Interplay of TC and CT system

reduced compared with the untreated condition. Even in the case that had a lesion in layer V only, the (because of the intact layer VI CT connections) still present granular sink was blurred, indicating that indeed the loss of CT neurons located in layer V was specifically responsible for this effect.

Though speculative, the most plausible explanation for the mechanisms behind these activations is the release from inhibition, i.e., the neuronal population targeted by layer V CT cells may normally contribute to inhibit certain brain regions, so that their elimination leads to a disturbance of the balance between excitation and inhibiton (shifts towards excitation), and in consequence to a more long-lasting activation of synapses in the AC. This activation, considering that there are different populations of (layer V) corticofugal cells [e.g., Ojima, 1994; Wong and Kelly, 1981; Bajo and Moore, 2005 (IC); Doucet et al., 2002 (SOC); Weedman and Ryugo, 1996 (cochlear nucleus); Budinger et al., 2000b, 2013 (various brainstem nuclei, CPu, amygdala); for review: Winer, 2006] is probably caused by the now dominant action of a corticofugal cell population that has been preserved and hence is still excitable by ICMS (see 4.4.2 and Fig.

4.4).

Since to our knowledge no distinct projections from layer V neurons to MGv have been characterised so far, it remains to be investigated whether the here probably observed minor connections from layer V are MGv-specific, or whether they represent collaterals of axons predominantly targeting other subcortical structures (e.g., MGd).

The minimal network includes afferent TC input to granular and to deep infragranular layers, mainly arising from MGv (LV, OV) (Huang and Winer, 2000; Hackett, 2011; Saldeitis et al., 2014), intracolumnar excitation through local intracortical recurrent microcircuits (Liu et al., 2007), intercolumnar short- and long-range horizontal intracortical connections (Ojima et al., 1991, 1992;

Kadia and Wang, 2003; Kurt et al., 2008; Moeller et al., 2010) (Fig 4.3, intracortical acitivity is represented by blue arrows) . In addition, a fast-acting CTC-loop is involved. Thus, the contribution of the TC projections to spectral processing depends not solely on the ascending inputs from brainstem, but also on the influence by CT feedback. The infragranular TC terminals, on their turn, probably reflected by iS1 in the CSD pattern, could modulate directly and fast, i.e. by bypassing the sequential cascade of information processing, the activity of CT neurons (see 4.3.2).

Our results indicate that (under normal conditions) a recurrent transthalamic loop emanating from layer VI CT neurons projecting to MGv is most likely responsible for sharpening frequency tuning of S1, doing this particularly by amplifying BF responses, rather than by additionally enhancing suppressive mechanisms to inhibit nonBF responses (i.e., lateral inhibition). We nevertheless included an inhibitory connection providing lateral inhibiton via the Rt or local inhibitory interneurons into our model, as it still may play a role in similar or other processes.

As aforementioned, the action of the recurrent transthalamic feedback loop becomes evident essentially for enhancing the ratio of BF to nonBF responses, while nearBF responses are not separately affected (i.e., at most along with a possible amplification of BF responses). This may be largely due to the divergence of TC projections (see 4.4.3).

Besides the involvement of CT neurons in frequency tuning of the initial input, they also seem to modulate the canonical interlaminar cascade of intracortical processing, which was manifested by a relative enhancement of the late extragranular sinks in lesioned animals. The influences they have on local intracortical processing may be accomplished by intrinsic collaterals to layer III-V (see 4.4.2).

4.4.2 Model of CTC interaction upon ICMS

To explain the electrically evoked phenomenons or differences that occurred after layer VI and/or V lesions, any suggested connectivity model had to stand all constellations, i.e., to account for the respective activation patterns in both lesioned and non-lesioned animals, both before and after cortical silencing.

There was no feasible model meeting all criteria that excludes the innervations of subcortical structures. Activation of AI via secondary auditory fields is also unlikely, since muscimol is expected to infuse and thus silence also the adjacent auditory fields.

Thus, we established a coherent explanatory model (Fig. 4.4) that includes the innervation of subcortical structures. How the different corticofugal circuitries interact is largely unknown. So, the model includes hypothetical mechanisms to both explain the generation of unreduced long-lasting (blurry) activity due to loss of layer V CT cells, and a mechanism which determines how the cortex represents this inherited activity.

We included in our model a further corticofugal target, namely the IC, in particular its cortices (in gerbil corresponding to DCIC/ECIC): (1) It receives considerable input from AC (Coleman and Clerici, 1987; Herbert et al., 1991;

Saldaña et al., 1996; Druga et al., 1997; Budinger et al., 2000b, 2013). (2) MGB and IC are bidirectionally interconnected (Senatorov and Hu, 2002; Kuwabara and Zook, 2000; Mellott et al., 2014). (3) Some of the CT axons that descend from layer V neurons (which are supposed to be eliminated here) may have collaterals that extend further to IC (Ojima, 1994; Lee et al., 2011; Slater et al., 2013). (4) IC neurons exhibit a row of distinct sustained firing patterns (Smith, 1992; Li et al., 1998; Sivaramakrishnan and Oliver, 2001), which may be required for the sustained cortical activation. (5) There are inhibitory mechanisms within the IC (e.g., Merchán et al., 2005; Nakamoto et al., 2013).

According to our model, ICMS excites all types of present corticofugal neurons.

When layer V CT cells are present, they will, via collaterals, activate GABAergic interneurons in IC, which in their turn will inhibit principle neurons of IC (projecting to MGB). As these neurons could also have been excited by an unbranched descending pathway to IC, they could shortly fire before being suppressed. In both cases of intact and lesioned layer V CT projections, the activated neuron will transmit its activity to AC via the thalamus, either in phasic

(possibly not distinguishable from direct CTC loop) or in more sustained manner, respectively. While MGd/MGm neurons can be driven by these colliculothalamic projections, the rather sparse projections from DCIC/ECIC to MGv (Mellott et al., 2014) suggest that these colliculothalamic projections are unable to drive MGv neurons, but sufficient to modulate (i.e., prolong) their response when activated by layer VI CT projections. Thus, the elimination of both layer V and VI CT neurons may cause a non-lemniscal-like cortical activation pattern (i.e., no granular sink).

Alternatively, it is currently discussed to which extent branched projections to both MGB and IC exist (Lee et al., 2011; Slater et al., 2013), though they were frequently considered to represent only a minor fraction of largely independent auditory corticofugal projection streams (Wong and Kelly, 1981; Ojima, 1994).

Therefore terminals of layer V CT cells could also activate thalamic inhibitory interneurons to quench the afferent signals from IC via presynaptic inhibiton.

However, the percentage of GABAergic neurons in MGB was found to be <1%

in rodents – although it has not been investigated in gerbils – (Winer and Larue, 1996), which may decrease the likelyhood and effectiveness of such a mechanism. The inclusion of additional synaptic relays, on the other hand, could make the circuitry too slow to account for the observed cortical patterns.

In AC then, the afferent input could lead to both (direct) excitation of cortical (e.g., excitatory pyramidal) cells and (indirect) inhibition, e.g., disynaptic feed-forward shunting inhibition (Sun et al., 2006) or feedback inhibition. This would explain why the elongated afferent input emerges only after muscimol (no intracortical inhibiton anymore), and in layer V lesioned animals (generation of persisting activation). In non-lesioned animals, the afferent input will be weaker and shorter even without intracortical inhibition, thus the model works for both animal groups.

As mentioned above, the differential CSD patterns of animals that have layer VI lesions and those which have not could be mediated by Rt (or inhibitory interneurons of MGB). In detail, collaterals of layer VI CT neurons could innervate Rt sectors projecting to MGd/MGm (Crabtree, 1998; Yu et al., 2004), which subsequently suppress these thalamic divisions. In this way, the electrically elicited cortical acitivity pattern of non-lesioned animals is largely determined by MGv inputs (acoustic-like appearance).

4.4.3 General principles governing auditory CT projections and functional implications for the CTC loops

Generally spoken, the CT system originating from layer VI neurons acts more locally (tonotopic, providing largely reciprocal feedback), while the system originating in layer V acts more globally (non-tonotopic, spreads activity widely across cortical fields in a feedforward manner) (see 1.2). Also their inputs (circumscribed vs. widespread; Llano and Sherman, 2009) and their intracortical axonal collaterals (vertical axonal arbors vs. horizontal wide ranging collaterals;

e.g., Ojima et al., 1992) fit into this scheme, as well as the distribution of terminals arising from the respective thalamic divisions (see 3.12 and 3.13).

Specifically, layer VI CT neuons are pivotal in providing fast feedback particularly to MGv. These CT cells, possibly being excited by direct TC synapses, are likely to mediate rapid adjustments related to various processing mechanisms (e.g., receptive field properties, gating, gain control, sound specific plasiticity (see 1.1). In particular, we found these cells to be involved in sharpening the frequency tuning in AI, probably via a CTC-loop through MGv (LV, OV).

Layer V neurons connecting to the non-lemniscal thalamus have for example been related to attentive functions (Yu et al., 2004). For instance, slow oscillations in non-lemniscal thalamus can be altered by AC, thereby possibly controling global alertness (He, 2003). Here, a supposed global role of layer V neurons may have appeared as the prolonged activity seen after cortical silencing in layer V lesioned animals, because they probably affected both lemniscal and non-lemniscal thalamus (nucleus-specific locations of electrically evoked cortical sinks depending on whether additional layer VI lesions were present or not). However, more animals with lesions restricted to layer V are required to ultimately confirm this effect.

Together, though the different CT systems hold specific functions (not at least determined by their distcinct intra- and subcortical connectivites), they do not appear totally separable. This implies that an interaction of several corticofugal systems on the (intra- and) subcortical level exist (see Figs. 4.2 and 4.3).