Neuroestrogen synthesis modifies neural representations of learned song without altering vocal imitation in developing songbirds

Abstract

Birdsong learning, like human speech, depends on the early memorization of auditory models, yet how initial auditory experiences are formed and consolidated is unclear. In songbirds, a putative cortical locus is the caudomedial nidopallium (NCM), and one mechanism to facilitate auditory consolidation is 17β-estradiol (E2), which is associated with human speech-language development, and is abundant in both NCM and human temporal cortex. Circulating and NCM E2 levels are dynamic during learning, suggesting E2's involvement in encoding recent auditory experiences. Therefore, we tested this hypothesis in juvenile male songbirds using a comprehensive assessment of neuroanatomy, behavior, and neurophysiology. First, we found that brain aromatase expression, and thus the capacity to synthesize neuroestrogens, remains high in the auditory cortex throughout development. Further, while systemic estrogen synthesis blockade suppressed juvenile song production, neither systemic nor unilateral E2 synthesis inhibition in NCM disrupted eventual song imitation. Surprisingly, early life neuroestrogen synthesis blockade in NCM enhanced the neural representations of both the birds' own song and the tutor song in NCM and a downstream sensorimotor region, HVC, respectively. Taken together, these findings indicate that E2 plays a multifaceted role during development, and that, contrary to prediction, tutor song memorization is unimpaired by unilateral estrogen synthesis blockade in the auditory cortex.

Figure 1

Changes in neuronal density and aromatase and parvalbumin expression in NCM across development. (A) Aromatase, parvalbumin, aromatase parvalbumin co-expression, respectively, from an exemplar sensory-aged male bird (26 dph; right hemisphere; ventral NCM). Pseudo-colored: yellow, aromatase; cyan, DAPI; magenta, parvalbumin. Each image from a single slice of a z-stack taken at 60x magnification. Scale bar = 30 µm. White arrowheads indicate aromatase and parvalbumin co-expression. (B) Expression of aromatase, parvalbumin, and aromatase/parvalbumin co-expression, respectively, relative to the number of DAPI-positive nuclei (%), and parvalbumin co-expression relative to total aromatase expression (%). Overall, there are no significant differences in expression by age or NCM subregion. Circles = dorsal NCM; triangles = ventral NCM; green = sensory-aged birds; orange = sensorimotor-aged birds. (C) Total cell counts (DAPI-positive nuclei) across development; top row: sensory-aged bird (25 dph; right NCM); bottom row: sensorimotor-aged bird (71 dph; right NCM). 10x images taken from a 4 × 4 stitched image. Dorsal and ventral NCM images taken from a z-project max intensity 60x image. Note, only 60x images were quantified. (D), Cell density (total cell counts; DAPI-positive nuclei) by region and age. Dorsal NCM shows higher cell density than ventral NCM. Similarly, sensory-aged birds have higher overall cell density across subregions compared to sensorimotor-aged subjects. *p < 0.05; **p < 0.001.

Figure 2

Experimental timeline.

Figure 3

Systemic estrogen synthesis inhibition suppresses song production without impacting tutor song copying. (A) Daily number of song bouts before and across the tutoring/treatment period. (B) Birds sing amounts before treatment/tutoring; however, systemic FAD treatment reduces song production (p = 0.012). Circles/orange = saline-treated birds (n = 3); triangles/blue = FAD birds (n = 3). (C) Song similarity is lowest at 49 dph despite treatment (effect of age: p = 0.005; * is relative to 49 dph). (D) At 130 dph, tutor song similarity, accuracy, and sequence similarity, respectively, are all similar across treatments. (E) Change in Wiener entropy at 49 dph (post-tutoring day #5) relative to pre-tutoring values predicts eventual percent song similarity to the tutor at 130 dph, independent of treatment (r² = 0.903; p = 0.004). *p < 0.05.

Figure 4

Song copying is unaffected when neuroestrogen production is inhibited via in vivo microdialysis. 130 dph (A) song similarity, (B) accuracy, and (C), sequence similarity, respectively, are all comparable across aCSF- and FAD-treated birds. Cannula ‘surgery controls’ are graphed for visual comparison. Orange = aCSF; blue = FAD; grey = cannula; circle = left NCM; triangle = right NCM.

Figure 5

Juvenile male songbirds are similarly attentive to the tutor during microdialysis. (A) The time a bird spent near a live adult male tutor during in vivo microdialysis is similar across treatments, targeted hemispheres, and tutoring day. Behavior presented is from the first 10 minutes of song playback alongside live male presentation (see Methods). Orange = aCSF; blue = FAD; circle = left NCM; triangle = right NCM. (B) Correlogram of tutoring behavior and song similarity measurements reveal significant correlations (more time spent near the tutor negatively associated with time spent away from the tutor; tutor zone time positively correlated with tutor preference ratio), and novel findings (positive correlation of head scratching and drinking); p < 0.0005 (adjusted α; Bonferroni correction). Behavior data presented is from the first 10 minutes of tutor playback across days 1 and 2 of tutoring.

Figure 6

Single-unit recordings in NCM reveal modest differences in auditory responses in adulthood. (A) Representative NCM single-unit recordings from an aCSF and FAD in response to presentations of birds’ own song (BOS) and tutor song. Each recording includes a song spectrogram (Top), and raster plot (Middle) with corresponding peri-stimulus time histogram in 10 ms bins (Bottom) across a 6 second period. The same unit is presented for each treatment across the two stimuli. (B) Spontaneous firing rates were unaffected by developmental microdialysis treatment. Orange = aCSF; blue = FAD; grey = cannula; circle = contralateral hemisphere (relative to microdialysis site); triangle = ipsilateral hemisphere (relative to microdialysis site). (C) Stimulus-evoked firing rates were significantly lower for WN and overall higher for CON1. A recording hemisphere × treatment interaction was significant; however, post hoc analyses limited to CON1 found no statistical differences for either treatment. (D) Analysis of normalized auditory response (z-score) yielded a significant stimulus × recording hemisphere effect: contralateral NCM responded less to WN compared to all other stimuli, whereas forward conspecific stimuli elicited higher responses in the ipsilateral NCM, irrespective of treatment. (E) Ipsilateral d’ values relative to WN. BOS selectivity was higher in FAD songbirds in the ipsilateral hemisphere. BOS = birds’ own song; CON1; CON2 = conspecific song; REV-BOS = reverse bird’s own song; REV-TUT = reverse tutor song; TUT = tutor song. *p < 0.05.

Figure 7

Tutor song selectivity is elevated in single HVC neurons of formerly estrogen-suppressed adult songbirds. (A) Representative HVC single-unit recordings from an aCSF and FAD in response to presentations of birds’ own song (BOS) and tutor song. Each recording includes a song spectrogram (Top), and raster plot (Middle) with corresponding peri-stimulus time histogram in 10 ms bins (Bottom) across a 6 second period. The same unit is presented for each treatment across the two stimuli. (B) Spontaneous firing rates were similar across treatments. Orange = aCSF; blue = FAD; grey = cannula; circle = contralateral hemisphere (relative to microdialysis site); triangle = ipsilateral hemisphere (relative to microdialysis site). (C) Stimulus-evoked firing rates were significantly higher for BOS compared to all other stimuli except for TUT. Further, ipsilateral HVC displayed higher overall stimulus-evoked firing rates compared to contralateral HVC, independent of treatment. (D) Analysis of normalized auditory response (z-score) yielded similar results as with firing rate; namely, a significantly higher response to BOS over all other stimuli independent of treatment, as well as a significantly suppressed response to WN compared to CON2 and TUT. (E) Contralateral d’ values relative to CON1. TUT selectivity is significantly higher in FAD subjects solely in the contralateral hemisphere. BOS = bird’s own song; CON1; CON2 = conspecific song; REV-BOS = reverse bird’s own song; REV-TUT = reverse tutor song; TUT = tutor song. *p < 0.05.

Figure 8

Neural adaptation to learned song is reduced in adult NCM independent of post-training E2 synthesis inhibition. (A) An exemplar multiunit response in the NCM of an untreated hemisphere. Adjusted RMS declines at a faster rate (steeper slope) for novel song (CON3, CON4, and CON5) compared to a shallower slope (slower adaptation) for the recently exposed song (CON1). Slopes for each stimulus is shown at the bottom of each panel. (B) Average slope per stimulus in aCSF or non-treated hemispheres compared to FAD-treated hemispheres in NCM; slope derived from multi-unit RMS. Orange = aCSF or no treatment; blue = FAD; circles = novel stimuli (three unique CON per bird); triangles = trained stimulus (a single unique CON). The y-axis has been compressed for clarity and five slope data points were omitted (−3.37, −2.72, −2.56, −1.58, 0.51). * = significant main effect of stimulus type (novel vs. trained); p < 0.05. CON = conspecific song.