Neuronal Dystroglycan regulates postnatal development of CCK/cannabinoid receptor-1 interneurons

The development of functional neural circuits requires the precise formation of synaptic connections between diverse neuronal populations. The molecular pathways that allow GABAergic interneuron subtypes in the mammalian brain to initially recognize their postsynaptic partners remain largely unknown. The transmembrane glycoprotein Dystroglycan is localized to inhibitory synapses in pyramidal neurons, where it is required for the proper function of CCK+ interneurons. However, the precise temporal requirement for Dystroglycan during inhibitory synapse development has not been examined. In this study, we use NEXCre or Camk2aCreERT2 to conditionally delete Dystroglycan from newly-born or adult pyramidal neurons, respectively. We then analyze forebrain development from postnatal day 3 through adulthood, with a particular focus on CCK+ interneurons. In the absence of postsynaptic Dystroglycan in developing pyramidal neurons, presynaptic CCK+ interneurons fail to elaborate their axons and largely disappear from the cortex, hippocampus, amygdala, and olfactory bulb during the first two postnatal weeks. Other interneuron subtypes are unaffected, indicating that CCK+ interneurons are unique in their requirement for postsynaptic Dystroglycan. Dystroglycan does not appear to be required in adult pyramidal neurons to maintain CCK+ interneurons. Bax deletion did not rescue CCK+ interneurons in Dystroglycan mutants during development, suggesting that they are not eliminated by canonical apoptosis. Rather, we observed increased innervation of the striatum, suggesting that the few remaining CCK+ interneurons re-directed their axons to neighboring areas where Dystroglycan expression remained intact. Together these findings show that Dystroglycan functions as part of a synaptic partner recognition complex that is required early for CCK+ interneuron development in the forebrain.


Background
Proper function of neural circuits requires precise connections between specific populations of excitatory pyramidal and inhibitory neurons. GABAergic interneurons are a highly diverse group of neurons that control brain function by synchronizing and shaping the activity of populations of excitatory pyramidal neurons (PyNs) [1][2][3][4][5]. In mice and humans, the majority of interneurons in the cortex and hippocampus are produced in the medial and caudal ganglionic eminences (MGE and CGE) of the ventral forebrain, and migrate long distances to their final destinations [6][7][8]. The importance of interneurons for brain function is underscored by their involvement in a wide variety of neurodevelopmental and neurological disorders including autism, schizophrenia, seizures, and Alzheimer's disease [9][10][11][12].
The proper integration of inhibitory interneurons into neural circuits during development relies on multiple processes such as proliferation, migration, axon guidance, cell death, synaptic target selection, synapse formation (synaptogenesis) and synaptic maintenance. Although much progress has been made in identifying candidate molecules that regulate inhibitory synaptogenesis, our understanding of how molecularly defined subtypes of inhibitory interneurons initially identify specific postsynaptic target cells is lacking [13,14]. One prominent hypothesis for explaining how diverse interneuron subtypes recognize one another during synapse development is the "molecular code" hypothesis, whereby different cell types use unique pairs or complexes of cell adhesion molecules to select their target cells [15][16][17][18]. Cell adhesion molecules are ideally suited to regulate synaptic target recognition due to their large diversity and presence at pre-and postsynaptic membranes. Several recent studies support the idea that cell adhesion molecules are key players in regulating subcellular targeting and synaptic specificity [19][20][21][22]. Although many families of cell adhesion molecules have been implicated in controlling synapse development, they are often involved in multiple aspects of neural circuit development, making it difficult to determine their precise role in mediating synaptic specificity.
Dystroglycan is a cell adhesion molecule widely expressed throughout the body including the developing and adult brain. Dystroglycan is extensively glycosylated, and mutations in at least 19 genes that participate in synthesizing and elongating specific O-mannose sugar chains on Dystroglycan result in a form of congenital muscular dystrophy called dystroglycanopathy, characterized by muscle weakness and neurological defects of varying severity [23][24][25]. Dystroglycan (Dag1) is expressed by multiple cell types in the developing nervous system, including neuroepithelial cells, astrocytes, oligodendrocytes, and excitatory neurons [26,27]. Loss of Dystroglycan function in the nervous system phenocopies the most severe forms of dystroglycanopathy, and causes multiple structural brain and retinal abnormalities due to its indirect role in regulating neuronal migration and axon guidance [28][29][30][31][32][33]. However, some individuals with milder forms of dystroglycanopathy exhibit cognitive impairments even in the absence of detectable brain malformations, suggesting a possible role for Dystroglycan at later stages of brain development including synaptogenesis [34,35]. In PyNs, Dystroglycan is highly concentrated on the cell body and proximal dendrites where it is a major postsynaptic component of inhibitory synapses (Fig. 1A [27,[36][37][38]). However, because of its importance in early aspects of brain development, the role of Dystroglycan at synapses has remained obscure. Using a mouse genetic approach to selectively delete Dystroglycan from PyNs, a recent study showed that Dystroglycan is required for the formation and maintenance of CCK+ interneuron (CCK+ IN) synapses in adult animals, but its specific role in the early development of these interneurons has not been examined [39].
In this study, we show that postsynaptic Dystroglycan on PyNs is required for the proper development of presynaptic CCK+ INs throughout the forebrain. In mice lacking Dystroglycan in PyNs, CCK+ INs fail to elaborate their axons during the first postnatal week and are largely absent by P10. CCK+ INs were not rescued by genetic deletion of Bax suggesting that CCK+ INs may undergo Bax-independent cell death or fail to differentiate in the absence of Dystroglycan. Some remaining CCK+ INs retarget their axons into the striatum, where Dystroglycan expression is retained, suggesting that Dystroglycan functions to allow CCK+ INs to recognize their synaptic partners. Collectively, these results demonstrate that Dystroglycan is a critical regulator of CCK+ IN development.

CCK+ interneurons are largely absent in mice lacking Dystroglycan from pyramidal neurons
To investigate the role of neuronal Dystroglycan in forebrain development, we used a conditional genetic approach to delete Dystroglycan selectively from pyramidal neurons (PyNs). We crossed Dystroglycan conditional mice (Dag1 Flox/Flox ) with Nex Cre driver mice to delete Dystroglycan in all postmitotic excitatory neurons of the forebrain except Cajal-Retzius cells, beginning at E12.5 [40][41][42][43]. Control (Nex Cre ;Dag1 F/+ ) and conditional knockout mice (Nex Cre ;Dag1 F/− ) are hereafter referred to as Dag1 Control and Dag1 cKO mice, respectively (Fig. 1B). We verified the recombination specificity of the Nex Cre line by crossing it with a reporter mouse that expresses mCherry in the nuclei of Cre-recombined cells (R26 LSL-H2B-mCherry [44]). mCherry+ nuclei were detected in excitatory neurons of the forebrain including the cortex, hippocampus, amygdala, and nucleus of the lateral olfactory tract (nLOT) (Fig. S1A). Importantly, mCherry+ nuclei did not overlap with markers for interneurons (CB 1 R, PV, Calbindin) or astrocytes (GFAP), confirming the specificity of the Nex Cre mouse (Fig. S1B, C). In Dag1 Control mice, Dystroglycan staining was observed as puncta concentrated primarily on the cell bodies and proximal dendrites of PyNs, as well as blood vessels (Fig. 1C). In Dag1 cKO mice, Dystroglycan staining was absent from PyNs but was still present on blood vessels, confirming the specificity of the conditional knockout.
Deletion of Dystroglycan from neuroepithelial cells results in disrupted neuronal migration, axon guidance, and dendrite development in the brain, spinal cord and retina [28][29][30][31][32][33]. In contrast, deletion of Dystroglycan from PyNs with Nex Cre did not affect overall brain architecture, consistent with previous results [32]. Cortical lamination in Dag1 cKO mice was normal based on CUX1 immunostaining of layer 2-4 PyNs and labeling of layer 5-6 and hippocampal PyNs with a Thy1 GFP-H transgenic line (Fig. 1D-F). Therefore, neuronal Dystroglycan is not required for PyN migration in the forebrain.
Forebrain interneurons (INs) are a remarkably diverse population, with multiple molecularly and morphologically distinct IN subtypes forming synapses onto different subcellular domains of PyNs [2,45,46]. Since Dystroglycan is localized to inhibitory synapses on the soma and dendrites of PyNs, we examined whether IN development is affected in Dag1 cKO mice. We performed immunostaining with a panel of molecular markers that label IN subpopulations in the hippocampus of adult mice (Fig. 2). In Dag1 Control mice, parvalbumin (PV) and somatostatin (SOM) positive INs, which label the majority of interneurons that originate from the medial ganglionic eminence (MGE), were abundant throughout the hippocampus. The distribution of PV+ and SOM+ cell bodies and their synaptic targeting patterns appeared Mouse breeding scheme for generating pyramidal neuron-specific Dag1 conditional knockout mice using Nex Cre driver mice. (c) Immunostaining for Dystroglycan in the hippocampal CA1 region of P30 Dag1 Control mice (left panel) shows punctate Dystroglycan protein on the soma and proximal dendrites of pyramidal neurons, whereas Dag1 cKO mice (right panel) lack perisomatic staining. Asterisks denote Dystroglycan staining on blood vessels which is retained in Dag1 cKO mice. (d) Coronal sections from P15 Dag1 Control and Dag1 cKO cortex were immunostained for upper layer marker CUX1 (L2-4). (e) Coronal sections of the cortex from P30 Dag1 Control and Dag1 cKO mice crossed with a Thy1 YFP reporter mouse to sparsely label layer 5-6 pyramidal neurons (green) and stained for Calbindin (magenta) to label layer 2-3 pyramidal neurons. (f) Coronal sections of the hippocampus from P30 Dag1 Control and Dag1 cKO mice crossed with a Thy1 YFP reporter mouse to label excitatory neurons (green) in the CA regions and dentate gyrus (f) Quantification of CB 1 R pixels for each CA layer of the CA1 and CA3 shows a significant reduction in CB 1 R staining in Dag1 cKO mice (*P < 0.05, unpaired two-tailed Student's t-test; n = 4 mice/genotype). Data are presented as mean values ± s.e.m. Data are normalized to Dag1 Control signal in each CA layer. CA layers: SO, stratum oriens; SP, stratum pyramidale; SR, stratum radiatum the same in Dag1 cKO mice, suggesting these populations are unaffected by the loss of Dystroglycan ( Fig.  2A, B).
We next stained the hippocampus for IN subtypes that originate from the caudal ganglionic eminence (CGE). The distribution and synaptic targeting of Calretinin interneurons, which target other INs as well as PyN dendrites, appeared normal ( Fig. 2C [47,48]). In contrast, we found a dramatic reduction in cannabinoid receptor-1 (CB 1 R) staining in the hippocampus, which labels the axon terminals of cholecystokinin (CCK) + INs (Fig. 2D [ [49][50][51]). CB 1 R+ terminals were significantly reduced in all CA subregions (Fig. 2E, F). In both the CA1 and CA3, the magnitude of the reduction varied by layer. CB 1 R+ terminals were most strongly reduced (> 95%) in the stratum pyramidale (SP) where CCK+ INs form basket synapses onto PyN cell bodies, and more moderately reduced in the stratum radiatum (SR) and stratum oriens (SO) where CCK/CB 1 R+ INs synapse onto PyN dendrites (Fig. 2E, F). In contrast, CB 1 R+ terminals were abundant in the dentate gyrus of Dag1 cKO mice (Fig. S2). This is likely because Nex Cre recombination is restricted to the outer third of granular layer neurons (Fig. S1C [41]).
The loss of CB 1 R staining in the hippocampus of Dag1 cKO mice could reflect either downregulation of CB 1 R expression or a loss of CCK+ INs. To distinguish between these possibilities, we examined whether other independent markers of CGE-derived CCK+ INs were similarly reduced. These include NECAB1, a calcium binding protein that specifically labels CCK+ IN cell bodies (Fig. 3A) [52], and VGLUT3, a vesicular glutamate transporter enriched at CCK+ IN synapses (Fig. 3C) [53][54][55]. Both NECAB1+ cell bodies and VGLUT3+ synaptic terminals were reduced in the hippocampus of Dag1 cKO mice (Fig. 3B, D). Based on the loss of all three markers, we conclude that CCK+ INs are largely absent from the hippocampus of Dag1 cKO mice.
In addition to the hippocampus, Dystroglycan is present in PyNs of the cortex, amygdala, and nucleus of the lateral olfactory tract (nLOT) [27], which all receive extensive innervation from CCK+ INs [56][57][58]. Therefore, we assessed whether deletion of Dystroglycan from PyNs affects CCK+ INs and their terminals in these forebrain regions. We first performed immunostaining for CB 1 R on sagittal sections from Dag1 Control and Dag1 cKO mice. CB 1 R terminals were largely absent throughout the entire forebrain of Dag1 cKO mice (Fig. 4A, B). Next, we stained P30 Dag1 Control and Dag1 cKO mice for NECAB1 and CB 1 R to label the cell bodies and terminals of CCK+ INs, respectively ( Fig. 4C-E). In Dag1 Control mice, NECAB1+ cell bodies were numerous and CB 1 R innervation was extensive in the cortex, amygdala, and nLOT. In contrast, NECAB1+ cell bodies were dramatically reduced, and CB 1 R staining was almost completely absent in all regions of Dag1 cKO mice ( Fig.  4C-E). In each region, a few NECAB1+ cell bodies remained in Dag1 cKO mice, and these co-localized with CB 1 R. Therefore, Dystroglycan expressed in PyNs is required broadly in the developing forebrain for the proper integration of CCK+ INs. Ai9 mice (b) immunostained for CB 1 R (green; right panels) and tdTomato/Ai9 (magenta; middle panels). In Dag1 cKO ;Ai9 mice, CB 1 R staining is lacking in all the forebrain regions where Nex Cre drives recombination in excitatory neurons (tdTomato expression, middle panels) including the cortex (CTX), hippocampus (HC), and olfactory bulb (OB). Note the absence of tdTomato signal in the striatum (STR) and midbrain (MB), which are not targeted by Nex Cre . (c-e) Immunostaining for CB 1 R (green) and NECAB1 (magenta) in the cortex (c), amygdala (d), and nucleus of the lateral olfactory tract (e) shows the reduction of CCK+ interneuron markers in the forebrain of P30 Dag1 cKO mice (right panels). Enlarged images (yellow boxed regions) show individual NECAB1+ cell bodies (magenta) co-localized with CB 1 R (green)

Postnatal development of CCK+ interneurons is impaired in the forebrains of Dag1 cKO mice
Our results showing that deletion of Dystroglycan from PyNs leads to a reduction in CCK+ IN innervation is consistent with previous work [39]. However, the temporal onset of this phenotype has not been determined. During embryonic development, CCK+ INs generated in the caudal ganglionic eminence (CGE) begin populating the hippocampus around E14.5 [59,60] (Fig. 5A). At postnatal ages, CCK+ INs settle into their final positions within the hippocampus and initially extend axons throughout the hippocampal layers before refining their projections to form characteristic basket synapses onto PyN somas (Fig. 5B, D) [61][62][63]. We first examined the development of CB 1 R+ terminals in Dag1 Control mice during the first two postnatal weeks (P3-P15), as CB 1 R staining is largely absent from CCK+ INs before birth [64][65][66][67]. At early postnatal ages (P3-P5), the majority of CB 1 R+ terminals were observed in the stratum radiatum (SR) layer of the hippocampus, where immature PyN dendrites are located (Fig. 5B, D). Between P5 and P10, CB 1 R+ terminals increased in the stratum pyramidale (SP) where PyN cell bodies are located. From P15 through adulthood (15 months), CB 1 R+ terminals became progressively concentrated in the SP.
Next, we examined CB 1 R+ terminal development in Dag1 cKO mice. At P3, the earliest age we were able to conclusively identify CCK+ INs, CB 1 R+ staining was already reduced in the hippocampus of Dag1 cKO mice. This reduction persisted throughout the period of postnatal development and into adulthood, as late as 15 months (Fig. 5C, E, F). To further confirm this finding, we stained the hippocampus for VGLUT3, an independent synaptic marker for CCK+ IN terminals that is upregulated during early postnatal ages (Fig. S3A). In Dag1 Control mice, VGLUT3+ terminals increased in the hippocampus during the first two postnatal weeks, and showed a similar pattern of innervation as CB 1 R+ staining (Fig. S3B). In contrast, VGLUT3+ terminals were reduced at all ages examined in Dag1 cKO mice (Fig. S3B). PV staining, which increases between P10 and P15 [68], was unaltered in Dag1 cKO mice compared with controls ( Fig. S3C).
We next determined whether the reduction of CCK+ INs in the cortex, amygdala, and nLOT followed the same developmental time course as the hippocampus. In Dag1 Control mice, CB 1 R+ terminals gradually increased in density in all regions between P3 and P15, and remained stable beyond this age into adulthood (15 months) (Fig. 6). In contrast, CB 1 R+ terminals in Dag1 cKO mice failed to elaborate during the first two postnatal weeks, and remained sparse in adult animals. Collectively, these results demonstrate that Dystroglycan in PyNs is critical during the first two postnatal weeks for the development and integration of CCK+ INs throughout the forebrain.

Post-developmental maintenance of CCK+ interneurons does not require Dystroglycan
Inhibitory synaptogenesis increases between P5-P15, and is largely complete by P30 [20,22,54]. Therefore, we wanted to assess whether deletion of Dystroglycan after inhibitory synapse formation impairs the maintenance of CCK+ INs. To achieve temporal control of Dystroglycan deletion from PyNs, we generated mice expressing tamoxifen-inducible Cre recombinase under the control of an excitatory neuron-specific promoter Camk2a, (Calcium/calmodulin-dependent protein kinase II alpha [69]). Control (Camk2a CreERT2 ;DG F/+ ;Ai9) or inducible-cKO (Camk2a CreERT2 ;DG F/− ;Ai9) mice were administered tamoxifen at P23 via oral gavage, which induced recombination in the majority of PyNs in the hippocampus (Fig. 7A, B). We then analyzed CB 1 R+ innervation 6 weeks later at P65. No differences were found between the Dag1 inducible-cKO and controls, suggesting that Dystroglycan is not required for the post-developmental maintenance of CCK+ INs (Fig. 7C, D).
The number of PyNs and INs in the forebrain is tightly regulated during early postnatal development, with excess or inappropriately connected neurons pruned by Bax-dependent apoptosis [70][71][72][73]. PyN apoptosis is largely complete by P5, followed by IN apoptosis which peaks at P7-P9. We hypothesized that in the absence of PyN Dystroglycan, CCK+ INs are unable to recognize their postsynaptic targets and are therefore eliminated by apoptosis. We tested this hypothesis by generating Dag1 Ctrl and Dag1 cKO mice that lack either one (Dag1 Ctrl ;Bax Ctrl and Dag1 cKO ;Bax Ctrl ) or both copies of Bax (Dag1 Ctrl ;Bax KO and Dag1 cKO ;Bax KO ) to block apoptosis (Fig. 8A). Deletion of Bax from control mice (Dag1 Ctrl ;Bax KO ) did not alter CB 1 R+ innervation in the CA1 subregion of the hippocampus (Fig. 8B-C, F). In line with our previous results, Dag1 cKO ;Bax Ctrl mice lacking one copy of Bax had a similar reduction in CB 1 R+ terminals as Dag1 cKO mice (Fig. 8D, F). Surprisingly, we found that complete deletion of Bax in Dag1 cKO mice (Dag1 cKO ;Bax KO ) was not sufficient to rescue CB 1 R+ innervation (Fig. 8E, F). Staining for an additional CCK+ IN synapse marker (VGLUT3) further confirmed this result (Fig. S4). Finally, we examined whether deletion of Bax could rescue CB 1 R+ innervation in the cortex, amygdala, and the nucleus of the lateral olfactory tract (nLOT) of Dag1 cKO mice (Fig. S5). In all regions examined, CB 1 R+ terminals were reduced in mice lacking Dystroglycan (Dag1 cKO ;Bax Ctrl ). Similar to our   as they are not targeted by Nex Cre (Fig. 4A-B [39,41]). We therefore examined CB 1 R innervation of the striatum in Dag1 Control and Dag1 cKO mice. In Dag1 Control mice, CB 1 R innervation in the striatum was present, but sparse compared with neighboring regions of the cortex (Fig. 9A) [74,75]. In contrast, CB 1 R innervation in the striatum of Dag1 cKO mice was noticeably increased (Fig.  9C, I). The lateral regions of the striatum closest to the cortex exhibited dense CB 1 R innervation, which decreased towards the medial striatum. Global deletion of Bax from Dag1 Control or Dag1 cKO mice did not alter the pattern of CB 1 R innervation in the striatum (Fig. 9B, D).
Examination of the developmental timecourse of CB 1 R+ innervation in the striatum revealed sparse CB 1 R+ terminals at P10 in both Dag1 Control and Dag1 cKO mice (Fig. 9E), which increased in Dag1 cKO mice compared with controls between P15 and P30 ( Fig. 9F-G). This coincides with the period of CB 1 R+ innervation of forebrain targets in Dag1 Control mice. Occasionally, CB 1 R+ cell bodies could be seen in the cortex near the striatal boundary, with their axon terminals projecting into the striatum (Fig. 9H). These results suggest that some CCK+ INs in the cortex of Dag1 cKO mice may redirect their axons into the

Discussion
Dystroglycan plays a critical role in maintaining the integrity of the neuroepithelial scaffold during early stages of brain development, which has made it difficult to assess its function within neurons at subsequent stages. In the current study, we show that Dystroglycan in pyramidal neurons regulates the development of a subset of their pre-synaptic partners. When Dystroglycan is selectively deleted from PyNs, CCK+ INs throughout the entire forebrain fail to properly integrate, and largely disappear during the first postnatal week. Surprisingly, we found that deletion of Bax did not rescue CCK+ INs in Dag1 cKO mice, suggesting  their disappearance is not due to apoptotic cell death. The few remaining CCK+ INs redirect their axons into neighboring regions of the brain in which Dystroglycan is still present, suggesting that Dystroglycan functions as a part of a synaptic partner recognition complex.

What stage of CCK+ interneuron development requires Dystroglycan?
The localization of Dystroglycan to inhibitory synapses in forebrain pyramidal neurons has been described by multiple studies, while its function at these synapses has remained obscure [36][37][38]76]. Recently, it was found that Dystroglycan is required for the formation and function of CCK+ inhibitory basket synapses, but not PV+ basket synapses onto the same PyNs [39]. This finding is significant, because very little is known about the molecules and mechanisms involved in orchestrating the formation of specific subtypes of inhibitory synapses [17]. However, since the earliest timepoint examined in this previous study was P21, it was unclear what stage of synapse development requires Dystroglycan. During neural circuit development, neurons must first migrate and direct their axons to their appropriate targets, then recognize the appropriate synaptic partners from a myriad of potential choices, then finally form functional synapses [13]. Our data suggest that Dystroglycan is required for the least understood of these processes: synaptic partner recognition. This is supported by the observation that CCK+ INs are present at the earliest stages they can be conclusively identified in the forebrain of Dag1 cKO mice (P3), but then fail to elaborate their axons and integrate into these circuits during the first postnatal week (Figs. 5, S3, Fig. 6). Interestingly, the few remaining CCK+ INs appear to project their axons into regions that continue to express Dystroglycan in Dag1 cKO mice (Figs. 5,9). Taken together, this data suggests that CCK+ INs in Dag1 cKO mice fail to recognize their normal postsynaptic PyN targets in the hippocampus and cortex during early postnatal development, and instead re-route to neighboring Dag1+ neurons, discussed further below.
The process of synaptic partner recognition in mammals has been difficult to study due to our inability to precisely identify and genetically manipulate the specific neuronal populations during development. Determining whether loss of Dystroglycan impairs CCK+ IN development before birth is technically challenging due to the lack of immunohistochemical and genetic tools for detecting CCK+ INs prenatally [60]. The cannabinoid receptor-1 (Cnr1) and cholecystokinin (Cck) genes are both expressed in PyNs at prenatal timepoints, limiting their usefulness for detecting CCK+ INs. Transcription factors such as Prox1 are also of limited usefulness due to its broad expression in multiple CGE-derived IN subtypes [77]. VGLUT3, which labels a subset of CCK+ INs, does not increase in expression until the first postnatal week [54]. Other IN subtypes exhibit delayed expression of selective molecular markers as well. For instance, MGE-derived Parvalbumin INs do not begin to express Parvalbumin until P10, well after the period of initial synaptic partner recognition [68,78].

What happens to CCK+ interneurons in the absence of Dystroglycan?
Our results show that deletion of Dystroglycan from PyNs resulted in a loss of all of the markers we used to identify CCK+ INs in the forebrain (Figs. 2, 3, 4). What happens to these neurons in the absence of PyN Dystroglycan? One possibility that we examined is that CCK+ INs undergo apoptosis. During the first 2 weeks of development (P5-P10), a significant number of excitatory and inhibitory neurons are pruned by Bax-dependent apoptotic cell death [70][71][72][73]. This ensures the proper number of neurons and removes neurons that fail to integrate into the developing circuit. Whereas Baxdependent developmental cell death has been described for most MGE and CGE-derived interneuron subtypes, whether CCK+ INs normally undergo the same process has not been directly examined [72,73]. We tested whether the loss of CCK+ INs in Dag1 cKO mice could reflect premature or amplified developmental apoptosis, which peaks around P9 for other IN subtypes. However, constitutive deletion of Bax, which is sufficient to block developmental apoptosis in other neuronal populations, did not rescue CCK+ INs (Figs. 8, S4, S5). This suggests that canonical apoptosis is not responsible for the loss of CCK+ INs in Dag1 cKO mice. It is possible that CCK+ INs are eliminated in a Bax-independent manner, similar to some populations of Cajal-Retzius cells in the cortex and astrocytes in the developing retina [79,80]. CCK+ INs comprise a molecularly and morphologically diverse group of cells that include both cell body targeting (perisomatic) and multiple dendrite targeting subtypes [54,81,82]. In the hippocampus, CCK+ INs frequently express one of two non-overlapping markers, VGLUT3 (~45%) and VIP (~16%) [53]. In Dag1 cKO mice, all synaptic and cell body markers selective for CCK+ INs (CB 1 R, VGLUT3, NECAB1) that we examined were reduced at the onset of their expression. While it is formally possible that Dystroglycan in PyNs is required for CCK+ INs to fully differentiate into their mature, molecularly defined subtype, we consider this unlikely. In this situation, Dystroglycan present on PyNs would be required to transmit a retrograde signal to CCK+ INs to direct their differentiation. We are unaware of any cell adhesion molecules that function in this manner. Rather, fate switching or failure to fully differentiate is usually observed upon cell-autonomous loss of specific transcription factors [83].
Our data also indicate that Dystroglycan is not required to maintain CCK+ INs after the period of synapse formation (Fig. 7). This is in contrast to a previous study that showed a gradual reduction in the number of Vglut3+ puncta when Dag1 was deleted in adult mice using AAV-Cre [39]. Aside from the different approaches used for adult deletion, this difference may arise from the level of analysis: in our study, we saw no difference in the cellular organization of CCK+ INs following adult deletion, whereas the previous study was focused specifically on presynaptic puncta. It is possible that in our inducible-cKO (Camk2a CreERT2 ;DG F/− ;Ai9) mice, synaptic inputs from CCK+ INs are reduced without altering the survival of these neurons. Alternatively, there may still be some residual Dystroglycan protein remaining in Camk2a CreERT2 ;DG F/− ;Ai9 mice, at levels sufficient to support CCK+ IN maintenance.
Although CCK+ INs and their terminals were dramatically reduced throughout the brains of Dag1 cKO mice, some CCK+ IN terminals were still present, particularly along the cortico-striatal boundary and in the upper dendritic layers of the cortex (layer 1) and hippocampus. Importantly, striatal neurons and Cajal-Retzius cells, which are located in superficial cortical layers during postnatal development, are not targeted by Nex Cre [84]. This suggests that in the absence of Dystroglycan on their normal postsynaptic targets (PyNs), CCK+ INs may direct their axons to secondary synaptic targets that retain Dystroglycan expression.
A number of studies have indicated that synaptic partner recognition and targeting may be "stringent" or "flexible", depending on the cell type involved. Studies in the Drosophila visual system have shown that synaptic cell adhesion molecules such as DIP/Dprs can promote either stringent or flexible outcomes among synaptic partners depending on the cellular context and the molecules involved. For instance, postsynaptic Dm8 neurons containing the receptor DIP-γ undergo cell death if not innervated by a matching R7 photoreceptor containing the cognate ligand Dpr11 [85]. In contrast, loss of DIP-β from L4 neurons does not impair synapse formation or cause cell death, but instead leads to ectopic synapses onto alternative synaptic partners [86]. Synaptic partner recognition "flexibility" and "stringency" has also been demonstrated in the mammalian nervous system. In the developing retina, On-alpha retinal ganglion cells will re-wire to increase inputs from neighboring bipolar cell types when their normal presynaptic inputs (Type 6 bipolar cells) are genetically ablated [87]. In contrast preG-ABA INs in the developing spinal cord retract their processes when their primary targets (proprioceptor axons) are not present, rather than forming synapses onto secondary targets [88]. Despite retracting their axons, preGABA INs do not undergo cell death, suggesting that loss of neurons is not a necessary consequence of losing synaptic partners. In Dag1 cKO mice, CCK+ INs may stringently require Dystroglycan for their ability to recognize their primary synaptic targets and die in a Bax-independent manner in its absence. The observation that some CCK+ INs near the cortico-striatal boundary survive and innervate the striatum suggests that they may exhibit some degree of flexibility to make contacts onto secondary targets. Determining whether the remaining CCK+ INs in Dag1 cKO mice exhibit normal morphological and physiological properties will require fate mapping these neurons, which is difficult with currently available genetic tools.
Why are CCK+ interneurons selectively affected in Dag1 cKO mice? CCK+ INs appear to be the only interneuron subtype affected by deletion of Dystroglycan from PyNs. Compared to other IN populations, CCK+ INs express high levels of CB 1 Rs, which can play important roles in neuronal proliferation, migration, and axon outgrowth [89]. In utero exposure to exogenous cannabinoids results in a specific loss of CCK+ INs through unknown mechanisms [90]. However, conditional deletion of the cannabinoid receptor-1 gene Cnr1 from CCK+ INs does not affect interneuron migration or neurochemical specification, but rather increases the number of perisomatic VGLUT3+ inhibitory synapses on cortical PyNs [67]. In addition, CB 1 R signaling is not necessary for the survival of CCK+ INs [91]. Therefore, it is unlikely that alterations in CB 1 R activity underlie the selective loss of CCK+ INs in Dag1 cKO mice.
Biochemical experiments have identified α-DG as a major interaction partner of αand β-neurexins in whole brain lysates, and these interactions are dependent upon the lack of splice inserts in LNS2 and LNS6 of neurexin [98][99][100][101]. Conditional deletion of all three Neurexins from interneurons revealed distinct outcomes depending on the IN population examined [102]. Deletion of all Neurexin isoforms from PV+ INs results in a significant decrease in the number of PV+ synapses in the cortex, whereas it does not affect inhibitory synapse numbers when deleted from SST+ INs. While PV+ IN numbers were not affected by conditional deletion of Neurexins, this could reflect the timing of deletion, which is unlikely to occur before three weeks of age based on the onset of Cre expression [68,78]. Nrxn1α and Neurexin 3α/β are expressed at significantly higher levels in CCK+ INs than in PV+ INs, and CCK+ INs predominantly express Neurexin isoforms lacking splice inserts in LNS6 [100]. Therefore, CCK+ INs may show a larger degree of Nrxn: Dystroglycan interaction than other IN subtypes. Mice harboring a mutation in Dystroglycan that exhibits reduced glycosylation, and thus Neurexin binding capacity (Dag1 T190M ), showed no impairments in CCK+ IN terminal development [39,103]. However, these mice do not display the cortical migration phenotypes associated with a complete loss of Dystroglycan, suggesting that Dystroglycan retains some residual function, which may be sufficient for CCK+ IN terminal development. Whether Neurexins are required cell autonomously in CCK+ INs for their development has not been directly tested, in part due to a lack of genetic tools.

Limitations in studying CCK+ interneuron development
Our understanding of CCK+ IN development and function has lagged behind other interneuron subtypes (PV, SOM, VIP, etc) due in part to the lack of viral and mouse genetic tools available for selectively labeling and manipulating CCK+ INs. All major markers of CCK+ INs (CCK, CB1R, and VGLUT3) are also expressed at lower levels in PyNs, limiting the usefulness of single promoter/recombinase approaches for targeting CCK+ INs [104][105][106]. Specific targeting of CCK+ INs therefore requires dual recombinase-based intersectional approaches, including CCK-Cre;Dlx5/6-Flp double transgenic mice [107][108][109][110][111], dual CCK-dsRed;GAD67-GFP reporter mice [60], or VGLUT3 Cre mice which label approximately half of CCK+ INs [54,112]. Other reporter lines (5HT3AR EGFP ) target the entire CGE-derived interneuron population, of which CCK+ INs only comprisẽ 10% [113,114]. A recently developed Sncg FlpO mouse line appears to provide selective genetic access CCK+ basket cells by taking advantage of the fact that Sncg is specifically expressed in CCK+ INs [115]. However, it is not clear when the onset of recombination occurs in this line, and whether it will be useful for studying the early development of CCK+ INs. Indeed, many of the genes used for targeting IN subtypes are not significantly expressed until after the first postnatal week in mice, when much of the process of synaptic partner recognition and initial synapse formation has already occurred ( Fig. S3 [54,68,78]).

Conclusion
In this study, we identified a critical role for excitatory neuron Dystroglycan in regulating the development of forebrain CCK+ interneurons during the first postnatal week. Given the emerging role for CCK+ INs and cannabinoid signaling in controlling neural circuit activity, Dag1 cKO mice may be useful for studying the consequences of losing a major IN population.

Animal husbandry
All animals were housed and cared for by the Department of Comparative Medicine (DCM) at Oregon Health and Science University (OHSU), an AAALACaccredited institution. Animal procedures were approved by OHSU Institutional Animal Care and Use Committee (Protocol # IS00000539) and adhered to the NIH Guide for the care and use of laboratory animals. Animals older than postnatal day 6 (P6) were euthanized by administration of CO 2, animals <P6 were euthanized by rapid decapitation. Animal facilities are regulated for temperature and humidity and maintained on a 12 h light-dark cycle and animals were provided food and water ad libitum.

Tamoxifen administration
Tamoxifen (Sigma; Cat# T5648-1G) was dissolved 1:10 in sunflower seed oil. Each mouse was orally gavaged with 200 μL of tamoxifen at a final concentration of 5 mg/ml tamoxifen.

Perfusions and tissue preparation
Brains from mice younger than P15 were dissected and fixed in 4% paraformaldehyde (PFA) in phosphate buffered saline (PBS) overnight for 18-24 h at 4°C. Mice P15 and older were deeply anesthetized using CO2 and transcardially perfused with ice cold 0.1 M PBS for two minutes to clear blood from the brain, followed by 15 mL of ice cold 4% PFA in PBS.

Immunohistochemistry and antibodies
Single and multiple immunofluorescence detection of antigens was performed as follows: Free-floating vibratome sections (50 μm) were briefly rinsed with PBS, then blocked for 1 h in PBS containing 0.2% Triton-X (PBST) plus 10% normal donkey serum. Sections were incubated with primary antibodies (

Experimental design and statistical analysis
All phenotypic analyses were conducted using tissue collected from at least three mice per genotype from at least two independent litters unless otherwise noted. The number of mice used for each analysis ("n") are indicated in the figures and figure legends. No specific power analyses were performed, but sample sizes were similar to our previous work and other published literature [28,29,33]. Male and female mice were analyzed together. In many cases, highly penetrant phenotypes revealed the genotypes of the mice and no blinding could be performed. Significance between groups was determined using unpaired two-tailed Student's t-test. Data are presented as mean ± standard error of the mean (s.e.m) and statistical significance was set at alpha = 0.05 (P < 0.05). Graphical representations of data and statistical analyses were performed in GraphPad Prism 8 (San Diego, CA).
Additional file 2: Fig. S2. CCK+ interneuron innervation of the dentate gyrus is minimally altered in Dag1 cKO mice. (A) Immunostaining of CB 1 R in the dentate gyrus from P30 Dag1 Control (left panels) and Dag1 cKO mice (right panels). Single channel images of CB 1 R (gray) are shown below. (B) Quantification of CB 1 R pixels for each dentate gyrus layer (*P < 0.05, unpaired two-tailed Student's t-test; n = 4 mice/genotype). Data are presented as mean values ± s.e.m. Data are normalized to Dag1 Control signal in each dentate gyrus layer. OML, outer molecular layer; IML, inner molecular layer; GCL, granule cell layer.