Synapse density and total number

The HVC volume did not differ significantly between 60 dph and 90 dph, regardless of the tutoring conditions. The only noticeable change in volume was significant growth between 30 dph and 60 dph. Therefore, for most of the groups, the HVC synapse density and synapse number reflected similar developmental dynamics.

Therefore, the HVC synapse density and synapse total number measurements and estimations (the total synapse number in HVC was not directly measured) will be discussed together in this section.

143

The synapse density in HVC changed significantly at 30 dph, 60 dph, and 90 dph during development in both the song-isolated and song-tutored conditions (see Table 3. 7). The total synapse number in HVC exhibited more distinct differences between 30 dph and 60 dph because both the HVC volume and HVC synapse density were low at 30 dph.

In the song-isolated groups, the HVC total synapse density first increased between 30 dph and 60 dph and then decreased at 90 dph to an even lower level than the density at 30 dph. The HVC total synapse number showed a significant (almost one-fold) increase between 30 dph and 60 dph and a significant decrease between 60 dph and 90 dph. In the song-tutored groups, the HVC total synapse density, interestingly, exhibited the opposite developmental pattern in which the density first decreased between 30 dph and 60 dph and then increased until 90 dph, with value still lower than the density at 30 dph. The HVC total synapse number exhibited a similar developmental pattern as was shown for HVC synapse density. The number first decreased between 30 dph and 60 dph and later increased between 60 dph and 90 dph. However, neither the initial decrease nor the later increase was significant (see Table 3. 16). At both 60 dph and 90 dph, the HVC synapse densities were significantly higher in the song-isolated group than in the song-tutored group. This experience-dependent synapse pruning in HVC recapitulated the findings of Experiment I (60 dph groups, see Results section 2.3.2 and Discussion sections 2.4.2 and 2.4.5) and further confirmed with the HVC total synapse number estimation results (see Figure 3. 16 and Table 3. 16). The experience-dependent pruning were still observed at later stages of song development (90 dph), but they were much less pronounced. The density and number of the synapses in HVC seems converging onto a similar level near the end of the song learning period, regardless previous song learning experiences of the birds.

The synapse density and number changes in HVC during song development will be further discussed for asymmetric and symmetric subtypes in the following discussions.

The HVC asymmetric synapse density remained at almost the same level between 30 dph and 60 dph under song-isolated conditions and only afterwards decreased significantly by 24% when it was measured at 90 dph. Considering that the HVC volume increased significantly between 30 dph and 60 dph (see Results section 3.3.2), there should be a matching and significant asymmetric synapse formation in HVC between 30 dph and 60 dph to compensate for the volume increase-induced density decrease. This can be seen in the estimations of HVC total asymmetric synapse numbers (see Results section 3.3.5). In summary, between 30 dph and 60 dph, there was a balanced increase in both the HVC volume and HVC asymmetric synapse number, resulting in the asymmetric synapse density almost unchanged.

After 60 dph, the asymmetric synapses in HVC was developmentally pruned to a lower number.

144

In the song-tutored groups, the asymmetric synapse density in HVC exhibited an initial drastic (up to 50%) decrease between 30 dph and 60 dph, and it then increased at 90 dph, although it was still much lower than the 30 dph value. The later increase in asymmetric synapse density in the song-tutored conditions seems in line with a very early study (Gilbert and Soderstrom 2011) that reported much higher HVC dendritic spine density in older adult birds. The age-matched comparison between song-isolated and song-tutored groups at 60 dph confirmed the findings in Experiment I (ISO and LONG group, see Results 2.3.2 and 2.3.3), that the asymmetric synapse density in HVC is much lower in the tutor song exposed birds than the song deprived birds. The total number estimation (see Results section 3.3.5, for ISO_60 and TUT_60) further supported the hypothesized tutor song experience-dependent structural pruning of HVC asymmetric synapses.

Notably, the asymmetric synapses in HVC were pruned first, and the synapses then started to increase in both density and number from 60 dph and 90 dph. Another study reported similar findings that, at 120 dph, the HVC dendritic spine density are 24% higher in the song-tutored condition compared with the song-isolated condition (Lauay et al. 2005), and the authors hypothesized this increase is a result of sensorimotor learning. In the mice motor cortex, motor training and enduring motor memories are observed to be associated with the rapid formation of excitatory synapses (Ma et al. 2016; Xu et al. 2009). In-vivo studies further supported this structure-behavior correlation that the formation of dendritic spines can be induced by repetitive motor training (Fu et al. 2012), and vice versa, the newly acquired motor skill can be disrupted by artificial shrinkage of the newly formed task-specific dendritic spines (Hayashi-Takagi et al. 2015). Thus these results established a direct link between new motor skills and the formation of excitatory synapses. Therefore, the increased asymmetric synapse density and number between 60 dph and 90 dph I observed could be a result of sensorimotor learning of the birds. It can be hypothesize that in juvenile zebra finches the song motor practice and memory are tightly associated with the newly formed asymmetric (excitatory) synapses in HVC during the sensorimotor learning period. Partially supported by the findings in the current study, this hypothesis can be further tested with artificial removal of the newly formed song-specific synapses in HVC and subsequent examining of the song performance in juvenile zebra finches. Alternatively, it can also be tested with a correlative LM and EM approach (Oberti et al. 2011) that would allow for precise identification of both the pre- and post- synaptic target of given synapses in HVC.

As hypothesized, one would expect to find that in HVC of song-tutored zebra finches between 60 dph and 90 dph, the major population of newly formed asymmetric synapses should originated from the HVCRA neurons, which are the premotor neurons of birdsong production (Vicario 1994). In addition, age-matched song-isolated birds should also exhibit such increase in the asymmetric synapse number from HVCRA neurons in HVC, which should further correlate with their isolated song production.

145

The HVC symmetric synapse density was initially (30 dph) very low, and it increased three-fold in the song-isolated condition and two-fold in the song-tutored condition at 60 dph. Later, at 90 dph, the HVC symmetric synapse density slightly decreased in both conditions. Therefore, in both the song-isolated and song-tutored conditions, the HVC symmetric synapse density exhibited a clear “peak-decline”

developmental pattern. These results indicate that HVC symmetric synapses develop slower than HVC asymmetric synapses and that there is a lot of synaptogenesis of symmetric synapses between 30 dph and 60 dph. Similarly delayed development of inhibitory circuits was also observed in developing rodent auditory cortex (Dorrn et al. 2010) that suggested to undergo progressive remodeling to sharpen the receptor field, balance the synaptic activity in local network, and limits further sensory exposure induced modifications near the end of the developmental critical period.

The age-matched comparisons between the song-isolated and song-tutored groups at both 60 dph and 90 dph indicated that song tutoring had a significant impact on the HVC symmetric synapses and always resulted in a much lower synapse density and number. These results indicated that tutoring also induced significant synapse pruning on the symmetric synapses in HVC at 60 dph and 90 dph. Especially at 60 dph, the HVC symmetric synapses were pruned by half (see Figure 3. 18) under the tutored condition. These differences were smaller in the later stage of song development (90 dph), which seems similar as observed in the asymmetric synapses that also converging onto a similar level near the end of the song learning period. It need to note that, in Experiment I the 1-day tutoring induced significant increase in the symmetric synapse density (see Results section 2.3.4 and Discussion section 2.4.2), which seems contradict the conclusion of experience-dependent synapse pruning of the symmetric synapses. Therefore, I should clarify here that the findings so far on the experience-dependent pruning are mainly obtained from the regular (24 days) song-tutored birds. For the findings from the birds in the 1-day tutored group (Experiment I, SHORT group), they should be treated specially as the ‘acute’ effect of the critical sensory experience exposure. I observed pruning in the asymmetric synapses in this group, but not in the symmetric synapses. On the contrary I observed an increase in the symmetric synapses which reflected an acute formation of inhibitory synapses in HVC shortly after initial tutor song exposure. This point has been discussed in the last chapter (see Discussion section 2.4.3 and 2.4.5), and will be mentioned in the summary in the end (Summary and Conclusion section 4.4).

Inhibitory neurons are added to HVC during song learning (Scott and Lois 2007), and not renewed in adulthood (Scotto-Lomassese et al. 2007). The time window of the formation of the inhibitory neurons coincides well with the peak formation of inhibitory (symmetric) synapses in HVC, which occurs from 30 dph to 60 dph.

During song learning, the HVC inhibitory interneurons reliably and precisely inhibit the premotor neurons when the learned portion of the song has been auditory perceived, which has been deduced to function as a mask to protect the learned information from sensory input (Vallentin et al. 2016). The estimation of HVC total inhibitory synapse number showed that total inhibitory synapse number increased more than two-fold at 60 dph compared to the number at 30 dph upon exposure to

146

the tutor song for 24 days, but the number decreased by half when compared to the tutor-deprived conditions. These results indicated that a dynamic remodeling process involving both inhibitory synapse formation and elimination shaped the HVC inhibitory networks during the early phase of song learning.

In adults, the HVC inhibitory networks interact with the excitatory networks to produce the learned song (Kosche et al. 2015; Mooney and Prather 2005). Therefore, the HVC inhibitory network is involved in both song learning and song production, and it should have access to both the tutor song information and the auditory feedback of the bird’s own vocalizations. A recently proposed model of song learning in zebra finches has suggested that HVC inhibitory networks act as a

‘stabilizing critic’ during song learning. When juvenile birds start producing their own songs, the HVC inhibitory networks synchronize its firing to reinforcing the correct motor output (Nick 2015).

Summarizing the synapse density and total number results of Experiment II and related studies together, HVC synapses generally exhibit a peak-decline developmental pattern in which they first increase to a higher value and then decrease. Significant experience-dependent synapse pruning on both excitatory and inhibitory synapses was observed in HVC at both 60 dph and 90 dph. The HVC excitatory synapses have been hypothesized to also be involved in song motor practice in the later stages of song learning, which result in the addition of new excitatory synapses. The HVC inhibitory synapses were rapidly formed between 30 dph and 60 dph, which coincided with the addition of new HVC inhibitory interneurons and the initial critical period of song learning. Both the HVC excitatory and inhibitory synapses seem to stabilize around 90 dph, when the initial changes at 60 dph are balanced and the differences between the song-isolated and song-exposed conditions are smaller or have disappeared. Because the HVC excitatory and inhibitory synapses undergo different temporal developmental patterns, the HVC excitatory and inhibitory balance is very likely to significantly shift during the song learning period. The HVC E/I balance and its role in song learning will be discussed with the symmetric synapse percentage estimation results in the next section.

3.4.4. Percentage of symmetric synapses

The percentage of HVC symmetric (inhibitory) synapses was very low at the beginning of song development (ISO_30), which was mainly due to the low HVC symmetric synapse counts. The percentage of symmetric synapses quickly increases from 30 dph on. In the song-isolated condition, the value quickly increased to 20%

and was stable afterwards. In the song-tutored condition, the HVC symmetric synapse percentage increased initially to 23% at 60 dph and then decreased significantly to a lower value at 90 dph. Comparisons between age-matched

147

song-isolated and song-tutored groups showed significant differences at both 60 dph and 90 dph (see Table 3. 14), indicating that song tutoring significantly affected the percentage of symmetric synapses in HVC during song development.

The measurement of HVC mean synapse size in Experiment I (see Results section 2.3.6) indicated that there were no significant differences between song-isolated and song-tutored birds at 60 dph. The mean synapse volumes and Feret diameters of both asymmetric and symmetric subtypes in HVC at 60 dph were similar regardless of song experience except for a nonsignificant increase in the size of HVC excitatory synapses 24 hours after the initial tutor song exposure. This transient sensory event was not present in any of the groups examined in Experiment II.

Therefore, these results suggest that there were no song experience-associated synapse size difference among the groups at the same developmental stages. The perentage of the HVC symmetric (inhibitory) synapses can be assumed to directly reflect the HVC E/I balance without needing to further consider the average synaptic strength. It can then be assumed in the following discussion that the percentage of the symmetric synapses, as estimated in Results section 3.3.4, reflects the HVC E/I level.

The maturation of HVC synaptic inhibition onto HVCRA premotor neurons correlates with learning but not age (Vallentin et al. 2016). This could partially explain the observed significant differences between the age-matched song-isolated and song-tutored groups. However, two additional observations of the current data need to be mentioned. First, significant age-related changes were observed in the HVC inhibitory synapse percentage, density, and total number, especially in the song-tutored conditions. Second, the tutored group showed different deviation directions at 60 dph and 90 dph compared with the song-isolated groups: the percentage of inhibitory synapse was higher at 60 dph but lower at 90 dph. Thus I observed the maturation of the inhibitory network in HVC are influenced by both song learning and age. The noted difference between the current anatomical data and the previous observation on the HVC inhibition might result from inhibitory synapses onto non-premotor neurons in HVC. In adults, pairwise electrophysiological recordings have revealed that HVC inhibitory interneurons target both HVCRA (premotor, 8 pairs found) and HVCX (nonpremotor, 20 pairs found) projection neurons as well as other HVCI interneurons (only 1 pair found) (Mooney and Prather 2005). The observed discrepancy could arise from synapses onto HVCX and HVCI neurons in HVC of juvenile zebra finches during song development, which are ignored in the Vallentin et al. study but should be sampled (as well as synapses onto HVCRA neurons, hypothetically. I do not have clue of the identity of the pre- and post- synaptic neurons in my data) in Experiment II.

The percentage of the HVC inhibitory synapse in the song-isolated condition was mostly stable at 60 dph and 90 dph, but it exhibited a significant developmental increase between 30 dph and 60 dph. In the song-tutored condition, this percentage value was much more dynamic. It initially increased to a higher level at 60 dph and

148

then decreased to a very low level at 90 dph. This pattern could reflect a general E/I balance shift of the HVC network during song learning that is modulated by song experience. In Experiment I, the HVC inhibitory synapse percentage reacted quickly (increased from 20% to 27%) to the sensory experience of the tutor song (see Results section 2.3.5 and Discussion section 2.4.3). Taken together this experience-dependent E/I balance changes might indicate the structural plasticity changes in HVC, as indicated by similar findings in the developing mouse visual cortex (Hensch 1998). This point is further supported by an observation in zebra finches, that increased pharmacological inhibition in HVC prematurely closed the song learning plasticity window in juvenile birds (Yazaki-Sugiyama et al. 2009).