Revolutions in Neuroscience: Tool Development: John Bickle

HYPOTHESIS AND THEORY
published: 08 March 2016

doi: 10.3389/fnsys.2016.00024
Revolutions in Neuroscience:
Tool Development
John Bickle 1,2 *
1
Department of Philosophy and Religion, Mississippi State University, Mississippi, MS, USA, 2 Department of Neurobiology
and Anatomical Sciences, University of Mississippi Medical Center, Jackson, MS, USA
Thomas Kuhn’s famous model of the components and dynamics of scientific revolutions
is still dominant to this day across science, philosophy, and history. The guiding
philosophical theme of this article is that, concerning actual revolutions in neuroscience
over the past 60 years, Kuhn’s account is wrong. There have been revolutions, and
new ones are brewing, but they do not turn on competing paradigms, anomalies, or
the like. Instead, they turn exclusively on the development of new experimental tools.
I adopt a metascientific approach and examine in detail the development of two recent
neuroscience revolutions: the impact of engineered genetically mutated mammals in
the search for causal mechanisms of “higher” cognitive functions; and the more recent
impact of optogenetics and designer receptors exclusively activated by designer drugs
(DREADDs). The two key metascientific concepts, I derive from these case studies are
a revolutionary new tool’s motivating problem, and its initial and second-phase hook
experiments. These concepts hardly exhaust a detailed metascience of tool development
experiments in neuroscience, but they get that project off to a useful start and distinguish
the subsequent account of neuroscience revolutions clearly from Kuhn’s famous model.
I close with a brief remark about the general importance of molecular biology for a current
Edited by:
Mikhail Lebedev, philosophical understanding of science, as comparable to the place physics occupied
Duke University, USA when Kuhn formulated his famous theory of scientific revolutions.
Reviewed by:
Yoshio Sakurai, Keywords: Thomas Kuhn’s model of scientific revolution, tool development experiments, metascience, gene
Doshisha University, Japan targeting techniques, optogenetics and DREADDs, motivating problem, hook experiments
Victor De Lafuente,
Universidad Nacional Autónoma de
México, Mexico
KUHN’S MODEL AND NEUROSCIENCE
Ralph Adolphs,
California Institute of Technology, Kuhn’s (1962) model of scientific revolution still pervades much discussion in science,
USA philosophy, and history. A dominant paradigm eventually emerges within a scientific field.
*Correspondence: All practitioners learn it—its practices, models, exemplars, and most prominent applications.
John Bickle Normal science ensues: these resources get developed further and the paradigm gets
jb1681@msstate.edu extended beyond its initial applications. Yet anomalies start to arise—applications which
seem intuitively promising for the paradigm, but which prove recalcitrant to absorption.
Received: 22 December 2015 Anomalies pile up, and some practitioners begin to notice patterns among them, which
Accepted: 22 February 2016
suggest the outlines of a new paradigm. Conservative practitioners hold to the still-dominant
Published: 08 March 2016
paradigm, and continue pursuing normal science; radicals, often younger scientists, flock
Citation: to the emerging new paradigm, to develop its components and extend its applications. At
Bickle J (2016) Revolutions in
Neuroscience: Tool Development.
this stage revolutionary science ensues, and evidence and rational argument are only two of
Front. Syst. Neurosci. 10:24. numerous coercive techniques employed. In successful revolutions the emerging new paradigm
doi: 10.3389/fnsys.2016.00024 wins out—often because the older defenders of the previously dominant paradigm retire or die.
Frontiers in Systems Neuroscience | www.frontiersin.org 1 March 2016 | Volume 10 | Article 24

Bickle Revolutions in Neuroscience: Tool Development
Its basic resources get fleshed out and extended, and a new period experiment planning in current neuroscience, with the potential
of normal science ensues, now under the direction of the new ‘‘to make scientific revolutionaries of us all’’ (Silva et al., 2014,
paradigm. p. 177). Landreth and Silva (2013) provide useful solutions to
This narrative is familiar, easy to recite, and since Kuhn’s first practical questions about implementing the program.
book has been applied almost algorithmically to numerous cases Like Silva et al. (2014) my goal in this article is to
from science’s history. Expectedly, it guides much thinking when answer a metascientific question. By ‘‘metascience’’ I mean
new scientific revolutions are heralded. Everybody knows what the attempt to understand scientific practice and products on
to look for: components of the dominant paradigm, troublesome scientists’ terms, as unencumbered by philosophical assumptions
anomalies, any patterns across them which suggest initial or presuppositions about what those practices or products
sketches of the revolutionary paradigm, and the various extra- ‘‘have to be or do’’ (Bickle, 2003, chapter 1; Bickle, 2008).
rational persuasive techniques that characterize the revolutionary In short, I am asking questions about science, only here my
stage. question is: what kick-starts actual neuroscience revolutions? Of
In light of its continuing influence, this article’s principal course, this question presupposes another. Which developments
philosophical theme is that, concerning neuroscience over constitute neuroscience revolutions—the real ‘‘game changers’’
the past 60 years, Kuhn’s model is wrong. There have been in the discipline which set the stage for so much novel work that
revolutions in its fields, and new ones are brewing. But these followed? Consider four developments, from the past 60 years:
episodes do not adhere to Kuhn’s model. When it comes to one
• single-cell recordings in vivo, and the subsequent assignments
of science’s hottest interdisciplinary endeavors, real revolutions
of ‘‘field’’ properties (sensory, motor, memorial, etc.) to
turn on a different dynamic. The failure of Kuhn’s model is also
individual neurons based upon their responses;
not due to neuroscience’s languishing in some ‘‘pre-paradigm’’
• manipulating individual protein components of intra- and
phase. It is easy to cast current mainstream neuroscience in
inter-cellular signaling pathways in genetically engineered
Kuhnian terms. The basic resources are there.1 The problem is
mammals (‘‘knock-outs’’, ‘‘transgenics’’), and measuring their
deeper. Kuhn’s famous account misses the principal hinge.
effects in vivo using well-established behavioral protocols;
My metascientific approach in this essay focuses on two
• imaging increasingly smaller regions of the functioning human
recent cases of neuroscientific revolution, both drawn from
brain; and
cellular/molecular neurobiology, and behavioral neuroscience
• most recently, using light stimuli to activate or inhibit
using animal models. These are the fields I know best. However, I
specific selected neurons in vivo, while the animal engages in
will suggest revolutionary cases from other areas of neuroscience,
behavioral tasks.
including one from cognitive neuroscience, which I suspect also
illustrate the metascientific points I will develop below. Each of these developments was a bonafide revolution.
The first came to define reductionistic neuroscience from
REVOLUTIONS IN NEUROSCIENCE: THE the late 1950’s through the 1980’s. The second enabled
ROLE OF TOOL DEVELOPMENT manipulation of hypothesized cellular and molecular causal
mechanisms with a precision that far outstripped the capacities
Silva et al. (2014) provide a recent foray into revolutionary of stimulating electrodes and pharmacological interventions, and
neuroscience. The authors derive a Framework of types of has subsequently become part of the practices across virtually
experiments, and a set of procedures for integrating results all of behavioral neuroscience. The third now practically defines
across experiments and labs, by analyzing landmark publications the field of cognitive neuroscience.2 And I’ll have much to
from the field of molecular and cellular cognition (MCC). Using say about the revolutionary nature of the fourth below. Each
rudimentary causal graph theory they then develop research of these revolutions stemmed directly from the development
maps of published results, show how to derive causal maps and justification of a new experiment tool—at least one novel
of pathways between neuroscientific phenomena from them, to neuroscience at its time. The first revolution stemmed
and demonstrate the potential for automating these procedures. from fine-tip, high-impedance conducting microelectrodes; the
Together these resources provide a huge aid for rational second from the application of gene targeting techniques to
mammals; the third on positron-emission tomography (PET)
1 These resources include: a set of textbooks that govern education into and less invasively, functional magnetic resonance imaging
the discipline; a set of methods taught routinely to aspiring trainees, and
(fMRI); and the fourth on technologies both for inducing
researchers learning new techniques; a set of agreed-upon experimental
results which inform new experiments (c.f. citation numbers for ‘‘classic’’ translation and synthesis of microbial opsin genes in mammalian
articles). What then explains the common impression, even among neurons and delivering the light stimulus to the manipulated
neuroscientists that no ‘‘governing paradigm’’ exists? Probably this attitude is
due to the explicitly interdisciplinary scope of the field. Within given fields of 2 There is a genuine dispute about whether cognitive neuroscience armed
neuroscience-certainly within cellular/molecular/behavioral neurobiology- with functional brain imaging methods is revolutionary. There is the well-
many of Kuhn’s basic conditions on ‘‘paradigm’’ are evident; and I suspect known ‘‘computational’’ approach in cognitive (neuro-) science, and the
this is true for other fields in the discipline with which I am less familiar. It is many problems of consciousness. Perhaps the safest remark is that I am least
also worth noting in this content how vague Kuhn’s concept is Masterman’s confident that the argument to be developed in this article correctly applies
(1970) well-known critical notice of Kuhn (1962) counted 21 distinct to cognitive neuroscience. A serious investigating of this case is an important
meanings in that work alone! I thank an anonymous reviewer for pushing me project for future metascientific research. I thank an anonymous reviewer for
to clarify my quick dismissal pf neuroscience’s still being ‘‘pre-paradigmatic.’’ reminding me about these complexities.

neurons. In short: understanding tool development is the key to This work eventually won Capecchi a share of the 2007 Nobel
understanding real revolutions in actual neuroscience. Prize for Physiology or Medicine, shared with co-contributors
Silva et al. (2014) recognize tool development as one of Martin Evans and Oliver Smithies. The press release for the
the three basic categories of experiments in MCC, at the most Prize heralded their work as ‘‘the beginning of a new era
abstract level of their Framework. But they then focus exclusively in genetics’’ (http://www.nobelprize.org/nobel_prizes/medicine/
on the nature of Connection experiments (a second of their three laureates/2007/press.html).
basic categories), which seek evidence for hypothesized causal By the late 1980’s long-term potentiation (LTP), a form of
relations between neuroscientific kinds, because Connection activity-dependent synaptic plasticity (enhancement), provided
experiments constitute the bulk of the published research in a beguiling hypothesized neurobiological mechanism for
MCC. The authors thus leave a metascientific treatment of Tool learning and memory. Originally discovered as a physiology
Development experiments for future work. But note that their laboratory oddity in the 1960’s, the authors of its first systematic
entire Silva et al. (2014) book is an argument for developing a new experimental investigation speculated explicitly about its
tool for revolutionizing experiment planning in neuroscience, potential role in memory (Bliss and Lømo, 1973). Much work,
namely, research maps. Their approach is illuminating. When both slice-physiological and behavioral in vivo, followed quickly.
neuroscientists self-consciously seek to revolutionize some aspect The circumstantial case for the LTP → memory connection
of their discipline, their principal strategy is to develop a new tool. was well known (Lynch, 1986). Then-recently discovered post-
synaptic N-methyl-D-aspartate receptors (NMDAR), activated
CASE STUDY #1: GENE TARGETING IN only during periods of strong membrane depolarization (to
MAMMALIAN NEURONS remove a magnesium ion channel block), which permitted
calcium (Ca++ ) and sodium (Na= ) influx when activated,
As a first step toward developing a metascience of revolutionary seemed ideally suited to be part of the mechanism for LTP
tool development in neuroscience, I examine two detailed case induction (although the experimentally confirmed molecular
studies, both drawn from the four examples I noted above. details of NMDAR-dependent LTP were then quite sparse). Still,
I will derive two concepts from investigating these cases: many neuroscientists doubted that the LTP → learning and
motivating problem and hook experiments. Silva et al. (2014, memory causal connection had been established experimentally,
chapter 3) coined the term ‘‘hook experiment’’ to denote the first with a reasonable level of scientific confidence.
experiments and publications which gained the field of MCC its Morris’s (1989) systematic, multiple-experiment study using
initial notice. I am building on that usage, although I will have D,L-2-amino-5-phosphonopentanoic acid (AP5; also known
more to say about some specific features of hook experiments. as APV) was a big step forward. AP5 is a potent and
These two concepts hardly exhaust a metascientific analysis selective NMDAR antagonist, blocking the influx of Ca++
of tool development in neuroscience, but they get an analysis into the post-synaptic neuron. Morris (1989) administered
started—and distance the subsequent account of revolution in AP5 intraventricularly into rats prior to their performing
neuroscience immediately from Kuhn’s model. a variety of hippocampus- and non-hippocampus-dependent
Consider first the application of gene targeting technology learning and memory tasks (mostly water maze tasks, known
into mammalian neuroscience.3 The basic outline of this case by Morris’s name). In rodent hippocampus slices, AP5 was
study is relatively well known within neurobiology (although known to block LTP. Morris (1989) showed that it also decreased
some of the details I’ll stress often get ignored). This is an memory performance in the hidden platform version of his
important reason why I start a metasceintific analysis of tool water maze task, but not the visual platform version. In the
development with this case: an unfamiliar project is often best hidden platform task, hydrophobic rodents must learn to find
illustrated initially by working with a well-known case. The a platform submerged below the surface of opaque water in
Benzer lab at Caltech had begun manipulating specific genes in a pool by learning its location relative to numerous distal
flies, and tracking their behavioral effects, as early as the late- visual cues on the walls of the room containing the pool.
1960’s; by the late-1980’s they and other labs had developed over This task is hippocampus-dependent; rodents with bilateral
20 specific learning and memory fly mutants (with expressive hippocampus lesions are impaired in their ability to learn
names such as ‘‘rutabaga’’ and ‘‘dunce’’). Might it be possible to the hidden platform’s location relative to the visual cues, as
extend such work to mammals, with their far richer behavioral compared to sham-lesioned controls. In the visual platform
repertoires? This possibility seemed intuitively remote, due to task, a single visual cue at the location of the submerged
the myriad complexity between gene expression and protein platform, like a flag on the platform emerging out of the water’s
synthesis, and behavior in mammals (as compared to flies); but surface, marks the location. This second water maze task is
was enhanced by the work of Capecchi at the University of Utah, not hippocampus-dependent; although rodents with bilateral
who first developed mammalian gene ‘‘knock-out’’ technologies hippocampus lesions are slower than sham-lesioned animals to
in the mid-1980’s. Thomas and Capecchi’s (1987) lab targeted learn the task initially, by the end of standard training periods
developmental genes in the mouse, but he asserted confidently their performance is statistically identical. These were exactly
that the technique should be applicable to any cloned gene. the pattern of results Morris (1989) obtained with AP5-infused
rats.
3 See Silva et al. (2014, chapter 3) for additional details of this case study and Intraventricular AP5 infusion also produced motor deficits in
its aftermath, aimed at developing a metascience of Connection experiments. the rats, but a careful subsequent experiment suggested that these

deficits were independent of water maze performance. Morris just mentioned lay others. Was Capecchi’s speculation correct,
(1989) fifth experiment showed that the dosages of AP5 used in that his gene targeting technique could work for any cloned
the memory experiments were sufficient to block hippocampus gene of interest? In particular, could it work for a gene
LTP in vivo without interrupting normal synaptic transmission. in neurons in vivo? Neurons are relatively delicate cells,
Nevertheless, selective pharmacological NMDAR antagonists susceptible to cell death in a variety of ways. The gene
like AP5 inevitably disrupt synaptic function in subtle ways, target chosen had to code for a protein significant enough
potentially interfering with activity in hippocampus circuitry. As in neuronal signaling pathways to block LTP if eliminated,
Silva et al. (1992b, p. 201) put this concern, perhaps the failure and have downstream consequences all the way to behavior
of learning Morris had painstakingly demonstrated ‘‘results not if it was to provide evidence for the LTP → (rodent spatial)
from the deficit in LTP, but simply from some other incorrect learning and memory hypothesis. These effects had to follow
operation of hippocampus circuits that lack NMDA receptor from the elimination of a single gene and its subsequent
function’’. Obviously, a selective NMDAR antagonist like AP5 protein product. Yet this disruption had also to be specific
could not unravel that potential confound. enough not to disrupt other aspects of synaptic function
Taken together, the predicted general applicability of in excitatory forebrain neurons. Experimenters would have
Capecchi’s gene targeting techniques to mammalian nervous to verify that the targeted gene’s transcription and protein
tissue, and the experimental demand to block LTP without production truly was eliminated. The targeted gene could
disrupting other aspects of synaptic function in order to not interfere with normal development, from the embryonic
test the alluring LTP → (rodent spatial) learning and stem cell stage when the mutation is engineered, up through
memory hypothesis, constituted the motivating problem for the adult life stage when the behavioral tests would be
the application of gene targeting techniques in mammalian administered. Brain development in the mutants had to be
behavioral neuroscience. Other possible solutions to the second normal, from functioning excitatory synapses all the way up
feature of this motivating problem seemed reasonable to pursue. to gross anatomy of hippocampus circuitry. The behaving
One might try to develop more specific pharmacological mutants had to possess normal vision, motor capacities, and
agents, along with more precise means of their delivery motivation to solve the behavioral tasks, so the specific
to specific neuronal targets in vivo. But this strategy was mutation had to leave all those phenomena intact. And perhaps
known to face limits. Even the most selective receptor most challenging: could all this be accomplished, in one fell
agonists and antagonists inevitably affect synaptic function; swoop?
receptor activation is the first (post-synaptic) step of that Notice that the centrality of this first metascientific concept
complex process. Drugs can leak from injection sites into of motivating problem for a new tool’s development introduces
surrounding tissue, particularly in behaving animals, and distinctively un-Kuhnian aspects to the subsequent account of
especially if larger doses are required to elicit desired behavioral neuroscience revolutions. First, experimental tools are just one
effects. Even the most specific drug typically has ‘‘off-target’’ component of Kuhn’s multi-component concept of a paradigm;
effects. A more promising strategy would seem to be to my claim here is that this one component is central to actual
manipulate individual proteins more directly, in specific intra- neuroscience revolutions. But notice that a new tool’s motivating
or interneuronal molecular pathways underlying quite specific problem is even more specific: on a specific tool developed for
physiological processes. Capecchi’s gene targeting procedure a specific experimental purpose. Second, a revolutionary new
seemed ideal for this task. By ‘‘knocking out’’ the specific tool’s motivating problem is not a Kuhnian anomaly. In the
gene coding for some judiciously chosen protein product, case at hand all serious practitioners in the field recognized the
one could abolish that protein’s specific contribution to the limits of the best existing pharmacological tool (AP5) put to use
phenomena of interest—LTP, rodent spatial learning and to achieve the overarching experimental goal of testing directly
memory, whatever—hopefully without disrupting other aspects the LTP → (rodent spatial) learning and memory hypothesis.
of synaptic function, due to the vastly increased specificity of It was widely accepted that the existing tools failed to achieve
the molecular intervention compared to the best interventions this specific experimental end; further applications of the existing
pharmacology could offer. But an immediate question loomed: tools were not expected to succeed beyond what Morris (1989)
which neuronal genes/proteins to target, to test experimentally had accomplished. Even the most methodologically conservative
the enticing LTP → (rodent spatial) learning and memory experimentalists recognized the need to develop some new
hypothesis? tool—at a minimum, some new pharmaceutical compound other
Like most motivating problems which generate revolutionary than a post-synaptic receptor agonist or antagonist, and a more
tool development in neuroscience, this one was multi-faceted: specific way to deliver it in vivo. This is not to say that no further
it was a combination of multiple technical problems with a experimental work with compounds like AP5 would be done. It
daunting interactive dynamic.4 Beyond the looming question still had numerous experimental uses. But its use alone could
4 We will see this same feature of ‘‘motivating problem’’ illustrated in our the tool’s invention. A broader metascientific study of tool development
second case study in the next section. However, it is an open and interesting must address this potential alternate dynamic; suffice it to say that the two
metascientific question whether most tool development in neuroscience, cases of revolutionary tool development discussed in this article answered
revolutionary or otherwise, is guided by similarly structured motivating to a recognized motivating problem. I thank an anonymous reviewer for
problems. Perhaps in some tool development histories all applications follow mentioning this alternative account.

not further confirm the hypothesized LTP → (rodent spatial) inserted the blastocytes into pseudo-pregnant females, bred
learning and memory causal connection; and no one was puzzled the resulting chimeric males with wild-type females, and
about why.5 after multiple crosses confirmed the expected Mendelian
The combination of features of a potentially revolutionary ratios (wild-type homozygous, wild type-mutation heterozygous,
new tool’s motivating problem illustrated in this case study mutation homozygous) for a non-lethal mutation. Homozygous
also explains why its initially successful hook experiments are α-CaMKII mutants completely lacked α-CaMKII messenger
so surprising to the research community. Even scientists ‘‘in RNA (mRNA) and protein in forebrain tissue, or any truncated
the know’’ about early attempts to develop the new tool assign form of them, but showed normal mRNA and forebrain
a low probability to these experiments succeeding. Due to the protein levels for the closely related β-CaMKII isoform. Coronal
subsequent reaction to these experiments’ publication, not to sections through hippocampus revealed no gross anatomical
mention the number of citations which follow quickly and abnormalities in cell types, distributions, or axonal pathways.
sustain, there is usually not much controversy over which Aside from ‘‘increased jumpiness’’ or ‘‘nervousness’’ when
experiments constitute a revolutionary new tool’s initial ‘‘hooks.’’ handled by humans, mutant mice behaved normally. The
They are the first published results using the tool: (1) typically in mutation had no effect on long-term survival under standard
a top journal in the field; which (2) specifically apply the tool to laboratory housing. In light of all these preserved features, wild-
the targeted population of experimental interest; and (3) address type littermates could be used as controls for experimental
a phenomenon in the targeted field of inquiry. For gene targeting mutant mice, for both slice-physiological and behavioral studies.
techniques in mammalian behavioral neuroscience, these initial Electrophysiological studies using hippocampus slices found
hook experiments were published in Silva et al. (1992a,b) no differences between α-CaMKII mutants and controls in
and Grant et al. (1992). All three articles were published in synaptic currents, in either NMDA or non-NMDA components.
Science (condition 1), all targeted a specific gene/protein in mice Measures included peak size, time course, and ratio of NMDA
(condition 2), and all tested effects of the genetic intervention to non-NMDA components. Nor were there any differences in
on LTP in hippocampus slices using then-state-of-the-art the dependency of NMDAR channel conductance on neuronal
electrophysiological techniques and measures, and long-term membrane voltage potentials (Silva et al., 1992b, Figures 3, 4).
learning and memory in water maze and other hippocampus- NMDAR function in α-CaMKII mutant mice slices was normal.
dependent (declarative, explicit) tasks (condition 3). So this initial hook experiment delivered successfully on one key
Alcino Silva, working in Susumu Tonegawa’s lab, used desideratum of the motivating problem: the targeted mutation
Capecchi’s gene targeting technique to knock out the gene for did not disrupt synaptic function. The key confound plaguing the
the α-isoform of the calcium-calmodulin-dependent kinase II most careful and specific pharmacological studies of the LTP →
α-CaMKII). Some features of this protein were then known. It (rodent spatial) learning and memory connection was resolved.
is highly enriched in the post-synaptic densities of mammalian But that part of the motivating problem was, relatively speaking,
forebrain excitatory neurons, including hippocampus and the easy part. Could this single targeted gene mutation produce
neocortex. It plays a role in NMDAR-dependent LTP. It is the needed physiological and behavioral effects?
activated by calmodulin loaded with intracellular Ca++ , whose Silva et al. (1992b) next investigated the second desideratum,
influx into the post-synaptic cell is through activated NMDARs. impaired LTP in hippocampus slices, using both field potential
Activated α–CaMKII phosphorylates numerous other post- recordings to survey populations of hippocampus neurons
synaptic proteins, thereby activating them, and itself remains and more sensitive whole-cell recordings. They demonstrated
activated via autophosphorylation after Ca++ influx ceases. In deficient LTP in mutant hippocampus slices. Slices from
these ways α-CaMKII already met some key requirements of a littermate controls showed normal tetanus-driven LTP for all
computational model of a molecular mechanism to strengthen time periods measured (up to 1 h after tetanus). Aside from a
synapses developed a few years earlier by Lisman (1985) and brief post-tetanus stimulus potentiation (about 1 min), synaptic
Lisman and Goldring (1988). Lisman’s model itself was a more strength in mutant slices was unchanged from baseline levels,
specific version of Hebb (1949) famous ‘‘fire together-wire and remained so even after a second tetanus with increased
together’’ speculation. pulse trains was delivered. In whole-cell recordings most all
Silva et al. (1992b) constructed the plasmid that disrupted hippocampus neurons in control slices exhibited normal LTP,
the α-CaMKII sequence, transfected the plasmid into mice while only a small fraction of α-CaMKII neurons did. Quantal
embryonic stem cells, injected the stem cells into blastocytes, analysis did reveal that the LTP induced in the small fraction of
α-CaMKII neurons displaying it was normal (Silva et al., 1992b,
5 An anonymous reviewer drew a potentially intriguing distinction between Figures 6–8, and Table 1). So this single gene/protein mutation
tool development, along the un-Kuhnian lines suggested here, and concept reliably diminished LTP in hippocampus neurons.
development in neurobiology, which might better match Kuhn’s account of Silva et al. (1992a) investigated the third desideratum on the
paradigm shift and anomalies. I admit that my focus on the neurobiology motivating problem, behavior in vivo in rodent hippocampus-
of memory in these case studies might be occluding the discussion from and non-hippocampus-dependent learning and memory tasks.
conceptual revolutions, because plastic synapses have dominated theorizing
α-CaMKII mutants were slower to learn the non-hippocampus-
and experimental research on memory and the brain for more than a century.
This point is worth serious consideration, but would take this article beyond dependent visible platform version of the Morris water maze
its intended scope of showing how central experimental tool development has task initially, but over a standard 2-day, 12-trial training period
been in neurobiological revolutions. quickly matched control performance. Interestingly, this pattern

mimicked both hippocampus-lesioned and AP5-administred four nonreceptor tyrosine kinases in different mouse mutants.
animals (discussed above). On the hippocampus-dependent Synaptic transmission in hippocampus slices encompassing
hidden-platform version, α-CaMKII mutants never learned to CA3-to-CA1 circuitry from each of the four mutants was normal
locate and mount the platform as quickly as controls, over compared with littermate controls, including maximum field
either a standard 3-day or 5-day training period. In probe trials EPSPs and paired-pulse facilitation (Grant et al., 1992, Figure
after training, where the platform was removed, mutants spent 2). LTP in CA1 neurons was normal in src, yes, and abl
significantly less time in the maze quadrant where the platform mutants, but impaired in fyn mutants, in both field EPSP and
had been located during training, and crossed the platform’s population spike. Strong intensity tetanus stimulation produced
training location significantly fewer times, than did littermate a reduced form of LTP in fyn mutants, and was blocked
controls. They also crossed the training location of the maze by applications of AP5. Whole-cell patch clamp recordings
significantly fewer times compared to controls in a small number indicated that the NMDA component of excitatory currents in
of random-platform trials interspersed during training, where the fyn mutants was normal; synaptic depolarization of the post-
the hidden platform was changed to a new location (Silva et al., synaptic cells produced by the input was adequate to activate
1992a, Figures 1–4). However, α-CaMKII mutants performed NMDAR function (Grant et al., 1992, Figures 4, 5). Behaviorally,
normally compared to controls on a water-filled plus maze task, just like the α-CaMKII mutants (and hippocampus-lesioned
which requires animals to use a single distal visual cue to learn and AP5-administered rodents), fyn mutants were initially
which arm holds the hidden platform, (in contrast with the slower to learn the visual-platform version of the Morris water
spatial relations between platform location and numerous distal maze task, but eventually matched control performance during
visual cues, required on the hidden-platform Morris water maze standard training durations. However, they never matched
task; Silva et al., 1992a, Figure 5). Thus the mutants’ failures to control performance on the hidden-platform version, either
learn the hidden-platform Morris water maze task were not due during standard (7-day) training, or on probe trials after training
to an inability to see the distal cues, or to learn an association on either time in target quadrant or number of target crossings
between escape and the distal visual environment. Interestingly, (Grant et al., 1992, Figure 6). LTP and behavior in hippocampus-
α-CaMKII mutants also demonstrated other subtle behaviors dependent spatial learning both were normal in src, yes, and
similar to mice with hippocampus lesions (in addition to the abl mutants. However, fyn mutants displayed a developmental
‘‘jumpiness’’ to human contact noted above), such as increased deficit, in the arrangements of hippocampus dentate gyrus
exploration and activity in open fields and enclosed Y-mazes. granule neurons and their target CA3 pyramidal neurons (Grant
Silva et al.’s (1992a) own words about their results illuminate et al., 1992, Figure 7).
the dual metascientific notions of motivating problem and These authors were likewise enthusiastic about the general
initial hook experiments developed here. They assert that applicability of this new gene targeting tool, now demonstrably
their work ‘‘strengthens considerably the contention that the feasible for mammalian behavioral neuroscience by work from
synaptic changes exhibited in LTP are the basis for spatial two labs: ‘‘In addition to their role in the study of behavior
memory.’’ Perhaps more surprisingly, and for the first time, and learning, targeted disruption of genes provides a powerful
it ‘‘demonstrates that a mutation in a known gene is linked tool for examining the role of specific proteins in the function
to a specific mammalian learning deficit, and indicates that of the brain’’ (Grant et al., 1992, p. 1908). And these initial
single genetic changes can have a selective but drastic impact hook experiments stuck. In October 1992, 4 months after
on learning and memory.’’ Finally, and with an eye to the the Silva et al. (1992a,b) articles had been published and
future use of the tool they had demonstrated as applicable to 2 months before the Grant et al. (1992) article appeared,
mammalian behavioral neuroscience, they predict that ‘‘other Morris himself, with Mary Kennedy, published a review, ‘‘The
similarly constructed mice with mutations in judiciously chosen Pierian Spring’’ in Current Biology, with the subtitle ‘‘mutant
genes will be useful for studying mammalian behavior’’ (Silva mice engineered to lack an enzyme critical for long-term
et al., 1992a, p. 210). synaptic plasticity are deficient in spatial learning’’ (Morris and
Learning and memory neurobiologists did not have to wait Kennedy, 1992, p. 511). They outlined the genetic engineering
long for this prediction to be met. Five months later another procedures used, and the electrophysiological and behavioral
set of initial hook experiments was published. Grant et al. investigations employed, for non-specialists, and closed with a
(1992) working in Eric Kandel’s lab, engineered mutations in section on ‘‘Implications and potential.’’ The α-CaMKII knock-
mice to the genes for each of the four nonreceptor tyrosine out was ‘‘an ingenious piece of molecular engineering’’; the
kinases, src, fyn, yes, and abl. The motivating problem for LTP and behavioral deficits ‘‘were by no means a foregone
Grant et al. (1992) work was that for Silva and collaborators: conclusion’’; the work ‘‘should be recognized as the considerable
the widely accepted limits of pharmacological agents and achievement it truly represents.’’ The findings ‘‘vindicate and
their method of application adequately to test the LTP → extend earlier results’’ (namely, Morris’s own!) (Morris and
(rodent spatial) learning and memory connection. Tyrosine Kennedy, 1992, p. 513). The authors note that this work
kinase inhibitors had been shown to block the induction of was not without its own problems. The learning capacities in
LTP without affecting normal synaptic transmission, but the mutants might have been underestimated. The subtlety of the
existing drugs lacked the specificity necessary to identify the effects of eliminating so significant a post-synaptic protein in
specific tyrosine kinases involved. Grant et al. (1992) used forebrain excitatory neurons might reflect compensatory effects
Capecchi’s gene targeting technique to knock out each of these of other CaMKII isoforms. α-CaMKII was eliminated completely

from the mutants’ brains, including from all regions and pre- developed a homozygous knock-out mouse deleting the gene
synaptically, and so these specific mutants could not be used for the α- and δ-isoforms of cAMP-response element-binding
to address the then-still raging controversy over whether LTP protein (CREB). CREB is a transcriptional enhancer prominent
was mediated pre-synaptically, post-synaptically, or both. α- in many tissues. Earlier work with flies, including gene targeting
CaMKII is also prominent in neocortical excitatory neurons, work, and with the sea slug Aplysia californica, suggested that
and consolidation and long-term storage of spatial memories CREB, when activated via phosphorylation, enhanced expression
probably occurs there, in addition to hippocampus; that phase and synthesis of a variety of genes and proteins important for
of memory induction may too depend on α-CaMKII activation, LTP (or for ‘‘long-term facilitation,’’ as the process is called in
which would also be blocked in the mutants. Despite these invertebrates) and for invertebrate forms of associative learning.
problems, however, these first hook experiments were ‘‘an Working with Schütz’s CREB mutants in Silva’s lab at Cold
auspicious beginning and likely to fund a small industrial Spring Harbor, Bourtchuladze et al. (1994) showed they were
revolution.’’ Rather than treading lightly with this new tool, deficient in long-term memory tasks (24-h delay), but intact in
neuroscientists interested in the brain mechanisms of learning learning and short-term memory (30–60 min delay) on these
and memory ‘‘should, as Pope went on to write, ‘‘Drink deep, same tasks. Behaviors included both hippocampus-dependent
or taste not the Pierian Spring’’6 Morris and Kennedy, (1992, (single-training trial aversive contextual fear conditioning, where
p. 514). the animal is conditioned with a foot shock after initial exposure
Neuroscientists, and not just those working on learning to a new environment, and the hidden-platform version of
and memory, certainly heeded Morris and Kennedy (1992) the Morris water maze) and non-hippocampus-dependent tasks
recommendation. Gene targeting quickly became standard (single-training trial cued Pavlovian aversive conditioning to
practice across neurobiology, which greatly expanded the then- an auditory tone). These results suggested that CREB plays a
nascent search for molecular mechanisms of ‘‘higher’’ functions. key role in the consolidation of memory from short-term to
This new experimental tool brought molecular neuroscience to long-term form. Hippocampus slice-physiology work proved
the prominence in the discipline it maintains to the present day. consistent with this interpretation of the behavioral results. LTP
But the initial hook experiments which expose a new tool to in mutant slices was smaller than littermate controls and declined
the purview of professional scientists are just one component to baseline by 90 min (Bourtchuladze et al., 1994).
of genuinely revolutionary tool development. A second kind CREB mutants immediately became the target of extensive
of hook experiment garners the tool even wider appeal and behavioral neuroscience investigations, and the same pattern
application. Identifying these ‘‘second-phase’’ hook experiments of results emerged for an large variety of rodent memory
for any given revolutionary tool is more controversial, as tasks: intact short-term performance but impaired long-term
performance on the same tasks. This pattern held both for
there are often a number of viable candidates. For gene
standard memory consolidation and for reconsolidation after
targeting techniques in behavioral neuroscience, however, the stimulus re-presentation. The importance that CREB mutants
general shape of these second-phase hook experiments was played for learning and memory research over the next few
anticipated by both Silva et al. (1992a) and Grant et al. (1992): years after Bourtchuladze et al.’s (1994) second-phase gene
engineered genetic mutations more specific than whole-gene targeting hook experiments is nicely summarized by Dudai
knock-outs. Silva et al. (1992a, p. 210) write: ‘‘Perhaps even (2002, p. 65) in his introductory sourcebook for memory
more useful would be . . . mice with mutations directed to research:
specific regions of the brain. Construction of such mutant
mice may be feasible’’. Grant et al. (1992, p. 1908): ‘‘To CREB is one of the most commonly used acronyms in neurobiology
strengthen the links between Fyn, hippocampus LTP, an these days, and also one of the few words in the jargon of molecular
spatial learning, it will also be necessary to specifically biology that even experimental psychologists and computational
neuroscientists might have encountered. And if they didn’t, they
manipulate the expression of mutant forms of Fyn, restricted
should. Because the more we advance our knowledge in molecular
only to the hippocampus’’. The continued development and
biology the more we realize that CREB plays a pivotal role in the
use of gene targeting in mammalian behavioral neuroscience response of neurons to external stimuli.
quickly exceeded both groups’ expectations. Not only did the
regional specificity both envisioned follow quickly, but so did Dudai’s remarks demonstrate one crucial feature of second-
temporal spexcificity, the capacity to express engineered gene phase hook experiments for genuinely revolutionary tools:
mutations only during specific developmental phases, e.g., in widespread dissemination of the tool and its results to a
adult rodents, and even at specific times during behavioral wider public, beyond the specialists working in the field that
protocols. developed it.7
A new molecular target for learning and memory studies Neuroscientists were also soon attracted to a different
emerged first. Günther Schütz’s developmental biology lab kind of gene engineering technique, the transgenic approach.
This approach involves inserting an extra copy or copies
6 Morris and Kennedy (1992) drew their title from this line of Alexander of a cloned gene into the DNA of mammal embryonic
Pope’s early-18th century poem, ‘‘An Essay on Criticism.’’ The line of the stem cells, often attached to a promoter region which limits
couplet that precedes this quoted line, also quoted by Morris and Kennedy
(1992) at the beginning of the review, is perhaps the most famous line from 7 We will also see this point illustrated in the second case study of
the poem: ‘‘A little learning is a dangerous thing’’. neuroscience revolution in the next section.

expression to specific neurons. Every cell in the mutants’ on short-term (1-h delay) contextual fear conditioning,
bodies contains the extra transgene copy or copies; but significantly impaired on the long-term (24-h delay) version,
transgene transcription (and subsequent protein synthesis) but unimpaired on both short-term and long-term versions of
only occurs in those tissues or specific cells possessing tone-foot shock (Pavlovian) conditioning compared to control
the promoter molecule in sufficient abundance. An early mice. The latter task is amygdala-dependent, where the R
influential use of this approach in behavioral neuroscience transgene was not expressed significantly due to the CaMKII
was Abel et al. (1997) in Kandel’s lab. They inserted promoter.
extra copies of the gene for regulatory subunits of the Two more recent second-phase hook experiments bring us
cAMP-dependent protein kinase A (PKA) molecule into up to date on the development of gene targeting targeting
mouse embryonic stem cells, with a CaMKII promoter techniques in behavioral neuroscience. The first used conditional
which limited transgene expression to forebrain regions, gene knock-outs, which temporarily and reversibly repress a
including hippocampus (but excluding significant expression targeted gene’s expression and protein synthesis. An important
in amygdala). PKA plays a prominent role in the cAMP- hook experiment for this development was Kida et al. (2002),
PKA-CREB intracellular signaling pathway. In mammalian who fused a CREB α-isoform repressor (IR, with an alanine
post-synaptic neurons, increased intracellular cAMP (driven residue substituted for the serine residue at positon 133) to a
by activation of dopaminergic modulatory neurons) binds ligand-binding domain of a human estrogen receptor mutated
to regulatory submits of PKA molecules. This binding frees to become activated by the drug tamoxifen (TAM). Kida et al.
catalytic PKA subunits to translocate to the neuron’s nucleus (2002) derived transgenic mice expressing this CREBIR mutation
and phosphorylate CREB molecules. Phosphorylated CREB in under the control of an α-CaMKII promoter to render it active
turn drives gene expression and synthesis for both regulatory only in excitatory forebrain neurons. The CREBIR isoform
and effector proteins which restructure active synapses, leading protein competes with phosphorylated CREB for binding sites
to increased EPSPs to subsequent glutamate pre-synaptic on target regulatory and effector genes, and represses expression
release. In Abel et al. (1997) R transgene mutants, the extra at those sites, but only for the duration of systematic TAM
R PKA (regulatory) subunits available in neurons in which administration. At all other times the CREBIR transgene is silent
the transgene is expressed quickly bind up PKA catalytic with no effect on CREB activity (including throughout brain
subunits freed by increased cAMP, blocking that early step development). CREBIR activation did not disrupt short-term
in the cAMP-PKA-CREB pathway. At the time that these memory (2-h delay) for (hippocampus-dependent) contextual
second-phase gene targeting hook experiments were carried fear conditioning, but disrupted its consolidation into long-
out, CREB’s role in LTP and memory consolidation had term form (24-h delay). CREBIR activation also impaired the
already been established, with Bourtchuladze et al.’s (1994) stability of reactivated fear conditioned memories in mutants,
CREB− mutants (see above). But CREB is phosphorylated Mice were returned to the context paired with the foot shock
through numerous intracellular signaling pathways, and the 24 h after initial training (the ‘‘reactivation’’ phase) and were
CREB knock-out mice couldn’t distinguish between which of then retested in the context 24 h later. Transgenics whose
these pathways was crucial for its role in late-phase LTP CREBIR repressor isoform was activated by TAM during
and memory consolidation. Attempts to knock out the PKA memory reactivation froze significantly less to the context at
gene using Capecchi-style gene targeting techniques had been retest than did the three control groups: wild-type littermate
inconclusive. controls, transgenics whose CREBIR repressor was not activated,
Abel et al.’s (1997) founder mutant mice bred successfully and transgenics whose CREBIR was reactivated but whose
and transmitted the transgene to offspring. The promoter context-fear memories were not reactivated by context re-
limited significant transgene expression to forebrain areas, presentation. The results from this second-phase gene targeting
including all regions of the hippocampus, but hippocampus hook experiment with a temporally conditional knock-out
gross anatomy was otherwise unaffected. Hippocampus mutation provided a crucial piece of evidence that CREB
PKA activity was reduced in R transgenic mice, but basal is part of the molecular mechanism for both consolidation
synaptic transmission, paired-pulse facilitation, and post- and reconsolidation of hippocampus-dependent (‘‘declarative’’)
tetanus potentiation to one- or two-train tetanus stimuli memory.
were all unaffected in mutant hippocampus Schaffer The use of engineered virus vectors to deliver transgenes
collateral (CA3→CA1) pathways. However, late-phase via the usual process of the virus replication cycle also enabled
(L-) LTP was reduced significantly to four-train stimuli scientists to bypass possible developmental abnormalities which
in mutant slices for all time periods tested (starting at result from insertions in embryonic stem cells, and to insert genes
about 40 min and measured up to 180 min post-tetanus). of interest directly into the neurons of adult mammals. Delivery
Behaviorally, while escape latency during training on the of virus vectors via microinjection further limited infection, and
hidden-platform version of the Morris water maze task hence transgene transfer, to specific neural regions. An early
in R transgenic mutants did not differ from littermate application of this technique to behavioral neuroscience was
control performance, mutants were deficient in both time- Josselyn et al.’s (2001) work, in Michael Davis’s lab. Rodents
in-target quadrant and number of target crosses in probe show increased L-LTP, and increased long-term learning, in
trails administered after training (Abel et al., 1997, Figure tasks requiring multiple training sessions when training sessions
5). More importantly, R transgenic mutants were intact are spaced, with non-training intervals in between, rather than

when massed into a single multiple-exposure session. Josselyn some of its resources have been standard in molecular biology
et al. (2001) showed that rats who received infusion of a for three times that long). This is an especially interesting case
transcription enhancer isoform of the CREB gene, with the of scientific revolution, not only because the science itself is so
virus vector microinjected bilaterally into the basolateral nucleus intriguing and was initially so implausible, but also because it
complex of the amygdala, showed significantly enhanced long- is a revolution still in the making. Yet it is difficult to deny
term memory with massed training on a fear-potentiated startle that optogenetics’ impact on neuroscience has already been
paradigm involving four light-shock pairings. No control groups revolutionary. Since it was named ‘‘Method of the Year’’ in 2010
showed any increase in long-term memory performance to by Nature, the number of optogenetics publications searchable
massed training: rats microinjected with a control substance, annually has increased from roughly 100 to already close to 900,
rats microinjected with a mutated inactive form of the CREB through only the first half of 2015 (Deisseroth, 2015). Recent
gene, rats microinjected with a Lac-Z gene with no learning reviews discuss experimental results in mammalian behavioral
and memory effects, rats whose microinjections had missed or neuroscience ranging from learning and memory, sleep/wake
whose infections had extended beyond the amygdala basolateral transition, addiction, motivation, reward, social interactions,
nucleus, and rats microinjected with the CREB gene directly anxiety, and models of neurological disease and trauma.8
into the caudate nucleus. There were no differences across Being an ongoing, still potential revolution implies, of course,
groups in immediate reactions to shocks, explicitly unpaired that the optogenetics revolution might not come to fruition.
training failed to produce long-term memory in the CREB- Otchy et al. (2015) report results which challenge an assumption
boosted rats, and increased long-term memory was retained in behind the experimental use of rapid and reversal manipulations
CREB-boosted rates 14 days after massed training. As Josselyn of neural activity like optogenetics, that the observed behavioral
et al. (2001, p. 2410) remark, their results were first to show that effects reflect the function of the manipulated circuits. This
‘‘overexpression of CREB in a specific mammalian brain region assumption is problematic because of indirect effects of the
at a specific time enhances LTM [long-term memory]’’. These manipulation on independent downstream circuits, which are
types of experiments, labeled ‘‘mimicry’’ by Sweatt (2009) and difficult to control. So noted. Despite its role in kick-starting
‘‘positive manipulations’’ by Silva et al. (2014), provide a kind of neurobiological revolutions, tool development is just one aspect
evidence widely recognized as necessary for scientific confidence of progressing science, revolutionary or otherwise.9
that a hypothesized causal connection, e.g., CREB expression in One feature that makes optogenetics appealing as an
specific brain regions → enhanced LTM, has genuinely been experimental tool is the unprecedented control it offers over
established. the activation (or inhibition) of specific targeted neurons in
The gene targeting revolution across mainstream living, behaving animals. Various classes of light-responsive
neuroscience is now fully accomplished. The technique is proteins, most prominently the microbial opsins from algae,
a mainstay of work in cellular, molecular, and behavioral serve as membrane-spanning channels for cation (e.g., sodium,
neuroscience across all phenomena. Any research institutions Na+ ) or anion (e.g., chloride, Cl− ) influx, thereby depolarizing
of any serious standing now has a core facility to develop or hyperpolarizing the cell when activated. The DNA coding
conditional knock-out and transgenic rodents (and other for these opsins is removed and engineered to be expressed
mammals), not just for neurobiologists, but for all fields of effectively in mammalian DNA. It is inserted via a virus vector,
bio-medical inquiry. The vast increase in our knowledge of nowadays typically an engineered adeno-associated virus whose
direct cellular and molecular mechanisms for ‘‘higher-level’’ own genetic material has been removed to allow insertion of
biological phenomena of all types can be traced to the application the engineered opsin DNA. The virus vector is then micro-
of these techniques. To deny this scientific development the injected directly into specific regions of mammalian brains,
status of a revolution would be simply to misunderstand its typically mice or rats. Specific neurons infected by the virus
significance in current scientific inquiry. But the actual dynamics take up the engineered opsin DNA by the usual mechanisms
of this revolution, in recent neuroscience as in all other fields, of the viral replication cycle. They are thus poised to begin
hinged on the development of this tool, especially the roles of expressing the engineered gene and synthesizing the new protein,
its motivating problem, and initial and second-phase hook under the control of the promoter region of the engineered gene.
experiments. These features, rather than anything deemed These synthesized opsin proteins embed in the infected neurons’
crucial on Kuhn’s famous model, were this scientific revolution’s membranes. They now can be activated by a light stimulus to pass
vanguard. the specific cation or anion into these neurons, thereby activating
or inactivating them at the experimenters’ command—by
switching on the laser or light-emitting diode (LED) light source
CASE STUDY #2: OPTOGENETICS AND
DESIGNER RECEPTORS EXCLUSIVELY 8I will have more specific things to say in defense of optogenetics’
ACTIVATED BY DESIGNER DRUGS revolutionary nature after I explain the technique, canvass some results, and
apply the notions of motivating problem and hook experiments to these
(DREADDs) details.
9 See Silva et al. (2014, Chapter 1) for a broader perspective of how Tool
For a second detailed case study of an actual neuroscientific Development experiments fit with other kinds of experiments in science. I
revolution, consider optogenetics, an experimental technique in thank an anonymous reviewer for reminding me about this concern, and for
existence in neuroscience for scarcely one decade (even though pointing me to the Otchy et al. (2015) article.

embedded in the animal’s brain. The stimulating light source range of the stimulating light source are affected. This extent
is inserted through chronically implanted cannulae cemented of the drug’s effect is a two-edged sword, however. It puts
permanently into the animals’ skulls. Activating (or inactivating) a premium on micro-injection placement of the DREADD-
specific neurons can have downstream effects of activating (or containing virus and subsequent verification of this placement.
inactivating) entire neuronal circuities. Typically the engineered Leakage of the virus into other non-target neurons, and
DNA for an imaging protein (mCherry, green florescent protein, subsequent expression of the designer receptor in them, can
yellow fluorescent protein) is also coupled to the engineered confound behavioral results. On the other hand, DREADDs’
DNA for the opsin protein in the virus vector. All neurons temporal control is tied to the pharmacokinetics and -dynamics
infected by the virus will also express the imaging protein, and of the designer drug—its absorption, distribution, metabolism
glow a particular color (red, green, yellow) in brain tissue slices into inactive metabolites, and excretion. This slow and imprecise
under standard light microscopy. Experimenters can thus verify temporal effect contrasts sharply with optogenetics’ temporally
exactly which neurons expressed the opsin protein, and were much more tightly controlled on/off light switch. Both
activated (or inhibited) by the light stimulus (this is all now techniques permit anterograde-retrograde interactions, in which
standard procedures for the use of viral vectors in gene targeting, the specific neuronal targets of the axons of the optogenetics-
introduced toward the end of the previous section). and DREADD-activated neurons can be identified and in turn
Those are the basics. Further refinements to these techniques manipulated.
typically depend upon the specific experimental questions The two metascientific concepts illustrated in the case
optogenetics is used to address. For example, memory of revolutionary gene targeting techniques in neuroscience
researchers often insert the opsin transgene against an already- are also present in the ongoing optogenetics revolution. The
genetically engineered rodent background that activates the motivating problem for optogenetics stemmed from the causal-
expressed opsin transgene only in infected neurons that are also mechanistic explanatory goals of mainstream neurobiology.
highly activated by the target memory stimulus. Many types of To test hypotheses about the causal role played by specific
highly active neurons express c-fos, the protein product of an neurons to produce behaviors routinely taken as indicators
immediate early gene, which in turn can drive the production of particular cognitive functions in mammalian models,
of an engineered tetracycline-responsive element promoter experimenters need the capacity reliably to intervene into
protein, which turns on opsin transgene expression only in the hypothesized mechanisms in vivo—to activate or inhibit
infected neurons which are also the most highly active. At all them in the behaving organism, as directly and as efficiently
other times this background tetracycline-transactivator system is as possible. They can then measure the effects of these
kept silent by doxycycline in the animals’ diets. This trick enables experimental interventions on the behaviors of the organism,
experimenters later to turn on via the light stimulus all and only in tightly controlled experimental protocols routinely taken
those infected neurons which were most active during memory to be indicators of the target phenomenon. This task has
formation—the hypothesized cellular engram for the memory become more difficult in light of continued discoveries of the
stimulus. sometimes sparse anatomical distributions of the hypothesized
Designer receptors exclusively activated by designer drugs cellular mechanisms. Pharmacological interventions easily can
(DREADDs) technology is a more recent refinement which be confounded by the spread of the drug to nearby non-
dispenses with optogenetics’ inserted light stimulus. Instead of an target neurons; nature doesn’t always cooperate to clump target
engineered microbial opsin gene, the gene for an entire protein neurons together into discrete cortical columns or microcolumns
receptor for a designer drug—a compound, such as clozapine- to make stimulation by microelectrodes feasible. The intriguing
N-oxide (CNO), which nowhere occurs naturally in biological suggestion of using light stimuli to activate or inhibit target
tissue—is inserted via the virus vector. These engineered neurons was already on offer (Crick, 1999). The ‘‘obvious’’
receptors are metabotropic (coupled to a G-protein), to induce solution was to develop some way to induce the distributed
either bursting (Gq -type protein) or silencing (Gi -type protein) specific target neurons themselves to synthesize light-sensitive
in infected neurons. The designer drug is then administered ion channels. But that solution explodes quickly. For speed of
systemically, either through peritoneal injections or diet. It binds activation or inhibition the expressed ion channels should be
to the engineered receptors in the infected neurons, activates the ionotropic, which simply open direct channels for a specific
attached G-protein complex, and induces bursting or silencing cation or anion influx or efflux. Engineered gene expression
in those neurons for the extent of the pharmacological activity of and protein synthesis would have to be at levels both safe
the designer drug. and in sufficient numbers to have enduring effects on neuron
Neurobiologists are now furiously exploring the relative membrane potentials. The activating light stimulus had to be
advantages and disadvantages of optogenetics vs. DREADDs. deliverable to induced neurons safely, often deep in the brain,
The lack of need for an implanted light source is one yet with sufficient strength to induce sufficient activity in enough
technological advantage of DREADDs. So can be the number of the target neurons to affect circuit activity. And all of this
of neurons activated by the drug. Systemic administration of had to be implemented in vivo. But if successful, neurobiology
the designer drug means that all the neurons infected with would then have a great new experimental interventionist tool
and expressing the transgene for the designer receptor will be to explore causal-mechanistic hypotheses relating activity in
affected, no matter where they are located in the brain. In specific neurons to behaviors—including behaviors routinely
optogenetics, only those neurons infected, expressed, and in the taken to indicate the occurrence of particular cognitive functions.

This was optogenetics’ motivating problem. Like most hypothalamic neurons. Electrophysiological studies with tissue
motivating problems which generate revolutionary tool slices in vitro showed that ChR2-expressing Hcrt neurons
development in neuroscience, including the example of gene reliably responded to light stimuli with volleys of action
targeting techniques discussed above, this was a composite of potentials (Adamantidis et al., 2007, Figure 1). Optogenetic
many technical problems with a daunting interactive dynamic.10 photostimulation of ChR2-expressing lateral hypothalamus Hcrt
Deisseroth (2015), in whose lab optogenetics in its current form mice also activated infected neurons in vivo, as measured by
was first successfully developed, summarizes this situation nicely. increased c-fos expression specifically in these neurons. And
The three core features of the optogenetics technique include optogenetic photostimulation to lateral hypothalamus in vivo
the engineered microbial opsin genes and proteins; general in ChR2-Hcrt mice outfitted for chronic electroencephalogram
methods for targeting opsin expression to well-defined neurons; (EEG) and electromyogram (EMG) recordings showed that
and methods for guiding light stimuli to brain regions, often direct activation of these neurons significantly decreased latency
deep below the cortical surface, while the organism carries out to wakeful state, from both slow wave sleep (SWS) and rapid
the behaviors of interest. Each component problem individually eye movement (REM) sleep (condition 3), as compared to both
faced serious difficulties; but together they constituted ‘‘a baseline (non-photostimulated) latency in these same mice and
biological three body problem in which it was hard to resolve photostimulation in control mice (who received the virus vector
(or, even more importantly, to motivate attempts to resolve) injections minus the ChR2 transgene; Adamantidis et al., 2007,
any one of the three challenges without first addressing the Figure 3). Effects of photostimulation appeared limited to sleep-
other components’’ Deisseroth (2015, p. 1214). Ion currents wake transitions since duration of waking EEG and EMG events
through microbial rhodopsin proteins were predicted to be did not differ significantly across groups. These effects were
very small; hence high gene expression and light-intensity levels dependent upon frequency of the photostimulation, however,
seemed necessary. These levels of expression and synthesis had as 15 ms light pulses delivered at 1 Hz did not elicit them,
to be sustainable in neurons, biological cells well known to be while all frequencies greater than 5 Hz did. And the mechanism
detrimentally sensitive to membrane protein overexpression, of these photostimlation-induced effects influenced the sleep-
and to side effects of heat and light. All this plus the specificity wake circuitry by affecting Hcrt release in the infected neurons.
of expression had to be achieved while minimizing cell toxicity, Administration of a single dose of an Hcrt receptor (type 1)
in order to be applicable in vivo. Furthermore, attempts were antagonist (SB334867) blocked this photostimulation effect on
already underway to use engineered metazoan rhodopsin genes sleep-wake transition latency, from both SW and REM sleep
and channel proteins (which mediate photon transduction in (Adamantidis et al., 2007, Figure 4).
vertebrate retina), which reduced initial enthusiasm for trying to As Deisseroth himself notes, the Adamantidis et al. (2007)
solve these technical problems using a family of genes so foreign study was the first experiment to demonstrate that
to the mammalian genone.
Again, this combination of features of the motivating it was possible to selectively target a microbial opsin gene with
problem for a revolutionary tool development illustrates why high specificity and penetrance to a defined population of neurons
deep in the brain of adult mice . . . to play in a broad range of
the successful initial hook experiments were so surprising.
spike patterns throughout an optical fiber to these cells, to collect
Even scientists ‘‘in the know’’ about early attempts to develop simultaneous multimodal system readouts during freely moving
optogenetics in the form it took assigned a low probability behavior, . . . and to demonstrate a causal role for defined activity
to these experiments succeeding. To repeat from above, thee patterns in specific brain cells in natural behavior.
experiments are the first published experiments using the —(Deisseroth, 2015, p. 1216)
technique: (1) usually in a top journal in the field; (2) applying
it to the targeted experimental population of research interest; Subsequent applications of developments in gene targeting,
while (3) addressing a phenomenon in the target field of inquiry. and opsin genomics and engineering, quickly contributed to
For optogenetics, this was the Adamantidis et al. (2007) study the rapid increase in published optogenetics results over the
from the Deisseroth lab. Condition (1) was met: the article was subsequent few years.
published in Letters to Nature. As we saw in our first case study, the initial hook experiments
Adamantidis et al. (2007) engineered a lentivirus using which expose a new experimental tool to the purview of a
the promoter region of the mouse (condition 2) prepro- given field’s research specialists are important; but genuinely
hypocretin gene (Hcrt). This limited expression of the engineered revolutionary tools garner even wider appeal by way of second-
microbial opsin gene, channelrhodopsin2 (ChR2) to hypocretin- phase hook experiments. As it was for the case of gene targeting
(also known as orexin-) producing cells. They micro-injected techniques, identifying the key second-phase hook experiments
this virus vector into mouse lateral hypothalamus. Expression is more controversial, as there are a number of viable candidates.
persisted and was highly specific to Hcrt-producing lateral In the case of optogenetics, however, arguments for identifying
some specific second-phase hook experiments are aided because
10 Note
the technique has penetrated into the awareness of the general
that, intuitively, the motivating problems for each of the other two
instances of revolutionary tool development in contemporary neuroscience I
(scientifically literate) public. One set of recent optogenetic
mentioned above-high-output electrodes for single-cell recording in vivo and experiments in particular stands out: work on reactivating
functional brain imaging -also share this feature. One might be starting to the cellular engrams of specific memories, and pairing these
sense a pattern . . . reactivations with new memorial information to create false

memories in rodents. This work came from Tonegawa’s lab. kinds of experimental evidence—‘‘correlation’’ of the parallel
One behavioral protocol the lab employs is contextual fear occurrences of hypothesized mechanism and effect; ‘‘blockage’’
conditioning, a hippocampus-dependent (‘‘declarative’’) rodent of the hypothesized mechanism, which demonstrates its necessity
memory task. This is the same behavioral protocol used by Abel for the effect; and ‘‘mimicry’’ of the hypothesized mechanism,
et al. (1997) with their R-subunit PKA mutants, discussed in the which demonstrates its sufficiency12 —Ramirez et al. (2014,
section above. The rodent is placed in a novel context, a cage with p. 3) note that ‘‘mimicry experiments for memory engram
an internal environment it has never encountered previously, studies remained a considerable challenge’’. Experimenters
allowed to explore for a short duration (1–2 min), and then is ‘‘widely recognized and agreed’’ that mimicry experiments were
foot-shocked. Later exposures to that environment elicit freezing, essential toward testing the cellular engram hypothesis. But ‘‘the
the characteristic rodent fear response: complete suppression lack of tools that could precisely label and control selected
of all movement except breathing and a stereotypic crouched neurons involved in a particular memory posed formidable
posture. Time spent freezing indicates strength of memory for obstacles to carry out these experiments’’ (Ramirez et al.,
the context-shock association. 2014, p. 3). Gene targeting techniques using microinjections
Liu et al. (2012) used a transgenic mouse line in which of virus vectors for overexpressing genes like CREB and
an excitatory opsin (ChR2) was expressed only during fear α-CaMKII in specific neurons prior to training had some
conditioning to the novel context, in the most active neurons limited successes (e.g., Josselyn et al., 2001, discussed above).
in the dentate gyrus region of the hippocampus. Expression was Optogenetics, however, is the first tool which could reliably
limited by the tetracycline-controlled transcriptional activation ‘‘selectively label and activate the memory engram-bearing
system described above. ChR2 opens a direct channel for cells to induce the predicted behavioral changes caused by
sodium (Na+ ) influx when photostimulated, to depolarize learning,’’ and so reliably provide this final, essential kind of
infected neurons. Hence only the specific infected dentate gyrus evidence for direct cell activity → cognitive function causal
neurons most active during the context-shock pairing, which hypotheses. Experiments in the neurobiology of learning and
constitute the hypothesized cellular engram for the association, memory could now reliably provide mimicry evidence. And
were reactivated by the photostimulation. Liu et al. (2012) they did, quickly. Two years after Liu et al.’s (2012) first
ingenious experimental design was to reactivate the infected results were published a review article appeared in Trends in
cells encoding the context-shock association to the first context Neurosciences, appropriately titled ‘‘The optogenetic revolution
while the animal was exploring a second and different novel in memory research, ’’ which referenced 100 publications
context. No actual foot shock was administered in that second (Goshen, 2014).
context. However, photostimulated experimental mice later
demonstrated significant freezing to the second, non-shocked, CONCLUDING REMARKS: MOLECULAR
context. BIOLOGY VS. PHYSICS AS
A follow-up study by Ramirez et al. (2013) in Tonegawa’s PARADIGMATIC SCIENCE
lab also expressed the opsin only in the most active dentate
gyrus neurons, but this time during non-shocked exploration of Real revolutions in contemporary cellular/molecular/behavioral
the first novel context. They reactivated those specific neurons neuroscience have little to do with the components and dynamics
by photostimulation when the animals underwent context- of Thomas Kuhn’s famous model, and mostly to do with novel
fear (foot shock) conditioning in a second novel context. tool development. The two concepts developed and illustrated
The experimental mice demonstrated the fear response when here, motivating problem and hook experiments, are just first
placed back in the first, neutral context, in which they never steps toward a metascience of experimental tool development,
were shocked. These experimenters interpret their results the key components of such scientific revolutions. They in no
straightforwardly: ‘‘we created a false memory in mice by way exhaust all the interesting metascientific features of tool
optogenetically manipulating memory engram-bearing cells in development in neurobiology, much less in science generally.
the hippocampus’’ (Ramirez et al., 2013, p. 387).11 Both components need additional support from other case
The Tonegawa lab’s second phase optogenetics hook studies. But those two alone show ways in which revolutions
experiments not only brought opogenetics to wider scientific in neurobiology depart significantly from Kuhn’s model. It is
notice, but also attracted huge public interest: ranging from also not clear how many other contemporary sciences share
Smithsonian Magazine to The Guardian to Time Magazine this revolutionary dynamic. But it is illuminating to note that
to neurogadgit.com, to name just a few popular sources that cellular and molecular biology in general, and biochemistry,
covered it. Google ‘‘false memory optogenetics’’ to get a quick from which much of contemporary mainstream neuroscience
feel for the number and range of popular media coverage. Its stems, has now outdistanced physics as the most influential
scientific revolutionary credentials were perhaps best expressed science of our time, in terms of research funding, number of
in these scientists’ own words. Noting that a case for the
12 These labels are from neurobiologist Sweatt (2009). The attendant concepts
biological mechanism for a specific process involves three
are similar to Silva et al. (2014) notions of ‘‘non-intervention,’’ ‘‘negative
manipulation,’’ and ‘‘positive manipulation’’ experiments, which combine
11 Philosopher Robins (forthcoming) makes interesting use of these and to make up ‘‘Convergent Three integration,’’ although there are subtle
other results from the Tonegawa lab to challenge ‘‘constructionist’’ views of differences, especially concerning ‘‘mimicry’’ and ‘‘positive manipulation.’’
memory, now prevalent in psychology and cognitive neuroscience. That discussion is beyond the scope of this essay.

publications, and number of practicing scientists. Obviously, The recent growing influence of the ‘‘new mechanist’’ approach
numerous social forces drive a great deal of this current influence. in philosophy of science, built most centrally on biology rather
But in light of it, one might legitimately wonder whether the than physics, might be one important, albeit still implicit step in
still-lasting—and as I have argued here, mistaken—influence of this direction (Machamer et al., 2000; Craver, 2007; Craver and
Kuhn’s model of scientific revolution even on these fields is one Darden, 2014).
lingering relic of mid-20th century philosophy of science, when
physics was assumed to be the paradigmatic science, against AUTHOR CONTRIBUTIONS
which all others were compared and judged. Perhaps it is now
time for cellular and molecular biology to assume that role? JB is the sole author of this manuscript.
REFERENCES
Abel, T., Nguyen, P. V., Barad, M., Deuel, T. A. S., Kandel, E. R., and Liu, X., Ramirez, S., Pang, P. T., Punyear, C. B., Govindarajan, A., Deissereoth, K.,
Bourtchouladze, R. (1997). Genetic demonstration of a role for PKA in the late et al. (2012). Optogenetic stimulation of a hippocampal engram activates fear
phase of LTP and in hippocampus-based long-term memory. Cell 88, 615–626. memory recall. Nature 484, 381–385. doi: 10.1038/nature11028
doi: 10.1016/s0092-8674(00)81904-2 Lynch, G. (1986). Synapses, Circuits, and the Beginnings of Memory. Cambridge,
Adamantidis, A. R., Zhang, F., Aravanis, A. M., Deisseroth, K., and de Lecea, MA: MIT Press.
L. (2007). Neural substrates of awakening probed with optogeneticcontrol of Machamer, P., Darden, L., and Craver, C. F. (2000). Thinking about mechanisms.
hyocretin neurons. Nature 450, 420–424. doi: 10.1038/nature06310 Philos. Sci. 67, 1–25. doi: 10.1086/392759
Bickle, J. (2003). Philosophy and Neuroscience: A Ruthlessly Reductive Account. Masterman, M. (1970). ‘‘The nature of a paradigm,’’ in Criticism and the Growth
Dordrecht: Kluwer (now Springer). of Knowledge: Proceedings of the International Colloquium in the Philosophy of
Bickle, J. (2008). ‘‘Real reduction in real neuroscience: metascience, not philosophy Science, London, 1965 (Vol. 4), eds I. Lakatos and A. Mugrave, (Cambridge:
of science (and certainly not metaphysics!),’’ in Being Reduced, eds J. Hohwy Cambridge University Press, 59–90.
and J. Kallestrup (New York, NY: Oxford University Press), 34–51. Morris, R. G. M. (1989). Synaptic plasticity and learning: selective impairment
Bliss, T. V., and Lømo, T. (1973). Long-lasting potentiation of synaptic of learning in rats and blockade of long-term potentiation in vivo by the
transmission in the dentate area of the anesthetized rabbit following N-methyl-D-aspartate receptor antagonist AP5. J. Neurosci. 9,
stimulation of the performant path. J. Physiol. 232, 331–356. doi: 10. 3040–3057.
1113/jphysiol.1973.sp010273 Morris, R. G. M., and Kennedy, M. B. (1992). The pierrian spring. Curr. Biol. 2,
Bourtchuladze, R., Frenguelli, B., Blendy, J., Cioffi, D., Schütz, G., and Silva, A. J. 511–514. doi: 10.1016/0960-9822(92)90001-Q
(1994). Deficient long-term memory in mice with a targeted mutation of the Otchy, T. M., Wolff, S. B. E., Rhee, J. Y., Pehlevan, C., Kawai, R., Kempf, A., et al.
cAMP-responsive element-binding protein. Cell 79, 59–68. doi: 10.1016/0092- (2015). Acute off-target effects of neural circuit manipulations. Nature 528,
8674(94)90400-6 358–363. doi: 10.1038/nature16442
Craver, C. F. (2007). Explaining the Brain. New York, NY: Oxford University Press. Ramirez, S., Liu, X., Lin, P.-A., Suh, J., Pignatelli, M., Redondo, R. L., et al. (2013).
Craver, C. F., and Darden, L. (2014). In Search of Mechanisms. Chicago, IL: Creating a flase memory in the hippocampus. Science 341, 387–391. doi: 10.
University of Chicago Press. 1126/science.1239073
Crick, F. (1999). The impact of molecular biology on neuroscience. Philos. Trans. Ramirez, S., Tonegawa, S., and Liu, X. (2014). Identification and optogenetic
R. Soc. Lond. B Biol. Sci. 354, 2021–2025. doi: 10.1098/rstb.1999.0541 manipulation of memory engrams in the hippocampus. Front. Behav. Neurosci.
Deisseroth, K. (2015). Optogenetics: 10 years of microbial opsins in neuroscience. 7:226. doi: 10.3389/fnbeh.2013.00226
Nat. Neurosci. 18, 1213–1225. doi: 10.1038/nn.4091 Robins, S. (forthcoming). Optogenetics and the mechanisms of false memory.
Dudai, Y. (2002). Memory from A to Z. New York, NY: Oxford University Press. Synthese 1–23. https://sarahkrobins.files.wordpress.com/2014/06/robins-
Goshen, I. (2014). The optogenetic revolution in memory research. Trends optogenetics.pdf doi: 10.1007/s11229-016-1045-9
Neurosci. 37, 511–522. doi: 10.1016/j.tins.2014.06.002 Silva, A. J., Landreth, A., and Bickle, J. (2014). Engineering the Net Revolution in
Grant, S. G. N., O’Dell, T. J., Karl, K. A., Stein, P. L., Soriano, P., and Kandel, E. R. Neuroscience. New York, NY: Oxford University Press.
(1992). Impaired long-term potentiation, spatial learning and hippocampal Silva, A. J., Paylor, R., Wehner, J. M., and Tonegawa, S. (1992a). Impaired spatial
development in fyn mutant mice. Science 258, 1903–1910. doi: 10.1126/science. learning in α-calcium-calmodulin kinase II mutant mice. Science 257, 206–211.
1361685 doi: 10.1126/science.1321493
Hebb, D. (1949). The Organization of Behavior. New York, NY: John Wiley & Sons. Silva, A. J., Stevens, C. F., Tonegawa, S., and Wang, Y. (1992b). Deficient
Josselyn, S. A., Shi, C., Carlezon, W. A., Neve, R. L., Nestler, E. J., and Davis, M. hippocampal long-term potentiation in α-calcium-calmodulin kinase II mutant
(2001). Long-term memory is facilitated by cAMP response element-binding mice. Science 257, 201–206. doi: 10.1126/science.1378648
protein overexpression in the amygdala. J. Neurosci. 21, 2404–2412. Sweatt, J. D. (2009). Mechanisms of Memory. 2nd Edn. Burlington, MA: Academic
Kida, S., Jossely, S. A., Peña de Ortiz, S., Kogen, J. H., Chevere, I., Masushige, Press.
S., et al. (2002). CREB required for the stability of new and reactivated fear Thomas, K. R., and Capecchi, M. R. (1987). Self-directed mutagenesis by gene
memories. Nat. Neurosci. 5, 348–355. doi: 10.1038/nn819 targeting in mouse embryo-derived stem cells. Cell 51, 503–512. doi: 10.
Kuhn, T. S. (1962). The Structure of Scientific Revolutions. Chicago, IL: University 1016/0092-8674(87)90646-5
of Chicago Press.
Landreth, A., and Silva, A. J. (2013). The need for research maps to navigate Conflict of Interest Statement: The author declares that the research was
published works and inform experiment planning. Neuron 79, 411–415. doi: 10. conducted in the absence of any commercial or financial relationships that could
1016/j.neuron.2013.07.024 be construed as a potential conflict of interest.
Lisman, J. E. (1985). A mechanism for memory storage insensitive to molecular
turnover: a bistable autophosphorylation kinase. Proc. Natl. Acad. Sci. U S A Copyright © 2016 Bickle. This is an open-access article distributed under the terms
82, 3055–3057. doi: 10.1073/pnas.82.9.3055 of the Creative Commons Attribution License (CC BY). The use, distribution and
Lisman, J. E., and Goldring, M. A. (1988). Feasibility of long-term storage of graded reproduction in other forums is permitted, provided the original author(s) or licensor
information by the Ca2+ /calmodulin-dependent protein kinase molecules of are credited and that the original publication in this journal is cited, in accordance
the postsynaptic density. Proc. Natl. Acad. Sci. U S A 85, 5320–5324. doi: 10. with accepted academic practice. No use, distribution or reproduction is permitted
1073/pnas.85.14.5320 which does not comply with these terms.

Revolutions in Neuroscience: Tool Development: John Bickle

Uploaded by

Copyright:

Available Formats

Revolutions in Neuroscience: Tool Development: John Bickle

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Revolutions in Neuroscience: Tool Development: John Bickle

Uploaded by

Copyright:

Available Formats

HYPOTHESIS AND THEORY

published: 08 March 2016

Frontiers in Systems Neuroscience | www.frontiersin.org 1 March 2016 | Volume 10 | Article 24

Frontiers in Systems Neuroscience | www.frontiersin.org 2 March 2016 | Volume 10 | Article 24

Frontiers in Systems Neuroscience | www.frontiersin.org 3 March 2016 | Volume 10 | Article 24

Frontiers in Systems Neuroscience | www.frontiersin.org 4 March 2016 | Volume 10 | Article 24

Frontiers in Systems Neuroscience | www.frontiersin.org 5 March 2016 | Volume 10 | Article 24

Frontiers in Systems Neuroscience | www.frontiersin.org 6 March 2016 | Volume 10 | Article 24

Frontiers in Systems Neuroscience | www.frontiersin.org 7 March 2016 | Volume 10 | Article 24

Frontiers in Systems Neuroscience | www.frontiersin.org 8 March 2016 | Volume 10 | Article 24

Frontiers in Systems Neuroscience | www.frontiersin.org 9 March 2016 | Volume 10 | Article 24

Frontiers in Systems Neuroscience | www.frontiersin.org 10 March 2016 | Volume 10 | Article 24

Frontiers in Systems Neuroscience | www.frontiersin.org 11 March 2016 | Volume 10 | Article 24

Frontiers in Systems Neuroscience | www.frontiersin.org 12 March 2016 | Volume 10 | Article 24

Frontiers in Systems Neuroscience | www.frontiersin.org 13 March 2016 | Volume 10 | Article 24

You might also like