Anatomical and molecular characterization of dopamine D1 receptor-expressing neurons of the mouse CA1 dorsal hippocampus

Male and female, 8–12-week old, Drd1a-EGFP (n = 11 C57BL/6N background, founder S118), Drd2-Cre (C57BL/6J background, founder ER44) heterozygous mice and RiboTag:loxP [The Jackson Laboratory, (Sanz et al., 2009)] were used in this study. BAC Drd1a-EGFP and Drd2-Cre mice were generated by GENSAT (Gene Expression Nervous System Atlas) at the Rockefeller University (New York, NY, USA) (Gong et al. 2003). Homozygous RiboTag female mice were crossed with heterozygous Drd2-Cre male mice to generate Drd2-Cre::RiboTag mice (Puighermanal et al., 2015). Animals were maintained in a 12 hour light/dark cycle, in stable conditions of temperature and humidity, with food and water ad libitum. All experiments were in accordance with the guidelines of the French Agriculture and Forestry Ministry for handling animals (authorization number/license D34-172-13).

Parvalbumin (PV). PV-positive cells are widely distributed in the CA1 subfield (Fig. 1a). Depending on their location in the different layers they allow the classification of various GABAergic inhibitory interneurons (Klausberger 2009; Pawelzik et al. 2002). Thus, PV-expressing cells identified axo-axonic, basket, and bistratified interneurons in both strata pyramidale and oriens. In this latter layer, it also marked the horizontal axo-axonic and oriens-lacunosum-moleculare (O-LM). Our analysis revealed that GFP/PV-positive cells represented ~37 and ~31 % of the total GFP labeled in strata oriens and pyramidale, respectively (Fig. 1a, b; Table 2). Low or no co-localization was found in strata radiatum (~2 %) and lacunosum-moleculare (0 %) where PV-positive cells identified perforant path-associated QuadD, quadrilaminar, and R-receiving apical targeting interneurons (Fig. 1a, b; Table 2).

Calbindin-D28k (CB). In CA1 subfield, CB-immunoreactivity is found in both principal glutamatergic cells in strata pyramidale and radiatum as well as in GABAergic interneurons located in strata oriens, radiatum, and lacunosum-moleculare (Jinno and Kosaka 2002) (Fig. 1c). In stratum pyramidale, we found that among the 272 GFP-immunoreactive cells quantified, 55 co-localized with CB (~20 % of total of GFP-positive neurons) (Fig. 1c, d; Table 2). In stratum oriens, where CB-positive cells identified recurrent O-LM, oriens alveus, and SO–SO cells, ~22 % of GFP-labeled neurons co-expressed CB (Fig. 1c, d). Finally, CB immunolabeling also marked LMR-projecting, radiatum, and Schaffer collateral associated classes of interneurons in strata radiatum and lacunosum-moleculare, in which ~20 and ~13 % of CB/GFP-positive neurons were detected (Fig. 1c, d; Table 2).

Calretinin (CR). CA1 CR-positive cells are distributed in all the layers where they allow the identification of several classes of interneurons (Wheeler et al. 2015) (Fig. 1e). Overall, our analysis revealed a low degree of co-localization between GFP and CR immunoreactivity whatever the layers analyzed. The highest percentage of co-localization was found in stratum pyramidale (~11 %) where CR-positive cells marked interneuron specific LMO-O, interneuron specific O-targeting QuadD, interneuron specific R-O, and interneuron RO-O, a class of interneurons specialized in the control of other interneurons (Fig. 1e, f; Table 2). In addition to interneurons specific, CR-positive cells were expressed in oriens-bistratified in stratum oriens and perforant path-associated QuadD, quadrilaminar, and Schaffer collateral receiving R-targeting cells in stratum radiatum. In both layers, GFP-positive cells expressing CR was rather low, representing only ~4 % in stratum oriens and ~5 % in stratum radiatum (Fig. 1e, f; Table 2). Finally, in stratum lacunosum-moleculare where CR cells stain Cajal–Retzius cells and quadrilaminar interneurons, only three GFP/CR-positive cells were detected among the 173 GFP-immunoreactive cells (Fig. 1e, f; Table 2). These co-labeled cells which most likely correspond to quadrilaminar interneurons represented only ~2 % (Fig. 1e, f; Table 2).

Neuropeptide Y (NPY). NPY/GFP-positive neurons were found in all the layers of the CA1 subfield (Tricoire et al. 2011). Co-localized GFP and NPY immunoreactive cells represented ~40 % and of ~38 % of GFP-positive cells in strata oriens and pyramidale, respectively. Within these two layers, NPY marked back-projection, O-LM, recurrent O-LM, SO–SO interneurons as well as bistratified and ivy cells (Fig. 2a, b; Table 2). Co-localization was also high in strata radiatum (~31 %) and lacunosum-moleculare (~28 %) in which ivy, LMR, perforant path-associated QuaD, radiatum, and radial trilaminar interneurons as well as neurogliaform interneurons are distributed (Fig. 2a, b; Table 2).

Somatostatin (SOM). The highest percentage of GFP/SOM-positive neurons was detected in stratum oriens (~34 %) where SOM is expressed by several classes of interneurons including O-LM, recurrent O-LM, O-LMR, oriens-bistratified, oriens-bistratified projecting as well as trilaminar (Chittajallu et al. 2013; Tricoire et al. 2011) (Fig. 2c, d; Table 2). In stratum pyramidale, only ~6 % of GFP-positive cells co-expressed SOM, a marker of bistratified interneurons. It should be noted that a significant fraction (~57 %) of these neurons appeared to express D1R. Finally, in strata radiatum where SOM-containing cells identify LMR, perforant path-associated QuaD, quadrilaminar and radiatum interneurons, GFP/SOM co-expressing cells represented only ~2 % of GFP-positive cells, but 53 % of SOM-positive neurons (Fig. 2c, d; Table 2). No co-labeling was found in stratum lacunosum-moleculare (Fig. 2d; Table 2).

Neuronal nitric oxide synthase (nNOS). nNOS-expressing neurons represent one of the largest subclasses of interneurons present in the CA1 subfield of the hippocampus. Highly concentrated in strata oriens and lacunosum-moleculare, they allow the identification of neurogliaform and ivy interneurons (Armstrong et al. 2012; Price et al. 2005; Tricoire et al. 2010). As shown in Fig. 3, percentages of nNOS/GFP-immunoreactive cells were high in all the CA1 layers reaching ~62 and ~57 % in strata pyramidale and radiatum, and being slightly lower in strata oriens and lacunosum-moleculare (~51 and ~47 %, respectively) (Fig. 3a, b; Table 2).

Reelin (RLN). In CA1 subfield, RLN allows the identification of both glutamatergic and GABAergic interneurons (Wheeler et al. 2015). Our analysis revealed that GFP was never found in small RLN-positive cells located at the border of strata radiatum/lacunosum-moleculare, which correspond to glutamatergic Cajal–Retzius cells. In contrast, a high level of co-localization was found in strata oriens (~57 %), radiatum (~64 %), and lacunosum-moleculare (~45 %) where RLN identified O-LM and neurogliaform interneurons (Fig. 3c, d; Table 2). In stratum pyramidale, RLN-immunoreactive cells represented only ~12 % of GFP-expressing neurons (Fig. 3d; Table 2).

Metabotropic glutamate receptor type 1α (mGluR1α). The largest density of mGluR1α-positive cells was found in stratum oriens (Tricoire et al. 2011). Within this layer, mGluR1α marked preferentially trilaminar, recurrent O-LM, and O-LM interneurons and co-expressed within GFP in ~42 % of the case (Fig. 4a, b; Table 2). mGluR1α/GFP-expressing cells were also found to a lesser extent in stratum radiatum (~14 %), where they identify hippocampo-subicular projecting ENK⁺ interneurons (Fig. 4a, b; Table 2). Low (~7 %) and no co-localization were detected in strata pyramidale and lacunosum-moleculare, respectively (Fig. 4a, b; Table 2).

Cannabinoid type 1 receptor (CB1R). CA1 CB1R-expressing interneurons are preferentially found in strata radiatum and lacunosum-moleculare identifying LMR projecting, Schaffer collateral-associated, and trilaminar interneurons. They also correspond to CCK-positive basket cells distributed in strata oriens, pyramidale, and radiatum (Freund and Buzsaki 1996) (Fig. 4c). Because CB1R are mainly presynaptically expressed, hippocampal CB1R immunoreactivity did not allow us to quantify the percentage of CB1R-positive cells among the D1R-expressing population. However, a few scattered CB1R/GFP-positive cells were clearly identified in stratum radiatum and at the border of strata radiatum/lacunosum-moleculare (Fig. 4d) as well as in stratum pyramidale (Fig. 4e).

Vesicular glutamate transporter type 3 (VGLUT3). Only four different types of interneurons located in strata oriens, pyramidale, and radiatum express VGLUT3 (Wheeler et al. 2015). Among them, two classes of VGLUT3-expressing interneurons are also CB1R-positive. These include CCK-positive basket and radial trilaminar interneurons. The two other subtypes are negative for CB1R and identify perforant path-associated QuaD and horizontal basket interneurons. As shown in Fig. 4, a dense plexus of VGLUT3-immunoreactive fibers surrounding the stratum pyramidale was detected in the CA1 subfield. Interestingly, most of the sparse VGLUT3/GFP-positive cells detected in strata radiatum and pyramidale were also positive for CB1R (Fig. 4c–e).

The present analysis of the distribution of GFP in Drd1a-EGFP mice suggests that diverse classes of GABAergic interneurons express D1R. Because the distribution of D1R-expressing cells was reminiscent to the one recently described for CA1 D2R-containing neurons (Puighermanal et al. 2015), we analyzed whether GFP/D2R co-expressing cells were present in the CA1 dorsal hippocampus of Drd1a-EGFP. The analysis of endogenous D2R distribution, using anti-D2R antibody (see Table 1), revealed a pattern of expression of D2R-positive cells that resembles to the one recently described (Puighermanal et al. 2015). Indeed, in the dentate gyrus most of the D2R-positive neurons were located in the hilus identifying the hilar mossy cells (Fig. 5a). In the CA1 subfield, D2R-labeled cells were predominantly detected in strata oriens and radiatum (Fig. 5a). In addition, an intense D2R immunoreactivity was detected in stratum lacunosum-moleculare most likely corresponding to the terminals of O-LMs interneurons (Fig. 5a). On the other hand, they were rarely found in stratum pyramidale (Fig. 5a). The pattern of distribution of endogenous D2R-expressing cells was further confirmed by analyzing the degree of co-localization between D2R and HA immunoreactivity in Drd2-Cre::RiboTag mice. As illustrated, all HA-expressing cells located in strata oriens, radiatum, and at the border of strata radiatum/lacunosum-moleculare were also positive for D2R (Fig. 5b, yellow arrows). Only a few neurons were D2R⁺/HA⁻ suggesting that the expression of endogenous D2R might not be fully recapitulated in Drd2-Cre::RiboTag mice (Fig. 5b).

We next examined the degree of co-localization of GFP-labeled cells with D2R in the dorsal CA1 hippocampus of Drd1a-EGFP. In stratum oriens a large majority of GFP-expressing cells were also D2R positive (~82 %). The percentage of GFP/D2R-immunoreactive neurons was also high in strata pyramidale and radiatum, reaching ~64 and ~68 %, respectively (Fig. 5c, d; Table 2). By contrast, in stratum lacunosum-moleculare only 15 GFP/D2R-positive cells have been detected among the 110 GFP-immunoreactive cells (Fig. 5c, d; Table 2).

To further confirm that both receptors were expressed at least by a fraction of hippocampal cells, we took advantage of the Drd2-Cre::RiboTag mice that express tagged-ribosomes selectively in D2R-containing cells (Fig. 6a). After homogenization of the hippocampus, tagged-ribosomes and their bound mRNAs were captured by HA immunoprecipitation (Fig. 6a). The analysis by quantitative RT-PCR (qRT-PCR) of purified mRNAs compared to the input fraction revealed a de-enrichment of glial markers, including Gfap for astrocytes, Cnp1 for oligodendrocytes, and Iba1 for microglia as well as of glutamatergic pyramidal cells markers such as Camk2a and Slc1a1 (EAAT3) (Fig. 6b). By contrast, the glutamatergic Cajal–Retzius and hilar mossy cells marker Calb2 (CR) was highly enriched in mRNAs purified following HA immunoprecipitation (Fig. 6b). Similarly, GABAergic markers including Gad1, Slc32a1 (VIAAT), and Sst (SOM) were also clearly enriched (Fig. 6b) confirming our previous observations (Puighermanal et al. 2015). Finally, the presence of Drd2 mRNA was confirmed as expected, but also Drd1a mRNAs were isolated following HA immunoprecipitation (Fig. 6c), in agreement with the co-localization of GFP and D2R in Drd1a-EGFP mice (Fig. 5). Taken together, these results indicate that in CA1 subfield a large proportion of D1R-expressing cells also contain D2R.

An early tract-tracing study and double immunofluorescence analyses reported that VTA DA neurons projecting to the hippocampus were preferentially localized in strata oriens and pyramidale, with sparse fibers in stratum radiatum and barely any innervation of stratum lacunosum-moleculare (Gasbarri et al. 1994a, b; Kwon et al. 2008). This heterogeneous laminar distribution of DA fibers within CA1 was recently confirmed by analyzing hippocampal efferents from genetically defined VTA DA neurons using Cre-inducible AAV-expressing ChR2-EYFP (Broussard et al. 2016; McNamara et al. 2014; Rosen et al. 2015). Interestingly, our analysis revealed that D1R-expressing cells are found in a large proportion in stratum oriens and to a lesser extent in stratum pyramidale. Thus, in these two layers DA terminals largely overlap with D1R-containing neurons as illustrated by the presence of GFP-labeled neurons in the vicinity of TH-positive fibers (Supplemental Figure 1, Inset). D1R-expressing cells are also present in strata radiatum and at the border of radiatum/lacunosum-moleculare. However, despite the detection of TH immunoreactivity (Supplemental Figure 1), VTA DA neurons do not innervate these two layers (Broussard et al. 2016; McNamara et al. 2014; Rosen et al. 2015). In fact, strong evidence indicate that the dense plexus of TH-labeled fibers observed within strata radiatum and lacunosum-moleculare corresponds to noradrenergic (NE) axons arising from the locus coeruleus (LC) (Kwon et al. 2008). Of interest, a recent study showed that these LC NE fibers co-release DA and NE, suggesting that they constitute the only source of DA in the vicinity of D1R-expressing neurons located in these two layers (Smith and Greene 2012; Walling et al. 2012). Therefore, depending on their laminar location, D1R-expressing neurons could be controlled by DA arising from two distinct sources: the VTA for strata oriens and pyramidale and the LC for strata radiatum and lacunosum-moleculare.

If the expression of D1R by the granule cells in the dentate gyrus is well admitted and has been demonstrated by in situ hybridization, binding, and immunofluorescence studies (Boyson et al. 1986; Fremeau et al. 1991; Gangarossa et al. 2012; Huang et al. 1992; Mansour et al. 1990, 1991; Rocchetti et al. 2015; Sarinana et al. 2014), the distribution/identity of D1R-expressing cells in the CA1 subfield remains unclear. Indeed, although the presence of D1R in pyramidal cells has been suggested (Huang et al. 1992; Kern et al. 2015; Ladepeche et al. 2013a, b), little or no signal for D1R at the transcript level was detected (Fremeau et al. 1991; Mansour et al. 1990; Rocchetti et al. 2015; Sarinana et al. 2014). Consistent with these latter findings, our analysis revealed a weak and sparse distribution of GFP-labeled neurons in stratum pyramidale. This finding strongly suggests that the D1R staining detected in this layer and initially thought to label the plasma membrane of CA1 pyramidal cells (Huang et al. 1992) most likely corresponds to D1R-positive terminals arising from another cell type, possibly GABAergic interneurons. Supporting this hypothesis, D1R-expressing cells were found in strata oriens and radiatum/lacunosum-moleculare (Gangarossa et al. 2012; present study), three layers populated by a large diversity of GABAergic interneurons (Wheeler et al. 2015). The distribution of GFP-labeled cells reported in the present study is consistent with the early description of the D1R expression pattern. Thus, although at low density, autoradiography studies revealed the presence of D1R binding sites in stratum oriens (Mansour et al. 1990, 1991). Moreover, cells containing D1R mRNA have been detected in both strata oriens and radiatum (Fremeau et al. 1991). Finally, both layers also exhibit a strong D1R immunoreactivity (Huang et al. 1992). Combined with previous findings, our results strongly suggest that D1R are preferentially expressed by GABAergic interneurons and not by pyramidal cells within the CA1 subfield.

The use of different neurochemical markers including calcium-binding proteins (parvalbumin, calbindin-D28k, calretinin), neuropeptides (somatostatin, NPY), receptors/transporters (mGluR1α, CB1R, VGLUT3) and miscellaneous markers (nNOS, reelin) allowed us to evaluate the distribution of GFP among 33 out of the 37 types of interneurons known to be present in CA1 (Wheeler et al. 2015). Based on the laminar localization and the percentage of co-localization, we estimate that D1R are expressed by at least eight distinct classes of GABAergic interneurons. For instance, the high percentage of GFP/nNOS-positive cells in all layers indicate that D1R might be expressed by both ivy and neurogliaform cells (Armstrong et al. 2012; Price et al. 2005; Tricoire et al. 2010). Their presence in this latter population is further supported by the strong percentage of GFP/NPY- and GFP/reelin-positive neurons estimated in the stratum lacunosum-moleculare (Fuentealba et al. 2008; Tricoire et al. 2011). In stratum oriens, the co-expression of GFP with SOM/mGluR1α suggests that D1R are expressed by O-LMs and trilaminar interneurons (Chittajallu et al. 2013; Klausberger 2009; Matyas et al. 2004; Tricoire et al. 2011). The presence of GFP/PV-expressing cells also favors the hypothesis that axo-axonic, basket, and bistratified interneurons contain D1R. This observation is further strengthened by the recent demonstration of the critical role played by D1R signaling in PV cells for the consolidation of long-term memory (Karunakaran et al. 2016). Finally, although not quantified, the presence of GFP/VGLUT3/CB1R-positive neurons suggests that basket CCK⁺ might express D1R (Wheeler et al. 2015). Interestingly, most of the D1R-containing cells located in the stratum pyramidale correspond to GABAergic interneurons. Based on the combination of the molecular marker they express, one can reasonably conclude they comprise axo-axonic, basket, bistratified, and ivy cells. Finally, a small fraction of GFP-positive neurons were co-labeled with calbindin-D28k, confirming the scarity of pyramidal neurons expressing D1R. Our cross analysis also allowed us to identify CA1 cell types devoid of D1R. Thus, at least two types of interneurons, the perforant path-associated QuaD located in stratum radiatum and the quadrilaminar interneurons found in both strata radiatum and lacunosum-moleculare (Pawelzik et al. 2002; Tricoire et al. 2011). Finally, the lack of co-localization between GFP and calretinin/reelin-positive cells localized in stratum lacunosum-moleculare supports the absence of D1R in Cajal–Retzius cells (Marchionni et al. 2010; Tricoire et al. 2011). Further experiments using double fluorescent in situ hybridization and/or cell-type specific mRNA profiling should help to further confirm the presence of D1R transcripts in these distinct classes of GABAergic interneurons.

Although BAC transgenic mice expressing fluorescent proteins represent a useful tool to characterize genetically identified cell populations, caution should be taken when analyzing the expression pattern. Indeed, during the course of the characterization of our Drd2-Cre::RiboTag mouse line, which express tagged-ribosomes selectively in D2R-containing cells, we found that in the hippocampus D2R-expressing cells displayed a much widespread pattern than the one initially described in Drd2-EGFP mice (Gangarossa et al. 2012; Puighermanal et al. 2015). The difference was particularly evident in CA1 where HA-positive cells of Drd2-Cre::RiboTag mice identified diverse classes of GABAergic interneurons (Gangarossa et al. 2012; Puighermanal et al. 2015). This observation, together with our present analysis, led us to re-examine whether, in the CA1 subfield, D1R and D2R-expressing cells were fully segregated, as initially thought (Gangarossa et al. 2012; Puighermanal et al. 2015), or could partially overlap. The present findings argue in favor of this last hypothesis. Thus, in strata oriens, pyramidale, and radiatum, our double immunofluorescence analysis revealed a high degree of co-localization between GFP and D2R. The presence of cells co-expressing both D1R and D2R was further confirmed by the enrichment of Drd1 transcripts isolated from tagged-ribosomes expressed in D2R-containing cells. Interestingly, the presence of both receptors on diverse GABAergic interneurons, which for some of them have antagonistic activity onto CA1 pyramidal cells, could account for the complexity and variability of DA action following bath application in hippocampal slices. Indeed, while DA bath application does affect excitatory Schaffer collateral (SC) drive onto CA1 pyramidal cells, a depression of the synaptic transmission of temporoammonic (TA) pathway has been reported (Ito and Schuman 2007; Otmakhova and Lisman 1999). This latter effect, which requires both D1R and D2R, also involves local GABAergic interneurons located at the border of strata radiatum and lacunosum-moleculare (Ito and Schuman 2007; Otmakhova and Lisman 1999). However, at this synapse, following high-frequency stimulation, DA facilitates excitatory drive to CA1 pyramidal cells mainly through the decreased feedforward inhibition (Ito and Schuman 2007; Otmakhova and Lisman 1999). Therefore, one can envision that the ability of DA to gate TA synaptic transmission would not only depend on the excitatory inputs frequency but also depend on the dual action of DA on GABAergic interneurons co-expressing both D1R and D2R. Because these different types of D1R/D2R-expressing interneurons innervate specific and distinct domains of pyramidal cells and other interneurons, future experiments will be necessary to understand whether their pattern of activity will change depending on tonic, phasic, or ramping DA signals.

In conclusion, our study revealed that in the CA1 subfield of the hippocampus, distinct classes of GABAergic interneurons express D1R. Contrasting with the dorsal striatum where D1R and D2R are highly segregated (Bertran-Gonzalez et al. 2010; Valjent et al. 2009), a high degree of D1R-containing neurons also express D2R. Future studies using cell-type specific invalidation of D1R and/or D2R are promptly required to untangle the complexity of DA signals within the hippocampus.