Generation of hM3Dq-expressing Mice
Previously, we showed in vitro
that hM3Dq signals exclusively via the canonical Gq
pathway in non-neuronal cell types (Armbruster et al., 2007
; Conklin et al, 2008
). To assess whether hM3Dq couples to Gq
in mixed neuronal-glial cultures, we visualized calcium flux with a calcium-sensitive dye in primary cultures of mouse postnatal day 1 (P1) hippocampal cells infected with hM3Dq-transducing lentivirus. CNO stimulated calcium transients only in those neurons expressing hM3Dq (identified by an IRES-driven mCherry reporter) and was without effect in uninfected neurons (Figure S1
and see Supplemental Videos SV1
Having thus demonstrated that hM3Dq conferred the ability to stimulate neuronal Gq
-signaing in vitro
, we sought to determine whether expression of hM3Dq could alter neuronal function in vivo
. We first generated transgenic mice expressing tetracycline-sensitive HA-tagged hM3Dq using the Tet-off system (i.e
. transgene expression is repressed upon administration of tetracycline or its analog, doxycycline). This system allows for inducible, spatiotemporally-regulated transgene expression. Pronuclear injection of B6SJL hybrid mouse oocytes with Tet Response Element (TRE)-hM3Dq DNA () resulted in the birth of 35 mouse pups, of which ten carried the transgene. To determine whether we could achieve brain region-specific expression of hM3Dq protein, we crossed TRE-hM3Dq founders with CaMKIIα-tTA mice in which tTA expression is targeted to principal neurons mainly in cortex, hippocampus, and striatum (Mayford et al., 1996
) (). We removed the brains from F1 adults and immunoprecipitated HA-hM3Dq protein from membrane fractions of whole brain (minus cerebellum) homogenates. Initial immunoblot analysis of the immunoprecipitates revealed expression of HA-hM3Dq in double-transgenic (both TRE-hM3Dq and tTA transgenes) F1 progeny from two of the founder mice (data not shown). Subsequent studies of mutant F1 progeny from both sets of founder mice did not reveal any differences in receptor expression or activity; therefore, we chose only one of the founder lines (Line 1) for subsequent studies. Hereafter, we refer to the mice from this line, which express HA-hM3Dq driven by the CaMKIIα promoter, as hM3Dq mice.
Generation of transgenic mice with inducible expression of HA-hM3Dq
Having determined that the CaMKIIα-tTA transgene effectively drives hM3Dq expression in the brain, we next sought to repress hM3Dq expression by doxycycline administration. Immunoprecipitation and immunoblot analyses similar to those described above revealed undetectable hM3Dq expression in both hM3Dq F1 progeny fed doxycycline-containing chow (200 mg/kg) for one month and single-transgenic (TRE-hM3Dq transgene only) mice fed normal chow (). These findings demonstrated, first, that doxycycline successfully repressed tTA-driven hM3Dq expression in vivo and, second, that hM3Dq expression could be temporally regulated.
Characterization of hM3Dq Expression
To examine the spatial distribution of transgene expression, we prepared coronal brain sections for immunofluorescence and probed them with an anti-HA antibody. We found that immunofluorescence was most intense in the cortex and hippocampus (), as expected for CaMKIIα-tTA-driven transgene expression (Mayford et al., 1996
). Within the hippocampus, immunofluorescence was most intense in CA1 strata radiatum
(), a pattern consistent with expression in the apical and basal dendrites of CA1 pyramids. Confocal immunofluorescent microscopic analysis of single CA1 pyramidal cells confirmed immunoreactivity in the soma and apical and basal dendrites (). Apical dendrite immunoreactivity was also detected in cortex (). Importantly, inclusion of doxycycline in the diet resulted in no detectable immunoreactivity in either cerebral cortex or CA1 (compare ). Additionally, a Nissl stain revealed no overt structural differences between WT and hM3Dq mice (Figure S2
Receptor expression and localization in transgenic mouse brains
To obtain independent evidence of hM3Dq expression in the hM3Dq mice, we measured the Bmax
H] quinuclidinyl benzilate (QNB) binding in cortical and hippocampal membranes from wild-type and hM3Dq mice. Radioligand competition binding isotherms were first modeled based on prior affinity estimates obtained with cloned wild-type and hM3Dq receptors expressed in vitro
=0.05 nM at hM3; KD
=1.6 nM at HA-hM3Dq; Armbruster et al., 2007
). We conducted computerized simulations with Prism 4.0 using different concentrations of [3
H]QNB and various ratios of native to transgenic receptor expression in order to determine the optimal concentration of [3
H]QNB, a non-selective muscarinic antagonist, for detection of both endogenous muscarinic receptors and hM3Dq receptors. Having determined the optimal assay conditions, competition curves were then obtained using unlabeled acetylcholine (which binds hM3Dq albeit with lower affinity compared to wild type mAChRs; Armbruster et al., 2007
) and [3
H]QNB. Data were fit simultaneously by LIGAND (Munson and Rodbard, 1980
) using weighted non-linear regression analysis and a two-site model, which best described all data sets (by F-test). In hM3Dq tissue, a distinct site emerged for which ACh had low affinity (1.0 mM in cortex, 1.7 mM in hippocampus) (, Table S1
). The affinity of ACh for this site was similar to that of ACh for cloned hM3Dq, but not wild-type muscarinic, receptors (Armbruster et al., 2007
). The inclusion of doxycycline in the diet of hM3Dq mice eliminated this low-affinity site (Table S1
). We concluded that (1) this low-affinity ACh site accounted for the difference in Bmax
between hM3Dq and wild-type or doxycycline-treated tissue, and (2) the difference in Bmax
was due to hM3Dq transgene expression. Indeed, computerized non-linear least-squares regression analysis taking into account the difference in affinity of [3
H]QNB for the hM3Dq and native mAChRs revealed an increase of four- and seven-fold in the Bmax
H]QNB binding to membranes isolated from cortex and hippocampus, respectively, of hM3Dq compared to WT mice (, Table S1
). These data indicate that hM3Dq was expressed in both cortex and hippocampus at levels greater than endogenous muscarinic receptors.
Behavioral Analyses of hM3Dq Mice in the Absence of CNO
A major goal of the evolution of the second generation of RASSLs was to minimize constitutive activity of hM3Dq and also to eliminate the activation of hM3Dq by endogenous ACh, thus limiting hM3Dq signaling in vivo
in the absence of CNO. When expressed in yeast, cultured mammalian cells, and mouse pancreatic β-cells, no detectable constitutive activity of hM3Dq in the absence of CNO was found (Armbruster et al., 2007
; Guettier et al
., submitted). To test whether hM3Dq expression in vivo
modified behavior in the absence of its exogenous activating ligand (CNO), we examined mouse appearance and behavior through a battery of neurobehavioral tests on single-transgenic (TRE-hM3Dq transgene only) and hM3Dq mice (see Supplementary Methods and Figures S3
; n=11 littermate pairs). No qualitative differences were evident between control and hM3Dq mice. Specifically, the two groups were similar in observations regarding size, overall morphology, coat condition, body posture and gait, reflex response to gentle touch with a cotton swab on the whiskers or eyes, balance in an empty, shifting plastic cage, ability to climb a pole or walk across an elevated dowel, grip-strength in a wire-hang test, and responses during 20 seconds of tail-suspension.
A variety of quantitative assessments of behavior were also performed. We quantified motor coordination using an accelerating rotarod, the results of which revealed no significant differences between hM3Dq and control mice (Figure S3
). Likewise, no significant differences were detected in the elevated plus maze, acoustic startle response, pre-pulse inhibition of startle responses (PPI), latency to locate a food reward buried in bedding, the Morris water maze visual cue test, or acquisition and reversal learning in the Morris water maze (Figures S4
In contrast to these measures that revealed no differences between the two genotypes, we did find subtle but unremarkable differences with respect to activity in an open field (Figure S7
). hM3Dq mice demonstrated a tendency toward reduced locomotor activity as measured by total distance traveled in a locomotor chamber [F(1,20)=6.15, p<0.05] (Figure S7A
). They further demonstrated fewer fine, stereotypic movements [F(1,20)=6.63, p=0.0181] (Figure S7B
). However, no differences in rearing movements or in time spent in the center of the open field were detected (data not shown). Apart from the mild reduction of spontaneous locomotor activity, the hM3Dq mice were similar to controls on an extensive battery of tests of diverse behaviors, implying that overexpression of a Gq
-coupled receptor has minimal untoward effects.
In Vitro CA1 Pyramidal Cell Response to CNO
After characterizing the behavior of hM3Dq mice in the absence of exogenous ligand, we sought to determine the effects of hM3Dq receptor activation by CNO. We first asked whether neurons from hM3Dq animals would respond to CNO. To address this question, we isolated acute hippocampal slices from hM3Dq and littermate control animals and performed whole-cell recordings from CA1 pyramidal neurons in the absence and presence of bath applied CNO. To isolate membrane potential responses from synaptic responses, we performed these recordings in the presence of tetrodotoxin (TTX). We found that bath application of CNO (500 nM) to hippocampal slices isolated from hM3Dq mice depolarized CA1 pyramidal cells by 8.0 ± 1.8 mV compared with baseline membrane potentials (p<0.01; n=7 neurons from 5 animals; ). In contrast to these findings in hM3Dq mice, we found no significant change in membrane potential of CA1 pyramidal neurons of hippocampal slices isolated from control (either single transgenic or wild-type) mice (p>0.05, n=7 neurons from 5 animals; ). We additionally asked whether bath application of CNO to hippocampal slices isolated from hM3Dq mice would increase the firing rate of CA1 pyramidal neurons. To that end, we performed whole-cell recordings from CA1 pyramidal cells in the absence of TTX. We found that bath application of CNO (500 nM) in the absence of TTX increased the firing rate of CA1 pyramidal cells (n=8 neurons from 4 animals; ).
CNO effects on CA1 pyramidal neurons recorded in vitro
Because Gq signaling can affect membrane conductances through phospholipase C (PLC)-dependent mechanisms, we tested the dependence of the CNO-evoked depolarization on PLC. We performed whole-cell recordings from CA1 pyramidal cells in acute slices isolated from hM3Dq animals in the presence of TTX and the PLC inhibitor, U73122 (10 μM), and then bath applied CNO (500nM). We found that in the presence of U73122, CNO had no significant effect on the membrane potential of CA1 pyramidal cells (p>0.05, n=6 neurons from 4 animals; ). By contrast, in the presence of U73343 (10 μM), the inactive analog of U73122, CNO produced a significant depolarization of CA1 pyramidal neurons by 5.0 ± 0.72 mV (p<0.01, n=4 neurons from 4 animals; ). These findings indicate that CNO-induced activation of hM3Dq depolarizes and increases the firing rate of CA1 pyramidal neurons through a PLC-dependent mechanism.
Behavioral Response of hM3Dq mice to CNO
Treatment of hM3Dq but not WT mice with CNO induced striking behavioral effects. Using a cohort of adult mice distinct from that in which baseline behavioral characterization was performed, we measured ambulation and episodes of repetitive beam breaks (generally representing fine movements or stereotypic behaviors) following peripheral administration of vehicle (saline) and low dose (0.1 and 0.3 mg/kg) CNO (n=10 hM3Dq; n=14 controls). Among hM3Dq mice, significant behavioral effects were noted. When CNO was administered, activity (both total ambulation and repetitive beam breaks) increased in a dose-dependent fashion among hM3Dq mice and exceeded control levels. The administration of 0.1 mg/kg CNO resulted in a significant increase in repetitive beam breaks (p<0.01) and a trend toward a significant increase in total locomotion in hM3Dq mice relative to controls. Administration of 0.3 mg/kg CNO produced a statistically significant increase in both total locomotion and repeated beam breaks in hM3Dq mice relative to controls (p<0.01; ). Furthermore, the increases in locomotion and repetitive beam breaks following CNO administration occurred in a time-dependent manner such that locomotor activity gradually increased with time relative to controls (), and the increase in locomotor activity persisted >9 hours (; n=4 hM3Dq; n=4 controls). These data provide behavioral evidence of neuronal activation by CNO in hM3Dq mice and demonstrate the long duration of CNO activity in hM3Dq animals.
Behavioral consequences of CNO administration
Behavioral and in vivo electrophysiological timecourse of CNO effects in hM3Dq animals
To assess the behavioral responses of hM3Dq mice to higher doses of CNO, we observed control and hM3Dq mice treated in parallel with saline or increasing doses of CNO. We found that treatment of hM3Dq mice with 0.5 mg/kg CNO reliably evoked limbic seizures of behavioral Class 1 (); no behavioral evidence of seizure activity was detected after administration of lower doses (0.03, 0.1, or 0.3 mg/kg CNO). Treatment with doses of CNO higher than 0.5 mg/kg evoked continuous seizure activity of classes three through seven, namely status epilepticus (SE) and death. CNO at doses of 1 or 5 mg/kg evoked SE in 2 of 4 animals and 5 of 7 animals (). Among the 5 animals in which SE was elicited by 5 mg/kg CNO, an orderly progression of behavioral seizure intensity was observed over time following CNO treatment (). The latency to onset of the first Class 1 seizure was 15.4 ± 3.7 minutes. SE was lethal in 4 of the 7 animals during the 3 hr following SE onset. Whereas CNO reliably evoked seizures in hM3Dq animals, CNO evoked no behavioral seizures in control animals at any dose tested (0.03 – 5 mg/kg; n=4 carrying the tTA transgene alone; n=2 carrying the TRE-hM3Dq transgene alone; n=5 wild-type animals).
In Vivo Recordings from Hippocampus
We next sought to examine the electrophysiological correlates of the behavioral effects of CNO in the hM3Dq mice. Because the behavioral features of the seizures were typical for seizures arising from hippocampus and because hM3Dq is expressed in hippocampal neurons, we focused on the hippocampus. Towards this end, we implanted control and hM3Dq mice with multi-electrode arrays to monitor both local field potentials (LFPs) and spike activity of multiple individual neurons in the hippocampus. These in vivo recordings were performed during the same sessions as the behavioral seizure observations described above.
CNO increased neuronal activity in a dose- and time-dependent fashion as measured by both LFP and single unit firing rate (Figures -). With respect to the dose-dependent change in neuronal activity, we found that treatment of hM3Dq mice with saline () or 0.03 mg/kg () CNO produced no significant change in the LFP or in single unit firing, consistent with the lack of behavioral affects of these doses. However, doses of 0.1 mg/kg to 1 mg/kg elicited successively greater increases in power of the gamma frequency band of the LFP (). At the sub-convulsive doses of 0.1 mg/kg and 0.3 mg/kg, increased power in the gamma frequency band was seen in the LFP, with 0.3 mg/kg eliciting a greater magnitude of gamma power increase (). In addition, during the period of increased gamma power we found an increase in the firing rate of hippocampal interneurons (Figures -). Treatment of hM3Dq mice with convulsant doses of CNO (0.5, 1 and 5 mg/kg) induced striking alterations in neuronal activity evident in both LFP and single unit activity (). Notably, spectral analyses of LFP activity revealed an increase in gamma power preceding the onset of seizures (Figure , ). Analyses of single neuron firing frequency revealed an increase in interneuron firing frequency that coincided with the increased gamma power in the LFP. This pattern of single neuron firing persisted until the onset of seizures, at which point interneuron firing dramatically decreased (Figures -). By contrast, treatment of hM3Dq mice with saline induced no significant change in LFPs or the firing frequency of interneurons (). Likewise, treatment of control mice with either CNO or saline induced no significant change in LFP spectral parameters or firing rate of interneurons ().
CNO dose-dependently increase neuronal activity and gamma power in hM3Dq mice
Local field potential and single unit recordings from hippocampus of control and hM3Dq animals administered either saline or 1 mg/kg CNO subcutaneously
We also assessed the time-dependent changes in neuronal activity following CNO administration. The effect of CNO on the LFP, as measured by changes in gamma power, was first seen between 5 and 10 minutes after CNO administration and peaked between 45 and 50 minutes for all doses tested (). To determine the duration of CNO effects on neuronal activity, we administered CNO (0.3 mg/kg or 0.5 mg/kg) and recorded LFP for 24 hours (n=3 mice for 0.3 mg/kg; n=2 mice for 0.5 mg/kg). At both doses tested, the CNO-induced increase in gamma power persisted for approximately 9 hr (), paralleling the pattern we observed in locomotor activity (). These data suggest that the offset kinetics are not dependent on CNO dose. Moreover, 24 hours after the first 0.3 mg/kg dose, we readministered CNO (0.3 mg/kg) to the same mice and found no significant difference in peak gamma power response or duration of this response between day 1 and day 2, indicating that sensitivity to CNO, at least with 24 hours between administrations, was unaltered by such long-lasting effects ().
Efficiency of Tet-Off System in hM3Dq Mice
As described above, doxycycline virtually eliminated hM3Dq protein expression as assessed by radioligand binding assays, Western blot analysis and immunofluorescence microscopy. To assess whether doxycycline also eliminates the behavioral and electrophysiological responses to CNO, hM3Dq mice were treated with doxycycline for 4 weeks and behavioral and electrophysiological responses to CNO were assessed (n=5 hM3Dq; n=5 control). CNO (0.3 and 5 mg/kg) induced no locomotor response with respect to either total distance traveled () or repetitive beam breaks () in hM3Dq mice treated with doxycycline (n=5 mice). Moreover, behavioral seizures and alterations in LFP, as measured by spectral analysis, were notably absent from hM3Dq mice administered CNO (5 mg/kg) during doxycycline treatment (; n=3 mice). To assess the reversibility of doxycycline suppression of transgene expression on the electrophysiological response, administration of doxycycline was terminated and four weeks later the same mice were treated again with CNO (5 mg/kg) (). In this case, CNO induced increased power in the gamma frequency band as detected by LFP recordings as well as behavioral seizures culminating in SE and death, thereby demonstrating both the reversibility of doxycycline suppression of transgene expression and the specificity of CNO’s effects for the hM3Dq receptor.
Doxycycline treatment of hM3Dq animals prevents CNO induced behavioral and electrophysiological changes