Tinnitus-like “hallucinations” elicited by sensory deprivation in an entropy maximization recurrent neural network

Sensory deprivation has long been known to cause hallucinations or “phantom” sensations, the most common of which is tinnitus induced by hearing loss, affecting 10–20% of the population. An observable hearing loss, causing auditory sensory deprivation over a band of frequencies, is present in over 90% of people with tinnitus. Existing plasticity-based computational models for tinnitus are usually driven by homeostatic mechanisms, modeled to fit phenomenological findings. Here, we use an objective-driven learning algorithm to model an early auditory processing neuronal network, e.g., in the dorsal cochlear nucleus. The learning algorithm maximizes the network’s output entropy by learning the feed-forward and recurrent interactions in the model. We show that the connectivity patterns and responses learned by the model display several hallmarks of early auditory neuronal networks. We further demonstrate that attenuation of peripheral inputs drives the recurrent network towards its critical point and transition into a tinnitus-like state. In this state, the network activity resembles responses to genuine inputs even in the absence of external stimulation, namely, it “hallucinates” auditory responses. These findings demonstrate how objective-driven plasticity mechanisms that normally act to optimize the network’s input representation can also elicit pathologies such as tinnitus as a result of sensory deprivation.

Genetic predisposition to tinnitus in the UK Biobank population

Tinnitus, the phantom perception of noise originating from the inner ear, has been reported by 15% of the world’s population, with many patients reporting major deficits to cognition and mood. However, both objective diagnostic tools and targeted therapeutic strategies have yet to be established. To better understand the underlying genes that may preclude tinnitus, we performed a genome-wide association study of the UK Biobank’s 49,960 whole exome sequencing participants to identify any loci strongly associated with tinnitus. We identified 17 suggestive single nucleotide polymorphisms (p < 1e−5) spanning 13 genes in two sex-separated cohorts reporting chronic, bothersome tinnitus (control males n = 7,315, tinnitus males n = 226, control females n = 11,732, tinnitus females n = 300). We also found a significant missense mutation in WDPCP (p = 3.959e−10) in the female cohort, a mutation which has been previously implicated in typical neuronal functioning through axonal migration and structural reinforcement, as well as in Bardet-Biedl syndrome-15, a ciliopathy. Additionally, in situ hybridization in the embryonic and P56 mouse brain demonstrated that the majority of these genes are expressed within the dorsal cochlear nucleus, the region of the brain theorized to initially induce tinnitus. Further RT-qPCR and RNAScope data also reveals this expression pattern. The results of this study indicate that predisposition to tinnitus may span across multiple genomic loci and be established by weakened neuronal circuitry and maladaptive cytoskeletal modifications within the dorsal cochlear nucleus.

Development of spontaneous firing of fusiform neurons from the dorsal cochlear nucleus of mice occurs after hearing onset

The dorsal cochlear nucleus (DCN) in the auditory brainstem integrates auditory and somatosensory information. Mature fusiform neurons express two qualitative intrinsic states in equal proportions: quiet, with no spontaneous regular action potential firing, or active, with regular spontaneous action potential firing. However, how these firing states and other electrophysiological properties of fusiform neurons develop during early postnatal days to adulthood is not known. Thus, we recorded fusiform neurons from mice from P4 to P21 and analyzed their electrophysiological properties. In the pre-hearing phase (P4-P13), we found that fusiform neurons are mostly quiet, with the active state emerging after hearing onset at P14. Subthreshold properties present more variations before hearing onset, while action potential properties vary more after P14, developing bigger, shorter, and faster action potentials. Interestingly, the activity threshold is more depolarized in pre-hearing cells suggesting that persistent sodium current (INaP) increases its expression after hearing. In fact, INaP increases its expression after hearing, accordingly with the development of active neurons. Thus, we suggest that the post-hearing expression of INaP creates the active state of the fusiform neuron. At the same time, other changes refine the passive membrane properties and increase the speed of action potential firing of fusiform neurons.

Sparsely Distributed, Pre-synaptic Kv3 K+ Channels Control Spontaneous Firing and Cross-Unit Synchrony via the Regulation of Synaptic Noise in an Auditory Brainstem Circuit

Spontaneous subthreshold activity in the central nervous system is fundamental to information processing and transmission, as it amplifies and optimizes sub-threshold signals, thereby improving action potential initiation and maintaining reliable firing. This form of spontaneous activity, which is frequently considered noise, is particularly important at auditory synapses where acoustic information is encoded by rapid and temporally precise firing rates. In contrast, when present in excess, this form of noise becomes detrimental to acoustic information as it contributes to the generation and maintenance of auditory disorders such as tinnitus. The most prominent contribution to subthreshold noise is spontaneous synaptic transmission (synaptic noise). Although numerous studies have examined the role of synaptic noise on single cell excitability, little is known about its pre-synaptic modulation owing in part to the difficulties of combining noise modulation with monitoring synaptic release. Here we study synaptic noise in the auditory brainstem dorsal cochlear nucleus (DCN) of mice and show that pharmacological potentiation of Kv3 K+ currents reduces the level of synaptic bombardment onto DCN principal fusiform cells. Using a transgenic mouse line (SyG37) expressing SyGCaMP2-mCherry, a calcium sensor that targets pre-synaptic terminals, we show that positive Kv3 K+ current modulation decreases calcium influx in a fifth of pre-synaptic boutons. Furthermore, while maintaining rapid and precise spike timing, positive Kv3 K+ current modulation increases the synchronization of local circuit neurons by reducing spontaneous activity. In conclusion, our study identifies a unique pre-synaptic mechanism which reduces synaptic noise at auditory synapses and contributes to the coherent activation of neurons in a local auditory brainstem circuit. This form of modulation highlights a new therapeutic target, namely the pre-synaptic bouton, for ameliorating the effects of hearing disorders which are dependent on aberrant spontaneous activity within the central auditory system.

3D Reconstruction of the Clarified Rat Hindbrain Choroid Plexus

The choroid plexus (CP) acts as a regulated gate between blood and cerebrospinal fluid (CSF). Despite its simple histology (a monostratified cuboidal epithelium overlying a vascularized stroma), this organ has remarkably complex functions several of which involve local interaction with cells located around ventricle walls. Our knowledge of CP structural organization is mainly derived from resin casts, which capture the overall features but only allow reconstruction of the vascular pattern surface, unrelated to the overlying epithelium and only loosely related to ventricular location. Recently, CP single cell atlases are starting to emerge, providing insight on local heterogeneities and interactions. So far, however, few studies have described CP spatial organization at the mesoscale level, because of its fragile nature and deep location within the brain. Here, using an iDISCO-based clearing approach and light-sheet microscopy, we have reconstructed the normal rat hindbrain CP (hCP) macro- and microstructure, using markers for epithelium, arteries, microvasculature, and macrophages, and noted its association with 4th ventricle-related neurovascular structures. The hCP is organized in domains associated to a main vessel (fronds) which carry a variable number of villi; the latter are enclosed by epithelium and may be flat (leaf-like) or rolled up to variable extent. Arteries feeding the hCP emerge from the cerebellar surface, and branch into straight arterioles terminating as small capillary anastomotic networks, which run within a single villus and terminate attaching multiple times to a large tortuous capillary (LTC) which ends into a vein. Venous outflow mostly follows arterial pathways, except for the lateral horizontal segment (LHS) and the caudal sagittal segment. The structure of fronds and villi is related to the microvascular pattern at the hCP surface: when LTCs predominate, leaflike villi are more evident and bulge from the surface; different, corkscrew-like villi are observed in association to arterioles reaching close to the CP surface with spiraling capillaries surrounding them. Both leaf-like and corkscrew-like villi may reach the 4th ventricle floor, making contact points at their tip, where no gap is seen between CP epithelium and ependyma. Contacts usually involve several adjacent villi and may harbor epiplexus macrophages. At the junction between medial (MHS) and lateral (LHS) horizontal segment, arterial supply is connected to the temporal bone subarcuate fossa, and venous outflow drains to a ventral vein which exits through the cochlear nuclei at the Luschka foramen. These vascular connections stabilize the hCP overall structure within the 4th ventricle but make MHS-LHS joint particularly fragile and very easily damaged when removing the brain from the skull. Even in damaged samples, however, CP fronds (or isolated villi) often remain strongly attached to the dorsal cochlear nucleus (DCN) surface; in these fronds, contacts are still present and connecting “bridges” may be seen, suggesting the presence of real molecular contacts rather than mere appositions.

Trigeminal Contributions to the Dorsal Cochlear Nucleus in Mouse

The dorsal cochlear nucleus (DCN) is the first site of multisensory integration in the auditory pathway of mammals. The DCN circuit integrates non-auditory information, such as head and ear position, with auditory signals, and this convergence may contribute to the ability to localize sound sources or to suppress perceptions of self-generated sounds. Several extrinsic sources of these non-auditory signals have been described in various species, and among these are first- and second-order trigeminal axonal projections. Trigeminal sensory signals from the face and ears could provide the non-auditory information that the DCN requires for its role in sound source localization and cancelation of self-generated sounds, for example, head and ear position or mouth movements that could predict the production of chewing or licking sounds. There is evidence for these axonal projections in guinea pigs and rats, although the size of the pathway is smaller than might be expected for a function essential for a prey animals’ survival. However, evidence for these projections in mice, an increasingly important species in auditory neuroscience, is lacking, raising questions about the universality of such proposed functions. We therefore investigated the presence of trigeminal projections to the DCN in mice, using viral and transgenic approaches. We found that the spinal trigeminal nucleus indeed projects to DCN, targeting granule cells and unipolar brush cells. However, direct axonal projections from the trigeminal ganglion itself were undetectable. Thus, secondary brainstem sources carry non-auditory signals to the DCN in mice that could provide a processed trigeminal signal to the DCN, but primary trigeminal afferents are not integrated directly by DCN.

Activity of CaMKIIa+ dorsal cochlear nucleus neurons are crucial for tinnitus perception but not for tinnitus induction

The dorsal cochlear nucleus (DCN) is a region known to integrate somatosensory and auditory inputs and is identified as a potential key structure in the generation of phantom sound perception, especially noise-induced tinnitus. Yet, how altered homeostatic plasticity of the DCN induces and maintains the sensation of tinnitus is not clear. Here, we chemogenetically decrease activity of a subgroup of DCN neurons, Ca2+/Calmodulin kinase 2α (CaMKIIα) positive DCN neurons, using Gi-coupled human M4 Designer Receptors Exclusively Activated by Designer Drugs (hM4Di DREADDs), to investigate their role in noise-induced tinnitus. Mice were over-exposed to loud noise (9-11kHz, 90dBSPL, 1h, followed by 2h of silence) and auditory brainstem responses (ABRs) and gap prepulse inhibition of acoustic startle (GPIAS) were recorded two days before and two weeks after noise exposure to identify animals with a significantly decreased inhibition of startle, indicating tinnitus but without permanent hearing loss. Neuronal activity of CaMKIIα+ neurons expressing hM4Di in the DCN was lowered by administration of clozapine-N-oxide (CNO). We found that acutely decreasing firing rate of CaMKIIα+ DCN units decrease tinnitus-like responses (p = 0.038, n = 11 mice), compared to the control group that showed no improvement in GPIAS (control virus; CaMKIIα-YFP + CNO, p = 0.696, n = 7 mice). Extracellular recordings confirmed CNO to decrease unit firing frequency of CaMKIIα-hM4Di+ mice and alter best frequency and tuning width of response to sound. However, these effects were not seen if CNO had been previously administered during the noise overexposure (n = 6 experimental and 6 control mice). 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 perception in mice.

Different Neurogenic Potential in the Subnuclei of the Postnatal Rat Cochlear Nucleus

In patients suffering from hearing loss, the reduced or absent neural input induces morphological changes in the cochlear nucleus (CN). Neural stem cells have recently been identified in this first auditory relay. Afferent nerve signals and their impact on the immanent neural stem and progenitor cells already impinge upon the survival of early postnatal cells within the CN. This auditory brainstem nucleus consists of three different subnuclei: the anteroventral cochlear nucleus (AVCN), the posteroventral cochlear nucleus (PVCN), and the dorsal cochlear nucleus (DCN). Since these subdivisions differ ontogenetically and physiologically, the question arose whether regional differences exist in the neurogenic niche. CN from postnatal day nine Sprague-Dawley rats were microscopically dissected into their subnuclei and cultivated in vitro as free-floating cell cultures and as whole-mount organ cultures. In addition to cell quantifications, immunocytological and immunohistological studies of the propagated cells and organ preparations were performed. The PVCN part showed the highest mitotic potential, while the AVCN and DCN had comparable activity. Specific stem cell markers and the ability to differentiate into cells of the neural lineage were detected in all three compartments. The present study shows that in all subnuclei of rat CN, there is a postnatal neural stem cell niche, which, however, differs significantly in its potential. The results can be explained by the origin from different regions in the rhombic lip, the species, and the various analysis techniques applied. In conclusion, the presented results provide further insight into the neurogenic potential of the CN, which may prove beneficial for the development of new regenerative strategies for hearing loss.