The Human MNTB: The Denial of its Existence and the Implications of this Claim

Many of the investigations of the human auditory brainstem left considerable doubt about the existence of the MNTB in humans, suggesting that it is absent or vestigial.

The first detailed investigation of the human SOC was published by Moore and Moore (1971), who also examined other primates. These authors utilized cresyl violet staining to investigate every seventh section (thickness = 30 μm) through a single human brainstem and found only occasional round or oval cells in the area ventral and medial to the (easily identified) MSO. They noted that these cells only occasionally formed a definable nucleus. A schematic figure of the SOC from a series of primates demonstrated the MNTB in all species examined (loris, marmoset, galago, owl monkey, macaque, and gibbon) except human. However, no images of putative MNTB neurons were shown nor was a figure of the human SOC provided. Strominger and Hurwitz (1976) examined every 10th section (thickness = 20 μm) from the brainstems of five adults, two infants and one fetus (using cresyl violet and/or luxol fast blue staining) and concluded that an “NTB” could not be defined. Notably, none of the figures in this manuscript depicted the region ventral and medial to the MSO and the reader is left only with the author’s interpretation. Strominger (1978) later went on to report some neurons “ventromedial” to the MSO which may represent the MNTB but indicated that these cells appeared cytoarchitecturally different from neurons in other species. Strominger thus concluded that the MNTB was poorly defined in human.

Moore (1987), in a further study of four human brainstems (examining every 10th or 20th tissue section [thickness = 50 μm] and using Nissl and myelin staining techniques), noted that a small number of globular neurons were found in the region where the MNTB should be located but that they did not form a nucleus. It is also stated emphatically that there was “no clearly definable nucleus of the trapezoid body” and that “no distinct cell group comparable to the trapezoid nucleus is evident in the human SOC.” The schematics of the SOC provided (Figures 8 and 9) labeled only the MSO and LSO. Again, no Nissl stained images of the SOC as a whole were shown (only Woelcke stained) and only a single high-magnification view of “periolivary” cells which were located along the rostral pole of the SOC was provided. Moore et al. (1999) later examined brainstem tissue from 12 human brains [ranging in age from 16 weeks prenatal (section thickness = 40 μm) to 17 years (section thickness = 75 μm)] with immunohistochemistry for choline acetyltransferase and calcitonin-gene-related peptide. Again, the authors stated that there was no identifiable MNTB in the human brain. Their schematics of the SOC include the MSO, LSO and periolivary regions. Again, no Nissl stained sections of the SOC were provided. Furthermore, in a review article (Moore, 2000) it was again noted that there was no apparent MNTB in the human SOC. Here, it was unequivocally stated that the MNTB “cannot be identified in over 70 human brainstems” and that small clusters of NTB-like neurons were observed in the fetus. In a textbook chapter (Moore and Linthicum, 2004), the MNTB was noted to be vestigial in humans and to consist of only a few scattered oval cells. The MNTB was not labeled in their MAP2 stained section of the SOC.

Qualitative immunohistochemistry was used by Bazwinsky et al. (2003) to study the human SOC in three subjects (the number or spacing of sections was never stated; section thickness = 50 μm). These authors described only scattered CB-IR neurons in the region ventromedial to the MSO and claimed they could not further characterize the neurons in this region. They conclude: “In our material, we could not identify a coherent nuclear region that might correspond to the MNTB of other mammals. Neither Nissl stained nor CB immunolabeling of brainstem sections provided evidence of the existence of this nucleus.”

Hilbig et al. (2009) examined Nissl stained brainstem sections (thickness = 20 μm) from bonobo, chimpanzee, gorilla, orangutan, gibbon, macaca, and human (n = 12). They examined every 10th tissue section (e.g., only 10% of the SOC) and described only a few scattered neurons ventromedial to the MSO in the human. Only a single, small low-magnification view of the human SOC was provided. Based on this study, these authors also denied the existence of the MNTB in the human brain. The authors further noted a decrease in the size of the LSO from orangutans to humans. They provided an estimate of the number of neurons in the human MSO (3,891) and LSO (1,980) but indicated there were no MNTB neurons. Interestingly, they also provided photographic evidence of a single fusiform neuron in the area ventromedial to the MSO surrounded by a perineuronal net, although the methods use to label this neuron were not provided.

Reviewing all these studies of the human SOC, we found few figures illustrating the region ventral and medial to the MSO that had sufficient resolution to actually identify or negate the claims made about the existence of the MNTB. It was typical in these studies for only every 10th or 20th tissue section to have been examined without any sort of quantitative analysis. Thus, the conclusion that the MNTB does not exist in human was made from examination of less than 10% of the SOC. Regardless, it is apparent from these limited studies of the human SOC that a number of globular/round neurons are indeed present in region ventral and medial to the MSO cell column. We declare that these studies have not provided adequate evidence to support their conclusion.

There is no doubt that the MNTB is the major source of directly sound driven synaptic inhibition in the SOC affecting all major nuclei (LSO, MSO, SPN/SPON) and the LL (citations see above) which in turn send significant projections to next level within the ascending auditory pathway, the inferior colliculus in the auditory midbrain. The fact that early in development MNTB neurons release GABA (with some co-transmission of glutamate; Gillespie et al., 2005; Noh et al., 2010) but after hearing onset release almost exclusively glycine (Kotak et al., 1998), relates nicely to the function: this circuit requires speed and precision (Moore and Caspary, 1983; Park et al., 1996). The GBC pathway with its comparatively thick axons (Morest, 1968), axonal peculiarities (Ford et al., 2012) the calyx of Held synapse with its morphological and physiological synaptic specializations (von Gersdorff and Borst, 2002) and the tendency to directly target the post-synaptic soma rather than dendrites (Clark, 1969; Cant, 1984; Kapfer et al., 2002) provides temporally secure and very fast transmission from the CN to the MNTB allowing for a basically 1:1 input:ouput relation (Mc Laughlin et al., 2008; Englitz et al., 2009). This results in fast and precise glycinergic transmission to the target nuclei in SOC and LL. Moreover, the post-synaptic properties in particular of MSO and LSO with almost unmeasurably low input resistances (only a few MOhms) ensure unusually short temporal integration times on the post-synaptic side. The post-synaptic effects of the glycinergic inhibition of the MNTB depend on three parameters. (1) The post-synaptic properties, e.g., membrane time constant and temporal integration properties (Magnusson et al., 2005; Couchman et al., 2010; Mathews et al., 2010; Scott et al., 2010); (2) the interactions with other, excitatory inputs and (Pecka et al., 2008; Myoga et al., 2014); (3) the frequency domain [(a) within the phase-locking domain below or above 2–3 kHz and (b) the animals need and ability to process interaural time differences for localizing low-frequency sounds – these two aspects co-vary but are not identical; Grothe and Pecka, 2014].

The most prominent of the MNTB projections is directed toward the LSO. This projection was also the first to be described functionally and studied developmentally (Moore and Caspary, 1983; Kandler and Friauf, 1995). As far as we know today, this circuit applies to all non-aquatic mammals (there the anatomy and, even more, the functionality of the SOC is more or less unknown and therefore not discussed in this review) and its one and only purpose is sound localization. For higher sound frequencies, the MNTB provides the inhibitory side of one of the most simple, although highly affective, neuronal computations known: a simple subtraction of the inhibitor effect driven by the contralateral side from the ipsilaterally driven excitation (via spherical bushy cells in the cochlear nucleus) that results in the well-known sigmoid interaural level difference functions and it has been found in every single mammalian species studied (reviewed in Grothe et al., 2010). In order to allow the inhibition to affect the onset of the ipsilateral, excitatory drive and to enable the circuit to completely inhibit LSO output if the sound arises from the contralateral side, the inhibitory pathway via GBC and MNTB needs to compensate for the much longer axonal distance (the LSO is in close proximity to the ipsilateral cochlear nucleus) and for the additional synapse in the MNTB. It seems apparent that this most prominent and among mammals universal function of the MNTB directly relates to the peculiar morphological and physiological specializations observed in this nucleus as well as its universal appearance in all mammals including humans (although during evolution ITDs became more relevant for us, as a low-frequency hearing mammal we are actually quite sensitive to ILDs with a resolution of sometimes below 1 dB; Grothe and Pecka, 2014).

Besides the need for a fast transmission and conductance from the contralateral side, the LSO also provides evidence for the need of extraordinary accuracy. Even bats, that do not hear low frequencies and do not use interaural time differences, where it is comparably safe to say that the function of the LSO is ILD coding and nothing else, the LSO neurons function on a millisecond by millisecond basis (Moore and Caspary, 1983; Park et al., 1996). At first glance this may be surprising. However, consider multiple sound sources that partially overlap in spectrum and time [e.g., in a group of bats flying in a cave, but similarly relevant for us (for instance at a cocktail party)]. Such overlaps (whether direct sounds or echoes) result in wired ILDs that rapidly fluctuate (Meffin and Grothe, 2009). Hence, the LSO must be fast enough to provide reliable information about the sometimes very short periods of unambiguous ILDs and temporal summation of ILDs over time need therefore to be avoided. In any case, the fact that the LSO processes ILDs with sub-millisecond precision creates an epiphenomenon: a rough ITD sensitive to the onsets of the sound envelope. Therefore it is not surprising that the same excitation–inhibition interaction that creates ILD sensitivity also provides a basic mechanism for processing ITD (Moore and Caspary, 1983). As far as we know, the low-frequency limb of the LSO seems to exclusively process ITDs based on the interaction of the two known inputs (Moore and Caspary, 1983; Tollin and Yin, 2005) and is almost certainly the basis for the widely known trough-type ITD sensitive cells at higher levels of the ascending auditory pathway (Kuwada et al., 1997; Siveke et al., 2006).

If we take into account that the early mammals after inventing the middle ear lived in a high-frequency world (“ultra-sound” is a very anthropocentric term since most mammals can hear far beyond 40 kHz and, at the same time, many do not hear low frequencies at all), and therefore relied on ILD but not ITD processing (review: Grothe and Pecka, 2014), it does not come as a surprise, that the ITD processing structure, the MSO, receives the same inputs as LSO, only complemented by mirrored inputs from the contralateral CN (excitation) and the LNTB (glycinergic). The MSO is even faster, meaning shorter membrane times constants and shorter synaptic currents (Magnusson et al., 2005; Couchman et al., 2010; Mathews et al., 2010; Scott et al., 2010), and with an even clearer spatial limitation of the inhibitory (MNTB and LNTB) inputs to the soma only (Kapfer et al., 2002; Werthat et al., 2008). Therefore it seems unlikely that the temporal aspect of the MNTB input does not matter, even if the in vivo results (which to date are indirect if it comes to the cellular mechanisms underlying ITD processing) are controversial. On the other hand, there is new evidence for even more peculiar morphological adaptations in the low-frequency GBC inputs to the MNTB in the form of an unusual myelination pattern deviating from the general assumption of structural similarity (Ford et al., 2012).

Less controversial is the fundamental role of the MNTB mediated glycinergic inhibition in temporal processing. The earliest evidence came from mustached bat homolog of the MSO that shows highly precise phasic on and off responses to contralateral sounds (Covey and Casseday, 1991). Later it was shown that these responses result from an interaction of a tonic SBC excitation and glycinergic, MNTB driven inhibition. Depending on the relative timing (in some cells the excitation is faster, in some the inhibition – note again that the additional synapse in MNTB is compensated for) either an off or on response results and can be turned into a sustained response by blocking the inhibition (Grothe, 1994). Later, similar interactions of contralaterally driven excitation and inhibition were shown in the rabbit, rat and mouse SOC (Kulesza et al., 2007). It is particularly prominent in the SPN/MNTB (here the MNTB inhibition competes with a precise onset input derived from cochlear nucleus octopus cells) and is believed to function in gap detection (Kulesza et al., 2007; Kopp-Scheinpflug et al., 2011).

Much less is known about the VNLL where, again, octopus cells from the contralateral CN interact with MNTB inputs (Spangler et al., 1985; Sommer et al., 1993; Smith et al., 1998). One function seems to be the creation of an inhibitory onset delivered to the inferior colliculus that has an unusual intensity independent fixed delay (Covey and Casseday, 1999).

Besides MNTB derived glycinergic inhibition there is a very similar, so far largely under-appreciated glycinergic inhibition driven by the ipsilateral ear, which seems to by and large mirror/compliment the contralaterally driven MNTB inhibition. In fact, it shows some comparable structural adaptations (GBC-input; large end-bulb of Held synapses – large terminals or calyx-like; Spirou and Berrebi, 1997; Spirou et al., 1998) and has, at least in the MSO, in vitro very similar features as the MNTB inputs (Grothe and Sanes, 1994). Besides this there are intrinsic connections in SOC that are GABAergic, many of them, however, acting via GABA-B receptors and, hence, rather modulating responses than being a main component of the precise spatio-temporal computations (Magnusson et al., 2005; Stange et al., 2013).

Taken together, the MNTB is the main source of precise inhibitory inputs targeting all prominent SOC and LL nuclei of the ascending auditory pathway. This glycinergic input contributes an essential component to all spatio-temporal computations known in the SOC. From this point of view, the idea of a missing MNTB in humans is neither plausible nor explainable.

Although novel sources for inhibition are not unthinkable and there is circumstantial evidence that other inhibitory neurons may contact the usual MNTB targets in a transgenic mouse model missing MNTB (Jalabi et al., 2013; Altieri et al., 2014), the extraordinary anatomical and biophysical specializations characteristic of the MNTB make it highly unlikely that other sources of inhibition could fully compensate for MNTB function.

Therefore, we propose consideration of the principle of parsimony. This principle states that, in consideration of historical homology, the explanation that requires the fewest assumptions and phylogenic transformations is most likely to be correct. As the MNTB has been demonstrated (without question) to exist in other primates (rhesus, baboon, bonobo, chimpanzee, gorilla, orangutan, gibbon; Bazwinsky et al., 2005; Hilbig et al., 2009; Kim et al., 2014), and exhibits rather unique structural and biophysical adaptations to serve its function, it seems most likely that this nucleus would also be present in the human SOC. To suggest otherwise would require that humans have miraculously developed, de novo, new mechanisms for instance for sound localization.

The authors would like to thank Nell Cant and Conny Kopp-Scheinpflug for their comments and suggestions on a draft version of the manuscript.