you have to give to undergraduates interested in neuroscience research? Dr. Meaney: Every undergraduate, while they’re here at Penn, should do some kind of research. Now this could be research at the bench, this could be research in the clinic, this could be research reading a series of text and doing some historical survey in the arts, sciences, or humanities. I think that research is an important thing for undergraduates to experience, because you don’t know the answer until you do research. Your education for the most part, up until the point of college has always been: study these boxes and we’ll evaluate it. Research is very different. Research is “Here’s the problem, and we’re not really sure what the best way to approach this problem scientifically is. So here’s a few approaches, try a few out, and be happy if one of them works, because most of them will likely fail.” Nothing is so well circumscribed in the world anyway, so what better place to learn that than college. So I would say utilize Penn and the research opportunities you can while you’re here, because there arelot of great people you can work with. There’s a lot of great resources.
FALL 2015 | PENNSCIENCE JOURNAL
23
RESEARCH Fabrication of Novel Nanomaterial Tungsten Disulfide (WS2) Nanopores for Solid-State DNA Sequencing Laura Beth Fulton Drndić Condensed Matter Physics Lab, University of Pennsylvania Nano/Bio Interface Center 2015 Research Experience for Undergraduates PI: Dr. Marija Drndić Mentors: Gopinath Danda, Paul Masih Das Tungsten disulfide (WS2) is a novel nanomaterial that offers promise to create nanopores for easy and efficient detection of DNA translocation with high spatial resolution. In this work, nanopore membranes are fabricated, a method is designed for the transfer of WS2, and WS2 flakes are characterized for nanopore application and transferred to membranes. Semiconductive properties of WS2 indicate promise for WS2 nanopores, which are drilled in the WS2 flakes. Silicon wafers were prepared using a standard method to form ~50 um x 50 um membranes for nanopore drilling. The created membranes were observed for uniformity and their size noted using optical microscopy. Scanning electron microscopy (SEM) was employed to visualize the membranes and focused ion beam (FIB) ~50 nm holes were sculpted. Flakes of WS2 were acquired and images were captured using optical microscopy, height profile characterized by atomic force microscopy (AFM), and Raman spectroscopy used to observe monolayer or multilayer flakes. A stamp transfer method was designed to achieve optimal alignment of one flake to a membrane. Testing of the stamp transfer indicated a touch, release set-up that provided visual observation using optical microscopy. This provided the ability to visually align WS2 flakes over membranes. Using minute motion, the flake was lowered to the microscope stage and contact was made with the membrane. The results of my contribution - wafer fabrication, SEM imaging, FIB drilled holes, designed stamp transfer, capturing optical microscopy images, atomic force microscopy (AFM) and Raman profiles for WS2 flakes, and transfer of WS2 to membrane indicate that the suspended WS2 is ready for nanopore drilling. Once a nanopore is drilled in the suspended WS2 flake, the Drndic group will conduct further experiments - Raman spectroscopy to check quality of 2D suspension, atomic force microscopy (AFM) to observe the suspended thickness and height profile, and methods to determine device noise and electronic frequency - to assist in determining the viability of WS2 nanopores as gateways for DNA sequencing.
Background
There is a renewed interest in nanopores in the field of nanotechnology as a new, efficient way of sequencing DNA, and as a priority for improving personalized medicine 1-3.DNA translocation measurements are obtained as an applied field drives strands of DNA in salt solution through a membrane pore. As the bases pass through, a sensor reads each base. Nanopores provide a gateway for sequencing DNA, which is a powerful method to reveal genetic variations at the molecular level, including gene fusion and insertion/ deletion, and is relevant to improving the understanding disease mechanism and genetic diagnosis 2. Solid state nanopore sensors consist of thin, highly insulated membranes constructed from synthetic materials which proffer pore flexibility 1. Silicon nitride (SiNx), an insulator, is a common membrane material used for nanopore device fabrication 5. At the University of Pennsylvania, experiments in the Drndic Laboratory have demonstrated DNA and single molecule translocation through 2D materials like graphene, 4-10 and experiments in the Radenovic group have worked on molybdenum disulfide (MoS2) nanopores 11. The major advantage of using 2D materials is that their signal is very high compared to the signal obtained from the nanopore membranes of thick materials like Si3N4 (thickness ~50-100 nm) 12, 13. Graphene nanopores are thin and flexible with good electronic conductivity and robust mechanical properties. However, graphene nanopores have relatively high noise; single nitrogenous bases are often
24 PENNSCIENCE JOURNAL | FALL 2015
not detected or are detected improperly. Thus, applications requiring intermediate metallic behavior are better served by nanopores fabricated from semiconductor materials 1, 5. MoS2 nanopores are an inorganic analogue of graphene with semiconductor properties that can be used for DNA biosensing 11. Nanopore membranes fabricated from tungsten disulfide (WS2), a novel 2D material, 14-16 can be used to detect DNA translocations. WS2 is a semiconductor and member of the transition metal dichalcogenides group. The semiconductor properties of WS2, similarly may allow for a viable alternative to graphene nanopore sequencing. MoS2 and WS2 have attracted attention from the scientific community due to their semiconductor band gaps, which permit intermediate metallic and insulating behavior due to strong covalent bonds and weak van der Waals stacking 16. Flakes of WS2 are being produced and studied, but nanopore experiments regarding electronic transport and nano-electric application have not been conducted with the material. Although MoS2 and WS2 are both members of the metal dichalcogenide group and have similar chemical properties, we expect to see differences in signal and noise characteristics.
Results and Discussion
Membranes were produced on SiO2 wafers using the MA6 mask aligner lithography to pattern the wafer. Reactive Ion Etching was used to etch the patterned wafer. Using optical
RESEARCH Key Words: WS2 , fabrication, characterization, DNA sequencing, 2D-materials Defined Terms: Dichalcogenide - containing two atoms of chalcogen (group 16 elements) per molecule or unit cell; layered materials with strong in-plane bonding and weak out-of-plane bonding interactions enabling exfoliation into 2D layers Semiconductor -material whose ability to conduct electricity is intermediate between that of a metal (conductor) and that of an insulator and is strongly temperature-dependent. Translocation - substitution, displacement of a chromosomal segment to a new position, especially one on a nonhomologous chromosome
Figure 1: Optical microscope images of membranes illustrating side size lengths (um). On the left, the membrane is etched fully as shown by the symmetrical geometry of its interior. The middle membrane is half etched; it’s right geometry matches that of the etched membrane on the left, but it’s left side is not symmetrical due to incomplete etching. The membrane at right unetched as seen from its lack of inner geometrical structure. All scale bars are: 34.13um.
microscopy, images of the membranes were captured, and s. Sizes of the membranes were measured. Distribution of membrane size was graphed for perspective; average size of the membranes was ~50 um withand standard deviation of ±4.2. The wafer was checked to see that the membranes were etched through. Membranes are etched through when an inner curvature is slightly risen on the square base. Images of the membranes were captured. Microscopy images illustrate that membranes were etched through and show membrane size (Figure 1). Membranes were observed using scanning electron microscopy (SEM) and, using a forced ion beam (FIB), pores were drilled in the membranes. With the focused beam, pores were drilled - diameters ranging from ~20 - 100 nm - in the membrane windows. Due to the possibility of the beam stigmation if thewith shaking of the FIB shook, some pores created were slanted rather than perfectly round. Adjusting the duration of the beam on the sample and HV could minimize slant. For the purpose of this experiment, it wais not necessary to have perfectly round FIB holes. However, minimizing slant could contribute to decreased noise in translocation for future experiments. Minimal slant was observed in creating the 50 nm pores using 1pA for 1 second (Figure 2). Flakes of WS2 used in this experiment were cultured by the Johnson lab at the University of Pennsylvania. Previously, in the Drndic lab, characterization and experiments were performed, creating nanopores with flakes of MOS2 . WS2 is a semiconductor structure similar to MOS2. Visualization of the WS2 molecular structures provides insight into its conducting behavior. WS2 exhibits van der Waals stacking between layers and laterally it exhibits strong covalent bonds laterally (Figure 3). WS2 The semiconductive nature of WS2 is contributed to the band gaps between layers promoting metallic behavior combined along with the insulating
behavior of the lateral bonding. In preparation for the WS2 flake transfer, a method was designed to ensure precise transfer and positioning of a flake on a membrane. Traditionally, a chip with flakes of 2D material is placed on the wafer with membranes. The transfer is random, so f. Flakes are not guaranteed to Figure 2: FIB hole 77nm in SiN 1pA land on membranes. for 1sec. Using this method, if contact is made, placement can be skewed, in which case only part of the flake covers the membrane. To increase the chance that flake transfer is successful, a stamp method was designed (see methods). The method was tested with flakes of graphene. Graphene flakes on a chip were suspended on the PDMS gel (Figure 4). The PDMS with the graphene flakes easily separates from the chip in a copper solution. The gel with the graphene is secured to the tip of a micromanipulator. The micromanipulator, which was positioned adjacent to the optical microscope at an angle that positioned the tip of the manipulator with the gel under the microscope lens. Under 10x magnification, flakes were observed. The wafer with membranes wais positioned beneath the suspended manipulator tip on the microscope stage; the two components weare separate (Figure 6). Adjusting microscope focus, a membrane wais identified. The fine motion of the micromanipulator alloweds the gel to be lowered, and this. The lowering wais observed on a screen hooked up to the microscope. Flake wais touched to
FALL 2015 | PENNSCIENCE JOURNAL
25
RESEARCH
Figure 3: WS2 molecular structure exhibiting lateral bonding and stacking. Created by Paul Masih Das, Drndic lab using Blender
Figure 4: On the left, is a silicon chip with graphene flake placed on the spin coater; in the middle, the PDMA-chip separation is illustrated - the clear PDMS with graphene flakes lies on top of the copper solution, the wafer at the bottom of the solution; on the right, is the resultant PDMS gel with graphene flakes
Figure 5: The micromanipulator was positioned adjacent to the optical microscope. Using minute motion, the apparatus is lowered to the microscope stage and the gel with flakes contacts with the membrane.
26 PENNSCIENCE JOURNAL | FALL 2015
the membrane and transfer wais complete. Testing of this method with graphene flakes demonstrates that this method will work with WS2 flakes. Flakes of WS2 were imaged using optical microscopy. To capture flake sizes, a 10x magnification was used. Sizes of WS2 flakes averaged ~20 um with a distribution with ranging from a minimum ~7 um flakes to a maximum of ~ 38 um (Figure 6). Flake size depended on distance from tungsten seeds (Figure 7a). The Johnson lab cultured the WS2 pby placing tungsten seed on a wafer with sulfide. The largest flakes were imaged closest to the seeds. The largest flakes were mostly multilayer and were not ideal for this experiment due to their thickness. Single layer flakes of WS2 were found at a distance from the seed. Size distribution of the single layer flakes were recorded using the optical microscope measurement tools. Coloration of the flakes under the microscope indicated thickness. The thicker flakes were observed to be bright blue; single layer flakes were not bright. Clusters of thick and single layer flakes were observed on the sample (Figure 7b.). Ideal flake clusters were imaged on the sample edges and towards the center of the sample, areas which were further from the tungsten seeds (Figure 7c.). Once sizes were recorded using the 10x magnification, the 100x magnification was used. Images captured under 100x showed magnified features of the flakes (Figure 7d). Most magnified flakes showed perfect triangles, buthowever, the edges on some flakes were damaged. Damaged flake edges appeared like swiss cheese. For this experiment, the flakes with undamaged edges will be were used. Atomic force microscopy (AFM) showed a height profile of the WS2 flakes. Topography of the flake surface was characterized with AFM. A height profile was obtained as the AFM cantilever oscillated, scanning the surface of a WS2 flake. Generated noise from the scan and roughness of the WS2 flake corresponded to thickness of the flake. Height profiles were obtained for monolayer and multilayer WS2 flakes (Figure 8). Raman spectrometry was compared to known values for WS2 (Figure 9), which showed the monolayer and multilayer distribution of the WS2 flakes (Figure 10). Transfer of the WS2 flakes to the membranes occurred using the designed stamp transfer method. The triangular
RESEARCH
Figure 6: Top box and whisker plot and belo histogram representations of size distributions of the side lengths of the WS2 flakes observed using optical microscopy
WS2 flake lays on the PDMS gel ready for transfer to the FIB hole in the membrane (Figure 11). Ultimately, transfer of flake to membrane proved unsuccessful. It is speculated that acetone was applied too soon to the PDMS with the flake once contact was made with the membrane. This did not allow the flakes to adhere to the membrane and leave the PDMS. In future experiments, a procedure will be designed where application of acetone is delayed so that the flake when the flake attaches to the membrane, it does so in a dry environment.
Materials and Methods
Wafer processing: SiO2 wafers weare prepared for this experiment. A Si wafer was cleaned using an O2 plasma clean for 5 minutes on both sides. To wash the wafer, acetone was applied to each side followed by isopropanol. Wafer sides were dried with N2. Resists weare applied to the wafer through a spin coating process. The wafer was spin coated on one side with an S1818 positive resist at 4000 rpm for 45 seconds. Following spin coating, the wafer was baked at 115°C for three minutes, with the uncoated side on the hot plate. After baking, the wafer was spin coated on the other side with negative resist (NR7). The wafer was then baked at 115°C for three minutes, with the S1818 side on the hot plate. Making windows using lithography and reactive ion etching (RIE): Using the MA6 mask aligner, one side of the wafer was exposed and patterned. The recipe was followed: 3.4 seconds at 365 nm at 5mW/cm2. The wafer was baked at 115°C for three minutes. Development occurred by placing the NR7 side of the wafer in RD6 for 7 seconds. After 7 seconds, the wafer was rinsed with DI water and dried with N2. RIE is used to etch the exposed nitride on the wafer. For a 50 nm wafer, etching occurred for 2 minutes; for a 100 nm wafer, etching occurred for 4 minutes. Following RIE, buffered oxide etching ( BOE) was performed on the exposed SiO2 layer. This layer was etched by subjecting the wafer to buffered oxide etching (BOE). The wafer was then submerged in a BOE bath. The wafer was tapped gently to remove bubbles and ensure the wafer surface was fully wet. The wafer was subject to the BOE for 70 minutes. After 70 minutes, an F-40 reflectometer was used to check if there was any SiO2 remaining. If needed, additional etching time was performed to remove SiO2 . Once the F-40 reflectometer indicated the SiO2 layer was removed, acetone was used
to strip photoresists. To etch the exposed silicon layer, a KOH etch was used. The KOH etch was prepared using a 40% potassium hydroxide (KOH) solution consisting of 1000 mL water and 666.67 g KOH. The solution was heated at 62°C and stirred at 120 rpm. An external probe monitored the temperature of the KOH solution. To etch 1 mm, KOH etching occurred for 22.5 - 23 hours. After 20 hours, the wafer was checked manually for visibility of light shining through membranes. When white light shined through the membrane, the etching was terminated. The wafer was cleaned using an acetone rinse followed by an isopropanol rinse. Membranes were observed using optical microscopy. S1818 photoresist was applied without spinning and let dry overnight to protect the unetched SiN layer. To remove the exposed SiN and SiO2 layers, the wafer was placed in a HF BOE bath for 100 minutes. Following etching, acetone was used to strip the S1818. Creating focused ion beam (FIB) holes: Using a forced ion beam, a focused beam of electrons was used to drill holes in membrane windows that ranged from 20 - 100 nm in diameter. The FIB machine was used following in accordance with a common procedure. The ion current was set to 2.2 ± 0.1 um, and t. The ion column HV was set to 5kV. Using a raster speed of 0.339 the membrane was scanned with SEM and positioned at a 52° angle. Angling the stage positioned the ion beam perpendicular to the membrane, allowing for the beam to enter straight on. Transfer of WS2 onto FIB hole: In the interim of the Drndic lab acquiring WS2 flakes, a stamp-transfer process was designed to improve the transfer and achieve optimal alignment of one flake of WS2 onto a FIB hole. To test this method, a SiO2 chip was used. On the chip, PMMA was spun at 1000 rpm for 30 seconds, and t. The chip was baked at 180°C for 3 minutes. PDMS mixture was made in a 10: 1 ratio of base to hardener . The PDMS acted as a gel that will be used to hold the flakes. PDMS was applied to coat the top of the chip and desiccated for 1 hour. Desiccation removed air bubbles. After 1 hour, the chip was removed and heated at 150°C for 10 minutes. A polyurethane quick cast was applied to secure the PDMS gel to a micromanipulator at a 30° angle. Once transfer had occurred, with the flake meeting the membrane, the gel wais removed with acetone. WS2 flake characterization: Flakes were characterized pre-transfer following optical microscopy, Raman spectrometry, and AFM protocols.
FALL 2015 | PENNSCIENCE JOURNAL
27
RESEARCH
Figure 7: Optical Microscopy of WS2 flakes. a.) Tungsten seeds (green/black) and surrounding WS2 flakes. Further from the seed, the flakes are primarily monolayer. b.) Clusters of multilayer (illuminated a light blue) and single layer (less bright) WS2 flakes. c.) Ideal, single layer WS2 flakes. d.) 100x magnification of WS2 flakes
Figure 9: Raman spectroscopy of WS2 flakes. This is a diagram of the expected spectra for WS2 with 532 nm layer. (Zhao et al. Nanoscale, 2013, 5, 9677 - 9683).
28 PENNSCIENCE JOURNAL | FALL 2015
Figure 8: Atomic force microscopy (AFM) of WS2 flakes. A flake is pictured with its width and height characterized. Generated frequency diagram corresponds to the flake’s layer thickness.
Figure 10: Raman spectroscopy. Based on the peak at roughly 310 cm-1, our spectra is most likely 2 layers or more.
RESEARCH Experience for Undergraduates (REU). Thank you to the National Science Foundation (NSF) for providing funding to support this research experience.
References and Notes
Figure 11: 10x optical microscopy images of WS2 flakes suspended on PDMS. Flakes are ready for transfer.
Transfer of WS2 onto FIB hole: A thin, PMMA gel was created to suspend the WS2 flakes pre-transfer. The gel was placed on a microscope slide with WS2 flakes open to air contact. The glass slide was attached to the micromanipulator at a 30° angle using the polyurethane quick cast. Once transfer had occurred, flake meeting the membrane, the gel wais removed with acetone.
Conclusions
This research demonstrated fabrication of membranes, drilling of FIB holes, setup for the transfer of WS2, AFM and optical microscopy of WS2 flakes. In future work, once the transfer of WS2 flakes has occurred, transmission electron microscopy (TEM) can be used to sculpt nanopores, Raman spectroscopy to check the quality of the 2D suspended membrane, atomic force microscopy (AFM) to observe thickness and height profile, and methods for determination of the device noise and electronic frequency employed. WS2 offers potential for the creation of nanopores as an easy, efficient way to sequence DNA at a low cost. The development of WS2 nanopores can be used to detect DNA translocation with high spatial resolution.
Acknowledgments
Thanks to the Drndic physics lab group, specifically Dr. Drndic for hosting me this summer and to my lab mentors Gopinath Danda and Paul Masih Das for encouraging my research curiosity through experiments. I would like to acknowledge the University of Pennsylvania Nano/ Bio Interface Center for hosting the 10 week Research
1. Edel, J.; Albrecht, B.;Andrew, W. Engineered Nanopores for Bioanalytical Applications. 2013. Norwich, N.Y. : Oxford: William Andrew. 2. Wang, Y.; Yang, Q; Wang, Z. The Evolution of Nanopore Sequencing. Front Genet. 2015, 5, 449. 3. Wanunu, M. Nanopores: A Journey Towards DNA Sequencing. Phys. Life Rev. 2013, 9, 125-158. 4. Fischbein, M. D.; Drndic, M. Electron Beam Nanosculpting of Suspended Graphene Sheets. App. Phys. Lett. 2008, 93, 113107. 5. Merchant, C. A.; Healy, K.; Wanunu, M.; Ray, V.; Peterman, N. DNA Translocation through Graphene Nanopores. Nano Lett. 2010, 10, 2915 - 2921. 6. Merchant, C. A.; Drndic, M. Graphene Nanopore Devices for DNA Sensing. Nanopore-Based Technology, 2012, 211 - 266. 7. Puster, M.; Rodriguez-Manzo, J. A.; Balan, A.; Drndic, M. Toward Sensitive Graphene Nanoribbon Nanopore Devices by Preventing Electron Beam - Induced Damage. ACS Nano, 2013, 7, 11283 - 11289. 8. Drndic, M.; Rodriguez-Manzo, J. A.; Qi, Z. J.; Johnson, A. T. Making and Electrical Biasing Graphene Nanoribbon Devices Inside the TEM. Bull.of the American Phys. Society, 2015, 60. 9. Drndic, M. Sequencing with graphene pores. Nature Nanotechnology, 2014, 9, 743. 10. Rodriguez- Manzo, J. A.; Balan, J.; Carl, N; Parkin, W; Puster, M; Johnson, A. T.; Drndic, M. The Effect of Defects Produced by Electron Irradiation on the Electrical Properties of Graphene and MoS2. Bull. of the American Phys. Society 2015, 60. 11. Liu, K. Atomically Thin Molybdenum Disulfide Nanopores with High Sensitivity for DNA Translocation. ACS Nano, 2014, 8, 2504-2511. 12. Schneider, G.; Kowalczyk, S.; Calado, V.; Pandraud, G.; Zandbergen, H.; Vandersypen, L; Dekker, C. DNA Translocation through Graphene Nanopores. Nano Lett., 2010, 10, 3163-3167 13. Liu, S.; Lu, B.; Zhao, Q. Boron Nitride Nanopores: Highly Sensitive DNA Single-Molecule Detectors. Adv. Matter 2013, 25, 45-54. 14. Zhao, W.; Ghorannevis, Z.; Chu, L.; Minglin, T.; Kloc, C.; Tan, P.; Eda, G. Evolution of Electronic Structure in Atomically Thin Sheets of WS2 and WSe2. ACS Nano 2013, 7, 791-797. 15. Zhang, Y. Controlled Growth of High-quality Monolayer WS2 Layers on Sapphire and Imaging its Grain Boundary. ACS Nano 2013, 7, 8963-8971. 16. Gutiérrez, H.; Perea-López, N; Elías, A.; Berkdemir, A.; Wang, B. Extraordinary Room-Temperature Photoluminescence in Triangular WS2 Monolayers. Nano Lett., 2013, 13, 3447-3454.
FALL 2015 | PENNSCIENCE JOURNAL
29
RESEARCH Social Stress Induces Dendritic Growth in Layer V Pyramidal Neurons of the PFC Eric Geng, Rita Valentino PhD, Kim Urban PhD Children’s Hospital of Philadelphia Department of Anesthesiology and Critical Care Medicine The prefrontal cortex (PFC) regulates high-level cognition, such as long-term planning and goal-oriented behavior. The PFC, however, is particularly vulnerable to stress. Even mildly acute stress can cause significant changes in PFC function and plasticity (Arnsten, 2009). This study examined the effects of repeated social stress on the plasticity, morphology, and membrane properties of layer V PFC pyramidal neurons. Neurons from male and female early-adolescent, mid-adolescent, and adult Sprague-Dawley rats were examined. The resident-intruder model was used as a social stress paradigm (Snyder, 2014). Neurons were Golgi-stained (FD NeuroTechnolgies, Inc. Elliot City, MD) and reconstructed using Neurolucida (MBF Bioscience inc, Williston, VT). Currently, this study has demonstrated that repeated social stress increases complexity and length of basil and apical dendrites for male adults. Furthermore, apical dendrite surface area was increased and basal dendrite surface area was reduced following social stress.
Introduction The prefrontal cortex (PFC) is responsible for higherlevel cognition, such as personality expression, long-term planning, and social behavior. The PFC has extensive connections with the amygdala, locus coeruleus, ventral tegmental area, and hippocampus. The PFC, however, is particularly susceptible to stress. The innervation of the PFC by norepinephrine (NE) and dopamine (DA) is powerful (Arnsten, 2009). Both catecholamines exhibit an inverted “U-shaped” influence on the PFC; too little or too much of either neurotransmitter impairs PFC function (Arnsten and Li, 2005). Under normal conditions, the PFC receives an optimal level of catecholamine input, maintaining regular function. Under stress, the amygdala triggers response pathways in the hypothalamus and brainstem, resulting in high levels of norepinephrine and dopamine release (Arnsten 2009). Cognitive effects depend on whether the stressor is acute or chronic. Chronic stress rapidly deteriorates PFC functionality. High levels of NE stimulate loweraffinity α1 and β receptors, which reduces PFC firing. Abnormally high or low levels of D1 receptor stimulation by DA lead to suppression of PFC neuronal firing in all directions, resulting in loss of spatial tuning and responsiveness. (Vijayraghavan, S. et al., 2007). These effects result in higher order cognitive impairments, such as reduced spatial working memory performance and attention regulation (Birnbaum, 2004; Arnsten, 2009). Blocking α1 and D1 receptors has been shown to alleviate these deficits (Arnsten, 2009). In contrast, acute stress strengthens memory consolidation and emotional associations facilitated by the amygdala (Elliot and Packard 2008). For example, high catecholamine levels strengthen amygdala-mediated fear conditioning and habitual behavior (Debiec, J. & LeDoux, 2006). Therefore, under stress, the brain’s responses shift from the thoughtful higher order processes to more habitual and emotional responses. Morphologically, stress has been shown to induce dendritic spine loss, reduce apical dendritic length, and eliminate axospinous synapses 30 PENNSCIENCE JOURNAL | FALL 2015
in PFC pyramidal neurons, which may account for impairments of PFC plasticity and function (Radley et al., 2006). Previous research has shown that repeated social stress increases corticotropin releasing factor (CRF) release on the locus coeruleus, resulting in increased NE release on the PFC (Van Bockstaele EJ, Valentino RJ, 2013). Earlyto mid-adolescents are most vulnerable to stressors due to slow maturation of the PFC (Arnsten and Shansky, 2004). This age marks a critical period of susceptibility to mental illness (Christie et al., 1988). Adolescent males, for example, are more prone to schizophrenia, and adolescent females face higher susceptibility to depression. During this developmental period, the PFC undergoes significant reorganization (Abel, 2010; Crews et al., 2007). For example, absolute PFC volume decreases (Sowell et al. 1999), along with a significant decline in excitatory glutamatergic inputs to the PFC (Huttenlocher, 1984). This period of remodeling may be a critical period of developmental plasticity, in which neurological circuitry is reshaped according to environmental needs, rendering adolescents more vulnerable to stressors (Crews et al., 2007). This research seeks to examine the effects of repeated stress on the plasticity, morphology and membrane properties of PFC pyramidal neurons. These properties were examined in early-adolescent, mid-adolescent, and adult male and female Sprague-Dawley rats. The resident intruder paradigm was used as a social stress model (Koolhaas et al., 2013). The Golgi impregnation method was used to stain tissue slices for neurons, so that morphology could be reliably compared between stress groups and ages.
Methods Animals Male and female Sprague-Dawley rats were used as intruders (Charles River, Wilmington, MA). Male LongEvans retired breeder rats were used as residents. All were housed individually on a 12h light/dark cycle.
RESEARCH
Social Stress Social stress was conducted using the residentintruder model (Snyder, 2014). Male and female rats at early-adolescence, mid-adolescence, and adulthood were exposed to social stress or control conditioning for five consecutive days. Intruder rats were placed into the cage of a resident and allowed to interact. The two were separated with a wire mesh divider after the intruder assumed a defeated posture or 15 minutes elapsed; the divider remained in place for the remainder of the 30-minute period. The defeat latency, number of attacks, and latency to first attack was measured and averaged across all five days of social stress for each intruder. Defeat latencies were used to determine coping strategy of the intruder — previous research has shown that repeated social stress affects resilient versus susceptible coping phenotypes differently, with susceptible rats showing greater tendency to later anxiogenic and depressive behaviors (Snyder et al., 2015).
Golgi Staining Golgi staining was done using an improved and simplified version of the Golgi-Cox technique (FD NeuroTechnolgies, Inc. Elliot City, MD). Animals were anesthetized using isoflurane. The animal was decapitated and the brain was rinsed with double distilled water. Tissues were then placed in the impregnation solution, and stored at room temperature for two weeks. The impregnation solution was replaced after the first day. The container was swirled twice a week. Tissue was then transferred into rehydration solution and kept at room temperature for 72 hours; the solution was refreshed once after 24 hours. 100 Âľm sections were cut on cryostat at -20 oC and then mounted on Colorfrost plus microscope slides (Fisher Scientific). Sections were dried at room temperature and then rinsed in double distilled water two times, four minutes each. Sections were then stained for 10 minutes and then rinsed in double distilled water two times, four minutes each. Sections were dehydrated
FALL 2015 | PENNSCIENCE JOURNAL
31
RESEARCH
in 50%, 75%, and 95% ethanol, four minutes each and again in pure ethanol, four times for four minutes each, and finally, cleared with xylene three times, four minutes each and cover-slipped using Permount.
BSA (1:1000) for two hours at room temperature. After, slices were rinsed with PBS four times for 40 minutes and cover-slipped with Fluoromount G (Electron Microscopy Sciences, Hatfield, PA).
Immunohistochemical Staining Slices were rinsed in 0.1 M Potassium Buffered Saline (PBS) three times for 30 minutes. They were then permeabilized in 0.1 M PBS-Tx for 1 hour. Next, were slices rinsed again with PBS three times for 30 minutes and later blocked in 10% PBS-Tx-BSA for one hour. Slices were then placed in primary antibody solution (rabbit anti-phospho Glur1) diluted with PBS-Tx-BSA (1:500) and incubated overnight at 4 oC for 24 hours. Slices were then rinsed four times in PBS for 40 minutes and after placed in secondary antibody (goat anti-rabbit conjugated with alexa fluor 488) diluted with PBS-Tx-
Morphologic Analysis Layer V pyramidal cells of the PFC were reconstructed and analyzed using Neurolucida (MBF Bioscience Inc, Williston, VT). Cell somas and processes were all traced manually by a person blinded to the group. Measurements were obtained using the Neurolucida Explorer program. Branch order was determined using centrifugal branch ordering. The dendrites, apical dendrites, and axons were analyzed for length of longest tree, number of nodes, number of branches, surface area, volume, and complexity. Complexity was derived from the Dendritic Complexity Index (Pillai, de Jong, Kanatsu
32 PENNSCIENCE JOURNAL | FALL 2015
RESEARCH
and al., 2012). Complexity was calculated as the sum of the terminal orders and number of terminals multiplied by the ratio of total dendritic length to the number of primary dendrites. Sholl analysis was conducted by measuring the number of intersections of processes with concentric circles radiating from the cell body in 20 µm increments. These concentric circles ranged from 10 µm to 910 µm distance from the cell soma. Cell somas were analyzed for volume and surface area. Statistical Analysis The morphologic analyses were done using independent samples t-tests to compare adult control and defeat rats. Outliers were defined as values more than two standard deviations from the mean. For Sholl analysis, the number of intersections at each concentric circle was averaged across all animals in each group. A paired t-test was used to compare Sholl analyses between control and stressed rats.
Results
A total of 80 Layer V pyramidal neurons were reconstructed and analyzed from eight male adult rats. Four rats underwent five consecutive days of social stress, and four underwent control manipulation for the same amount of time. Apical dendrite lengths ranged from 1357.4 µm to 4468.7 µm for defeated rats and 726.1 µm to 4848.9 µm for control rats. Apical dendrite length was significantly longer in stressed rats than control rats (Control, 1661.53 ± 538.64 µm; Stress 2779.63 ± 988.55 µm, P = 0.00095). Basal dendrite lengths ranged from 762 µm to 3105.1 µm for defeated rats and 533.8 µm to 3428.4 µm for control rats. Basal dendrite length was significantly longer in stressed rats than control rats (Control, 1090.3 ± 429.00 µm; Stressed, 1696.73 ± 710.68 µm, P = 0.0095). Stressed rats, however, showed reduced axonal length (Control, 100.1 ± 35.31 µm; Stressed, 71.2 ± 39.71 µm, P = 0.049). Lengths ranged from 34.3 µm to 278.8 µm for control rats and 30 µm to 335.9 µm for stressed rats. There was not a trend towards increase in number of dendritic nodes in stressed rats. (Control, 6.87 ± 2.90; Stressed, 9.13 ± 4.19 P = 0.097). No difference was found
in the number of apical dendritic nodes (Control, 12.60 ± 7.60; Stressed, 16.29 ± 5.77, P = 0.15) or axonal nodes (Control, 0.13 ± 0.35; Stressed, 0 ± 0 P = 0.16). There was a near significant increase in apical dendritic branches in stressed rats (Control, 24.50 ± 13.47; Stressed, 34 ± 11.77, P = 0.057), and a trend towards increased dendritic branching (Control, 16.73 ± 6.27; Stressed, 22.06 ± 9.46, P = 0.081). No difference was evident in axon branching (Control, 1.56 ± 1.5; Stressed, 1.27 ± 0.99, P = 0.50). Sholl analysis indicated no significant difference in apical dendrite intersections (P = 0.56). However, there was a significant increase in intersections between basal dendrites and sholls in stressed rats (P = 0.0026). Complexity measurements did reveal increased complexity in both apical dendrites (Control, 139842.2 ± 153947.8; Stressed, 334333.7 ± 237699.3, P = 0.021) and basal dendrites (Control, 7444.47 ± 3806.10; Stressed, 16917.3 ± 12912.96, P = 0.015) for stressed rats. Apical dendrite surface area was increased in stressed rats (Control, 5855.18 µm² ± 1520.84 µm²; Stressed, 7139.55 µm² ± 1748.159 µm² P = 0.048). However, no difference was found in apical dendrite volume (Control, 3668.19 µm³ ± 1545.04 µm³; Stressed 3719.72 µm³ ± 1561.21 µm³, P = 0.39). Basal dendrites, on average, had a reduced volume in stressed rats (Control, 172.086 µm³ ± 108.68 µm³; Stressed, 90.20 µm³ ± 53.77 µm³, P = 0.030), but no difference was evident in average surface area (Control, 632.79 µm² ± 320.51 µm²; 486.28 µm² ± 205.66 µm², P = 0.18). No differences were evident in cell soma surface area (Control, 3155.78 µm² ± 692.00 µm²; Stressed 3460.23 µm² ± 900.85 µm²) or volume (Control, 9131.76 µm³ ± 2395.29 µm³; Stressed, 10321.4 µm³ ± 3169.82 µm³). Discussion Prior studies have suggested that other forms of chronic stress, such as restraint stress or social isolation stress lead to dendritic atrophy. Post-weaning social isolation stress in adult male Sprague-Dawley rats caused a reduction in dendritic spine density (Silvia-Gomez et al., 2005). Following three weeks of repeated restraint stress, rats of the same age and breed exhibited a 20% decrease in total apical dendritic length of layer II/III pyramidal neurons
FALL 2015 | PENNSCIENCE JOURNAL
33
RESEARCH
(Radley et al., 2004), suggesting a decline in synaptic transmission to those layers. This effect, however, was limited to just apical dendrites. The atrophy is reversible following a period of stress-free recovery (Wellman, 2001). Several studies have indicated that repeated stress increases glutamate release onto the PFC (Moghaddam, B., 2002). Prolonged exposure to high glutamate levels has been shown to induce dendritic degeneration though enhanced activation of extracellular-regulated kinase (ERK) (Stanciu et al., 2000). Abnormal ERK signaling has been correlated to cytoskeletal destabilization (Runden et al. 1998). Phosphorylated ERK is also expressed in high levels in distal dendrites of PFC pyramidal neurons following repeated foot-shock stress (Kuipers et al., 2003). This present study used social stress as a paradigm for repeated stress. In contrast to prior studies, social stress appears to increase apical and basal dendrite length and complexity in layer V pyramidal neurons of the PFC for adult males. No differences were observed in the number of branches or nodes, implying that increased complexity is mainly due to changes in length and possibly direction of processes. This study suggests that social stress may cause different neurochemical changes in contrast to other forms of chronic stress. A disruption in ERK signaling may not be caused by social stress in adult males, although further investigation is needed for confirmation. It has, however, been previously shown that repeated social stress increases corticotropin releasing factor (CRF) release on the locus coeruleus, resulting in increased norepinephrine (NE) release on the PFC (Van Bockstaele EJ, Valentino RJ, 2013). This finding suggests that morphological changes may arise from a NEdependent mechanism. In the case that neurochemical changes are identical to those induced from restraint stress and isolation stress, it would imply that layer V pyramidal cells respond differently to chronic stress than those of layer II/III. Social stress appears to induce morphological changes 34 PENNSCIENCE JOURNAL | FALL 2015
that increase input. Increases in dendritic surface area, complexity and length suggest the possibility of more inputs to the cell. However, the number of excitatory synapses was not measured. Interestingly, axon length was reduced suggesting a reduction in output from layer V pyramidal cells. These changes can likely impact cognition. For example, layer II/III pyramidal neurons provide a significant amount of excitatory input to layer V pyramidal neurons (Thomson and Bannister 2003). This connection is thought to play a major part in goaldirected behaviors (Fuster 1991; Goldman-Rakic, 1995). Therefore, morphological changes following social stress may alter goal-oriented behavior. These changes also result in heightened brain reserves through increases in dendritic complexity and length, which may convey resilience to disease. For example, basilar dendrites of layer V pyramidal cells in schizophrenia subjects had 40% fewer intersections in Sholl analysis, suggesting an association between the disease and reduced dendritic complexity (Black, 2004). It is unknown if these changes are permanent. Dendritic atrophy, for example, has been found to be reversible in hippocampal and layer II/III PFC neurons given a stressfree period of recovery. Future studies are needed to determine the extent to which morphological changes in layer V PFC cells are permanent following social stress. The PFC undergoes significant changes during midadolescence, which may heighten sensitivity to stress (Crews et al., 2007). Women are more prone to stressinduced disorders as well, suggesting that gender may be a factor. Future studies would have to use lactating female Sprague-Dawley rats for residents to assess whether gender and age also play a role in morphological restructuring.
References Abel KM, Drake R, Goldstein JM (2010) Sex differences in
RESEARCH schizophrenia. Int Rev Psychiatry 22:417–428 Arnsten AF, Li BM (2005) Neurobiology of executive functions: catecholamine influences on prefrontal cortical functions. Biological Psychiatry 57:1377-1384. Arnsten, Amy FT. “Stress signalling pathways that impair prefrontal cortex structure and function.” Nature Reviews Neuroscience 10.6 (2009): 410-422. Arnsten AFT, Shansky RM. Adolescence: Vulnerable period for stress-induced prefrontal cortical function? Ann NY Acad Sci 2004; 1021: 143–147. Birnbaum, S. B. et al.. Protein kinase C overactivity impairs prefrontal cortical regulation of working memory. Science 306, 882–884 (2004)
in developing human cerebral cortex. Am J Ment Defic 1984;88:488–96. Moghaddam, B., 2002. Stress activation of glutamate neurotransmission in the prefrontal cortex: implications for dopamine-associated psychiatric disorders. Biol. Psychiatry 51, 775–787. Radley, J.J., Sisti, H.M., Hao, J., Rocher, A.B., McCall, T., Hof, P.R., McEwen, B.S., Morrison, J.H., 2004. Chronic behavioral stress induces apical dendritic reorganization in pyramidal neurons of the medial prefrontal cortex. Neuroscience 125, 1–6. Radley, J.J., Rocher, A.B., Janssen, W.G.M., H
Black, James E., et al.. “Pathology of layer V pyramidal neurons in the prefrontal cortex of patients with schizophrenia.” Pathology 161.4 (2004).
Runden, E., Seglen, P.O., Haug, F.M., Ottersen, O.P., Wieloch, T., Shamloo, M., Laake, J.H., 1998. Regional selective neuronal degeneration after protein phosphatase inhibition in hippocampal slice cultures: evidence for a MAP kinase-dependent mechanism. J. Neurosci. 18, 7296– 7305.
Christie KA, Burke JD Jr, Regier DA, Rae DS, Boyd JH, Locke BZ (1988) Epidemiologic evidence for early onset of mental disorders andhigherriskofdrugabuseinyoungadults. AmJPsychiatry145: 971–975
Silvia-Gomez, A.B., Rojas, D., Juarez, I., Flores, G., 2003. Decreased dendritic spine density on prefrontal cortical and hippocampal pyramidal neurons in postweaning social isolation rats. Brain Res. 983, 128–136.
Crews F, He J, Hodge C. Adolescent cortical development: A critical period of vulnerability for addiction. Pharmacol Biochem Behav 2007; 86: 189–199
Snyder KPB, M.; Valentino, R.J. (2014) Cognitive impact of social stress and coping strategy throughout development. Psychopharmacology (Berl) epub ahead of print.
Debiec, J. & LeDoux, J. E. Noradrenergic signaling in the amygdala contributes to the reconsolidation of fear memory: treatment implications for PTSD. Ann. NY Acad. Sci. 1071, 521–524 (2006).
Sowell ER, Thompson PM, Holmes CJ, Batth R, Jernigan TL, Toga AW. Localizing age-related changes in brain structure between childhood and adolescence using statistical parametric mapping. Neuroimage 1999;9: 587– 97.
Elliott, A. E. & Packard, M. G. Intra-amygdala anxiogenic drug infusion prior to retrieval biases rats towards the use of habit memory. Neurobiol. Learn. Mem. 90, 616–623 (2008). Fuster JM. The prefrontal cortex and its relation to behavior. Prog Brain Res 87: 201–211, 1991. Goldman-Rakic PS. Architecture of the prefrontal cortex and the central executive. Ann NY Acad Sci 769: 71–83, 1995. Koolhaas, J.M., Coppens, C.M., de Boer, S.F., Buwalda, B., Meerlo, P., Timmermans, P.J.A. The Resident-intruder Paradigm: A Standardized Test for Aggression, Violence and Social Stress. J. Vis. Exp. (77), e4367, doi:10.3791/4367 (2013) Kuipers, S.D., Trentani, A., Den Boer, J.A., Ter Horst, G.J., 2003. Molecular correlates of impaired prefrontal plasticity in response to chronic stress. J. Neurochem. 85, 1312–1323. Huttenlocher PR. Synapse elimination and plasticity
Stanciu, M., Wang, Y., Kentor, R., Burke, N., Watkins, S., Kress, G., Reynolds, I., Klann, E., Angiolieri, M.R., Johnson, J.W., DeFranco, D.B., 2000. Persistent activation of ERK contributes to glutamate-induced oxidative toxicity in a neuronal cell line and primary cortical neuron cultures. J. Biol. Chem. 275, 12200–12206. Thomson AM and Bannister AP. Interlaminar connections in the neocortex. Cereb Cortex 13: 5–14, 2003 Van Bockstaele EJ, Valentino RJ (2013) Neuropeptide regulation of the locus coeruleus and opiate-induced plasticity of stress responses. Adv Pharmacol 68:405-420. Vijayraghavan, S. et al.. Inverted-U dopamine D1 receptor actions on prefrontal neurons engaged in working memory. Nature Neurosci. 10, 376–384 (2007). Wellman, C.L., 2001. Dendritic reorganization in pyramidal neurons in medial prefrontal cortex after chronic corticosterone administration. J. Neurobiol. 49, 245–253.
FALL 2015 | PENNSCIENCE JOURNAL
35
www.pennscience.org PennScience is sponsored by the Science and Technology Wing at the University of Pennsylvania.