Input dependent modulation of olfactory bulb activity by GABAergic basal forebrain projections

Basal forebrain modulation of central circuits is associated with active sensation, attention and learning. While cholinergic modulations have been studied extensively the effect of non-cholinergic basal forebrain subpopulations on sensory processing remains largely unclear. Here, we directly compare optogenetic manipulation effects of two major basal forebrain subpopulations on principal neuron activity in an early sensory processing area, i.e. mitral/tufted cells (MTCs) in the olfactory bulb. In contrast to cholinergic projections, which consistently increased MTC firing, activation of GABAergic fibers from basal forebrain to the olfactory bulb lead to differential modulation effects: while spontaneous MTC activity is mainly inhibited, odor evoked firing is predominantly enhanced. Moreover, sniff triggered averages revealed an enhancement of maximal sniff evoked firing amplitude and an inhibition of firing rates outside the maximal sniff phase. These findings demonstrate that GABAergic neuromodulation affects MTC firing in a bimodal, sensory-input dependent way, suggesting that GABAergic basal forebrain modulation could be an important factor in attention mediated filtering of sensory information to the brain.


Introduction
The basal forebrain (BF) is a complex of subcortical nuclei with projections to various brain areas and has been implicated in attention and cognitive control. It constitutes the primary source of cholinergic projections to limbic structures, the cortical mantle and olfactory areas 1 . Cholinergic neuromodulatory systems are thought to enhance sensory processing and amplify the signal-to-noise ratio of relevant responses 2-5 e.g. the running-induced gain increases evident in sensory cortex 6,7 or the dishabituation of odor responses in the olfactory system 8 . Furthermore, they have been identified as key players in mediating attentional modulation of sensory processing as well as in coordinating cognitive operations 9,10 . However, the concept of a prevalent role of cholinergic cells in the BF was recently challenged as activity of non-cholinergic neurons was shown to strongly correlate with arousal and attention [11][12][13][14][15] . Despite the knowledge of BF subpopulations containing neurotransmitters different from acetylcholine [16][17][18][19] no direct comparison of modulation effects caused by cholinergic and non-cholinergic projections is currently available. Especially for sensory processing which is strongly influenced by attentional states 20 , effects of non-cholinergic BF modulation have been sparsely investigated.
The olfactory system in mice is heavily innervated by centrifugal inputs from the BF with the majority of bulbopetal neurons located in the horizontal limb of the diagonal band of Broca (HDB) [21][22][23][24][25][26] .
Though only about one fifth of BF neurons are cholinergic 25 studies on olfactory processing have mainly focused on cholinergic effects 8,[27][28][29][30][31][32][33][34][35][36][37][38][39][40][41][42][43][44][45] . In in vivo studies a specific activation of cholinergic HDB cell bodies was shown to inhibit spontaneous mitral tufted cell activity 27 while optogenetically activating cholinergic axons directly in the OB added an excitatory bias to OB output neurons: the enhancement of mitral/tufted cell odorant responses occurred independent of the strength or even polarity of the odorant-evoked response 28 . The effect of cholinergic fiber stimulation is reminiscent of sensory gain modulation in the form of baseline control 46 , which fits well to behavioral effects of nicotinic acetylcholine modulation in the OB 36 reported to increase behavioral discriminability.
Despite 30 % of the bulbopetal projections neurons in the BF being GAD-(glutamic acid decarboxylase, the rate-limiting enzyme in the synthesis of GABA) positive 25 , less attention has been directed towards GABAergic BF OB projections 23,47,48 . Using predominantly in vitro OB slice recordings, studies identified periglomerular interneurons 48 and granule cells 47 as targets of GABAergic projection.
Here, we used electrophysiological and optogenetic approaches to examine how cholinergic or GABAergic projections from BF modulate MTC output from the OB in vivo. We found marked differences between these projections; centrifugal cholinergic fibers from BF lead to an enhanced excitation of MTCs both at rest and in response to weak or strong sensory inputs. Effects of GABAergic BF axon stimulation in the OB on the other hand were sensory input strength dependent and mainly caused suppression of spontaneous MTC activity while predominantly enhancing odor evoked MTC spiking.
These results suggest that both, cholinergic and GABAergic projections from the same area, rapidly modulate sensory output but might have markedly different impacts on sensory information processing.

Differential expression of ChR2 in basal forebrain projection neurons
To selectively target cholinergic or GABAergic projections from BF to the OB we used mouse lines expressing Cre under control of the ChAT (ChAT-Cre mice; 49 ) or the GAD2 promotor (GAD2-Cre, 50 ). We expressed channelrhodopsin specifically in cholinergic or GABAergic BF neurons using a Cre-dependent viral expression vector targeted to BF by stereotaxic injection (Supp. Fig. 1). As reported previously 28,51,52 viral injection in ChAT-Cre animals led to ChR2-EYFP expression on the somata and processes of neurons throughout HDB and, to a lesser extent, the vertical limb of the diagonal band of Broca (Fig. 1A). In few preparations sparsely labelled neurons could be additionally observed in the magnocellular preoptic nucleus (MCPO, data not shown). BF-injected GAD2-Cre animals displayed ChR2-EYFP expression predominantly in the HDB (Fig. 1B). In fewer cases the MCPO showed a sparser cellular expression.
Four weeks after virus infection, ChR2-EYFP protein was apparent in BF fibers throughout the OB (Fig. 1C, D). In ChAT-Cre mice labelled axon terminals were visible in all layers of the OB (Fig. 1C left, Fig. 1D), consistent with earlier reports about cholinergic fibre distribution 25,28,[53][54][55][56] . The fluorescence intensity of EYFP per area unit was uniform across higher OB layers and declined in the granule cell layer. In GAD2-Cre animals the OB was also densely innervated by labelled fibres. Here, the fluorescence intensity per area unit was especially high in the glomerular and the granule cell layer ( Fig. 1C right, Fig 1D) recapitulating previous finding 47 but also strong in the mitral cell layer.
Fluorescence intensities were distinctly lower in the external plexiform layer, the main location of MTC / GC dendrodendritic synapses. The normalized fluorescence intensity per area unit of single fibers in ChAT-Cre and GAD-Cre OB was not significantly different (1.00 ± 0.06 and 1.06 ± 0.06, respectively; n = 3 mice, p = 0.52). Therefore, the differences in average fluorescence intensities reflect the difference in fiber density rather than ChR2-EYFP expression levels.

Optogenetic activation of cholinergic and GABAergic axons in the OB modulates MTC spontaneous spiking
To investigate BF modulation effects on early olfactory processing, we directed 473 nm light (1- To access the impact of cholinergic and GABAergic fiber stimulation on MTC excitability in the absence of sensory input, we optically activated BF axons without ongoing inhalation ( Fig. 2A). In this condition, MTC display an irregular firing pattern 28,57,58 . As shown previously 28 , optogenetic activation of cholinergic fibers in ChAT-Cre mice lead to a significant increase of spontaneous MTC spiking from 2.05± 2.34 Hz (mean ± SD) before stimulation to 2.40± 2.24 Hz during stimulation (n = 27 units from 5 mice; p = 0.0157 Wilcoxon signed rank test). 8 of these units (30%) showed a significant stimulationevoked increase in firing activity when tested on a unit-by-unit basis (Mann-Whitney U test); none showed a decrease (Fig. 2B, left).
In contrast, optical stimulation in GAD2-Cre mice led to a significant decrease in MTC spontaneous spiking, from 5.43 ± 4.07 Hz (mean ± SD) before stimulation to 3.45 ± 3.54 Hz during stimulation (n =44 units from 5 mice; p = 1.07 x 10 -8 , Wilcoxon signed rank test, Fig. 2A, B). When tested on a unit-by-unit basis 20 of the 44 recorded units showed a significant reduction in firing activity while none showed a significant increase (Fig. 2B, left). The median reduction in spike rate across these cells was 1.78 ± 1.36 Hz. Across the population of all recorded units, the decrease in spontaneous firing rate persisted for the duration of the 10 s optical stimulation (Fig. 2B, right). Following the stimulation, an increase in spiking was observed that returned to prestimulation levels within 20 s after stimulation ceased. Thus, optogenetic activation of GABAergic BF fibers at the level of the OB leads to a reduction of MTC activity while activating cholinergic fibers causes output neuron excitation, demonstrating that these subpopulations cause opposing effects on spontaneous MTC firing.
In order to rule out optical activation artifacts we stimulated the OB of uninjected control mice with the same parameters as before during the no sniff condition, since in this condition even small changes could have been detected (Supp. Fig. 2). We found that optical stimulation led to no significant change in spontaneous firing rate (n = 19 units from three mice; 6.10 +-4.15 Hz before stimulation, 6.20 +-4.25 Hz during stimulation (mean ± SD); p = 0.365, Wilcoxon signed rank test). Thus, the light- The averaged time course depicted an initial decrease in MTC firing rate that, in contrast to the time course in the no sniff condition, returned to prestimulation levels already during the stimulation period ( Fig. 2D). Following stimulation, spike rate increased above baseline levels for approx. 20 s before returning to baseline. Taken together, while optogenetic stimulation of cholinergic fibers in both conditions was qualitatively similar, activating GABAergic fibers leads to mixed effects of MTC spiking in the sniff condition that were not observed during spontaneous spiking.

MTC spiking
Since optical activation of BF inputs to the bulb modulates inhalation-linked MTC spiking consistent with modulating weak sensory-evoked responses, we next evaluated the impact of bulbar The BF receives input from different olfactory areas 45 and it has been shown that even during sleep and anesthesia, cholingergic and GABAergic BF neurons are rhythmically discharging [61][62][63][64] . We therefore tested the effect of inhibiting cholinergic and GABAergic BF projections to the OB using the light-gated chloride pump Halorhodopsin as an optogenetic silencer (Supp. Fig. 3). Despite robust, yet sparser expression of Halo-YFP, labelling could be observed in the OB for both ChAT-Cre and GAD-Cre mice four weeks after viral injection (Supp. Fig. 3A). Optogenetic stimulation during recording of presumptive MTCs (Supp. Fig. 3B) showed, when tested on a unit-by-unit basis, no significant effects in ChAT-Cre animals (8 units, 2 mice) while only two out of 28 recorded cells showed a weak but significant decrease in odor-evoked spiking (0.47 spikes/sniff/s) in GAD-Cre animals (Supp. Fig. 3C).
No significant modulation effects were observed in the other tested conditions (no sniff and sniff, data not shown). This was also true for units showing strong sensory evoked spiking, rendering it unlikely that effects went undetected due to low spike counts. The surprisingly weak effects of optogenetic inhibition might be the result of only weak spontaneous cholinergic and GABAergic OB fibre activity in the anesthetized animal.

GABAergic projections modulate OB output dependent on sensory input
Unlike spontaneous spiking, which got suppressed by optical stimulation in GAD-Cre mice, the same optical stimulation during odor stimulation predominantly caused MTC excitation (Fig. 3 F). In a separate set of experiments we therefore investigated if this suppression to excitation transition can also be observed on a single unit basis, or might be caused by a recording bias e.g. through different populations of output neurons being detectable in the different conditions. We recorded GABAergic modulation effects in individual MTC tested in both the spontaneous as well as the odorant evoked condition in one continuous session (Fig. 4 A and B depict recordings from the same unit).
As shown previously (Fig. 2), optical stimulation in the no sniff condition led to a significant reduction in MTC spontaneous spiking across all recorded units (3.61 ± 3.43 Hz before stimulation; 3.14 ± 3.37 Hz during stimulation; n =53 units from 3 mice; p = 0.014 Wilcoxon signed rank test). When tested on a unit-by-unit basis 5 of the 53 units recorded showed a significant reduction in firing activity and only two units showed an increase (Fig. 4C). Similar to the previous findings (Fig. 3), GABAergic axon activation during odor presentation had a predominantly excitatory effect on MTC activity (7.

Discussion
The basal forebrain is critical for many cognitive processes 65

Effects of basal forebrain projections on OB output cell activity
Recording from olfactory bulb output neurons, we show that, in contrast to a local and specific activation of cholinergic fibres, that add an excitatory bias to mitral/tufted cell firing, a selective activation of GABAergic BF fibres leads to bimodal, sensory input dependent effect on OB output: whereas optogenetic stimulation mainly inhibited spontaneous MTC firing, odor evoked MTC cell spiking was predominantly enhanced; an effect that could also be observed on a single neuron level.
Additionally, MTCs showed a reduction of firing outside and an increase of firing within the preferred sniff phase. This modulation is strongly reminiscent of a model of a bimodal gain change evoked by attention 9 also referred to as filtering. These filter processes are reported to dampen activity to nonattended or background stimuli while enhancing relevant sensory input. Indeed, in the odor condition, the size of the GABAergic modulation effect was dependent on the odor evoked firing activity for both excitatory and inhibitory odor responses, pointing to a multiplicative population firing change for relevant olfactory stimuli while decreasing background activity.
Our direct comparison of cholinergic and GABAergic OB fiber activation suggest that both BF derived fiber systems might have a role in gain modulation of OB output, but in a very distinct way: while cholinergic modulation seems to be rather similar to baseline control 46 , GABAergic modulation seems to lead to a filtering of weak signals in the OB.

Possible circuit mechanisms underlying bulbar GABAergic modulation
The circuit mechanisms underlying basal forebrain derived modulation effects remain to be elucidated since a comprehensive list of targets for BF fibers in the OB is missing especially for

Conclusion
Our recordings provide, for the first time, a detailed comparison of BF cholinergic and GABAergic influence on early sensory processing and highlight the potential of the noncholinergic BF population to modulate perception. By inhibiting weak and facilitating strong inputs, GABAergic BF fibres in the OB likely increase MTCs signal-to-noise ratio, a hallmark of attentional processes that have been previously attributed mainly to cholinergic processes [2][3][4]9,10 . Our findings are in line with recent data indicating that the classical view on the (cholinergic) BF system might be oversimplified: activity of non-cholinergic BF neurons was more strongly correlated with arousal and attention [11][12][13][14][15] , whereas cholinergic neuron activity was correlated with body movements, pupil dilations, licking, punishment 75,103 as well as primary reinforcers and outcome expectations 11 . The distinct early sensory modulation effects of cholinergic and GABAergic BF neurons observed in this study might therefore be owed to the different functions of these two basal forebrain systems. Addressing the exact interplay between ACh and GABA in olfactory bulb sensory processing in the awake animal will be critical to fully understand the relative contribution of each system.

Animals strain and care
We

Olfactometry
Odorants were presented as dilutions from saturated vapor in cleaned, humidified air using a custom olfactometer under computer control 28,106,107 . Odorants were typically presented for 10 seconds.

Extracellular recordings and optical stimulation
MTC unit recordings and optical OB stimulation were performed as described previously 28 with several modifications. Briefly, mice were anesthetized with pentobarbital (50 mg/kg) and placed in a stereotaxic device. Mice were double tracheotomized and an artificial inhalation paradigm used to control air and odorant inhalation independent of respiration [108][109][110] . Extracellular recordings were obtained from OB units using sixteen channel electrodes (NeuroNexus, A1x16-5mm50-413-A16, Atlas Neuro, E16+R-100-S1-L6 NT) and an RZ5 digital acquisition system (TDT, Tucker Davis Technologies). Recording sites were confined to the dorsal OB. Action potential waveforms with a signal-to-noise ratio of at least 4 SD above baseline noise were saved to a disk and further isolated using off-line spike sorting (Open-Sorter; TDT, Fig. 1F). Sorting was done using the Bayesian or (in fewer cases) K-Means cluster cutting algorithms in OpenSorter. Units were defined as "single units" if they fell within discrete clusters in a space made up of principle components 1 and 2. Units with interspike intervals lower than the absolute refractory period (< 2.5 ms) were excluded from further analysis 111 (Fig. 1G). For units to be classified as presumptive MTCs, units additionally had to be located in the vicinity of the mitral cell layer, show spiking activity in the absence of odorants, and a clear sniffmodulation (the maximum spike rate in a 100 ms bin had to be at least two times the minimum spike rate for the PSTH (Fig. 1G insets)); similar as described in 57 . Subsequent analyses were performed using custom scripts in Matlab. Odorant alone ('baseline') and odorant plus optical stimulation trials (at least 3 trials each) were interleaved for all odorants (inter-stimulus interval 40-50 s). Recordings were subject to unit-by-unit statistical analysis as described below.
For optical OB stimulation, light was presented as a single 10 -sec pulse either alone or simultaneous with odorant presentation using a 470 or 565 nm LED and controller (LEDD1B, Thorlabs) and a 1 mm optical fiber positioned within 3 mm of the dorsal OB surface as in earlier studies 28 . The light power at the tip of the fiber was maximal 3 and 10 mW for the 565 nm and 470 nm LED, respectively.

Extracellular Data analysis
Basic processing and analysis of extracellular data followed protocols previously described for multichannel MTC recordings 28 . Responses to optical or odorant stimulation were analyzed differently depending on the experimental paradigm. Stimulation effects on spontaneous spike rate (no artificial inhalation, "no-sniff" condition) were measured by calculating spikes / second (Hz) for the 9 sec before or during stimulation. Selection of 'sniff modulated' units was performed as described previously 28 .
Inhalation-evoked responses during inhalation of clean air ("sniff" condition) were measured by averaging the number of spikes per 1-sec period following each inhalation in the 9 inhalations prestimulation or during stimulation and across multiple trials (minimum of 3 trials in each condition for all units). Odorant-evoked responses were measured as changes in the mean number of spikes evoked per 1-sec inhalation cycle (Δ spikes / sniff) during odorant presentation, relative to the same number of inhalations just prior to odorant presentation. For statistical analysis, significance for changes in firing rate for baseline versus optical stimulation was tested on a unit-by-unit basis using the Mann-Whitney U test on units tested with 5 or more trials per condition.

Statistical analysis
Significance was determined using paired Student's t-test, Wilcoxon signed rank test and Mann-Whitney U test, where appropriate. Significance was defined as *P<0.05, **P<0.005, ***P<0.0005, ****P<0.0001. All tests are clearly stated in the main text.  G. Inter-spike-interval histograms for two units. One is a single unit (SU) and one a multi unit (MU).
Only single units were analysed further. For units to be classified as presumptive MTCs also a clear sniff-modulation in sniff triggered spike averages had to be present (inset).  A. Spike raster and rate histogram (bin width, 50 ms) depicting a spike rate decrease for a MTC during optical stimulation of the dorsal OB ("stim", blue shaded area).
B. The same unit plotted in A was also tested in the odor condition. Odorant-evoked spiking is enhanced by optical OB stimulation in this unit.
C. Plot of spontaneous firing rate in the 9 s before (no stim) and during (stim) optical stimulation for all units tested in both the no-sniff and the odor condition (n= 53 units). Squares indicated significantly modulated units subjected to a unit-by-unit test. 38 D. Plot of odorant-evoked spiking changes (∆ spikes/sniff) in the absence of (no stim) and during (stim) optogenetic stimulation of the same units tested in C.
E. Quantitative comparison of stimulation-evoked spiking changes (∆ spikes/sec) in the no-sniff and odor condition. Open circles, spiking changes for individual units; filled bars, mean value. Lines connect the same unit across conditions.