|
|
||||||||
Department of Neurobiology, Duke University Medical Center, Durham, North Carolina
Submitted 28 January 2005; accepted in final form 13 September 2005
| ABSTRACT |
|---|
|
|
|---|
| INTRODUCTION |
|---|
|
|
|---|
A specialized network of brain nuclei in oscine songbirds mediates learned vocal control and contains neurons strongly and selectively responsive to playback of the bird's own song (BOS) (Bottjer et al. 1984
; Doupe 1997
; Margoliash and Konisti 1985
; Nottebohm et al. 1976
; Vicario and Yohay 1993
), providing an attractive system to investigate stimulus-dependent synaptic interactions relevant to learned vocal communication. Among these nuclei, the telencephalic nucleus HVC is an especially important site for analyzing synaptic mechanisms underlying specific representations of self-generated vocal sounds because BOS-evoked firing patterns in HVC PNs are more sparse than in HVC's major auditory afferent, the interfacial nucleus of the nidopallium (NIf), and because HVC PNs respond more exclusively to the BOS than do neurons in NIf (Coleman and Mooney 2004
). Furthermore, HVC contains two different PN types (HVCRA and HVCX) that innervate functionally specialized pathways for song patterning or audition-dependent vocal plasticity, raising the possibility that auditory information transmitted by these two output cell types may be distinct (Fig. 1A) (Bottjer et al. 1984
, 1989
; Brenowitz 1991
; Gentner et al. 2000
; Nottebohm et al. 1976
). Indeed, although both HVC PNs discharge sparse bursts of action potentials to BOS playback (Mooney 2000
; Rosen and Mooney 2003
), BOS-evoked subthreshold activity differs in the two HVC PNs: HVCRA exhibit sustained synaptic depolarizations, whereas HVCX undergo sustained hyperpolarizations punctuated by brief depolarizing postsynaptic potentials (PSPs) (Mooney 2000
). These contrasting subthreshold response patterns likely reflect different synaptic specializations in HVC's two PN types that may relate to their distinct functional roles and that could modify auditory information transmitted to RA and area X.
|
To characterize functional synaptic interactions in HVC, we made dual intracellular recordings from pairs of HVC neurons in vivo and used spike-triggered averaging (STA) methods to examine their spontaneous and BOS-evoked synaptic connections. Consistent with the idea that functional interactions in HVC change in a stimulus-dependent manner, we found that all HVC neurons receive common excitation that becomes more tightly correlated across projection neurons during BOS playback. To determine the extent to which the common excitation seen across all HVC neuron types arises from HVC's local network or from HVC's afferents and to establish that BOS-evoked inhibition in HVCX cells derived from local sources, we compared BOS-evoked synaptic activity in HVCX and HVCRA cells while pharmacologically inactivating the entire HVC circuit or while selectively disrupting G-protein-coupled inhibition in HVCX cells. Either treatment abolished song-evoked hyperpolarizations in HVCX cells and increased the similarity of BOS-selective subthreshold response patterns in the two PNs. Therefore the two HVC PNs receive part of their common excitation from a BOS-selective extrinsic source, and BOS playback selectively recruits local inhibition onto HVCX neurons. These results indicate that the functional connectivity within HVC is altered distinctly for each PN type in a stimulus-dependent manner, which may enable HVC to transform information received from a common source (i.e., NIf) into two distinct representations transmitted to premotor and basal ganglia pathways.
| METHODS |
|---|
|
|
|---|
Subjects and stimuli
Experiments used 56 adult (>120 posthatch-days) male zebra finches (Taeniopygia guttata) in accordance with a protocol approved by the Duke University Institutional Animal Care and Use Committee. Prior to the experiment, birds were placed in a small recording chamber with a female, and song was recorded and edited to include two exemplary motifs using LabView software (all custom software for this study was written by M. Rosen and R. Balu). Stimuli were 1.53.0 s in duration and included the BOS and reversed BOS ((BOS-REV) i.e., song played backward, perturbing local and global temporal order, while maintaining spectral information). Prior studies have shown that HVC neurons are differentially responsive to these two stimuli both at a suprathreshold and subthreshold synaptic level (Margoliash 1983
, 1986
; Mooney 2000
).
Preparatory surgery, in vivo electrophysiology, song presentation, and drug application
On the day of electrophysiological recording, birds were anesthetized with 20% urethan in dH2O (90 µl total), further anesthetized with halothane for 1 min, and placed in a stereotaxic device, and a stainless steel post was mounted to the rostral skull with dental cement and reinforced with cyanoacrylate. Birds were immobilized via the mounted post in a sound-attenuating chamber on an air table, and temperature was maintained at 39°C using an electric warming blanket (Harvard Apparatus, Holliston, MA). The top layer of caudal skull was removed and HVC was localized by visual inspection and stereotaxic measurements, a small craniotomy (<500 µm) was made over HVC, and a small tear was made in the dura with a minuten pin.
Sharp electrodes were 60125 M
when filled with 2 M K-acetate, which in some cases was supplemented with 20 mM GTP
S or GDP
S to block BOS-evoked hyperpolarizations in HVCX neurons (Rosen and Mooney 2003
). (These nonhydrolyzable GTP or GDP analogues constitutively activate or inactivate, respectively, G-protein activated inward-rectifying potassium [GIRK] channels in other systems [Kurachi et al. 1986
], ultimately occluding or blocking endogenous GIRK-mediated synaptic inhibition.) The electrode was lowered into the nucleus (
50600 µm) using a one-dimensional hydraulic manipulator (Soma Scientific). For dual intracellular recordings, two AxoClamp 2B intracellular amplifiers were used in bridge mode to measure intracellular potentials, which were low-pass filtered at 3 kHz, digitized at 10 kHz, and stored on a PC. HVC neuronal types were identified on-line by their characteristic spike shape and firing patterns in response to positive current injection (Mooney 2000
). HVC shelf neurons were identified by the location ventral to the nucleus or by their DC-evoked action potential responses, which were different from HVC neurons. Cells were tested with 1030 iterations (at
70 dB) of the auditory stimuli if their resting potentials were negative of 55 mV and robust spontaneous synaptic activity was present. As noted in the results for certain cells, tonic negative or positive currents were injected through the recording electrode to shift the resting membrane potential of the cell.
For GABA inactivation experiments, a second micropipette [1020 µm tip, filled with 250 mM GABA (RBI, Natick, MA) dissolved in 0.9% NaCl] was lowered at a 30° angle through a second craniotomy to a point just ventral to HVC. The GABA was pressure-ejected with a Picospritzer (General Valve) in 10- to 200-ms pulses at 40 psi. GABA was applied during each interstimulus interval while recording, as it washed out very quickly (10120 s) otherwise. Concentrated GABA application strongly activates chloride and potassium conductances, shunting positive currents, thus greatly reducing excitatory synaptic drive. During GABA treatment, remaining synaptic activity in HVC was greatly reduced but still detectable, and the cells from which we recorded were unable to spike, thus removing their action-potential dependent contributions to local circuit activity. Effectiveness of the GABA inactivation was assessed during and after the application period by monitoring changes in spontaneous PSP amplitude as well as by monitoring the ability of a positive current pulse to evoke action potentials from the impaled cell. To ensure effective inactivation of the entirety of HVC, we used parameters of GABA application that were previously effective in entirely inactivating the song nucleus LMAN, which is of similar size to HVC. For LMAN, we confirmed that the whole nucleus was inactivated by recording from cells in disparate locations across the nucleus during GABA application (Rosen and Mooney 2000
). This spatial verification was not possible in HVC as the nucleus is close to the surface and the size of the craniotomy had to be kept to a minimum to maintain stable recordings. However, the positioning of the GABA pipette immediately ventral to HVC where diffusion should encompass all of HVC, along with application of identical concentrations and similar quantities of GABA as those used effectively on LMAN, suggest that HVC was effectively inactivated by this treatment.
Data analysis
SPIKE-TRIGGERED AVERAGES.
Briefly, a spike-triggered average (STAs) of one of the two cell's membrane potentials was calculated by averaging the median-filtered neuronal membrane potential of one cell relative to action potential activity in the other cell that exceeded a user-set trigger threshold (Fig. 1B); STAs were calculated for a time window ±200 ms relative to the trigger event. (Median-filtering the membrane potential removes high-frequency events such as spikes by replacing each point by the median value of the surrounding 50 points, equivalent to 5 ms at a 10-kHz sampling rate.) These STAs could reveal membrane potential deflections likely to reflect excitatory or inhibitory synaptic events associated with the timing of spikes in the other neuron of the pair (Fig. 1, C and D). To determine significance, spontaneous and stimulus-evoked STAs were compared with control STAs calculated by averaging the cell's membrane potential with respect to simulated regular spike trains of 115 Hz (set to match the average spontaneous or stimulus-evoked firing rate of the spiking neuron). These control STAs indicate the maximum amplitude over which uncorrelated events would appear, effectively capturing the variability inherent in the membrane potential of each cell, which was usually
1 mV. Deflections in the raw STA that exceeded the maximum or minimum of the control STAs were classified as significant events that could be used to further assess peaks and onsets of the raw STA; these STAs were generally >2 mV, well above the level of the controls (Fig. 1, C and D). The onset times of STA peaks were estimated by calculating the cumulative sum of the mean STA from 20 ms before and 30 ms after spiking in the other cell and designating the peak onset to be 5% of the maximum response (Coleman and Mooney 2004
) (Fig. 1, C and D, average of white circles). The onset times of STA troughs, which fell outside this time window, were estimated as the average of the time at which the membrane potential leading up to the trough minimum crossed the mean resting potential of the prespike portion of the STA (Vrest) and the maximum of the cumulative sum of the mean STA between a chosen time point clearly prior to the membrane potential descent and the time point of the trough minimum (these 2 measures often yielded identical onset times). The timing of the maxima or minima of STAs (i.e., the STA peaks or troughs) were measured relative to the trigger spike time. STAs were calculated only for those traces in which both cells remained stable and healthy (spike height was maintained and resting Vm remained negative to 50 mV). Two examples of typical paired recordings are depicted in Fig. 1E. The prespike portion of the STA depicts the mean resting potential (Vrest) of the cell the membrane potential of which was being averaged. The amplitude of an STA peak or trough is dependent on the difference between Vrest and the reversal potential of the activated currents. Therefore amplitude comparisons were always made between prestimulus and stimulus epochs that were immediately sequential to avoid any effects of any slow fluctuations in Vrest.
To fully characterize the nature of functional interactions observed for various neuron-pair types, STAs were calculated separately during silence (i.e., spontaneous activity), auditory stimulation with a behaviorally-relevant stimulus (BOS-evoked activity), and auditory stimulation with a control stimulus, BOS-REV (BOS-REV-evoked activity) (e.g., Figs. 4, B and E, and 6). All of the action potentials generated by the trigger cell during stimulus presentation were used to calculate each STA, and STAs for all three stimulus conditions were calculated across all cells within each neuron-pair type. To visualize the values representing the shapes of the STAs for all pairs, scatter plots were constructed where the onset times of the STA peaks/troughs were plotted at 0 amplitude, and these onset times were connected by a line to a point marking the time and amplitude of their corresponding peak or trough. Note that in some cases, significance was achieved for only one peak or trough per STA, thus scatter plots did not always contain equivalent numbers of peaks and troughs. These onset-to-peak scatter plots are included to show the variability across all cell pairs that contributed to the average STA (e.g., the variability of the average STA in Fig. 2B is represented by the scatter plot of contributing onsets and peaks in Fig. 2C).
|
|
|
For a better estimate of changes in stimulus-evoked neuronal interactions, additional comparisons were made on normalized STAs, where the time-varying effect on the membrane potential due to the auditory stimulus was first removed by normalizing each stimulus-evoked STA. Specifically, the median-filtered, mean membrane potential in response to several repetitions of the stimulus (BOS or BOS-REV) was subtracted from each individual trace, and STAs were calculated as described in the preceding text from these normalized traces. This allowed a more accurate estimate of cellular interactions than that obtained by comparing un-normalized STAs, which would be contaminated by activity driven by the auditory stimulus rather than by cell-cell interactions. To compare normalized traces to one another, BOS-evoked STAs were compared with BOS-REV-evoked STAs as a control stimulus rather than with spontaneous STAs, which could not be similarly normalized. Within-cell t-tests were applied to compare the peak amplitudes and latencies of STAs obtained during BOS-evoked and BOS-REV-evoked activity, and were computed on the subset of cells in which significant peaks/troughs occurred in both stimulus-evoked conditions (Fig. 6, Table 2).
|
To quantify each neuron's song selectivity, the psychophysical measure d' (Green and Swets 1966
), which represents the discriminability between two stimuli, was used to compare BOS with BOS-REV responses, where d'supra represents suprathreshold responsiveness and d'Vm represents subthreshold responsiveness (Rosen and Mooney 2003
; Solis and Doupe 1997
). A d' value of 0 indicates equal responsiveness to the two stimuli; values of 0.7 or 0.7 indicates a cell was significantly more or less selective to BOS than BOS-REV. Averages throughout the text are reported with the standard error of the mean (±SE).
CROSS-CORRELATIONS.
To quantify the similarity of the shape of the BOS-evoked membrane potential response either across or within cells before and after drug treatment, cross-correlations of HVCX versus HVCRA responses were calculated using MatLab software (MathWorks, Natick, MA) from the averages of the median-filtered membrane potential responses during 1030 iterations of BOS playback. Briefly, cross-correlation provides a measure of the correlation between two signals as a function of the temporal displacement between them and is useful for determining whether signals emanate from a common source and with what temporal delay. Before cross-correlating the membrane potential records of two cells, we subtracted the mean resting potential of each cell during a 2-s silent baseline period prior to the stimulus from that cell's record to remove the resting potential offset (i.e., BOS-evoked changes in membrane potential were measured from 0 mV). For some comparisons of the cross-correlations of subthreshold responses between the different HVC PN types before either local circuit inactivation or inhibitory disruption at the level of single HVCX cells, half-wave or full wave rectification was applied to the HVCX trace prior to calculating the cross-correlation to control for the effects of BOS-evoked hyperpolarizations on the cross-correlation value (see RESULTS for more specific details). In half-wave rectified traces, points negative to the mean resting potential were replaced by the mean resting potential, resulting in traces with depolarizing peaks interspersed with "flat" spots. In full-wave rectified traces, nonzero points were replaced by their absolute values, effectively reflecting hyperpolarizations around the mean resting potential. All cross correlations were divided by their autocorrelations (1 signal correlated with itself), yielding values between 1 (perfectly correlated) and 1 (perfectly anti-correlated, i.e., mirror symmetrical waveforms, with respect to membrane polarity). Thus the normalized cross-correlation R was calculated as
![]() |
| RESULTS |
|---|
|
|
|---|
STAs of the membrane potentials of one neuron of a pair were calculated relative to action potentials in the other neuron of the pair (Fig. 1B; see METHODS), to detect functional connectivity of the two cells (Fig. 1, C and D). Qualitatively, two major classes of interactions were observed. In the first class, the onset of a depolarizing STA occurred before the trigger spike time, and the STA peak occurred either slightly before (
5 ms) or after the spike time (Fig. 1C). This behavior could be explained by excitatory inputs to the two cells, wherein excitation sufficient to drive spiking in the trigger cell also simultaneously depolarized the other cell. The second major class of interaction we observed was one in which the onset and the peak negativity of a hyperpolarizing STA followed the spike time (Fig. 1D). This behavior could be explained by either a common source of excitation to both the trigger cell and an inhibitory interneuron innervating the other cell in the pair or by direct inhibitory interactions between the two cells.
Simultaneous dual intracellular in vivo recordings: all HVC cell types receive common excitatory inputs
Simultaneous intracellular recordings were made from 9 HVCRA-HVCRA pairs, 11 HVCX-HVCX pairs, 13 HVCRA-HVCX pairs, and 4 HVCX-HVCInt pairs. Examples of spontaneous and BOS-evoked activity from each pair type are depicted in Figs. 25. All pair types showed clear evidence of common excitatory inputs, i.e., the onset of a significantly depolarizing membrane potential movement in at least one cell of the pair reliably occurred prior to the spike time in the other cell (Fig. 1C; see METHODS). STAs revealed evidence of common excitation in all pairs of HVCRA neurons where at least one member of the pair was suprathreshold (n = 7 of 9 pairs; Fig. 2, B and E), in all 11 HVCX-HVCX pairs (Fig. 3, B and E), in 11 of 13 HVCX-HVCRA pairs (Fig. 4, B and E), and in 3 of 4 HVCX-HVCInt pairs (Fig. 5, B and E). Scatter plots of individual STA onset times and peak times (see METHODS) show that all positive peaks occurred near or past spike time, and that most positive onsets occurred prior to spike time (for onsets that lagged after the spike, a depolarizing STA onset occurred prior to spike time in the other cell of the pair, indicative of common excitatory input; Figs. 25, C and F). These data indicate that the majority of HVC cells receive excitatory inputs from a common source, while a minority may also receive additional lagging excitatory inputs. This excitation may arise extrinsically from NIf, intrinsically via connectivity within HVC, or from a combination of the two.
|
|
In addition to common excitation, these dual recordings helped us further elucidate the nature of BOS-evoked hyperpolarizing activity in HVCX cells. Indeed, HVCX neurons respond to BOS playback with intermittent, phasic excitation occurring on a background of sustained membrane hyperpolarization, clearly distinguishing these cells from HVCRA neurons. Sequential recordings from different HVC cell types in individual birds have shown that the timing of HVCInt excitation correlates with this hyperpolarization, indicating that HVCInt neurons may provide hyperpolarizing inhibition onto HVCX cells (Mooney 2000
). Inhibition onto HVCX cells from these HVCInt neurons also may be functionally recruited by one or both HVC PN types, an organizational feature which would result in negative STAs of HVCX membrane potential lagging after spikes in any HVC neuron type. To identify potential direct and/or indirect sources of inhibition onto HVCX neurons, we made dual recordings in HVC where at least one cell was of the HVCX type and exhibited BOS-evoked hyperpolarization.
BOS-evoked hyperpolarizing responses were detected in 13 of 22 HVCX neurons from 11 spiking HVCX-HVCX pairs, in 8 of 10 HVCX neurons from 10 spiking HVCX-HVCRA pairs (3 additional HVCRA neurons were subthreshold), and in 3 of 3 HVCX neurons from 3 spiking HVCX-HVCInt pairs (1 additional HVCInt neuron was subthreshold), yielding a total of 24 cell pairs for the analysis (see Figs. 3D, 4D, and 5D for BOS-evoked hyperpolarization). Of these 24 cell pairs, STAs measured during both spontaneous and BOS-evoked activity revealed a total of 23 HVCX neurons with significant negative peaks (12 of 13 from HVCX-HVCX pairs, 8 of 8 from HVCX-HVCRA pairs, and 3 of 3 from HVCX-HVCInt pairs; Figs. 3, B and E, 4, B and E, 5, B and E). The mean peak negativity of these STAs lagged after spikes by 20.24 ± 4.16 ms (raw data visible in scatter plots in Figs. 3, C and F, 4, C and F, 5, C and F; see Table 1 for breakdown by cell type). These data show that the timing of action potential activity in HVC PNs of both types as well as interneurons is correlated with a delayed hyperpolarization in HVCX neurons, and thus all three cell types could provide direct or indirect sources of inhibition onto HVCX cells. Similar functional organization has been revealed recently in paired recordings made from different HVC neuron types in brain slices (Mooney and Prather 2005
).
|
Simultaneous dual intracellular in vivo recordings: Interactions between neurons are significantly altered during BOS playback
In addition to analyzing the sign of interactions between cells in HVC, the dual recording methods allowed us to assess whether these functional connections changed in a stimulus-dependent manner. Prior studies have shown that stimulus-driven changes in synaptic interactions that may reflect or underlie stimulus processing exist in several systems (Ahissar et al. 1992
; Lampl et al. 1999
; Prut et al. 1998
; Vaadia et al. 1995
). Therefore we also examined whether the BOS, a particularly salient behavioral stimulus, might dynamically alter synaptic activity within HVC neurons by measuring the timing and amplitude of STA peaks in cell pairs before and during presentation of the BOS and other auditory stimuli. We compared BOS-evoked STA peaks with STA peaks during two control conditions: spontaneous activity and BOS-REV. The latter comparison allowed us to normalize both BOS and BOS-REV traces by removing stimulus-driven membrane potential movement, with the result that the STAs generated from these normalized traces more accurately represent functional connections between cells. Comparisons of unnormalized STAs evoked by BOS versus spontaneous activity are depicted as average STAs and individual peaks and troughs in Figs. 25, B and C, versus E and F. Comparisons of normalized STAs evoked by BOS versus BOS-REV are depicted in Fig. 6, which shows an example of BOS and BOS-REV-evoked supra- and subthreshold activity for a HVCX-HVCRA pair and the resulting normalized STAs from that cell pair, as well as average STAs and individual peaks and troughs across all cell-pairs. The unnormalized BOS versus spontaneous comparison and the normalized BOS versus BOS-REV comparison produced similar results.
While common excitation could be detected during spontaneous, BOS-REV, and BOS-evoked activity, BOS playback significantly shortened the onset time of the excitation in HVCRA-HVCX neuron pairs when compared with BOS-REV (Fig. 6C, top), and marginally shortened the onset time compared with spontaneous activity [Fig. 4, B and C, vs. E and F; normalized BOS-REV = 14.5 ± 0.5 ms, normalized BOS = 12.5 ± 1.1 ms, paired t(13) = 2.8, P = 0.008. Unnormalized spontaneous = 14.9 ± 0.4 ms, Unnormalized BOS = 13.7 ± 0.8 ms, paired t(15) = 1.6, P = 0.06]. This effect was significant across all cell pair types in BOS versus BOS-REV comparisons (normalized BOS-REV = 15.0 ± 0.2 ms, normalized BOS = 14.3 ± 0.4 ms, paired t(44) = 2.5, P = 0.008). In other cell-pair types (HVCRA-HVCRA, HVCX-HVCX, HVCX-HVCInt), the onset times of excitatory peaks in BOS-evoked STAs did not differ from those in control STAs. Compared with spontaneous activity, BOS playback also significantly reduced the peak amplitude in HVCRA-HVCRA pairs [unnormalized spontaneous = 6.7 ± 1.6 mV, unnormalized BOS = 4.5 ± 1.1 mV, paired t(8) = 3.4, P = 0.005], and across all cell pair types [unnormalized spontaneous = 3.5 ± 0.4 mV, unnormalized BOS = 2.7 ± 0.3 mV, paired t(46) = 2.7, P = 0.004], perhaps due to the previously described sustained positive shift in membrane potential evoked in these cells by BOS playback (Fig. 2, D and E) (Mooney 2000
). Consistent with this idea, the STA membrane potential 200 ms prior to the spike time was significantly more positive in the BOS playback than silent condition in HVCRA-HVCRA pairs [visible in Fig. 2, B vs. E; spontaneous = 67.6 ± 4.9 mV, BOS = 64.2 ± 5.3 mV, paired t(11) = 4.09, P < 0.001]. A breakdown of the timing and magnitude of excitatory peaks by pair-type across stimuli is shown in Tables 1 and 2. In summary, the functional synaptic interactions revealed by both normalized and unnormalized STAs indicate that the robust common excitation onto HVC PN neurons is temporally sharpened in HVCX-HVCRA cell-pair types during BOS playback, with depolarizing STA onsets occurring with shorter latencies relative to the action potential trigger time.
As inhibition onto HVCX neurons is most potently evoked by the BOS, we also examined whether in given cell pairs BOS playback alters the timing or magnitude of these hyperpolarizing events in comparison with spontaneous and BOS-REV-evoked activity. Similarly to depolarizing STAs where onset latencies decreased, the latencies of trough maxima decreased in HVCRA-HVCX pairs but only when BOS STAs were compared with spontaneous STAs [unnormalized spontaneous = 32.5 ± 10.2 ms, unnormalized BOS = 34.3 ± 14.5 ms, paired t(9) = 2.1, P = 0.04]; this effect was not significant across the population of cell-pair-types in BOS versus spontaneous comparisons nor was it significant in any BOS versus BOS-REV comparisons. Additionally, across all cell-pair-types as well as in several individual cell-pair comparisons, we noted that the amplitude of hyperpolarizing STAs in HVCX traces increased during BOS playback when compared with both spontaneous activity [Figs. 3, B and C, vs. E and F, 4, B2 and C2 vs. E2 and F2, 5, B2 and C2 vs. E2 and F2; across all pair-types: unnormalized spontaneous = 1.5 ± 0.3 mV, unnormalized BOS = 1.9 ± 0.3 mV, paired t(29) = 1.8, P = 0.037] and with BOS-REV-evoked activity [Fig. 6C, bottom; across all pair-types: normalized BOS-REV = 0.8 ± 0.2 mV, normalized BOS = 1.0 ± 0.2 mV, paired t(24) = 2.1, P = 0.02; see Tables 1 and 2 for comparisons across all cell pair types]. This change was not due to an increase in driving force, as the HVCX membrane potential did not differ during spontaneous activity versus BOS playback [Vrest in HVCX cell measured prior to spike onset in other cell: spontaneous = 67.4 ± 1.4 mV, BOS = 65.0 ± 2.6 mV, paired t(29) = 1.1, P = 0.14] and was actually more hyperpolarized in normalized BOS-evoked traces than in BOS-REV-evoked traces, which would decrease the inhibitory driving force [normalized Vrest: BOS-REV = 0.1 ± 0.3 mV, BOS = 0.2 ± 0.3 mV, paired t(24) = 1,9, P = 0.03]. For individual cell-pair comparisons, a significant negative peak in the HVCX-trace STA either emerged or increased in amplitude during BOS playback compared with either BOS-REV or spontaneous activity in 70% of HVCX-HVCRA pairs [a significant increase for BOS vs. BOS-REV comparisons: normalized BOS-REV = 1.1 ± 0.3 mV, normalized BOS = 1.4 ± 0.4 mV, paired t(7) = 2.0, P = 0.04], 71% of HVCX-HVCX pairs [significant increases for both comparisons: normalized BOS-REV = 0.7 ± 0.2 mV, normalized BOS = 0.9 ± 0.2 mV, paired t(13) = 2.0, P = 0.037; unnormalized spontaneous = 0.9 ± 0.2 mV, unnormalized BOS = 1.4 ± 0.3 mV, paired t(16) = 2.1, P = 0.028], and 67% of HVCX-HVCInt pairs (effective statistical analysis was not possible as only 3 pairs yielded STAs). A breakdown of the timing, magnitude, and significance levels of these inhibitory deflections by pair type across stimuli is shown in Tables 1 and 2. Therefore our data overall indicate a stimulus-driven increase in the amplitude of inhibition onto HVCX neurons, with some evidence for tighter timing in the peaks of hyperpolarizing STAs.
GABA inactivation reveals selectivity of afferents and differential effects of local circuitry
Our STA data identify distinctive subthreshold responses exhibited by HVCRA and HVCX neurons, consistent with common excitatory synaptic inputs onto both cell types and an additional lagging inhibitory synaptic input onto HVCX neurons. However, dual recording experiments are not sufficient to deduce the sources of these excitatory and inhibitory inputs on HVC neurons. Previous experiments have shown that abolishing activity in NIf can suppress much or all spontaneous and song-evoked synaptic activity in HVC, and simultaneous recordings in NIf and HVC suggest that NIf supplies either direct or indirect excitatory input onto all three HVC neuron types (Coleman and Mooney 2004
). These experiments could not rule out a model where NIf supplies direct excitatory input to only some HVC cell types and excitatory connections local to HVC supply other HVC neuron types with all of their excitation. In addition, these experiments could not exclude the possibility that NIf supplies qualitatively distinct (auditory) input to different HVC cell types. More specifically, the excitation common to both HVC PN types could be due to common input arising from NIf onto both cell types, with the differences between them due to differences in local (inhibitory) activity, or functionally distinct NIf inputs could segregate onto the two cell types.
These alternate models make specific predictions about the effects of inactivating the HVC local circuit. First, if some HVC cell types receive only indirect auditory input from NIf, inactivating the local HVC circuit should abolish their auditory responses altogether. On the other hand, subthreshold responses should persist and become more similar in HVC neurons on local circuit inactivation if NIf provides a relatively homogeneous source of auditory input to all HVC cell types. Finally, if local inactivation of HVC reveals persistent auditory-evoked excitation in both HVC PNs, then blockade of inhibition in HVC neurons at the single-cell level should also cause subthreshold responses in the two PNs to become more similar.
To test these various ideas, we first made intracellular recordings from individual HVC cells and then eliminated the contribution of HVC's local circuitry to their synaptic responses by extracellular application of GABA to the nucleus. Consistent with prior studies, we found that before GABA application, all cell types showed sub- and suprathreshold selectivity (as measured using d' analysis, see METHODS) for the BOS with hyperpolarizations readily detectable in the HVCX membrane potential responses. On application of GABA to HVC, both spontaneous and stimulus-evoked firing was abolished completely in all HVC cell types, disabling local, action potential-dependent contributions to synaptic activity in HVC. Despite such local circuit inactivation, auditory-evoked synaptic activity persisted in all neuronal types (Fig. 7). Moreover, these synaptic responses remained BOS-selective, confirming that all HVC neuron types receive at least some extrinsic sources of BOS-selective input. Furthermore, these extrinsic inputs are likely to be excitatory, because they depolarized the cell and were not occluded by the GABA application.
|
We quantified changes in HVCX neuron subthreshold selectivity in two ways. First, as previously mentioned, we used d' analysis to show that HVCX but not HVCRA cells display altered selectivity on inactivation of the local circuit (Fig. 7C, left vs. middle). Second, we noted that HVCRA and HVCX d' values spanned the same range during local circuit inactivation [Fig. 7C, left vs. middle, comparing GABA selectivity: t(19) = 1.04, P = 0.31], whereas the d' values for these two cell types were marginally distinct before inactivation [Fig. 7C, left vs. middle, comparing pre-GABA selectivity: t(19) = 2.00, P = 0.059]. These two analyses indicate that the relative bias of subthreshold responses in the two PNs become more similar when local circuit contributions to HVCX neurons are removed.
We also noted that the shape of the subthreshold response pattern changed following local inactivation and did so in a cell-type specific manner. Qualitatively, with the local circuit inactivated, BOS-evoked depolarizations in HVCX neurons became more sustained (Fig. 7A, 3 vs. 5) and more similar in shape to control responses in HVCRA cells recorded in the same bird (Fig. 7A, 3 vs. 2). In contrast, the overall shape of HVCRA BOS-evoked responses did not change markedly on local circuit inactivation (Fig. 7A, 1 vs. 2). Notably, the change in HVCX responses was not merely an effect of altered driving forces due to GABA-induced hyperpolarization, as the profile of the HVCX BOS-evoked response during DC-induced hyperpolarization in the absence of GABA (effectively removing the BOS-evoked hyperpolarization) did not closely mimic the profile of the subthreshold depolarizations recorded in the HVCRA cell from the same bird (Fig. 7A, 1 and 2 vs.7A5).
To more completely quantify changes in subthreshold response patterns in GABA-inactivated HVCX neurons, we calculated normalized cross-correlations at zero time lag of BOS-evoked median-filtered averaged membrane potential records of HVCRA neurons in control conditions and GABA-inactivated HVCX neurons within the same bird (see METHODS). These normalized cross-correlations were significantly higher than normalized cross correlations of membrane potential records from the same HVCRA neuron in control conditions and the same HVCX neuron prior to GABA treatment [with the HVCX cell held in a tonically hyperpolarized state; Fig. 9B, black bars; control hyperpolarized HVCX: 0.14 ± 0.07, GABA-treated HVCX: 0.35 ± 0.10, n = 9, paired t-test: t(8) = 2.2, P = 0.03]. These results suggest that the shapes of the BOS-evoked responses in the two PNs become significantly more similar when local circuit contributions to HVCX neuron responses are removed. To estimate an upper limit for membrane potential cross-correlation values between different HVC neurons, we also calculated subthreshold cross-correlations within homotypic pairs of either untreated HVCX or untreated HVCRA neurons within the same bird (the white bars in Fig. 9B). Notably, these homotypic pair cross-correlation values are quite similar to the across-PN-type cross-correlation values of GABA-treated HVCX neurons with control HVCRA neurons [Fig. 9B, compare right-most black bar with white bars; GABA-treated HVCX: 0.35 ± 0.10; HVCX vs. HVCX: 0.28 ± 0.05; HVCRA vs. HVCRA: 0.33 ± 0.03; 1-way ANOVA: F(3) = 0.78, P = 0.51]. Therefore local circuit activity accounts for much of the differences in subthreshold response patterns in the two HVC PN types that are detected by cross-correlation methods.
|
|
Both dual intracellular recordings and GABA inactivation of HVC suggest that a major distinction between the inputs onto HVCRA and HVCX neurons is the addition of a locally-generated, lagging, BOS-enhanced inhibition onto HVCX cells. If this is indeed the case, then blocking BOS-enhanced inhibition within single HVCX neurons should mimic the effects of blocking all local HVC activity with exogenous GABA application. Prior studies have shown that a G-protein-coupled potassium (i.e., GIRK)-mediated inhibition is a major component of song-evoked hyperpolarizations in HVCX cells and that this form of slow inhibition is likely to be driven by HVCInt cells (personal observations; Mooney and Prather 2005
; Rosen and Mooney 2003
). To directly test the extent to which GIRK-mediated inhibition could account for differences in subthreshold responses in the two HVC PN types, we recorded intracellularly from single HVCX cells with GTP
S or GDP
S in the recording electrode (see METHODS), nonhydrolyzable analogues of GTP and GDP, which, respectively, occlude or block GIRK-mediated inhibition in HVCX cells. With GIRK-mediated inhibition effectively disrupted, we then compared BOS-evoked subthreshold activity in these treated cells with subthreshold activity of HVCRA neurons recorded with control solutions in the same birds. Note that although these drugs have been used to block or occlude GIRK-mediated inhibition in HVCX cells in a previous report (Rosen and Mooney 2003
), the comparison of subthreshold responses in treated HVCX neurons with subthreshold responses in HVCRA neurons in the same bird is novel.
We observed that disrupting inhibition in HVCX cells unmasked an underlying excitation to BOS that was similar in shape to the excitation evoked by BOS in sequentially-recorded HVCRA cells of the same bird (n = 50 pairs; example depicted in Fig. 9A ). To quantify this effect, we again calculated cross-correlations of the BOS-evoked median-filtered averaged membrane potential records in HVCRA cells with membrane potential records in HVCX cells from the same bird at either early or late time points during inhibitory blockade. These cross-correlation values became significantly more positively correlated when inhibition was blocked or occluded by intracellular dialysis with nonhydrolyzable analogues of GTP and GDP [Fig. 9B, 1st gray vs. 4th gray bars; n = 51 pairs, inhibition intact: 0.25 ± 0.03, inhibition disrupted: 0.24 ± 0.04, t(50) = 10.0, P < 0.0001]. Interestingly, HVCRA-HVCX cross-correlation values following disruption of GIRK-mediated inhibition in single HVCX neurons were quite similar to those obtained following GABA inactivation of the entire HVC network [compare right-most black with right-most gray bars, Fig. 9B; GABA-treated HVCX: 0.35 ± 0.10; inhibition-disrupted HVCX: 0.24 ± 0.04; unpaired t-test, t(58) = 1.18, P = 0.24]. This similarity suggests that GIRK-mediated inhibition could account for much of the differences in subthreshold response patterns in the two HVC PNs. Indeed, cross-correlations indicate that the subthreshold depolarizing BOS-evoked responses in GTP
S- or GDP
S-dialyzed HVCX cells are as similar to BOS-evoked responses in control HVCRA cells as the responses recorded across homotypic PN pairs in control conditions are to one another [Fig. 9B, compare rightmost black and gray bars with white bars; GABA-treated HVCX: 0.35 ± 0.10; inhibition-occluded HVCX: 0.26 ± 0.03; HVCX vs. HVCX: 0.28 ± 0.05; HVCRA vs. HVCRA: 0.33 ± 0.03; 1-way ANOVA: F(3) = 0.78, P = 0.51]. As with the GABA inactivation experiments, these comparisons suggests that selectively removing local inhibitory components of the HVCX neuronal response can decrease, if not entirely abolish, the different response patterns in the two classes of HVC PNs.
We also investigated two models that may account for why the subthreshold response patterns of the two PN types become more similar on disruption of GIRK-mediated inhibition in