|Home | About | Journals | Submit | Contact Us | Français|
Our understanding of current treatments for depression, and the development of more specific therapies, is limited by the complexity of the circuits controlling mood and the distributed actions of antidepressants. Although the therapeutic efficacy of SSRIs is correlated with increases in cortical activity, the cell types crucial for their action remain unknown. Here we employ bacTRAP translational profiling to show that layer 5 corticostriatal pyramidal cells expressing p11 (S100a10) are strongly and specifically responsive to chronic antidepressant treatment. This response requires p11 and includes the specific induction of Htr4 expression. Cortex-specific deletion of p11 abolishes behavioral responses to SSRI’s, but does not lead to increased depression-like behaviors. Our data identify corticostriatal projection neurons as critical for the response to antidepressants, and suggest that the regulation of serotonergic tone in this single cell type plays a pivotal role in antidepressant therapy.
Mood is governed by the actions of a complex, anatomically dispersed circuit comprised of specific subpopulations of neurons arranged into functional units. Functional imaging studies of patients suffering from major depressive disorder (MDD), bipolar disorder and anxiety indicate that a fundamental neural circuit controls emotion, with different elements of the circuit contributing to specific aspects of emotive behavior (Drevets et al., 2008; Mayberg, 2009). For example, fMRI and PET imaging of patients suffering from MDD has consistently demonstrated decreased prefrontal cortex function, and alterations in the activity of subcortical structures that include the basal ganglia, amygdala and thalamus (Drevets, 2000; Mayberg et al., 1999). There is general agreement that in MDD, disease and treatment mechanisms impact the same basic circuitry controlling emotional state. However it is evident from deep brain stimulation studies in a number of neurological and psychiatric disorders that the pathological mechanisms causing dysfunction and the immediate targets of clinical interventions most effective in alleviating symptoms may involve different brain structures. Given this realization, our understanding of the pathophysiology of depression as well as the development of improved therapies for this disorder will be advanced by identification of cell types and molecular mechanisms responsible for generating depression-like phenotypes, as well as those mediating responses to antidepressant treatment.
Recently, p11 (the protein product of the S100a10 gene) was found to be an important factor mediating antidepressant responses and depression-like states (Svenningsson et al., 2006). P11 is an adaptor protein that is expressed specifically in the CNS (Egeland et al., 2011). It regulates serotonin signaling by binding to serotonin receptors (Htrs) 1b, 1d, and 4 and stabilizing the localization of these receptors at the cell surface (Svenningsson et al., 2006; Warner-Schmidt et al., 2009). Decreased p11 levels were found in the cortex of MDD patients, suicide victims, and a mouse model of depression (Anisman et al., 2008; Svenningsson et al., 2006). Chronic antidepressant treatment, electroconvulsive therapy, and BDNF treatment all result in increased p11 expression in the cerebral cortex (Svenningsson et al., 2006; Warner-Schmidt et al., 2010). Importantly, mice lacking p11 exhibit depressive-like behaviors, increased anxiety, and a blunted behavioral response to antidepressant treatment (Svenningsson et al., 2006; Warner-Schmidt et al., 2009).
Antidepressant drugs target neuromodulatory systems that have widespread effects throughout the CNS, engaging receptors that are broadly expressed in the brain. While the pharmacological effect of these drugs is immediate, there is a therapeutic delay of weeks to months before antidepressant activity is evident. This delay is thought to reflect neuroadaptive changes in pre- and postsynaptic cells, including long-term changes in gene expression and protein translation (Krishnan and Nestler, 2008). Although the cell types and precise molecular mechanisms mediating the efficacy of antidepressant drugs have not been identified, neuroimaging studies have shown that the clinical effects of antidepressant drug therapy and deep brain stimulation correlate with increased activity in the cerebral cortex (Drevets et al., 2008; Mayberg, 2009). Given these observations, and the dramatic regulation of p11 in the frontal cortex of depressed patients, we wished to determine whether p11 expressing cells in the cerebral cortex are critically important for antidepressant action, and to search for beneficial adaptations that may occur in these cells in response to chronic antidepressant treatment.
For this study, we created an S100a10 bacTRAP mouse line to characterize p11 expressing cells and measure their responses to antidepressant treatment (Doyle et al., 2008; Heiman et al., 2008). We report that p11 is highly enriched in layer 5 corticostriatal (CStr) projection neurons, that these cells respond preferentially to chronic SSRI treatment by altering serotonergic tone, and that loss of p11 in the cortex results in the inability to respond to an SSRI. Our data demonstrate that the beneficial actions of antidepressant therapy can be mediated by a single cell type in the cerebral cortex, and suggest that development of drugs that specifically target the activity of CStr neurons can result in improved therapies for depression.
Previous in situ hybridization (ISH) studies have shown that the S100a10 gene is expressed in the cerebral cortex (Arlotta et al., 2005; Svenningsson et al., 2006), although the cell types expressing S100a10 in layer 5 and their position in CNS circuitry have not been determined. To identify this cell population and study its properties using the Translating Ribosome Affinity Purification (TRAP) approach, bacTRAP transgenic mice were generated using a BAC clone containing the S100a10 gene. The EGFP-L10a ribosomal protein fusion (Heiman et al., 2008) was engineered into the S100a10 BAC by homologous recombination in E. Coli (Gong et al., 2002) so that it would be expressed under the control of S100a10 regulatory elements (Figure 1A). S100a10 bacTRAP founder line ES691 was chosen for further analysis because the pattern of transgene expression reliably recapitulated endogenous expression patterns of the S100a10 gene throughout the brain (Figure S1).
In the cortex, EGFP-L10a was expressed in a laminar pattern (Figures 1B, C, and S1A), consistent with previous reports (Arlotta et al., 2005; www.gensat.org; www.brain-map.org; www.stjudebgem.org). Transgene expression was primarily restricted to superficial layer 5 (layer 5a), although a few cells with low levels of EGFP-L10a expression could be found in layers 4 and 5b. EGFP-L10a expression was localized to the soma, proximal dendrites, and the nucleolus, as previously described for other bacTRAP mouse lines (Doyle et al., 2008; Heiman et al., 2008). At high magnification it was evident that somas of EGFP-L10a+ cells were pyramidal, with a clear apical dendrite (Figure 1D). As shown in Figure 1E, all EGFP+ cells were also labeled with anti-p11, and the vast majority of p11+ cells also expressed the fusion protein. These results confirm previous studies demonstrating laminar expression of the S100a10 gene in the cortex, identify these cells as layer 5a pyramidal cells, and demonstrate that the ES691 bacTRAP transgenic line faithfully reproduces expression of endogenous S100a10 in the mouse cortex.
Inspection of an S100a10 EGFP transgenic reporter line (www.gensat.org) revealed axons within the corpus callosum and striatum, and no labeling in the thalamus or pyramidal tract, suggesting that the S100a10 expressing cells project primarily to the striatum. To determine definitively the axonal targets of the pyramidal cells labeled in the S100a10 bacTRAP line, retrograde tracing experiments were performed using fluorescently-tagged cholera toxin β subunit (CTβ). These studies demonstrate that pyramidal cells labeled in the S100a10 bacTRAP line provide a major afferent input to the dorsal striatum (caudate putamen, CPu). Thus, 72±5% of the cortical neurons labeled by injecting CTβ into CPu were also EGFP-L10a+ (Figure 2A, C). Injections into different areas of the CPu resulted in labeling of S100a10 neurons in diverse areas of cortex, reflecting the topographic map of CStr projections (Figure S2D). A large proportion of cells that were double positive for CTβ and EGFP-L10a were also found in contralateral cortex following injections into CPu (Figure 2C), indicating that the S100a10 cells are capable of making both ipsilateral and contralateral projections to the dorsal striatum. The laminar distribution of EGFP-L10a+ and CTβ+ substantially overlapped (Figure 2B), consistent with previous studies demonstrating that a majority of striatally projecting cortical neurons are present in layer 5a.
Since the CPu and nucleus accumbens (NAc) are known to be functionally distinct, and since the majority of cortical inputs to the NAc arise from medial prefrontal cortical areas, we employed CTβ retrograde tracing to ask whether S100a10 expressing neurons project to the ventral striatum. Although CTβ and EGFP-L10a labeled cell bodies were intermingled in frontal cortex, there was minimal overlap between accumbal projecting neurons and S100a10 cells (1.9±1.7% ipsilaterally and 3.9±1.4% contralaterally; Figure 2A,C). Rather, medial prefrontal S100a10 cells co-labeled with injections to anteromedial CPu while those in lateral regions sent projections primarily to anterolateral CPu (Figure S2 and data not shown). We conclude that the major afferent input from the cortex to the CPu originates in S100a10 expressing pyramidal cells, but that direct projections from the frontal cortex to the NAc do not arise from this cell population.
As it is common for CStr neurons to send collateral projections to contralateral cortex, we examined whether this was the case for S100a10 cells. When CTβ was injected into somatosensory cortex, CTβ+ commissural projecting neurons (CPNs) were widely distributed across the cortex and showed no laminar patterning (Figure 2A, B). However, we found that 66±8% of CTβ+ cells in layer 5 of contralateral cortex were also EGFP+ (Figure 2C). Although we cannot definitively conclude that individual S100a10 expressing neurons that project to the striatum also send collaterals to contralateral cortex, strong support for this suggestion comes from the fact that the majority of layer 5 cells labeled by retrograde tracing with CTβ from either the dorsal striatum or the contralateral cortex are EGFP-L10a+.
Finally, we note that retrograde tracing from the cervical spinal cord revealed little overlap between corticospinal tract (CST) projection neurons in layer 5b and S100a10 expressing cells (Figures 2B and S2B), despite the fact that long distance projection neurons of this class are known to send axon collaterals to the striatum. Taken together, our results demonstrate that S100a10 expressing pyramidal cells constitute a major class of layer 5 neurons whose projections are limited to forebrain structures (Figure 2D). The fact that these neurons send afferents specifically to medial and lateral CPu, and that they do not provide inputs to the NAc, is interesting given the concept of multiple, parallel, and functionally segregated pathways from the cerebral cortex to the striatum that has emerged from rodent and primate neuroanatomical studies over the last several decades (Voorn et al., 2004).
To compare the molecular properties of a second well characterized pyramidal cell type to the S100a10 CStr cell population, we chose Glt25d2 expressing corticopontine neurons. Glt25d2 expressing pyramidal cells are long-distance projection neurons that reside in layer 5b and send axons to pons, thalamus, and spinal cord (Doyle et al., 2008; Groh et al., 2010). Their dendritic architecture and electrophysiological properties distinguish them from CStr cells (Groh et al., 2010). Glt25d2 cells are present deeper in layer 5 than S100a10 neurons in multiple areas of cortex and cell somas of Glt25d2 cells are larger than those of S100a10 cells (Figure S3), consistent with previous observations (Groh et al., 2010).
The TRAP strategy allows identification of all proteins being synthesized in any genetically targeted cell population, and alterations of this translational profile in response to pharmacological perturbations (Doyle et al., 2008; Heiman et al., 2008). To identify molecular distinctions between S100a10 and Glt25d2 expressing cells, we isolated polysome-bound mRNA from these cell types and assayed global differences in gene expression using microarrays. Because the S100a10 gene and the corresponding EGFP-L10a bacTRAP construct are also expressed in the pia mater and some blood vessels in the brain (Figure S4A–D), we limited our analysis to a subset of 12,235 probe sets (representing 8,390 genes) that are expressed specifically in neurons (Table S1). As shown in Figure 3A, comparative analysis of mRNAs immunoprecipitated (IP) from S100a10 cells versus those present in the whole cortex (input) demonstrates selective enrichment of S100a10 probe sets in the IP (7.5 and 5.9 fold). Quantitative RT-PCR (qRT-PCR) of TRAP RNA showed that mRNA for the vesicular glutamate transporter, Slc17a7, was expressed (although not enriched compared to neighboring cells), confirming that the S100a10 cells are indeed excitatory neurons (Figure 3B). In contrast, mRNAs specific to astrocytes (Aldh1l1), oligodendrocytes (Cnp), and GABAergic interneurons (Gad1) were expressed at very low levels compared to the input (Figure 3B). mRNA for genes known to label other pyramidal cell types (Doyle et al., 2008), including layer 6 corticothalamic (Ntsr1), a subpopulation of layer 5 CStr (Etv1), and Glt25d2 were not enriched in IP mRNA (Figure 3B, D), suggesting that the S100a10 cells are a unique, previously uncharacterized cell population.
To assess whether other genes enriched in the S100a10 IP distinguish these cells from Glt25d2 expressing pyramidal cells, mRNAs purified from these two cell types using the TRAP protocol were compared. This analysis resulted in identification of 1,112 differentially expressed mRNAs (data not shown). Genes previously shown to be expressed in CST projection neurons were more highly enriched in the Glt25d2 cells, whereas those demonstrated to be expressed in CPNs in cortical layer 5 were preferentially expressed in the S100a10 cells (Figure 3C and Table S2). To demonstrate that the bacTRAP data collected from the S100a10 and Glt25d2 pyramidal cell populations provide cell type specific information even for those genes that have previously been identified as expressed in layer 5 of the cerebral cortex, and to identify useful additional markers for CStr neurons, we compared our data with genes annotated as specifically expressed in layer 5 by the Allen Brain Atlas. Although the ISH data demonstrate laminar expression in layer 5 for each of the six examples shown (Figure 3E–J), comparative analysis of the translational profiles of S100a10 and Glt25d2 pyramidal cells identify Akap12, Depdc6, D6Wsu176e, Dexi, and Emb as specifically expressed in CStr cells, and D330050I23Rik as enriched in corticopontine neurons (Figure 3D). These results demonstrate the value of bacTRAP data for definitively identifying the specific cell type expressing a gene of interest and for determining the distinguishing molecular characteristics of closely related cell types, as well as illustrate the utility of ISH databases as tools for validation of bacTRAP datasets.
Given human neuroimaging studies implicating CStr circuits in the response to antidepressants (Drevets et al., 2008), previous studies demonstrating the induction of p11 in the cerebral cortex by antidepressants and electroconvulsive therapy (Svenningsson et al., 2006; Warner-Schmidt et al., 2010), and the present demonstration that S100a10 expressing cells provide a major and specific pathway for communication between the cortex and dorsal striatum, we were interested in examining the possibility that these cells respond preferentially and specifically to antidepressant treatment. We chose the selective serotonin reuptake inhibitor (SSRI) fluoxetine (FLX) because it is one of the most highly prescribed antidepressants and is known to have antidepressant activity in rodent models of depression (Cryan et al., 2005; Dulawa et al., 2004). It was previously reported that chronic administration of tricyclic antidepressants imipramine or tranylcypromine increases the amount of p11 in the cortex in mice (Svenningsson et al., 2006). To assess whether FLX had a similar effect, we treated wildtype (WT) mice with FLX or saline for 14 days, and measured levels of p11 mRNA in the cortex by ISH. As expected, p11 expression increased ~25% in response to FLX (Figure 4A), confirming the previous observation that p11 expression in cortex is sensitive to antidepressant action (Svenningsson et al., 2006).
To identify long term neuroadaptations that occur in response to SSRI treatment, we administered FLX (0.167 mg/ml) or vehicle (VEH) to S100a10 and Glt25d2 bacTRAP mice in their drinking water for 15–18 days. No change in the distribution of cells expressing EGFP-L10a occurred in either bacTRAP line in response to treatment (data not shown). Polysome-bound RNA from the cortex of each line was collected using TRAP, and microarray analysis was used to examine the effect of chronic FLX on the profile of proteins being synthesized by each cell type. Sixty-two genes (65 probe sets) were differentially regulated by FLX in the S100a10 cells, as opposed to only four in the Glt25d2 cells (Figure 4B; Table S3). These data confirmed the hypothesis that the S100a10 (and thus p11) expressing cells are preferentially sensitive to antidepressant treatment, and suggest that this population of CStr cells might play an important role in antidepressant therapy.
Given the facts that p11 modulates Htr function (Svenningsson et al., 2006; Warner-Schmidt et al., 2009), and that chronic treatment with SSRIs results in elevated levels of serotonin in the nervous system, we asked whether the response of S100a10 expressing cells to FLX included long term changes in the expression of Htrs. We restricted our analysis to Htrs expressed in cortical neurons (see Table S1). As shown in Figure 4C, a dramatic and specific increase in Htr4 mRNA (16-fold) occurred in S100a10 cells following chronic FLX treatment. Elevated Htr4 expression was not observed in Glt25d2 neurons and no increases in the expression of other Htr subtypes were observed in either cell type. Further, this increase in Htr4 expression was not detectable in whole cortex (input) samples (Figure 4D). Taken together, our data both demonstrate that layer 5 CStr projection neurons are highly responsive to antidepressant treatment, and suggest that the induction of Htr4 expression in this critical cell population may play an important role in restoring the balance between cortical and striatal activity that occurs in response to SSRIs.
p11 directly interacts with multiple Htrs, and its loss leads to reduced responses of cortical cells to serotonin at striatal synapses (Svenningsson et al., 2006). To determine whether p11 is involved in the molecular adaptations that we have observed in CStr neurons in response to SSRIs, we crossed S100a10 bacTRAP mice with p11 knockout (KO) animals (Svenningsson et al., 2006) to obtain animals (S100a10 bacTRAP/p11KO) in which the molecular consequences of p11 deletion could be assessed in CStr pyramidal cells. EGFP-L10a transgene expression was not significantly altered in the cortex of KO mice (Figures 1E and and5A),5A), and retrograde tracing revealed that EGFP expressing cells still accounted for a majority of projections to CPu and contralateral cortex (Figure S5A). As shown in Figure 5B, deletion of p11 resulted in severe dysregulation of Htrs in these neurons. Htrs 1a, 1b, 2c, 5a, and 7 were all significantly down-regulated in the KOs. No significant change was observed in the translation of Htr1d or Htr4, both of which bind directly to p11. However, given previous studies demonstrating that the cell surface expression of Htr1b and Htr4 is critically dependent on p11 (Svenningsson et al., 2006; Warner-Schmidt et al., 2009), we expect that signaling through these receptors is also reduced. Taken together, our data suggest that deletion of p11 results in a loss of serotonergic tone in CStr pyramidal cells.
We next tested whether the alterations in Htr expression in S100a10 cortical neurons in p11 KOs resulted in dampened responses to FLX. S100a10 bacTRAP/p11KO mice received chronic administration of FLX or VEH as described above, and translational profiles from these cells were assayed on microarrays. As shown in Figure 5C, p11 KO animals failed to show the robust response to FLX observed in S100a10 CStr neurons. Thus, the majority of genes shown to be significantly regulated in the S100a10 cells of WT animals were no longer regulated by FLX in the p11 KO, and the induction of Htr4 noted in WT cells was significantly blunted (Figure 5D; p<0.05, Student’s t-test). We conclude that the actions of SSRI’s on S100a10 CStr projection neurons are critically dependent on p11.
Given our bacTRAP molecular phenotyping data establishing S100a10 CStr pyramidal cells as strongly and specifically responsive to chronic SSRI treatment, and our demonstration that these cells provide the majority of afferents to the dorsal striatum from the cortex, we were interested in testing the role of S100a10 cortical neurons in the therapeutic actions of antidepressants. Since CStr cells in p11 KOs failed to show molecular responses to FLX and have altered expression of Htrs, we generated cortex specific p11 KO animals (Emx1/p11KO) to assess the functional consequences of loss of these adaptations on behavior. Immunohistochemistry was employed to confirm the loss of p11 from the cortex, and its continued expression in the striatum and in other subcortical sites (Figure 6A; data not shown).
We employed the novelty suppressed feeding (NSF) paradigm and the tail suspension test (TST) to assess the therapeutic actions of chronic FLX in Emx1/p11KO mice. In NSF, food-deprived mice are placed in an open field that contains food in the center. The mice must decide to approach and consume the food, or avoid the novel environment. Animals treated with antidepressants for several weeks before testing consistently exhibit shorter feeding latencies than those that received acute or no treatment, making NSF a valuable assay for evaluating responses to chronic antidepressant administration (Dulawa and Hen, 2005). We found that Emx1/p11KO mice failed to exhibit behavioral responses to chronic FLX treatment in NSF (Figure 6B). WT mice (carrying the floxed p11 allele but not Cre) showed a significantly decreased latency to feed (436±36 s) following chronic FLX administration compared to VEH treated controls (561±25 s). In contrast, there was no difference in latency between VEH and FLX administration (523±32 vs. 469±41 s) in Emx1/p11KO mice. VEH-treated Emx1/p11KOs showed no difference in baseline anxiety (as measured by latency to feed) compared to WT mice (Figure 6B, white bars). Further supporting the conclusion that the behavioral differences in the NSF test likely reflect an inability of the cortex specific KOs to respond to drug treatment rather than a change in generalized anxiety, mice of all genotypes displayed no difference in thigomaxis in an open field test.
We also tested Emx1/p11 KO mice in the TST, which is a classic behavioral test widely employed to screen antidepressant efficacy in mice (Cryan et al., 2005). The TST measures the time mice spend immobile while being suspended by their tail, an inescapable stress. Following chronic FLX treatment, WT mice exhibited a significant reduction (~20%) in immobility in the TST and this effect was abolished in Emx1/p11KOs (Figure 6C). There was no difference between genotypes in locomotor activity (Figure S5B), and importantly there was also no indication of a depressive-like effect in Emx1/p11 KO on immobility in the TST (Figure 6C, white bars), or anhedonia in the sucrose preference test (Figure 6E). These data further support the notion that the behavioral phenotype of Emx1/p11KO mice lies in their lack of response to FLX treatment and not to an elevated depressive-like state.
Emx1/p11KO mice also failed to display a behavioral response to chronic FLX in the forced swim test (FST; Figure S5C). WT animals showed the expected decrease in immobility following chronic FLX treatment, but it did not reach statistical significance (p=0.08). This is likely due to genetic background, since our mice were backcrossed onto the C57BL/6 strain and it has been shown that chronic, but not acute, FLX treatment in drinking water reduced immobility in the FST in BALB/c mice but not C57BL/6 mice (Dulawa et al., 2004). We conclude that proper function of S100a10 CStr pyramidal cells is required for normal behavioral responses to chronic antidepressant treatment.
Human neuroimaging studies have consistently demonstrated alterations in cortical activity in depressed patients and normalization of this activity in response to antidepressants, although neither the cell types altered in response to treatment nor the nature of the neuroadaptations that occur in these cell types have been identified. In this report, we show that CStr projection neurons expressing S100a10 (p11) are essential for both the molecular and behavioral responses to the SSRI FLX. We demonstrate that S100a10 pyramidal cells project primarily to the dorsal striatum and contralateral cortex, and that they experience long term adaptive changes in response to chronic administration of FLX that do not occur in other layer 5 projection neurons. These adaptations require p11 and include the induction of Htr4 receptor expression. Finally, we show that deletion of p11 in these cells abolishes behavioral responses to FLX, but does not lead to a depressive-like phenotype. Our data identify S100a10 CStr neurons as essential for the actions of SSRI antidepressants, reveal specific molecular adaptations that occur in these cells in response to chronic treatment, and suggest that modulation of serotonergic tone in this single cortical cell population can result in effective antidepressant therapy.
Molecular phenotyping of cell types that comprise the neural circuits underlying pathological neuropsychiatric conditions and their responses to therapy provides one avenue for development of more specific and effective treatments. Until recently, the histological complexity of the CNS has provided a barrier for such studies, especially in the cerebral cortex where few molecular markers exist to distinguish between subpopulations of morphologically similar pyramidal cells. In this study, we took advantage of the discovery that p11 is regulated in the cortex following antidepressant treatment (Svenningsson et al., 2006) and of the recently developed bacTRAP approach for translational profiling of specific CNS cell types (Doyle et al., 2008; Heiman et al., 2008) to study long term changes that occur in specific cortical neuron populations in response to antidepressants. Our results demonstrate that bacTRAP translational profiling can be used to identify cell types that respond to therapeutic drugs, and provide a first illustration that those cells most responsive to treatment may also participate in the therapeutic actions of the drug. It will be of great interest to employ this strategy in other mouse models of psychiatric and neurological disease to explore further the relationship between molecular phenotype and cellular function in the setting of complex, circuit based disorders.
Our data argue strongly that the specific molecular changes that occur in relevant cell types in the context of effective treatment of disease may provide critical insights into mechanisms that determine therapeutic efficacy. For example, we have shown here that chronic treatment with FLX results in the specific induction of Htr4 expression in S100a10 CStr pyramidal cells, and that loss of p11 causes both a decrease in the expression of Htrs in this cell type and a failure to respond behaviorally to antidepressants. The induction of Htr4 expression is consistent with previous studies showing that Htr4 agonists have antidepressant effects (Lucas et al., 2007) which are dependent on p11 expression (Warner-Schmidt et al., 2009). This suggests the specific hypothesis that the increase of Htr4 expression following chronic FLX treatment enhances the sensitivity of CStr neurons to the elevated serotonin levels, thereby increasing outflow from the cerebral cortex to the striatum and contributing to the efficacy of antidepressants. The fact that the loss of p11 from these cells resulted in a downregulation of many Htrs suggests a wider impact of p11 on serotonergic tone. Although additional experimentation is needed to test this hypothesis, our data strongly argue that the information obtained from TRAP profiling of individual cell types in mouse models of disease can significantly advance our understanding of molecular and cellular mechanisms of disease.
A confluence of evidence has emerged from studies as diverse as human neuroimaging (Drevets et al., 2008; Mayberg et al., 1999) and immediate early gene transcription (Covington et al., 2010) to implicate altered activity in several brain areas in depression. These studies form the basis of our current understanding of depression, in which mood is governed by a distributed brain circuit that does not function normally in depressed individuals (Ressler and Mayberg, 2007). Given this understanding, how is it that the molecular adaptations that occur in a single circuit element (S100a10 CStr projection neurons) can be so important for the antidepressant activity of SSRIs? Our data identify several factors that may help to answer this query. First, S100a10 pyramidal cells occupy a critical position in the mood circuitry. We have shown by retrograde tracing that this cell population provides a major conduit for communication between the cortex and the dorsal striatum. Although early studies emphasized a role for the ventral striatum in reward processing and depression, recent structural and functional imaging studies have revealed that the dorsal striatum is both active during reward related tasks and altered in MDD (Delgado, 2007; Pizzagalli et al., 2009). In particular, reduced prediction error signals in both the bilateral caudate and the NAc correlate with increased anhedonia (a core feature of MDD) in depressed subjects (Gradin et al., 2011). In addition, MDD subjects with elevated anhedonic symptoms or late-onset depression showed a reduced caudate volume (Krishnan et al., 1992; Pizzagalli et al., 2009). It is possible that S100a10 CStr cells projecting to dorsomedial striatum (DMS) underlie antidepressant responses. This idea is supported by a growing body of evidence demonstrating that the DMS plays a significant role in reward learning and motivated behavior (Voorn et al., 2004).
The adaptations that occur in this cell type in response to FLX, including the increase in Htr4 expression, occur specifically in S100a10 CStr cells. Thus, we were not able to detect these changes in TRAP data from Glt25d2 corticopontine neurons or in data collected from whole cortex samples. These results highlight the unique biochemical properties of distinct classes of pyramidal cells, and confirm our previous demonstration that the genome wide pharmacokinetic responses of even closely related cell types can vary dramatically (Heiman et al., 2008).
Finally, it is important to emphasize that we chose to analyze p11 expressing cortical neurons because previous studies had shown that this protein is expressed in a laminar pattern in the cortex, and that its expression in cortex is reduced in depressed individuals and restored in response to treatment (Anisman et al., 2008; Svenningsson et al., 2006). We report here that p11 is required for the molecular adaptations seen in CStr cells in response to an SSRI, and that loss of p11 expression in the cortex abolishes behavioral responses to the SSRI FLX. Although the large number of distinct serotonin receptors and their broad distribution across the nervous system (and periphery) suggest that serotonin elicits a variety of responses in vivo, it seems possible that effective long-term, functional adaptation to SSRIs may require a constellation of factors present only in S100a10 cortical neurons.
Our data demonstrate that the beneficial actions of SSRI antidepressant therapy may be mediated by a single cell type in the cerebral cortex, and suggest that development of drugs that specifically target the activity of CStr neurons may result in improved therapies for depression. However, it is important to note that our results do not predict that the pathophysiological changes that elicit depression occur in this cell type, or that modulation of other critical cell types might not provide equally effective therapies. The observation that deletion of p11 in the cerebral cortex results in a lack of responsiveness to FLX but does not elicit depressive-like or anxiety behaviors strongly suggests that loss of p11 in other cell types must be involved in generation of the behavioral phenotypes evident in constitutive p11 KO mice. While the separation of the anatomical locations responsible for generation of the behavioral phenotypes from those governing treatment responses in this animal model of depression may need clarification, we believe that such a dichotomy is typical of complex circuit-based disorders. For example, deep brain stimulation studies of Parkinson’s disease have shown that even circuits that are irreversibly altered as a result of neurodegeneration can be phenotypically normalized by stimulation of circuit elements that are not lost as a result of the pathophysiology of the disease (Limousin and Martinez-Torres, 2008). Given the genetic complexity of MDD and the circuitry that is involved in the regulation of mood, we expect that the causes of MDD will alter function at many points in the mood circuitry. However, the identification of a single cell type in the cerebral cortex that is both positioned to normalize activity between cortical and subcortical sites, and that is required for the response to SSRI antidepressants, is an exciting step forward in the march toward development of more effective treatments for depression.
All procedures involving animals were approved by the Rockefeller University Institutional Animal Care and Use Committee and were in accordance with the National Institutes of Health guidelines. The Glt25d2 bacTRAP mice, constitutive p11 KO mice, and Snap25 bacTRAP mice were generated and maintained at Rockefeller University and have been described previously (Doyle et al., 2008; Svenningsson et al., 2006). The S100a10 bacTRAP transgenic mouse line was generated by modifying a BAC clone containing the S100a10 gene to insert an EGFP-L10a fusion protein into the translation start site, and the modified BAC was purified and used for transgenesis as described previously (Gong et al., 2002). See Extended Experimental Procedures for details. Cortex-specific p11 KO mice were generated by breeding floxed p11 mice (Warner-Schmidt et al., 2011) with Emx1-Cre mice purchased from The Jackson Laboratory (Stock #005628). All mice were maintained on a 12 h light/dark cycle and given ad libitum access to food and water. For chronic drug treatments, mice were housed 1–2 per cage and a FLX (0.167 mg/ml)/saccharine (1%) mixture or saccharine alone (VEH) was administered in the drinking water for 15–18 days and replaced with fresh solution every 3–4 days.
Alexa-555-conjugated cholera toxin β-subunit (Invitrogen) was stereotaxically injected into adult (8–12 week old) transgenic mice under ketamine/xylazine (100/10 mg/kg) anesthesia as described in Supplemental Information. Quantification of double-labeled cells and cell depth was done using Image J (NIH) as detailed in Extended Experimental Procedures.
Tissue was prepared as described in Extended Experimental Procedures. Immunofluorescent staining was carried out on 35 μm thick sections using chicken anti-GFP (1:2000, Abcam), mouse anti-EGFP (1:500; Heiman et al., 2008), rabbit anti-CTβ (1:1000, Abcam), goat anti-S100a10 (1:200, R&D Systems), and/or mouse anti-NeuN (1:1000, Millipore) followed by Alexa-fluor conjugated secondary antibodies (Invitrogen). All sections were imaged on a Zeiss LSM700 confocal microscope.
S100a10 (p11) in situ hybridization was perfomed as described previously (Warner-Schmidt et al., 2010) and in Extended Experimental Procedures.
Three male mice were pooled for each sample and three biological replicates were collected for each condition. All polysome purifications and mRNA extractions were carried out as described previously (Heiman et al., 2008) and in Extended Experimental Procedures. RNA quantity was measured with a Nanodrop 1000 spectrophotometer and quality was assayed on an Agilent 2100 Bioanalyzer.
For microarrays, RNA was amplified and labeled using the GeneChip Expression 3′ Amplification 2-Cycle cDNA Synthesis Kit (Affymetrix) and hybridized to GeneChip Mouse Genome 430 2.0 microarrays (Affymetrix) according to manufacturers’ protocols. CEL files were imported into GeneSpring GX 7.3.1 and pre-processed with the GCRMA algorithm. For each experiment, data were filtered to remove probe sets with low intensities (normalized expression <50 in at least one condition) and those identified as background (Doyle et al., 2008). Only probe sets expressed in neuronal cells were used for analysis (see Supplemental Information). Differentially expressed genes were identified by performing a moderated t-test (Smyth, 2004) as described in Extended Experimental Procedures. P-values were adjusted for multiple hypothesis testing, controlling the false discovery rate with the Benjamini-Hochberg procedure. Only genes with an FDR <0.05 (5%) and a fold change >1.4 were used for analyses. MIAME-compliant raw data are available from Gene Expression Omnibus.
cDNA was synthesized using the WT-Ovation RNA Amplification kit (NuGEN Technologies) according to manufacturer’s protocol. For Htrs, PCR was carried out in an Applied Biosystems StepOnePlus RT-PCR System using Taqman expression assays. All other reactions were performed in a Bio-Rad iQ5 Multicolor RT-PCR Detection System using the SYBR green method (BioRad). Data were normalized to β-actin and relative expression changes between conditions were calculated by the comparative CT (2−ΔΔCT) method. Student’s t-test was used for all statistical analyses. Detailed methods are provided in Extended Experimental Procedures.
Detailed descriptions of all behavioral methods can be found in Extended Experimental Procedures. Novelty suppressed feeding (NSF) was performed exactly as described (Egeland et al., 2010) and the tail suspension test, forced swim, sucrose preference, and open field were performed as described previously (Warner-Schmidt et al., 2009). Statistical comparisons were made by ANOVA using Prism 5 software (GraphPad). In experiments comprised of more than two groups, data were first analyzed by two-way ANOVA followed by post-hoc Bonferroni test. Statistical significance was set at p<0.05.
This work was supported by the Howard Hughes Medical Institute (NH), NIH/NIDA ARRA Grand Opportunity Grant PHS DA028968 (NH/PG), Conte Center PHS MH090963 (PG/NH), USAMRAA Grant W81XWH-09-0401 (JWS), The JPB Foundation (PG), and the Fisher Foundation (PG). We wish to thank the Rockefeller University Genomics Resource Center, and Neuroscience Associates. We would further like to thank Tanya Stevens, Sujata Bupp, and George Skabardonis for their excellent advice and assistance, and Joseph Dougherty for providing the Snap25 bacTRAP mice.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.