Patients that suffer from chronic tinnitus complain of an ongoing perception of a phantom sound in the absence of any physical source for it. About 5–15% of the population in western societies experience a phantom tinnitus sound and 1–3% of the population suffer from severe tinnitus that affects their daily life and is accompanied in 50% of the cases by depression, in 40% of the cases by insomnia and about 20% of the patients complain of an important decrease in their quality of life [1, 2]. Unfortunately, the underlying mechanisms responsible for the tinnitus perception is currently not known. Tinnitus therapies typically concentrate on coping with the tinnitus but there is no therapy that reliably reduces the perception of tinnitus.

Tinnitus is often accompanied by damage to the peripheral hearing system and a series of plastic changes in the central auditory system are observed in parallel to that. It is thought that a deafferentation of the hearing system triggers a series of reorganization processes at all levels of the auditory system [3]. Indeed, abnormal neuronal activity in tinnitus has been demonstrated for the auditory nerve fibers, the dorsal cochlear nucleus, the inferior colliculus, the primary and the secondary auditory cortex (see a review of this in [3]). Furthermore, it has been found that a dissection of the auditory nerve in tinnitus patients does not lead to relief in tinnitus and most of the patients still experience tinnitus after surgery [4, 5]. Thus, there is an agreement that the tinnitus phantom sound is generated in the central nervous system – most likely as a result of the reorganization that is going on in the auditory system after hearing loss.

However, there are also studies that demonstrated tinnitus-related cortical abnormalities outside the auditory system. Using methods as different as Positron Emission Tomography (PET), Voxel Based Morphometry (VBM) and Magnetoencephalography (MEG) differences in cortical activity have been shown for the frontal cortex, the parietal lobe, mesial posterior regions and the subcollosal region including the nucleus accumbens [6–8]. As hypothesized earlier by Jastreboff [9] it might be that tinnitus is generated within the auditory system while non-auditory regions are involved in encoding the conscious percept well as the emotional evaluation of it. This idea also fits with a recently established model of the global neuronal workspace by Deheane and colleagues [10]. This group suggests the existence of workspace neurons that are located mainly in the parietal lobe, the frontal, the cingulate cortex and the sensory systems. In order to form a conscious percept of a stimulus, two conditions are required: First, neuronal activity of the sensory cortex of the respective modality. Second, an entry into the global neuronal workspace and thus long-range coupling between the widely distributed workspace neurons. According to this model, coupling within this fronto-parietal-cingulate network is needed for conscious perception (i.e. awareness of the stimulus). Activity of the sensory areas without this coupling would remain preconscious.

Different brain regions need to communicate with each other in order to integrate information, perform their specific function and distribute information to other brain areas. It has been suggested that this communication is performed by neuronal synchronization between those brain areas and the functional importance of this inter-areal coupling has been shown in several studies [11–18]. In this literature, the terms 'coherence', 'synchrony', and 'coupling' are used with slightly different connotations. To avoid misunderstandings we want to use the term 'coupling' throughout this manuscript to describe the functional interaction between distant Neuronal Cell Assemblies.

The importance of long-range functional coupling has been shown recently by many authors in different fields of neuroscience. For instance, Supp et al. [17] demonstrated different patterns of long-range coupling in the gamma band between visually presented familiar and unfamiliar objects and Miltner et al. [15] found enhanced gamma band coupling during associative learning. Melloni et al.[14] used different masks to manipulate whether a test stimuli was visible or invisible to the participants. They found significant differences of gamma phase locking between the 'visible' and the 'invisible' condition. Hummel and Gerloff [13] showed an increase of alpha band coupling between occipital and left central areas correlates with behavioral performance in a visuotactile integration task. Uhlhaas and colleagues reviewed abnormal neuronal coupling in a large variety of brain disorders, namely schizophrenia, epilepsy, autism, Alzheimer's disease and Parkinson's disease [18]. In a behavioral experiment, Gross et al. [12] showed that changes in the inter-regional coupling vary with changes of the behavioral task demands.

To the best of our knowledge there is currently no study on long-range functional coupling in chronic tinnitus. In previous studies we investigated abnormal power changes in the spontaneous activity of tinnitus patients and found an increase in delta-power (<4 Hz) and a decrease in alpha-power (8–12 Hz). These changes were most prominent in the temporal region, however abnormalities were also found in the left frontal and right parietal cortex. This already suggested a frequency-specific long-range cortical network, however no measure of functional coupling was applied. In another study using Magnetoencephalography to describe power changes in the temporal cortex we showed an increase of gamma band activity in chronic tinnitus patients [19]. However, these changes were only calculated for time windows around slow-wave peaks and we did not investigate long-range coupling. Theoretically, power changes of Neuronal Cell Assemblies and coupling between them can be completely independent [20, 21] and thus we were not able to deduce knowledge about inter-regional coupling from these studies.

With the present study we aimed to investigate inter-areal functional coupling of spontaneous activity in tinnitus patients and to compare them with normal controls. Functional coupling was measured in a broad frequency range from 1 to 90 Hz by means of phase locking analysis. We used a phase locking method described by Lachaux et al. [22], which measures the phase difference between two recorded signals to quantify whether this phase difference is constant over time. A perfect coupling of the two signals results in a constant phase difference and is operationalized with a phase locking value of one. Lower values indicate weaker phase locking and the value of zero reflects no phase coupling at all.

Here, we first present substantial differences in the resting-state long-range functional coupling in chronic tinnitus sufferers. Specifically, two networks of different architectures and anti-correlated activity primarily account for the group differences. First, tinnitus patients are characterized by a decrease of phase couplings in the alpha frequency. Second, they display enhanced phase coupling in the 48–54 Hz gamma range. In both the tinnitus and the control group, there was a significant negative correlation between the alpha and the gamma network activity, suggesting an interplay of alpha and gamma coupling on an individual level. Furthermore, the duration of tinnitus seems to have an impact on the network architecture. In patients with tinnitus of short duration, gamma network changes are concentrated on the left temporal cortex. In contrast in the group with longer tinnitus duration, this network appears more widespread distributed over the entire cortex with lower impact of temporal areas.

Because the source montage that we used in this study covers only main areas of interest in the cortex, we are not able to interpretation of the precise location of the coupled sources. This is also because of technical constraints that are inherent to the inverse modeling used in MEG. However, the rough coverage of the brain however does not diminish the frequency-specific findings reported here. Also, we analyzed the power spectra of all sources to check whether they match with findings that were reported elsewhere [8]. Power spectra analysis of spontaneous resting-state data (eyes open) in tinnitus usually shows a reduction of alpha power and an enhancement of slow-wave power [8]. In this study, the alpha reduction was most pronounced for the temporal areas and to a smaller amount in the parietal and poster regions. The enhancement of the slow-wave power was localized mainly in the left temporal cortex. Overall, the alpha reduction was stronger than the enhancement of the slow-waves. In the additional material to this paper we report the grand average power spectrum over all sources (additional file 1) and the power spectra of all source locations (additional file 2) of the current analysis. The effects that we found in earlier studies were also observed in this analysis. Additionally, there was a slight increase in gamma power that was also found in another study of our group [32]. Furthermore, the additional file 3 gives a graphical illustration of the alpha power distribution (9–12 Hz) over the sources for the tinnitus and the control group. Occipital alpha power in resting state recordings with eyes open is usually smaller than in recordings with eyes closed. With this source montage, occipital alpha power is largely represented by the PCC-source. There was no significant group difference in the alpha power of this source (P > .2).

In this study, we found evidence for abnormal functionality in long-range cortical networks between tinnitus and control participants in the resting state, which are specific to the alpha and gamma frequency band. A general interaction between alpha and gamma power in the brain has already been postulated earlier [33]. It is assumed that alpha directly or indirectly reflects an intrinsic mechanism that prevents the build-up of gamma coupling within neural cell assemblies during deprivation from input. Functionally, such a mechanism appears to be necessary, as strongly interconnected excitatory oscillators would have a natural tendency to synchronize their activity. A deficiency of this mechanism is putatively an important prerequisite for the emergence of phantom perceptions. Our finding suggests that this relationship between alpha and gamma frequency is not limited to local power changes, but also might apply to inter-areal phase coupling.

We found that the inter-regional coupling of the alpha and gamma frequency bands discriminate well between the tinnitus and the control participants. Participants with a tinnitus perception are characterized by a decrease of long-range alpha coupling and an increase of long-range gamma coupling. Even though the discrimination of 83% is not sensitive enough to use it as an objective diagnostic tool for tinnitus, this is a strong argument that long-range couplings play an important part in the neuronal mechanisms associated with the tinnitus perception.

Here we propose a tinnitus model that integrates this finding with current knowledge on the tinnitus. On a first level the tinnitus is generated within the central auditory system and is most likely a result of reorganization processes triggered by damage to the hearing system. This is supported by numerous studies that show functional reorganization of the auditory system in tinnitus patients [3, 8, 34–36]. On a second level, abnormal coupling with higher-order brain regions outside the auditory system underlies its conscious perception [9, 10]. We assert that both levels are necessary for an ongoing perception of the tinnitus phantom sound.

Even though the alpha and gamma coupling discriminated well between tinnitus and control participants an association between the long-range coupling and the subjective degree of tinnitus distress was lacking. In an earlier study – also with resting-state recordings in the MEG – we found moderate correlations of the subjective tinnitus rating with alpha power decrease in temporal regions [8]. It is likely that we investigated two different neuronal mechanisms: One mechanism that is involved in the general perception of tinnitus and the other mechanism that is associated with tinnitus distress. The former study reported an association between temporal power changes and tinnitus distress. The current study on long-range coupling discriminated well between tinnitus perception and no tinnitus perception. Both mechanisms do not necessarily have to be associated.

With respect to tinnitus duration we found that longer-lasting tinnitus (> 4 years) accompanies marked changes in the pattern of the gamma network compared to shorter-lasting tinnitus. The most obvious difference between long and short-lasting tinnitus is a decrease in importance of the left temporal part of the network, i.e. there are fewer connections formed within this brain region. On the other side, functional connections between non-auditory areas are increased in tinnitus of longer duration. Based on the data that we present here we cannot decide whether this is only a change of functional coupling or whether structural changes also occur. A study using diffusion tensor imaging could help to clarify this question.

Notwithstanding whether the change in the network architecture is structural or functional, the results offer an explanation for a so far unresolved riddle in treating chronic tinnitus with Transcranial Magnetic Stimulation: It has been shown in a series of clinical studies that the efficacy of TMS treatment strongly depends on the duration of tinnitus [24, 25, 27, 37]. In these studies shorter tinnitus duration predicts better treatment outcome for therapeutical application of Transcranial Magnetic Stimulation applied while the treatment efficacy declines with longer duration of tinnitus. A tinnitus duration of 3 to 4 years seems to represent the turning point and patients below this point benefit only little from the treatment. The intriguing detail about this is that TMS is traditionally applied to the left temporal cortex. In the light of our findings these negative treatment effects make sense: as the gamma network shows a major hub in the left temporal cortex of patients with short tinnitus duration, stimulation of this region exhibits a potentially great impact on this network. However, since the gamma network is more widespread in patients with a long history of tinnitus, the impact of the stimulation to the left hemisphere is largely reduced. This idea of an alteration of the tinnitus-related neural network over time was hypothesized earlier [25] and the data that we presented here are the first experimental support for this idea.

Here we demonstrate for the first time alterations in the long-range network during spontaneous activity in tinnitus patients. The results can be described by an overall decrease of coupling in the alpha frequency band together with an increase of gamma coupling. This pattern of phase coupling discriminates with a high percentage (83%) the tinnitus patients from the healthy controls.

Here we suggest a tinnitus model that incorporates 1) altered activity of the central auditory system that most likely generates the tinnitus sound and 2) the coupling across distant brain regions that is needed for a conscious perception of the tinnitus sound.