UNIVERSIDADE FEDERAL DO RIO GRANDE DO NORTE
PROGRAMA DE P ´OS GRADUAC¸ ˜AO EM NEUROCI ˆENCIAS
The role of activity of the dorsal cochlear nucleus in tinnitus perception in mice
O papel da atividade do n´ucleo coclear dorsal na percep¸c˜ao de tinnitus em camundongos
Thawann Malfatti Borges
Natal-RN 2020
THAWANN MALFATTI BORGES
The role of activity of the dorsal cochlear nucleus in tinnitus perception in mice
O papel da atividade do n´ucleo coclear dorsal na percep¸c˜ao de tinnitus em camundongos
Tese de doutorado apresentada ao Programa de P´ os-Gradua¸c˜ao em Neurociˆencias da UFRN como parte dos requisitos para obten¸c˜ao do t´ıtulo de Doutor em Ciˆencias. ´Area de concentra¸c˜ao: Neurociˆencias
Orientadora: Prof. Dra. Emelie Katarina Svahn Le˜ao
Natal-RN 2020
Borges, Thawann Malfatti.
The role of activity of the dorsal cochlear nucleus in tinnitus perception in mice / Thawann Malfatti Borges. - Natal, 2020.
116f.: il.
Tese (Doutorado em Neurociências) - Instituto do Cérebro, Universidade Federal do Rio Grande do Norte, 2020.
Orientador: Emelie Katarina Svahn Leão. Coorientador: Richardson Naves Leão.
1. Tinnitus. 2. Cochlear nucleus. 3. Chemogenetics. 4. Optogenetics. I. Leão, Emelie Katarina Svahn. II. Leão, Richardson Naves. III. Título.
RN/UF/Biblioteca Setorial Árvore do Conhecimento CDU 612.8
Universidade Federal do Rio Grande do Norte - UFRN Sistema de Bibliotecas - SISBI
Catalogação de Publicação na Fonte. UFRN - Biblioteca Setorial Árvore do Conhecimento - Instituto do Cérebro - ICE
The role of activity of the dorsal cochlear nucleus in tinnitus perception in mice
O papel da atividade do n´ucleo coclear dorsal na percep¸c˜ao de tinnitus em camundongos
THAWANN MALFATTI BORGES
Apresentada em 27 de mar¸co de 2020
BANCA EXAMINADORA
Prof. Dr Sergio Neuenschwander
Prof. Dr Rodrigo Neves Romcy Pereira
Profa. Drª Marine Diniz da Rosa
Prof. Dr Nivaldo Antonio Portela de Vasconcelos
Prof. Dr Richardson Naves Le˜ao
Acknowledgments
First of all, none of this work would be done without my brilliant and lovely wife, Barbara Ciralli. Thank you for all the help, support and unconditional love.
I would like to thank my supervisor Katarina Le˜ao for guiding me through this process, as my co-supervisor Richardson Le˜ao and my evaluation committee, Jo˜ao Bacelo, Claudio Queiroz, Tarciso Velho, Marine Rosa, Nivaldo Vasconcelos, Sergio Neuenschwander and Rodrigo Pereira.
Furthermore, I would like to thank the professors and collegues from the Brain Institute for all the shared knowledge and insightful conversations.
Next, it was awesome to be selected to go to the Transylvanian Experimental Neuro-science Summer School! I would like to thank IBRO-PERC, TINS, the TENSS committee, and all the great friends I met there for all the impressively condensed knowledge shared.
Also this work would not be possible without the financial support from the American Tinnitus Association, UFRN, CAPES and CNPq.
Contents
1 Introduction 1
1.1 The cellular organization of the dorsal cochlear nucleus . . . 1
1.2 The dorsal cochlear nucleus is implicit in tinnitus . . . 3
1.3 Why so hard to treat or cure tinnitus? . . . 4
1.4 Generation and detection of tinnitus perception in mice . . . 6
1.4.1 Generating tinnitus - Noise exposure . . . 6
1.4.2 Assessing hearing thresholds - Auditory brainstem responses . . . . 8
1.4.3 Quantification of tinnitus - gap prepulse inhibition of acoustic startle 9 1.5 Targeting neuronal populations using gene promoters . . . 10
1.6 Controlling neuronal activity - Optogenetics and chemogenetics . . . 13
2 Aims 15 3 Methods 17 3.1 Hardware and software . . . 17
3.1.1 Timestamps markers . . . 17
3.1.2 Sound calibration . . . 18
3.1.3 Light calibration and stimulation . . . 21
3.1.4 Software availability . . . 21
3.2 Animals . . . 22
3.3 Experimental design . . . 23
3.4 Gap prepulse inhibition of acoustic startle reflex . . . 25
3.5 Viral injection . . . 27
3.6 Auditory brainstem responses . . . 29
3.7 Noise exposure . . . 30
3.8 in vivo unit recording . . . 31
3.9 Histological evaluation . . . 33
3.9.1 Transcardial perfusion . . . 33
3.9.2 Brain sectioning . . . 33
3.9.3 Acquisition of images . . . 33
4 Summary of studies 35 4.1 Dorsal cochlear nucleus unit activity can be controlled using constructs containing the CaMKIIα and ChRNA2 promoters . . . 35
4.2 CaMKIIα+ dorsal cochlear nucleus neurons are necessary for tinnitus main-tenance . . . 38
6 Manuscript 1 - Dorsal cochlear nucleus unit activity can be controlled using constructs containing the CaMKIIα and ChRNA2 promoters 59 7 Manuscript 2 - CaMKIIα+ dorsal cochlear nucleus neurons are
List of Figures
1.1 Schematic of the DCN circuitry . . . 3 1.2 Examples of gap prepulse inhibition of acoustic startle (GPIAS) protocols . 10 1.3 Markers for specific neuronal populations in the cochlear nucleus . . . 12 3.1 Schematics of sound and light timestamps using a two channel sound card 19 3.2 Sound calibration . . . 20 3.3 Experimental design . . . 23 3.4 GPIAS stimulation paradigm . . . 25 4.1 Graphical abstract summarizing study 4.1: CaMKIIα-ChR2-eYFP positive
DCN cells optogenetic manipulation alters DCN units firing . . . 37 4.2 Graphical abstract summarizing study 4.2: CaMKIIα-hM4Di positive DCN
cells have an important role in noise-induced tinnitus . . . 39 6.1 CaMKIIα-ChR2-eYFP positive neurons in the DCN following local viral
injection . . . 60 6.2 Blue light stimulation of CaMKIIα-ChR2-eYFP expressing neuron in the
DCN can increase or generate delayed decreased unit firing . . . 62 6.3 Green light stimulation of CaMKIIα-Arch-eYFP expressing neurons
gen-erates delayed excitation or delayed inhibition . . . 64 6.4 Inhibition of CaMKIIα+ cells in the DCN can both increase and decrease
excitation of DCN units . . . 65 6.5 Confocal image showing tdTomato expression and DIO-ChR2-eYFP
ex-pression in transversal brainstem sections from Chrna2cre/tomatolx mice . . 66
6.6 Chrna2-cre positive neurons of the VCN can be targeted to drive activity of DNC neurons . . . 68 6.7 Normal auditory brainstem response in injected animals . . . 70 7.1 Noise exposure causes tinnitus in mice without causing hearing loss . . . . 86 7.2 Tinnitus-like behaviour can be recovered by decreasing DCN
CaMKIIα-hM4Di positive cells activity . . . 87 7.3 Decreasing CaMKIIα-hM4Di positive cells activity in the DCN changes
firing rate of the circuitry . . . 88 7.4 Tinnitus without hearing loss cannot be prevented by decreasing
CaMKIIα-hM4Di positive cells activity in the DCN during noise exposure . . . 90 7.5 Decreasing CaMKIIα-hM4Di positive DCN cells activity during noise
ex-posure abolish hM4Di-dependent recovery . . . 91 7.6 CNO reduces firing frequency and best frequency and increases broadness
List of Tables
3.1 Hardware . . . 22 3.2 Titer of viral vectors . . . 28 7.1 Viral vectors with the Ca2+/Calmoduline kinase 2α promoter . . . 84
Abbreviations, acronyms and initialisms
ABR auditory brainstem response
ADC analog-digital converter
eArch3.0 enhanced archaerhodopsin 3.0
CaMKIIα calcium/calmodulin-dependent protein kinase type II α
ChR2 channelrhodopsin-2
ChRNA2 choline receptor nicotinic alpha 2 subunit
CNO clozapine-n-oxide
CNS central nervous system
Cre Cre-recombinase
DAC digital-analog converter
DAPI 4’,6-diamidino-2-phenylindole dBSPL decibels sound pressure level
DCN dorsal cochlear nucleus
DREADD designer receptor exclusively activated by designer drug
EEG eletroencephalography
eYFP enhanced yellow fluorescent protein
GlyT2 glycine transporter 2
GPIAS gap prepulse inhibition of acoustic startle
GUI Graphical User Interface
hM3Dq human muscarinic 3 designer receptor hM4Di human muscarinic 4 designer receptor
IC inferior colliculus
i.p. intraperitoneal
MBCT Mindfulness-based cognitive therapy
PBS phosphate buffered saline
PPI prepulse inhibition
PSD power spectral density
SD standard deviation
SNR signal-to-noise ratio
SOC superior olivary complex
tDCS transcranial direct current stimulation
VCN ventral cochlear nucleus
VGAT vesicular GABA/glycine transporter VGluT vesicular glutamate transporter
Abstract
Noise-induced tinnitus is the perception of an acute or chronic phantom sound, with-out a physical source, caused by over-exposure to loud noise. Tinnitus patients that report bothersome tinnitus often suffer from increased anxiety, stress, lack of concentration, in-somnia and depression, conditions that can severely decrease quality of life. There are still no cure or a reliable treatment option for tinnitus highlighting that further studies of cellu-lar mechanisms behind the generation and maintenance of tinnitus are much needed. The dorsal cochlear nucleus (DCN), a region of the brainstem known to integrate somatosen-sory and auditory input, has been identified as a potential key structure in the generation of tinnitus. However, there is still a lack of studies investigating the effect of manipulating subgroups of DCN neurons in tinnitus in vivo. Here, we first show how optogenetic stim-ulation of DCN cells expressing proteins for activating (CaMKIIα-ChR2) or inhibiting (CaMKIIα-eArch3.0) DCN neurons expressing the genetic promoter calcium/calmodulin-dependent protein kinase type II α can modulate different DCN units responses to simple sounds. Our results show that these neurons are highly interesting for modulating sound processing of the auditory brainstem. Next, in mice with noise-induced tinnitus, we show that our noise exposure does not lead to hearing loss, but changes the perception of a silent gap in background noise before an acoustic startle. This indicates that mice have tinnitus in a specific frequency that is matched by the background sound tested. Next we show that chemogenetic inhibition of DCN cells expressing CaMKIIα-hM4Di, a synthetic protein that lowers excitability of neurons, can significantly decrease tinnitus (p = 0.038, n = 11 mice), compared to control mice that showed no improvement in tinnitus (control virus; CaMKIIα-eYFP, p = 0.696, n = 7 mice). Unit recordings confirmed chemogenetic inhibition of CaMKIIα-hM4Di positive DCN cell activity. Specifically, decreasing DCN units excitability using chemogenetic g-protein receptors activated by low dose clozapine-N-oxide decreased DCN firing frequency, modulated best frequency and tuning width of unit response to sound. Next, to examine if this population is necessary for generating noise-induced tinnitus, CaMKIIα-hM4Di positive DCN neurons were instead chemoge-netically inhibited during tinnitus induction (n = 6 experimental and 6 control mice), but such manipulation did not prevent tinnitus, nor did chemogenetic inhibition 2 weeks later decreased tinnitus perception. Our results suggest that CaMKIIα-hM4Di positive cells in the DCN are not crucial for tinnitus induction but play a significant role in maintaining tinnitus in mice. This study adds understanding of the role of a subgroup of DCN cells in tinnitus, pointing to a potential target for tinnitus interventions.
Resumo
Tinnitus induzido por ru´ıdo ´e a percep¸c˜ao aguda ou crˆonica de um som fantasma, sem fonte f´ısica, causada pela exposi¸c˜ao excessiva ao ru´ıdo intenso. Os pacientes com tinnitus que relatam incˆomodo causado pelo tinnitus frequentemente sofrem de aumento de ansiedade, estresse, falta de concentra¸c˜ao, insˆonia e depress˜ao, condi¸c˜oes que podem diminuir severamente a qualidade de vida. Ainda n˜ao h´a cura ou uma op¸c˜ao de tratamento confi´avel para o tinnitus, destacando que mais estudos dos mecanismos celulares por tr´as da gera¸c˜ao e manuten¸c˜ao do tinnitus s˜ao necess´arios. O n´ucleo coclear dorsal (DCN), uma regi˜ao do tronco encef´alico conhecida por integrar inputs somatosensoriais e auditivos, foi identificado como uma potencial estrutura chave na gera¸c˜ao do tinnitus. No entanto, ainda faltam estudos investigando o efeito da manipula¸c˜ao de subgrupos de neurˆonios do DCN no tinnitus in vivo. Aqui, mostramos primeiro como a estimula¸c˜ao optogen´etica de c´elulas do DCN expressando prote´ınas para ativar (CaMKIIα-ChR2) ou inibir (CaMKIIα-eArch3.0) neurˆonios expressando o promotor gen´etico calcium/calmodulin-dependent protein kinase type II α pode modular diferentes respostas de unidades do DCN a sons simples. Nos-sos resultados mostram que esses neurˆonios s˜ao altamente interessantes para modular o processamento do som do tronco encef´alico auditivo. Em seguida, em camundongos com tinnitus induzido por ru´ıdo, mostramos que nossa exposi¸c˜ao ao ru´ıdo n˜ao leva `a perda auditiva, mas muda a percep¸c˜ao de uma lacuna de silˆencio no ru´ıdo de fundo antes de um sobressalto ac´ustico. Isto indica que os camundongos tˆem tinnitus em uma freq¨uˆencia espec´ıfica que ´e igualada pelo som de fundo testado. A seguir mostramos que a inibi¸c˜ao quimiogen´etica das c´elulas do DCN expressando CaMKIIα-hM4Di, uma prote´ına sint´etica que diminui a excitabilidade de neurˆonios, pode diminuir significativamente o tinnitus (p = 0.038, n = 11 camundongos) em compara¸c˜ao com os camundongos controle, que n˜ao mostraram melhora no tinnitus (v´ırus controle; CaMKIIα-eYFP, p = 0.696, n = 7 ca-mundongos). Os registros de unidades confirmaram inibi¸c˜ao quimiogen´etica da atividade de c´elulas do DCN expressando CaMKIIα-hM4Di. Especificamente, a diminui¸c˜ao da excitabilidade das unidades do DCN usando receptores quimiogen´eticos de prote´ına g ati-vados por baixas doses de clozapina-N-´oxido diminuiu a freq¨uˆencia de disparo, modulou a melhor freq¨uˆencia e largura da curva de resposta por frequˆencia de unidades do DCN que respondem a som. Em seguida, para examinar se esta popula¸c˜ao ´e necess´aria para gerar tinnitus induzido pelo ru´ıdo, os neurˆonios do DCN expressando CaMKIIα-hM4Di foram inibidos quimiogeneticamente durante a indu¸c˜ao do tinnitus (n = 6 camundongos experimentais e 6 camundongos controle), mas tal manipula¸c˜ao n˜ao impediu o surgimento do tinnitus, nem a inibi¸c˜ao quimiogenetica 2 semanas depois diminuiu a percep¸c˜ao do tin-nitus. Nossos resultados sugerem que as c´elulas do DCN expressando CaMKIIα-hM4Di n˜ao s˜ao cruciais para a indu¸c˜ao do tinnitus, mas desempenham um papel significativo na manuten¸c˜ao do tinnitus em camundongos. Este estudo acrescenta a compreens˜ao do papel de um subgrupo de c´elulas do DCN no tinnitus, apontando para um alvo potencial para interven¸c˜oes de tinnitus.
1.
Introduction
The auditory system allows us to better connect with our surrounding, helping us to navigate safely and communicate with others. Obviously losing our ability to hear is debilitating, but also to hear excessive sounds that are generated by our own brain can be very stressful. Tinnitus, the perception of a phantom sound, can severely decrease quality of life, and still has no cure or well estabilished treatment. One of the reasons is that we do not fully understand at circuitry level how tinnitus arises and remains chronic. This thesis investigated the role of a subgroup of neurons from the beginning of the central auditory pathway in tinnitus, and we found that this subgroup has an important role in mantaining tinnitus in mice.
1.1
The cellular organization of the dorsal cochlear
nucleus
The cochlear nucleus is the first part of the ascending auditory system receiving sound information directly from the auditory nerve. It is located on the lateral edges of the brainstem and can be subdivided into the dorsal cochlear nucleus (DCN) and ventral cochlear nucleus (VCN). The VCN is specialized in temporal processing, while the DCN is specialized in processing spectral contrasts (for example, differences between voices or music types). The DCN also receives somatosensory inputs from skin of the head and the neck, for example linking the sound of the wind blowing with the sensation of wind on your head, thereby being the first sensory bimodal structure in the auditory pathway (Schnupp et al., 2011). Another difference between the ventral and dorsal parts of the cochlear nucleus is their projection targets, where the VCN projects output largely to the superior olivary complex (SOC) of the brainstem, crucial sound localization, while the DCN projects directly to the inferior colliculus (IC), a structure that communicates both with the thalamus and the auditory cortex for complex sound processing and regulation
of ascending and descending signals (Slater et al., 2019). Interestingly, the DCN has a layered organization, structurally different from the other auditory nuclei, probably due to the complex integration of sensory and sound afferents. Specifically, the DCN can be divided into three layers: the molecular layer, the superficial layer containing mostly mossy fibers, having only a few cell bodies, including granule, golgi and small stellate cells; the fusiform cell layer, which contains mainly fusiform and cartwheel cells; and the deep layer, which contains several different types of cells, including giant cells, granule cells and tuberculoventral cells (Willott, 2001; Oertel and Young, 2004; Figure 1.1).
The DCN cells, similar to the cochlea and most auditory structures, are tonotopically organized, meaning they are spatially distributed from lower to higher frequencies they respond better to. The output of the DCN is provided by excitatory fusiform cells some-times referred to as pyramidal cells of the DCN (Mancilla and Manis, 2009), and giant cells, sending excitatory fibers through the dorsal acoustic stria, on the most dorsoven-tral edge of the DCN, mainly to the condorsoven-tralateral inferior colliculus and medial genicu-late nucleus of the thalamus. The input to these cells comes from the auditory nerve, that provides excitatory inputs to fusiform, giant and tuberculoventral (or vertical) cells; also the VCN provides important input to the DCN in the form of two different types of stellate cells. D-stellate cells provides inhibitory inputs to giant and tuberculoven-tral cells, and T-stellate cells provide excitatory inputs to giant cells. Next also mossy fibers provide somatosensory-related excitatory and inhibitory inputs to fusiform cells and the inhibitory cartwheel cells in bundles of parallel fibers from the dorsal column nuclei (somatosensory input from limbs and tail) and the pontine nuclei (a precerebellar nucleus), the pinnae, the trigeminal ganglion, among others (Oertel and Young, 2004; Shore, 2005). Internally, fusiform cells receive inhibitory inputs from cartwheel tuber-culoventral and DCN stellate cells; and giant cells receive inhibitory input from cartwheel and tuberculoventral cells (Figure 1.1, Oertel and Young, 2004; Schofield et al., 2014; Trussell and Oertel, 2018). It is clear that the DCN contains a complex circuitry with several different excitatory cells receiving input from different sources, information inte-grated also by complex actions of different inhibitory cells. The DCN is even compared to cerebellar circuits, due to its parallel fiber/mossy fiber input, and the fusiform cells that receive different input onto its apical and basal dendrites (Oertel and Young, 2004). Therefore it is not surprising that the DCN has been pointed as a key structure in tinnitus pathophysiology (Kaltenbach et al., 2005).
MGN IC Molecular layer Fusiform cell layer Deep layer Somato-sensory Inferior colliculus D-stellate cells (VCN) T-stellate cells (VCN) Midline Auditory nerve
Figure 1.1: Schematic of the DCN circuitry. Excitatory connections ends with a black dot, while inhibitory connections ends with a bar. Schematics show granule cells (cyan), fusiform cells (green), giant cells (blue), cartwheel cells (red), tuberculoventral cells (purple), purkinje-like cells (orange) and stellate cells (yellow) (based onOertel and Young, 2004)
1.2
The dorsal cochlear nucleus is implicit in tinnitus
Tinnitus is defined as the perception of a high-pitched sound in one or both ears or in the head when no external sound is present (the American Tinnitus Association, www.ata.org). It has been shown that bilateral DCN lesions prevent noise-induced tinni-tus development (Brozoski et al., 2012). Another evidence for the DCN being implicated in tinnitus is that excitatory fusiform cells of the DCN, tuned to frequencies of tinni-tus, show increased spontaneous activity (Baizer et al., 2012) and increased synchrony and bursting activity after noise exposure (Wu et al., 2016). Also inhibitory neurons of the DCN shows alterations in tinnitus models. Bursting neurons, supposedly cartwheel cells, showed an increased excitability after noise-induced tinnitus, where increasing doses of carbachol (a cholinergic agonist known to increase excitability) generated higher fir-ing rates in tinnitus-model animals than in control animals (Chang et al., 2002). Since cartwheel cells inhibits fusiform cells but also other cartwheel cells, as well as the fact that the two cell types have different glycinergic and gabaergic receptor composition (Mancilla and Manis, 2009), noise-induced tinnitus could be correalted to an unbalance of inhibition within the DCN circuit.
Besides excitatory and inhibition unbalance, it is also possible that tinnitus can cause unbalance in auditory and somatosensory integration in the DCN (Dehmel et al., 2008; Dehmel et al., 2012; Basura et al., 2015). Actually, somatosensory stimuli can modulate tinnitus in some patients, related to the multimodal activity of the DCN (Baizer et al.,
2012). Also, unilateral cochlear damage can alter the distribution of vesicular glutamate transporters (VGluTs) on synaptic terminals. Specifically, VGluT 1 and 2 in the cochlear nucleus showed labeling of VGluT1 (relating to auditory signals) to be decreased and of VGluT2 (relating to somatosensory input) to be increased following cochlear damage (Zeng et al., 2009). These results were replicated by overexposing guinea-pigs (Heeringa et al., 2018) or rats (Han et al., 2019) to noise. Since these transporters are associated to auditory and somatosensory input respectively, and since noise-induced tinnitus can cause cochlear damage, it is likely that exposure to a loud noise for a long time cause an un-balance between auditory and somatosensory modalities, seen simply by a reorganization of the amount of fibers innervating the cochlear nucleus. This is supported by swelling of axonal terminals in response to overexcitation, leading to deformed presynaptic terminals and decreased reattachment to post-synaptic densities following acoustic trauma (Kujawa and Liberman, 2009). Therefore it is clear that the DCN plays a role in tinnitus.
1.3
Why so hard to treat or cure tinnitus?
Several forms of tinnitus are known, and are generally separated by its probable cause, such as noise exposure, hearing loss, head and neck trauma, secondary to other disorders (Meniere’s disease, Lyme disease, fibromyalgia), tumors, jaw misalignment, use of ototoxic drugs, cardiovascular and pulsatile tinnitus (www.ata.org). It is estimated that about one third of adults experience tinnitus occasionally while 10-15% have prolonged tinnitus requiring medical evaluation (Heller, 2003). In patients with severe tinnitus the common comorbidities are insomnia, anxiety, depression and cognitive dysfunction (Shi et al., 2014), conditions all leading to a decreased quality of life.
There is still no cure for tinnitus, and more so, there are no firmly estabilished proto-cols for treating tinnitus. Cognitive-behavioural therapy shows promising results (Cima et al., 2014), while variations of auditory training, such as sound-masking or sound therapy, has not yet provided clinically relevant data to show its usefulness in decreasing tinnitus perception (Hoare et al., 2010). Recently, it was shown that Mindfulness-based cognitive therapy (MBCT) can significantly decrease patients reported Tinnitus Functional Index, a scoring protocol that measures tinnitus effect on different life aspects, 8 weeks after therapeutic intervention (Husain et al., 2019). The intervention lead to decreased con-nectivity between the amygdala and frontal and parietal areas, which may be underlying
the observed Tinnitus Functional Index improvement (Zimmerman et al., 2019). Pharma-cological treatments aims merely at treating comorbidities and may for example include anxiolytics or antidepressants (Shi et al., 2014) although this is recommended against as tinnitus treatment strategy (Tunkel et al., 2014). Furthermore, low frequency repetitive transcranial magnetic stimulation of the auditory cortex has shown some reduction in local excitability that can relieve tinnitus symptoms short term (Minami et al., 2011). Interestingly, a treatment consisting of auditory-somatosensory bimodal stimulation was shown to reduce tinnitus perception in humans and in a noise-induced tinnitus animal model (Marks et al., 2018).
Limbic structures are also involved in tinnitus perception (Kraus and Canlon, 2012). The hippocampus, which mediates memories but also emotions, such as risk-taking and anxiety (Mikulovic et al., 2018; Winne et al., 2018), has been shown to be excessively activated in tinnitus patients, suggesting a relation between tinnitus and the limbic system (Lockwood et al., 1998). Several nuclei within the limbic system have been shown to have the capability to inhibit neurons within the auditory thalamus (M¨uhlau et al., 2006; Yu et al., 2009). Also, alteration of the auditory system alone may not be sufficient to cause tinnitus, as there are numerous patients with sensorineural hearing loss and no tinnitus, and non-auditory factors that can modulate tinnitus perception, as modelled by Rauschecker et al. (2010)and later reviewed byShore et al. (2016). There can be dynamic changes in neuronal activity within the auditory and limbic system that cause tinnitus with normal audiograms (Qu et al., 2019). Therefore, it is important to understand the earliest stages of tinnitus generation in the auditory pathway, to decrease interplay by other than auditory brain areas.
This doctoral thesis is focused on the common form of tinnitus: noise-induced tinni-tus, which can present as an acute or chronic conditions depending on the severity (Davis, 2014). One difficulty in understanding noise-induced tinnitus is that it is closely related to noise-induced hearing loss, making it difficult to determine whether alterations like hyperexcitability, increased spontaneous firing and increased synchrony are related to one of them, or both. There are studies demonstrating the relation between noise-induced hearing loss and increased neural synchrony, excitability and spontaneous firing rate (e.g. Seki and Eggermont, 2003; Yang et al., 2011; Manzoor et al., 2013), triggering homeo-static plasticity, tonotopic reorganization of the auditory cortex, and then finally leading to tinnitus. However, several evidences add complexity to this hypothesis, for
exam-ple, loud noise exposure cause a permanent damage in the auditory system (synaptic loss and neuronal degeneration) but cochlear function (and hearing thresholds) can fully recover after two weeks (Kujawa and Liberman, 2009). Also, response of the auditory nerve is decreased after noise damage, but the IC responses appears normal (Schaette and McAlpine, 2011). Together, this shows the importance of decreasing the amount of confouding factors when studying noise-induce tinnitus.
In all cases of tinnitus, the major hinder for treatment of tinnitus is the lack of un-derstanding of how the condition starts and what brain areas are most important for its persistence. In summary, the DCN make up a complex circuit with a fine balance of activity between different cell populations, and this balance appears altered in tinnitus. The balance between excitation and inhibition is essential for system functionality, and a disturbance of this balance (increased excitation and decreased inhibition, for example) is thought to be one of the main causes of tinnitus (Kaltenbach and Manz, 2012). There-fore, to treat or cure tinnitus it is possible that one has to find ways to specifically alter the activity of subsets of neuronal populations, to decrease activity of excitatory cells for example. So far, the hunt for specific pharmacological agents (Langguth et al., 2019) or specific sensory or electrical stimulation first rely on having adequate animal models of tinnitus.
1.4
Generation and detection of tinnitus perception
in mice
1.4.1
Generating tinnitus - Noise exposure
The animal model of noise-induced tinnitus involves exposing an animal to a loud noise for a prolonged period of time. This procedure is called noise (over)exposure, and it appears to be able to induce tinnitus permanently depending on the level and duration of the noise. Interestingly, adequate noise exposure often generate a temporary threshold shift (temporary hearing loss) that can recover to normal levels indicating that it is possible to generate a model of tinnitus without the complexity of concurrent hearing loss. The details vary along the literature, where studies use different subjects, intensities from 100 to 124 decibels sound pressure level (dBSPL), pure tones or filtered white noise, for 0.5 to 4h, unilateral or bilateral exposure (e.g.,Bauer and Brozoski, 2001;Heffner and
Harrington, 2002;Basta and Ernst, 2004;Kujawa and Liberman, 2009;Yang et al., 2016).
For example, Bauer and Brozoski (2001) exposed long-evans rats to a gaussian white noise containing frequencies from 5 to 40kHz, centered at 16kHz, at loud levels (105dBSPL), for one or two hours, unilaterally. Both animals exposed for one or two hours presented an 50 dBSPL level of threshold shift (meaning that if they first could detect sound at 30dBSPL they would only detect sound as loud as 80dBSPL shortly after the noise ex-posure). Animals exposed for 1h had a more tonal tinnitus, around 20kHz; and animals exposed for 2h showed a broader tinnitus, from 10 to 30kHz, and had a greater cochlear hair-cell loss. In another example, Heffner and Harrington (2002) exposed hamsters to a 10kHz pure tone, at 124dBSPL for 0.5, 1 or 4h; or at 127dBSPL for 2h. Animals exposed for 1, 2 and 4h showed the same threshold shift around 25dBSPL at 20kHz. Animals exposed for 1h did not develop tinnitus, while the 2h-exposed animals presented the smallest overlap with control animals in behavioral tests (eg. supposedly developed tinnitus). Pace and Zhang (2013) showed that after unilaterally exposing mice to 120dB 10kHz pure tone for 2h animals developed tinnitus at 6-8kHz and 26-28kHz, and had no spatial learning deficits nor increased anxiety comparing to control animals. Here it is important to note that a single pure-tone is known to be more processed by the VCN and affect DCN circuits differently from broadband noise, where inhibitory tuberculoven-tral cell respond more to pure tones (Oertel and Young, 2004). Therefore it is likely that there are different mechanisms involved in pure-tone generated tinnitus compared to other bands of noise-induced tinnitus. For human subjects, broadband noise is more likely to cause tinnitus, as sources of persistent loud pure tones are rare in our natural surroundings and something we naturally avoid being exposed to.
For auditory neuroscience, the literature is far from consistent about the details of noise-induced tinnitus (Kaltenbach and Afman, 2000;Bauer and Brozoski, 2001;Heffner and Harrington, 2002; Seki and Eggermont, 2002; Basta and Ernst, 2004; Shore et al., 2008; Finlayson and Kaltenbach, 2009; Kujawa and Liberman, 2009; Wang et al., 2009; Longenecker and Galazyuk, 2011;Middleton et al., 2011;Yang et al., 2011;Dehmel et al., 2012; Li et al., 2013; Manzoor et al., 2013; Yang et al., 2016). For this thesis, we have chosen gaussian white noise containing frequencies from 9 to 11kHz; with intensity of 90dBSPL, trying to minimize persistent threshold shifts/hearing loss; with 1h exposure, followed by 2h in a quiet environment, since a noisy environment can interfere in tinnitus development (Nore˜na and Eggermont, 2006; Le Prell et al., 2012; Sturm et al., 2017).
1.4.2
Assessing hearing thresholds - Auditory brainstem responses
Auditory brainstem responses (ABRs) involve a procedure that objectively assess the hearing threshold of subjects. ABRs represents electrical potentials from several struc-tures of the auditory pathway in response to sound stimuli. Such stimuli can be sound ”clicks”, around 3ms of noise containing all frequencies the speaker/phone can produce, with unpredicted intensity for each frequency but controlled total intensity, which are nor-mally used to verify anatomical structure functionality (cochlea, auditory nerve, cochlear nucleus for example); or ”tone pips”, around 5ms pure-tones with controlled intensity for each frequency tested, normally used for assessing hearing thresholds for different frequen-cies. In general, an ABR stimuli consists of several intensities presented to the subject in decreasing order (for example intensities starting at 80dBSPL next decreasing stepwise by 10dBSPL). In humans the neuronal response generated to clicks/tone pips are recorded with scalp electrodes (such as for any eletroencephalography; EEG), while for animals electrodes can be placed subdermally onto the scalp and neck region. The electrical re-sponse, the ABRs, are represented by electrical waves of activity and are quantified by each wave peak. The largest five peaks of an ABR after stimulus onset corresponds to spi-ral ganglion neurons (the neurons forming the auditory nerve), the cochlear nucleus, the cochlear nucleus/trapezoid body, the SOC and the IC, respectively. These responses are expected to be intensity-dependent, where lower intensities evoke responses with smaller amplitude and greater latencies (Willott, 2001).
For assessing hearing levels in mice, a common approach is to place a recording and a reference electrode subdermally onto the animal’s head and a ground electrode connected to a ground sink, which can be the animal’s skin or the recording system ground. Then, sound is played repeatedly at a certain frequency, usually 10∼30kHz. The responses to each stimulus intensity are then averaged and bandpass-filtered from 300 to 3000Hz. The integrity of the auditory pathway is assessed by the amplitude, absolute latencies and inter-peak latencies for the five largest peaks in the first 10ms time window of the resultant trace after sound onset in a suprathreshold intensity, generally 80dBSPL. In other words, if a peak is missing, this can indicate that a particular structure is not functioning normally. Hearing threshold is usually defined as the least intense stimulation that could evoke discernible peaks in the ABR traces (Willott, 2001).
In this study, we used 3ms narrow-band gaussian white noise pulses at five different 2kHz-wide frequency bands, which is a middle-term between testing a pure tone or a
broadband noise; repeated 529 times, to increase signal-to-noise ratio (SNR) 23x (increase in SNR equals the square root of number of trials averaged); repeated at 10Hz, since it provides good peak amplitude without leading to adaptation, i.e., responses would be smaller for each trial (ASHA, 1988). See General Methods section 3.6 for technical details.
1.4.3
Quantification of tinnitus - gap prepulse inhibition of
acous-tic startle
A recent behavioural test for evaluating tinnitus perception in mice is the GPIAS test, first described for rats by Turner et al. (2006). The GPIAS test is a variant of the classical prepulse inhibition (PPI) test used for measuring sensorimotor gating deficits sometimes seen in schizophrenia. Here, the GPIAS instead measure the capability of the animal to detect a warning cue, a gap of silence during persistent background noise, to decrease the animal’s startle reflex in response to a sudden loud sound. The usefulness of this test for tinnitus perception is that it measures a reflex and preferably do not depend on training or motivation of an animal. The logic behind the GPIAS test for tinnitus perception is that if the intensity and frequency of a certain background noise matches intensity and frequency of tinnitus, the animal will not perceive a silent pause (gap) in the background noise as a warning cue, and thereby startle more to a sudden loud sound pulse. If the mouse detects the gap this is expected to decrease its startle reflex. Thus, it is possible to separate animals into groups of no tinnitus and tinnitus perception, based on the acoustic startle reflex. Still, studies report that not all mice are capable of perceiving the gap in background noise, despite normal hearing threshold, and therefore it is crucial to pre-screen animals to exclude mice that fail the task (Li et al., 2013).
Similar to noise overexposure, parameters vary between GPIAS studies. For exam-ple,Turner et al. (2006) used a 60dBSPL 1kHz-wide filtered white noise centered at 10 or 16kHz; a 50ms gap; followed by 50ms of background noise; followed by a 20ms loud broad-band noise pulse at 105dBSPL. Mwilambwe-Tshilobo et al. (2015) used 60dBSPL white noise filtered at several different 2kHz-wide frequency bands and broadband white noise; a 40ms gap; followed by 60ms of background noise; followed by a 50ms loud broadband noise pulse at 115dBSPL.
The frequency bands of the background noise are different between studies intending to match the tinnitus frequency perceived by the animal, and each study had its own parame-ters for tinnitus generation; and usually the loud sound pulse is not filtered. The intensity
0 50 100 150 200 Time [ms] 0 20 40 60 80 100 120 In tensit y [dB] Turner et al., 2006 0 50 100 150 200 Time [ms] Mwilambwe et al., 2015 A B
Figure 1.2: Examples of GPIAS protocols. Trial portion used byTurner et al. (2006) (A) and Mwilambwe-Tshilobo et al. (2015) (B). Note how the gap portion, the noise before loud pulse and the loud pulse can vary between the two studies.
of the background noise is usually fixed at 60 or 70dBSPL, and intensity of the loud pulse used varies from 105 to 116dBSPL (Turner et al., 2006;Longenecker and Galazyuk, 2011; Middleton et al., 2011; Turner et al., 2012; Mwilambwe-Tshilobo et al., 2015). In sum-mary, it is important to carefully choose parameters for optimal tinnitus experiments (Longenecker and Galazyuk, 2012) and more so to screen mice for GPIAS detection ca-pabilities before starting experiments, as well as, measure the hearing threshold of mice to make sure the background noise is audible.
1.5
Targeting neuronal populations using gene
pro-moters
As mentioned earlier, it would be highly beneficial to target specific cells within the DCN to be able to investigate if the presumed unbalance in neuronal activity is crucial for both tinnitus induction and general mechanisms of tinnitus. A widely adopted strategy for target the expression of a synthetic protein that can be controlled to modulate neuronal activity is local injection of viral constructs containing a promoter sequence specifically expressed in a cell population of interest, the genetic code of a protein of interest, and a gene of a fluorescent reporter for histology and confirmation of infected cells (Chow et al., 2012). The incorporation of the genetic material by the virus allows expression only in cell types that carries the factors that can bind to the specific promoter sequence
(cells that inherently express the promoter of choice), and thereby activate transcription of a particular gene. Another popular strategy for targeting specific cell types is the use of transgenic animals with the bacterial gene Cre-recombinase (Cre) coupled to a neuronal or neuronal population-specific promoter. Cre is an enzyme with the function of cutting or spinning DNA sequences at specific short base-pair regions called lox-sites (locus of crossing over site). The logic here is that a gene of interest can be delivered to the nervous system using virus, but instead of carrying a specific promoter region, the viral construct is designed so that gene expression needs activity of Cre (Chow et al., 2012). The Cre/lox system was originally developed for agricultural use but is now part of the standard toolbox of molecular neuroscience (Sauer, 1998) for targeted expression of optogenetic proteins (Tsien, 2016).
In order to study neuronal populations, unique cell markers would be highly use-ful to drive protein expression in the particular population of interest. For example, Calcium/calmodulin-dependent protein kinase type II α (CaMKIIα) is an enzyme acti-vated by increase in intracellular Ca2+ concentration, involved in regulation of various physiological processes (Yamauchi, 2005). This enzyme is expressed by pyramidal cells of the neocortex as well as pyramidal and granular cells of the hippocampus (Wang et al., 2013). Interestingly, the CaMKIIα expression in the DCN of rats was demonstrated by Ochiishi et al. (1998) to be concentrated in the superficial layers of DCN (molecu-lar and fusiform cell layer, Figure 1.3A). Since the cellu(molecu-lar properties of the superficial layers and the deep layer are different, it may be possible to use CaMKIIα as a target promoter for expressing proteins for modulating neuronal activity, such as optogenetic and chemogenetic proteins, in fusiform cells also in mice. In line with these findings, using the Allen Mouse Brain atlas and Allen Mouse Brain Connectivity Atlas (avaliable at http://mouse.brain-map.org, see Lein et al., 2007) show CaMKIIα mRNA to indeed target the fusiform cell layer but also some cells of the deep layer in the DCN of mice (Figure 1.3B, upper part DCN, lower part VCN).
A recently described transgenic animal, the ChRNA2-Cre (expressing Cre after choline receptor nicotinic alpha 2 subunit promoter,Le˜ao et al., 2012; Wootz et al., 2013; Perry et al., 2015), has been useful for targeting several unique neuronal population depending on area within the central nervous system (OLM cells of the hippocampus, Renshaw cells of the ventral horn of the spinal cord, and layer 5 Martinotti cells of the cortex). Inter-estingly, what these populations appear to have in common is that they provide recurrent
B
CaMKIIa CaMKIIa
A
C
ChRNA2 Figure 1.3: Markers for specific neuronal populations in the cochlear nucleus. A∼B) CaMKIIα+ cells in the molecular and fusiform cell layer of the rat DCN (A, Ochiishi et al., 1998) and of the mouse DCN (B, Allen Mouse Brain atlas). Note that some cells in the deep layer of the mouse DCN are also labeled (B). C) Choline receptor nicotinic alpha 2 (ChRNA2) positive cells of a ChRNA2-Tomato mouse expressing Tomato reporter (Malfatti, 2015). Soma of cells can be observed in the VCN, with fibers projecting to intensely innervate deep and fusiform cells of the DCN. A second cell population (bushy cells) of the VCN also innervates the SOC, with projections going ventro-medially in the image. Scalebars: A, 100µm; B, 80µm; C, 250µm. Lateral right, dorsal up (A∼C).
inhibition, thereby serving as a presumed ”break” or switch of input to nearby large exci-tatory neurons. In the auditory brainstem we have observed ChRNA2-Cre/tdTomato-lox positive animals to show (where Cre-expressing cells are labeled in fluorescent red, tomato) distinct cell bodies in the VCN projecting fibers into the DCN and medially into the SOC (Figure 1.3C), and preliminary data of fiber projections as well as literature has indicated that these cells are likely to be T-stellate cells (as D-stellate cells do not respond to cholin-ergic agonists;Fujino and Oertel, 2001) and bushy cells (based on projection patterns, M. Hilscher and R. Leao personal communication). Interestingly, both the labeled ChRNA2+ neurons of the VCN are excitatory glutamatergic neurons, differently from previously de-scribed ChRNA2+ populations. Thus, it is likely that the ChRNA2-Cre mouse may be highly useful for manipulating DCN cells but also that promoters may differ to what cell type they are expressed in (e.g. excitatory or inhibitory neurons) depending on region within the central nervous system, and this needs to be further investigated.
1.6
Controlling neuronal activity - Optogenetics and
chemogenetics
Optogenetics was first tested in cultured neurons in 2005 and proved to be a technique that allows for millisecond-timescale control of cellular activity (Boyden et al., 2005). This technique has since revolutionized neuroscience and allowed for genetic dissection of a multitude of circuits (Luo et al., 2008; Luo et al., 2018). The technique builds on expressing photosensitive transmembrane proteins (channels or pumps) that allows certain ions to cross the cell membrane when exposed to a specific light wavelength band (Nagel et al., 2003;Boyden et al., 2005; Deisseroth et al., 2006). Two of the most used optogenetic proteins are channelrhodopsin-2 (ChR2) and enhanced archaerhodopsin 3.0 (eArch3.0). ChR2 is derived from a green-algae light-activated cation-selective ion channel, which can depolarize cells by passive inward cation conductance, mainly Na+, when exposed to blue light. Derived from bacteria, eArch3.0 is an outward proton pump that can conduct a large and stable photocurrent (flux of positively charged protons, H+, out from the cell) to inhibit spontaneous or evoked activity, when exposed to yellow-green light (Mattis et al., 2012). Thereby optogenetics involves light-controlled activity of neurons using thin optic fibers to shine light into specific brain regions. Control of specific neuronal populations in slice preparation of the cochlear nucleus has been carried out using transgenic mice, where the VGluT2 has been used to target and control DCN fusiform cell firing (Apostolides and Trussell, 2014), and the promoter for the glycine transporter 2 (GlyT2-Cre mice) have been used to control DCN cartwheel cell firing in vitro (Apostolides and Trussell, 2013;Lu and Trussell, 2016). Transgenic mice expressing ChR2 coupled to the promoter for the vesicular GABA/glycine transporter (VGAT) has been used to excite inhibitory interneurons and study inhibitory neurotransmission in slices of the VCN (Xie and Manis, 2014). Still, there is a lack of studies showing subpopulation control of cochlear nucleus neurons in vivo.
Chemogenetics serves a similar purpose as optogenetics, but does not allow precise temporal control of neuronal firing, instead chemogenetics can increase or decrease ex-citability of cells over longer time periods, without the need of implanting optic fibers. It consists in expressing mutated muscarinic metabotropic receptors modified to respond only to clozapine-n-oxide (CNO), a synthetic drug with no biological function in mice. Thus, CNO can be injected locally or systemically as it only binds to its mutated
re-ceptors. Administration of CNO will then trigger a cascade of cellular events (G-protein coupled signaling) that leads to increased or decreased cell excitability/activity. Recently, a commonly used chemogenetic tool is the designer receptor exclusively activated by de-signer drugs (DREADDs), and comprise excitatory and inhibitory receptors. Specifically the human muscarinic 3 designer receptor (hM3Dq), when expressed for example using the Cre/lox system, increases cell excitability, while the human muscarinic 4 designer re-ceptor (hM4Di) leads to decrease in cell excitability (Rogan and Roth, 2011). The effect of CNO was shown to peak ≈10min after intraperitoneal (i.p.) injection of CNO, and the molecule is removed from the system after 2h, even though effects could be observed up to 8h in pancreatic β cells after CNO injection (Guettier et al., 2009).
The combination of correct delivery strategy, the appropriate cellular expression and the correct stimulation protocol allows millisecond-timescale control and long-term control of cell population activity which is crucial for better understanding complex neuronal abnormalities such as tinnitus.
2.
Aims
The aim of this thesis is to investigate the role of different DCN neuronal populations in the generation and/or maintenance of tinnitus perception. For the purposes of this thesis, “CaMKIIα+” refers to neurons infected with viral constructs to express optogenetic or chemogenetic proteins. The following three hypothesis were examined:
• Hypothesis A) Optogenetic control of cochlear nucleus CaMKIIα+ neurons can modulate DCN unit responses to sound stimulation (Study 4.1);
• Hypothesis B) Decreasing activity of DCN CaMKIIα+ neurons can decrease noise-induced tinnitus perception (Study 4.2);
• Hypothesis C) Decreasing activity of DCN CaMKIIα+ neurons during noise expo-sure can prevent tinnitus (Study 4.2).
3.
Methods
3.1
Hardware and software
3.1.1
Timestamps markers
Timestamps refer to a digital record of the time of occurrence of a particular event, such as sound pulses or light pulses, and to be able to analyze neuronal responses it is important to time neuronal events to the stimuli (sometimes randomized). For auditory structures such as the DCN, millisecond-timescale precision is crucial for stimulation and analysis of neural responses to sound. Here we controlled timestamp markers using a combination of a sound card and an Arduino board, taking advantage of GNU/Linux audio real-time capabilities. In detail, the two sound card outputs were connected to the sound amplifier, then one of the amplifier outputs was connected to a speaker while the other was connected to an analog input in the Open-ephys board. In addition, one digital output pin of Arduino was connected to a digital input of the Open-ephys board, exclusively used for triggering start/stop of the recording system.
Initially we compared digital and analog timestamp accuracy. In our system, we quantified the jitter of digital timestamps to test the temporal accuracy of the digital inputs. We recorded 5000 square pulses of 5V amplitude delivered to both analog and digital inputs to the acquisition board. When compared to analog traces, we found 15% of the digital timestamps to be delayed by >150µs, which is a jitter of 5% of the 3ms pulse width (Figure 3.1A), making analog timestamps our choice for stimulation timestamps.
Next we implemented simple coding rules for transforming the analog pulses into timestamps. To avoid producing capacitive-like traces when using square pulses in a sound card, a square wave (both positive and negative part) was used for every sound pulse written to a sound card output. The square wave was twice the sound pulse duration with 0.5V amplitude and was written simultaneously in a second sound card output. A
diode was connected to the sound card outputting the square wave, thus only the positive part of this signal is conducted, resulting in a square pulse with the same width of the sound pulse providing timestamp markers for onset and offset of the stimulus (Figure 3.1B). For experiments using sound synchronized with light stimulation (optogenetic ex-periments) three outputs would be required (one carrying the sound signal to a speaker, one carrying the sound square waves/timestamps and the third carrying square waves for light trigger/timestamps), although most sound cards only provide two analog output channels. To solve this, a square wave of twice the length for light stimulation was used with 4V amplitude, so when simultaneous sound and light stimulation is required, the sound pulse is written to one output channel while the sum of the sound and light square waves is written to another output channel (Figure 3.1D). Thereby the second output channel triggers the laser (amplitude >3.3V), as well as provides edges for sound and light timestamp detection needed for subsequent analysis or responses. Here, the detec-tion of onset and offset of stimuli was defined as M ean + (3 × SD) as threshold, where M ean and SD are the mean and standard deviation (SD) of the square pulses signal, respectively. Thereby, onsets were detected when a value was greater than the threshold, and the previous value was smaller than it. Inversely, offsets were detected when a value was less than the threshold and the previous value was greater than it.
3.1.2
Sound calibration
Sound cards generally need to be calibrated, as the output signal may undergo pre-processing, such as filtering. Here, the sound card (Sound Device USBPre2, Thomann GmbH, Germany) was calibrated so the amplitude of the output signal matches the gener-ated signal written to the card; and the amplitude of the signal read by the board matches that of the signal sent to its input. The sound card volume control was placed at unit level (the level where the signal is not amplified or attenuated, i.e., no filters are applied if any). A 10kHz sine wave of 1V amplitude was written to the sound card and we verified the amplitude of the output signal using an oscilloscope (for our system 1.7V). The sound card output amplification factor was recorded as 1/Output, where Output is the measured amplitude, so every generated signal is multiplied by it before being written to the card. We connected the sound card output to the sound card input, and a 1V 10kHz sine wave was played and recorded. The input amplification factor was recorded as 1/Input, where Input is the amplitude of the recorded signal (for our system 0.486V), and signals read
B
C
D
Sound pulse Speaker 1 mV 5 ms Sound pulse Speaker 1 mV 5 ms Soundsq. wave timestampSound
Diode 4 mV 5 ms Acq. board analog in Laser sq. wave Laser timestamp Diode Laser 2 V
5 ms Acq. boardanalog in
Sound+Laser
sq. wave Sound+Lasertimestamp
Diode Laser 2 V 5 ms Acq. board analog in Delay to analog edge [ms] 0 10 20 30 40 50 % digital marks Analog Digital
A
Observed Expected Ch 2 Ch 1 Ch 2 Ch 2 Ch 1 0 0.5 1 3msFigure 3.1: Schematics of sound and light timestamps using a two channel sound card. A) Left, schematic showing analog pulses and a jitter of digital timestamps. Right, Histogram showing percentage of recorded pulses and delay to analog edge. B) Illustration of channel 1 producing a sound wave to the speaker; channel 2 producing the corresponding square wave with the positive portion having the same duration as the sound pulse (gray shading). A diode cuts the negative portion and allows for detection of the sound timestamp. C) Channel 2 writes the light square wave and the positive portion is split to provide a laser trigger and the acquisition board analog input indicates the light timestamp. D) Channel 1 as ‘B’, and channel 2 illustrates the sum of light and sound square waves. As in ‘C’, both timestamps can be extracted from the signal recorded from the analog input of the acquisition board.
Frequency [Hz] Frequency [Hz] Frequency [Hz] Frequency [Hz] Frequency [Hz] A B C D E F
Figure 3.2: Sound calibration. Sound pulses were played at several amplification factors and simultaneously recorded at the position of the animal’s left ear during experiments. A) Intensity curve per amplification factor for each frequency band. As expected, sound intensity decreases logarithmically as the amplification factor of the signal is decreased. B-F) power spectral density (PSD) of each frequency at 80, 70, 60, 50, and 40dBSPL. The last trace of each frequency is a sound pulse with amplification factor of 0, showing that environment noise level has no frequency peaks from 5 to 20kHz and oscillates between 29 and 31dBSPL.
from the board were multiplied by it before any further processing. A loudspeaker (Super tweeter ST400 trio, Selenium Pro) was calibrated using a microphone (4939-A-011, Br¨uel and Kjær, Denmark) 4.5 or 10cm in front of the speaker, depending on the experimental setup. Next, sound pulses of 2s duration were generated at the desired frequency bands with 300 logarithmically decreasing amplification factors (voltage output to the speaker) and simultaneously recorded using a personal computer.
We calculated the power spectrum density (PSD) of the recorded signal using a Hann window with no overlap. Root mean square (RMS) was calculated as
v u u t n X i=1 P SDi× BinSize
where P SD is a 1 × n array and BinSize is the spectral resolution (frequency bin size). The intensity in dB sound pressure level (dBSPL) was finally calculated as
20 × log
RM S M icSensV P a
2 × 10−6
where M icSensV P a is the microphone sensitivity in mV/Pa. These calculations allow
control of sound intensity with at least 0.5dBSPL precision, and allow further verification of other acoustic features, such as room background noise and speaker frequency response (see Figure 3.2). All data was written to disk and loaded for every experiment using sound stimulation, in a way that any intensity in dBSPL has a corresponding amplifica-tion factor. Figure 3.2B-F shows that the frequency band generated corresponds to the frequency band of greater power in the signal spectrum, with other frequencies strongly attenuated. Sound calibrations, full or reduced, were routinely done before each set of experiments. The reduced calibration test consisted of only 4 intensities (80, 60, 40 and 0dBSPL) per frequency and were applied to verify that the intensities recorded matches the expected intensity.
3.1.3
Light calibration and stimulation
Light intensity calibration was performed before each experiment. Optic fibers of 200µm diameter were cleaned with lens cleaning tissue and ethanol (99.5%). The light intensity was measured and the laser lens position adjusted until light power at the tip of the fiber was 5-7mW/mm2 measured by an optical power meter (Thorlabs PM20). Light
stimuli triggers were generated in Python and written to the sound device (USBPre2), in which the output was splitted and connected to the laser input and the data acquisition board. Light stimulus was delivered using a 473nm laser (for ChR2 stimulation) and a 532nm laser (for eArch3.0). For ChR2 experiments, laser stimuli consisted of at least 5 blocks of 200 light pulses at intensity of 5-7mW/mm2 with 10ms duration, presented
at 10Hz (10ms on and 90ms off) 473nm blue light, with 10s between each block. Light stimulation was presented alone or simultaneously with sound stimulus. For eArch3.0 ex-periments, laser stimuli consisted of 5 pulses of 20s of 543nm green light, with 10s between each pulse. Light stimulus was presented only simultaneously with sound stimulation.
3.1.4
Software availability
All hardware and software used are shown in Table 3.1. All scripts used for con-trolling devices, stimulation control and data analysis are available at https://gitlab. com/malfatti/SciScripts. The operating system of choice was Gentoo GNU/Linux, due to its flexible management of libraries (Ioanas, 2017). The Sounddevice python library (Geier, 2015) was used to read and write signal from/to the sound card. Recordings
were done using Open-ephys Graphical User Interface (GUI) (Siegle et al., 2015). Cal-culations were done using Scipy (Jones et al., 2001) and Numpy (Van Der Walt et al., 2011), and all plots were produced using Matplotlib v2.2.4 (Caswell et al., 2019; Hunter, 2007). Spikes were detected and clusterized using Klusta (Rossant et al., 2016) and SpyKING Circus (Yger et al., 2018), and visual inspection was performed using Phy (Rossant et al., 2016). Images were adjusted for brightness and contrast and zStacks projections were generated by ImageJ (Schneider et al., 2012).
Table 3.1: Hardware used for measurement, recording and stimulation. ADC = analog-digital converter; DAC = digital-analog converter.
Hardware Details
Accelerometer MMA8452Q 12bit 3-axis ±2g resolution
Acquisition board Open-ephys v2.2, Opal Kelly XEM6010-LX150, 30kHz sampling rate, 16bit, 8 ADCs, 8 DACs
Computer HP Z220 Intel Xeon 8-core 3.6GHz, 16GB
RAM
Fiber optic patch cables Thorlabs M81L01 200µm core fiber optic cable
Headstage Intan RHD2132 16 unipolar channels
Lasers CNI MBL-III-473 (blue) and
MGL-III-532 (green)
Microcontroller Arduino Uno or Due 54 digital pins, 12 12bit ADCs, 2 12bit DACs
Micromanipulator Sutter 1µm precision
Microphone Br¨uel and Kjær 4939-A-011
1/4” free-field microphone, sensi-tivity of 4.23643mV/Pa from 0.1-100kHz
Multichannel electrode NeuroNexus Single shank, 16 channels, 25 or 50µm channel spacing, 177µm recording site, 5mm length
Piezoelectric sensor MEAS-DT 10mV per microstrain
Sound amplifier Marantz PM8004 see Figure 3.2
Sound card USBPre2 192kHz sampling rate, 24bit ADC
Speakers Selenium Trio ST400 see Figure 3.2
3.2
Animals
For Study 4.1, Male C57Bl/6J mice, ChRNA2-Cre or ChRNA2-Cre mice crossed with the reporter line Ai14 tdTomato (ChRNA2-Cre/R26tom) mice age P21-P75 (n=14) were used. For Study 4.2, male C57Bl/6J mice (n=30), 21 days old at first experiment, were used, so at the last experiment the animals were approximately 2 months old. All animal
z + + + + + + + + + z
Figure 3.3: Experimental design of study 4.2. A) Two groups (experimental and control, re-ceiving viral vector injections containing the hM4Di gene sequence or only eYFP as fluorescent reporter, respectively) went through noise exposure, and hearing thresholds and tinnitus be-haviour were tested before and after noise exposure. B) Two separate groups, also experimental and control receiving viral injections as for A), went through noise exposure 30 minutes after CNO i.p. injection, and hearing thresholds and tinnitus behaviour were tested similarly.
procedures were approved and followed the guidelines of the Ethical Committee in Use of Animals from the Federal University of Rio Grande do Norte (CEUA - protocol number 051/2015).
3.3
Experimental design
In Study 4.1, different optogenetic tools were tested to control activity of a subgroup of DCN units. For this wild-type C57Bl/6J mice received viral injection of CaMKIIα-ChR2-eYFP or CaMKIIα-eArch3.0-eYFP, and ChRNA2-Cre mice received viral vector injection of DIO-ChR2-eYFP. Four weeks later, extracellular neuronal activity in the DCN was recorded using in vivo electrophysiology in anesthetized mice while the DCN was stimulated by either blue or green light using an optic fiber. At the end of recording sessions the animals were sacrificed for histological procedures.
In Study 4.2, the experimental design was meticulously laid out to obtain the most amount of data from each animal at fixed time points (see Figure 3.3). See specific details for each procedure in the specific subsection of General Methods. Over all, animals were habituated for 3 days and screened for up to 2 sessions (one per day) for gap-detection capabilities using the GPIAS test, and animals failing to detect the silent gap before the loud pulse under control conditions were excluded from further experiments (6 mice). Next, mice were injected with viral vectors (chemogenetic constructs), and one month later tested for hearing threshold (both approximately 3 days before and 3 days after noise exposure). Six days after noise exposure, animals were retested with the GPIAS test to evaluate tinnitus perception 30 minutes after NaCl injection (control/vehicle) and 30 minutes after CNO injection (to alter neuronal firing) in the following day (see Figure 3.3). At the end, extracellular neuronal activity in the DCN was recorded in the presence of NaCl and CNO using in vivo electrophysiology in anesthetized mice, and lastly animals were sacrificed for histological procedures. To answer the two experimental questions of whether tinnitus perception can be reduced, and if tinnitus induction can be prevented, by decreasing the activity of a subpopulation of DCN cells at different specific time points, animals were separated into four experimental groups:
1. Groups designed to evaluate if tinnitus perception can be decreased by lowering the activity of CaMKIIα+ DCN cells during the GPIAS test using CNO and how unit responses are altered by CNO application to cells expressing hM4Di receptors;
(a) The first group received viral vector injection of CaMKIIα-hM4Di at day 6 and CNO injections at day 46;
(b) The second group received viral vector only carrying the fluorescent marker enhanced yellow fluorescent protein (eYFP) (CaMKIIα-eYFP) at day 6 and CNO injection at day 46;
2. Groups designed to evaluate if tinnitus-induction can be avoided when decreasing CaMKIIα+ DCN cells activity during the noise exposure.
(a) The third group also received viral vector injection of CaMKIIα-hM4Di at day 6 but CNO injections at day 38-40 30 minutes before noise exposure and at day 46;
(b) The fourth group received viral vector injection of CaMKIIα-eYFP as control and also injection of CNO 30 minutes before noise exposure and at day 46.
0 60 105 In tensit y [dB] No-gap trial 0 1 2 3 Time [s] 0 60 105 In tensit y [dB] Gap trial A B C
Figure 3.4: GPIAS stimulation paradigm. A) Photograph of the sound-shielded box showing the acrylic tube with an accelerometer fixed underneath for vibration measurement. B-C) The fixed portion of the No-gap (B) and Gap (C) trials are intercalated with 10-20s of noise, pseudo-randomly selected, matching the frequency of the current trial, to avoid prediction responses.
3.4
Gap prepulse inhibition of acoustic startle reflex
The gap prepulse inhibition of acoustic startle (GPIAS, Turner et al., 2006) test, based on an animal startle reflex to loud sounds, was conducted in a sound shielded room with acoustic foam covering walls and rubber carpet covering the floor, in order to reduce reverberation and attenuate external noise. Inside the room a sound-shielded box, internal dimensions 37x45x37cm or 44x33x24cm, with LED lights was used for further isolation of ambient noise. During recordings, the animal was placed in a small custom-made clear acrylic tube, restricting it from standing on the back paws, dimensions 6.1x5.9x5.1cm. The acrylic tube was placed inside the sound-shielded box that was fitted with a speaker (Selenium Trio ST400, JBL by Harman, Brazil) placed 4.5cm away from the acrylic tube, connected to the sound amplifier connected to the sound card. In order to measure the animal’s startle reflex, a piezoelectric sensor or a digital accelerometer was mounted to the base plate of the acrylic tube. For initial recordings using the piezoelectric sensor, it was connected directly to an analog input of the Open-ephys board. For recordings using the accelerometer, it was connected to an Arduino Uno board using the I2C pins, and accelerometer data was outputted through PWM pins connected to three Open-ephys analog inputs.
back-ground level (60-70dBSPL or 10dBSPL above hearing threshold), loud (at 105dBSPL) or silence. Specifically, the sound stimulus was a concatenation of: a random integer value between 10 and 20 seconds of noise at background level (background noise between trials); 2.3s of noise at background level (initial background noise); 40ms of noise at back-ground level for No-Gap trials or 40ms of silence for Gap trials (Gap portion); 100ms of noise at background level (background noise before loud pulse); 50ms of noise at 105dB-SPL (loud pulse); and 510ms of noise at background level (final background noise; Figure 3.4). Timestamp marks were used only for the loud pulse. The bands of frequencies tested were 8-10, 9-11, 10-12, 12-14 and 14-16kHz. Background noise level was, for initial GPIAS test, 60dBSPL. For GPIAS after noise exposure, background noise level was individually adjusted to 10dBSPL above hearing threshold for the frequency tested.
The acrylic tube was always cleaned with ethanol 70% and then with water to remove smell of ethanol, which could stress the mouse, before each animal was placed in the GPIAS chamber. In order to habituate animals to manipulation, the home-cage was placed on a small table and mice were handled for 10 minutes in the test room for at least three days, until the mice freely moved in and out of its home-cage and explore the space surrounding it. Then, during the acclimatization session, the animals were placed in the GPIAS chamber and exposed to the test without the loud pulses, and next returned to the home-cage. A successful acclimatization was considered when animals decrease urination and defecation in the tube. After the habituation/acclimatization period, animals were screened for gap detection capability. In detail, animals were placed in the acrylic tube and left in the recording chamber for 5 minutes, allowing the animal to stay calm and stop exploring the chamber (Valsamis and Schmid, 2011). Next the test consists of 18 trials per band of frequency tested, 9 with gap (Gap trials) and 9 with noise filling the gap portion of the stimulus (No-gap trials), presented pseudo-randomly.
The recorded portion of each trial includes all steps from initial background noise to final background noise. The signal was bandpass filtered from 70 to 400Hz for piezoelectric recordings and lowpass filtered below 50Hz for accelerometer recordings. Data was cut 200ms around the loud pulse onset to only analyze the portion contains the gap or No-gap trials. For accelerometer recordings, the absolute values of the three axes were averaged. The 9 Gap trials of the same frequency band were averaged, as were the 9 No-gap trials. The instantaneous amplitude of the signal was calculated as the magnitude of the analytic representation of the averaged signal using the Hilbert transform. The amplitude of the
response was defined as the mean instantaneous amplitude of 100ms after the loud sound pulse subtracted by the mean instantaneous amplitude of 100ms before the loud pulse, which corrects for baseline offsets. The GPIAS index was calculated as
Gap N oGap − 1 × 100
where N oGap is the amplitude of response to No-gap trials and Gap is the amplitude of response to Gap trials. Tests recorded after noise exposure had the most affected frequency for each animal calculated as the frequency with the greatest index shift from before to after noise exposure. Comparisons between treatments were done using two-tailed paired Student’s t-test, Bonferroni-corrected for the number of frequency bands tested.
On screening of gap detection capabilities at day 3-6 (GPIAS initial test), animals that did not show a decrease of at least 30% in Gap vs NoGap trials for all frequencies were retested on the next day only on those frequencies. Animals that, after retest, did not show a decrease of at least 30% in Gap vs NoGap trials in at least two frequencies were excluded from further experiments.
3.5
Viral injection
For local injections of viral vectors carrying optogenetic or chemogenetic constructs mice were anesthetized with an i.p. injection of ketamine-xylazine combination at 90/6 mg/kg. When necessary, additional ketamine at 45 mg/kg was applied during surgery. The mouse was next mounted into a stereotaxic device while resting on a heating block at 37°C. The animal’s eyes were covered with dexpanthenol to prevent ocular dryness and povidone-iodine 10% was applied onto the skin of the animal’s head to avoid infections. The skin on the scalp was anesthetized with lidocaine hydrochloride 3% before a straight incision was made. After the incision, hydrogen peroxide 3% was applied onto the exposed skull to remove the connective tissue and to allow a clear visualization of the skull sutures. Then, the head was aligned so bregma and lambda have the same dorsoventral (DV) position, as for the symmetric coordinates at -3mm anteroposterior (AP) and ±2mm mediolateral (ML) positions. A small mark was made at AP=-6.24mm and at ML=2.3mm (coordinates of the DCN, according toPaxinos and Franklin, 2004), and a small hole was carefully drilled with a dental microdrill. All coordinates were previously normalized by
