Discussion
In this report, we have demonstrated that noise-induced tinnitus is
associated with an increased ECM density in primary auditory cortex
contralateral to the trauma side. This result may point to trauma
induced neuroplastic changes in AC, where an ECM density increase may be
an indication of tinnitus chronification.
The induction of a mild acoustic noise trauma – here monaurally – is a
reliable method of inducing cortical plasticity (e.g., Jeschke et
al. , 2021) and tinnitus-related behavioral changes in our rodent animal
model (e.g., Grimm et al. , 2022; Kiefer et al. , 2022;
Tziridis & Schulze, 2022). To ensure an objective analysis of the data,
only automated and parameter-free evaluation methods were used
(Schilling et al. , 2019) to rule out human evaluation bias
(Zaitoun et al. , 2016). With a mean HL of nearly 10 dB in the
exposed ear (12 dB in the most affected frequency of 2 kHz), the effect
of the trauma applied here was comparably mild when compared to other
studies (e.g., Tziridis & Schulze, 2022). This fact could also explain
the relatively low percentage of animals with behavioral signs of
tinnitus (4/9, 44%). These behavioral signs of tinnitus were determined
with a well-established statistical method (Schilling et al. ,
2017) but are no proof that the animals really perceive a human-like
tinnitus percept (Eggermont, 2013). On the other hand, the animals with
behavioral signs of tinnitus tend to show less hearing loss compared to
animals without these behavioral signs. This is in line with reports of
differences in human tinnitus patients compared to non-tinnitus patients
with hearing loss (e.g., Gollnast et al. , 2017).
Nevertheless, we always combine the behavioral data of possible signs of
tinnitus in our animal model with further physiological markers and
often find correlations between the tinnitus status of the animals and
those markers (e.g., Tziridis et al. , 2021), which indicates a
causal connection between behavior and physiological changes. In this
report, the further physiological marker is the density of the
extracellular matrix of the primary auditory cortices of the animals.
The ECM itself occupies between 10% and 20% of the entire brain volume
and plays an important role in genesis, plasticity and regeneration of
the central nervous system (Chelyshev et al. , 2022). Furthermore,
it is known that it plays a significant part in physiological as well as
pathological conditions (Vargova & Sykova, 2014; Krishnaswamy et
al. , 2019). The ECM can be divided into different subtypes:
perineuronal nets and perisynaptic extracellular matrix (Chelyshevet al. , 2022). Specifically for Mongolian Gerbils it has been
shown that the intact ECM in its entirety is important for the correct
function of the cortical layer dynamics and cross-columnar frequency
integration of the auditory cortex (El-Tabbal et al. , 2021) but
its density modulation (i.e., removing it) can also be used for inducing
neuronal plasticity and re-learning (Happel et al. , 2014).
With the knowledge of the ECM being that important for neuronal
plasticity (or possibly mal-plasticity in the case of tinnitus), we
investigated the ECM density of the auditory cortices of Mongolian
gerbils 13 d after acoustic noise or sham exposure by using a specific
ECM affine marker (WFA-FITC). This marker is a lectin that specifically
binds to chondroitin sulfates and thus labels chondroitin sulfate
proteoglycans within the ECM (Celio & Blumcke, 1994; Pizzorussoet al. , 2002). The fluorescence luminance of the marker is a
direct measurement of the ECM density. To correct for possible
differences in marker efficacy on the different histological slices, we
corrected the luminance by relating it to a reference region in the
brainstem. The 13 days after trauma were selected as a possible ECM
rearrangement is finalized after that time (Happel et al. , 2014)
and the neurophysiological changes in the auditory cortex – measured by
electrophysiological recordings – leading to a chronification of the
tinnitus percept are likely to be completed (Ahlf et al. , 2012;
Tziridis et al. , 2014). We found the clear evidence that animals
with behavioral signs of tinnitus have a much more solidified ECM
surrounding of their auditory cortex neurons than animals without these
signs or healthy control animals. This effect seems to be stronger in
the cortical hemisphere contra-lateral to the trauma ear, which can be
explained by the functional crossing of excitatory projections within
the auditory pathway to the contra-lateral side.
From the described effect of the ECM being more solidified in tinnitus
animals compared to control or non-tinnitus animals, several conclusions
can be drawn regarding potential mechanistic tinnitus models,
respectively. As it is not possible to provide a complete overview of
all existing tinnitus models, we here focus on the most common ones.
First, the effect points to the assumption that in contrast to former
models, chronic tinnitus is not a result of a re-organization of the
tonotopic map after a cochlear damage as claimed by Mühlnickel et al.
(Mühlnickel et al. , 1998), but is instead characterized by a
fixation of the wiring scheme after transient re-wiring in the cortex
(Ahlf et al. , 2012). Indeed, the hypothesis raised by Mühlnickel
et al. has been falsified by several other groups in various recent
studies, where huge tinnitus cohorts were investigated with fMRI (Koopset al. , 2020; Koops & van Dijk, 2021).
Second, our findings suggest that during tinnitus chronification the
responses of the neural system enter a neural attractor, which is
further reinforced by stiffening of the ECM in the auditory cortex. In
this context, the combined stochastic-resonance-predictive-coding model
introduced by Schilling and co-workers might provide an explanation
(Schilling et al. , 2023b). In a nutshell, the model describes
tinnitus as a result of intrinsically generated neural noise coming from
outside the auditory system (see also Zeng, 2013; Koops & Eggermont,
2021) used to compensate for reduced hearing thresholds resulting from
decreased cochlear output due to noise-induced cochlear damage (Krausset al. , 2016; Gollnast et al. , 2017; Krauss et al. ,
2017; Krauss et al. , 2018; Krauss et al. , 2019; Tziridiset al. , 2021; Schilling & Krauss, 2022; Schulze et al. ,
2023). Thus, a feedback loop implemented in the brainstem tunes for the
optimal noise level to maximize information transmission along the
auditory pathway by means of adaptive stochastic resonance (for details
see Krauss et al. , 2016; Schilling et al. , 2021). This
part of the model has already led to several testable hypotheses and was
tested in computer simulations (Schilling et al. , 2022). It
delivered a mechanistic explanation for the little brother of tinnitus,
the Zwicker tone (Zwicker, 1964; Schilling et al. , 2023a;
Schilling et al. , 2023c) and explains increased auditory
sensitivity after simulated hearing loss in an animal model (Krauss &
Tziridis, 2021), and provided individualized treatment approaches for
human patients (Schilling et al. , 2020; Tziridis et al. ,
2022). Nevertheless, this part of the model is a pure brainstem model
and does not account for the expected and described cortical effects. To
that means, it was assumed that the intrinsically generated neural noise
is amplified along the auditory pathway via central gain effects and is
then transmitted to higher brain structures such as the thalamus and the
auditory cortex.
Complementary to this bottom-up information flow, top-down mechanisms
seem to play a crucial role in tinnitus development and especially
tinnitus manifestation (Sedley et al. , 2016; Hullfish et
al. , 2019). Thus, the brain and especially the cortex is assumed to
operate as prediction machines trying to predict the cause of certain
signals transmitted via the auditory pathway. Thus, the brain makes a
default prediction (predictor for the Bayesian brain formalism), which
is silence under normal conditions. The product of predictor (silence)
and the top-down neuronal signal (likelihood) are the actual percept.
According to the combined stochastic-resonance-predictive-coding model
which unifies the bottom-up and top-down mechanisms (Schilling et
al. , 2023b), the increased bottom-up neural activity (neural noise)
caused by the stochastic resonance effect is mis-interpreted as neuronal
signal evoked by a real auditory stimulus. The noise thus leads to
increased activity and might also cause a decreased sensory precision –
i.e., higher mean and lower variance of the likelihood – as brain
states with low activity become very unlikely due to the continuous
neural noise. In other words, the neural noise leads to continuous
mis-predictions of the cortex with regards to incoming auditory signal.
Thus, the brain updates the predictions in a pathological manner and
starts to manifest a continuous auditory input as default prediction
(updated predictor), or attractor.
On the other hand, it is assumed that certain plastic changes in the
cortex could decrease the precision again or even prevent it from
happening at all and therefore reset / keep the default prediction to
silence (Sedley et al. , 2016; Hullfish et al. , 2019). The
here described results support this idea as in animals without tinnitus
the ECM density is not increased, neuronal plasticity is therefore still
possible. In tinnitus animals however, the new synaptic weights and the
mis-predictions (falsely updated predictor) are consolidated by the
solidified ECM.
In conclusion, we assume the tinnitus percept to be caused by increased
neural noise, mis-interpreted as auditory input and manifested via an
updated predictor (attractor) of higher brain areas. If this
mis-prediction is manifested via an increased ECM density there is no
possibility for the system anymore to change this wrong predictor, and
consequently tinnitus is perceived chronically.