|Home | About | Journals | Submit | Contact Us | Français|
Tools for suppressing synaptic transmission gain power when able to target highly selective neuron subtypes, thereby sharpening attainable links between neuron type, behavior, and disease; and when able to silence most any neuron subtype, thereby offering broad applicability. Here we present such a tool, RC::PFtox, that harnesses breadth in scope along with high cell-type selection via combinatorial gene expression to deliver tetanus toxin light chain (tox), an inhibitor of vesicular neurotransmission. When applied in mice, we observed cell-type specific disruption of vesicle exocytosis accompanied by loss of excitatory postsynaptic currents and commensurately perturbed behaviors. Among various test populations, we applied RC::PFtox to silence serotonergic neurons, en masse or a subset defined combinatorially. Of the behavioral phenotypes observed upon en masse serotonergic silencing, only one mapped to the combinatorially defined subset. These findings provide evidence for separability by genetic lineage of serotonin-modulated behaviors; collectively, these findings demonstrate broad utility of RC::PFtox for dissecting neuron functions.
Uncovering in vivo functions served by different neuron classes is of clinical and fundamental interest. Tools enabling such functional mapping are in need and under development in various forms (Nakashiba et al., 2008; Yamamoto et al., 2003; Yu et al., 2004; and reviewed in (Dymecki and Kim, 2007; Luo et al., 2008). Here we present RC::PFtox (Fig. 1A–C and Supplemental Fig. 1A–C), a broadly applicable genetic tool for silencing virtually any neuron subtype in the living mouse. Maximized in RC::PFtox is the attainable cell-type specificity of neuronal silencing, thus sharpened is the attainable link between neuron type and function served. Also offered is breadth in applicability across most neuron types.
RC::PFtox, a knock-in allele of the ROSA26 (R26) locus (Zambrowicz et al., 1997), exploits the powerful and highly selective dual-recombinase methodology of intersectional gene activation (Awatramani et al., 2003; Farago et al., 2006; Jensen et al., 2008)(Fig. 1A–C and Supplemental Fig. 1A–C) for conditional expression of a GFPtox fusion protein (Yamamoto et al., 2003). Tox suppresses vesicle-mediated neurotransmitter release by cleaving the synaptic vesicle-associated membrane protein VAMP2/synaptobrevin2 (Schiavo et al., 2000). In the absence of VAMP2, assembly of the SNARE protein complex, needed for exocytic fusion of synaptic vesicles with plasmalemma, is inhibited (Schiavo et al., 2000). Tox is quite potent, requiring fewer than ten molecules intracellularly to block fifty percent of synaptic vesicle exocytosis in Aplysia neurons (Schiavo et al., 2000). Expression of GFPtox from RC::PFtox requires removal of two stop cassettes, a loxP-flanked cassette excisable by Cre recombinase and an FRT-flanked cassette, by Flpe (Fig. 1A–C). GFPtox action, therefore, should restrict to just those cells having expressed both Cre and Flpe recombinase. Requiring two recombination events allows for defining the targeted cell subtype with great specificity (cartooned in Supplemental Fig. 1A, B), that is, by pair wise combinations of expressed genes, rather than solely by a single-gene profile afforded by single-recombinase or conventional transgenic strategies. Thus, confounding interference from silencing too heterogeneous a cell population is minimized, resulting in an improved capacity for delineating which neuron subtypes underlie specific behaviors or physiological processes. To maximize scope of neuron subtypes amenable to silencing by RC::PFtox, and thus the breadth of applicability, we exploited a set of broadly active enhancer sequences – from R26 (Zambrowicz et al., 1997) and CAG (Niwa et al., 1991) – that, when coupled with the potency of tox, would be expected to equip RC::PFtox with the ability to suppress vesicular neurotransmission in a wide range of neuron subtypes (Farago et al., 2006; Muzumdar et al., 2007; Zong et al., 2005). Thus, RC::PFtox can leverage immediately the extensive panel of existing recombinase mouse lines by endowing them with the new capacity to tease out neuron function upon partnering with RC::PFtox.
We validated this RC::PFtox method of neuronal silencing by demonstrating, through a range of assays, effective and selective inhibition of the granule-to-Purkinje cell synapse in vivo. Breadth of applicability of RC::PFtox was established through partnering with different Cre and Flpe recombinase drivers to target GFPtox delivery, in separate experiments, to disparate neuron subtypes. Among the different populations tested, were drivers capable of efficiently activating GFPtox expression in near all serotonergic neurons or in just a combinatorially defined subset. Parcelation of a panel of serotonin (5HT) related behavioral phenotypes was suggested. These findings not only validate RC::PFtox and the utility of intersectional silencing, but also provide evidence that the neural substrates involved in contextual fear conditioning, sensorimotor gating, and certain anxiety-related behaviors may be separable based on genetic cell lineage, offering hope to the possibility for selective therapeutics.
Testing RC::PFtox required first applying it to an already well-defined neural system as proof of concept, and one not essential for viability, so as to allow postnatal analyses. We therefore targeted a major cerebellar circuit for silencing by partnering RC::PFtox with both Math1-cre (Matei et al., 2005) and hβact::Flpe transgenes (Rodriguez et al., 2000). In this test case, cell specificity was determined by the Math1-cre transgene, given that the β-actin (hβact) sequences driving Flpe act broadly and efficiently during embryogenesis, especially early in development resulting in an animal in which virtually all cells harbor the target allele in a Flpe-recombined form (Rodriguez et al., 2000). Although the high cell-type selectivity offered by the combinatorial platform of RC::PFtox was not fully exploited in this one example, given this breadth of Flpe recombination, this genotype allowed us to assess not only the dependency of GFPtox expression on Cre and Flpe events but also the efficacy of the remaining loxP-flanked stop cassette in preventing GFPtox transcription in cells having undergone only Flpe-mediated recombination – a feature critical for restricting GFPtox expression to only cells having undergone both Flpe- and Cre-recombinations.
Stop cassette function indeed proved adequate. Only triple transgenic Math1-cre, hβact::Flpe, RC::PFtox animals expressed GFPtox (Fig. 1), and only in neurons in which β-actin and Math1 enhancer elements were active at some point in their history, such as in mossy fiber precerebellar neurons, cerebellar granule cells, and neurons of the cerebellar nuclei (Fig. 1I–K). Recombination and consequent GFPtox delivery to Math1-cre-descendants was robust and highly efficient and specific, as evidenced by GFPtox detection (brown staining) in (1) the pontine grey nuclei (Fig. 1I versus control panel 1E), (2) the cerebellar granule cell layer (GCL) and the molecular layer (ML) which harbors the GFPtox-containing granule cell axons called parallel fibers (PFs), respectively (Fig. 1J versus control panel 1F) – note that Purkinje cells are devoid of GFPtox as expected given their lack of Math1 enhancer activity (Farago et al., 2006; Matei et al., 2005; Wang et al., 2005) – and (3) certain large output neurons in cerebellar nuclei (Fig. 1K versus control 1G). Furthermore, control double transgenic hβact::Flpe, RC::PFtox siblings showed no GFPtox expression, indicating that the remaining loxP-flanked stop cassette was sufficiently robust across cell types (Supplemental Fig. 1E). Animals of the other possible double transgenic genotype– Math1-cre, RC::PFtox – showed mCherry expression in the expected Math1-cre descendant territories (Fig. 1E–G insets) indicating efficient recombination of RC::PFtox by Math1-cre (see overlap between mCherry and DAPI signal in Fig. 1H); again, no GFPtox expression was detected, indicating that the remaining FRT-flanked stop cassette (Fig. 1B) was also sufficiently robust to block GFPtox expression, at least in the Math1-cre lineage. Thus GFPtox expression from RC::PFtox was strictly dependent on the combined actions of Cre and Flpe.
Granule cell axons, called parallel fibers (PFs), pack the molecular layer (ML) of the cerebellar cortex and synapse primarily onto Purkinje cells (PCs) (Supplemental Fig. 1D and Fig. 2). PF-to-PC synapses, and thus granule cell-produced VAMP2, is typically enriched in the ML. In triple transgenic Math1-cre, hβact::Flpe, RC::PFtox animals, little VAMP2 immunoreactivity was detectable in the ML as compared to control (white signal in Fig. 2B vs 2D), while the overall cytoarchitecture appeared normal as assayed by cresyl violet stain (Fig. 2A, C) as did Purkinje cells as assayed by Calbindin immunodetection (Fig. 2B, D insets and Supplemental Fig. 3). Moreover, we detected proteolytic cleavage fragments of VAMP2 only in protein lysates from triple transgenics but not control siblings (data not shown). Thus, the loss of VAMP2 immunoreactivity observed in triple transgenic animals likely resulted from GFPtox-mediated VAMP2 cleavage within granule cells.
Loss of VAMP2 predicts a disruption in exocytic fusion of synaptic vesicles, leading to impaired synaptic transmission. We tested this by determining the properties of the PF-to-PC synapse in acute brain slices. Stimulation of PFs with a 100 Hz train (Fig. 2E) evoked facilitating excitatory postsynaptic currents (EPSCs) in PCs (Fig. 2F). By contrast, in triple transgenics, the same stimulus evoked little detectable response in PCs (Fig. 2G). On average the synaptic responses were reduced more than 10-fold (in control and triple transgenics the averages were 324 pA and 22 pA and the medians were 118 pA and 3pA, respectively, n=11 each, p=0.0013 Mann-Whitney-Wilcoxon test (MWW)). In slices from triple transgenics it was sometimes possible to evoke EPSCs following PF activation, but extremely high stimulus intensities were required (Supplemental Fig. 2). The decrease in EPSC amplitude did not result from an inability to stimulate PFs in the triple transgenics because the evoked presynaptic volleys produced by propagating action potentials in PFs had similar properties in triple transgenics and littermate controls (Supplemental Fig. 2B, C). Next, we assessed the selectivity of synapse suppression. Climbing fibers (CFs) from neurons in the inferior olive also synapse onto PCs; in triple transgenics, however, CFs should be GFPtox-negative and unaffected. Indeed, we found that CF-to-PC EPSCs in both control and triple transgenic animals were similar in amplitude (Fig. 2H–J) (averages were 2.2 nA in control [n=7] vs. 2.4 nA in triple transgenic animals [n=10], p=0.31 MWW test) and exhibited short-term depression that is characteristic of this synapse (paired pulse ratio=0.29 in control and 0.27 in triple transgenics)(Konnerth et al., 1990). Collectively, these findings indicate a selective defect in neurotransmission between PFs and PCs. This defect is the result of granule cell dysfunction: (1) GFPtox-mediated depression of PF neurotransmitter release and (2) perhaps blockade of the mossy fiber-to-granule cell synapse (an upstream synapse), given that mossy fiber neurons are also Math1-cre-descendants and positive for GFPtox.
Consistent with these electrophysiological deficits and the known roles played by the various Math1-descendant neuron populations, we observed in triple transgenic Math1-cre, hβact::Flpe, RC::PFtox animals, robust and reproducible defects in gait, general motor coordination and balance (Fig. 2M, 2N and Supplemental movies). Also predicted is an accumulation of unreleased synaptic vesicles in PFs from triple transgenics (Schiavo et al., 2000); indeed, this was the case (Fig. 2K versus 2L). Thus, in triple transgenic Math1-cre, hβact::Flpe, RC::PFtox animals, PF neurotransmission appeared disrupted as judged by molecular, electrophysiological, and ultrastructural criteria.
Next, we partnered RC::PFtox with various other Cre and Flpe drivers to further sample its utility across different neuron types. For example, selective GFPtox expression in cerebellar PCs resulted in tremor upon movement and abnormal locomotion, while perinatal lethality resulted from broad GFPtox expression either throughout the midbrain and cerebellum or throughout the entire nervous system (summarized in Supplemental Table 1). In addition to these expected phenotypes, silencing serotonergic neurons also proved informative. Selective expression of GFPtox in central serotonergic neurons (Pet1-descendant neurons (Hendricks et al., 1999)) was efficiently and reproducibly achieved in triple transgenic ePet1::Flpe (Jensen et al., 2008), hβact-cre, RC::PFtox animals, with excellent concordance between GFPtox and 5HT immunodetection (Fig. 3A–H, and Supplemental Fig. 4). Furthermore, immunocytochemical analyses revealed that the serotonin (5HT)-positive axon varicosities, typical of serotonergic neurons (Maley and Elde, 1982), were enlarged in triple transgenics as compared to controls (Fig. 3H versus D, 5HT immunodetection), suggesting a tox-dependent build-up of 5HT-loaded vesicles and diminished neurotransmitter release. Both kinds of varicosities (Agnati et al., 2006) were likely affected: those at the axon terminal involved in synaptic vesicular neurotransmission, as well as those along the axon length involved in vesicle-mediated volume transmission because enlarged varicosities were observed both at target regions as well as throughout axon tracts. Given the diffuse nature of serotonin projections and the heterogeneity of input to the postsynaptic target neurons, our phenotyping focused on behavioral output rather than additional histological analyses or electrophysiological recordings. We found that triple transgenic animals (n=32) as compared to littermate controls (n=34) were more exploratory and less averse to open, brightly lit spaces, suggestive of a lowering of anxiety-associated behaviors (Fig. 3I, p<0.05). Triple transgenic females (n=16), especially, showed an anxiolytic response as compared to female controls (n=17), not only in the open field test but also in the zero-maze and light-dark tests (F(1, 62)= 3.95, p< 0.05 and F(1,62)=4.11, p< 0.05, respectively; data not shown). Additionally, triple transgenics (n=32) showed enhanced conditioned freezing to contexts (Fig. 3J, p=.044), suggestive of enhanced associative learning, and showed enhanced prepulse-mediated inhibition of the acoustic startle reflex, indicative of enhanced sensorimotor gating (Fig. 3K, p=.007).
Next, we used RC::PFtox to selectively deliver GFPtox to a subset of serotonergic neurons rather than to the entire 5HT system. In particular, we targeted the subset of Pet1-descendant 5HT neurons that arise from serotonergic progenitors in rhombomere (r) 1; in other words, those 5HT neurons that have a history of both Pet1 and En1 expression (Jensen et al., 2008). Consistent with our previously generated intersectional fate maps (Jensen et al., 2008), we observed, in triple transgenic En1-cre, ePet1::Flpe, RC::PFtox mice, GFPtox expression selectively in 5HT neurons of the dorsal raphe nucleus (DRN) in its entirety (B4, B6, and B7 nuclei using the Dahlstroem and Fuxe B1–B9 nomenclature (Dahlstroem and Fuxe, 1964)) along with certain 5HT neurons considered classically to be part of the median raphe nucleus (MRN, also referred to as the B5, B8, and B9 nuclei) (Fig. 4E, F and Supplemental Fig. 4). Thus, using RC::PFtox, we were able to efficiently and reproducibly mark with GFPtox the full contingent of r1-(En1-) derived 5HT neurons. Consistent with this GFPtox expression pattern, we observed enlarged 5HT varicosities associated with rostral projections (Fig. 4H); by contrast, caudal serotonergic fibers (non-r1-derived and thus GFPtox-negative in this experiment) showed varicosity sizes similar to littermate controls (Fig. 4H inset, versus D and D inset).
Phenotyping revealed behaviors consistent with lowered anxiety levels in triple transgenic En1-cre, ePet1::Flpe, RC::PFtox mice (n=20) as compared to littermate controls (n=21) (Fig. 4I). However, no behavioral differences were observed between triple transgenic En1-cre, ePet1::Flpe, RC::PFtox and control mice with respect to contextual fear conditioning and prepulse-mediated inhibition of the acoustic startle reflex (Fig. 4J, K). Thus, of the three behavioral phenotypes revealed upon expressing GFPtox in all Pet1-descendant 5HT neurons (Fig. 3), only the anxiety-related abnormalities were evoked upon silencing the r1-derived subset (Fig. 4I).
Altering the synaptic activity of select, genetically defined neuron subsets in an otherwise undisturbed mouse offers powerful means for delineating neuron functions. Here we describe the generation and validation of RC::PFtox mice that allow for the delivery of GFPtox, an inhibitor of synaptic vesicle exocytosis, to highly select neuron subtypes, while also being applicable across numerous, if not all, neuron subtypes. Activation of GFPtox expression was observed to be efficient and cell-type selective – consistent, in all cases, with the different Cre and Flpe drivers partnered with RC::PFtox. Importantly, GFPtox appeared well able to suppress vesicular neurotransmitter release in vivo, as determined by multiple independent means.
Upon applying RC::PFtox to Math1-cre-descendant neurons including granule cells and their PF fibers in vivo, we observed: (1) PF-specific cleavage and loss of VAMP2, the molecular target of tox action; (2) synaptic vesicle accumulation in PFs, diagnostic for inhibition of vesicle exocytosis; (3) a greater than ten-fold reduction in EPSCs in PCs following PF stimulation; (4) normal EPSCs in PCs following CF stimulation, indicating that the synaptic suppression was specific for the PF-PC synapse; and (5) dysfunctional gait and motor coordination consistent with disruption of cerebellar and precerebellar circuitry. Collectively, these findings provide strong support for the utility of RC::PFtox as a neuronal silencing tool. Further, our findings not only support but also extend the important studies of Yamamoto and colleagues (Yamamoto et al., 2003) in which they used a tetracycline inducible, less broadly applicable, system to deliver GFPtox to cerebellar granule cells (Yamamoto et al., 2003). A milder phenotype was reported likely reflecting many attribute differences including level and duration of GFPtox expression via the RC::PFtox approach and the more extensive cell populations targeted (cerebellar and precerebellar) in this particular proof-of-principle example. Importantly, RC::PFtox, as applied here to silence PF-to-PC synapses, offers now the ability to ascertain changes in Purkinje cells that follow selective blockade of just the PF input while maintaining intact all other input classes, such as from climbing fiber neurons, stellate cells, and basket cells. Indeed, many such exciting experiments are made possible.
Application of RC::PFtox to silence serotonergic neurons also proved informative, with the results both validating the utility and versatility of RC::PFtox and revealing of 5HT neuron functions as relates to genetic cell lineage. Expression of GFPtox in Pet1-descendant 5HT neurons resulted in mice that exhibited behaviors consistent with lowered anxiety levels, enhanced associative learning (enhanced conditioned freezing to contexts), and enhanced sensorimotor gating (as reflected in enhanced prepulse-mediated inhibition of the acoustic startle reflex). Validating these findings, and thus RC::PFtox, are reports of phenotypes reciprocal to these for mice in which the extracellular level of 5HT is expected to be increased, as opposed to the expected decrease here: among examples, mice null for Slc6a4, the gene encoding the serotonin reuptake transporter, show elevated levels of anxiety-like behavior (Bengel et al., 1998; Holmes et al., 2003); mice treated perinatally with a selective serotonin reuptake inhibitor (an SSRI) show elevated levels of anxiety-like behavior in adulthood (Ansorge et al., 2004); and mice given MDMA (3,4 methylenedioxymethamphetamine; Ecstasy), a serotonin releaser, show diminished prepulse inhibition (Dulawa and Geyer, 1996).
Interestingly, triple transgenic ePet1::Flpe, hβact-cre, RC::PFtox animals are most similar to mice null for the 5HT receptor 1B with respect to anxiety, exploratory behavior, startle reactivity, and sensorimotor gating (Dulawa et al., 2000; Dulawa et al., 1997; Malleret et al., 1999; Saudou et al., 1994; Zhuang et al., 1999). This commonality opens an avenue for thinking about behaviors that might be differentially served by 5HT1B receptors depending upon whether they are acting as heteroreceptors or autoreceptors – functions that have otherwise been challenging to delineate. The 5HT1B receptor localizes to axon terminals and upon activation inhibits neurotransmitter release; it acts either as an autoreceptor to limit 5HT release when expressed on serotonergic neurons or as a heteroreceptor to limit release of other neurotransmitters when expressed on nonserotonergic neurons (Maroteaux et al., 1992; Moret and Briley, 2000). Triple transgenic ePet1::Flpe, hβact-cre, RC::PFtox animals likely have diminished extracellular 5HT; those behaviors characterizing 5HT1B null mice that are shared with the triple transgenics are likely the result of functionally similar conditions. Loss of 5HT1B heteroreceptor function renders heterologous neurons deaf to 5HT, a situation functionally similar to loss of extracellular 5HT in triple transgenics; by contrast, loss of 5HT1B autoreceptor function would be expected to result in excessive extracellular 5HT. Thus, our data and the degree of phenocopy with 5HT1B null animals supports a larger heteroreceptor rather than autoreceptor role for the 5HT1B receptor in anxiety, exploratory behavior, and sensorimotor gating.
Of the phenotypes observed on silencing all Pet1-descendant 5HT neurons (reduced anxiety, enhanced conditioned freezing to contexts, and enhanced prepulse inhibition), only a reduction in anxiety-like behaviors was evoked upon silencing intersectionally the r1 (En1)-derived 5HT neuron subset. As we established previously (Jensen et al., 2008), and confirmed here by GFPtox expression, r1 (En1)-derived 5HT neurons account for all DRN 5HT neurons (also referred to as the B4, B6, B7 clusters) and an intermingled subset of MRN 5HT neurons (intermingled within the B5, B8, B9 clusters). The MRN also receives contributions from two other progenitor pools – the 5HT progenitors that situate in r2 and those in r3 (Jensen et al., 2008). Thus it appears that contextual fear conditioning and prepulse inhibition map away from the r1 (En1)-derived subset of serotonergic neurons, and, by subtraction, map onto the r2- and/or r3-derived 5HT neuron subsets in the MRN. Note that these behaviors are not thought to be modulated by the more caudal 5HT neurons of the medulla (the B1–B3 classes).
Nongenetic approaches, such as focal lesioning of 5HT neuron subsets by pharmacological means (injection of 5,7-dihydroxytryptamine, a neurotoxin preferentially, although not exclusively, taken up by 5HT neurons) have resulted in varied and often conflicting findings as to the relative importance of MRN (Avanzi et al., 1998); (Kusljic et al., 2003); (Melik et al., 2000) versus DRN (Maier et al., 1993; Maier et al., 1995; Sipes and Geyer, 1995) 5HT neurons in modulating contextual fear conditioning and prepulse inhibition. Our genetic findings support the MRN model, and in particular, the likely importance of the r2- and/or r3-derived MRN neuron subtypes. It now becomes possible to test this hypothesis further through partnering RC::PFtox and ePet1::Flpe with various rhombomere specific Cre drivers (Jensen et al., 2008). In this way it should be possible to better resolve different brain structures – different genetically defined 5HT neuron subsets – involved in specific 5HT modulated behaviors. Indeed, these initial intersectional findings are exciting because they support the notion that the neural substrates associated with these different behaviors may be separable, at least in part, by differences in genetic cell lineage (5HT populations defined by differences in developmental gene expression histories). These molecular differences, along with association with functional differences, suggest that selective therapeutics might one day become a possibility. This is of intense interest given that fear-conditioned learning is relevant to phobia, panic, and posttraumatic stress disorders (Garakani et al., 2006), that altered sensorimotor gating is likely affected in disorders such as schizophrenia, obsessive compulsive disorder, Tourette’s syndrome, and Huntington’s disease (Braff et al., 2001), and given the prevalence of and considerable impairment associated with generalized anxiety disorder, including comorbidity with clinical depression.
Now that tool efficacy has been established, it will be important to map other 5HT-modulated behaviors onto these different genetically defined subsets of 5HT neurons. One especially intriguing example is offspring nurturing, which recently has been shown to be dependent upon the level of serotonergic neuron function (Lerch-Haner et al., 2008).
In summary, RC::PFtox is the first broadly applicable tool for silencing highly selective, molecularly defined neuron populations in vivo. By endowing the hundreds of already existing recombinase mouse lines with the capacity to uncover neuron functions in vivo it is a powerful resource. Moreover, the capability to select cells for silencing based on a combinatorial gene expression signature has multiple advantages: (1) highly selective cell subtypes can be manipulated, sharpening the capacity for mapping behaviors and/or physiological processes to specific neuron subtypes; (2) the generated “neuron-to-behavior” map, having an associated molecular signature, may prove informative with respect to relating gene functions to the uncovered cellular functions; and (3) it offers added temporal resolution in two ways. First, the two recombinase driver genes do not have to coincide temporally, rather they may be expressed at different times in a cell’s developmental history with activation of GFPtox triggered only after the second recombination event has been completed. Second, inducible forms of Cre or Flpe can be employed, such as CreERT2 or FlpeERT2 (Dymecki and Kim, 2007), making it possible to silence cells at later points within the profile of a dynamically expressed driver gene. For example, using a Slc6a4:: CreERT2 (Gong et al., 2007) driver, we see induced postnatal recombination in virtually all 5-HT-staining neurons (Brust and Dymecki, unpublished data). While RC::PFtox offers molecular, spatial, and temporal control over GFPtox delivery in vivo, it does not, in its current form, offer reversibility of silencing – such enablement is the next step having established here the general efficacy and utility of the RC::PFtox approach across many neuron types.
While the intersectional aspect of RC::PFtox is an important advantage, some experiments nonetheless may warrant GFPtox delivery following a single recombination event. RC::PFtox readily offers this feature as well through the ready derivation of two additional mouse lines: one that requires only Cre recombination to activate expression of GFPtox – we refer to this derivative allele as RC::Ptox, because the loxP cassette remains while the FRT cassette has been removed by germline deletion; and one that requires only Flpe recombination to activate GFPtox expression – we refer to this derivative allele as RC::Ftox, because the FRT cassette remains following germline removal of the loxP cassette. These three tools – RC::PFtox, RC::Ptox, RC::Ftox – coupled with the present and ever growing number of Cre and Flpe expressing mouse strains has considerable potential for advancing studies of neuron function, fate, and behavioral output.
Annealed oligonucleotides containing multiple cloning sites (SnaBI, AsiSI, SrtI, BbvCI, and AvrII) and flanked by FRT sites were subcloned into the complementary AgeI site of pBig-T (Srinivas, 2001) to generate our pPF vector. His3-SV40pA STOP sequences were amplified by PCR from pBS302 (Sauer, 1993) and subcloned into the AvrII site of pPF to generate pPF-His3-SV40pA. Mouse codon-optimized cDNA encoding tox was PCR amplified from pUAST (gift from Dr. Sean Sweeney) and ligated with cDNA encoding EGFP (ClonTech). The resulting GFPtox fragment was subcloned into the SacII and NotI site of pPFII-His3-SV40pA. The resulting vector was refered as pPFII-His3-SV40pA-GFPtox. cDNA encoding mCherry was subcloned into the AsiSI site of pPFII-His3-SV40pA-GFPtox by blunt end ligation. The resulting vector was designated pPFII-Cherry-His3-SV40pA-GFPtox. The DNA region containing two stop cassettes and GFPtox was excised from pPFII-Cherry-His3-SV40pA-GFPtox using PacI and AscI sites and subcloned into the AsiSI and AscI sites of a CAG-MCS vector. The resulting vector was designated CAG-PFII-Cherry-His3-SV40pA-GFPtox. The fragment containing the CAG sequences, two stop cassettes, and GFPtox was excised using PacI and AscI and subcloned into the AscI and PacI site of pRosa26-1 (gift from Dr. Philippe Soriano) to generate pR26-CAG-PF-Cherry-GFPtox.
Linearized pR26-CAG-PF-Cherry-GFPtox (20 μg) was electrophorated into Tc-1 P5 ES cell (1–2 ×107 cells). ES cells were cultured and subjected to G418 selection. Mouse embryonic fibroblasts (cat # PMEF-N) and LIF ESGRO were purchased from Chemicon. Genomic DNA was isolated from 48 G418-resistant ES cell clones and analyzed by Southern. 14 clones were identified as homologous recombinants and one used in blastocyst injections to generate chimeric mice, with approval of the HMA Standing Committee on Animals. Triple transgenic mice expressing GFPtox in specific types of neurons were obtained by combining RC::PFtox alleles with various Cre and Flpe transgenes including Math1-cre (Matei et al., 2005), En1-cre (the cre knock-in allele, En1Cki) (Kimmel et al., 2000; Zervas et al., 2004), L7::cre (Barski et al., 2000), hβact::Flpe (Rodriguez et al., 2000), and ePet1::Flpe (Jensen et al., 2008) transgenes. Transmission of cre, Flpe, and RC::PFtox allele was assessed by PCR-based genotyping using the following primers.
Triple transgenic animals and control litter mates used in behavioral analysis were generated by crossing female Cre mice having FVB/N genetic background (hβact-cre) or C57B6 genetic background (En1-cre) with double transgenic male mice (ePet1::Flpe, RC:PFtox) having mixed genetic backgrounds of C57BL/6 and 129/SV. In all experiments, the pool of control animals was comprised of littermates to the experimental triple transgenics and were of either single transgenic (cre, Flpe, or RC::PFtox), double transgenic (any of the possible combinations), or nontransgenic genotypes.
Test was done as previously described (Kadotani et al., 1996). Mice were place on wood bar (5×5mm width, 70cm in length, 40cm above ground) and measured time that the animal spends on the bar. Each mouse was tested three times, and maximum 60secs were allowed for each session.
Animals were brought into the darkened testing area a minimum of 30 minutes before testing and allowed to acclimate. Mazes were dimly illuminated by 15W red bulbs suspended above the maze. To begin the test, animals were placed into a closed quadrant of the maze. The test session was 5 minutes in duration. Once an animal had been tested, it was placed in a holding cage until all animals from the home cage had been tested.
Animals were brought into testing area a minimum of 30 minutes (but ideally 45 minutes to 1 hour) before testing. Animals were allowed to acclimate to area. Open field session was 20 minutes in length. To begin the test, individual animals were removed from the cages and placed into the center of an open field monitor. Once an animal had been tested, it was placed in a holding cage until all animals from the home cage had been tested.
Animals were acclimated to the darkened room for a minimum of 30 minutes. A lamp with 15W bulb was located directly above the light portion of the light/dark box. To begin the 10 minutes test, animals were placed in the light half of the box. The guillotine door was then removed to allow animals to freely move between the two halves of the box. The amount of time spent in the light versus dark compartment was measured. Once an animal had been tested, it was placed in a holding cage until all animals from the home cage had been tested.
Animals were placed in the startle chamber with a 65 db background white noise and allowed to habituate. Over an approximately 15 minute session, 55 db pseudo-random trials were given. A 120 db white noise burst was used as the acoustic startle stimulus. Pre-pulses were 70, 80, and 85 db white noise bursts which preceded the startle stimulus by 10 ms. Startle response to the startle stimulus and to each of the pre-pulse db levels was measured.
The first part of fear conditioning (Training) was carried out approximately 1 ½ to 2 hours after the startle and prepulse inhibition test. Animals were placed in the fear conditioning chambers and allowed to habituate for 2.5 minutes. Animals were then presented with three pairings of an 85 db tone and 0.36mA foot shock. The tone was 30 seconds in duration and the shock was presented during the last 2 seconds of the tone. There was a 2.5 minute interval between each of the tone + shock pairings. On the day following the training session, animals were placed back into the same chambers where they underwent training. During this 6 minutes session, activity (beam breaks) per 30 second bin was measured and compared to activity during the habituation period on the training day. This procedure was used as a measure of conditioning to the context. Approximately 2 hr later, the behavior of the mice was tested in an altered context. The fear conditioning chambers were altered by placing a grey, square tile over the grid floor, placing a black Plexiglas insert over walls of the chambers, and attaching a small cup containing orange oil diluted in water in the upper corner of the box. Animals are allowed to explore the altered environment for 2.5 minutes, after which time, the conditioned stimulus (tone) was presented for 2.5 minutes. Activity (beam breaks) was evaluated in 30 second bins. This procedure was used as a measure of conditioning to the auditory cue.
All animals were tested in the elevated zero-maze, open field, light/dark, acoustic startle/prepulse inhibition, and fear conditioning tasks. On the first day of testing animals were tested in the elevated zero-maze. Starting the following day, and over a 4 day period animals were tested in the remaining tasks in the aforementioned order as described elsewhere (Cook et al., 2007).
All data were first analyzed using a two variable (genotype, sex) between subjects Analysis of Variance (ANOVA). Where significant results were obtained, post-hoc comparisons were performed using Tukey’s Least significant difference (LSD) test. A ‘p-value’ of p<0.05 was considered to be significant statistically.
Animals were transcardially perfused with 0.1 M phosphate-buffered saline (PBS) (pH 7.4) followed by 4% paraformaldehyde in 0.1 M PBS. Brains were harvested from the perfused animals, immersion-fixed in 4% paraformaldehyde in 0.1 M PBS at 4• 12 hours, and cryo-protected in 30% sucrose/PBS for 48 hours. 40 μm frozen sections were sliced using a cryostat, collected to PBS, and processed as free-floating sections for immunohistochemistry experiments using the Vectastain ABC Elite peroxidase rabbit IgG kit (Vector, USA, PK6101) for DAB/peroxidase-mediated antigen detection. For fluorescent immunohistochemistry, Cy2, Alexa 488, Cy3, or Cy5-conjugated secondary antibody was used for detecting primary antibody. Imaging of the fluorescent samples was performed on a Leica SP2 inverted confocal microscope using excitation with a 488nm laser for Cy2 and Alexa 488 fluorophore, a 594nm laser for Cy3, and a 633nm laser for Cy5.
Animals were transcardially perfused with 0.1 M phosphate-buffered saline (PBS) (pH 7.4) followed by 2% paraformaldehyde/2% glutaraldehyde fixative. Brains were harvested from the perfused animals, immersion-fixed in 2% paraformaldehyde/2% glutaraldehyde fixative at 4• for 12 hours. 100 μm sections were sliced using a vibrotome and collected to PBS. Samples were then washed several times in 0.1 M cacodylate buffer (pH 7.4) and osmicated in 1% osmium tetroxide/1.5% potassium ferrocyanide solution for 3 hours, followed by several washes with distilled water. 1% uranyl acetate in maleate buffer (pH 5.2) was added for one hour then washed several times with maleate buffer. This was followed by a graded cold ethanol series up to 100% which is changed three times over one hour. The samples were treated three times with propylene oxide over one hour, and then placed in 1:1 mixture of propylene oxide and plastic embedding resin including catalyst overnight. Samples were embedded in pure plastic the next day and put into 60• oven for one or two days. Sample blocks were cut for 95nm sections with Leica ultracut microtome, picked up on 100m formvar coated Cu grids, stained with 0.2% Lead Citrate, and imaged under the Philips Technai BioTwin Spirit Electron Microscope.
All procedures were approved by the Harvard Medical Area Standing Committee on Animals. Whole cell voltage clamp recordings were performed in Purkinje cells as previously described for parallel fiber (Dittman et al., 2000) and climbing fiber (Foster et al., 2002) synapses. Slices from P13–14 triple transgenic mice and age-matched littermate controls were maintained at 33–35° C. The extracellular ACSF contained: 125 mM NaCl, 26 mM NaHCO3, 25 mM glucose, 2.5 mM KCl, 1.25 mM NaH2PO4, 1 mM MgCl2, 2 mM CaCl2, and was bubbled with 95% O2/5% CO2. Glass electrodes (1–1.5 MΩ) were filled with an internal solution consisting of: 35 mM CsF, 100 mM CsCl, 10 mM EGTA, 10 mM HEPES. Bicuculline (20 μM) was added to the ACSF to block inhibitory currents. Parallel fiber EPSCs were recorded from Purkinje cells held at −60 mV in 275μm-thick transverse cerebellar slices. Glass stimulating electrodes (0.7–1MΩ) filled with ACSF were placed 250–300 μm from the cell body in the inner third of the molecular layer and care was taken to do this consistently across experiments. CGP55845A was added to the ACSF to block GABAB receptor activation. Climbing fiber EPSCs were recorded from Purkinje cells held at −30 mV in 250μm-thick parasagittal slices. NBQX (300 nM) was included in the external solution to minimize series resistance errors and allow quantification of the otherwise prohibitively large climbing fiber EPSCs.
Supplemental Figure 1. Intersectional selectivity of neuronal silencing by RC::PFtox. (a,b) Unique gene expression codes distinguish select neuron subtypes. A grid represents an anatomical area populated by different types of neurons. A neuron type positive for expression of gene A is depicted in yellow, gene B in blue (A). A select neuron subtype (the intersectional population) is positive for expression of both gene A and gene B and is depicted in green. When geneA::cre and geneB::Flpe are coupled with RC::PFtox, only cells expressing both Cre and Flpe produce GFPtox (B). When geneA::cre alone is combined with RC::PFtox, cells are marked by mCherry. (C) Tetanus toxin light chain (tox) has been used extensively to study neuron function in transgenic fruit flies and now transgenic mice. At the molecular level, tox suppresses vesicle-mediated neurotransmitter release by cleaving the synaptic vesicle-associated membrane protein VAMP2/synaptobrevin2. In the absence of VAMP2, assembly of the SNARE protein complex, needed for exocytic fusion of synaptic vesicles with plasmalemma, is inhibited. (D) Schematic diagram of aspects of the main cerebellar circuit. Activity in the mossy fiber axon, originating from brainstem nuclei such as the PGN (pontine gray nuclei), excites granule cells (GC) in the cerebellar cortex. Granule cells project parallel fiber axons to the molecular layer of the cerebellar cortex, where they form excitatory synapses with Purkinje cells (PC) dendrites. PCs form inhibitory synapses with neurons in the cerebellar nuclei (DCN). Arrows indicate the direction of axon depolarization. Plus and minus signs indicate the excitatory and inhibitory synaptic connection, respectively. (E) GFPtox and mCherry expression is undetectable in double transgenic hβact-cre, RC::PFtox controls tissue. (tissue analyses similar to that in Fig. 1E–K).
Supplemental Figure 2. Characterization of synaptic currents in control and triple transgenic hβact::Flpe, Math1-cre, RC::PFtox mice. Experiments were conducted as in Fig. 2E. PFs were activated with a range of stimulus intensities and EPSCs were measured in Purkinje cells (n=20 triple transgenic; n=16 control). (A) In slices from control animals, increases in stimulus intensity evoked progressively larger EPSCs (black). At intensities above 40 μA the EPSCs were so large that series resistance errors made them difficult to quantify. In slices from triple transgenic mice (red), stimulus intensities below 60 μA typically failed to evoke synaptic currents, but higher stimulus intensities evoked small EPSCs. Box plots display the 10th, 25th, 50th (thick line), 75th and 90th percentile responses. Statistical significance was determined using the Mann-Whitney-Wilcoxon test. (B–D) The number of PF synaptic inputs activated can depend upon several factors including the angle of the slice, introducing variability in EPSC amplitude measured for a given stimulus intensity across slices. PF stimulation triggers action potentials in presynaptic fibers that produce currents known as presynaptic volleys, which provide a measure of the number of activated PFs. Presynaptic volleys (prespikes) were quantified in whole-cell recordings. (B) We found no statistical difference in the amplitude of prespikes evoked in control and triple transgenics (mean ± SEM, t-test). (C) Representative traces showing the presynaptic volleys (indicated with arrows) and EPSCs evoked in a control animal with a 20 μA stimulus intensity (black, left) and in a triple transgenic with 20 μA (red, center) and 60 μA (red, right) stimuli. Note that for the same stimulus intensities the amplitudes of the prespikes were similar, but the EPSC was much larger in the control. In the triple transgenic, EPSCs were observed at high stimulus intensities that recruited more fibers and greatly increased the amplitude of the prespike. (D) To take into account the number of activated fibers, we determined the ratio of EPSC to prespike amplitude for control (black, n=17) and triple trangenics (red, n=13). The large currents evoked in control conditions (and therefore the control ratio) are likely underestimated due to series resistance errors, but the difference in these ratios provides a conservative estimate that the EPSC is reduced by about a factor of 9 in triple transgenic animals compared to controls. (E,F) If tetanus toxin reduced the probability of release then facilitation would be much more prominent in triple transgenic animals. If this were the case, high frequency stimulation could in essence rescue transmission. To determine whether use-dependent plasticity was altered in triple transgenic animals, we examined responses in Purkinje cells from those animals for which higher intensity stimulation was able to evoke responses. (E) Representative traces showing EPSCs evoked by a 100 Hz train in control (black) and triple transgenic (red) animals. (F) The extent of facilitation of the second (left) and 10th (right) EPSCs in the train increased slightly but not significantly in triple transgenic animals (n=15) compared to control (n=7; mean ± SEM, t-test).
Supplemental Figure 3. Cerebellar Purkinje cells remain intact upon disruption of synaptic transmission from granule to Purkinje cells. Cerebellar Purkinje cells were examined by calbindin immunostaining (red signal) of sagittal sections taken from triple transgenic animals (Math1-cre, hβact::Flpe, RC::PFtox; top row) versus control littermates (hβact::Flpe, RC::PFtox; bottom row) at age P21. No significant difference was noticed in either Purkinje cell number or morphology judged by calbindin immunoreactivity between triple transgenics and control littermates.
Supplemental Figure 4. Recombination in both triple transgenic animals (hβact-cre, ePet::Flpe, RC::PFtox; versus En1-cre, ePet::Flpe, RC::PFtox) is highly efficient and reproducible. A. Dorsal raphe region in coronal section from two different triple transgenic genotypes (hβact-cre, ePet::Flpe, RC::PFtox, first, third, and fourth row; versus En1-cre, ePet::Flpe, RC::PFtox, second row) was examined by double immunostaining with GFP (green signal), 5HT (red signal) antibody, DAPI (white signal). In both transgenic genotypes, co-labeling shows excellent concordance, suggesting that Cre- and Flpe-mediated recombination were very efficient. B. GFPtox (brown signal) expression was not only highly specific but also very reproducible between animals as shown in four representative examples of GFPtox expression in dorsal raphe region.
Supplemental Table 1. Summary of neuronal silencing achieved by a variety of Cre and Flpe lines.
Supplemental Movies 1–4. Movies capturing representative behaviors of triple transgenic animals (Math1-cre, hβact::Flpe, RC::PFtox) and their control littermates (Math1-cre, RC::PFtox) during assessments of motor coordination and balance. Movies A and B, gross walking patterns of triple transgenics (A) versus controls (B); movies C and D, fixed-bar test for triple transgenics (C) versus controls (D). Triple transgenic animals showed an ataxic walking pattern with a significant delay in initiating movements and performed poorly in the fixed-bar test.
We thank members of the Dymecki lab and SIDS Program Project Grant (P01 HD036379) for input, Drs Sweeney and Yamamoto for plasmids encoding tetanus toxin light chain, Dr. D. Rowitch for providing the Math1-cre transgenic, Dr. Alexandra Joyner for providing the En1Cki. This work was supported by grants from the Foundation for Fighting Blindness - Canada (to J.C.K.), the Helen Hay Whitney Foundation (to M.R.C.), and from the US National Institutes of Health (R01 DK067826, R21 MH083613-01, and NIH P01 HD036379 to S.M.D.; R37 NS032405 to W.G.R.; and R01 DA020677 with subcontract to M.N.C.).