Structural and molecular heterogeneity of calretinin‐expressing interneurons in the rodent and primate striatum

Abstract Calretinin‐expressing (CR+) interneurons are the most common type of striatal interneuron in primates. However, because CR+ interneurons are relatively scarce in rodent striatum, little is known about their molecular and other properties, and they are typically excluded from models of striatal circuitry. Moreover, CR+ interneurons are often treated in models as a single homogenous population, despite previous descriptions of their heterogeneous structures and spatial distributions in rodents and primates. Here, we demonstrate that, in rodents, the combinatorial expression of secretagogin (Scgn), specificity protein 8 (SP8) and/or LIM homeobox protein 7 (Lhx7) separates striatal CR+ interneurons into three structurally and topographically distinct cell populations. The CR+/Scgn+/SP8+/Lhx7− interneurons are small‐sized (typically 7–11 µm in somatic diameter), possess tortuous, partially spiny dendrites, and are rostrally biased in their positioning within striatum. The CR+/Scgn−/SP8−/Lhx7− interneurons are medium‐sized (typically 12–15 µm), have bipolar dendrites, and are homogenously distributed throughout striatum. The CR+/Scgn−/SP8−/Lhx7+ interneurons are relatively large‐sized (typically 12–20 µm), and have thick, infrequently branching dendrites. Furthermore, we provide the first in vivo electrophysiological recordings of identified CR+ interneurons, all of which were the CR+/Scgn−/SP8−/Lhx7− cell type. In the primate striatum, Scgn co‐expression also identified a topographically distinct CR+ interneuron population with a rostral bias similar to that seen in both rats and mice. Taken together, these results suggest that striatal CR+ interneurons comprise at least three molecularly, structurally, and topographically distinct cell populations in rodents. These properties are partially conserved in primates, in which the relative abundance of CR+ interneurons suggests that they play a critical role in striatal microcircuits.

In both rodents and primates, a more detailed quantification of the molecular identity of CR1 neurons with different structural properties is needed to better differentiate subtypes of CR1 interneuron. Such a characterization could also provide greater understanding of the role of these interneurons in patients with Huntington's disease and Tourette syndrome, where there is a preferential loss of large-sized striatal CR1 interneurons and relative sparing of the medium-sized cells (Cicchetti et al., 2000;Kataoka et al., 2010). Using a combination of immunohistochemistry and stereological cell counting, we demonstrate that the selective expression of Scgn, as well as the transcription factors SP8 and Lhx7, within the CR1 interneuron population can be used to identify three cell "types" that can be distinguished from one another on the basis of their structural properties and distribution within the dorsal striatum of the rat and mouse. We also demonstrate that one of these markers, Scgn, also identifies a subpopulation of CR-expressing interneurons that is unevenly distributed throughout the caudate-putamen in primates. Together these results provide further evidence of functionally distinct subpopulations of CR1 interneurons in the rodent and primate striatum.

| Preparation of rat and mouse brain tissue for immunofluorescence and cell counting
The experimental procedures described below were carried out using 14 adult (3 months old, 280-350 g) male Sprague Dawley rats (Charles River) and 6 adult, 3-month-old C57Bl/6J male mice (Charles River) in accordance with the Animals (Scientific Procedures) Act, 1986 (UK).
After being deeply anaesthetized using isoflurane (4% v/v in oxygen), each rat was given a lethal dose of pentobarbitone (1.3 g/kg; i.p.) followed by transcardial perfusion with approximately 50 mL of 0.05 M phosphate-buffered saline, pH 7.4 (PBS), followed by 300 mL of fixative (4% w/v paraformaldehyde with 0.1% w/v glutaraldehyde in 0.1 M phosphate buffer, pH7.4 (PB)). This was followed by a third perfusion of approximately 200 mL of fixative (4% w/v paraformaldehyde in PB).
Mice were deeply anesthetized with pentobarbitone and perfused transcardially using 20 mL of PBS, followed by 20 mL of fixative (4% w/v paraformaldehyde in 0.1 M PB). For both species, once the brain was removed, the tissue was post-fixed in this solution for 24 hr at 48C. Using a vibrating microtome (Leica VT1000S), 50-mm-thick coronal sections containing the striatum were cut and collected for immunofluorescence processing. For some images, 50-mm-thick parasagittal sections were used to visualize the medial striatum.  Table 1 for details about the sources and dilutions of antibodies used). After exposure to primary antibodies, sections were washed in PBS and incubated overnight at room temperature in Triton-PBS which contained a mixture of secondary antibodies (all raised in donkey) which were conjugated to the following fluorophores: DyLight 649 (1:500; Jackson ImmunoResearch Laboratories); Cy3 (1:1,000; Jackson ImmunoResearch Laboratories); AlexaFluor-488 (1:500; Invitrogen); or AMCA (1:250 dilution; Jackson ImmunoResearch Laboratories). To ensure minimal cross-reactivity, these antibodies were cross adsorbed by the manufacturers. After further washing in PBS, sections were mounted on glass slides (VWR Super Premium Microscope slides) using fluorescence mounting medium (Vectashield; Vector Laboratories), followed by the addition of a coverslip. West, 1999), was used to generate unbiased cell counts, determine the relative expression of molecular markers, and map distributions of striatal interneurons. In all procedures performed, the accuracy of these estimates was ensured by taking absolute counts of all neurons expressing a given molecular marker, thereby allowing for a nearly precise definition of their distribution within the striatum.
Once boundaries were delineated for a given section that had undergone a given immunofluorescence protocol, the selected area was subsequently captured by imaging a series of completely tessellated, zstacked images (with an optical step size of 1 mm) at depths 2-12 lm from the upper surface of each section at the level of the striatum. This was done using the 203 1.8 NA objective lens of an epifluorescence microscope (as above). The thickness of the section was measured at each counting site and then averaged to obtain a correction factor for tissue shrinkage. The average section thickness was found to be 47.8 6 0.73 mm in rats meaning that the calculated shrinkage factor was $4. 4%. In mice, the average section thickness was found to be 46.5 6 0.66 mm, resulting in a calculated shrinkage factor of $7.0%. To minimize confounds arising from surface irregularities, neuropil within a 2 lm "guard zone" at the upper surface was not imaged. By sampling sections in this manner, a 10 mm-thick "optical disector" was generated which had abutting, unbiased 2D counting frames (320 3 420 mm) consisting of two perpendicular exclusion lines and two inclusion lines. This was used to generate all stereological cell counts and molecular expression profiles presented in this study (Garas et al., 2016;Glaser, Greene, & Hendricks, 2007;West, 1999West, , 2012, and allowed for the generation of robust and unbiased stereological cell counts.
Captured images were analyzed and labeled neurons were counted using StereoInvestigator 9.0 software (MBF Biosciences). A labeled  (2007) 1 -Goat anti-Scgn and rabbit anti-Scgn were raised against different secretagogin peptide antigens but resulted in a significant overlap in immunoreactive neurons (Garas et al., 2016). neuron was counted if the top of its nucleus came into focus within the optical disector. If the nucleus was already in focus at the top of the optical disector, the neuron was excluded (West, 1999

| Calculations of section area, volume and estimates of total cell number
Once the number of neurons expressing a marker or markers had been counted and recorded for a given coronal plane, the volume of the striatum in which these neurons had been counted was calculated using StereoInvestigator 9.0 software (MBF Biosciences). The software uses a method that is derivative of the point counting method (Oorschot, 1996). The volume of tissue within the optical disector was calculated by multiplying the cross-sectional area by 10 mm. The density of labeled neurons was then computed by dividing the total number of counted neurons by the volume of the optical disector. By assuming that the density of immunoreactive neurons is constant within the 50-mm-thick tissue section (corrected for tissue shrinkage in mice and rats), then the calculated value for absolute density applies to the entire section. This absolute density value was calculated for every coronal plane studied for all counted neurons, and the values were plotted to demonstrate changes in density along the rostro-caudal axis of the dorsal striatum. A calculation for an estimate of the mean density of counted neurons within the striatum was performed by computing the mean of all densities for each coronal plane that was studied.
In order to calculate an estimate for the total number of a given type of neuron in the striatum, the following equation (Oorschot, 1996) was used: where N is the estimate for the total number of neurons in the structure, Nv is the numerical or volume density, and V(ref) is the volume of that structure.
The volume of the dorsal striatum in rats and mice was calculated using the Cavalieri direct volume estimate (Gundersen & Jensen, 1987), which involves multiplying the sum of the cross-sectional area of every nth serial section multiplied by the fixed distance between each of the sampled sections. The volume of the striatum in both rats and mice was calculated using 13 and 9 equally spaced 50-mm-thick coronal sections, respectively (Gundersen & Jensen, 1987;Rymar et al., 2004). The point counting method option of the StereoInvestigator 9.0 software was used to obtain an estimate of each of the section's cross-sectional area. Briefly, this involved overlaying a grid of equally spaced points over each delineation of the structure, and then counting the number of points that lie within the contour of the structure (West, 1999). The spacing of the points is often set such that approximately 150 points lie within the entire of set of contours for a given structure. This allows for an accurate measurement of each section's area.
Once complete, the volume of the structure, V(ref), was calculated according to the following (Oorschot, 1996;West, 1999): where P is the number of points counted within a delineated contour, a (p) is the area represented by each point, and t is the fixed distance between each section. When calculated in this manner across the striatum of six rats, the mean volume of the dorsal striatum was determined to be 24.2 6 2.1 mm 3 , a value that resembles previous estimates (Oorschot, 1996;Rymar et al., 2004). The mean volume of the mouse striatum (n 5 6 mice) was determined to be 9.4 6 0.5 mm 3 . Once the value of V(ref) is determined, multiplying that volume by the mean neuronal density yields an estimate for the total number of neurons for that animal. Because most neuronal counts were performed across three animals (a sample that is typical of stereological studies), in depth statistics were deemed inappropriate, and so the value calculated from each animal is plotted along with the mean.
In order to estimate the precision of this estimate for the volume of striatum, the coefficient of error (CE) was calculated using the Stereo investigator software. In all such studies, a CE (m 5 1) less than 0.1 is the agreed upon value which indicates that the level of variation in striatal volume calculation between different animals is a product of biological, rather than methodological variation (Gundersen, Jensen, Kieu, & Nielsen, 1999). In this study, the coefficient of error was kept below the value of 0.1 for each animal by first ensuring that each section's cross sectional area was calculated using approximately 150 points as per the point counting method. Secondly, because each volume estimate was performed using 13 striatal sections in the rat and 9 striatal sections in the mouse, this ensured that the striatum cross-sectional area was sampled at approximately every 9th section in the rat and every 7th section in the mouse. This degree of sampling has previously been shown to ensure a CE of less than 0.1 when estimating the volume of complex structures (Schmitz & Hof, 2005;Slomianka & West, 2005).

| Topographical and statistical analysis of interneuron distribution
neurons in the dorsal striatum, horizontal, and vertical lines were first placed from the position of each interneuron within a given coronal plane. Each interneuron was then assigned a value of between 21 and 1 according to its relative distance from the medio-lateral and dorsoventral borders of striatum at that same coronal plane. Notably, these two distance values were calculated according to the first points along the striatal contours that were contacted by the horizontal and vertical lines drawn from each neurons position. In this manner, the definition of "medial" and "lateral" striatum was shifted depending on the position of the interneuron in the dorso-ventral axis and vice versa. This method of normalization therefore accommodates the irregular, ever changing shape of the dorsal striatum when viewed over multiple coronal planes.
As each counted interneuron in a section was given a value for each axis, and at least 3-6 animals were used for each protocol, this sample size was deemed large enough to use the Wilcoxon signed-rank test in order to determine if the mean relative value of a neural population in a given coronal plane differed significantly from 0, a value indicating an unbiased distribution along a given axis. For coronal planes in which 5 or fewer neurons of a given class were counted within the confines of the optical disector, the sample size was deemed too small for the generation of reproducible statistics. The minimum significance level for these statistical tests was taken to be p .05, and is denoted in plots by the presence of an asterisk (*). Mann-Whitney U tests where used to compare the mean positional values of two different interneuron populations along the medio-lateral or dorso-ventral axis within a given coronal plane. The minimum significance level for these statistical tests was taken to be p .05, and is denoted by the presence of a box (w). Multiple comparisons across sections were corrected using the false discovery rate (FDR) method (Noble, 2009).

| Measurements of mean somatic diameters of interneurons
Immunofluorescent detection of markers that labeled the entirety of the somata of certain classes of CR1 interneurons was used in order to measure the somatic diameters of individual neurons in the dorsal striatum of the rat, mouse and primate. A neuron was only measured if the contours of its entire soma in all three dimensions were contained within the 50-mm-thick section of tissue. As previously described (Rymar et al., 2004), this was done in order to ensure that the longest axis of a chosen neuron's soma was imaged. The somatic diameter of a labeled neuron was defined as the length of its longest axis, and was measured using StereoInvestigator software. For a given class of interneurons, the mean of all somatic diameters measured in this manner was calculated. Statistical comparisons of multiple mean somatic diameters (of multiple classes of interneuron) were performed using a Kruskal-Wallis ANOVA followed by post hoc Dunn tests (where appropriate).
As such, the 403 lens (NA: 0.8) of a laser scanning confocal microscope (Zeiss LSM 710) was used to image well-labeled CR-expressing interneurons at multiple optical planes in order to form a z-stack to highlight their somatodendritic structure. Single-plane images highlighting the somata of these neurons were also taken in order to determine whether these neurons co-expressed combinations of secretagogin, SP8, and Lhx7. In order to illustrate the rostral bias of those CR1 interneurons that co-expressed SP8, parasagittal sections containing the medial striatum, the lateral ventricle and the subventricular zone were co-labeled for CR and SP8. Imaging of these sections was performed Electronic Design). As described previously, brain state was defined based on the oscillatory content of the ECoG and categorized as either slow-wave activity (SWA) or cortical activation (Garas et al., 2016;Magill, Bolam, & Bevan, 2000;Magill, Sharott, Bolam, & Brown, 2004;Sharott et al., 2012) Once a neuron was recorded, it was juxtacellularly labeled with neurobiotin (Garas et al., 2016;Sharott et al., 2012). Such labeling was achieved by sending positive current pulses (2-10 nA, 200 ms, 50% duty cycle) through the electrode until the recorded single-unit activity was "entrained" by the injection of current. Approximately 2-6 hr after labeling, animals were given a lethal dose of ketamine (150 mg/kg) followed by a transcardial perfusion with PBS, followed by fixative, as described above. Life Technologies). Sections were analyzed for Cy3 signal, and those containing neurobiotin-labeled cells were isolated for further molecular characterization by indirect immunofluorescence (Garas et al., 2016;Sharott et al., 2012). Neurobiotin-labeled neurons appearing to possess dendritic spines were defined as SPNs. In certain cases, they were also tested for their expression of Ctip2, a nuclear marker that has been shown to be expressed in the SPNs, but not in interneurons, in rodents.

| Tissue processing for identification of recorded and juxtacellularly labeled interneurons
Neurons that did not express Ctip2, or had aspiny dendrites, were tested for their expression of one or more of the classical markers of striatal interneurons: CR, parvalbumin, choline acetyltransferase, and nitric oxide synthase. The initial molecular marker tested was guided by the labeled cell's somatodendritic structure and position within the striatum. Once a positive expression of one of these markers was established, no other classical marker was tested since these molecules are rarely co-expressed in rodent striatal interneurons (Kawaguchi, 1993). Interneurons that expressed CR were further tested for their expression of secretagogin (Garas et al., 2016;Mulder et al., 2009) and Lhx7 (Pangas et al., 2006). All indirect immunofluorescence and subsequent epifluorescence and/or confocal imaging of neurobiotin-labeled interneurons was performed using protocols described above for stereological cell counting and structural imaging of different subtypes of cells.

| Data selection and analysis regarding the firing rate and regularity of identified striatal interneurons
Electrophysiological data were visually inspected, and epochs of robust cortical SWA or cortical activation were selected based on previous descriptions of these brain states (Garas et al., 2016;Sharott et al., 2012;Sharott, Vinciati, Nakamura, & Magill, 2017). Recordings were included for further analysis only when they were artifact free, and had a minimum duration of 50 s regardless of the brain state (185 6 37.4 s; range 53-404 s). Spike sorting procedures such as template matching, principal component analysis and clustering (Mallet et al., 2008) were applied using Spike 2 software in order to isolate single-unit activity.
Isolation of a single unit was verified by the presence of a distinct refractory period in the interspike interval (ISI) histogram. Further analysis was later conducted after having converted single-unit activity into a binary digital event (Spike 2) which was subsequently imported and analyzed using MATLAB software (Mathworks, version R2014b). Due to the low probability of "detecting" CR1 neurons using "blinded" in vivo recordings (see Results), only 4 CR1 interneurons were recorded and successfully labeled in the manner described above. As such, the average spike waveforms, firing rates and ISIs of all these recorded interneurons are individually presented.  (Paxinos, Huang, & Toga, 2000) from one monkey was similar to that in the second animal.

| Tissue preparation and indirect
These sections were washed with PBS and pre-incubated for 2 hr in a solution consisting of 10% v/v NDS (normal donkey serum) and 0.3% Triton-PBS.
Sections were incubated over three nights at room temperature in a solution of Triton-PBS which contained primary antibodies against Scgn and CR. Table 1 contains the details, source and the dilutions used for these antibodies, as they did not differ from those used for rat and mouse tissue. The tissue was then washed in PBS before being incubated overnight at room temperature in a solution of Triton-PBS and one of the fluorophore-conjugated secondary antibodies. After further washing in PBS, sections were mounted on glass slides (VWR Super Premium Microscope slides) using fluorescence mounting medium (Vectashield; Vector Laboratories), followed by the addition of a coverslip. The caudate and putamen of each section were imaged using a Zeiss Imager M2 epifluorescence microscope equipped with a 203 objective and StereoInvestigator 9.0 software. The delineation of each structure was defined using the monkey brain atlas (Paxinos et al., 2000) and was done on a section-by-section basis. Once delineated, the acquisition and subsequent counting of labeled neurons was carried out in precisely the same manner as was done in rodent tissue (see above).

| Analysis of stereological cell counts in the rhesus macaque
For each combination of immunofluorescence markers studied, the number of labeled neurons was counted in the caudate or the putamen (treated as separate structures) and recorded for a given coronal plane using the same principles described for the rodent.

| R E SU LTS
3.1 | Topographically discrete populations of CR1 interneurons in the rat dorsal striatum can be identified by the selective expression of Scgn In the adult rat, the CBP secretagogin (Scgn) is expressed in a proportion of striatal interneurons (Garas et al., 2016;Kosaka et al., 2017;Mulder et al., 2009). We observed that some, but not all, CR1 interneurons co-express Scgn (Figure 1a, b). In the CR1/Scgn1 neuron population, the expression of these CPBs was relatively intense, such that immunoreactivity for both markers could often be detected in proximal dendrites as well as in somata (Figure 1a). Previous descriptions of CR1 interneurons in the rat (Bennett & Bolam, 1993;Rymar et al., 2004) and mouse (Petryszyn et al., 2014)  We hypothesized that these groupings of CR1 interneurons on the basis of their molecular identity and somatic size would correspond to previous qualitative observations of heterogeneity in somatodendritic structure in the CR1 interneuron population (Petryszyn et al., 2014;Tepper et al., 2010). Confocal imaging of CR1 interneurons in the rat striatum suggested that the three groups we describe above broadly correspond with the three morphologically defined cell "types" suggested by Tepper and colleagues (Tepper et al., 2010). In addition to having relatively large somata, CR1/Scgn2/Lhx71 interneurons consistently had thick primary dendrites that branched infrequently ("Type 1", Figure 6a, b). In contrast, the medium-sized CR1/Scgn2/ Lhx72 interneurons has thinner, bipolar dendrites ("Type 2", Figure 6c).
Overall, these findings suggest that CR1 interneurons in the rat dorsal striatum are likely composed of at least three distinct cell types that can be identified on the bases of their molecular expression profiles, somatodendritic structures, and topographical distributions. interneurons, and were not strongly biased in their distributions ( Figure   8g, h). In accordance with previous estimates (Petryszyn et al., 2014), we report that the mean density of all CR1 interneurons in the dorsal striatum mouse is similar to that of the rat (Figure 8h).
As in the rat, the mean somatic diameters of the three molecularly

| D ISC USSION
The results presented in this study provide the most comprehensive correlations to date between the molecular expression profiles, morphological properties and topographical distributions of CR1 interneurons in the dorsal striatum of the rat, mouse, and primate. We demonstrate that, in rodents, the expression of different combinations of Scgn, SP8, and Lhx7 correlate strongly with the topographic and structural properties of CR1 interneurons. Together, these findings suggest that there are at least three distinct classes of CR1 interneurons in the striatum of the rat and the mouse. Furthermore, the spatial distribution of CR1/Scgn1 interneurons in primates has striking similarities to that of rodents, suggesting that these properties are conserved to some extent throughout phylogeny. In addition, we provide the first electrophysiological recordings from a small number CR1 interneurons, which demonstrate that some striatal CR1 interneurons can be "tonically active" in vivo.

| Secretagogin co-expression delineates smallsized rostrally biased CR1 interneurons that may be adult-born
The most conspicuous subtype of CR1 interneurons in rodents have small somata and tortuous, spiny dendrites, and are found almost exclusively in the rostromedial striatum ("Type 3"). These CR1 interneurons have been described previously (Petryszyn et al., 2014;Revishchin, Okhotin, & Pavlova, 2010;Tepper et al., 2010;Wei et al., 2011) and are thought to be cells that migrate postnatally from the SVZ into the rostromedial striatum, where they mature into CR1 interneurons (Dayer et al., 2005;Inta et al., 2015;Luzzati, De Marchis, Fasolo, & Peretto, 2006). In rodents, this process is thought to occur throughout adulthood (Revishchin et al., 2010) and is accelerated in response to stroke (Wei et al., 2011). We provide further evidence that these smallsized CR1 interneurons represent a distinct subpopulation, by   in governing striatal function in the rat as compared to mouse. The relatively small proportion of CR1/Scgn1 interneurons in primates suggests that the greater abundance of CR1 interneurons as a whole (Petryszyn et al., 2014;Petryszyn, Di Paolo, Parent, & Parent, 2016) is driven to a large extent by a selective proliferation of CR1/Scgn2 neurons. A greater understanding of other functional properties of the different classes of CR1 interneuron is needed to further interpret the relevance of these species differences.
The biased rostromedial distribution of these CR1/Scgn1 interneurons in rodent striatum is particularly striking. The broad function of neurons in the striatum generally corresponds to the type of cortical input to a given territory (Hintiryan et al., 2016;Hunnicutt et al., 2016).
The rodent rostromedial striatum receives cortical projections primarily from associative cortices in the rat (Hoover & Vertes, 2011;Mailly, Aliane, Groenewegen, Haber, & Deniau, 2013). However, to our knowledge, there is no information about the extent to which these putative newborn neurons are "functionally embedded" into the striatal microcircuit. Thus, in this case, proximity to the SVZ could be an equally or more important factor underlying the high density of this cell population in rostromedial areas of striatum. Interestingly, we found that CR1/Scgn1 interneurons in the primate striatum were also considerably more abundant in rostral areas, particularly the head of the caudate nucleus, that receive input from associative cortices (Yeterian & Pandya, 1993. As we could not test whether these monkey CR1/Scgn1 interneurons co-expressed SP8, further work will be needed to establish whether these interneurons are also putative adult-born cells or represent a topographically specific population with a function related to those associative cortical areas in primates. In either case, and as recently described for parvalbumin (PV)-expressing striatal interneurons (Garas et al., 2016), Scgn appears to delineate a specialized subpopulation of CR1 interneurons in rodents and primates.

| Medium-sized CR1 interneurons do not express Lhx7 or secretagogin
Early qualitative descriptions of CR1 interneurons in the rodent dorsal striatum suggested they were "medium-sized" (Bennett & Bolam, 1993;Jacobowitz & Winsky, 1991), an observation later extended to monkeys (Parent et al., 1996) and humans (Cicchetti et al., 1998). Our use of Scgn and Lhx7 immunoreactivity, as well as an in-depth investigation of somatic size, revealed that medium-sized CR1 interneurons (as defined by their somatic diameter and their lack of Scgn or Lhx7 expression) make up 20% of all CR1 interneurons in the rat and nearly 50% of all CR1 interneurons in the mouse. These "Type 2" mediumsized CR1/Scgn2/Lhx72 interneurons were preferentially distributed in the medial striatum in rats. In the primate striatum, we found that CR1/Scgn2 interneurons (irrespective of Lhx7, which was not tested) did not demonstrate any strong or consistent distributional bias within either the caudate or the putamen. This is in agreement with many studies of interneuron distribution in the primate striatum, where any preference for localizing to one particular region was not detected (Deng et al., 2010;Petryszyn et al., 2014Petryszyn et al., , 2016 4.3 | Medium-sized CR1 interneurons are "tonically active" during cortical activation in the rat The functional roles performed by GABAergic CR1 interneurons within the striatal microcircuit are unclear, which is partly due to the lack of data regarding their local synaptic connectivity and electrophysiological properties (Tepper et al., 2010). To our knowledge, we herein provide the first definition of the in vivo electrophysiological properties of identified CR1 interneurons. By chance, all of our single-cell recordings were made from medium-sized, Type 2 CR1 interneurons (CR1/ Scgn2/Lhx72). Our recordings suggest that CR1/Scgn2/Lhx72 interneurons can be "fast spiking" (i.e., have a spike duration of around 1 ms) and have generally heterogeneous firing rates and patterns, particularly during cortical slow-wave activity. The most consistent feature of these interneurons was there near continuous, regular firing at 3-12 spikes/s during cortical activation. These firing properties resemble those of cholinergic interneurons and some PV-expressing interneurons (Sharott et al. 2012;Doig et al. 2014;Garas et al, 2016), but not of SPNs or nitric oxide synthase-expressing interneurons (Sharott et al., 2012;Sharott et al., 2017), as also recorded in this brain state in anesthetized rats. If these firing properties of CR1/Scgn2/Lhx72 interneurons are maintained in awake, behaving animals, this cell type might be confused (erroneously grouped) with putative cholinergic interneurons, which are often assigned this identity on the basis of "tonically active" firing patterns (Goldberg & Reynolds, 2011). In rodents, the sparsity of these CR1 interneurons makes it likely that they would constitute only a small number of recorded neurons in behaving animals. In primates, however, where CR1 interneurons are far more numerous, extracellular recordings of tonically active neurons could encompass both cholinergic interneurons and a significant number of CR1 interneurons. It should be noted that, if the waveform of CR1 interneurons were to differ from the long, triphasic action potential of putative cholinergic interneurons, careful examination of the action potential waveform could partly resolve this ambiguity (Adler, Katabi, Finkes, Prut, & Bergman, 2013). Given the high density of CR1 neurons in the primate striatum and the importance of primate data in elucidating the role of different striatal interneuron types in relation to behavior, our data suggests that this issue at least warrants further exploration in these primate data sets.

| Lhx7 expression in the large-sized CR1
interneurons of the rodent may link them to the large CR1/ChAT1 expressing interneurons of the primate In the striatum of squirrel monkeys (Parent et al., 1996) and humans (Cicchetti et al., 1998), many large-sized (>20 mm) CR1 interneurons co-express ChAT and are highly analogous in somatodendritic structure to the large striatal cholinergic interneurons that do not co-express CR (Cicchetti et al., 2000;Parent et al., 1996). Although ChAT and CR are not co-localized in interneurons of the rodent striatum, relatively large (>15 mm) CR1 interneurons have been described in the striatum of both mice (Petryszyn et al., 2014) and rats (Rymar et al., 2004). Here, we show that these larger CR1 interneurons selectively express the transcription factor Lhx7, which is required for the development of cholinergic interneurons from embryonic medial ganglionic eminence (MGE)-derived cell progenitors (Fragkouli et al., 2009). This finding suggests that CR1 interneurons co-expressing Lhx7 could constitute a separate subtype that is related in lineage to ChAT1 interneurons. If so, the relatively large-sized CR1/Scgn2/Lhx71 interneurons of the rodent could be related to the large CR1/ChAT1 interneurons of the primate. However, although the CR1/Scgn2/Lhx71 interneurons are the largest of CR1 interneurons, they are smaller than ChAT1 interneurons in rodents and primates (Petryszyn et al., 2016;Petryszyn et al., 2017). This observation supports the alternative premise that ChAT1/CR2 and ChAT1/CR1 interneurons in primates are more strongly related to ChAT1 interneurons in rodents, with CR1/Scgn2/ Lhx71 interneurons comprising an entirely distinct cell type. Further molecular characterization of striatal interneurons across the life courses of rodents and primates will be needed to resolve this issue.
Whatever the case, the expression of Lhx7 by large "Type 1" CR1 interneurons in the adult rat is by itself significant, as it suggests that these neurons are also derived from the MGE, whereas CR1/Scgn1/ Lhx72 interneurons expressing SP8 are likely to be derived from the SVZ or embryonic caudal ganglionic eminence (Inta et al., 2015). This implied difference in developmental linage, as revealed by the selective expression of Scgn or Lhx7, provides further evidence that the molecular characterization described here identifies a minimum of three types of striatal CR1 interneuron. In mice, the density of CR1/Scgn2/ Lhx71 interneurons was around half that of CR1/Scgn2/Lhx72 interneurons, whereas these proportions were approximately reversed in rats. This could reflect further differences between the striatal microcircuits of mice and rats, in addition to those we have recently described for parvalbumin-expressing interneurons (Garas et al., 2016).
The present study provides a novel correlative approach to identify different populations of morphologically, topographically, and molecularly distinct CR1 interneurons in the striatum of both mice and rats.
One of these populations, namely rostrally biased CR1/Scgn1 interneurons, was partially conserved between the rodent and primate. Our molecular characterization of striatal CR1 interneurons using Scgn and Lhx7 may also provide a useful starting point for the design of celltype-selective manipulations that can be used to disentangle the specialized contributions of different populations of CR1 interneurons to activity dynamics within the striatal microcircuit.