Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Exploring sensory responses in the different subdivisions of the visual thalamus
(USC Thesis Other)
Exploring sensory responses in the different subdivisions of the visual thalamus
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
EXPLORING SENSORY RESPONSES IN THE DIFFERENT
SUBDIVISIONS OF THE VISUAL THALAMUS
by
Ulas M. Ciftcioglu
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPY
(BIOLOGY (NEUROBIOLOGY))
August 2019
Copyright 2019 Ulas Mustafa Ciftcioglu
ii
To my family
iii
ACKNOWLEDGEMENTS
I would like to start with thanking my advisor Dr. Judith Hirsch for her guidance and
support throughout my Ph.D. studies. Her experience, knowledge, enthusiasm and presence has
done invaluable contributions to me as a professional and beyond. None of this work would be
possible without her mentorship. I am also very grateful to my co‐advisor Dr. Fritz Sommer for
his guidance and contribution over the years. His presence has provided extra dimensions to my
development and research. I would also like to thank my committee members Dr. Bartlett Mel,
Dr. Andrew Hires, Dr. Bosco Tjan and Dr. Hong Wei Dong for their valuable insight and feedback
throughout my studies. I would like to express my gratitude to my previous advisor Dr. Didem
Gökçay for her contributions and encouragement for me to pursue a PhD in neuroscience.
I would like to express my deep appreciation for working and interacting with amazing
friends and colleagues in the lab; Dr. Vandana Suresh, Alexis Gorin and Seohee Ann. A special
thanks to Dr. Vandana Suresh for all her mentorship. I am very grateful to the previous members
of our lab, Dr. Michael Halassa, Dr. Ralf Wimmer, other members of the Halassa lab and members
of the Hires lab for their contributions. I would also like to thank my friends at the vision journal
club and USC for their companionship throughout my PhD. Finally, I would like to send my love
and appreciation to my wife Merve, my parents and my brother Ertugrul for their continuous
love and support.
iv
Table of Contents
Dedication ii
Acknowledgements iii
List of Figures v
Abstract vii
Chapter 1. Introduction 1
Chapter 2. Visual information processing in the sensorimotor division of the lateral 11
geniculate nucleus of the thalamus
2.1. Introduction 11
2.2 Materials and Methods 14
2.3 Results 24
2.4 Discussion 45
Chapter 3. Spatial and temporal processing of visual signals in the mouse thalamic 51
reticular nucleus
3.1. Introduction 51
3.2 Materials and Methods 54
3.3 Results 60
3.4 Discussion 75
Chapter 4. Conclusion 82
References 85
v
List of Figures
Figure 1.1. Major afferent and efferent projections of the LGN and TRN. 2
Figure 2.1. Retinal ganglion cell types projecting to different subdivisions of the LGN. 13
Figure 2.2. Quantification of spatial receptive fields via membrane currents. 17
Figure 2.3. Receptive fields in the dLGN are usually smaller than those in the vLGN. 26
Figure 2.4. Morphological differences between neurons in dLGN vs vLGN. 28
Figure 2.5. The shapes of membrane currents recorded from the dLGN and vLGN. 31
Figure 2.6. Afferent and efferent projections of the vLGN. 32
Figure 2.7. Receptive field maps and sizes in the vLGN and SCs. 33
Figure 2.8. Temporal precision and information content of spike trains evoked by natural 36
scene movies for the vLGN, dLGN and SC.
Figure 2.9. Visual oscillations in the superior colliculus from earlier studies. 39
Figure 2.10. Visually evoked oscillatory spike trains in the vLGN and SC. 40
Figure 2.11. Flash‐evoked gamma band oscillatory currents for a sample ON cell in the 42
vLGN.
Figure 2.12. Changes in the strength of membrane oscillations during the presentation 44
of natural scene movies in the vLGN.
Figure 2.13. Modeling of oscillatory activity based on visual stimulus using ridge 45
regression during natural scene movies in the vLGN.
Figure 3.1. The connections and hypothesized sensory roles of the TRN. 53
Figure 3.2. Targeting and identification of visual reticular neurons in vivo. 56
vi
Figure 3.3. Receptive field maps and sizes in the cat LGN and TRN. 63
Figure 3.4. Receptive field maps in the mouse LGN and TRN. 64
Figure 3.5. Receptive field sizes in the mouse LGN and TRN. 65
Figure 3.6. Receptive field sizes across visual space in the mouse TRN. 67
Figure 3.7. Membrane currents and associated receptive field in the TRN. 68
Figure 3.8. Dual modes of firing in the thalamus and associated temporal receptive fields. 70
Figure 3.9. Receptive fields associated with burst and tonic firing in the mouse LGN and 71
TRN.
Figure 3.10. Temporal receptive fields in the mouse LGN and TRN and their analysis. 73
Figure 3.11. Population statistics of temporal receptive fields associated with bursts 74
and tonic spikes in the mouse LGN and TRN.
Figure 3.12. Influence of TRN firing mode on the thalamocortical pathway. 80
vii
ABSTRACT
The visual role of thalamus is more than a passive relay of visual information to the
primary visual cortex. Indeed, thalamus has various visual structures with different roles. To
begin with, the lateral geniculate nucleus (LGN) is composed of substructures which have
different functions. Its well studied dorsal division (dLGN) relays visual information from the
retina to the primary visual cortex. On the other hand, the ventral division (vLGN) and associated
intergeniculate leaflet (IGL) do not project to cortex. These two structures (vLGN/IGL) connect
with many subcortical structures such as superior colliculus (SC) and pretectum, hence they likely
have both sensory and motor roles, also referred as non‐image forming visual functions. In
addition to the LGN, the visual thalamus includes a small portion of the thalamic reticular nucleus
(TRN), a network of GABAergic cells. The TRN is usually studied in the context of sleep and
attention, but our knowledge of its contributions to visual function is sparse. How do these
different structures in the thalamus process visual information? What functions do they serve in
the visual system?
To address these questions, I first explored on the visual processing in the vLGN/IGL via
electrophysiological recordings from individual neurons in mice and how that compares to the
dLGN and SC. The size of receptive fields in the vLGN was larger than the ones in the dLGN. The
dendritic arbors in the vLGN was commensurately larger and the size of individual post‐synaptic
currents were smaller. When compared to SC, the temporal precision observed in the vLGN was
lower, along with coarser spatial receptive fields. The vLGN also exhibited visual gamma band
viii
oscillations, similar to SC. The strength of these oscillations also changed during the presentation
of naturalistic stimulation, suggesting a role for oscillations during natural vision.
Then, I moved on to investigate the visual receptive field structure in the TRN in mice. The
receptive fields in TRN were mostly ON – OFF, though one contrast was generally dominant. In
terms of spatial scale, around half of the cells in TRN had receptive fields as small as the ones in
dLGN, providing a means for localized reticular inhibition on LGN. At the same time, the temporal
profile of the receptive fields revealed that bursts encode different information than tonic spikes
in both TRN and dLGN. Bursts were preceded with a stronger phase of non‐preferred contrast
leading the stimuli of preferred contrast, than tonic spikes. This suggests that visual stimulus can
change the mode of firing and bursts likely have a distinct role during vision. When compared to
previous studies, our results highlight conserved aspects of visual processing in TRN in mouse, a
leading model system to study attention and neural circuits.
Overall, this work provides a fuller understanding of the role of various visual thalamic
nuclei and how they process visual information.
1
CHAPTER 1
INTRODUCTION
The first station in the visual system is the retina, which distributes the visual information
to various structures. Mainly, visual information follows the two streams. The first one goes
through the thalamus before reaching to the visual cortex. The second stream follows through
subcortical structures such as the superior colliculus before reaching various downstream motor
structures. Within thalamus, lateral geniculate nucleus (LGN) is the station which transmits visual
information to the primary visual cortex. The receptive fields of the thalamocortical cells (also
known as “relay cells”) in the LGN display center surround structure similar to retinal ganglion
cells (Hubel, 1960; Cleland et al., 1971). Hence, this created the impression that there is no
obvious transformation of visual information at the level of LGN. However, insights from many
studies challenged this view by suggesting that neural information is transformed in the LGN (see
(Wang et al., 2011; Hirsch et al., 2015) for reviews).
In fact, the role of transmitting the visual information to the cortex only reflects the dorsal
division of the LGN (dLGN). Contrary to the well known dLGN, the ventral division (vLGN) does
not project to the cortex (Figure 1.1). Indeed, it projects to several subcortical structures with
motor functions, including the superior colliculus (SC). In addition to the difference in the
connectivity, the composition of cell types in both divisions of the LGN also differs. In dLGN, both
excitatory and inhibitory cells are present. Excitatory neurons correspond the relay cells that
project to the cortex whereas inhibitory neurons are local and provides feedforward inhibition
2
to relay cells. However in the vLGN, all neurons seem to be inhibitory including the projection
neurons (Ohara et al., 1983; Gabbott and Bacon, 1994; Inamura et al., 2011). The difference in
the cell types can also be observed during development through gene expression profiles
(Golding et al., 2014).
Figure 1.1. Major afferent and efferent projections of the LGN and TRN. LGN is composed of three
subdivisions, the dLGN, vLGN and IGL, which all receive retinal input. dLGN projects to the primary visual cortex,
along the pathway for form vision (figure (top right) taken from (de Haan and Cowey, 2011)). However vLGN
(along with IGL) doesn’t project to cortex and is mainly connected with subcortical structures. VLGN is generally
associated with non‐image forming visual functions such as eye movements (figure taken from (Sparks, 2002))
and pupillary reflex (figure taken from (Schmidt et al., 2011)). TRN is reciprocally connected with LGN and also
receives cortical input. dLGN, dorsal lateral geniculate nucleus; vLGN, ventral lateral geniculate nucleus; IGL,
intergeniculate leaflet; TRN, thalamic reticular nucleus, V1, primary visual.
The visual thalamus not only includes the LGN but also a portion of the Thalamic Reticular
Nucleus which reciprocally connects with LGN. TRN includes inhibitory neurons and generally
studied in the context of attention and sleep (see (Pinault, 2004) for a review). The sensory
3
physiology of the TRN has critical influence for the geniculocortical pathway given that reticular
neurons provide feedback inhibition onto relay cells. Recent studies suggest that reticular
neurons display high level of visual selectivity (Vaingankar et al., 2012; Soto‐Sánchez et al., 2017).
Connectivity and associated roles for the sensorimotor division of the LGN (vLGN)
The connectivity profile of the vLGN and the intergeniculate leaflet (IGL), a thin layer of
tissue separating two main divisions of LGN, seems very different than the dLGN (Figure 1.1). The
dLGN projects to the primary visual cortex and this pathway is associated with form vision,
carrying high acuity visual information to process and interpret the visual scene. However, the
vLGN and IGL (referred collectively as vLGN in this dissertation due to their similarity of
connectivity and function) is connected with many subcortical structures. vLGN connects
reciprocally with the SC, pretectum, accessory optic system and contralateral vLGN. In addition,
it projects to suprachiasmatic nucleus, subthalamus and the dLGN. It also receives afferent
projections from the retina, higher order visual cortex, cerebellum and vestibular nuclei (Figure
2.6). All these connections plot a complicated picture for understanding the function of vLGN.
The fact that vGLN is connected to many structures suggests that vLGN likely has important role.
Most of the targets and inputs of vLGN are structures with motor roles.
Fortunately, recent studies that selectively characterized retinal ganglion cells types
provide great insights on the nature of visual information fed to the vLGN (Hattar et al., 2006;
Huberman et al., 2008; Kim et al., 2008; Ecker et al., 2010; Chen et al., 2011; Kay et al., 2011;
Rivlin‐Etzion et al., 2011) (Figure 2.1). The vLGN received dense innervation from intrinsically
photosensitive retinal ganglion cells (ipRGCs), which express melanopsin. This class of ganglion
cells were associated with a broad range of visual functions, collectively referred to as non‐image
4
forming visual functions. These functions include circadian rhythms and pupillary light reflex. The
connectivity of vLGN with motor structures and its retinal inputs suggest that vLGN is involved in
non‐image forming visual functions associated with ipRGCs.
In addition to its connectivity, some lesion studies provided into the functions of vLGN.
vLGN seemed to be involved in intensity discrimination, pupillary reflex and circadian rhythms
(see (Harrington, 1997) for a review). However the body of work in these lesion studies haven’t
explored several important visual motor behaviors. In addition, experimental challenges and
complications associated with lesions left a question mark on whether vLGN or other structures
was critical for impaired behaviors.
Exploring sensory processing using intracellular electrophysiology in vivo
The neural information transmitted by vLGN likely has critical consequences for various
visual functions given that vLGN projects to several important motor structures. Earlier attempts
to investigate the sensory physiology in vLGN were mainly limited to subjective assessments of
visual selectivity (see (Harrington, 1997) for a review). These studies left the impression that the
receptive fields were diffuse and most cells preferred bright stimulus. In order to advance our
understanding how visual information is processed in the vLGN, quantitative approaches to
analyze visually driven neural activity is critical.
The sensory physiology generally have some morphological correlates in the nervous
system. The morphological properties of neurons sometimes can provide great insights on
physiology, as in the case of retinal ganglion cells (Rockhill et al., 2002; Sanes and Masland, 2015).
Intracellular recordings in vivo provides a means to track these morphological correlates.
However studies exploring the sensory physiology of the vLGN in vivo were limited with
5
extracellular recordings. Intracellular recordings allows to fill dyes into the cells recorded, hence
one can link physiology with anatomy. In addition to that, one can record not only the action
potentials but also the membrane currents of a neuron. This provides a window into synaptic
physiology and can provide valuable insights on how cells process their inputs.
Oscillations driven by visual stimulus
Cells can respond to sensory stimulus not only by changes in firing rate but also with
oscillatory patterns. Oscillatory neural activity has been detected in the early stages of different
sensory modalities across various model systems (Koepsell et al., 2010). In the early visual
system, the visual oscillations are reported for retina (Munk and Neuenschwander, 2000;
Koepsell et al., 2009), dLGN (Munk and Neuenschwander, 2000; Koepsell et al., 2009; Saleem et
al., 2017), cortex (Saleem et al., 2017) and the SC (Brecht et al., 2001; Sridharan et al., 2011; Stitt
et al., 2013). Some of the proposed roles for visual oscillations are serving as a second information
channel, encoding spatially extensive features, encoding of temporal information induced from
eye movements and directing attention to specific features of the visual scene (Munk and
Neuenschwander, 2000; Koepsell et al., 2010). As previously stated, the vLGN connects
reciprocally with SC and also these two structures connect with common set of structures. Then
can the cells in the vLGN respond to visual oscillations by oscillations? If yes, what can be the
roles of these oscillations? Can oscillations have a function during natural vision?
Sensory physiology in the visual TRN
The visual sector of TRN receives excitatory input from both the LGN and the visual cortex.
Morphological studies have explored the properties of synaptic boutons from the LGN and cortex
onto reticular neurons. The synapses from thalamocortical cells are located on the proximal
6
dendrites and somas of the reticular neurons (Montero and Singer, 1984; Cucchiaro et al., 1991;
Bickford et al., 2008). However cortical projections innervate distal dendrites (Ide, 1982;
Cucchiaro et al., 1991; Bickford et al., 2008). This pattern suggests that thalamic inputs serve as
the driver input whereas cortical inputs modulate, as suggested in (Sherman and Guillery, 2001).
This scenario is in line with a study which explored the effects of cortex removal on the visual
response properties in the TRN (Xue et al., 1988). Orientation and spatial frequency selectivity,
along with ocular dominance in TRN were similar following the removal of cortex. In addition to
the aforementioned excitatory inputs, reticular neurons are interconnected via chemical
(Montero and Singer, 1984) and electrical synapses (Landisman et al., 2002). However the
presence of gabaergic connections between reticular neurons was challenged in a recent study
for mouse (Hou et al., 2016). The reported type of terminals between reticular neurons were
axonal F1 terminals (Ide, 1982; Pinault et al., 1997) and dendritic F2 appendages (Montero and
Singer, 1984). In terms of convergence, a study using rats (Burke and Sefton, 1966) suggests that
reticular cells are innervated by more relay cells than they project. Given the presence of various
inputs and factors, it seems challenging to dissect the contribution of each input by just recording
neural activity during visual stimulation.
The inhibition from TRN to LGN has topographical organization (Sanderson, 1971; Uhlrich
et al., 1991; Fitzgibbon, 2002), similar to the circuits of the auditory (Kimura et al., 2007) and
somatosensory (Pinault, 2004; Lam and Sherman, 2007) sectors of the TRN and associated
sensory thalamic nuclei. Earlier studies on the visual selectivity in TRN stated the presence of
binocular and ON‐OFF responses ((Sanderson, 1971; Dubin and Cleland, 1977; Ahlsen and
Lindstrom, 1982)), along with the impression of diffuse receptive fields (Sanderson et al., 1969;
7
Uhlrich et al., 1991). However, recent studies using state of the art analysis techniques reported
that receptive fields in the TRN were complex and diverse and can be spatially localized
(Vaingankar et al., 2012; Soto‐Sánchez et al., 2017).
Dual modes of firing in the TRN and LGN
Cells in the TRN and LGN both fire in two modes; bursts and tonic spikes (Jahnsen and
Llinas, 1984; Domich et al., 1986; Steriade et al., 1986; Huguenard and Prince, 1992; Contreras et
al., 1993; Destexhe et al., 1996). The firing mode of the reticular cells and the thalamocortical
cells are influenced by the membrane potential. At more depolarized membrane levels, tonic
spikes are fired whereas bursts are generated for more hyperpolarized membrane levels. The
bursts are initiated by the activation of T‐type Ca
+2
channels and contain several closely spaced
action potentials (Jahnsen and Llinas, 1984). The temporal structure of the bursts display
differences between LGN and TRN. In LGN, the interspike interval between adjacent spikes
increase towards the end of burst, however the interspike intervals first decrease and then
increase through the bursts in the TRN (Huguenard and Prince, 1992; Contreras et al., 1993). In
addition, the number of action potentials in a burst differs in the LGN and TRN. The bursts in the
TRN contain typically more action potentials than the ones in the LGN (Vaingankar et al., 2012).
Several studies have reported that visual stimuli can influence the mode of firing in the LGN
(Alitto et al., 2005; Wang et al., 2007) and TRN (Vaingankar et al., 2010). The spatial structure of
the receptive fields reconstructed from bursts and tonic spikes were similar. However, the
temporal structure of the receptive fields varied between the two modes of firing. Both type of
receptive fields generally included two temporal phases of opposite contrast. The phase of the
preferred (or dominant) contrast was closer to the spike and this phase was preceded by the
8
other phase of nonpreferred (or nondominant) contrast. For temporal receptive fields of bursts,
the earlier phase in time was stronger than the one observed in tonic fields. This was consistent
with the biophysical mechanism underlying burst generation because nonpreferred (or
nondominant) contrast can induce the inhibition necessary to initiate slow Ca
+2
currents
associated with bursts.
Role of TRN in visual processing
Given this position and connectivity pattern, TRN has the potential to significantly
influence visual information flowing to the cortex. Most of the related earlier studies focused on
how the inhibition TRN influences LGN activity in visual space. Two competing hypotheses have
been proposed to frame the spatial scale of the influence from the TRN to the LGN (Figure 3.1)
(Crick, 1984; Soto‐Sánchez et al., 2017). The thermostat hypothesis suggests that reticular cells
operate on a global level, hence feedback from TRN should serve as a gain control mechanism
onto the LGN. The alternate searchlight hypothesis is based on the idea that TRN provides top‐
down attentional influence through local modulation of LGN activity. A recent study quantifying
the receptive field scales in the TRN and LGN provided evidence favoring the searchlight
hypothesis (Soto‐Sánchez et al., 2017).
In addition to the visual selectivity governed by firing rate, a recent study focused on how
the modes of firing in TRN may influence visual information transmitted by thalamocortical cells
(Vaingankar et al., 2010). Since inhibition is a critical component in dictating the firing mode, the
inhibition and its strength from TRN can influence the mode of firing in thalamocortical cells of
the LGN. Given that visual stimuli can change the mode of firing in the TRN, TRN can influence
the firing mode in the LGN during visual stimulation. To illustrate, a burst of a reticular neuron
9
would induce stronger hyperpolarization than a tonic spike on a thalamocortical cell due to
temporal summation of postsynaptic currents. The hyperpolarized state of a thalamocortical cell
would increase the likelihood of a burst generation in the LGN. And eventually, when a burst from
a thalamocortical cell is fed to the cortex, it is more likely to trigger a spike because of temporal
summation of multiple excitatory postsynaptic currents.
Using mouse as a model system to investigate the visual processing in TRN
Visual response properties in TRN are studied in model systems such as the cat.
Unfortunately, the techniques for dissecting neural circuits are limited in these model systems.
However, mouse offers many advantages such as genetic tools, ease, cost effectiveness and rising
popularity to study the visual system. Along with increasing visual behavior assays using mice,
the behavioral outcomes of manipulating neural circuits can be determined. If aspects of sensory
physiology in TRN is conversed across species, mouse would be a favorable model system to
explore the visual processing in TRN and its influence on the neural information flowing through
the thalamocortical pathway. Mouse has already became a very popular model system to explore
functions of TRN such sleep and attention (Halassa et al., 2011; Wimmer et al., 2015). For
instance, a recent study stated that the activity of visual TRN neurons can reflect whether the
animal is attending to a particular sensory modality (Wimmer et al., 2015). When visual stimulus
is attended, the firing rate of visual reticular neurons decreased at the population level, thereby
increasing the activity of LGN. However, when the auditory stimulus is attended, the firing rate
of visual reticular neurons increased, hence decreasing the overall level of activity in the LGN.
The visual stimulus used was a wide field flash. Studies like the work presented in this dissertation
10
can provide insights on whether TRN can influence LGN activity as a function visual space,
providing a means for spatial attention.
11
CHAPTER 2
VISUAL INFORMATION PROCESSING IN THE SENSORIMOTOR DIVISION OF THE LATERAL
GENICULATE NUCLEUS OF THE THALAMUS
2.1 Introduction
There are two main routes the visual information travels from the eye to the brain in
mammals, one that involves the thalamus and the other that targets the tectum. The retino‐
thalamic pathway targets the lateral geniculate nucleus (LGN) and is usually discussed in the
context of form vision. However, only the dorsal part of the LGN (dLGN) projects to the neocortex,
where images in the visual scene are deciphered. The remaining, ventral, division (vLGN)
projects subcortically (Monavarfeshani et al., 2017). It is associated with structures linked to the
retinotectal pathway, whose function involves diverse aspects of sensorimotor processing, other
motor structures (Harrington, 1997; Kashef et al., 2014) and pathways associated with behavioral
responses such as mood (Huang et al., 2019) and fear (Salay et al., 2018). Thus, the vLGN is often
described as part of the “nonimage forming” pathway (Harrington, 1997; Monavarfeshani et al.,
2017).
The differences between the two divisions of the LGN are vast. The cellular composition
in the dLGN is like that of other primary thalamic nuclei, comprising a majority of excitatory relay
cells that project to cortex and a minority of local inhibitory interneurons with short‐range axons
(Bickford, 2018). By contrast the vLGN is mainly composed of interconnected gabaergic
projection cells that have a separate embryonic origin from cells in the dLGN (Golding et al., 2014;
12
Jager et al., 2016). There are also substantial differences in the cohort of ganglion cells that
innervate each division (Monavarfeshani et al., 2017) (Figure 2.1) and in the size and strength of
retinal synapses ((Hammer et al., 2014)). Last, studies of visual responses, though qualitative,
describe large receptive fields and a preference for bright rather than dark stimuli(Spear et al.,
1977; Harrington, 1997), profiles very different from the specific stimulus selectivity that
characterizes cells in the dLGN (Grubb and Thompson, 2004; Piscopo et al., 2013; Suresh et al.,
2016).
In highly visual animals like cat (Spear et al., 1977) and monkey (Babb, 1980), the size of
the dLGN often dwarfs that of the vLGN (called the pregeniculate nucleus in primate). By
contrast, the two divisions occupy similar territory in rodent (Harrington, 1997), suggesting
comparable importance. Thus, we chose to explore visual processing in the mouse vLGN. We
studied the nucleus from two perspectives, by comparing the vLGN to the principle nucleus of
the form vision pathway (dLGN) and that of the retinotectal pathway (superior colliculus (SC)).
Our main experimental approach included whole‐cell recording and anatomical labeling in vivo
combined with various computational tools. Consistent with past work (Spear et al., 1977;
Harrington, 1997), we found that receptive fields in the vLGN were more than twice as large, on
average, as those in dLGN. The disparity in receptive field size was mirrored, and perhaps
explained, by a dramatically larger breadth of dendritic arbors of cells in the ventral vs dorsal
divisions. Further, we explored temporal coding of the stimulus using artificial and natural
stimuli. Analogous to the larger receptive fields in the vLGN, spike timing with respect to the
stimulus was usually less precise in vLGN than in the SC and the dLGN. Our results also showed
that roughly a third of cell in vLGN encode information by spike timing relative to network
13
oscillations. That is, oscillation phase was entrained by stimulus onset and oscillation strength
waxed and waned during discrete segments of natural movies. Companion recordings from the
SC and dLGN showed that the characteristics of oscillations in the vLGN were closer to those in
SC (Brecht et al., 2001; Sridharan et al., 2011) than in dLGN (Koepsell et al., 2009; Storchi et al.,
2017). Thus, although the vLGN provides a spatiotemporally coarse version of visual images, it is
able to engage with the intrinsic rhythms of downstream targets. Taken together our results
describe the quality of information that the vLGN transmits to structures that influence
visuomotor and other behaviors.
Figure 2.1. Retinal ganglion cell types projecting to different subdivisions of the LGN. dLGN and vLGN receive
input from both common and separate classes of retinal ganglion cells (see (Kim et al., 2008; Huberman et al.,
2008; Kay et al., 2011; Rivlin‐Etzion et al., 2011; Hattar et al., 2006; Ecker et al., 2010; Chen et al.,2011)). dLGN,
dorsal lateral geniculate nucleus; vLGN, ventral lateral geniculate nucleus; RGC, retinal ganglion cell; DS,
direction selective; J, junctional adhesion molecule B; DRD4, dopamine receptor D4; TRHR, thyrotropin‐
releasing hormone receptor; BD, bistratified dendrite; Brn3b, Brn3b transcription factor.
14
2.2 Materials and Methods
Experimental Protocols
Preparation
Adult, pigmented mice (C57BL/6) were sedated with chlorprothixene (5 mg/kg, IM) after
which anesthesia was begun and maintained with urethane (0.5–1 g/kg, 10% w/v in saline, IM
(Niell and Stryker, 2008). After the head was cleaned and shaved, an incision was made to expose
the skull so that a metal headpost could be affixed to hold the animals in place. Next, a small
craniotomy centered around the LGN or SC was made. All wound margins were infiltrated with
lidocaine, the brain and eyes were kept moist with saline, and body temperature was maintained
at 37°C. All procedures were approved by the Institutional Animal Care and Use Committees of
the University of Southern California following guidelines from the National Institutes of Health.
Histology and verification of recording site
Brains were perfused with 3% paraformaldehyde, cut in coronal sections, 100µm thick,
and processed to visualize labeled cells using standard procedures (Hirsch et al., 1998). 3D
reconstructions of single cells were made using Neurolucida software (MBF Bioscience). In the
many instances for which we did not recover labeled cells, we estimated electrode position from
stereotaxic coordinates and depth measurements (Kopf, Model 2660) derived from positions of
cells labelled during this study and (Suresh et al., 2016). These estimates suggest that we have
sampled most of the vLGN. Our sample may, however, have been biased towards the external
zone of the vLGN. The external zone has larger cells than those in the medial zone and is far
wider as well. Further, the external zone receives dense retinal input whereas the medial zone
15
receives little or none (Monavarfeshani et al., 2017). Last, all cells we filled in the vLGN were
within the external zone.
Recordings
Whole‐cell and cell‐attached recordings were made using biocytin‐filled pipettes as
described previously (Wang et al., 2007); pipette resistances were 5‐20 MΩ. For whole‐cell
(intracellular) recordings, the membrane was held slightly below the firing threshold (in the
absence of visual stimulation) to resolve subthreshold excitatory and inhibitory membrane
currents. Neural signals were recorded using a Multiclamp 700B amplifier (Axon Instruments);
digitized and stored at 10‐20 kHz using a Power 1401 data acquisition system (Cambridge
Electronic Design).
Visual stimuli
Stimuli were generated using a ViSaGe (Cambridge Research Systems) stimulus generator
and displayed on gamma corrected Dell U2211H LCD monitor at a 70 Hz refresh rate and a
viewing distance of 180 mm. The stimulus set included sparse noise, dense noise, full‐field
luminance steps, expanding disks and natural movies. The sparse‐noise stimulus, (adapted from
(Jones and Palmer, 1987)) consisted of individual dark or bright squares (2‐20°) shown in
pseudorandom order 16 times each at 50% or 100% contrast on a 16 x 16 grid; grid spacing was
2‐5°. Dark or bright expanding disks were stepped from 0° to 40° (Zhao et al., 2014) or 5° to 100°
and displayed at 100% or 50% contrast. Full field luminance steps ranged from 50% to 100%
contrast, and natural movies were displayed at a single contrast.
16
Statistical Analyses
Construction of receptive fields
We recovered receptive fields from responses to sparse noise from membrane currents
or from spikes. To form On (bright stimuli) and Off (dark stimuli) maps of receptive fields from
subthreshold responses, we first removed action potentials using a median filter (medfilt2
function of MATLAB, The Mathworks) (Wang et al., 2007). We then averaged inward or outward
currents evoked from each On or Off pixel on the grid and integrated the response during a
manually selected time window {Martinez, 2005 #640} (Figure 2.2). For spikes, we used standard
spike‐triggered averaging of the stimulus ensemble (STA)(Schwartz et al., 2006) to generate
spatiotemporal maps. These were displayed as 2D contour plots made from the frame (or
average of nearby frames) with the peak pixel, or as time courses of the STA at the position of
the peak pixel. We quantified the sizes of receptive fields using the semi‐major and semi‐minor
axes as well as the area, of the 1 σ contours of the 2D Gaussian fits to the receptive field (Wang
et al., 2007). For cells that responded to both stimulus contrasts, the map of dominant contrast
was used to determine receptive field size.
17
Figure 2.2. Quantification of spatial receptive fields via membrane currents. (a) Average membrane currents
from a sample cell in the vLGN across time in response to sparse noise stimuli displayed for different grid
locations. Red and blue traces represent responses to bright and dark squares respectively. Currents from a
9x9 window (labeled with green in (b) on the entire 16x16 kernel) are displayed. Vertical dashed black lines
represent the onset of the stimulus. The box with dashed black outline (in (a) & (b)) represent the kernel
location with the strongest response. (b) Receptive field map of the same cell computed using the membrane
currents evoked by the bright stimuli (the dominant contrast), shown as a contour plot. Stimulus size is
indicated by the yellow square. Ovals overlay is a 1 sigma contour drawn from 2D Gaussian fit of the receptive
field.
Neuronal morphological analysis
We characterized the extents of the dendritic arbors in two ways. First, we used classical
Sholl analysis (Sholl, 1953) to measure the number dendritic intersections at progressive
distances (5 µm step size). To quantify the symmetry of neuronal arbors, we devised a metric
called “Arbor Isotropy Index” as follows:
Arbor Isotropy Index
,
Eq (1)
18
where n
distal
is the number of distal points anchoring the convex hull and d
max
is the maximal
distance of any point from the soma in the neuronal arbor. The Arbor Isotropy Index is defined
as the convex hull volume (CHV) (calculated with Neurolucida software (MBF Bioscience)) of the
arbor, normalized by a hypothetical volume designed to provide an upper limit to the CHV, given
the extent of the arbor and number of anchoring points. The CHV represents the total volume
spanned by a given neuronal arbor. To determine the hypothetical upper limit for each neural
CHV, we constructed a synthetic arbor. The number of distal points in this arbor is the number of
points in the convex set anchoring neuronal CHV. The distance of these points from the soma is
assigned as the maximum distance of any point from the soma. Calculating the maximal possible
CHV in 3D was non‐trivial. Thus, we used a similar and well‐studied approach, the Thomson
problem (Thomson, 1904) to compute the maximal CHV (MATLAB Poblano toolbox, Sandia
National Laboratories). The nonlinear nature of the problem and the practical limitations
associated with numerical optimization caused each different initialization of variables to yield a
slightly different solution. Hence, we repeatedly ran the optimization process using different
initial values and selected the configuration that generated the largest CHV. The ratio of the
neuronal CHV to the maximum possible CHV yields the Arbor Isotropy Index, with values ranging
from 0 to 1. Index values are low for neurons whose processes extend for different lengths along
different axes and are high for cells with symmetrically distributed processes.
Temporal precision and information content of neuronal responses
To quantify the temporal precision of neuronal responses across stimulus repetitions, we
used the Reliability Measure devised in (Schreiber et al., 2003). This metric reflects the
correlation between pairs of filtered spike trains and is computed as follows:
19
Reliability
∑∑
. |
||
|
Eq (2)
where N represents number of stimulus repetitions and s is the number of filtered spike trains
for individual repetitions; s is computed as the convolution of the binary spike train with a
Gaussian filter for which we set the standard deviation to 10 ms.
Neuronal information was calculated in bits per spike as described in (Brenner et al.,
2000). This information metric is defined as:
Bits per Spike log
dt Eq (3)
where T, t, r(t), r ̅ denote stimulus duration, time, average firing rate as a function of time, and
average firing rate throughout the stimulus, respectively. Data limitations (finite number of
repetitions of finite length) and narrow time bins can lead artificially high values of the metric.
Thus, we used the linear extrapolation method to address this type of error (see (Brenner et al.,
2000; Koepsell et al., 2009)).
To estimate lower bounds for temporal precision and information content within a given
spike train, we created artificial spike trains based on a homogeneous Poisson process, for which
firing rate remains constant over time. We repeated simulations to generate spike trains that
spanned the range of physiological firing rates and compared the results to calculations made
with our data.
Model fitting and comparison of temporal precision and information across populations of cells
Both the reliability and the information metrics are influenced by firing rate. Intuitively,
higher firing rates should increase the value of reliability metric and reduce the number of bits
per spike. Simulations with Poisson spike trains lack temporal structure and so are not useful for
testing these assumptions and there is not an analytical framework that relates the firing rate to
20
temporal precision and information content. Hence, to compare these quantities across cells, we
devised statistical analyses that took firing rate into account. We chose fit functions based on
their simplicity (one free parameter), the necessary boundary conditions for the firing rate (0 and
infinity), and qualitative fit. For reliability, our metric approached 0 as the firing rate goes to 0
and 1 as rate approaches infinity. Values for bits per spike approach infinity as firing rate nears
0 and, conversely, approach 0 as firing rate climbs towards infinity. Taking these properties into
account we chose the following fit functions:
Reliability
Eq (4)
Bits per Spike √ Eq (5)
where r ̅ and α denote average firing rate and the free parameter (α) respectively. To compute
the fit functions for a particular set of data points, we used maximum likelihood estimation. The
likelihood for each data point was defined by a Gaussian probability density function around the
fit. The standard deviation parameter of this Gaussian function was also a free parameter, along
with α. This created an optimization problem as follows:
L x
…x
,α,σ ∏ Nx
f α, r
,σ
Eq (6)
where L, x, i, n, α, σ, N, f, and r ̅ denote likelihood, metric value (reliability or bits per spike),
individual data point, number of data points, free fit parameter, free standard deviation
parameter, normal probability density function, fit function, and average firing rate, respectively.
Optimization was computed using a gradient descent approach with a line search strategy.
After fitting datasets as above, we compared different models statistically using the
Akaike information criterion (Akaike et al., 1974), which is defined as follows:
Akaike information criterion 2 k 2ln L
Eq (7)
21
where k represents number of free parameters and L
the maximum likelihood value for this
model. The model with the lower Akaike information criterion was defined as the better model.
Quantifying oscillatory neuronal activity
We explored oscillatory activity in both spike trains and membrane currents. To compute
the strength of oscillations, we using a statistical framework called the “oscillation
score”(Muresan et al., 2008). This analysis is based on the Fourier transform of the smoothed
autocorrelation function in which the central peak and the segment corresponding to the
refractory period are removed. The frequency with the highest amplitude in the range of interest
(the gamma band, 25 ‐ 100Hz for this study) selected and the coefficient for this frequency is
normalized by the averaged coefficient across frequencies to calculate the oscillation score. The
oscillation score along with a Confidence Score (the reliability of the oscillation score across
stimulus repetitions) is used test the significance of oscillatory activity. Significant oscillations
were defined as having oscillation scores >5 and confidence scores >0.65. We excluded responses
without a clear peak in the gamma range; this lack may have been due to data limitations (< 200
spikes total) in some cases.
To analyze oscillatory activity in membrane currents, it was necessary to adapt the
oscillation score, as follows. First, we performed a Fourier transform of spike subtracted
membrane currents. The amplitude of the frequency components of the resulting spectra
decreased with frequency, reminiscent of an exponential trajectory. This near exponential
pattern provided a challenge for the selection of the oscillation score and frequency when using
the original analysis, which relies on the assumption that there is little to no general trend in the
Fourier transform. Thus, we plotted amplitude versus frequency from membrane currents using
22
logarithmic axes, which produced a linear trend across datasets that could be fit using a line in
the logarithmic domain. Then it was possible to use this fit to normalize the amplitudes in the
frequency spectrum and so remove the confounding trend in the amplitudes with frequency. We
were able to discern the frequency with the highest amplitude in this normalized spectrum and
proceed as for the oscillation score for spikes. In addition, we could conduct the analysis with
sliding windows to characterize oscillation strength over time. Further, to explore whether or
not visual stimuli evoked oscillations, we compared oscillation strength before, during, and after
stimulus onset. For these experiments, we determined significance by bootstrapping with
resampled datasets that had the same number of repetitions as the original dataset. Thresholds
for the oscillation score and frequency were determined using the 200 ms window preceding
stimulus onset and using Bonferroni correction (for the number of samples in the window
analyzed) at significance level of 0.01. Thus, we were able to identify neural responses in which
changes in oscillation strength became significant.
We next correlated oscillatory behavior in membrane currents with spikes. First we
convolved the phase of oscillations in the membrane currents with a complex Morlet waveform
((Koepsell et al., 2009)). For the Morlet waveform, we used the oscillation frequency derived from
membrane currents as the frequency parameter and the period of that frequency as the temporal
width parameter. Then we formed a distribution of phases in which spikes occurred. This
distribution reflects how well oscillations can be propagated downstream.
To determine if and how the stimulus entrains oscillations, we computed oscillation phase
across repetitions. Thus, we computed the distribution of phases across trials before during and
after the stimulus. We fitted each distribution with a von Misses function to obtain a
23
concentration parameter as a measure for phase locking (Koepsell et al., 2009) and used
bootstrapping to determine the statistical significance of the concentration parameter. The next
step was to determine the statistical significance of phase entrainment. Thus, we first generated
a distribution of concentration parameter values for phases drawn from a uniform (0°‐360°)
distribution; this simulated distribution contained the same number of stimulus repetitions as for
the corresponding dataset. The threshold for statistically significant phase entrainment was
based on the significance level (α=0.01) along with the Bonferroni correction. We determined
the latency from stimulus onset to the initial significant peak entrainment (i.e. initial peak value
of the concentration parameter).
Linear model for oscillation score
To relate features of natural movies to oscillations in the neural response, we used ridge
regression analysis. The image statistics we considered included the average intensity of the full
image or the portion falling within the receptive field (determined from rectangular kernel tightly
enclosing the 1 σ contour of the receptive field fits). The oscillation score was modeled as the
weighted average of the image statistics computed from the preceding 214ms movie segment.
With the addition of a regularization term for ridge regression, the optimization problem can be
expressed as follows:
‖ AS B X ‖
‖ λS ‖
Eq (8)
where A, S, B, X and represent the linear term, image statistics, constant term, and oscillation
scores from the training data and ridge parameter, respectively. The solution for A and B were
obtained by minimizing this function. To estimate model performance, leave‐one‐out cross‐
validation was used (one stimulus repetition is reserved for testing and all other trials used for
24
training, with the process repeated to leave out each trial). Next we calculated the Pearson
correlation between oscillations scores that the model predicted and each repetition used for
testing. The mean correlation across trials was calculated for a wide range of values of the ridge
parameter and the value that gave the maximum mean correlation was selected for each model.
To provide an upper bound for assessing model performance, the average correlation between
oscillation scores of individual repetitions and the average oscillation score across the remaining
repetitions from training set was calculated. This entire analysis was done separately for mean
contrast in the full image and within the region demarcated by the receptive field.
Additional statistics
In addition to the analyses described above, statistical variation for a given data set is
reported as mean ± standard error, unless otherwise noted.
2.3 Results
We asked how the non‐image forming division of the LGN processes visual information.
In carnivore and primate, the dLGN dwarfs the vLGN. However, the two subnculei are almost the
same size in rodent, suggesting that the functions the vLGN serves are especially important in
murine species. Here, we used an interdisciplinary approach to studying the mouse vLGN,
including extracellular and whole‐cell recording in vivo, anatomy and computational tools.
Further, to gain a functional perspective, we compared the results from vLGN to those obtained
in the dLGN and superior colliculus (SC), subcortical structures that receive direct retinal input
and are involved in either the form vision (geniculostriate) or the visuomotor (retinotectal)
pathways. Our dataset includes 39 visually responsive neurons from the vLGN (or adjacent
25
intrageniculate leaflet (IGL)), 11 neurons from the SC, and 59 from the dLGN that were part of an
earlier study (Suresh et al., 2016) as noted in the text.
Comparison of receptive field structure and scale in the dLGN vs the vLGN
While receptive fields in the dLGN have been quantified in several taxonomic orders
(primate, carnivore and rodent (Shapley and Lennie, 1985; Piscopo et al., 2013)), there are only
scant and qualitative descriptions of receptive fields in the vLGN ((Spear et al., 1977; Sumitomo
et al., 1979)). Thus, we used a sparse‐noise stimulus to map On and Off components of neural
receptive fields in the murine vLGN and used 2D Gaussian fits to quantify their size. Then we
compared receptive fields in the vLGN with those in the dLGN. Typically, receptive fields of cells
in the dLGN were smaller and appeared less variable, Figure 2.3a, than those in the vLGN, Figure
2.3b, (receptive fields are shown as contour plots with an overlay representing 1σ of the Gaussian
fit) though there was a range of spatial extents from smaller to larger, Figure 2.3b, and see
(Suresh et al., 2016). To visualize population differences, the 1σ fits of maps for all cells in the
dLGN (purple) and vLGN (green) were aligned at their centers and superimposed, Figure 2.3c,
left. To quantify differences between receptive fields in the dorsal vs ventral LGN, we used two
measures. First, we compared the extents of the semi major and semi minor axes obtained from
the Gaussian fits, Figure 2.3c, top right; (vLGN: 33.99±2.05°; dLGN: 13.63±1.35°, p=2.3x10
‐8
,
Wilcoxon rank sum test, n=25 for each subdivision). Second, we compared the area, Figure 2.3c,
bottom right, (vLGN: 897.83 ± 102.24(°)
2
; dLGN: 173.75±41.51 (°)
2
, p=2.3x10
‐8
, Wilcoxon rank
sum test, n=25 for each subdivision). Both metrics show that receptive field size in the vLGN is
significantly larger than in the dLGN.
26
Figure 2.3. Receptive fields in the dLGN are usually smaller than those in the vLGN. (a) Contour plots of sample
receptive fields from dLGN derived from membrane currents of an On‐center (left) and Off‐center relay cell
(right). Overlays are 1 σ contours from 2D Gaussian fits of the receptive fields. The stimulus was sparse noise
and the yellow box at the upper left of each plot indicates stimulus size. (b) Sample receptive fields in the
vLGN for three On cells (top left, top right, bottom right) and an Off cell (bottom left), conventions as in a. (c)
(left) Plot of the 1 σ contours for receptive fields of relay cells in the dLGN, purple, and cells in the vLGN/IGL,
green; the contours are aligned at each receptive field center. (top, right) Bar graph comparing the average
size of receptive fields in vLGN and dLGN. (bottom, right) Bar graph comparing the area of receptive fields in
vLGN and dLGN. Error bars indicate SE. * denotes p<0.01.
27
Most of the cells we recorded in the vLGN preferred bright stimuli. All in all, there were
17 On, 3 Off, and 4 On‐Off (3 On dominant, 1 Off dominant) cells. Four remaining cells were
inhibited rather than excited by the presentation of visual stimuli, 3 were inhibited by bright
stimuli and an additional cell was inhibited by bright and by dark stimuli. The receptive fields of
some of these cells appear in later in this section.
Neuronal Morphologies in different division of the LGN
The large size of receptive fields in the vLGN suggested that single cells were able to pool
input arriving from broad retinal territories. By labeling and reconstructing cells in the vLGN, we
found that the dendritic arbors were wide‐ranging and that the same was true for axons. Of the
8 cells we labeled in full, 4 had dendrites that spanned the full mediolateral extent of the nucleus
and another spanned the entire dorsolateral axis. Furthermore, 6 cells had processes that spread
outside of the vLGN. In dLGN, by contrast, previous work ((Krahe et al., 2011)), including our own
(Suresh et al., 2016), had shown that the arbors of all subtypes of relay cells are compact,
sampling smaller portion of retinotopic space (Dhande et al., 2011). Reconstructions for 2 sample
cells from the dLGN (Figure 2.4a) and vLGN (Figure 2.4b) illustrate the difference in arbor size
between nuclei. We also filled 3 cells from the IGL; their arbors spanned the full leaflet and are
pooled with vLGN in the analyses below.
28
Figure 2.4. Morphological differences between neurons in dLGN vs vLGN. (a) Reconstructions of a Y (top) and
an X relay cell (bottom), in the dLGN projected in the coronal plane. Somas are outlined in black and the
contour surrounding each cell indicates the boundary of the dLGN at the location of the soma. (b) Drawings
of two different cells in the vLGN; conventions as in (a). (c) Plot of the distribution of processes as a function
of distance from the soma computed using Sholl analysis for the dLGN (purple) and vLGN (green). The values
for each neuron were normalized before calculating population statistics; shading indicates the SE. Dashed
vertical lines indicate the mean distance from the soma for each LGN subdivision, and horizontal error bars the
SE. The inset at the upper right plots the percentage of neurons that had processes that reached the distances
indicated along the abscissa. Arbors of all neurons in the vLGN extended further than those of relay cells in
the dLGN. (d) (left) Plot of Arbor Isotropy Index (AII) values in the vLGN and the dLGN and (right) a graphical
depiction the AII in 2D. For a neuron with four dendrites of equal length along four cardinal directions, the AII
value is 1, with progressively smaller values when the relative lengths of dendrites extending along each axis
differ as illustrated; the dashed lines indicate the convex hull. For reference, the AII value for the neuron drawn
in a (top) was 0.339, (a, bottom), 0.279, (b, top), 0.042, and (b, bottom), 0,042.
In order to quantify the difference between the extent of neuronal processes in each
subnucleus, we first used the Sholl analysis (Sholl, 1953), which counts how many processes
intersect each of a series of concentric rings centered on the soma. Cells in the vLGN sampled
29
distances as far as 600 µm from the soma, whereas arbors of relay cells rarely extended beyond
150 µm. The average intersection distance for neurons in the vLGN was roughly double that for
cells in the dLGN (vLGN: 116.1 ± 14.8 µm; dLGN: 52.7 ±1.3 µm, p=8.2x10
‐5
, Wilcoxon rank sum
test, n=11 for each subdivision). Histograms of the results of the Sholl analysis for each neuron
were normalized before calculating population statistics, Figure 2.4c. A companion analysis
shows the percentage of neurons with processes that reached a given distance from the soma,
Figure 2.4c, inset. These analyses showed that the arbors of relay cells in the dLGN tended to
have similar diameters whereas there was considerable variation among cells in the vLGN.
Note that Sholl analysis is only a 2D metric and fails to give a sense of the anisotropy of
projections. Yet, while dendrites in the dLGN seemed to extend, with modest variation (Krahe et
al., 2011), for similar distances along different directions, dendrites in the vLGN appeared to have
marked directional preferences. Thus, we devised a metric to capture elements of arbor
architecture that Sholl analysis overlooks ‐‐ the degree to which arborizations of cells extended
for substantially longer or shorter distances along particular axes. We call the new measure the
“Arbor Isometry Index”. It is defined as the convex hull volume enclosing the arbor, normalized
by a hypothetical upper limit to this volume, based on the extent of the arbor and number of
points anchoring a given convex hull volume (see Materials and Methods). The value of the
index is 1 for perfect symmetry and decreases as the preference for some axes grow stronger. A
flattened (2D) graphical depiction of how the Arbor Isotropy Index reflects directional biases in
the dendritic arbor is provided for three sample configurations, shown to the left of the
population data in Figure 2.4d. As per our initial impression; values in the vLGN were significantly
lower than in the dLGN (vLGN: 0.072 ± 0.015; dLGN:0.229 ± 0.024, p=3.0x10
‐4
, Wilcoxon rank sum
30
test, n=11 for each subdivision). All told, our anatomical analyses were consistent with the
observation that receptive field sizes were larger in the ventral vs dorsal LGN and suggest other
differences between nuclei as well.
Synaptic physiology in different division of the LGN
As above, receptive fields are larger and dendritic arbors wider in the ventral vs dorsal
division of the LGN. We asked if these observations were accompanied by differences in synaptic
physiology. To explore this possibility, we analyzed recordings of membrane currents. Sample
recordings from the dLGN and the vLGN reveal differences in the pattern of excitatory
postsynaptic currents (EPSCs) in each subdivision, Figure 2.5. In the dLGN, intracellular currents
recorded from relay cells were dominated by trains of sharp, unitary EPSCs that can be
distinguished by eye and easily labeled using tools such as a support vector machine (Suresh et
al., 2016), Figure 2.5a. These EPSCs are almost certainly retinogeniculate in origin (Koepsell et
al., 2009; Suresh et al., 2016). The recordings from the vLGN were more complicated, however.
Localizing individual EPSCs in the recordings from the vLGN was challenging because events were
small, slow or overlapping, when visible at all, Figure 2.5b. This observation is consistent with
the idea that if neurons in the vLGN pool a larger number of inputs than in dLGN, then the
contribution of each individual inputs is commensurately less (Turrigiano et al., 1998). Our results
are also consistent with the idea that many inputs are made at remote positions on the dendrite
(Fatt and Katz, 1951; Bloomfield et al., 1987), as might be expected given the large extents of
dendritic arbors in the vLGN.
31
Figure 2.5. The shapes of membrane currents recorded from the dLGN and vLGN. (a) Membrane currents
recorded from two cells in the dLGN (purple) and (b) two cells in the vLGN (green). Retinogeniculate EPSCs are
easily visible in traces from relay cells in the dLGN, where asterisks mark the first EPSC. By contrast, individual
synaptic events are difficult to discern in records from the vLGN.
Receptive field structure and response timing in the vLGN and the SC
Unlike the dLGN, which innervates cortex, the vLGN projects subcortically, with a main
target the SC—a hub for sensorimotor behaviors (Monavarfeshani et al., 2017; Cang et al., 2018)
(see Figure 2.6). The SC projects back to the vLGN in turn (Harrington, 1997). To understand how
the vLGN might interact with visuomotor brainstem regions, we compared receptive field
structure and spike timing in vLGN to that in the superficial (retinorecipient) SC. For this analysis,
we estimated receptive fields using spike‐triggered averaging of responses evoked by sparse
noise and then quantified size using 2D Gaussian fits (while receptive field structure had been
mapped previously in the mouse SC (Wang et al., 2010a; Gale and Murphy, 2014; Ellis et al.,
2016), it was important for us to collect our own dataset using a standardized stimulus). The
32
receptive fields in the vLGN were larger than those in the superficial SC, Figure 2.7, (2D extent
LGN: 32.48 ± 2.47° (n=24 cells); SC:15.45 ±1.96° (n=11 cells), p=3.5x10
‐5
, Wilcoxon rank sum test;
mean area; vLGN: 853.70 ± 160.10 (°)
2
(n=24 cells); SC: 176.30 ± 35.85 (°)
2
(n=11 cells), p=1.3x10
‐
5
, Wilcoxon rank sum test).
Figure 2.6. Afferent and efferent projections of the vLGN. vLGN is connected with many structures which have
motor functions, including the superior colliculus (for a review see (Harrington, 1997)). dLGN, dorsal lateral
geniculate nucleus; vLGN, ventral lateral geniculate nucleus.
33
Figure 2.7. Receptive field maps and sizes in the vLGN and SCs. (a) Sample receptive field of an On cell in the
vLGN mapped with flashed bright spots. (b) Sample receptive field of an On cell in the SCs mapped as in (a).
Ovals overlays are 1 σ contours drawn from 2D Gaussian fits of the receptive fields. The spacing of the stimulus
grid is 5° and stimulus size is indicated by the yellow squares. Membrane currents across repetitions are used
to generate receptive field maps. (c) Bar graph comparing the size of receptive fields in vLGN and SCs. Size is
calculated as the average of the extent in the semi‐major and semi‐minor axes of the 1 σ contours. Error bars
indicate the standard error around the mean. Wilcoxon rank‐sum test was used to assess significance. *
denotes p<0.01. vLGN, ventral lateral geniculate nucleus; SCs, superficial layers of SC.
Our next question was whether or not the size of receptive fields in a given area and the
degree of temporal precision go hand in hand. That is, since small receptive fields often correlate
with high visual acuity, we wondered if they might also correlate with high temporal precision
(reliability across trials). Temporal precision has implications for the speed of sensory processing
as well as neural coding, including the amount of information conveyed within spike trains (de
Ruyter van Steveninck et al., 1997; Brenner et al., 2002; Schreiber et al., 2003). Hence, after
mapping the receptive field, we recorded responses to natural movies, a stimulus that roughly
mimics the animal’s experience in the environment. Sample receptive fields and raster plots of
responses to repeated presentations of three different movies for cells from the vLGN, Figure
2.8a, the dLGN, Figure 2.8b, and the SC, Figure 2.8c, illustrate the approach. The level of temporal
precision changed throughout the stimulus and from cell to cell. At the population level,
however, the spike trains recorded from the SC and dLGN seemed more reliable.
34
In order to quantify potential differences in temporal precision, we used a correlation‐
based reliability measure (Schreiber et al., 2003), see Materials and Methods. This measure
assesses jitter in the spike train from one stimulus trial to another; values range between 0 and
1, with 1 indicating maximal precision. Note the measure is influenced by firing rate, so that we
chose to plot index values against firing rate for best comparison of reliability between different
brain structures. The resulting scatterplot, Figure 2.8d, left, indicated that cells in the SC (orange)
and dLGN (purple) fire more reliably than those in the vLGN (green) at a given firing rate. Each
dot in the plot represents an index value computed for a single cell’s response to each movie
stimulus (maximum number of points per cell is 3).
To perform a statistical comparison between populations, we used an approach that
models the reliability score as a function of firing rate. We fitted the distribution of points for
each brain structure (dashed lines in Figure 2.8d, left) and computed the likelihood of each fit
(see Materials and Methods). The curves for the SC and dLGN lay above that for the vLGN,
suggesting a higher degree of temporal precision in both the dLGN and tectum than the vLGN.
To determine if the results from the SC and dLGN were significantly different from vLGN,
we used the Akaike information criterion (Akaike, 1974), first comparing the vLGN with the SC
and then with the dLGN. This standard tool estimates the relative performance of statistical
models for a given dataset; low values correlate with better model quality. We first compared
models that fit the data from the SC and the vLGN either as separate or as joint sets and found
that model built using a separate set for each regions performed best. The Akaike information
criterion for separate models was ‐123.4 (green and orange dashed lines) and for a single model
it was ‐82.1; vLGN, n=37 responses, SC n=18 responses (Figure 2.8d). This means that the
35
increase in likelihood for the separate models was greater than the penalty introduced by the
number of parameters required for an additional model. When we restricted the analysis to
datasets made from responses to a single movie only, the performance of the separate versus
joint models was similar; the value for separate models was, ‐67.6, and for the single model, ‐6.4;
vLGN, n=25 cells, SC n=10 cells.
To address any biases that might have been introduced by differences in firing rates
across nuclei, we performed additional analyses. First, we analyzed a subset of data for firing
rates (4.4 ‐ 28.7 Hz) for which there was substantial overlap between regions. Again, the separate
model outperformed the single model (Akaike information criterion: vLGN, ‐85.6, n=19
responses; SC, 39.7; n=15 responses). Second, we calculated scores for reliability; these were
also significantly higher in the SC (0.560 ± 0.035, n=15) than the vLGN (0.318 ± 0.030, n=19),
p=1.2x10
‐4
, Wilcoxon rank sum test. Further, to give a sense of reliability in both regions
compared to responses with random temporal structure, we fit a curve to spike trains generated
by a Poisson process Figure 2.8d, left, gray dashed line. Last, we compared temporal precision in
the ventral and dorsal divisions of the LGN using the Akaike information criterion. The responses
from the dLGN were more reliable, the criterion values for the separate models were ‐137.0 and
for the single model = ‐97.3; n=37 for both subdivisions. For a single type of movie only, the
criterion for separate and single models were ‐84.5 and ‐65.8 respectively; n=25 for both
structures. For the dataset associated with overlapping range of firing rate (2.5 – 28.7 Hz), the
criterion was ‐130.2 and ‐79.4 for separate and joint models respectively (vLGN, n=28 responses,
dLGN=36 responses). And difference in reliability between vLGN (0.270 ± 0.025, n=28) and dLGN
(0.503 ± 0.033, n=36) was significant (p=8.2x10
‐6
, Wilcoxon rank sum test).
36
Figure 2.8. Temporal precision and information content of spike trains evoked by natural scene movies for the
vLGN, dLGN and SC. (a) Responses of two cells in vLGN (green), (b) one cell in dLGN (purple), and (c) one cell in
the superficial SC (orange) to three different natural movies. Receptive fields (STAs) made with sparse noise
are shown to the left of raster plots showing spiking responses to repeated stimulus trials for three different
movies for each cell. (d) (left) Plot of the Schreiber reliability measure against spike rate. Each point was
calculated from a given cell’s response to one of the three movies. Dashed lines denote the best fits (see
legend in plot for color codes and the fitting function used). The gray line plots results from a simulation with
homogenous Poisson spike trains. (right) Plot of bits per spike against firing rate, conventions as for d (left).
37
Information in Spikes of the vLGN and the SC
Here, we computed the information in bits per spike ((Brenner et al., 2002)) (see
Materials and Methods) using the dataset obtained with natural movies. The results, displayed
as points in a scatter plot, Figure 2.8d, right, conventions as in Figure 2.8d, left, show that for
responses with similar firing rates, the information per spike was usually higher for SC than vLGN.
In general, the amount of information that a single spike contributes decreases as firing rate
increases, and we quantified this relationship for our data. We modeled bits‐per‐spike as a
function of firing rate and plotted the results as dashed lines on the scatterplot. We estimated
the likelihood of these fits, again using the Akaike information criterion. The fits for the SC lay
above those fit for the vLGN. Correspondingly, the criterion values for the separate model was
45.4 and for the single model, 68.2; n=37 for vLGN, and n=18 for SC. Further, when we split the
datasets into responses to individual movies, the values obtained for the Akaike information
criterion were similar with the value for separate models, 32.6, and for single model, 50.0; n=25
for vLGN, n=10 for SC.
We next recalculated these values using the range of firing rates (4.4 ‐ 28.7, Hz) for which
there was substantial overlap between the SC and vLGN. Again, the criterion value for the
separate model, 10.9, was better than for the single model, 49.2; n=19 for vLGN, n=15 for SC.
Values of bits‐per‐spike, independent of firing rate, in the SC were higher than vLGN (vLGN:0.567
± 0.028 (n=19 responses); SC:1.245 ± 0.100 (n=15 responses), p=1.0x10
‐6
, Wilcoxon rank sum
test). For reference, we calculated information in bits‐per‐spike from simulated Poisson spike
trains, Figure 2.8d, right, dashed gray lines; in almost all cases, values from biological spike trains
were higher. We repeated the analyses for the two divisions of the LGN; as expected, values for
38
the dLGN were higher than the vLGN at comparable firing rates, Figure 2.8d, right. The Akaike
information criterion values were, 101.8, for the separate model and, 155.2, for the single model;
n=37 responses for both subnuclei. For a single movie type, the criterion values for separate and
single models were 68.5 and 101.8 respectively; n=25 for both structures. For the dataset
associated with overlapping range of firing rate (2.5 – 28.7 Hz), the criterion was 62.9 and 135.5
for separate and joint models respectively (vLGN, n=28 responses, dLGN=36 responses). And
difference in the information per spike between vLGN (0.689 ± 0.043, n=28) and dLGN (1.581 ±
0.079, n=36) was significant (p=2.98x10
‐10
, Wilcoxon rank sum test). Thus, single spikes in the
vLGN conveyed less information about the stimulus than do those in the dLGN or SC.
Oscillatory neural responses in the vLGN and the SC
Neurons can encode information about the stimulus in two ways; changes in spike rate
with respect to an extrinsic signal (the stimulus), and spike timing with respect to intrinsically
generated rhythms such as oscillations. The analyses that we illustrated in Figure 2.8 pertain to
the rate code. However, there was reason to explore potential roles oscillation in the vLGN. In
earlier work we (Koepsell et al., 2009; Koepsell et al., 2010) and others (Saleem et al., 2017;
Storchi et al., 2017) showed that some cells in the dLGN encode information both ways, and
others have demonstrated visually evoked oscillations in the SC (Brecht et al., 2004; Sridharan et
al., 2011; Stitt et al., 2013) (Figure 2.9). On balance, there seemed to be a difference between
oscillation‐based coding in the dLGN and SC, however. In the dLGN, oscillations typically reflect
ongoing activity in the retina and their phase is rarely entrained by visual stimuli. By contrast,
past work indicates that oscillations in the SC are evoked by visual stimuli and, hence, seem to be
39
generated by local networks (Sridharan et al., 2011) (Brecht et al., 2004; Stitt et al., 2013). Thus,
we explored the possibility of both types of oscillations in the vLGN.
We often recorded visually evoked oscillations resembling those reported in the SC in
vLGN. Full‐field stimuli evoked oscillations in the spike trains of 9 cells (6 of 19 On cells and 3 of
7 Off cells), Figure 2.10a. Bright but not dark expanding discs evoked oscillations in 2 of 3 cells.
By contrast, in SC, dark stationary and expanding (looming) discs drove oscillations ((Zhao et al.,
2014)) (Figure 2.10b). Here oscillation frequency and strength were computed using the
oscillation score method (Muresan et al., 2008).
Figure 2.9. Visual oscillations in the superior colliculus from earlier studies. (a) Oscillatory neural responses in
the superficial layers of ferret SC following a brief full field flash. (top) raster plot and (middle) peristimulus
time histogram from multiunit activity. (bottom) spectrogram computed from local field potential (figure taken
from (Stitt et al., 2013)). (b) Oscillatory neural responses in the superficial layers of barn owl optic tectum (avian
homologue of superior colliculus) following visual stimulus. (top) Unfiltered activity (gray) and filtered local
field potential (red). (bottom) average R‐spectrogram showing the relative change in power across time.
(bottom right) power ratio across frequency at 260ms (from R‐spectogram) (figure taken from (Sridharan et
al., 2011)).
40
Figure 2.10. Visually evoked oscillatory spike trains in the vLGN and SC. Examples of oscillatory spike trains in
the vLGN, green (a) and SC, orange, (c) evoked by a full‐field stimulus. Receptive fields are shown to the left
of spike rasters of responses to repeated trials of the stimulus that is depicted above each raster plot. Plots of
oscillation strength against frequency are shown to the right of the rasters, with insets containing values for
the oscillation score (OS) and frequency (f). (b & d) Same as a & c except the stimulus was an expanding spot,
as indicated in the figure. Spot‐size increased from 5° to 100° for vLGN and from 0° to 40° (at 50% contrast) for
SC.
Oscillatory membrane currents in response to visual stimulation in the vLGN
We assumed that the oscillatory activity recorded in spike trains reflected oscillations in
underlying membrane currents. To test this assumption, we obtained whole‐cell patch recordings
such as those shown in Figure 2.11a and modified the Oscillation Score method for subthreshold
responses (see Materials and Methods). Our modified analysis revealed oscillatory currents
similar in frequency to that observed for spike trains, Figure 2.11b. Next, we explored the
temporal relationship between membrane currents and spike timing by calculating the phase
coherence between the two types of signal (see Methods and Materials). Our results show that
spikes phase‐locked to membrane oscillations faithfully (Figure 2.11c).
Because intracellular responses provide a sensitive measure of neural response, we used
these to explore the evolution and duration of visually evoked oscillations. Oscillation strength
41
lagged the initial neuronal response, likely because it took time for the network dynamics to
change state, Figure 2.11d. During the stimulus, the oscillation score remained high (albeit with
some modulation) and then began a steady decline to baseline after the stimulus ended. Thus,
oscillations, once engaged, were maintained throughout stimulus duration. This example was an
On cell and the responses displayed were evoked by a full‐field stimulus. However, we have
observed similar responses for On‐inhibited and On‐Off cells by using dark stimuli.
The nature of the information that the oscillations convey depends, in part, on the degree
of precision with the stimulus entrains oscillations. Thus, we asked if oscillations phase‐lock to
the stimulus and found that this was most often the case. Shortly after stimulus onset, the phase
of membrane currents across repetitions became aligned, as illustrated in Figure 2.11e, and see
Materials and Methods. We observed significant entrainment for 72% of the 18 cells that
oscillated in response to the full field stimulus; the mean latency of entrainment was (mean ±
standard deviation = 98 ± 63 ms). Last, because we hyperpolarized cells slightly to visualize
membrane currents, firing rates were too low to provide sufficient data for a parallel analysis
with spikes.
42
Figure 2.11. Flash‐evoked gamma band oscillatory currents for a sample ON cell in the vLGN. (a) (top) Contour
plot of the receptive field, conventions as in Figure 2.3. (bottom) Icon depicting a full‐field bright stimulus is
shown above two individual responses to the stimulus, and the average across trials, bolded and at 2X gain. (b)
Illustration of how the oscillation score analysis was adapted for membrane currents. The oscillation frequency
is computed by taking the frequency that has the highest amplitude (dotted vertical line) with respect to the
linear trend (dashed line) in the spectrum (plotted on a logarithmic scale). The oscillation score is then
computed as the relative spectrum amplitude, normalized by the trend, at the oscillation frequency. (c)
Histogram of phase coherence between intracellular currents and spikes during the stimulus reveals temporal
locking between synaptic input and neural output. (d) Plot of the mean ± SE of the oscillation score of
membrane currents over time across trials; values were calculated using a 204.8 ms sliding window (n=25
repetitions). * denotes p<0.01. (e) Plot of the mean ± SE of the phase concentration parameter (κ) over time.
The dashed horizontal line marks the threshold for significant entrainment (α=0.01).
43
Oscillatory membrane currents during natural stimulation
Simple flashed stimuli evoke strong responses but are far from the spatiotemporal
patterns the animal experiences in the environment. Thus, we investigated the potential role of
oscillations for natural vision by making whole‐cell recordings during the presentation of natural
movies. We observed that oscillations in membrane currents waxed and waned during specific
stimulus sequences and quantified those changes using the oscillation score, Figure 2.12. Figure
2.12a, plots receptive field maps for a single cell above a graph of oscillation strength over time
made from responses to a movie of snow monkeys. Smooth transitions in the image and abrupt
changes, both, evoked preceded increases in oscillation strength, as seen in the membrane
responses to sample sequences shown in panels 7B, where time markers (t
1‐7
) indicate
corresponding points in the plot of oscillation score. We wondered if the oscillations were driven
locally from the portion of the image falling within the receptive field, or globally, by the entire
visible image. The oscillation score was modeled by a linear function of average stimulus
intensity for the entire image or portion within the receptive field. This model was fit using ridge
regression and the fitted model was used to generate an oscillation score waveform. The
correlation between this prediction and oscillation scores from individual repetitions spared for
testing was calculated, along with realistic upper bounds of performance, Figures 2.13b, d and
see Materials and Methods. This analysis suggested that stimulus feature in the receptive field
had more predictive power for oscillations than features of the full field. Overall, this simple
model was able to achieve performances close to the realistic upper bound. A similar analysis for
a second cell is also illustrated, Figure 2.13c, d.
44
Figure 2.12. Changes in the strength of membrane oscillations during the presentation of natural scene movies
in the vLGN. (a) On and Off maps of the receptive field of an On‐Off cell and an illustration of the stimulus are
shown above the oscillation score computed continuously from membrane currents recorded during the movie
(n=10 repetitions); the solid curve is the mean and the shading is the SE. The 1s segments of the curves marked
by the dashed and solid lines correspond to responses illustrated in (b) as do the vertical dashed lines labeled
t
1‐7
. (b) (Top panel), Image sequences (every 10
th
frame of the movie) for the interval marked in by the dashed
line in (a), with the position of the receptive field indicated by an overlay, are shown above sample membrane
currents (two individual traces with the average of all trials bolded at 2X gain). (Bottom panel), same as for the
top panel but for the interval marked by the solid line. (c‐d) Same as (a‐b), but for a cell that was inhibited by
On and Off stimuli, n=20 stimulus repetitions.
45
Figure 2.13. Modeling of oscillatory activity based on visual stimulus using ridge regression during natural scene
movies in the vLGN. (a) On and Off maps of the receptive field of an On‐Off cell and an illustration of the
stimulus are shown above the oscillation score computed continuously from membrane currents recorded
during the movie (n=10 repetitions); the solid curve is the mean and the shading is the SE. (b) Bar plots of
correlations between predicted oscillation scores made using ridge regression analysis and oscillation scores
computed from the biological data for the mean contrast of the full field or the mean contrast of the image
patch within the receptive field; error bars indicate mean ± SE. The dashed line indicates a theoretical upper
bound for model performance. (c‐d) Same as (a‐b), but for a cell that was inhibited by On and Off stimuli, n=20
stimulus repetitions.
2.4 Discussion
To learn about the role of the vLGN in visual processing, we compared it to the dLGN,
which projects to cortex to serve form vision, and the superficial SC, a subcortical hub for
sensorimotor processing that is reciprocally connected with vLGN. Compatible with past studies
in different species (Spear et al., 1977; Sumitomo et al., 1979; Harrington, 1997), we found that
receptive fields in vLGN were larger than those in the other two structures. To address the
46
structural basis of the difference in receptive field size, we labeled individual neurons in both
divisions of the LGN and found that the length of individual dendrites and the volume of the full
arbors were far larger in the ventral vs dorsal LGN, suggesting that former is able sample a wider
distribution of inputs than the latter. Further, the arbors of cells in the vLGN (and IGL) usually
extended along a preferred direction, in accord with the idea that there are hidden sublaminae
with the nucleus (Monavarfeshani et al., 2017). EPSCs in the vLGN seemed smaller and more
variable in shape that those recorded from relay cells in dLGN, supporting the view that cells in
the vLGN (and IGL) collect a great number of retinal inputs along the lengths of the dendrites
(Hammer et al., 2014). In addition to exploring spatial processing, we analyzed sensory
integration in time. Overall, neurons in the SC and dLGN fired with greater temporal precision
than those the vLGN, in keeping with differences in receptive‐field size across nuclei (Grubb and
Thompson, 2004; Wang et al., 2010a; Piscopo et al., 2013; Gale and Murphy, 2014; Suresh et al.,
2016). Further, a subset of cells in vLGN responded to visual images with gamma band
oscillations whose phase could be entrained by the stimulus and whose strength was modulated
by changes in artificial and natural visual patterns. This type of oscillation occurs in the SC (Brecht
et al., 2001; Stitt et al., 2013), suggesting that the vLGN can engage its targets using oscillation‐
based as well as rate codes. Finally, we note that the neurons in the vLGN usually prefer large
bright shapes, whereas expanding dark stimuli activate the tectum and drive escape behavior.
Thus, one role of the vLGN might be to release to enable rapid movement by releasing inhibition
and then to reset inhibitory tone to restore stability.
47
Structure function relationships in LGN
Our quantitative results showing that receptive fields in the vLGN are significantly larger
than those in the dLGN confirm qualitative descriptions from earlier studies in cat and rodent
(Spear et al., 1977; Sumitomo et al., 1979; Harrington, 1997). To explore the structural basis of
this difference in receptive field size we labeled and reconstructed single projection cells in each
main division of the LGN. The dendritic arbors of cells in the vLGN (and associated IGL) were
large and spanned the length and/or width of the nucleus. Our sample came from visually
responsive cells positioned throughout the nucleus, presumably concentrated in the
retinorecipient zones (Monavarfeshani et al., 2017). The arbors throughout the dLGN, by
contrast, were relatively compact. In addition, dendritic arbors in the vLGN or IGL, but not dLGN
often extended along a preferred direction, a feature we quantified by developing a new measure
of directional symmetry. These results suggest two conclusions. First, cells the vLGN are
positioned to receive convergent input from ganglion cells at a fuller range of retinotopic
distances than their counterparts in the dLGN. Second, the anisotropy of the arbors suggests
selectivity for a subset of retinal inputs and is particularly interesting in light of the finding that
some types of ganglion cells preferentially target specific zones within the vLGN (Monavarfeshani
et al., 2017).
The idea of greater convergence in the ventral vs dorsal LGN is in accord with differences
in the synaptic physiology and anatomy between subdivisions of the LGN. A corollary of principle
of synaptic scaling holds that the strength of individual inputs decreases as their number
increases (Turrigiano et al., 1998). Accordingly, recordings from brain slices show that
retinogeniculate EPSCs are smaller in the ventral vs dorsal LGN (Hammer et al., 2014;
48
Monavarfeshani et al., 2017), consistent with our finding in vivo that individual EPSCs are large
enough to detect in dLGN but that they can be difficult to resolve from each other or noise in
vLGN. The placement of retinal inputs input might also contribute to differences in EPSC shape
across nuclear subdivisions, as follows. Projection neurons in both divisions of the LGN receive
proximal retinal input, but only those in vLGN seem to synapse with retina at distal sites (Stelzner
et al., 1976); remote EPSCs smooth and shrink as they travel centrally, yielding a varied
complement of event shapes at the soma. Further, ultrastructural differences in the size and
complexity of retinal terminals correspond to the differences in synaptic physiology between
divisions of the LGN; usually, retinal boutons are large in the dLGN and small in vLGN (Hammer
et al., 2014). Last, synaptic input patterns in the IGL were similar to that of vLGN.
Temporal processing in the vLGN vs the the SC and dLGN—the main subcortical nuclei in the
sensorimotor and form vision pathways
We asked whether or not differences in the spatial scale of receptive fields paralleled
differences in temporal precision by comparing spike trains recorded from the vLGN, where
receptive field are usually large, with those from cells dLGN and superficial SC, where receptive
fields are usually smaller (Grubb and Thompson, 2004; Wang et al., 2010a; Piscopo et al., 2013;
Gale and Murphy, 2014; Suresh et al., 2016). We used a correlation‐based (Schreiber et al., 2003)
method to explore the temporal precision of responses to natural movies. Overall, responses in
vLGN were not as precise as in dLGN and SC. Consequently, the amount of information each
spike carried in the vLGN was lesser than in the other two nuclei. These differences in response
timing are consistent with the idea that the vLGN modulates rather than instructs downstream
targets.
49
Neurons can encode information by changes in spike timing with respect to the stimulus,
as above, or by spike timing with respect to network oscillations. Oscillations serve diverse
functions like increasing amount of information spike trains convey, resolving stimulus features
and context and routing information from one structure to another (Koepsell et al., 2010; Storchi
et al., 2017). Oscillations take different forms. Some, as in retina (Ishikane et al., 1999; Koepsell
et al., 2010), are ongoing and are modulated rather than evoked by the stimulus (Saleem et al.,
2017; Storchi et al., 2017). Other types occur after a stimulus activates a given network
(Sridharan et al., 2011). We recorded stimulus‐evoked (gamma band) oscillations from a subset
of cells in the vLGN. Oscillation strength waxed and waned during the presentation natural
movies, including image sequences with smooth or abrupt changes in pattern, a result that
suggests that oscillations can be driven both by object motion and self motion, respectively.
Further, our computational modeling suggested that the oscillations are driven by stimuli falling
within the receptive field and are thus driven by local stimulus features. Interesting, visually‐
evoked oscillations are also seen in subcortical motor structures the vLGN connects with, like the
SC (Brecht et al., 2001; Sridharan et al., 2011; Stitt et al., 2013). Thus, communications between
the vLGN and partner structures involve both rate‐based and oscillation‐based codes.
Potential functional roles of vLGN
Several lines of evidence support a view that the vLGN plays a role in visuomotor
integration. It is densely interconnected with subcortical structures that serve sensorimotor
function ((Monavarfeshani et al., 2017). Further, most if not all, of the projection neurons in the
vLGN are gabaergic (Monavarfeshani et al., 2017), resembling the basal ganglia, which also
contact remote targets via inhibitory connections (Nelson and Kreitzer, 2014). These types of
50
projections cause excitation indirectly, via release from inhibition, a mechanism that might help
to prevent runaway activity (A Whittington and Traub, 2004) and hence, unwanted movements.
Moreover, we find that the type of oscillations in vLGN are stimulus‐evoked, like those in the
superior colliculus, rather than ongoing, as in the dLGN. These oscillations, like changes in rate,
are preferentially evoked by bright stimuli for the vLGN but dark ones for the SC. This reverse
preference for stimulus contrast between structures is interesting in the context of the escape
response, a behavior mediated by the tectum that is triggered by looming dark stimuli (Gandhi
and Katnani, 2011; Cang and Feldheim, 2014; Zhao et al., 2014). The inhibition the vLGN provides
in sunlight would be reduced as a predator cast a shadow, disinhibiting the SC to allow rapid
flight.
In aggregate, our and others’ results suggest that the vLGN does not convey the type of
detailed spatial and temporal information that dLGN provides to cortex, but rather is wired to
provide an express route for information blended from diverse sources to coordinate and
modulate activity in downstream sensorimotor structures.
51
CHAPTER 3
SPATIAL AND TEMPORAL PROCESSING OF VISUAL SIGNALS IN THE MOUSE THALAMIC
RETICULAR NUCLEUS
3.1 Introduction
Thalamic sensory nuclei not only project to the cortex but also to the thalamic reticular
nucleus (TRN) (Pinault, 2004). The TRN is composed of gabaergic neurons that project back to
thalamocortical cells, providing feedback inhibition (Pinault, 2004). One segment of the TRN
connects reciprocally with the lateral geniculate nucleus (LGN) of the thalamus and receives
cortical input as well (Figure 3.1). Thus, the TRN is able to influence visual information
transmitted to the cortex. Previous studies revealed some insights about the organization of the
feedback provided by the visual TRN, as well as the visual response properties. The visual TRN to
LGN are topographically organized (Sanderson, 1971; Uhlrich et al., 1991; Fitzgibbon, 2002),
similar to corresponding regions of the auditory (Kimura et al., 2007) and somatosensory (Pinault,
2004; Lam and Sherman, 2007) thalamus. Reticular cells respond to binocular stimuli (Dubin and
Cleland, 1977; Ahlsen and Lindstrom, 1982; Uhlrich et al., 1991; Funke and Eysel, 1998;
Vaingankar et al., 2012) and are selective to complex features (Vaingankar et al., 2012).
How might these anatomical and physiological properties of the TRN contribute to visual
information flowing through the thalamocortical pathway? Two alternate hypotheses address
this question. The first one, referred as the “thermostat hypothesis”, is based on the notion that
reticular cells operate on a global level, hence inhibition from TRN should function as a gain
52
control mechanism throughout LGN (Figure 3.1) (Crick, 1984; Soto‐Sánchez et al., 2017). Second,
the competing “searchlight hypothesis” suggests that TRN provides top‐down attentional
influence, through local modulation of LGN activity. Small receptive fields in TRN would provide
a simpler explanation for local influence on LGN. A recent study using cat addressed this question
through quantifying the spatial scale of receptive fields in the TRN and LGN (Soto‐Sánchez et al.,
2017). The receptive field sizes at a given eccentricity in the TRN and LGN were similar, favoring
the searchlight hypothesis (Figure 3.2).
The studies described above have not taken an important property of the TRN. Like LGN,
cells in the TRN fire in two modes, burst and tonic (Huguenard and Prince, 1992). Bursts contain
several closely spaced sodium action potentials and ride on Ca
+2
currents mediated by T‐type Ca
+2
channels (Figure 3.8). This channel type closes shortly after it opens and must be strongly
hyperpolarized to be deinactivated. Hence bursts are fired during or following long periods of
hyperpolarization. In tonic mode, T‐type Ca
+2
channels are closed and cells fire conventional train
of sodium spikes (Figure 3.8). Studies in the LGN of the cat showed that the qualitative structure
of the receptive fields from burst and tonic spikes are similar (Alitto et al., 2005; Wang et al.,
2007). However, the temporal profiles of these receptive fields made from each type of spike are
clearly different (Alitto et al., 2005; Wang et al., 2007). Bursts are preceded by a stronger phase
of non‐preferred contrast before the stimulus phase pertaining to the preferred contrast in the
LGN. Then can visual stimuli influence the mode of firing in the TRN as well? A recent study using
cats suggests that this is the case (Vaingankar et al., 2010). However, the role of dual firing modes
at TRN in visual processing is yet to be determined. Unfortunately, further studies necessary to
53
dissect neural circuits for visual processing in TRN is experimentally challenging for previously
used animal models such as the cats.
Figure 3.1. The connections and hypothesized sensory roles of the TRN. (a) Block diagram showing the origin
and type of connections of the TRN and LGN (figure taken from (Hirsch et al., 2015)). (b) Illustrations of two
alternate hypotheses for the role of TRN in sensory processing: (left) the thermostat and (right) searchlight
hypotheses (figure taken from (Soto‐Sánchez et al., 2017), http://creativecommons.org/licenses/by/4.0/).
Boxes (shown with dashed outlines) show local circuits from adjacent regions of the visual space. The shape of
the type of connections are as shown in (a). The thermostat hypothesis suggests that the feedback from TRN
has an overall inhibitory effect on LGN, hence suppressing activity driven by the retina. This requires the
reticular receptive fields to be larger than the fields of relay cells. On the other hand, the searchlight hypothesis
suggest reticular input enhances local acidity of relay cells via rebound excitation. This predicts that the size of
receptive fields in TRN should be comparable to those in LGN, consistent with a role in top‐down attention.
TRN, thalamic reticular nucleus; LGN, lateral geniculate nucleus.
54
Towards the goal of establishing a more tractable and still useful model system to
investigate the visual role of TRN, we recorded neuronal responses in the TRN of mice during
visual stimulation. First, the receptive fields were mapped with noise stimuli and their sizes were
quantified with statistical approaches used earlier (Soto‐Sánchez et al., 2017). The diversity of
receptive field structure in the mouse TRN resembled that of the cat (Vaingankar et al., 2012). In
addition, a significant portion of the receptive fields in the TRN had spatial scales similar to that
LGN, as observed in cats (Soto‐Sánchez et al., 2017). Next, the bursts and tonic spikes of reticular
cells were detected and used to reconstruct respective spatiotemporal receptive fields. Similar
to cats, the receptive fields associated with dual modes of firing shared a common structure. But
the temporal profiles of these fields displayed significant differences, as also reported in cats
(Vaingankar et al., 2010). Overall, findings of this study suggest that mouse shares fundamental
aspects of sensory processing in the TRN and is a promising model system to further study the
visual roles of TRN.
3.2 Materials and Methods
Preparation
Mice (VGAT‐ChR2‐EYFP (Jackson Laboratories) or WT (C57BL/6)) were first sedated with
chlorprothixene (5 mg/kg); then anesthesia was induced and maintained with urethane (0.5–1
g/kg, 10% w/v in saline, IM) (Niell and Stryker, 2008). After the head was cleaned and shaved, an
incision was made to expose the skull so that a metal headpost could be affixed to hold the
animals in place. Next, a small craniotomy around TRN was made. All wound margins were
infiltrated with lidocaine, the brain and eyes were kept moist with saline, and body temperature
55
was maintained at 37°C. All procedures were approved by the Institutional Animal Care and Use
Committees of the University of Southern California following guidelines from the National
Institutes of Health.
Recordings
For multisite electrodes we used dual‐shank,32 channel, E‐2 acute probes, (Cambridge
NeuroTech, UK) connected with a 32 channel digital multiplexing headstage and a Digital Lynx
4SX – M acquisition system (Neuralynx, Bozeman, MT); filters were set to 0.1 Hz and 9 kHz and
the sampling rate was 30 kHz. Spike sorting from multisite recording was performed offline using
Kilosort (Pachitariu et al., 2016) and manual assessment of spike clusters.
Single cell site recordings were made with biocytin‐filled patch pipettes (Wang et al.,
2007) in whole‐cell or cell‐attached configurations using a Multiclamp 700B or Axopatch 200A
amplifier (Axon Instruments) and or digitized at 10 kHz using a Power 1401 data acquisition
system (Cambridge Electronic Design). Whole‐cell recordings were made in voltage‐clamp mode
to damp intrinsic conductances; the membrane potential was held slightly below the threshold
for firing to improve the visibility of synaptic currents. The placement of electrodes was guided
criteria developed by (Wimmer et al., 2015).
Optogenetic stimulation with was with blue light using an LED light source (Prizmatix,
Israel) or fiber‐optic light stimulation device (A‐M Systems, Sequim, WA); guided by a fiber optic
cable of 200µm diameter placed with a slight angle above the TRN (≥~1 mW power at the tip)
(see Figure 3.2).
56
Figure 3.2. Targeting and identification of visual reticular neurons in vivo. (a) Coronal section containing TRN,
from the Allen Mouse Brain Atlas. (b) Corresponding section displaying gene expression profile in VGAT‐ChR2‐
EYFP mice, from the Allen Mouse Brain Database. Inset was generated by linear transformation of the inset in
a. (c) Coronal section showing TRN cells retrogradely labeled following viral injection in LGN (figure taken from
(Wimmer et al., 2015)). (d) Inset in a shows TRN (cyan) and surrounding thalamic nuclei. (e) Sample spike train
induced by periodic blue light stimulation of a visual neuron in TRN of VGAT‐ChR2‐EYFP mice (top). A sample
burst‐firing pattern in TRN (bottom left). Plot of normalized ISI (interspike interval) distribution against
normalized ISI position (see (Vaingankar et al., 2012)) for the same cell (bottom right), derived from 846 bursts.
RT, thalamic reticular nucleus; LD, lateral dorsal thalamic nucleus; PO, thalamic posterior thalamic nucleus;
VPL, ventral posterolateral nucleus of the thalamus; VPM, ventral posteromedial nucleus of the thalamus.
57
Visual stimuli
Stimuli were generated using a ViSaGe (Cambridge Research Systems) stimulus generator
and displayed on gamma corrected Dell U2211H LCD monitor at a 70 Hz refresh rate and a
viewing distance of 180 mm. The stimulus set comprised dense noise, sparse noise and full‐field
stimuli. Two versions of dense noise were used, either a Gaussian noise stimulus (with the
luminance of each pixel drawn from a Gaussian distribution at 33% rms) or, if that stimulus was
ineffective, a binary version of the stimulus was substituted. The spatial resolution for dense
noise varied from 3.2° to 15.2°, the stimulus update rate was 29 or 57 ms, the stimulus sequences
ranged from 8000 to 16384 frames and were repeated 2 or 3 times. The sparse‐noise stimulus
consisted of, bright and dark squares, 5‐20°, shown at 50% contrast 16 times on a 16 x 16 grid (5°
grid resolution) (Jones and Palmer, 1987; Suresh et al., 2016).
Quantification of receptive fields and their sizes
We computed receptive fields from responses to dense noise using spike triggered
averaging (STA) to obtain a 3D spatiotemporal filter (Schwartz, 2006 #1246}; the STA was
computed from 0‐285ms segment of the stimulus preceding each spike. This 3D filter is displayed
as a spatial 2D map (shown in Figures 3.2 & 3.3), usually illustrating the time of the strongest
response.
The spatial structure of receptive fields in TRN was diverse so approaches that assume a
consistent shape (such as Gaussian functions) was not appropriate. Hence, a statistical approach
which assesses the significance of each voxel in the 3D spatiotemporal receptive field was used
(Soto‐Sánchez et al., 2017), similar to methods to analyze fMRI data (Genovese et al., 2002). First,
a bootstrapping approach (based on the original spike train) was used to generate z‐scores for
58
every voxel in the 3D spatiotemporal filter. Then, significant voxels were identified based on the
significance level (significance level was 0.001 in Figures 3.4‐3.6) using Bonferroni correction for
multiple comparisons. For display purposes (shown in Figures 3.2 & 3.3), the set of voxels in the
3D filter was reduced to a 2D spatial map referred to as “footprint”. The size of receptive fields
from the footprints were quantified using two metrics. The first was computed from the area of
the entire footprint (using pixels of both stimulus contrasts). The second was based on the area
of the footprint associated with a stimulus of the dominant contrast (as determined by that with
the highest significance). Finally, the square root of the footprint area was used to provide a 1D
measure (Figure 3.5).
Neuronal responses to sparse noise were quantified using one of two methods. To
analyze membrane currents we removed action potentials (medfilt2 function of MATLAB, The
Mathworks) (Wang et al., 2007), averaged the remaining inward or outward currents that each
spot evoked and then integrated the response over a manually selected time window (Martinez
et al., 2005; Suresh et al., 2016). This process yields separate 2D On and Off maps (shown in
Figure 3.7). Otherwise, for spikes, we computed the STA for each stimulus contrast (Schwartz et
al., 2006) and constructed 2D plots using the same protocol described for dense noise ( Figures
3.3 and 3.4).
Quantification of receptive fields associated with dual modes of firing and their temporal
profile
First, bursts and tonic spikes were detected in the spike trains from LGN and TRN. For
LGN, bursts were identified as two or more spikes, with a maximum of 4 ms between neighboring
spikes, that followed a silent period of at least 100 ms (Lu et al., 1992; Reinagel et al., 1999). For
59
TRN, bursts were defined as five or more spikes during a 70ms window, with a maximum of 30
ms between neighboring spikes, that followed a silent period of at least 70 ms (Domich et al.,
1986; Vaingankar et al., 2012). Then, for tonic spikes or the cardinal spike of each burst, we
computed the STA (from ‐514ms to 514 ms segment of the stimulus with respect to each spike
and displayed the maps for a 0‐257 ms segment of the stimulus preceding the spike, Figure 3.9.
The time course of the STAs are computed and displayed for the pixel with the strongest response
(sum of maxima for both contrasts within 0‐271 ms segment of the stimulus preceding the spike)
(Figure 3.9). The temporal profiles of receptive fields generally had a clear biphasic pattern
(Figure 3.10). The two phases of these waveforms are referred to as triggering and priming
phases. To describe each phase quantitatively, the temporal receptive fields are fitted with a
biphasic waveform composed of 2 triangular pulses (Figure 3.10d). A total of 7 parameters are
used to optimize the fit; 5 were associated with temporal information about the triangle points
and 2 for the height of these triangles. Two triangle pulses are linked; the time that point one
pulse finishes the same as the timepoint the second pulse begins. The fitting parameters were
computed using fminsearch function of MATLAB (The Mathworks, Natick, MA) to minimize the
squared error between two waveforms. To improve the chances of better qualitative fits, we
added constraints to the error function. The temporal parameters had to be separated by at
least the resolution of time signal, the order of temporal parameters was fixed, and the triangular
pulses were forced to have an opposite sign to ensure a biphasic waveform. After the
optimization procedure, only the triggering and priming phases with qualitatively good fits were
included in the population statistics. The features that are used to quantify the temporal
receptive fields are illustrated in Figure 3.10d. In rare cases for which the fit of a triangular pulse
60
extended beyond the t=0 line, the duration of that phase was assigned as the time from t=0 to
the end of the phase, and the magnitude of the phase was assigned the entire area of the triangle.
These rare cases had negligible effect on the population statistics.
3.3 Results
The reticular neurons (n=87) we used for analysis fired bursts with the statistical profile
unique to TRN (Vaingankar et al., 2012), responded to visual stimuli, and, for recordings from
vGAT mice, responded reliably with short latency bursts to pulses of blue light delivered near the
recording site. The dLGN dataset (n=57) was recorded for an earlier study (Suresh et al., 2016).
Comparison of receptive fields in the TRN and LGN
To explore the potential influence of the TRN on the LGN, we estimated receptive fields
in the TRN and compared these to a sample obtained from the dLGN. We focused on the
question of whether or not the receptive fields were spatially localized, as in cat and monkey, or
diffuse. Further, we were able to record intracellularly from a subset of cells to visualize the
synaptic basis of the receptive field.
Receptive field structure in the TRN
Receptive field structure in the dLGN has been quantified in many species (Shapley and
Lennie, 1985) (Piscopo et al., 2013). By contrast, visual receptive fields in the visual TRN have
been studied quantitatively only recently and only in cat (Vaingankar et al., 2012; Soto‐Sánchez
et al., 2017). While their shapes were diverse and irregular, their sizes scaled with position in the
visual field, small near the center and large in the periphery (Vaingankar et al., 2012; Soto‐
61
Sánchez et al., 2017); thus neurons in the TRN operate over local spatial scales. Qualitative
results in monkey support are consistent with this result (McAlonan et al., 2006).
In rodent, receptive field size does not increase with distance from the center of the visual
field; small and large fields are, largely scattered throughout visual space. Thus, we wondered if
there was a commensurate distribution of receptive field sizes in TRN.
We used the same stimuli (sparse and dense noise) and computational tools employed
for our earlier study of cat to approach this question. Most cells responded to both bright and
dark stimuli, though the strength and spatiotemporal structure of On and Off subregions varied
greatly among cells, as shown in the maps made using sparse noise, (Figures 3.4). For dense
noise, the polarity of each pixel in the receptive field maps is determined by relative weight of
On and Off responses. If these are equal, they cancel but if the response to one stimulus contrast
is greater than the other, then the stronger response alone is reported. Reponses to dense noise
show that most On‐Off cells in the TRN had a strong preference for a single stimulus polarity, eve
as the shape of the maps varied greatly in size and shape.
Comparison of receptive field scale in TRN and LGN
We estimated the sizes of receptive fields in the TRN by computing “footprints” made
using responses to dense noise (Soto‐Sánchez et al., 2017) (Figures 3.3 & 3.4) and compared
these results with a companion dataset from the dLGN. We made two sets of footprints. One
used stixels of both stimulus contrasts (combined footprint, Figure 3.5a). The other was made
only with the subset of stixels corresponding the preferred stimulus polarity (dominant footprint,
Figure 3.5b); this allowed us to compare the size of the fields in the TRN with the size of the
center of receptive fields in the dLGN that had a center surround structure. The receptive fields
62
were as small as 5‐10° for both structures (Figure 3.5), however there was a disparity in footprint
size at other extreme. The largest footprints were about 25° in LGN but reached ~80° (Figure 3.5)
in TRN. The 25
th
, 50
th
and 75
th
percentiles for the scale of combined maps were 9.2°, 13.1°, and
18.6° (n=57) for LGN and and 18.6°, 24.7°, and 37.8° (n=86) TRN. For dominant footprints, the
corresponding ranges were7.6°, 10.7°, and 15.2° (n=57) for LGN and 17.0°, ‐20.8°, and 32.2°
(n=86)for TRN.
Even though some receptive fields in TRN were very large, a substantial population of
reticular cells had footprints that overlapped those from LGN. In fact, 50% (n=43 of 86) of
combined footprints and 60% (n=52 of 86) of dominant footprints were smaller than those in top
(100
th)
percentile.
63
Figure 3.3. Receptive field maps and sizes in the cat LGN and TRN. (a) Sample spike triggered averages of the
stimulus ensemble (STA) made from responses to a sparse noise stimulus for 3 cells in the cat LGN. From left
to right, (1st and 2nd columns) contour plots made from responses to bright (red, ON) or dark (blue, OFF)
stimuli; each shaded rectangle contains maps for a single cell. Grid spacing was 1° and stimulus size is indicated
by yellow squares. (3rd column) STAs obtained using dense noise and (4th column) associated footprints made
from statistically significant responses to dense noise. Significant pixels of the receptive field are calculated
using a false discovery rate test (Benjamini and Hochberg, 1995; Soto‐Sánchez et al., 2017). (b) Sample set of
receptive field maps for 3 cells in cat TRN (cyan), mapped as for a. The sample cells were ordered by the size
of their receptive fields (size increased from top to bottom).The datasets for a and b are from (Soto‐Sánchez
et al., 2017). (c) Comparison of receptive field sizes in the TRN and LGN across eccentricities, for the combined
(i.e. entire field) (left) and dominant contrasts (right) (figure taken from (Soto‐Sánchez et al., 2017)). PGN,
perigeniculate nucleus (visual sector of TRN).
64
Figure 3.4. Receptive field maps in the mouse LGN and TRN. (a) Sample spike triggered averages of the stimulus
ensemble (STA) made from responses to a sparse noise stimulus for 6 cells in the mouse LGN. From left to
right, (1st and 2nd columns) contour plots made from responses to bright (red, ON) or dark (blue, OFF) stimuli;
each shaded rectangle contains maps for a single cell. Stimulus size is indicated by yellow squares. (3rd column)
STAs obtained using dense noise and (4th column) associated footprints made from statistically significant
responses to dense noise. Significant pixels of the receptive field are calculated using a false discovery rate test
(Benjamini and Hochberg, 1995; Soto‐Sánchez et al., 2017). (b) Sample set of receptive field maps for 6 cells
in mouse TRN (cyan), mapped as for a. The grid spacing ranged from 5° to 15.2°. The sample cells were ordered
by the size of their receptive fields (size increased from top to bottom).
65
Figure 3.5. Receptive field sizes in the mouse LGN and TRN. (a) Scatter plot of footprint sizes for cells in mouse
LGN (green) and TRN (cyan), from the entire footprint (combined contrasts). (b) Histograms for the sizes of the
entire footprint from the mouse LGN (green) (top) and TRN (cyan) (bottom). (c) Scatter plot of the size of the
footprint of dominant contrast for cells in mouse LGN (green) and TRN (cyan). (d) Histograms for the sizes of
the footprint of dominant contrast from the mouse LGN (green) (top) and TRN (cyan) (bottom). Size of the
entire footprint and the footprint of dominant contrast are calculated as the square root of the area of
corresponding footprints.
66
Receptive field scale across visual space
Given that receptive field size in the rodent retina does not scale with position in the
visual field, it was not surprising that we found no obvious relationship between the receptive
field size and location in visual space in the murine TRN (Figure 3.6). Figure 3.6 a&c illustrates
the scale (color coded) of each TRN receptive field made using the combined or dominant
footprints plotted as a function of its position in the visual field; the location of each point was
determined by the peak of the receptive field (i.e. the maximal STA coefficient). Separate plots
show the relationship between receptive field size and azimuth (Figure 3.6b&d) and elevation
(Figure 3.6c&f). To quantify the relationship between receptive field size and position in the
visual field, we calculated the Spearman’s rank correlation coefficient. The correlation values for
the combined footprint were ‐0.12 (p=0.28) for azimuth and ‐0.24 (p=0.03) for elevation. For the
dominant footprint the numbers were ‐0.09 (p=0.45), azimuth and ‐0.21 (p=0.06). It is possible
the values for elevation were smaller than those for azimuth because there was greater distance
from the eye to display monitor in that dimension.
67
Figure 3.6. Receptive field sizes across visual space in the mouse TRN. (a) Scatter plot of the sizes for the entire
footprint (combined contrasts) with azimuth (top) and elevation (bottom). (b) Scatter plot of the sizes for the
entire footprint (combined contrasts) (color coded with size) with azimuth and elevation. (c) Scatter plot of the
sizes for the footprint of dominant contrast with azimuth (top) and elevation (bottom). (d) Scatter plot of the
sizes for the footprint of dominant contrast (color coded with size) with azimuth and elevation. Size of the
entire footprint and the footprint of dominant contrast are calculated as the square root of the area of
corresponding footprints. For display purposes, the position in visual space is slightly jittered (≤2°, if needed)
to prevent the overlay of samples that had the same coordinate.
68
Synaptic physiology in the TRN
Morphological studies suggest input from LGN drives activity in the TRN and that cortical
feedback in largely modulatory. Past work in cat (Xue et al., 1988; Vaingankar et al., 2012)
supports this view. We wished to explore construction of fields in the TRN from a synaptic
perspective. For seven cells, we were able to record membrane currents during visual
stimulation, and for two of these cells were about to reconstruct receptive fields from membrane
currents. Figure 3.7 shows sample membrane currents evoked by sparse noise for a cell that
preferred dark stimuli. We are able to resolve unitary excitatory post synaptic currents (EPSCs)
in our records, consistent with the possibility that input from individual relay cells is strong. These
preliminary results also indicate that visual input is able to trigger the intrinsic currents that drive
bursts (see Figure 3.7d).
Figure 3.7. Membrane currents and associated receptive field in the TRN. (a) Sample ON (red) and OFF (blue)
receptive field of a cell in TRN computed using the spikes evoked by sparse noise stimuli; stimulus size is
indicated by yellow squares. (b) Receptive field maps of the same cell computed using the membrane currents
evoked by the same stimuli and displayed as in a. (c) Average membrane currents across time in response to
sparse noise stimuli displayed for every grid location. Red and blue traces represent responses to bright and
dark squares respectively. (d) 3 individual traces and cross‐trial average (bolded and displayed at 2X gain) for
a dark square centered at the inset shown in C; gray bar shows stimulus duration. EPSCs (Excitatory
Postsynaptic Currents) are visible in these traces (a sample shown with an asterisks), along with putative T
currents driving bursts (a sample shown with double headed horizontal arrow).
69
Comparison of receptive fields of bursts and tonic spikes in the TRN and LGN
Neurons in both TRN and LGN fire in two modes, burst and tonic (Figure 3.8). Bursts are
produced by T‐type calcium channels that close shortly after they open and must be strongly
hyperpolarized before they are able to reopen, thus bursts occur after or during long periods of
hyperpolarization (Figure 3.8). Previous studies show that visual stimuli can alter the mode of
firing in the LGN of cats (Alitto et al., 2005; Wang et al., 2007) (Figure 3.8). For relay cells in LGN,
the temporal profile of STAs reconstructed from both burst and tonic spikes are biphasic; the first
phase corresponds to the stimulus that triggers spikes and the second phase corresponds to the
stimulus of the non‐dominant contrast and primes firing. The relative amplitude of each phase
differs for tonic and burst firing, however for tonic spikes, the “triggering” phase is stronger that
the “priming” phase; for burst spikes the reverse is true. A similar pattern has been observed in
the TRN (Vaingankar et al., 2010). Here, we investigated whether the receptive fields associated
with burst and tonic spikes in mouse have properties similar to those observed in cat LGN and
TRN. Thus, we computed the spatiotemporal receptive fields for tonic and burst spikes evoked
by dense noise (Figure 3.9) for cells in the LGN and TRN. At first glance, the spatiotemporal
receptive fields made from bursts and tonic spikes look alike; the spatial positions of the stimulus
features that trigger or prime spikes are similar. However, as for cat, the relative strength and
duration of these features differed according to firing mode, as illustrated for three cells from
each structure (Figure 3.9). To visualize this result, the temporal profile of the strongest pixel
(see Methods) is plotted to the right of the spatiotemporal receptive field (displayed as a series
of 2D maps). The waveforms from burst and tonic STAS are superimposed to highlight the
differences in the strength and duration of the priming and triggering features.
70
Figure 3.8. Dual modes of firing in the thalamus and associated temporal receptive fields. (a) Voltage
dependent firing in the relay cells of the cat LGN (figure taken from (Sherman, 2001)). (top) When the cell is in
a relatively depolarized state, stimulation with a current pulse leads to a unitary stream of action potentials
also called as tonic spikes. (bottom) However, when the cell is in a relatively hyperpolarized state, stimulation
with a current pulse leads to a bundle of closely spaced action potentials called a burst, on top of a smoother
low threshold spike. (b) Temporal profile of spike triggered averages from 3 cells (each row) in the cat LGN
associated with bursts, tonic and long inter‐spike interval tonic spikes. The temporal profiles reflect the
differences in the stimuli evoking different modes of firing (figure taken from (Alitto et al., 2005)).
71
Figure 3.9. Receptive fields associated with burst and tonic firing in the mouse LGN and TRN. (a) Sample
spatiotemporal receptive fields computed from bursts and tonic spikes for 3 cells in the mouse LGN. Contour
plots show STAs made from dense noise displayed at consecutive points in time. The time course of the STA
for the pixel with the strongest response (sum of maxima for both contrasts) is plotted (right) for bursts
(magenta) and tonic spikes (black). Each shaded green rectangle contains the time series of STAs for bursts
(magenta) and tonic spikes (black). (b) Sample spatiotemporal receptive fields for 3 cells in the mouse TRN
(cyan), computed and displayed as for a; grid spacing ranged from 5.4° to 15.2°.
72
The priming phase for bursts seemed stronger that for tonic spikes. In addition, the peaks
of both phases for burst spikes seemed to have shorter latencies compared to the STAs made for
tonic spikes. To quantify these trends, we first fitted individual temporal receptive fields with a
biphasic function composed of two triangular pulses (see Methods). The population statistics
matched the qualitative observations about the time course of the receptive fields (Figure 3.11);
the following values are given as mean ± SD and statistical comparisons were calculated with
Wilcoxon signed‐rank test. The magnitude of the priming phases for bursts were higher than
those for tonic spikes in both LGN (burst, 53.9±22.7; tonic, 29.7±8.7, p=9.8x10
‐4
(n=12)) and TRN
(burst, 52.2 ± 21.7; tonic, 33.7 ± 19.0, p=4.0x10
‐4
(n=25)). Consistent with this result, the ratio of
the triggering to the priming peak was smaller for burst receptive fields both in LGN (burst, 1.51
± 0.76; tonic, 2.93 ± 1.63, p=0.0093 (n=12)) and TRN (burst, 1.74 ± 1.20; tonic, 2.81 ± 1.76,
p=0.0018 (n=24)). Further, the latency for the priming phase was shorter for burst than tonic
spikes, in both LGN (burst, 102.3 ± 23.7 ms; tonic, 114.8 ± 14.7 ms, p=0.0015 (n=12)) and TRN
(124.2 ± 21.1 ms; tonic, 150.4 ± 27.0 ms, p=3.6x10
‐5
(n=25)). The results for the latency of the
peak of the triggering phase was similar in LGN (burst, 62.0 ± 13.2 ms; tonic, 67.2 ± 8.9 ms,
p=0.0086 (n=17)) and TRN (burst, 61.9 ± 18.9 ms; yonic, 84.3 ± 21.3 ms, p=5.2x10
‐9
(n=45)) TRN).
73
Figure 3.10. Temporal receptive fields in the mouse LGN and TRN and their analysis. (a) Mean ± Standard Error
plot of temporal receptive fields in the LGN for bursts (magenta) and tonic spikes (black). (b) Mean ± Standard
Error plot of temporal receptive fields in the TRN for bursts (magenta) and tonic spikes (black). (c) Schematic
depiction of the triggering and priming phases of the temporal receptive field (figure adapted from (Alitto et
al., 2005)). (d) Fitting of temporal receptive fields by a biphasic function composed of two triangular pulses.
Following the fitting procedure, the temporal receptive fields are quantified using the metrics shown in the
plot by arrows.
74
Figure 3.11. Population statistics of temporal receptive fields of bursts and tonic spikes in the mouse LGN and
TRN. (a) The scatter plots of the latencies for peaks of the triggering and priming phases for bursts vs tonic
spikes in the LGN (left) and the TRN (right). (b) (left) Scatter plot of the duration of the triggering phase for
bursts vs tonic spikes in the LGN and TRN and (right) scatter plot of the magnitude of the triggering phase. (c)
Scatter plots for the duration and magnitude of the priming phase, displayed as in (b). (d) Scatter plot for the
ratio of the peak amplitudes of the triggering vs priming phases for bursts vs tonic spikes in the LGN and TRN.
75
3.4 Discussion
In this study, we investigated the role of the visual sector of the murine TRN in sensory
processing. First, we focused on the structure and size of receptive fields in visual space, in
comparison to LGN and to previous work in cat. In cat, we found that reticular cells were selective
for specific visual patterns and operated over local spatial scales. Our results in mouse were
similar, save that we found a population of cells in TRN with receptive fields far larger than those
in LGN. For some cells, we were also able to map receptive fields from membrane currents as
well as spikes. These recordings showed that some inputs to reticular cells are strong, consistent
with the suggestion from ultrastructural studies (Montero and Singer, 1984; Cucchiaro et al.,
1991; Bickford et al., 2008) that relay cells provide powerful input to TRN. Towards a fuller
understanding of how the TRN might influence its targets in LGN, we explored the role of dual
modes of firing (bursts and tonic spikes) during sensory stimulation in both structures. Past
studies of relay cells have shown that visual stimulation can toggle the membrane between tonic
and burst firing via interactions between synaptic inhibition and intrinsic membrane properties.
While some of the inhibition that promotes burst in LGN must come from local interneurons, we
wondered if the TRN might contribute as well. Thus, we explored the possibility of visually
evoked bursting in the TRN, using methods developed for LGN as a starting point. Studies of the
cat’s LGN show that the features that precede bursts divide into two parts, a priming and
triggering phase. Like cat, our results in mouse showed that the features that prime and trigger
bursts are roughly similar in shape but different in time. That is, the priming phases associated
with bursts are stronger and latencies of both phases were shorter for burst versus tonic spikes.
The inhibition provided by visually‐evoked bursts in TRN might work in concert with that provided
76
by local interneurons to promote bursts in LGN. Taken together, our results suggest that cells in
the TRN contribute feature selective inhibition over a broad range of spatial and temporal scales
to influence the signals relay cells send to cortex.
Receptive fields in TRN have diverse spatial arrangements
Receptive field structures in mouse TRN were diverse in size, shape, and preference for
stimulus contrast. This variability might, in part, reflect the fact that there are many types of relay
cells in murine dLGN (Piscopo et al., 2013; Suresh et al., 2016). However, nearly half of relay cells
in dLGN have center‐surround receptive fields (Piscopo et al., 2013; Suresh et al., 2016) whereas
we mapped few, if any reticular receptive fields of this type. This finding is reminiscent of cat,
where almost all relay cells in the LGN have receptive fields with a center‐surround structure
while most cells in the TRN are On‐Off, On or Off. Thus, it seems that receptive fields in both
rodent and carnivore TRN are formed by convergent input from different types of relay cells
(Vaingankar et al., 2012). While the ecological significance of the features encoded in TRN
remains unexplained, the complex receptive fields a recorded in the nucleus make it is clear that
it provides stimulus selective feedback to relay cells.
Relative spatial scale of receptive fields in TRN and dLGN
The TRN is variously hypothesized to play a role in global gain control or spatial attention
(Crick, 1984). A global action predicts large receptive fields while a role in spatial attention
predicts localized ones. Because the receptive fields in the TRN are oddly shaped, there sizes
cannot be quantified using standard fitting functions. Thus, in previous work, we devised a
statistical method to measure the size of receptive (Soto‐Sánchez et al., 2017) and found that
neurons in the cat’s dLGN and TRN, both, operate over local spatial scales. The receptive fields
77
near the center of the gaze were small and became larger with eccentricity; this relationship
between receptive field size and position within the visual field is common in the early visual
pathway of highly visual mammals like carnivores and primates. For rodent, however, cells with
small and large receptive fields are intermixed across retina and LGN and we found this to be the
case for cells in TRN as well. Thus we compared the distributions of receptive field sizes in rodent
LGN and TRN independent of position in visual space. Many cells in the mouse TRN had receptive
fields similar in size to those in LGN, comparable to the situation in cat and consistent with the
searchlight hypothesis. However, a substantial population of reticular cells had receptive fields
far larger than those in LGN. These might influence LGN on a coarse scale, as thermostat
hypothesis predicts. Thus, it is possible that subsets of neurons in the murine TRN exert a local
influence on LGN while others provide global modulation. Overall, the spectrum of responses in
TRN is consistent with local feature processing as in the case of searchlight hypothesis, but also
provides the flexibility to modulate LGN on a global scale suggested by the thermostat
hypothesis.
There are caveats to our interpretations, however. First, while it is likely that most of our
recordings were made from reticular cells that receive input from, and project to, the LGN, there
remains a possibility that we recorded from some cells connected to the lateral posterior nucleus,
the murine homolog of the pulvinar. We imagine that these reticular cells would have large
receptive fields, as this is the case for cells in the murine pulvinar. Also, our recordings were made
in anesthetized animals, where receptive fields can be larger than in awake animals (Worgotter
et al., 1998; Alitto et al., 2011; Sriram et al., 2016; Tschetter et al., 2018).
78
Synaptic inputs to TRN
With the goal of exploring how receptive fields in TRN are built, we made whole‐cell
recordings there. We were able to resolve individual EPSCs in the membrane currents. It seems
likely that these events derived from the LGN. First, relay cells contact the proximal dendrites of
reticular cells (Montero and Singer, 1984; Cucchiaro et al., 1991; Bickford et al., 2008) while
cortex innervates distal sites (Ide, 1982; Cucchiaro et al., 1991; Bickford et al., 2008). Further,
cortical inputs are typically too small to resolve. Second, past work has shown that removing
cortex does not change the shape of receptive fields in TRN (Sanderson, 1971; Xue et al., 1988;
Jones and Sillito, 1994). Indeed, we found that receptive fields computed from synaptic currents
resembled those made from spikes. Assuming these currents mainly derive from feedforward
input, it seems likely that cortex and other sources of input are likely to modulate rather than
drive activity in TRN.
Visually evoked bursts in TRN
Neurons in the LGN and TRN fire in two modes, tonic trains or rapid bursts of spikes. Past
work has in LGN shown that visual stimuli that evoke strong inhibition in relay cells can prime the
channels that trigger bursts, such that when an excitatory stimulus arrives, bursts fire (Wang et
al., 2007). Thus, bursts carry information about the stimulus (Reinagel et al., 1999; Lesica and
Stanley, 2004; Alitto et al., 2005; Wang et al., 2007). In addition, when relay cells burst, they are
better able to evoke firing in postsynaptic targets than when in tonic mode. For this reason,
bursts have been called the wake up call to cortex (Sherman, 2001; Swadlow and Gusev, 2001;
Usrey and Alitto, 2015). It seems certain that local interneurons in the LGN contribute to the
inhibition that primes bursts (Wang et al., 2010b; Suresh et al., 2016). We asked if TRN might
79
also help to prime bursts in the LGN. Just as bursts fired by relay cells drive stronger excitation
in their target than tonic trains of action potentials, bursts produced by reticular cells yield
stronger inhibition in relay cells than tonic spikes do (Kim et al., 1997). This stronger inhibition
can also result in suppressing LGN activity while priming bursts. Further, recent genetic studies
suggest that the sensory areas of TRN have higher densities of the channels that promote bursts
than found in extrasensory areas (Fernandez et al., 2018).
As is the case for LGN, we found that visual stimuli evoked bursts in TRN. The visual
features that precede firing (whether in burst or tonic mode) comprised two phases, an earlier
priming phase and a later triggering one. We found that the visual features that evoked both
tonic and burst spikes were similar in overall shape and position, but that the priming phase was
stronger for burst spikes in both LGN and TRN. The latencies to peak for both phases was also
shorter for burst vs tonic spikes. These results are similar to observations in cat (Alitto et al.,
2005). Hence, the TRN not only able to modulate spike rate in the LGN, but to alter spike timing
as well (See Figure 3.12 for an illustration).
80
Figure 3.12. Influence of TRN firing mode on the thalamocortical pathway. (a) In paired recordings in vitro,
bursts in TRN can induce significant hyperpolarization to trigger a rebound T type calcium current associated
with bursts in a relay cell of the LGN. However, tonic spikes in TRN provided weaker inhibition not strong
enough to prime a rebound burst (figure taken from (Kim et al., 1997)). (b) Illustration of a possible scenario
on how bursts in TRN may help promote bursts in LGN in vivo. (top) Membrane currents from a sample cell in
TRN during visual stimulation. (bottom) Membrane currents from a sample cell in LGN during visual
stimulation. Dashed vertical line represents the approximate time of the burst in the TRN recording and
horizontal lines represent the approximate holding level for the membrane voltage. A burst in a TRN cell may
provide strong inhibition to prime a burst in a relay cell during visual stimulation.
81
Comparison with other species and sensory systems
While we focused on the visual sector of the TRN, we believe that our work will
generalize to other sensory divisions of the nucleus. Studies in the auditory and somatosensory
systems indicate that reticular neurons there are selective for specific stimulus features (Simm
et al., 1990; Hartings et al., 2000; Cotillon‐Williams et al., 2008). These results like apply to a
role in the TRN for in forms of attention and arousal beyond visual spatial attentions (Wimmer
et al., 2015). Furthermore, given how greatly the murine visual system differs from that of
highly visual mammals, it was particularly interesting to find commonality of function in TRN
across species.
82
CHAPTER 4
CONCLUSION
There are two main pathways in the early visual system. One of them goes through
thalamus to the visual cortex and is associated with form vision. The other one is mainly
subcortical and organized around the superior colliculus (SC). Within thalamus, lateral geniculate
nucleus (LGN) is the station that is the main source of visual information to the cortex.
Historically, LGN was thought to passively relay visual information but recent studies challenged
this idea. In addition, LGN is composed of subdivisions with remarkably different connectivity.
Contrary to the well‐known dorsal division (dLGN) that projects to the cortex, the ventral division
(vLGN) does not project to the cortex. Indeed, it projects to several subcortical structures with
motor functions, including the SC. The visual thalamus not only includes the LGN but also a
portion of the thalamic reticular nucleus (TRN) which reciprocally connects with LGN. TRN is
made of inhibitory neurons and generally studied in the context of attention and sleep. In this
thesis, I presented work on investigating the role of vLGN and TRN in terms of sensory processing,
towards the goal of achieving a fuller understanding the visual functions of the thalamus.
First, I started with exploring the sensory physiology in vLGN, in comparison to dLGN and
SC. In the vLGN, the receptive fields were considerably larger and the temporal precision were
lower than the dLGN and SC. This suggests that inhibition from the vLGN to its downstream
targets modulates activity at coarse spatial and temporal scales. Commensurate differences are
also present in the dendritic morphology and synaptic physiology in the cells of vLGN. Apart from
83
the neural responses in the form of conventional firing rate based coding scheme, visual
oscillations similar to SC are present in the vLGN and these oscillations can have a role in natural
vision. Hence oscillations in vLGN might help synchronize activity in distributed visuomotor
networks. Taken together, this work provides insights on the sensory processing in the vLGN and
how it may influence downstream targets.
In the remaining part of this dissertation, I focused on the sensory physiology of TRN and
its function in vision. The structure of the receptive fields in the TRN were diverse and the spatial
scale of a significant portion of the receptive fields was comparable to LGN, providing a means
for localized reticular inhibition on LGN. Consistent with a role in spatial attention and local
feature processing, TRN operates on a local scale across species. Also, there were cells that had
larger receptive fields with the potential of exerting global influence on LGN. Then, I explored the
role of firing modes during vision in TRN, which fire bursts and tonic spikes similar to LGN. There
were clear differences in the temporal profile of the receptive fields from bursts versus tonic
spikes. The role of bursts in LGN has been proposed to provide a “wake up signal” to the cortex
to indicate sudden changes in the visual stimulus. Since reticular neurons inhibit relay cells, bursts
in the TRN may induce a stronger inhibitory influence on relay cells and hence help promote
bursts in the LGN. Hence bursts in TRN can ultimately influence the temporal pattern of input the
cortex receives.
Taken together, different inhibitory structures in visual thalamus can convey visual
information at either local or global scales. Further studies that perturb neural activity (such as
optogenetic stimulation) is needed to investigate the role of these thalamic structures in visual
84
functions using behaving awake models. And any attentional or visual dysfunction can be
explored during inactivation of bursts to investigate their role in vision.
85
REFERENCES
A Whittington M, Traub R (2004) Inhibitory interneurons and network oscillations in vitro.
Trends in neurosciences 26:676‐682.
Ahlsen G, Lindstrom S (1982) Excitation of perigeniculate neurones via axon collaterals of
principal cells. Brain Res 236:477‐481.
Akaike H (1974) A new look at the statistical model identification. IEEE Transactions on
Automatic Control 19:716‐723.
Alitto HJ, Weyand TG, Usrey WM (2005) Distinct properties of stimulus‐evoked bursts in the
lateral geniculate nucleus. J Neurosci 25:514‐523.
Alitto HJ, Moore BD, Rathbun DL, Usrey WM (2011) A comparison of visual responses in the
lateral geniculate nucleus of alert and anaesthetized macaque monkeys. The Journal of
Physiology 589:87‐99.
Babb RS (1980) The pregeniculate nucleus of the monkey (Macaca mulatta). I. A study at the
light microscopy level. J Comp Neurol 190:651‐672.
Benjamini Y, Hochberg Y (1995) Controlling the False Discovery Rate: A Practical and Powerful
Approach to Multiple Testing. Journal of the Royal Statistical Society Series B
(Methodological) 57:289‐300.
Bickford M, E (2018) Synaptic organization of the dorsal lateral geniculate nucleus. Eur J
Neurosci 49:938‐947.
Bickford ME, Wei H, Eisenback MA, Chomsung RD, Slusarczyk AS, Dankowsi AB (2008) Synaptic
organization of thalamocortical axon collaterals in the perigeniculate nucleus and dorsal
lateral geniculate nucleus. J Comp Neurol 508:264‐285.
Bloomfield SA, Hamos JE, Sherman SM (1987) Passive cable properties and morphological
correlates of neurones in the lateral geniculate nucleus of the cat. J Physiol 383:653‐692.
Brecht M, Goebel R, Singer W, Engel AK (2001) Synchronization of visual responses in the
superior colliculus of awake cats. Neuroreport 12:43‐47.
Brecht M, Schneider M, Sakmann B, Margrie TW (2004) Whisker movements evoked by
stimulation of single pyramidal cells in rat motor cortex. Nature 427:704‐710.
Brenner N, Agam O, Bialek W, de Ruter van Steveninck R (2002) Statistical properties of spike
trains: Universal and stimulus dependent aspects. Physical Review E 66, 031907.
Brenner N, Strong SP, Koberle R, Bialek W, de Ruyter van Steveninck RR (2000) Synergy in a
neural code. Neural Computation 12:1531‐1552.
86
Burke W, Sefton AJ (1966) Inhibitory mechanisms in lateral geniculate nucleus of rat. J Physiol
187:231‐246.
Cang J, Feldheim DA (2014) Developmental mechanisms of topographic map formation and
alignment. Annual Review Neuroscience 36:51‐77.
Cang J, Savier E, Barchini J, Liu X (2018) Visual Function,organization, and development of the
mousesuperior colliculus. Annual Review of Vision Science 4.
Chen SK, Badea TC, Hattar S (2011) Photoentrainment and pupillary light reflex are mediated by
distinct populations of ipRGCs. Nature 476:92.
Cleland BG, Dubin MW, Levick WR (1971) Simultaneous recording of input and output of lateral
geniculate neurones. Nature ‐ New Biology 231:191‐192.
Contreras D, Curro Dossi R, Steriade M (1993) Electrophysiological properties of cat reticular
thalamic neurones in vivo. J Physiol 470:273‐294.
Cotillon‐Williams N, Huetz C, Hennevin E, Edeline J‐M (2008) Tonotopic control of auditory
thalamus frequency tuning by reticular thalamic neurons. J Neurophysiol 99:1137‐1151.
Crick F (1984) Function of the thalamic reticular complex: the searchlight hypothesis. Proc Natl
Acad Sci 81:4586‐4590.
Cucchiaro JB, Uhlrich DJ, Sherman SM (1991) Electron‐microscopic analysis of synaptic input
from the perigeniculate nucleus to the A‐laminae of the lateral geniculate nucleus in
cats. J Comp Neurol 310:316‐336.
de Haan EHF, Cowey A (2011) On the usefulness of ‘what’ and ‘where’ pathways in vision.
Trends in Cognitive Sciences 15:460‐466.
de Ruyter van Steveninck RR, Lewen GD, Strong SP, Koberle R, Bialek W (1997) Reproducibility
and variability in neural spike trains. Science 275:1805‐1808.
Destexhe A, Bal T, McCormick DA, Sejnowski TJ (1996) Ionic mechanisms underlying
synchronized oscillations and propagating waves in a model of ferret thalamic slices. J
Neurophysiol 76:2049‐2070.
Dhande OS, Hua EW, Guh E, Yeh J, Bhatt S, Zhang Y, Ruthazer ES, Feller MB, Crair MC (2011)
Development of single retinofugal axon arbors in normal and beta2 knock‐out mice. J
Neurosci 31:3384‐3399.
Domich L, Oakson G, Steriade M (1986) Thalamic burst patterns in the naturally sleeping cat: a
comparison between cortically projecting and reticularis neurones. J Physiol 379:429‐
449.
Dubin MW, Cleland BG (1977) Organization of visual inputs to interneurons of lateral geniculate
nucleus of the cat. J Neurophysiol 40:410‐427.
87
Ecker JL, Dumitrescu ON, Wong KY, Alam NM, Chen SK, LeGates T, Renna JM, Prusky GT, Berson
DM, Hattar S (2010) Melanopsin‐expressing retinal ganglion‐cell photoreceptors: cellular
diversity and role in pattern vision. Neuron 67.
Ellis EM, Gauvain G, Sivyer B, Murphy GJ (2016) Shared and distinct retinal input to the mouse
superior colliculus and dorsal lateral geniculate nucleus. J Neurophysiol 116:602‐610.
Fatt P, Katz B (1951) An analysis of the end‐plate potential recorded with an intracellular
electrode. J Physiol 115:320‐370.
Fernandez LMJ, Vantomme G, Osorio‐Forero A, Cardis R, Béard E, Lüthi A (2018) Thalamic
reticular control of local sleep in mouse sensory cortex. eLife 7:e39111.
Fitzgibbon T (2002) Organization of reciprocal connections between the perigeniculate nucleus
and dorsal lateral geniculate nucleus in the cat: a transneuronal transport study. Vis
Neurosci, 19:511‐520.
Funke K, Eysel UT (1998) Inverse correlation of firing patterns of single topographically matched
perigeniculate neurons and cat dorsal lateral geniculate relay cells. Vis Neurosci, 15:711‐
729.
Gabbott PL, Bacon SJ (1994) An oriented framework of neuronal processes in the ventral lateral
geniculate nucleus of the rat demonstrated by NADPH diaphorase histochemistry and
GABA immunocytochemistry. Neuroscience 60:417‐440.
Gale SD, Murphy GJ (2014) Distinct representation and distribution of visual information by
specific cell types in mouse superficial superior colliculus. The Journal of Neuroscience
34:13458‐13471.
Gandhi NJ, Katnani HA (2011) Motor functions of the superior colliculus. Annu Rev Neurosci
34:205‐231.
Genovese CR, Lazar NA, Nichols T (2002) Thresholding of statistical maps in functional
neuroimaging using the false discovery rate. Neuroimage 15:870‐878.
Golding B, Pouchelon G, Bellone C, Murthy S, Di Nardo AA, Govindan S, Ogawa M, Shimogori T,
Luscher C, Dayer A, Jabaudon D (2014) Retinal input directs the recruitment of inhibitory
interneurons into thalamic visual circuits. Neuron 81:1057‐1069.
Grubb MS, Thompson ID (2004) Visual response properties in the dorsal lateral geniculate
nucleus of mice lacking the beta2 subunit of the nicotinic acetylcholine receptor. J
Neurosci 24:8459‐8469.
Halassa MM, Siegle JH, Ritt JT, Ting JT, Feng G, Moore CI (2011) Selective optical drive of
thalamic reticular nucleus generates thalamic bursts and cortical spindles. Nat Neurosci
14:1118‐1120.
Hammer S, Carrillo GL, Govindaiah G, Monavarfeshani A, Bircher JS, Su J, Guido W, Fox MA
(2014) Nuclei‐specific differences in nerve terminal distribution, morphology, and
development in mouse visual thalamus. Neural Development 9:16.
88
Harrington ME (1997) The ventral lateral geniculate nucleus and the intergeniculate leaflet:
interrelated structures in the visual and circadian systems. Neurosci Biobehav Rev 21.
Hartings JA, Temereanca S, Simons DJ (2000) High responsiveness and direction sensitivity of
neurons in the rat thalamic reticular nucleus to vibrissa deflections. J Neurophysiol
83:2791‐2801.
Hattar S, Kumar M, Park A, Tong P, Tung J, Yau KW, Berson DM (2006) Central projections of
melanopsin‐expressing retinal ganglion cells in the mouse. J Comp Neurol 497.
Hirsch JA, Alonso JM, Reid RC, Martinez LM (1998) Synaptic integration in striate cortical simple
cells. J Neurosci 18:9517‐9528.
Hirsch JA, Wang X, Sommer FT, Martinez LM (2015) How inhibitory circuits in the thalamus
serve vision. Annual Review of Neuroscience, Vol 38 38:309‐329.
Hou G, Smith AG, Zhang ZW (2016) Lack ofintrinsic GABAergic connections in the thalamic
reticular nucleus of the mouse. Journal of Neuroscience 36:7246‐7252.
Huang L, Xi Y, Peng Y, Yang Y, Huang X, Fu Y, Tao Q, Xiao J, Yuan T, An K, Zhao H, Pu M, Xu F, Xue
T, Luo M, So K‐F, Ren C (2019) A Visual Circuit Related to Habenula Underlies the
Antidepressive Effects of Light Therapy. Neuron 102:128‐142.e128.
Hubel DH (1960) Single unit activity in lateral geniculate body and optic tract of unrestrained
cats. J Physiol 150:91‐104.
Huberman AD, Manu M, Koch SM, Susman MW, Lutz AB, Ullian EM, Baccus SA, Barres BA
(2008) Architecture and activity‐mediated refinement of axonal projections from a
mosaic of genetically identified retinal ganglion cells. Neuron 59:425‐438.
Huguenard J, Prince D (1992) A novel T‐type current underlies prolonged Ca(2+)‐dependent
burst firing in GABAergic neurons of rat thalamic reticular nucleus. J Neurosci 12:3804‐
3817.
Ide LS (1982) The fine structure of the perigeniculate nucleus in the cat. The Journal of
Comparative Neurology 210:317‐334.
Inamura N, Ono K, Takebayashi H, Zalc B, Ikenaka K (2011) Olig2 Lineage Cells Generate
GABAergic Neurons in the Prethalamic Nuclei, Including the Zona Incerta, Ventral Lateral
Geniculate Nucleus and Reticular Thalamic Nucleus. Developmental Neuroscience
33:118‐129.
Ishikane H, Kawana A, Tachibana M (1999) Short‐ and long‐range synchronous activities in
dimming detectors of the frog retina. Vis Neurosci, 16:1001‐1014.
Jager P, Ye Z, Yu X, Zagoraiou L, Prekop HT, Partanen J, Jessell TM, Wisden W, Brickley SG,
Delogu A (2016) Tectal‐derived interneurons contribute to phasic and tonic inhibition in
the visual thalamus. Nature Communications 7:13579.
89
Jahnsen H, Llinas R (1984) Electrophysiological properties of guinea‐pig thalamic neurones: an
in vitro study. J Physiol 349:205‐226.
Jones HE, Sillito AM (1994) The length‐response properties of cells in the feline perigeniculate
nucleus. Eur J Neurosci 6:1199‐1204.
Jones JP, Palmer LA (1987) The two‐dimensional spatial structure of simple receptive fields in
cat striate cortex. J Neurophysiol 58:1187‐1211.
Kashef A, Campolattaro MM, Freeman JH (2014) Learning‐related neuronal activity in the
ventral lateral geniculate nucleus during associative cerebellar learning. J Neurophysiol
112:2234‐2250.
Kay JN, De la Huerta I, Kim IJ, Zhang Y, Yamagata M, Chu MW, Meister M, Sanes JR (2011)
Retinal ganglion cells with distinct directional preferences differ in molecular identity,
structure, and central projections. J Neurosci 31:7753‐7762.
Kim IJ, Zhang Y, Yamagata M, Meister M, Sanes JR (2008) Molecular identification of a retinal
cell type that responds to upward motion. Nature 452:478‐482.
Kim U, Sanchez‐Vives MV, McCormick DA (1997) Functional dynamics of GABAergic inhibition in
the thalamus. Science 278:130‐134.
Kimura A, Imbe H, Donishi T, Tamai Y (2007) Axonal projections of single auditory neurons in
the thalamic reticular nucleus: implications for tonotopy‐related gating function and
cross‐modal modulation. Eur J Neurosci 26:3524‐3535.
Koepsell K, Wang X, Hirsch JA, Sommer FT (2010) Exploring the function of neural oscillations in
early sensory systems. Front Neurosci 4:53.
Koepsell K, Wang X, Vaingankar V, Wei Y, Wang Q, Rathbun DL, Usrey WM, Hirsch J, Sommer FT
(2009) Retinal oscillations carry visual information to cortex. Front Sys Neurosci 3.
Krahe TE, El‐Danaf RN, Dilger EK, Henderson SC, Guido W (2011) Morphologically distinct
classes of relay cells exhibit regional preferences in the dorsal lateral geniculate nucleus
of the mouse. J Neurosci 31:17437‐17448.
Lam Y‐W, Sherman SM (2007) Different Topography of the Reticulothalmic Inputs to First‐ and
Higher‐Order Somatosensory Thalamic Relays Revealed Using Photostimulation. J
Neurophysiol 98:2903‐2909.
Landisman CE, Long MA, Beierlein M, Deans MR, Paul DL, Connors BW (2002) Electrical
synapses in the thalamic reticular nucleus. J Neurosci 22:1002‐1009.
Lesica NA, Stanley GB (2004) Encoding of natural scene movies by tonic and burst spikes in the
lateral geniculate nucleus. J Neurosci 24:10731‐10740.
Lu SM, Guido W, Sherman SM (1992) Effects of membrane voltage on receptive field properties
of lateral geniculate neurons in the cat: contributions of the low‐threshold Ca2+
conductance. J Neurophysiol 68:2185‐2198.
90
Martinez LM, Wang Q, Reid RC, Pillai C, Alonso JM, Sommer FT, Hirsch JA (2005) Receptive field
structure varies with layer in the primary visual cortex. Nat Neurosci 8:372‐379.
McAlonan K, Cavanaugh J, Wurtz RH (2006) Attentional modulation of thalamic reticular
neurons. J Neurosci 26:4444‐4450.
Monavarfeshani A, Sabbagh U, Fox MA (2017) Not a one‐trick pony: Diverse connectivity and
functions of the rodent lateral geniculate complex. Vis Neurosci, 34.
Montero VM, Singer W (1984) Ultrastructure and synaptic relations of neural elements
containing glutamic acid decarboxylase (GAD) in the perigeniculate nucleus of the cat. A
light and electron microscopic immunocytochemical study. Exp Brain Res 56:115‐125.
Munk MH, Neuenschwander S (2000) High‐frequency oscillations (20 to 120 Hz) and their role
in visual processing. Journal of Clinical Neurophysiology 17:341‐360.
Muresan RC, Jurjut OF, Moca VV, Singer W, Nikolic D (2008) The oscillation score: An efficient
method for estimating oscillation strength in neuronal activity. J Neurophysiol 99:1333‐
1353.
Nelson AB, Kreitzer AC (2014) Reassessing models of basal ganglia function and dysfunction.
Annu Rev Neurosci 37:117‐135.
Niell CM, Stryker MP (2008) Highly Selective Receptive Fields in Mouse Visual Cortex. J Neurosci
28:7520‐7536.
Ohara PT, Lieberman AR, Hunt SP, Wu JY (1983) Neural elements containing glutamic acid
decarboxylase (GAD) in the dorsal lateral geniculate nucleus of the rat;
Immunohistochemical studies by light and electron microscopy. Neuroscience 8:189‐
211.
Pachitariu M, Steinmetz N, Kadir S, Carandini M, Kenneth D H (2016) Kilosort: realtime spike‐
sorting for extracellular electrophysiology with hundreds of channels. bioRxiv:061481.
Pinault D (2004) The thalamic reticular nucleus: structure, function and concept. Brain Research
Reviews 46:1‐31.
Pinault D, Smith Y, Deschenes M (1997) Dendrodendritic and Axoaxonic Synapses in the
Thalamic Reticular Nucleus of the Adult Rat. J Neurosci 17:3215‐3233.
Piscopo DM, El‐Danaf RN, Huberman AD, Niell CM (2013) Diverse visual features encoded in
mouse lateral geniculate nucleus. J Neurosci 33:4642‐4656.
Reinagel P, Godwin D, Sherman SM, Koch C (1999) Encoding of visual information by LGN
bursts. J Neurophysiol 81:2558‐2569.
Rivlin‐Etzion M, Zhou K, Wei W, Elstrott J, Nguyen PL, Barres BA, Huberman AD, Feller MB
(2011) Transgenic mice reveal unexpected diversity of on‐off direction‐selective retinal
ganglion cell subtypes and brain structures involved in motion processing. J Neurosci
31:8760‐8769.
91
Rockhill RL, Daly FJ, MacNeil MA, Brown SP, Masland RH (2002) The Diversity of Ganglion Cells
in a Mammalian Retina. The Journal of Neuroscience 22:3831.
Salay LD, Ishiko N, Huberman AD (2018) A midline thalamic circuit determines reactions to
visual threat. Nature 557:183‐189.
Saleem AB, Lien AD, Krumin M, Haider B, Rosón MR, Ayaz A, Reinhold K, Busse L, Carandini M,
Harris KD (2017) Subcortical Source and Modulation of the Narrowband Gamma
Oscillation in Mouse Visual Cortex. Neuron 93:315‐322.
Sanderson KJ (1971) The projection of the visual field to the lateral geniculate and medial
interlaminar nuclei in the cat. J Comp Neurol 143:101‐108.
Sanderson KJ, Darian‐Smith I, Bishop PO (1969) Binocular corresponding receptive fields of
single units in the cat dorsal lateral geniculate nucleus. Vision Res 9:1297‐1303.
Sanes JR, Masland RH (2015) The types of retinal ganglion cells: current status and implications
for neuronal classification. Annu Rev Neurosci 38:221‐246.
Schmidt TM, Do MTH, Dacey D, Lucas R, Hattar S, Matynia A (2011) Melanopsin‐Positive
Intrinsically Photosensitive Retinal Ganglion Cells: From Form to Function. The Journal of
Neuroscience 31:16094.
Schreiber S, Fellous JM, Whitmer D, Tiesinga P, Sejnowski TJ (2003) A new correlation‐based
measure of spike timing reliability. Neurocomputing 52‐4:925‐931.
Schwartz O, Pillow JW, Rust NC, Simoncelli EP (2006) Spike‐triggered neural characterization.
Journal of Vision 6:484‐507.
Shapley R, Lennie P (1985) Spatial frequency analysis in the visual system. Annu Rev Neurosci
8:547‐583.
Sherman SM (2001) Tonic and burst firing: dual modes of thalamocortical relay. Trends in
Neurosciences 24:122‐126.
Sherman SM, Guillery RW (2001) Exploring the Thalamus. San Diego: Academic Press.
Sholl DA (1953) Dendritic organization in the neurons of the visual and motor cortices of the
cat. Journal of Anatomy 87:387‐406.
Simm GM, de Ribaupierre F, de Ribaupierre Y, Rouiller EM (1990) Discharge properties of single
units in auditory part of reticular nucleus of thalamus in cat. J Neurophysiol 63:1010‐
1021.
Soto‐Sánchez C, Wang X, Vaingankar V, Sommer FT, Hirsch JA (2017) Spatial scale of receptive
fields in the visual sector of the cat thalamic reticular nucleus. Nature Communications
8:800.
Sparks DL (2002) The brainstem control of saccadic eye movements. Nature Reviews
Neuroscience 3:952‐964.
92
Spear PD, Smith DC, Williams LL (1977) Visual receptive‐field properties of single neurons in
cat's ventral lateral geniculate nucleus. J Neurophysiol 40:390‐409.
Sridharan D, Boahen K, Knudsen E (2011) Space coding by gamma oscillations in the barn owl
optic tectum. J Neurophysiol 105:2005‐2017.
Sriram B, Meier PM, Reinagel P (2016) Temporal and spatial tuning of dorsal lateral geniculate
nucleus neurons in unanesthetized rats. J Neurophysiol 115:2658‐2671.
Stelzner DJ, Baisden RH, Goodman DC (1976) The ventral lateral geniculate nucleus, pars
lateralis of the rat. Synaptic organization and conditions for axonal sprouting. Cell Tissue
Res 170.
Steriade M, Domich L, Oakson G (1986) Reticularis thalami neurons revisited: activity changes
during shifts in states of vigilance. J Neurosci 6:68‐81.
Stitt I, Galindo‐Leon E, Pieper F, Engler G, Engel AK (2013) Laminar profile of visual response
properties in ferret superior colliculus. J Neurophysiol 110:1333‐1345.
Storchi R, Bedford RA, Martial FP, Allen AE, Wynne J, Montemurro MA, Petersen RS, Lucas RJ
(2017) Modulation of fast narrowband oscillations in the mouse retina and dLGN
according to background lightintensity. Neuron 93:299‐307.
Sumitomo I, Sugitani M, Fukuda Y, Iwama K (1979) Properties of cells responding to visual
stimuli in the rat ventral lateral geniculate nucleus. Exp Neurol 66:721‐736.
Suresh V, Ciftcioglu UM, Wang X, Lala BM, Ding KR, Smith WA, Sommer FT, Hirsch JA (2016)
Synaptic contributions to receptive field structure and response properties in the rodent
lateral geniculate nucleus of the thalamus. J Neurosci 36:10949‐10963.
Swadlow HA, Gusev AG (2001) The impact of 'bursting' thalamic impulses at a neocortical
synapse. Nat Neurosci 4:402‐408.
Thomson JJ (1904) On the structure of the atom: an investigation of the stability and periods of
oscillation of a number of corpuscles arranged at equal intervals around the
circumference of a circle; with application of the results to the theory of atomic
structure. The London, Edinburgh, and Dublin Philosophical Magazine and Journal of
Science 7:237‐265.
Tschetter WW, Govindaiah G, Etherington IM, Guido W, Niell CM (2018) Refinement of Spatial
Receptive Fields in the Developing Mouse Lateral Geniculate Nucleus Is Coordinated
with Excitatory and Inhibitory Remodeling. The Journal of Neuroscience 38:4531.
Turrigiano GG, Leslie KR, Desai NS, Rutherford LC, Nelson SB (1998) Activity‐dependent scaling
of quantal amplitude in neocortical neurons. Nature 391:892‐896.
Uhlrich DJ, Cucchiaro JB, Humphrey AL, Sherman SM (1991) Morphology and axonal projection
patterns of individual neurons in the cat perigeniculate nucleus. J Neurophysiol 65:1528‐
1541.
93
Usrey WM, Alitto HJ (2015) Visual functions of the thalamus. Annual review of vision science
1:351‐371.
Vaingankar V, Soto Sanchez C, Wang X, Sommer FT, Hirsch JA (2012) Neurons in the thalamic
reticular nucleus are selective for diverse and complex visual features. Front Integr
Neurosci 6.
Vaingankar V, Soto Sanchez C, Wang X, Bains A, Sommer F, Hirsch J (2010) Visual features evoke
reliable bursts in the perigeniculate sector of the thalamic reticular nucleus. Cosyne
Abstr:177.
Wang L, Sarnaik R, Rangarajan K, Liu X, Cang J (2010a) Visual receptive field properties of
neurons in the superficial superior colliculus of the mouse. J Neurosci 30:16573‐16584.
Wang X, Hirsch JA, Sommer FT (2010b) Recoding of sensory information across the
retinothalamic synapse. J Neurosci 30:13567‐13577.
Wang X, Sommer FT, Hirsch JA (2011) Inhibitory circuits for visual processing in thalamus Curr
Op Neurobiol 21:726‐733.
Wang X, Wei Y, Vaingankar V, Wang Q, Koepsell K, Sommer FT, Hirsch JA (2007) Feedforward
excitation and inhibition evoke dual modes of firing in the cat's visual thalamus during
naturalistic viewing. Neuron 55:465‐478.
Wimmer RD, Schmitt LI, Davidson TJ, Nakajima M, Deisseroth K, Halassa MM (2015) Thalamic
control of sensory selection in divided attention. Nature 526:705‐709.
Worgotter F, Suder K, Zhao Y, Kerscher N, Eysel UT, Funke K (1998) State‐dependent receptive‐
field restructuring in the visual cortex. Nature 396:165‐168.
Xue JT, Carney T, Ramoa AS, Freeman RD (1988) Binocular interaction in the perigeniculate
nucleus of the cat. Exp Brain Res 69:497‐508.
Zhao X, Liu M, Cang J (2014) Visual cortex modulates the magnitude but not the selectivity of
looming‐evoked responses in the superior colliculus of awake mice. Neuron 84:202‐213.
Abstract (if available)
Abstract
The visual role of thalamus is more than a passive relay of visual information to the primary visual cortex. Indeed, thalamus has various visual structures with different roles. To begin with, the lateral geniculate nucleus (LGN) is composed of substructures which have different functions. Its well studied dorsal division (dLGN) relays visual information from the retina to the primary visual cortex. On the other hand, the ventral division (vLGN) and associated intergeniculate leaflet (IGL) do not project to cortex. These two structures (vLGN/IGL) connect with many subcortical structures such as superior colliculus (SC) and pretectum, hence they likely have both sensory and motor roles, also referred as non-image forming visual functions. In addition to the LGN, the visual thalamus includes a small portion of the thalamic reticular nucleus (TRN), a network of GABAergic cells. The TRN is usually studied in the context of sleep and attention, but our knowledge of its contributions to visual function is sparse. How do these different structures in the thalamus process visual information? What functions do they serve in the visual system? ❧ To address these questions, I first explored on the visual processing in the vLGN/IGL via electrophysiological recordings from individual neurons in mice and how that compares to the dLGN and SC. The size of receptive fields in the vLGN was larger than the ones in the dLGN. The dendritic arbors in the vLGN was commensurately larger and the size of individual post-synaptic currents were smaller. When compared to SC, the temporal precision observed in the vLGN was lower, along with coarser spatial receptive fields. The vLGN also exhibited visual gamma band oscillations, similar to SC. The strength of these oscillations also changed during the presentation of naturalistic stimulation, suggesting a role for oscillations during natural vision. ❧ Then, I moved on to investigate the visual receptive field structure in the TRN in mice. The receptive fields in TRN were mostly ON – OFF, though one contrast was generally dominant. In terms of spatial scale, around half of the cells in TRN had receptive fields as small as the ones in dLGN, providing a means for localized reticular inhibition on LGN. At the same time, the temporal profile of the receptive fields revealed that bursts encode different information than tonic spikes in both TRN and dLGN. Bursts were preceded with a stronger phase of non-preferred contrast leading the stimuli of preferred contrast, than tonic spikes. This suggests that visual stimulus can change the mode of firing and bursts likely have a distinct role during vision. When compared to previous studies, our results highlight conserved aspects of visual processing in TRN in mouse, a leading model system to study attention and neural circuits. ❧ Overall, this work provides a fuller understanding of the role of various visual thalamic nuclei and how they process visual information.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Spatial and temporal precision of inhibitory and excitatory neurons in the murine dorsal lateral geniculate nucleus
PDF
Sensory information processing by retinothalamic neural circuits
PDF
Contextual modulation of sensory processing via the pulvinar nucleus
PDF
Neural circuits underlying the modulation and impact of defensive behaviors
PDF
Functional magnetic resonance imaging characterization of peripheral form vision
PDF
Synaptic mechanism underlying development and function of neural circuits in rat primary auditory cortex
Asset Metadata
Creator
Ciftcioglu, Ulas Mustafa
(author)
Core Title
Exploring sensory responses in the different subdivisions of the visual thalamus
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Biology (Neurobiology)
Publication Date
08/12/2019
Defense Date
04/30/2019
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
feedback,GABAergic,geniculate,in vivo,inhibition,LGN,neural coding,OAI-PMH Harvest,oscillations,receptive field,reticular,thalamus,TRN,visual,whole-cell
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Hirsch, Judith A. (
committee chair
), Hires, Andrew S. (
committee member
), Mel, Bartlett W. (
committee member
), Sommer, Friedrich T. (
committee member
)
Creator Email
ciftciog@usc.edu,ulas.ciftcioglu@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c89-210946
Unique identifier
UC11662943
Identifier
etd-Ciftcioglu-7766.pdf (filename),usctheses-c89-210946 (legacy record id)
Legacy Identifier
etd-Ciftcioglu-7766.pdf
Dmrecord
210946
Document Type
Dissertation
Format
application/pdf (imt)
Rights
Ciftcioglu, Ulas Mustafa
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
feedback
GABAergic
geniculate
in vivo
inhibition
LGN
neural coding
oscillations
receptive field
reticular
thalamus
TRN
visual
whole-cell