Retrograde Labeling Illuminates Distinct Topographical Organization of D1 and D2 Receptor-Positive Pyramidal Neurons in the Prefrontal Cortex of Mice

Abstract The cortex plays an important role in regulating motivation and cognition, and does so by regulating multiple subcortical brain circuits. Glutamatergic pyramidal neurons in the prefrontal cortex (PFC) are topographically organized in different subregions such as the prelimbic, infralimbic (IL), and orbitofrontal and project to topographically-organized subcortical target regions. Dopamine D1 and D2 receptors are expressed on glutamatergic pyramidal neurons in the PFC. However, it is unclear whether D1 and D2 receptor-expressing pyramidal neurons in the PFC are also topographically organized. We used a retrograde adeno-associated virus (AAVRG)-based approach to illuminate the topographical organization of D1 and D2 receptor-expressing neurons, projecting to distinct striatal and midbrain subregions. Our experiments reveal that AAVRG injection in the nucleus accumbens (NAcc) or dorsal striatum (dSTR) of D1Cre mice labeled distinct neuronal subpopulations in medial orbitofrontal or prelimbic PFC, respectively. However, AAVRG injection in NAcc or dSTR of D2Cre mice labeled medial orbitofrontal, but not medial prelimbic PFC, respectively. Additionally, D2R+ but not D1R+ PFC neurons were labeled on injection of AAVRG in substantia nigra pars compacta (SNpc). Thus, our data are the first to highlight a unique dopamine receptor-specific topographical pattern in the PFC, which could have profound implications for corticostriatal signaling in the basal ganglia.

The striatum is the main input center of the basal ganglia, receiving input from cortical, thalamic, limbic and dopaminergic nuclei (Ikemoto and Bonci, 2014). Glutamatergic signals from cortical areas, and dopamine signals from midbrain dopaminergic nuclei act on spiny projection neurons (SPNs) in the striatum. The SPNs integrate both glutamate and dopamine signals and coordinate various aspects of learning and behavior (Horvitz, 2009;Shiflett and Balleine, 2011;Bamford et al., 2018).
Afferent cortical glutamatergic inputs into the striatum originate within various medial or lateral subregions of the prefrontal cortex (PFC) such as the prelimbic, infralimbic (IL), orbitofrontal, and motor cortex. The PFC plays a critical role in motivation and cognition (Balleine and O'Doherty, 2010;Smith and Graybiel, 2014). Within the medial and lateral PFC, topographically-organized regions such as the dorsally located prelimbic corticostriatal neurons, and the ventral IL or orbitofrontal PFC (OFC) neurons have dissociable effects on motivated behavior and cognitive flexibility (Killcross and Coutureau, 2003;Rudebeck and Murray, 2011;Ahmari et al., 2013;Burguière et al., 2013;Gremel and Costa, 2013;Barker et al., 2017;Hart et al., 2018). Although, these studies elegantly outline the role of sub-regions in the PFC or striatum, few studies have explored the neuronal and molecular diversity of PFC pyramidal neurons involved in regulating motivation and cognition.
Dopamine and its receptors in the PFC also regulate motivated behavior and cognitive flexibility (Goldman-Rakic, 1998;Hitchcott et al., 2007;Barker et al., 2013;Natsheh and Shiflett, 2018;Ott and Nieder, 2019). Dopamine activates D1 and D2 class of receptors in the PFC that signal through stimulatory Gas or inhibitory Gai proteins, respectively, and through b -arrestins as well, which modulate the activity of both pyramidal neurons and interneurons O'Donnell, 2004, 2007;Beaulieu et al., 2007;Santana et al., 2009;Urs et al., 2016;Ferguson and Gao, 2018;Tomasella et al., 2018;Cousineau et al., 2020). Moreover, pharmacological targeting of D1Rs or D2Rs, or genetic deletion of D2Rs in the PFC can regulate dopamine-dependent behaviors such as locomotion, cognition, and goal-directed behavior (Del Arco et al., 2007;Hitchcott et al., 2007;Barker et al., 2013;Urs et al., 2016;Natsheh and Shiflett, 2018;Tomasella et al., 2018;Khlghatyan and Beaulieu, 2020). Although PFC dopamine receptors play an important role in motivation and cognition, the topographical distribution of D1R1 or D2R1 pyramidal neurons in specific subregions of the PFC is not known. Here, we use a retrograde adeno-associated virus (AAVRG)-based approach to identify distinct topographically-organized subpopulations of D1R1 and D2R1 pyramidal neurons in the PFC based on their target projection areas. Given the role of various subregions of the PFC in motivated behavior and cognition, and the heterogeneity of these regions, the effects of dopamine and its receptors within these subregions will expand our understanding of the molecular and neuronal mechanisms regulating motivated behavior and cognition.

Animals
All mouse studies were performed according to NIH guidelines for animal care and use, and were approved through the University Animal Care and Use Committee. All mice were housed in a 12/12 h light/dark cycle at a maximum of five per cage, provided with food and water ad libitum, and tested at 8-12 weeks of age. Mice were age matched and mice of both sexes were used, and all experiments were performed in naive animals. Dopamine D1 receptor Cre (D1Cre, EY262) and Dopamine D2 receptor Cre (D2Cre, ER44) transgenic mice were obtained from MMRRC. Cre1 hemizygous transgenics were used for all experiments.
Immunostaining, imaging, and quantification 40 mm thick vibratome cut sections of formalin-fixed mouse brains were processed for imaging. Sections from rostral, rostrocaudal, and caudal PFC, and striatal and midbrain sections were imaged using a Nikon AZ100 Zoom microscope, using the same exposure across genotypes for an injection pair. Captured images were used for quantifying number of fluorescent cells for each channel (GFP and TdT) in different subregions of rostral, rostrocaudal, and caudal PFC using ImageJ (NIH), and threshold was kept the same between genotypes and injection pairs. At least three sections from rostral, rostrocaudal, and caudal PFC were analyzed for each mouse, with an n = 5 mice per injection pair. For glutamatergic or GABAergic marker identification, we performed antigen retrieval in citrate buffer at 80°C on virally injected PFC sections, and colabeled with antibodies to GFP (Frontier Institute, catalog #AB_2571575), RFP (Rockland, catalog #600-401-379), CamKIIa (Enzo Life Sciences, catalog #ADI-KAM-CA002-D), parvalbumin (PV; Frontier Institute, catalog #AB_2571613), and GAD 65/67 (Frontier Institute, catalog #AB_2571698). Imaging for PV and GAD colabeling were done using the Nikon AZ100 zoom microscope. Imaging for CamKIIa labeling was done using a Nikon spinning disk confocal (CSU-X1, Yokogawa) with either 10Â or 60Â objective on an inverted microscope (Nikon Ti2-E), with a back-thinned sCMOS camera (Prime 95B, Photometrics).

Statistical analyses
Data were analyzed by a standard two-way ANOVA test for comparison between genotypes, and injection pairs. Individual genotypes, or injection pairs were compared using a post hoc Tukey's test. Data are presented as mean 6 SEM, p , 0.05 is considered as significant.

Results
For this study, we focused primarily on dopamine-related subcortical target regions such as the striatum and midbrain dopamine nuclei, i.e., SNpc. The striatum itself can be topographically divided along the dorsoventral or mediolateral axes, into the dorsal and ventral striatum, or DMS and DLS, respectively.
In both D1Cre and D2Cre mice, distinct projection fibers were observed in midbrain regions, i.e., SNpc and substantia nigra pars reticulata (SNpr). In the D1Cre mice, GFP1 projection fibers (from dSTR) were observed in the SNpr, whereas, TdT1 projection fibers were observed in the SNpc (Fig. 1Aii). In contrast, for the D2Cre mice only TdT1 projection fibers were observed in the SNpc, and no labeling in the SNpr (Fig. 1Bii).
PFC projection neurons are primarily glutamatergic, but some studies have shown that a small percent of these projection neurons can be GABAergic (Lee et al., 2014;Melzer et al., 2017), and contribute to physiological outcomes. To confirm whether these retrogradely labeled neurons are glutamatergic or GABAergic, we performed colocalization studies with known glutamatergic neuron marker CamKIIa, and known GABAergic neuron markers GAD 65/67 and parvalbumin (PV). As seen in Extended Data Figure 1-1, retrogradely labeled neurons in D1 or D2Cre mice predominantly colocalize with CamKIIa but not with GAD or PV, thus confirming a glutamatergic identity of these corticostriatal projection neurons.

Mediolateral topographical distribution of D1R1 and D2R1 neurons in the PFC
In the previous experiments, the injection sites were along the dorsoventral axis in the dorsocentral striatum and NAcc core. However, within the dSTR, both medial and lateral subregions have specific roles in motivated behaviors. The DMS is involved in the acquisition of goal-directed actions, whereas the DLS regulates acquisition of habitual behaviors (Yin et al., 2005(Yin et al., , 2006. Previous studies however show that mPFC pyramidal neurons primarily project to the DMS whereas more posterior lateral sensorimotor cortex neurons project to the DLS (Shiflett and Balleine, 2011;Kupferschmidt et al., 2017).
We next asked the question whether PFC D1R1 and D2R1 neurons have specific projection pattern to the DMS or DLS. Similar to Figures 1, 2, we injected 50 nl of Cre-dependent AAVRG GFP and AAVRG TdT in the DMS and DLS, respectively, of D1Cre and D2Cre mice. As seen in Figure 4, we observe distinct topographical patterns for D1R1 and D2R1 neurons projecting to DMS and DLS. For DMS injection in D1Cre mice, we observed robust GFP1 labeling predominantly in the PrL (67 6 12.5 neurons), whereas for DLS injection we observed robust TdT1 labeling predominantly in M1/ M2 (118.3 6 12 neurons) and AI (146 6 24.1 neurons; Fig. 4A,C). Overall, very few D2R1 neurons project to either DMS or DLS (Fig. 4B,D). Comparison of patterns of D1Cre and D2Cre mice for DMS show that only Cg1/ PrL has significantly greater labeling of D1R1 neurons (ppp , 0.01, two-way ANOVA; Fig. 4E). For DLS, however, Cg1/PrL, AI, and M1/M2 show significantly greater labeling for D1Cre compared with D2Cre mice (ppp , 0.01, two-way ANOVA; Fig. 4F).

Discussion
D1R1 and D2R1 neurons are found throughout all PFC subregions (Santana and Artigas, 2017;Anastasiades et al., 2019;Yu et al., 2019;Khlghatyan and Beaulieu, 2020), but this widespread expression pattern does not adequately explain how these neurons mediate distinct physiological and behavioral outcomes. In this study, we show a previously unappreciated distinct topographical organization of D1R1 and D2R1 neurons in the PFC of mice. As summarized in Figure 5, we observe distinct topographical organization patterns of D1R1 and D2R1 neurons along the dorsoventral and mediolateral axes, based on their projection target. Along the dorsoventral axis, PFC D1R1 neurons are topographically organized such that D1R1 neurons in prelimbic regions primarily project to the dSTR, whereas D1R1 neurons in MO and IL regions primarily project to the NAcc core. In contrast, PFC D2R1 neurons have a distinct pattern of organization, such that very few prelimbic D2R1 neurons project to dSTR, but MO and IL D2R1 neurons project to NAcc core, similar to D1R1 neurons. Along the mediolateral axis, D1R1 neurons in the medial prelimbic region primarily project to DMS, whereas D1R1 neurons in M1/M2 motor and agranular insular cortex primarily project to DLS. In contrast, very few D2R1 neurons project to either DMS or DLS. However, medial PFC D2R1 but not D1R1 neurons project to midbrain dopamine nuclei. Thus, our data provide, for the first time a detailed insight into the anatomic organization of D1R1 and D2R1 neurons in the PFC.
In this study, we use AAVRGs to identify afferent PFC inputs into various striatal and midbrain regions. One potential caveat with using AAVRGs is that we cannot control for variability of infection at the injection site, even if we inject the same volume of AAV. However, one advantage of using AAVRGs is that target neurons can be specifically labeled in a Cre-dependent manner, and therefore label cell bodies of specific populations of neurons using transgenic Cre mice. Our data are consistent with previous findings that D1R1 neurons are primarily corticostriatal, whereas D2R1 neurons are corticostriatal and also project to more caudal regions such as the thalamus . Other groups have also shown that PFC D1R1 and D2R1 neurons also project to other limbic areas such as the basolateral amygdala (BLA; Jenni et al., 2017). PFC neurons projecting to BLA, NAcc, or VTA are not only distinct subpopulations but also have distinct laminar distribution (Murugan et al., 2017). In this study we see distinct laminar distribution of both D1R1 and D2R1 PFC neurons. "D1-dSTR" prelimbic neurons are predominantly localized to layer 5, whereas "D1-NAcc" MO/IL neurons are predominantly localized to layer 2/3. Interestingly, "D2-NAcc" MO/IL neurons are predominantly localized to layers 2/3 and 5.
Our data also suggest that distinct predominantly D1R1 neuron subpopulations along the mediolateral axis (Cg1/PrL vs M1/M2/AI), project to the DMS and DLS, respectively. However, D2R1 PFC neurons do not project to either DMS or DLS. Thus, D1R1 PFC neurons might be directly involved in regulation of switching between goaldirected versus habitual actions (Yin et al., 2005(Yin et al., , 2006. Interestingly, a similar topographical pattern is maintained within the SNpc, where medial and lateral dopamine neurons project to DMS and DLS, respectively (Lerner et al., 2015).
The dSTR-NAcc injections (Fig. 1) reveal distinct projection fiber patterns in the SNpr and SNpc. For injections in dSTR, we observed projection fibers in the SNpr only in D1Cre and not D2Cre mice, consistent with AAVRG-GFP labeling striatal D1R1 direct pathway SPNs. However, for NAcc core injections, we observe projection fibers in the medial SNpc and not the SNpr in both D1Cre and D2Cre mice. Our findings are consistent with previous observations that D1R1 and D2R1 NAcc core SPNs do not follow the traditional direct/indirect dichotomy like dSTR SPNs, and instead send projections to ventral pallidum (VP) or midbrain (Sesia et al., 2014;Kupchik et al., 2015;Pardo-Garcia et al., 2019). Moreover, these studies suggest that both D1R1 and D2R1 NAcc core SPNs project to VP, but only D1R1 NAcc core SPNs project to the midbrain. Although we observe projection fibers in the SNpc of D2R1 mice, these are likely direct projections from the MO/IL PFC (Fig. 2B), and not from NAcc core SPNs. Thus, D1R1 pyramidal neurons in the MO/IL projecting to NAcc core, can not only release glutamate and regulate excitability of GABAergic SPNs, but also modulate dopamine release in the DMS by indirectly acting on medial SNpc dopamine neurons. In contrast, D2R1 pyramidal neuron in the MO/IL project directly to both NAcc core and SNpc. One possible caveat of this interpretation is that by using AAVRGs we are unable to establish whether these fibers in the SNpc are afferents on GABAergic or dopaminergic neurons. A more sophisticated approach with rabies virus retrograde labeling, with Cre-dependent labeling of target neurons is required to confirm our interpretation.
The dSTR is topographically divided into the DMS and DLS, which have been implicated in action-outcome learning and stimulus-response learning, respectively, whereas the NAcc has been implicated in reward perception (Shiflett and Balleine, 2011). D1R1 and D2R1 neurons in topographically organized regions in the PFC can thus have various effects on physiology and behavior depending on their striatal or midbrain projection target, and modulation by cortical dopamine.
Illuminating this unique pattern of organization of D1R1 and D2R1 neurons in the PFC will help us better understand the regional and global effects of cortical dopamine and dopamine receptors in the regulation of motivation and cognition.