A dominant role for the beta 4 nicotinic receptor subunit in nicotinic modulation of glomerular microcircuits in the mouse olfactory bulb

Abstract

Nicotinic acetylcholine receptors (nAChRs) regulate information transfer across the main olfactory bulb by instituting a high-pass intensity filter allowing for the filtering out of weak inputs. Excitation-driven inhibition of the glomerular microcircuit via GABA release from periglomerular cells appears to underlie this effect of nAChR activation. The multiplicity of nAChR subtypes and cellular locations raises questions about their respective roles in mediating their effects on the glomerular output. In this study, we address this issue by targeting heteromeric nAChRs using receptor knockouts (KOs) for the two dominant nAChR β-subunit genes known to be expressed in the central nervous system. KOs of the β₂-nAChR subunit did not affect nAChR currents from mitral cells (MCs) but attenuated those from the external tufted (ET) cells. In slices from these animals, activation of nAChRs still effectively inhibited excitatory postsynaptic currents (EPSCs) and firing on MCs evoked by the olfactory nerve (ON) stimulation, thereby indicating that the filter mechanism was intact. On the other hand, recordings from β₄-KOs showed that nAChR responses from MCs were abolished and those from ET cells were attenuated. Excitation-driven feedback was abolished as was the effect of nAChR activation on ON-evoked EPSCs. Experiments using calcium imaging showed that one possible consequence of the β₂-subunit activation might be to alter the time course of calcium transients in juxtaglomerular neurons suggesting a role for these receptors in calcium signaling. Our results indicate that nAChRs containing the β₄-subunit are critical in the filtering of odor inputs and play a determinant role in the cholinergic modulation of glomerular output.

NEW & NOTEWORTHY In this study, using receptor gene knockouts we examine the relative contributions of heteromeric nAChR subtypes located on different cell types to this effect of receptor activation. Our results demonstrate that nAChRs containing the β₄-subunit activate MCs resulting in feedback inhibition from glomerular interneurons. This period of inhibition results in the selective filtering of weak odor inputs providing one mechanism by which nAChRs can enhance discrimination between two closely related odors.

INTRODUCTION

Cholinergic innervation of the main olfactory bulb (MOB) arises from the horizontal limb of the diagonal band of Broca (HDB) in the basal forebrain (Macrides et al. 1981; Záborszky et al. 1986) and terminates, most densely, in the glomerular layer (Le Jeune and Jourdan 1993; Salcedo et al. 2011; Zheng et al. 1987). Autoradiographic studies reveal a laminar distribution of various cholinergic receptors with the nicotinic acetylcholine receptors (nAChRs) being largely expressed in the glomerular layer while subtypes of muscarinic acetylcholine receptors (mAChRs) are denser in the external plexiform layer of the MOB (Le Jeune et al. 1995, 1996; Le Jeune and Jourdan 1991). The heterogeneity of nAChRs in the MOB represents a puzzle, and it is necessary to disentangle the functional significance of these various subtypes at various locations to uncover their disparate roles, if any, in circuit computations.

Disruption of nAChRs using pharmacological or genetic tools results in loss of odor discrimination (Chaudhury et al. 2009) and working memory (Young et al. 2007). Conversely, olfactory working memory was enhanced upon subcutaneous administration of nAChR agonists (Rushforth et al. 2010). Optogenetic examination of endogenous cholinergic stimulation suggests that ACh reduces basal excitability and sharpens the tuning properties of mitral cells (MCs; Ma and Luo 2012), though this effect might vary depending on the locus of stimulation (Rothermel et al. 2014).

Results from studies using acute MOB slices allow us to draw the following conclusions on nAChR modulation of the glomerular microcircuit: Pharmacological studies suggest that the major nAChR subtypes that participate in modulation of glomerular output are the β₂*containing-nAChRs (perhaps α4β2*-nAChRs) and the α3β4*-nAChRs (* indicates the possible presence of other subunits; D’Souza et al. 2013; D’Souza and Vijayaraghavan 2012, 2014). Activation of these nAChRs results in rapid excitation-driven inhibition via GABA release from periglomerular (PG) cells and, perhaps, short-axon (SA) cells, mediated by glutamate release from external tufted (ET) cells and MCs. These results place functional nAChRs mainly on ET cells and MCs. No nAChR currents were detected from PG cells (Parsa et al. 2015), consistent with the idea that ACh/At-evoked spontaneous inhibitory postsynaptic current (sIPSC) frequency increases observed in both ET cells and MCs were eliminated by glutamate receptor antagonists (D’Souza et al. 2013; D’Souza and Vijayaraghavan 2012).

Interestingly, GABA has bimodal actions on PG cells, being both excitatory and inhibitory, depending on the activity state of the cells (Parsa et al. 2015). Activation of a few PG cells by glutamate can result in an amplified release of GABA via GABA-induced GABA release (GIGR), thus providing a means for effectively inhibiting MC signaling (Parsa et al. 2015). The activation of this excitation-driven inhibition sets up a high-pass intensity filter such that responses to weak inputs are suppressed while those for stronger ones are transmitted through the glomerular layer (D’Souza and Vijayaraghavan 2012).

In this study, we dissect the mechanism of nAChR modulation further. Using receptor gene knockouts (KOs), we examine the role that different nAChR subunits expressed on different neurons play in the cholinergic modulation of the glomerular circuitry. We show a dominant effect for α3β4*-nAChRs on the MCs in triggering the excitation-driven inhibitory feedback. Our data suggests that excitation of nAChRs on MC apical dendrites sets the threshold for information transfer across the glomerular network.

METHODS

Animals.

Wild-type C57 Bl/6 mice were obtained from Charles River Laboratories. KOs for β₂ (Picciotto et al. 1995) and β₄ (Salas et al. 2004) nAChR subunits were obtained from Dr. Jerry Stitzel and maintained at the Institute of Behavioral Genetics, University of Colorado, Boulder as part an NIH-funded Resource Center. All procedures were approved by the University of Colorado Institutional Animal Care and Use Committee. Experimental animals were genotyped post hoc using tail snips. Genotyping was done by Transnetyx (Cordova, TN).

Electrophysiology.

We prepared 280-µm horizontal slices from postnatal day (P)12–P18 mice olfactory bulbs as described (D’Souza et al. 2013; D’Souza and Vijayaraghavan 2012). Recording solutions contained (in mM) 120 NaCl, 2.5 KCl, 26 NaHCO₃, 1 NaH₂PO₄, 10 glucose, 1 MgCl₂, 2 CaCl₂, adjusted to 285–290 mosmol/kgH₂O. All solutions were constantly bubbled in 95% O₂, 5% CO₂. Recording pipettes ranged from 5 to 8 MΩ. Unless otherwise stated, internal solutions contained (in mM) 123 K-gluconate, 2 KCl, 2 EGTA, 2 Na-ATP, 0.5 Na-GTP. Recordings were acquired using Axograph X and a MultiClamp 700B amplifier (Molecular Devices). Data were low-pass filtered at 2 KHz and acquired at 10 KHz. Access resistances were monitored to ensure the stability of recordings. Focal application of drugs were performed using a Picospritzer III (Parker Instruments, Cleveland, OH). Pressures used were less than 5 psi. Electrical stimulation was carried out using a Stimulus Isolator (WPI, Sarasota, FL).

The internal solution contained 1 mg/ml Lucifer Yellow (Sigma, St. Louis, MO) to fluorescently label the recorded neurons. Slices were imaged at regular intervals using a Cooke Sensicam and the SlideBook software to identify the cell being recorded and, in the case of MCs, to identify the relevant glomerulus. Post hoc morphological reconstructions and Sholl Analysis were conducted using the Simple Neurite Tracer (Ferreira et al. 2014; Longair et al. 2011) package in ImageJ (NIH, Bethesda, MD).

Calcium imaging.

Loading of slices with the calcium-sensitive dye Fura-2 AM was done as described (Grybko et al. 2010; Parsa et al. 2015; Sharma et al. 2008). Briefly, slices were incubated with 20 µM Fura-2 AM in 0.2% pluronic acid. Imaging was performed with a Zeiss Axiosop II fitted with a Cooke Sensicam camera and a Sutter DG IV wavelength switcher. Images were acquired at 1 Hz using the SlideBook software (3i, Denver, CO).

Data analyses.

Axograph was used to measure peak amplitudes and charge transfer of nAChR currents and sIPSCs and spontaneous excitatory postsynaptic current (sEPSCs). Calcium signals were analyzed using an in-house routine written in MATLAB or by using Origin (Origin Laboratory, Northampton, MA). Values are represented as means ± SE and significance was determined using a two-tailed Student’s t-test. Alpha was taken as 0.05. Before and after comparisons on same cells were made using paired t-test and significance was taken as P < 0.05. Comparisons between two separate populations were done using unpaired t-test and significance was taken as P < 0.05. Multiple comparisons were done using single-factor ANOVA. Distributions were analyzed using the Kolomogorov-Smirnoff (K-S) test. Considering the small number of conditions (3), no Bonferroni correction was applied (Armstrong 2014; Perneger 1998) to account for familywise errors.

Drugs.

Atropine (2 µM, Sigma, St. Louis, MO) and QX-314 (1 mM, Tocris: Bio-Techne, Minneapolis, MN) were used throughout all experiments unless otherwise stated. Conotoxin AuIB (CTx; Alomone Laboratories, Jerusalem, Israel) was used at 10 µM. All other drugs were from Tocris (Bio-Techne). Other drugs were used at following concentrations: 10 µM DNQX, 50 µM APV, 1 µM TTx, and 10 µM GBz.

RESULTS

Stable whole-cell voltage clamp recordings were obtained from 95 cells (MC n = 54, ET n = 34, PG n = 23). Cell types were differentiated based on a combination of morphological and intrinsic electrical characteristics (Fig. 1). Mean input resistances (±SE) were 97.09 ± 5.09, 241.25 ± 17.79, and 1,488.63 ± 191.76 MΩ for recorded MC, ET, and PG cells, respectively (Fig. 1B). Mean whole-cell membrane capacitances were 66.16 ± 4.84 pF, 39.0 ± 2.74 pF, and 19.42 ± 1.04 pF for MC, ET, and PG cells, respectively (Fig. 1B). The extent of arborization for each cell type was determined using Sholl analyses (Fig. 1, C and D). ANOVA tests did not indicate any significant differences in electrical properties between knockout groups within any cell type (data not shown). Anatomically, ET cells were characterized by the lack of lateral dendrites. No further tests were done to look for subpopulations among each class of neurons.

Fig. 1.

Morphological and intrinsic properties of recorded cell types. A: example neuron reconstructions from each of the cell types used in this study. B: mean intrinsic electrical properties differed substantially between cell types. Note that R_in is plotted on a logarithmic scale due to the magnitude of difference between cell types. C: mean normalized Sholl analysis of dendritic reconstructions. Dendritic crossings were assessed at 15-µm intervals from the center of the soma. N = 7 for each cell type, translucent underlays denote SE. D: mean distance to peak arborization from analyses in C. Distances from soma to maximal dendritic densities (± SE) for each cell type were 271.97 ± 18.10, 91.56 ± 5.41, and 47.86 ± 8.92 µm for mitral (MC), external tufted (ET), and periglomerular (PG) cells, respectively.

Functional distribution of nAChR subtypes within the glomerular microcircuit.

Previous studies indicate nAChR-mediated excitation of both ET cells and MCs might contribute to excitation-dependent feedback inhibition from PG cells (D’Souza et al. 2013; D’Souza and Vijayaraghavan 2012). To determine the relative role of MCs and ET cells in the nicotinic regulation of the glomerular output, we used receptor knockout mice for the two major nAChR β-subunit genes (β₂ and β₄).

Recordings from MCs in wild-type (WT) mice revealed direct inward nAChR currents in response to a brief (1 s) application of 1 mM acetylcholine in the presence of 2 µM atropine (ACh/At) at the glomerulus. The use of atropine ensured that mAChRs were not activated in these experiments. Charge transfer was calculated by integrating the current trace for 5 s from the onset. The average charge transfer was 431 ± 81 pA·s (n = 9; Fig. 2, A and B). The nAChR response from MCs in β₂-KOs was not significantly different from that obtained from WT bulbs (418 ± 156 pA·s; n = 9; P = 0.938, two-tailed, unpaired t-test), suggesting that this receptor subtype plays a minor, if any, role in nAChR currents at the MC primary dendritic tufts. On the other hand, MC recordings from β₄-KOs showed that there was no significant response to ACh/At in these neurons (4.6 ± 4.15 pA·s; n = 10; compare with 431 ± 81 pA·s in MCs from WT mice, P = 0.0008, two tailed t-test, Fig. 2). One-way ANOVA for the three conditions gave a P value of 0.005, F value 6.4.

Fig. 2.

Effects of β-subunit knockout on nicotinic acetylcholine receptor (nAChR) currents A: example traces depicting nAChR currents from mitral (MC), external tufted (ET), and periglomerular (PG) recorded from wild-type (WT), β₂ knockout (KO), and β₄ KO mice. All cells were held at V_m = −70 mV and were subjected to a 1-s ACh/At puff directed at the dendritic tuft of the cell. ACh puffs are depicted with blue lines; columns are arranged so that all traces align with ACh puffs. Timeframe of charge transfer measurements (5 s) depicted by gray lines at the top of the figure. Scale bar 100 pA/2 s. B: mean charge transfer values measured over a 5-s period from the beginning of the ACh/At puffs. Charge transfer values are expressed as pA × seconds. Note the differences in y-axis values. **P < 0.01, ***P < 0.005, ****P < 0.001. ACh/At, 1 mM acetylcholine in the presence of 2 µM atropine.

nAChR activation also produced small and variable currents in ET cells (also see D’Souza et al. 2013). In MOB from WT animals the net charge transfer was 65 ± 18 pA·s (n = 9). This was reduced in β₂ KOs to 33 ± 11 pA·s (n = 8; P < 0.05 compared with WT) and in β₄ KOs (19 ± 4 pA·s; n = 7, P < 0.002, compared with WT; Fig. 2, A and B). One-way ANOVA gave a P value of 0.2 with F = 1.6. No significant currents were observed from PG cells in either the WT animals or the two KO animals (Fig. 2, A and B) as demonstrated previously.

Overall, and consistent with our previous work (D’Souza et al. 2013; D’Souza and Vijayaraghavan 2012; D’Souza and Vijayaraghavan 2014), the ET cells provide only a small fraction of the nAChR-mediated currents within the glomerular microcircuit, which is mainly driven by MC excitation. While previous pharmacological experiments suggested that the ET cells mainly express α4β2-nAChRs (D’Souza et al. 2013), the results presented here from β₄- KOs suggest that this subunit also participates in the nicotinic signal observed in ET cells, perhaps more so than the β₂-subunit. This issue is revisited below.

No significant currents were observed in PG cells, upon ACh/At application (Fig. 2, A and B, not significantly different from 0), consistent with our previous studies showing that nAChR driven excitation of PG cells is primarily triggered by glutamate release from MCs and/or ET cells (D’Souza et al. 2013; Parsa et al. 2015).

Calcium contributions from β₄-containing nAChRs in juxtaglomerular neurons.

nAChRs significantly contribute to calcium signaling in neurons (Sharma and Vijayaraghavan 2001, 2003; Vijayaraghavan et al. 1992; Zhang et al. 1994). To examine calcium signaling in juxtaglomerular cells expressing nAChRs, we examined responses to exogenously applied ACh/At in Fura-2 AM loaded glomeruli. These experiments were conducted in the presence of GluR blockers (10 µM DNQX, 50 µM APV), 1 µM TTx, and 10 µM GBz to block GABA_A receptors that could also contribute to calcium signals (Parsa et al. 2015). Under these conditions, one would nominally expect the signals to arise from cells expressing functional nAChRs.

Application of ACh/At resulted in calcium transients in a subpopulation of juxtaglomerular cells (Fig. 3, A and B). Similar number of cells responded in β₂ KOs. However, the number of cells responding to nAChR activation is dramatically reduced in β₄ KOs (Fig. 3, B and C). While ET cells from β₂ and β₄ KOs both express reduced nAChR currents (Fig. 2), the β₄- containing nAChRs seem to contribute disproportionately to the calcium signals. The calcium response in the β₂-KOs was drastically reduced by incubating the slices with 10 µM conotoxin AuIB, an inhibitor of the α3β4* class of receptors (87 ± 8.8% inhibition, n = 4 experiments; Fig. 3, D and E) suggesting that the signals in the β₂-KOs mainly arose from these β₄-containing nAChRs.

Fig. 3.

Effects of β-subunit knockouts on ACh/At-dependent increases in intracellular Ca²⁺ concentration in juxtaglomerular cells. A: example heat map depiction of 340/380 absorption ratio in a wild-type (WT) slice during 30-s baseline, 1-s ACh/At puff, and 30-s recovery periods. Baseline and recovery frames are averaged over 30 s, ACh/At frame averaged over 5 s from beginning of puff. Dashed line depicts the target glomerulus of the application pipette. Arrowheads in the ACh/At panel are used to highlight cells that show a higher Ca²⁺ signal than baseline. B: representative heat maps (normalized to baseline ratio) shown to demonstrate the gross effects of β-subunit knockouts on the ACh/At response. WT panel depicted was taken from the data set depicted in A. C: mean number of responsive juxtaglomerular cells from each slice used (*P < 0.05). D: net change in ACh/At-evoked calcium changes in juxtaglomerular cells from β₂-knockouts (KOs) with the agonist alone (Control) or agonist in the presence of 10 µM conotoxin (CTx) AuIB. The β₄-specific antagonist blocked ACh/At responses from these cells indicating that the responses in β₂-KOs are primarily due to contributions from nicotinic acetylcholine receptors (nAChRs) containing the β₄-subunit. E: average (± SE) trace of calcium transients from juxtaglomerular cells in β₂-KOs in response to ACh/At. Control response (black); response in the presence of 10 µM CTx AuIB (red). F, left: mean ratio traces (± SE) from cells that showed positive Ca²⁺ signals in response to ACh/At application. Ca²⁺ signals were recorded at 1 Hz over 80 s; a 1-s ACh/At puff was applied at 30 s (shown in blue). Control, black; β₂-KO, red; β₄-KO, gray. Right: averaged traces normalized to the peaks showing faster decay of calcium transients from β₂-KOs. G: average integrals of the calcium transient (red), peak response (green) and decay time constants (blue) of the ACh/At-driven calcium transients elicited from WT, β₂-KOs, and β₄-KOs.***P < 0.0002. ACh/At, 1 mM acetylcholine in the presence of 2 µM atropine.

The transients elicited by nAChR activation in WT glomeruli had a decay time constant of 15.9 s (mean from n = 54 cells from 4 experiments). In the β₄-KO (Fig. 3, E and F), while very few juxtaglomerular neurons responded with a calcium transient, the responses were not significantly different in their peak or decay kinetics from WT neurons (mean τ = 13.9 s; n = 16 cells from 6 experiments Fig. 3, F and G). In β₂-KOs, while there was no significant decrement in the peak response, the transients decayed significantly faster (mean τ = 7.3 s; n = 32 cells from 4 experiments, P < 0.0002, compared with WT, Fig. 3, F and G). The shorter duration of the signal in β₂-KOs implies that the total calcium signal, as described by the integral of the transient was significantly smaller (Fig. 3G). The sharp reduction in the number of responding neurons in the β₄-KOs suggests that this subunit is necessary for eliciting nAChR-activated calcium signals. The small number also precluded pharmacological tests to determine whether the remaining signals were mainly from β₂-expressing nAChRs. If one assumes that the calcium signals are triggered by depolarization-dependent activation of voltage-gated calcium channels, our results would imply that in the β₄-KOs, ACh/At activated depolarizations are subthreshold for activating these channels (see Fig. 2). The β₂-subunit might contribute to the duration of the calcium signals. Thresholded, all or none calcium responses are seen with nAChRs in other systems (Sharma et al. 2008; Sharma and Vijayaraghavan 2003), mainly because of downstream amplification by ER store calcium. Further, durations of nAChR-mediated calcium transients in astrocytes are regulated by the interplay between ryanodine- and IP₃-dependent calcium stores (Sharma and Vijayaraghavan 2001) . Thus, receptor locations relative to the contributing sources of calcium could determine the kinetics of these transients. The exact mechanisms in the glomerulus remain to be elucidated.

One puzzling finding is the contribution of the β₄-subunit to the ET cell nAChR response evidenced both from current measurement with the subunit knockout as well as calcium imaging. At the face of it, this finding appears to run counter to the previous pharmacological data that showed no significant inhibition by the α3β4* nAChR blocker CTx AuIB (D’Souza et al. 2013; D’Souza and Vijayaraghavan 2012). To rule out the simple explanation that this discrepancy might be due to strain differences (C57/Bl6 vs. FVB) we examined the effect of 10 µM CTx AuIB on ACh/At-induced currents from ET cells of WT mice. The total charge transfer upon ACh/At application in the presence of the toxin was 106 ± 23% of control responses (n = 4, P = 0.89; compare with 104 ± 29%, n = 6, in FVB mice, (D’Souza et al. 2013). The lack of significant inhibition in the WT mice and the results obtained with the β₂-KOs (Figs. 2 and 3) suggests the following possibilities to explain results obtained in Fig. 2. First, there might be a partial compensatory expression of the β₄-gene in the β₂-KOs. However, this would not explain the currents from the β₄-KOs, which are drastically reduced (Fig. 2). The minor effect of the CTx AuIB on WT currents from ET cells (in spite of a large reduction in currents from MCs) and the decrease in nAChR currents in ET cells from β₂-KOs (in spite of a large MC current; see Fig. 2) also argues against the contention that a large fraction of the toxin-sensitive currents in the ET cells arises from weak electrical coupling between these cells and MCs previously reported (Gire et al. 2012).

A parsimonious explanation might be that unlike the MCs, functional nAChRs on ET cells contain both structural subunits. Based on CTx AuIB affinity profiles, one would expect that a α3β2β₄-nAChR combination would have a much lower affinity for CTx AuIB than a subtype containing only the β₄-subunit (Luo et al. 1998).

β₄-nAChRs are necessary for feedback inhibition in the glomerular microcircuit.

nAChR activation has been shown to modulate glomerular output by strongly attenuating the postsynaptic responses in MCs to weak inputs from the olfactory nerve (ON) while allowing stronger signals to be transmitted (D’Souza and Vijayaraghavan 2012). Such a mechanism could potentially allow for better discrimination between two closely related odors (D’Souza and Vijayaraghavan 2014). The mechanism for this filtering effect involves the excitation of MCs and/or ET cells by nAChRs followed by a glutamate-driven excitation of PG cells and a feedback release of GABA inhibiting both MCs and ET cells (D’Souza et al. 2013; D’Souza and Vijayaraghavan 2012; D’Souza and Vijayaraghavan 2014).

Are these filter mechanisms intact in the nAChR KOs? We first examined nAChR-driven changes in sEPSCs in juxtaglomerular neurons. ET cells and PG neurons were held under voltage clamp. In ET cells from WT mice, application of ACh/At did not significantly alter the sEPSC frequencies (Fig. 4, A and B). The average frequencies upon nAChR activation (averaged over 5 s) were 3.625 ± 0.88 vs. 3.02 ± 0.79 Hz baseline frequency (NS, n = 7; P = 0.26).

Fig. 4.

ACh/At-mediated changes in spontaneous excitatory postsynaptic current (sEPSC) frequencies in external tufted (ET) and periglomerular (PG) cells. A: representative 10-s traces from PG and ET cells held at V_m = −70mV and given a 1 s ACh/At puff in slices from wild-type (WT) animals. Expanded sections following ACh application are shown in blue boxes (scales given by blue scale bars). B: frequency vs. time plots for sEPSC events in both cell types and each knockout group. Gross frequencies were determined within 1 s bins throughout the recordings and averaged between cells. A 1 s ACh/At puff was applied at 30 s (blue bar). C: ACh effect on mean sEPSC frequencies across both cell types and knockout conditions. Baseline frequencies were averaged over 30 s before ACh application, and ACh frequencies were averaged over 5 s from the beginning of application. D: comparison of net changes in mean sEPSC frequencies between knockout types (*P < 0.05, ns = no statistical significance). ACh/At, 1 mM acetylcholine in the presence of 2 µM atropine.

This is consistent with the idea that nAChRs do not modulate glutamate release at ON-ET synapses (D’Souza et al. 2013). Also supporting the idea, this remained true with both β₂- and β₄-KOs. No significant changes were observed in the sEPSC frequencies (Fig. 4, A and B). The mean sEPSC frequencies upon nAChR activation were 3.89 ± 0.79 vs. a 3.30 ± 0.70 Hz basal rate in β₂-KOs (NS, n = 7, P = 0.24) and 4.48 ± 1.17 vs. a 4.05 ± 1.54 Hz basal rate in β₄ KOs (NS, n = 7; P = 0.24).

A fraction of PG cells (6/8 WT and 5/7 β₂-KOs) showed increases in sEPSC frequencies. This fraction of responding PG cells(73%) is much more than that previously reported by us (35%; (Parsa et al. 2015). Whether this discrepancy is due to strain differences (previous studies were done in FVB mice) or whether there are subpopulations of PG cells that elude our current identification criteria is not known. Further study is required to resolve this issue.

In both WT and β₂-KO mice, the mean sEPSC frequencies increased by approximately four- and threefold, respectively (WT baseline 2.59 ± 0.84 vs. ACh/At 11.03 ± 2.97 Hz, n = 7, P < 0.05; β₂-KO baseline 2.96 ± 0.72 vs. ACh/At 9.18 ± 2.85, n = 6, P < 0.05; Fig. 4). However, this nAChR-dependent increase was abolished in β₄-KOs (baseline 1.96 ± 1.09 vs. ACh 2.36 ± 0.74, n = 7, P = 0.76). A single-factor ANOVA indicates that there is a significant knockout-dependent effect on ACh/At-mediated sEPSC increases [Fig. 4D; change in sEPSC frequency: WT = 10.79 ± 3.63, β₂-KO = 7.00 ± 2.47, and β₄-KO = 1.34 ± 0.91 Hz. F(2,17) = 3.84, P < 0.05].

On the other hand, the results from measurements of sIPSC frequencies on MCs and ET cells demonstrated a significant role for nAChRs in regulating excitation-driven inhibition at the dendrodendritic synapses between PG cells and the two populations of excitatory neurons. In WT neurons, a 1 s application of ACh/At causes a robust, transient increase in sIPSC frequencies in ET cells and MCs (Fig. 5). In WT mice sIPSC frequencies rose from a baseline of 6.03 ± 2.33 Hz to a mean value of 13.18 ± 2.14 Hz in MCs (n = 6) and from a baseline of 0.75 ± 0.42 Hz to a mean value of 7.5 ± 0.72 Hz in ET cells (n = 7). The mean frequency increase was approximately twofold in MCs (P < 0.005) and approximately tenfold in ET cells (P < 0.001 compared with baseline). This was maintained in β₂-KOs where the increase in sIPSC frequencies upon nAChR activation was not significantly different from that observed in WT mice. Mitral cells and ET cells demonstrated approximately four- and threefold increases in sIPSC frequencies, respectively (MC baseline = 3.51 ± 1.78 vs. ACh = 13.07 ± 1.78 Hz, P < 0.001; ET baseline = 1.63 ± 0.22 vs. ACh = 6.76 ± 1.14 Hz, P < 0.05). As predicted, the nAChR-mediated increase in sIPSC frequencies was completely abolished in MCs recorded from β₄-KOs (n = 6, baseline = 4.96 ± 1.63 vs. ACh = 7.62 ± 2.45 Hz, P > 0.05) and in ET cells (baseline = 1.31 ± 0.21 vs. ACh = 1.85 ± 0.06 Hz, P > 0.05). Our data, therefore, suggest that β₄-containing nAChRs are necessary mediators of feedback inhibition on both ET cells and MCs.

Fig. 5.

ACh/At-mediated spontaneous inhibitory postsynaptic current (sIPSCs) in external tufted (ET) and mitral (MC) cells. A: representative 10 s traces from wild-type (WT), MC, and ET cells held at V_m = −30 mV and given a 1-s ACh/At puff. Expanded 1-s traces shown in blue boxes follow ACh/At puffs. B: frequency vs. time plots for sIPSC frequencies in both cell types and each knockout condition. Gross frequencies were determined within 1-s bins and averaged between cells. A 1-s ACh/At puff was delivered at 30 s (blue bar). C: ACh effect on mean sIPSC frequencies across both cell types and knockout conditions. Baseline frequencies were averaged over 30 s before ACh/At puff. ACh/At frequencies were measured over a 5-s period from the beginning of ACh application. D: comparison of net ACh/At-mediated sIPSC frequency changes between knockout types (*P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.001, ns = not statistically significant). ACh/At, 1 mM acetylcholine in the presence of 2 µM atropine.

Single-factor ANOVAs indicate significant knockout-specific differences in sIPSC frequency gains in both ET cells and MCs [MC f(2,19) = 3.96, P < 0.05; ET f(2,21) = 8.10, P < 0.005]. Post hoc t-tests reveal that both WT and β₂-KO MCs demonstrate a significantly higher sIPSC frequency gain relative to β₄-KO (WT vs. β₄-KO P < 0.05; β₂-KO vs. β₄-KO P < 0.05) though they do not differ from each other significantly (WT vs. β₂-KO, P = 0.24). Similarly, both WT and β₂-KO ET cells demonstrate significant frequency increases relative to β₄-KO (WT vs. β₄-KO P < 0.001, β₂-KO vs. β₄-KO P < 0.01), and likewise do not differ significantly from each other (WT vs. β₂-KO; P = 0.19).

No significant changes in ET cell sIPSC amplitudes were observed upon ACh/At application. Mean sIPSC amplitudes under basal conditions were 12.67 ± 1.002 pA (n = 6), 13.43 ± 0.702 pA (n = 8), and 16.89 ± 4.02 pA (n = 6) for WT, β₂-KOs, and β₄-KOs, respectively (P = 0.39, F = 0.97, one-way ANOVA). Similarly, during the nAChR-mediated burst, the mean amplitudes were 16.89 ± 1.6 pA, 19.52 ± 2.5 pA, and 15.8 ± 3.63 pA for WT, β₂-KO, and β₄-KO, respectively (P = 0.54, F = 0.98, one-way ANOVA). The amplitude distributions did not reveal the appearance of a new class of large amplitude events upon nAChR activation (data not shown) under any of the above mentioned conditions.

Similarly, mean amplitudes of sIPSCs in MCs were 17.8 ± 0.2, 17.1 ± 0.31, and 20.45 ± 0.3 pA for WT, β₂-KOs, and β₄-KOs, respectively (n = 6, P = 0.49, F = 0.74 one-way ANOVA). Upon ACh/At application the mean amplitudes were 25.6 ± 0.64, 22.1 ± 0.62, and 19.4 ± 0.8 pA for WT, β₂-KOs, and β₄-KOs, respectively (n = 6, P = 0.56, F = 0.58). With both ET and MCs, sIPSCs showed a small increase upon ACh/At application in WT and β₂-KOs but this is likely due to random overlaps during the high-frequency bursts.

β₄-Containing nAChRs mediate the filtering of ON inputs.

Our previous studies indicate that a major role for nAChRs in the glomerular microcircuit is to act as high-pass intensity filters of ON inputs (D’Souza and Vijayaraghavan 2012). nAChR activation results in a GABAR-dependent suppression of MC responses to ON stimulation through an excitation-driven feedback inhibition on MCs (Fig. 5). To examine the relative roles of the β₂- and β₄-subunits in this filtering mechanism, we recorded the effects of ACh/At application on evoked EPSCs (eEPSCs) upon electrical stimulation of the ON. MCs were held at −60 mV. A stimulating electrode was placed in the ON layer above the glomerulus to which the MC projects (Fig. 6A). Stimulus intensities were progressively increased in 5-µA increments until saturation was achieved. A pair of stimuli was given 30 s apart at each intensity setting. Two seconds before the second stimulus in the pair, the glomerulus was exposed to a 1-s application of ACh/At. The EPSC amplitudes were integrated within a 100-ms window following the stimulus to determine nAChR-mediated inhibition of incoming signals (D’Souza and Vijayaraghavan 2012).

Fig. 6.

Nicotinic acetylcholine receptors (nAChRs) reduce olfactory nerve (ON) evoked excitatory postsynaptic current (eEPSC) amplitudes in mitral cells. A: schematic illustration of experimental setup. ACh/At pipette was placed at the dendritic tuft of the target mitral cells (MC) and stimulating pipette was placed in the adjacent ON layer ~30- to 50- µm from the target glomerulus. B: representative trace to demonstrate experimental protocol. Paired pulses were delivered 35 s apart (shown with gray arrows). A 1 s ACh/At puff was applied 2 s before the second stimulus (shown with short horizontal line). C: representative eEPSCs recorded from each knockout type (arranged vertically). Current input was increased in 5-µA intervals between protocol sweeps. Left traces show baseline eEPSCs and traces at right demonstrate eEPSCs following ACh/At application. Note: different scales are used in knockout traces due to variation of the saturated current between individual cells. D: scatterplot depicting eEPSC charge-transfer following ACh/At as a function of baseline eEPSC. Individual responses were normalized to the maximal baseline response within each recording. E: normalized eEPSC charge-transfer histograms demonstrating response distribution under baseline (left) conditions and following ACh/At application (right). Individual traces were normalized to mean baseline eEPSC integrals within each recording. ACh/At, 1 mM acetylcholine in the presence of 2 µM atropine. GL, glomerular layer; MCL, mitral cell layer; OSN, olfactory sensory nerve layer; Rec, recording electrode; Stim, stimulus electrode.

As eEPSC amplitudes could vary considerably between cells, responses were normalized to maximal baseline response (Fig. 6D) or mean baseline response (Fig. 6E). In WT cells, the mean eEPSC integral was reduced to 56.41 ± 6.37% of baseline following ACh application (n = 69 responses from 7 cells). Similarly the mean eEPSC integral in β₂-KO was reduced to 48.13 ± 33.69% of baseline (n = 60 responses from 6 cells). Conversely, in β₄-KO eEPSC integrals were not substantially changed with a post-ACh mean value of 103.21 ± 10.38% of baseline (n = 74 traces from 8 cells). K-S tests indicate that ACh/At puffs cause a significant leftward shift in eEPSC integrals in both WT and β₂-KO cells [WT D(69) = 0.29, P < 0.005; β₂-KO D(60) = 0.46, P < 0.001] but not in β₄-KO cells [D(74) = 0.076, P = 0.97].

We further dissected the eEPSC response by examining the effect of nAChR activation on the first 20 ms of the response and the next 80 ms. The results, shown in Table 1, suggest that nAChR modulates both components. If the effect on the fast response corresponds to nAChR-mediated actions on the ON and the slower one on the dendrodendritic circuitry (Vaaga and Westbrook 2017), it could imply that the actions of nAChRs are complex and operative at multiple loci. Countering this conclusion is our finding on ON-ET transmission that remains relatively unaltered by nAChR modulation (D’Souza et al. 2013). More work is needed to resolve this issue. It is clear, however, that effects on both components are attenuated in the β₄-KOs and not in the β₂-KOs (Table 1).

Table 1.

Modulation of eEPSCs on MCs by nAChRs

Mouse	nAChR/Basal (Low 0-20 ms) %	nAChR/Basal (High 0-20 ms) %	nAChR/Basal (Low 20-100 ms) %	nAChR/Basal (High 20-100 ms) %	N
WT	52 ± 13	30 ± 16	74 ± 7	16 ± 7	7
β2-KO	64 ± 15	16 ± 24	71 ± 15	12 ± 40	8
β4-KO	96 ± 2.3†	92 ± 5†	94 ± 3*	92 ± 3‡	6

eEPSC charge transfer was calculated from 0 to 20 ms and 20 to 100 ms for both basal (no ACh/At) and upon nAChR stimulation. This was done for both low- and high-stimulus intensity conditions. For each value, the ratio of the area upon receptor activation to its corresponding basal response was calculated (nAChR/Basal). ACh/At, 1 mM acetylcholine in the presence of 2 µM atropine; eEPSC, evoked excitatory postsynaptic current; KO, knockout; MCs, mitral cells; nAChR, nicotinic acetylcholine receptor; WT, wild-type. Values are expressed as %.

^*P < 0.05;

^†P < 0.01;

^‡P < 0.0001.

Based on the results presented here we would predict that functional β₄-containing nAChRs are essential for filtering ON inputs to weak stimuli. We compared the evoked firing of MCs in response to ON stimulation in WT and β₄-KOs under different stimulation intensities (Fig. 7). For these experiments QX-314 was omitted from the recording pipette. Net increase in spiking upon ON stimulation was measured by subtracting the baseline spike count (within 100 ms before stimulation) from that obtained 100 ms poststimulation. The baseline firing rate was 0 ± 0 Hz in control conditions and 19 ± 6 Hz upon nAChR activation (Fig. 7A, Table 2). In the WT mice, low-intensity stimulation resulted both in failures (Fig. 7B) and also a net decrease in firing frequency, often resulting in a negative firing rate (Fig. 7, B and C), suggesting an inhibition of ACh-induced firing. Similarly, at higher intensities, while the cells consistently reached firing threshold with electrical stimulation, there was still a decrease in net firing upon activation of nAChRs (12.50 ± 5 vs. 23.3 ± 2 Hz baseline, P < 0.05, paired t-test, n = 7; Fig. 7, C and D; Table 2). Both the failures with low intensity stimulation and the net decrease in firing were abolished in β₄-KOs (Fig. 7, B–D and Table 2). These results demonstrate that nAChR-mediated filtering of weak responses shows an obligatory dependence on the presence of the β₄-subunit.

Fig. 7.

β₄-Subunit mediates nicotinic acetylcholine receptor (nAChR)-dependent filtering of weak inputs to mitral cells. A: representative trace to demonstrate experimental protocol. Cells were held under current clamp. Paired pulses were delivered 35 s apart. Small vertical bars indicate stimulus application in all figure panels. The first stimulus (Stim 1) elicits the baseline response while the second stimulus (Stim 2), which arrives 1 s after a 1-s application of ACh/At represents the response after nAChR activation. B: example wild-type (WT) and β₄-knockout (KO) mitral cell responses to low and high stimuli. Low stimuli was the lowest current necessary to generate a single action potential under baseline conditions whereas high stimuli were minimal currents necessary to generate the maximal baseline AP rate in each cell. C: mean AP frequency increases following stimuli. It should be noted that frequencies depicted are net frequencies from prestimulus firing rates (*P < 0.05; ns, not statistically significant). D: scatterplots depicting net frequency increases following nAChR activation as a function of baseline net frequencies in both WT (left) and β₄-KO (right). The gray dashed line represents a slope of 1. The thin gray line in the left trace is at 0 to highlight negative responses. Note responses at all stimulus intensities are inhibited. In the β₄-KOs responses fall close to the dashed line, indicating loss of the nAChR-mediated effect. ACh/At, 1 mM acetylcholine in the presence of 2 µM atropine.

Table 2.

ACh/At-mediated modulation of MC firing

	Baseline Firing, Hz		Net Frequency Change upon ON Stimulation, Hz
	Control	ACh/At	Untreated	ACh/At	N
Wild-type
Low	0	19 ± 6	10.5 ± 0.5	−0.95 ± 5.9*	7
High	0	17.5 ± 5	23 ± 2	12.5 ± 5*	7
β4-KO
Low	0	0	12 ± 2	12 ± 2	5
High	0	0	34 ± 8.7	34 ± 9.2	5

MCs were held under current-clamp configuration. A stimulating electrode was placed in the ON layer. Two responses were elicited one in the absence of neuromodulation (Control) and the other 1s after a 1s application of ACh/At. The first two columns show changes in baseline frequencies upon nAChR activation while the next two columns are net change in firing frequencies upon ON stimulation. In all cases 100 ms prior to and after stimulation were taken as the baseline and response frequencies, respectively. ACh/At, 1 mM acetylcholine in the presence of 2 µM atropine; KO, knockout; MCs, mitral cells; ON, olfactory nerve; WT, wild-type.

^*P < 0.003.

DISCUSSION

Olfaction emerges from a shallow sensory system where a single synapse separates the environment from cortical regions. This synapse is the ON-MC synapse in the MOB glomerulus. Perhaps, necessitated by its shallowness, the MOB in general, and the glomerular circuit in particular, is subject to complex modulatory influence from glomerular interneurons and a number of centrifugal inputs. We have proposed that one such input, from the basal forebrain cholinergic system, regulates glomerular output via excitation-driven inhibitory mechanisms triggered by nAChR activation (D’Souza et al. 2013; D’Souza and Vijayaraghavan 2012, 2014; Parsa et al. 2015). In this study we have used nAChR subunit knockout mice to dissect the relative contributions from different nAChR subtypes and the cell types that form the glomerular microcircuit.

Recent studies have determined that the MOB is not a simple encoder of odor information to be decoded in higher order areas but that it performs a significant level of information processing by itself. For example, synchrony between MC spiking in response to odors can be modulated depending on whether the odors are rewarded or not (Doucette et al. 2011). This and other studies (Gao and Strowbridge 2009; Markopoulos et al. 2012) show that task-specific regulation of MOB circuits play an important role in information transfer to the cortex.

These findings also emphasize the importance of centrifugal inputs into the bulb in modulating circuit output. The bulb receives adrenergic (Devore and Linster 2012; Moriceau and Sullivan 2004; Sullivan et al. 2000), serotonergic (Liu et al. 2012; Petzold et al. 2009), and cholinergic (D’Souza and Vijayaraghavan 2014; Hamamoto et al. 2017; Ma and Luo 2012; Salcedo et al. 2011) projections. The cholinergic input to the bulb arising from the HDB might be involved in modulating odor sensitivity, odor discrimination, and perceptual learning (Bendahmane et al. 2016; Chaudhury et al. 2009; Hellier et al. 2010; Mandairon et al. 2006; Rushforth et al. 2010). Consistent with this, our work suggests that nAChRs in the MOB glomerulus preferentially select strong inputs to be transmitted (D’Souza and Vijayaraghavan 2012), potentially resulting in increased contrast between overlapping populations of activated glomeruli encoding closely related odors (D’Souza and Vijayaraghavan 2014). The threshold for this filter is set by excitation-driven GABA release from PG cells (Parsa et al. 2015).

The use of nAChR subunit KOs served a dual purpose. The first, and more obvious, is the elucidation of receptor subtypes that mediate this excitation-driven inhibition. The second is to examine the relative contributions of MCs and ET cells to the excitatory drive for inhibition. Pharmacology data suggested that ET cells predominantly express functional α4β2*-containing nAChRs while MC responses appear to be mediated by α3β4* subtype (D’Souza et al. 2013; D’Souza and Vijayaraghavan 2012). This conclusion was based on the finding that nAChR currents were blocked by low levels of mecamylamine but not by the α3β₄-nAChR blocker CTx AuIB. The use of constitutive knockout mice has its disadvantages, as it is not known whether there are specific developmental changes or compensatory mechanisms operative that might affect our results. However, the combination of pharmacology (D’Souza et al. 2013; D’Souza and Vijayaraghavan 2012, 2014; Parsa et al. 2015) and these models together provides a strong bases for the conclusions arrived at in this study.

Consistent with this conclusion, MCs in β₂-KOs continued to express large nAChR currents while nAChR currents on ET cells were attenuated. In these mice, the sIPSC increases with ACh/At were robust and not significantly changed from WTs. On the other hand, when MC nicotinic currents were abolished in the β₄-KO mice, the nAChR-mediated sIPSC increase was abolished as well. Interestingly, ET cells in the β₄-KOs had attenuated nAChR currents. This finding, taken together with the finding that there are residual currents in the β₂-KOs, suggests to us that portion of nAChR currents in the ET cells might arise from β₄-containing nAChRs. This might suggest that there are heteromeric nAChRs containing two different β-subunits as argued in results. However, the opposing effects of the two KOs on the changes in sIPSC frequencies, even though both elicit low-amplitude currents on ET cells, suggests to us that the primary driver for the excitation driven inhibition is functional α3β4*-nAChRs expressed on the MCs. At this point, the possibility that a small number of α3β4*-nAChRs on ET cells are strategically located and have a disproportionate effect on sIPSCs from PG cells cannot be ruled out. Based on our imaging studies, β₄-containing nAChRs probably play a dominant role in calcium signaling in these juxtaglomerular neurons.

Release of ACh in the central nervous system occurs at multiple time courses from tonic to phasic with time scales in the milliseconds to many seconds (Ballinger et al. 2016). The information carried at these multiple time scales are yet to be clearly understood. Relevant to this study, it has been shown that in the prefrontal cortex there is rapid and transient increase in ACh levels (lasting a few seconds) associated with rewarded cue detection, possibly linked to attentional/anticipation mechanisms (Parikh et al. 2007).

Our knowledge of nAChR subunits in the MOB is far from complete. While our data places the α3β4*-nAChRs on MCs and ET cells, involvement of other cell types and receptor subtypes remains to be elucidated. Recent studies have suggested α2-subunit containing nAChRs on deep short axon cells (Burton et al. 2017; Case et al. 2017). Similarly, autoradiographic studies reveal expression of receptors containing the α7-subunit in the glomerular neuropil (Hellier et al. 2010; Le Jeune et al. 1995). Further work aimed at integrating responses from these receptors and locations into our model could lead to a more complete understanding of nAChR modulation of the glomerular network. The existence of nine nAChR α-subunits and the possibility that many nAChRs might express multiple α-subunit genes makes the examination of all possible combinations a daunting task. This is further complicated by the fact that we have detailed pharmacological profile for only a few of these combinations. The task was somewhat simplified in this study using the β-subunit KOs. These are likely to target most, if not all heteromeric nAChRs. There are only three nAChR β genes (β₂–β₄) with one of them (β3) being very sparsely expressed in the brain (Kamens et al. 2015; Wu and Lukas 2011) and existing only in combination with one of the other two β-subunits (Boorman et al. 2003). Therefore, this study strongly implicates β₄-subunit-containing nAChRs in mediating the high-pass intensity filter. Adding evidence from our pharmacological studies and the results here, the likely candidate receptor is α3β4* receptor in MCs, and, perhaps, α3β2β4*-nAChRs on ET cells. However, these conclusions are only as strong as our knowledge of the existence and pharmacology of various possible subunit combinations, which is limited at present.

A parsimonious interpretation of our results leads us to the following testable model for nAChR effects on the glomerular microcircuit. The spike in transmitter levels seen during the attention/anticipation phase lasting a couple of seconds (Parikh et al. 2007; Sarter et al. 2009) activates the α3β4*-nAChRs on MCs. nAChR-mediated excitation of MCs sets up a firing pattern that marks the attentional/anticipational phase. This firing of MCs triggers feedback GABAergic inhibition from PG cells on to the MCs and inhibition at the level of ET cells. These effects result in a net decrease in the responses of MCs to an odor. Increased release of GABA from the PG cells would shunt MCs and also reduce feed-forward excitation mediated by the ET cells. Thus, for a few seconds, nAChR activation results in establishing a high-pass intensity filter that allows for removal of information from weak odor stimuli. Such a mechanism would be context specific, and not odor specific, and would result in better odor discrimination. This effect of filtering out weak inputs might be a consequence of the overall effect of nAChR activation to improve signal to noise ratio in the system. As shown in Fig. 7, the net firing is inhibited across all intensities of stimulation. This would be consistent with the idea that nAChRs sharpen MC receptive fields under a rate metric while mAChRs improve spike synchrony (Li and Cleland 2013). While failure with weak stimuli is consistent with a high-pass intensity filter, we need to know more about odor coding to understand the significance of net decrease in odor-evoked firing relative to an elevated background firing. It is possible that a change in the background template due to nAChR activation before odor arrival triggers different coding strategies that allow for better discrimination.

A requirement for this model is the efficient and uniform inhibition of all MCs impinging on a given glomerulus. This, we feel, is achieved by GIGR that serves to amplify inhibition (Parsa et al. 2015). GIGR explains how activation of a few PG cells (as would occur during a weak stimulus) can nonetheless effectively block the transfer of information to MCs across the entire glomerulus.

Both nAChRs and mAChRs participate in odor discrimination (Chan et al. 2017; Chaudhury et al. 2009; Devore et al. 2014; Hellier et al. 2010; Mandairon et al. 2006; Rushforth et al. 2010). Mouse KOs for M1 and M3 mAChR subtypes show significant deficits in odor discrimination (Chan et al. 2017). Autoradiographic studies demonstrate that mAChRs are densely expressed in the external plexiform layer and the granule cell layer (where they modulate dendrodendritic inhibition of MCs by granule cells (Pressler et al. 2007; Smith et al. 2015). It has been postulated that modulation by mAChRs serves to sharpen olfactory receptive fields (Chaudhury et al. 2009; Ma and Luo 2012).

While our results are consistent with a role for nAChRs in learned odor discrimination, the integrated response to endogenous cholinergic activation is still far from clear. Future work delineating the roles for different subtypes and conditions under which they can be activated by endogenously released ACh is needed to provide the appropriate framework and context in which to understand the role of this important centrifugal input in modulating olfactory behavior.

GRANTS

The work was funded by NIH R01DC 008855 and NIH R21 AG 053690 to SV.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

M.S.S., P.V.P., S.G.B., and R.D.D. performed experiments; M.S.S., P.V.P., S.G.B., R.D.D., and S.V. analyzed data; M.S.S., P.V.P., S.G.B., R.D.D., and S.V. interpreted results of experiments; M.S.S. and S.V. prepared figures; M.S.S., P.V.P., S.G.B., R.D.D., and S.V. edited and revised manuscript; M.S.S., P.V.P., S.G.B., R.D.D., and S.V. approved final version of manuscript; S.V. conceived and designed research; S.V. drafted manuscript.