Corticocortical connections of the rostral forelimb area in rats: a quantitative tract-tracing study

Abstract

The rostral forelimb area (RFA) in the rat is a premotor cortical region based on its dense efferent projections to primary motor cortex. This study describes corticocortical connections of RFA and the relative strength of connections with other cortical areas. The goal was to provide a better understanding of the cortical network in which RFA participates, and thus, determine its function in sensorimotor behavior. The RFA of adult male Long-Evans rats (n = 6) was identified using intracortical microstimulation techniques and injected with the tract-tracer, biotinylated dextran amine (BDA). In post-mortem tissue, locations of BDA-labeled terminal boutons and neuronal somata were plotted and superimposed on cortical field boundaries. Quantitative estimates of terminal boutons in each region of interest were based on unbiased stereological methods. The results demonstrate that RFA has dense connections with primary motor cortex and frontal cortex medial and lateral to RFA. Moderate connections were found with insular cortex, primary somatosensory cortex (S1), the M1/S1 overlap zone, and lateral somatosensory areas. Cortical connections of RFA in rat are strikingly similar to cortical connections of the ventral premotor cortex in non-human primates, suggesting that these areas share similar functions and allow greater translation of rodent premotor cortex studies to primates.

The motor areas of the cerebral cortex of mammals generally have been divided into the primary motor cortex (M1) and various premotor areas. By definition, premotor areas provide direct input to M1 (Dum and Strick 2002; Kunori and Takashima 2016). While several premotor areas have been identified in non-human primates, the status of premotor cortex in rodent species is still unclear. Rodents are used increasingly in studies focused on understanding the cortical control of movement, and the role of premotor areas in recovery after injury to M1 (Urban et al. 2012; Brown and Teskey 2014; Nishibe et al. 2015; Tennant et al. 2015; Kunori and Takashima 2016; Mohammed and Jain 2016; Okabe et al. 2016; Wang et al. 2016; Shiromoto et al. 2017; Jeffers et al. 2020; Averna et al. 2021; Gao et al. 2021; Forghani et al. 2023; Hudson et al. 2023). Therefore, it is important to understand the relationship of premotor areas in rodents and primates, and the similarity of their anatomical connectivity with other cortical areas.

Functionally, M1 is thought to be in the most direct control of movements of skeletal musculature. While the non-primary motor areas also have descending connections to the spinal cord (Nudo and Masterton 1990), premotor cortex is thought to be involved with higher order processing, affecting motor control primarily by modulating M1 (Dum and Strick 2002). Numerous interconnections between premotor areas and M1 could relay information about movement selection, preparation for movement, generation of movement sequences, or sensory guidance of movement. Thus, the cortical connectivity of premotor areas may provide critical information regarding the information flow within the cortex, and thus, the putative role that premotor areas play in sensorimotor control in rodents.

In rats, cortical motor areas have been delineated on the basis of cytoarchitecture, neural tract-tracing, and the motor response to electrical stimulation (Donoghue and Wise 1982; Strick et al. 2021). Using intracortical microstimulation (ICMS) techniques, 2 motor forelimb fields can be identified. The larger field is the region from which movements can be evoked by the lowest intensity of ICMS current. Cytoarchitechtonically, this zone is identified as the lateral agranular field containing large layer 5 neurons. Because of its similarity to M1 in primates, this area is typically considered to be the rat’s M1. The forelimb representation in rat M1 has been most studied and is called the caudal forelimb area (CFA).

A second region from which forelimb movements can be evoked by ICMS has been called the rostral forelimb area (RFA). Based on ICMS results, this area is separated from CFA by a thin strip of cortex where ICMS elicits neck or whisker movements. While cytoarchitectonically distinct from the CFA, layer 5 neurons in both areas send fibers to the spinal cord which terminate in the intermediate zone and ventral horn and make disynaptic connections with forelimb motoneurons (Rouiller et al. 1993; Strick et al. 2021), much like frontal motor areas in non-human primates. Similar to M1 and premotor cortex of non-human primates, CFA and RFA are strongly and reciprocally interconnected by corticocortical fibers.

These anatomical and physiological commonalities have led several investigators to propose that the RFA is either the rat premotor cortex, or the supplementary motor area (SMA; Rouiller et al. 1993; Nudo and Frost 2007). One approach to compare the functional importance of cortical areas in different species is to understand similarities and differences in the connection patterns with other cortical areas, especially with those areas outside of M1. While the basic connectivity of RFA has been described in a qualitative way, more quantitative details are needed to resolve this question. Due to the small size of the rat brain relative to larger non-human primate brains, injection cores have been relatively large in most tract-tracing studies, especially those conducted before the advent of labeled dextrans and other modern tracing techniques. Thus, maintaining tracer injections to a small region such as the RFA, occupying about 1 mm² of the cortical surface (Barbay et al. 2013), is challenging. As tract-tracing methods have evolved, smaller and more confined injection cores have allowed for a more precise picture of corticocortical connections. Few studies using more modern tracing techniques have been conducted in RFA, and none have used quantitative and unbiased stereological approaches. Lastly, although the laminar distribution of connectivity to/from RFA has been described in detail by examining labeling in coronal sections, tangential sectioning allows more precise definition of topographic relationships of terminal fields. This study employed ICMS to define the boundaries of RFA, utilized small injections of labeled dextrans, and tangential sectioning to understand the anterograde and retrograde connectivity of this functionally important region of rat cortex. Finally, using dextrans optimized for anterograde labeling, the distribution of terminal boutons was quantified throughout the cerebral cortex using unbiased stereological analysis.

The results of this study provide a quantitative description of the relative density of connections of the RFA with other cortical areas. We confirm that the ipsilateral corticocortical connectivity of RFA is not diffuse throughout the cortex but has clusters of connectivity with a preference for forelimb motor regions, frontal areas lateral to RFA, rostral insular areas, and higher order somatosensory processing areas. Further, we rank ordered the density of anterograde connections, providing further documentation of the corticocortical hierarchy. Retrograde labeling demonstrated a similar qualitative pattern and suggests that corticocortical connectivity with RFA is largely reciprocal.

Materials and methods

Subjects

Adult male Long-Evans hooded rats (n = 8 entered study; n = 6 analyzed; 370–450 g; 3–5 mo of age Harlan, Indianapolis, IN) were singly housed with a 12 h:12 h light:dark cycle. Food and water were provided ad libitum. Of the 8 animals that underwent surgical procedures, 6 were used for examination of the distribution and quantification of biotinylated dextran amine (BDA)-labeling and reported here. One additional animal died at the end of the surgical procedure. In another animal, the CO staining procedure failed to reveal the S1 representation. The Institutional Animal Care and Use Committee of the University of Kansas Medical Center approved all animal use. All experiments were conducted in accordance with the guidelines published in the NIH Guide for the Care and Use of Laboratory Animals.

Surgical procedure

Isoflurane sedation was followed by ketamine [100–80 mg/kg, IP] and xylazine [30 mg/kg, IM] anesthesia. Supplemental doses of ketamine (20 mg/kg IM) were provided throughout the procedure as needed to maintain stable anesthetic depth. After the rat was secured in a stereotaxic frame, Bupivacaine (2.5 mg, local anesthetic) was applied to the scalp. A homeothermic blanket system maintained physiological body temperature. The scalp was incised and reflected, and muscles attached to the temporal and occipital ridges were released. The cisterna magna was opened to relieve cerebrospinal fluid, and a craniotomy performed from +5 anterior to −4 mm posterior to Bregma, and from +1 mm lateral to the midline to the temporal ridge. The dura was reflected, and warm sterile silicone oil applied to the cortex.

Motor areas (CFA and RFA) were identified by ICMS methods (Urban 3rd et al. 2012; Barbay et al. 2013; Hayley et al. 2023). Briefly, a digital photomicrograph of the cortical surface vasculature was obtained through the surgical microscope (Fig. 1A) and overlaid with a grid pattern (250 μm) in image software (Canvas, Deneba Software, Miami, FL). A tapered and beveled glass micropipette (20 μm outside diameter) filled with concentrated saline solution (3.5 M) served as the microelectrode. Beginning with CFA, the pipette was inserted 1725 μm below the cortical surface at every other grid intersection to provide a resolution of 500 μm. A stimulation pulse train (40 ms duration), consisting of 13 monophasic cathodal pulses (200 μsec duration, 350 Hz) was delivered each second from an electrically isolated, charge-balanced (capacitively coupled), constant-current stimulation circuit (BSI-2, Bak Electronics Inc, Mount Airy, MD). The current was increased gradually from 0 μA until a movement was visible, then reduced until the movement was no longer visible. Stimulation of sites defined as “nonresponsive” did not elicit movements at the maximum current level of 80 μA.

Fig. 1.

Experimental overview. A. Photomicrograph of the cortex surface vasculature following craniotomy (case R11–34). Location of ICMS-defined RFA (green) and CFA (yellow) superimposed on surface vasculature. The black dots indicate microelectrode penetration sites. The size of the dots is inversely related to current threshold for evoking movements (see legend). The circle within RFA indicates site of BDA injection. The red dots indicate site of CTB647 injections used as fiducial markers. B. Inset to show relationship of RFA/CFA locations to sections in G–J. C. Photomicrograph of BDA injection core (black outline) within RFA in 50 μm thick, flattened, tangential histological section. D. Retrogradely labeled neuronal soma and processes stained with BDA (brown). E. BDA-labeled axonal projections with boutons (arrow heads). F. Photomicrograph of CTB647 injection site. Scale bar = 0.1 mm. G. Flattened, tangential section stained with cytochrome oxidase to indicate location of sensory areas. Somatosensory representations are outlined. H. Outline of sensory and other relevant areas based on cytochrome oxidase and BDA staining in G. The granular zone of S1 (light gray areas) is further divided into S1H, GZ, and PMBSF. I. Location of 100 μm X 100 μm (plotting area) x 50 μm (section thickness) voxels containing BDA-labeled boutons. Blue dot = voxel containing 2–30 boutons; red dot = voxel containing > 30 boutons. Note: Case R11–34 contained the 2^nd greatest number of labeled voxels. J. Location of voxels containing BDA-labeled somata (green dots). See Table 1 for abbreviations.

After defining the borders of both CFA and RFA, a micropipette containing the neuronal tract-tracer, BDA, 10,000 MW (BDA10kDa, 10% w/v in 0.9% sterile saline) was placed at approximately the center of RFA. The injection needle included a tapered glass micropipette cut to 60 μm outside diameter and was attached with beeswax to a 1 μL Hamilton syringe (30,100, Hamilton Company, Reno, NV). It was actuated by a microinjector (Micro4, World Precision Instruments, Sarasota, FL). Injection depth was controlled by a hydraulic Microdrive (650 Micropositioner, David Kopf Instruments, Tujunga, CA) on a stereotaxic arm. BDA10kDa was pressure injected in 3 boluses of 33.3 nL each (100 nL total) at 1500, 1250, and 1000 μm below the cortical surface. Cholera toxin beta subunit conjugated to AlexaFluor 647 (CTB647, 5 μg/μL in 0.9% sterile saline, C34778, Invitrogen, Grand Island, NY) was injected (with the same configuration and outside diameter as BDA10kDa) in 2 boluses of 75 nL each (150 nL total) at 1500 and 1000 μm below the cortical surface at 2 sites that were each roughly 1 mm caudal to the caudal border of CFA as defined by ICMS, serving as fiducial markers.

The cortical surface was rinsed with warm sterile saline (0.9%) and covered with a silicone sheet (Invotec International Inc, Jacksonville, FL), gel foam, (Surgifoam, Ethicon, Sommerville NJ), and dental acrylic and resin (Lang Dental Mfg Co Inc, Wheeling, IL) to form a protective cap over the craniotomy. The skin was sutured, penicillin injected (45,000 U, SQ) into the nape and local anesthetic (Bupivacaine, 2.5 mg, APP Pharmaceuticals, Schaumburg, IL) and topical antibiotic (Vetropolylycine gel, Dechra Veterinary Products, OP, KS) applied. Buprenorphine (0.05 mg/kg SQ, Reckitt Benckiser Farmaceuticals Inc, Richmond, VA) and acetaminophen (40 mg/kg oral) were given after the surgery for pain management. The rat was allowed to recover on the heating pad until it was alert and moving spontaneously and then returned to its home cage. Three additional doses of buprenorphine and acetaminophen were given during the subsequent 48 h.

Histology

Tissue harvesting. Seven days after the surgical procedure, rats were sedated with isoflurane, and euthanized with Beuthanasia-D (390 mg pentobarbital, 50 mg phenytoin sodium IP, Shering Plough Animal Health, Union, NJ). After thoracotomy and rib cage reflection, heparin sodium (500 USP Units, Hospira Inc, IL) was injected into the left ventricle, and exsanguination was achieved through transcardial perfusion of saline solution [0.9% saline in distilled water, heparin sodium (1,000 USP Units, APP Pharmaceuticals, Schaumburg, IL) and lidocaine HCl (20 mg, APP Pharmaceuticals, Lake Forest, IL)] followed by 3% paraformaldehyde in 0.9% saline. The brain was extracted, both hemispheres of cortex were separated from the underlying structures and flattened between glass slides. The flattened cortices were exposed to 4% paraformaldehyde-20% glycerol in 0.9% saline (2 h), 20% glycerol-2% dimethylsulfoxide in 0.9% saline (overnight), and 20% glycerol in 0.9% saline (24 h). Each flattened cortex was sectioned at 50 μm thickness on a freezing microtome chilled with dry ice. Individual sections were placed in 0.1 M PBS solution and maintained at 4°C.

Cytochrome oxidase staining. Cytochrome oxidase (CO) staining was used in selected sections to aid in the delineation of cortical sensory areas. Sections were placed in 0.1 M PBS solution and allowed to float. Then, sections were inspected with the unaided eye for the S1 representation, which is visible as several slightly opaque white areas within the translucent section. Sections with the most complete representations (3 to 4 sections/cortex) were chosen for CO staining. After rinsing (2 x 10 min in 0.1 M PBS), floating sections were reacted with CO solution at 37°C containing cytochrome c oxidase (20 mg, Sigma, #C2506-500MG), sucrose (4 g, Fisher Scientific), and 3,3’ Diaminobenzidine (50 mg, DAB, MP Biomedicals, Solon, OH, #980681) per 100 mL of 0.1 M phosphate buffered distilled water (pH 7.4). Sections were allowed to react for ~ 2–3 h until dark CO-rich areas were easily detectable against the lighter background. The sections were then rinsed (2 x 10 min) in 0.1 M PBS.

BDA10kDa visualization. All sections underwent a standard staining procedure using Avidin-Biotin Complex (ABC) linked to peroxidase with DAB reaction product as the chromogen. Sections were rinsed in 0.1 M PBS (2 x 10 min with agitation), then exposed to 0.4% Triton X-100 (Sigma, #X100-500ML) in 0.05 M PBS (1 h with agitation). Sections were rinsed in 0.1 M PBS (3 x 10 min, with agitation). Sections were incubated overnight in 0.1 M PBS with reagents “A” and “B” added according to Vectastain Elite Kit (Vector Laboratories, Burlingame, CA, #PK6100). Then, sections were rinsed (4 x 10 min, 0.1 M PBS), and exposed to DAB solution (0.05% w/v DAB and 0.01% v/v H₂O₂ in 0.1 M PBS). Sections were wet mounted in 0.05 M PBS onto subbed slides and allowed to dry overnight.

BDA10kDa signal intensification. Sections on slides were dehydrated in ascending alcohol concentrations (50%, 70%, 95%, and 100% for 5 min each), cleared with xylene (5 min), then rehydrated by reverse order of alcohol concentrations. Sections were exposed to 1.42% silver nitrate in distilled water (55°C, 1 h), rinsed (15 min, distilled water), exposed to 0.2% gold chloride (10 min), rinsed (15 min, distilled water), exposed to sodium thiosulfate (5 min), and rinsed (15 min, distilled water). Finally, sections were dehydrated again, as described above, cleared in xylene, and coverslipped with DPX mounting medium (Sigma, #44581-500ML).

Distribution of BDA-labeled boutons and somata

Regions of interest. The nomenclature, location, and description of the regions of interest (ROIs) utilized in this study were derived from various sources to achieve the most accurate and reliable description. Regions were identified using a variety of criteria including, response to ICMS, CO staining, obvious clustering of voxels, or spatial relationship to other regions, and are consistent with previously reported nomenclature (Donoghue and Wise 1982; Neafsey and Sievert 1982; Chapin and Lin 1984; Sanderson et al. 1984; Zilles 1985; Koralek et al. 1990; Fabri and Burton 1991; Harley and Bielajew 1992; Remple et al. 2003; Polley et al. 2007; Reep and Corwin 2009; Nishibe et al. 2010; Krubitzer et al. 2011; Tennant et al. 2011).

Section outlines were traced with the aid of a computerized microscope with motorized stage (Axiophot 2, Zeiss) and stereology program (StereoInvestigator, Microbrightfield). Because of the large number of ROIs, we used a semi-quantitative method previously developed for a survey of premotor cortex connections in squirrel monkey (Dancause et al. 2006a). This method provided an initial overview of cortical regions with connections to RFA so that unbiased quantitative stereological procedures could be applied to specific ROIs. A 100 μm square grid was overlaid on the flattened cortical outline. As the section thickness was 50 μm, voxels with dimensions 100 x 100 x 50 μm were examined systematically throughout the section. Acronyms used for the ROIs are defined in Table 1.

Table 1.

Abbreviations.

AGm	Medial agranular cortex	IN	Insular area, inclusive
ALBSF	Anterolateral barrel subfield of S1	INc	Insular area, caudal portion
Aud	Auditory cortex	INr	Insular area, rostral portion
BDA	Biotinylated dextran amine	M1	Primary motor cortex
CAA	Caudal association area	PirOl	Piriform and olfactory cortex
CFA	Caudal forelimb area	PMBSF	Posteromedial barrel subfield of S1
CFAc	Caudal forelimb area, caudal portion	PPCl	Posterior parietal cortex, lateral aspect
CFAov	Caudal forelimb area, overlap zone with S1	PPCm	Posterior parietal cortex, medial aspect
CFAr	Caudal forelimb area, rostral portion	PR	Parietal rhinal cortex
CO	Cytochrome oxidase	PRh	Perirhinal cortex
CTB	Cholera toxin beta subunit	PV	Parietal ventral area
DAB	Diaminobenzidine	RFA	Rostral forelimb area
DZ	Dysgranular zone of S1	RS	Retrosplenial area
DZo	Dysgranular zone of S1, other	S1	Primary somatosensory area
FL-ICMS	Forelimb ICMS area	S1FP	Forepaw area of S1
FR	Frontal cortical area	S1FPo	Forepaw area of S1, other
FRo	Frontal cortical area, other	S2	Second somatosensory area
GZ	Granular zone of S1	To	Temporal/occipital cortex
GZo	Granular zone of S1, other	TP	Temporal posterior cortex
ICMS	Intracortical microstimulation	Vis	Visual cortex

Axons labeled with BDA10kDa appeared as dark lines with varicosities or boutons (Fig. 1E). A bouton was defined as a dark (chromogen dense), round object about twice as wide as the thin dark fiber on either side of it (en passant bouton), or at the end of a thin projection off the main axon (terminal bouton). Boutons were counted and recorded semi-quantitatively within each 100 μm x 100 μm x 50 μm (section thickness) voxel. Voxels containing 2–30 boutons were marked with a blue dot (Fig. 1I), and voxels containing greater than 30 boutons were marked with a red dot (Fig. 1I).

BDA10kDa was used to optimize anterograde labeling of boutons for purposes of quantitative analysis. However, limited retrograde labeling occurs with BDA10kDa, and thus the distribution of BDA-labeled somata was described also. BDA-labeled neuronal somata were plotted and counted in the same section as boutons. A neuronal soma was defined as a uniformly dark, smooth-edged shape with evidence of at least one thin projection emanating from it (Dancause et al. 2006a) (Fig. 1D). BDA-labeled somata were marked with a green dot (Fig. 1J).

Within a given region, the density of boutons was greatest in deeper cortical laminae and least in superficial laminae. However, the distribution of BDA-labeled boutons and somata was similar throughout cortical depths, as reported in previous connectivity studies in rat (Reep et al. 1987; Rouiller et al. 1993) and squirrel monkey (Dancause et al. 2006a, 2006b). The current experiment used a superficial layer section to delineate between ROIs more easily, as deep layers have a larger more diffuse projection pattern. Thus, although variations in overall density exist from superficial to deep laminae, it was deemed that a single section was sufficient to qualitatively describe a particular animal’s bouton and soma distribution. A quantitative analysis of bouton distribution used sections throughout the cortical laminae.

Alignment procedure. Fiducial markers were used to align the ICMS maps from the surgical procedure with the section outlines, voxel counts, neuronal soma counts, and CO-rich zones drawn in StereoInvestigator. The location of 2 injections of CTB647 and 1 injection of BDA10kDa injection cores were marked on all ICMS maps and section outlines. Section outlines and ICMS maps were overlaid in Photoshop, and the ICMS map was scaled and rotated until the 3 injection cores were in register. Numerous symbols used to designate voxels (red and blue dots) were present in the region of RFA and FL-ICMS. Careful attention was paid to the relationship between the borders of RFA and FL-ICMS to the symbols, which helped maintain correct proportion while transferring the RFA and FL-ICMS outlines into StereoInvestigator.

Since this method provided a 2D overview of the distribution, we then measured the cortical surface area for each ROI to obtain the number of voxels with labeled boutons per square mm of cortical surface (Supplemental Table 1). The areal representations of the ROIs (as determined by alignment of CO-stained sections, ICMS maps and voxel clusters) were outlined in StereoInvestigator, and imported into a graphics software program (Illustrator, Adobe). The area (mm²) of each region was then measured with Image J (NIH).

Extent of BDA injections

In the flattened, tangentially sectioned cortex, the BDA injection core was defined as the area of dark, uniformly speckled staining with little identifiable cellular structure (Dancause et al. 2006a, 2006b; Dancause et al. 2007). The halo was identified as a larger area of dark staining extending beyond the core. While indistinguishable at low magnification, at high magnification, many BDA-labeled axons could be observed immediately around the injection core (Fig. 1C). This region was defined as the halo. The injection core identified in 9 BDA-labeled sections in each animal was aligned using fiducial landmarks and superimposed in the neurophysiological data. In each of the 6 cases, the BDA Injection core was within the borders of RFA as defined by ICMS (Fig. 4).

Fig. 4.

Extent of BDA injection cores relative to ICMS-defined RFA extent. In A and C–G, individual cases are illustrated. For each case, each tangential section was aligned and the vertically aligned BDA injection cores. Thus, darker areas represent a greater degree of overlap with other sections in the case. The injection core outline from the section used for bouton (voxel) counting is outlined in blue. Note that the BDA injection cores are largely confined to the ICMS-defined CFA territory. B. Photomicrograph of representative BDA injection core (black line). See methods for definition of core. All sections are scaled to the scale bar shown in B.

Unbiased stereological analysis of BDA-labeled boutons

After visual inspection of the sectioned tissue, sections 8 through 24 were determined to be consistently intact in all animals and were used for unbiased stereological analysis. Every other section (9 sections per animal) was used for determining BDA injection core size and performing stereological estimations of bouton counts. Both BDA and CTB647 (Fig. 1F) injection cores were outlined on sections. The section outlines and the ICMS maps were aligned using BDA cores as fiducial markers.

Terminal bouton quantification: unbiased stereology. Estimations of terminal boutons in each ROI identified in the distribution of voxels were performed using the stereological probe, the Optical Fractionator (StereoInvestigator, MBF Bioscience, Colchester, VT), according to methods described previously (Gundersen and Jensen 1987; Gundersen et al. 1988). Briefly, the Optical Fractionator is a 3-dimensional stereological probe that estimates population sizes by counting objects with optical disectors in a random, systematic sample within a volume of tissue located within an ROI (West et al. 1991). The counting target consisted of BDA-labeled terminal boutons (varicosity at the end of a thin projection attached to the main axon) and boutons en passant (varicosity approximately twice as wide as the axon on either side of it). At 100x magnification, the counting frame was a 14 x 14 μm square with 3 μm top and bottom guard zones. The systematic random sampling grid layout was set to distances that would give 15–30 counting frames per ROI. Once probe parameters were established for a particular ROI in an animal, they were constant through all sections of that animal. The parameters were adjusted for different animals’ ROIs to maintain a similar number of counting frames per ROI throughout the study. Tissue thickness was measured every other counting field. To be able to compare bouton estimates between ROIs of different volumes, bouton estimates were normalized by ROI volume for each animal to give an estimate of bouton density for each ROI (bouton density = bouton estimate/ROI volume). Since BDA10kDa was used, an anterograde tracer that optimizes bouton labeling but provides some retrograde somata labeling, stereological analysis was not performed on labeled somata.

Calculation of cortical volume for ROIs. The areal representations of the ROIs (as determined by CO-stained sections, ICMS maps, and voxel clusters) were outlined in StereoInvestigator on every tissue section analyzed. ROI volume (mm³) was calculated from the outline area and tissue thickness. The primary metric of interest for statistical purposes was the bouton density (number of BDA-labeled boutons per mm³) for each ROI.

Statistical analysis

For each ROI in each animal, bouton density values (boutons/mm³) were calculated. A Wilcoxon rank sum test was conducted for each pair of ROIs to test if the bouton density values of the 2 ROIs had similar distributions or not. The smaller the P value, the more different the 2 distributions were. A greater P value indicated that the distributions were more similar, an indication of connections between the 2 ROIs, as the bouton density values were distributed similarly between the 2 ROIs. To define categories for dense, moderate, and negligible connectivity with RFA, boxplots of the natural log of bouton density values were graphed for each of the 23 ROIs. In each boxplot, Q1 (25% quantile), median (50% quantile), Q3 (75% quantile), “minimum” (Q1–1.5*IQR), and “maximum” (Q3 + 1.5*IQR) were marked at the tip of the bottom whisker, the bottom edge of the box, midline in the box, top edge of the box, and the tip of the top whisker, respectively, where IQR = Q3-Q1. Dots outside the “minimum”-to-"maximum” range were considered outliers. The 3 connectivity categories were then assigned to ROIs that clustered into 3 “nearly nonoverlapping” groups.

Results

In each of the 6 animals used for analysis, BDA-labeling of terminal boutons and neuronal somata was clearly visible in various regions of the cortex. In the sections that follow, first the ROIs are defined. Next, the spread of the BDA injection cores is compared with the neurophysiologically defined RFA borders. Then, the distribution of BDA-labeled terminal boutons and neuronal somata is described with respect to the ROIs based on voxel distributions. Finally, the quantitative bouton density results based on stereological analysis are described.

Identification of ROIs

A total of 23 ROIs were identified based on a combination of features, including ICMS mapping data, cytochrome oxidase labeling, clustering of BDA labeling, relative topographical location, and neurophysiological/neuroanatomical overlap. The tangential sectioning of the cortex aided in the alignment of the multiple data sets, and ultimately in the standardization of nomenclature across the sample of animals. Since some of the 23 ROIs overlapped, broader designations (e.g. FRo, GZo, DZo and S1FPo) are used here to define mutually exclusive territories for quantitative analysis. For example, GZo refers to the cortical area corresponding to the granular zone of S1 (GZ) but excludes GZ cortex overlapping with CFA. GZo can be read “Granular Zone, other”.

Identification of CFA and RFA based on ICMS-evoked movements. ICMS maps of evoked movements were successfully obtained in each of the 6 animals used for analysis. Movements of the forelimb were evoked by ICMS within 2 clusters corresponding to the typical locations of the RFA and CFA using currents ≤ 60 μA. (Neafsey and Sievert 1982; Nishibe et al. 2010). Clusters of sites at which ICMS evoked forelimb movements were surrounded by sites at which ICMS evoked neck, jaw, orofacial, and vibrissae movements, or sites at which ICMS evoked no movement. The smaller cluster corresponding to RFA was located between +3.7 and + 2.7 mm anterior to Bregma and 2.0 to 4.0 mm lateral to the sagittal suture (Fig. 1A). The larger cluster corresponding to CFA primarily was located between +2.7 anterior and −1 mm posterior to Bregma and 2.0 to 4.5 mm lateral of the sagittal suture. In some cases, a narrow, caudal extension of the CFA movement map was observed. Since CFA was defined by the area from which forelimb movements could be evoked by ICMS currents ≤ 60 μA, we included this caudal extension as part of CFA.

ROIs identified by CO staining. CO staining revealed darkly stained zones within the granular cortex of flattened sections (Fig. 1G). These CO-rich zones were useful in identifying the primary sensory areas, S1, Vis, and Aud, as well as RS and TP cortex. CO staining was especially useful for identifying the prominent barrel field (PMBSF) within S1. S1 is a critical landmark in aligning data sets. These areas are consistent with Remple et al. (2003) and Chapin and Lin (1984).

Vis and TP are CO-dense regions coursing from the caudal edge of the cortex in a wide triangular shape. Both regions are thinnest at the rostral vertex. Vis is consistent with the pooled Oc1M and Oc1B (Zilles 1985). TP is the smaller triangular region directly lateral to and separated from Vis by a CO-sparse strip of cortex (Krubitzer et al. 2011). Aud is a large circular CO-dense region lateral and caudal to the S1 (Remple et al. 2003; Polley et al. 2007).

RS comprises a thin CO-dense region coursing along the medial edge of the caudal half of the flattened cortex (Harley and Bielajew 1992). The rostral aspect of RS stops at the caudal border of FR; the lateral border is concurrent with the CAA; the caudal aspect ends 1–2 mm from Vis; the medial border extends to the medial border of the cortex.

Identification of ROIs based on clusters of BDA-labeling. The locations of discrete clusters of voxels containing BDA-labeled terminal boutons superimposed on section outlines led to the identification of several ROIs. S2, PV, PR, PPCm, and PPCl, as defined here, are in accordance with Remple et al. (2003), and Fabri and Burton (1991), though PPCm and PPCl are sometimes called PM (parietal medial area) and PL (parietal lateral area), respectively (Koralek et al. 1990; Fabri and Burton 1991; Remple et al. 2003). Discrete dense clusters of labeling were identifiable lateral to S1, which spanned the area directly lateral to the caudal half of ALBSF and the rostral half of PMBSF. Proceeding from medial to lateral, S2, PV, and PR occupy the cortex lateral to S1, i.e. above the rhinal fissure, rostral to Aud and caudal to IN. These 3 areas each have generally the same dimensions, 1–3 mm in the mediolateral, and 2–3 mm in the rostrocaudal dimensions. As the rostrocaudal boundaries of labeled vs. unlabeled tissue were relatively sharp for these 3 areas, the extent of the regions was determined by the largest extent of the cluster of voxels with labeled boutons.

PPCm and PPCl were identified by small clusters of voxels with labeled boutons, which were located caudal to and caudolateral to S1, respectively. PPCm shares its rostral border with S1. The extent of the labeled area was ~1.2 mm anterior to posterior. PPCl shares a border with PPCm and S1 medially, S2 rostrally, and Aud laterally. This topographic pattern is similar to those reported in other published reports (Fabri and Burton 1991; Reep and Corwin 2009).

Identification of ROIs by topographic relationships to other identified regions. After aligning CO-rich regions with the distribution of BDA-labeling, additional regions were identified by their spatial relationship to other regions. FRo and IN were largely identified by their relationship to S1. FR occupies the rostral half of the medial cortex as well as the frontal pole, bordering S1 along its lateral caudal aspect. The definitions of FR and IN cortex follow Zilles (1985) with some modification. All of the Frontal regions of Zilles (Fr1, Fr2, and Fr3, and Cingulate cortex) were pooled into one region: FR. The term FRo is used here to distinguish the frontal rostral area that excludes RFA, the focus of the study. Zilles’ insular cortex (AID, AIV) was pooled with the rostral half of Vis to comprise IN in the present report. IN occupies the lateral aspect of the cortex and is bordered by FR and S1 at its medial aspect, shares its caudal border with S2, PV and PR, and the rhinal fissure laterally. The border between IN and FR was drawn from the rostromedial edge of S1 to a small consistent area of dense axons and boutons in the rostral pole. IN was further divided into rostral (INr) and caudal (INc) partitions with the border extending from the most lateral aspect of S1 to the rhinal fissure.

Identification of ROIs based on neurophysiological/neuroanatomical overlap. Some regions were identified through the overlap of ICMS mapping data with CO-rich areas. These overlapping zones were treated as separate regions in the quantitative analysis. Most importantly, several investigators have noted the overlap zone between CFA and S1FP, i.e. the forelimb portion of GZ (Sanderson et al. 1984; Tennant et al. 2011). Due to its potential functional importance in interpreting RFA connectivity, the CFA was divided into 3 partitions (Fig. 2). CFAr was designated as the cortex that includes the rostral portion of CFA, i.e. the portion of CFA anterior to the overlap zone with S1FP. Thus, CFAr overlaps with the rostral portion of DZ and the caudal portion of FR. CFAov is the CFA overlap zone with GZ, including S1FP. CFAc was designated as the cortex that includes the caudalmost portion of CFA, overlapping with a portion of DZ.

Fig. 2.

Overlap of CFA and S1. A. Caudal forelimb (motor) area as defined by ICMS. B. S1 subdivisions as defined by neurophysiological recordings and cytochrome oxidase staining. C. Location of cortical areas on flattened, tangential section. D. Enlarged view of ICMS-defined forelimb motor area and S1 subdivisions. Abbreviations: FL-ICMS = ICMS-defined caudal forelimb motor area; CFAr = rostral portion of CFA; CFAov = overlap zone between CFA and S1FP; CFAc = caudal portion of CFA; S1FP = forepaw somatosensory area. Note that the entire forepaw somatosensory area in D comprises S1FP and CFAov.

With respect to the histologically defined S1, the size and location of motor representations in CFAr and CFAov were similar across animals. CFAr was located along ~2 mm of the rostrolateral border of S1FP, extending rostrally ~2 mm from the S1FP border. In all 6 animals, CFAov was co-extensive with S1FP, arranged in an oblique angle along the rostral aspect of S1, and extended more caudally into non-FP areas of GZ. Its dimensions were ~2 x 3 mm. In contrast, there was variability in the extreme caudal extent of CFA. In 4 of the 6 animals, ICMS at low current levels (<60 μA) evoked forelimb movements in regions caudal to GZ (Fig. 3B, C, E, and F). This region extended caudally as a narrow strip of 1 to 4 sites at which ICMS evoked forelimb movements. The caudalmost sites overlapped with the histologically defined DZ (Chapin and Lin 1984). CFAc did not overlap PMBSF in any of the 6 animals. In the 4 animals in which CFAc was identified, forelimb movements were evoked at a total of 15 sites. Of these 15 sites, wrist extension was evoked at 13 sites, while finger flexion was evoked at one site, and elbow extension at one site.

Fig. 3.

Overlap of CFA and S1 in individual cases. The thick gray lines delineate S1 subdivisions as defined by neurophysiological recordings and cytochrome oxidase staining.

RFA formed a segregated cluster of sites where ICMS evoked movements of the forelimb, as described in a previous section. RFA did not overlap with any of the CO-positive regions.

Qualitative distribution of BDA-labeled boutons

Despite the relatively restricted injection cores, in each of the 6 animals, large numbers of BDA-labeled boutons were visible throughout frontal, parietal, insular, and peri-rhinal cortex. Distributions were qualitatively similar in different animals, despite differences in total numbers of labeled boutons. As described in methods, each 100 x 100 x 50 μm voxel throughout a single section through the middle layer of cortex was examined and coded separately if there were 0–1 (no color code), 2–30 (blue), or >30 (red) boutons per voxel. The distribution of voxels with labeled boutons (hereafter simply called “voxels”) in the animal with the fewest voxels (R11–29; Fig. 5A-B), and the animal with the most voxels (R11–09; Fig. 5C-D), is shown in Fig. 5 relative to the ROI borders. Regardless of differences in total number, the relative distribution was similar between animals. Numerous red (>30 boutons/voxel) and blue (2–30 boutons/voxel) voxels were consistently found in FRo, RFA, CFAr, CFAov, DZo, INr, S2, PV, and PR. Voxels present within GZ were confined to the rostral border centered around CFAov. Smaller numbers of primarily blue voxels were found in GZo, PMBSF, PPCm, PPCl, PRh, and INc. Relatively few scattered blue voxels were found within Aud, PirOl, Vis, TP, and RS. Total labeled voxels in each region are shown in Supplemental Table 1.

Fig. 5.

Spatial distribution of voxels with labeled boutons and labeled somata following BDA injection in RFA. A and B illustrate case R11–29, the case with the least number of labeled voxels; C and D illustrate case R11–09, the case with the greatest number of labeled voxels. Boutons (voxels) are shown in A, and somata in B. BDA injection core is shown as a filled black region within RFA. CTB647 injection site is shown as smaller black regions. Regions of unusually high bouton density are enclosed by green lines. A single, 50 μm thick, tangential section in the middle layers of cortex is shown. In case R11–29 (A), large numbers of voxels with >30 BDA-labeled boutons (red dots) were observed in FR, S2, PV, PR, DZo, GZo, PM, rostral IN, and PRh. Due to the 10 kDa BDA used in these studies, relatively few BDA-labeled somata were observed in this case. In case R11–09 (C), large numbers of voxels with >30 BDA-labeled boutons (red dots) were observed in FRo, CFAr, the rostral part of CFAov, S2, PV, PR, DZ, extreme rostral PMBSF, and rostral IN. More labeled somata were observed in this case (D), but the distribution was similar to that seen in (B). Qualitatively, the distribution was similar across cases.

Due to the perspective of the tangential sectioning, a clear distinction was found in S1. The caudal half of DZo was coextensive with an area of numerous voxels within the center of S1. The remainder of S1 (GZo, PMBSF, and S1FPo) contained relatively few voxels. This pattern was evident in each of the 6 rats.

Qualitative description of BDA-labeled somata

The distribution of retrogradely labeled somata paralleled the pattern of voxel distribution (Figs. 1 and 5). Sparsely distributed, labeled neuronal somata were distributed throughout the section with minimal clustering. Numerous labeled somata were found within RFA, FRo, CFAr, CFAOv, and DZo. Relatively smaller numbers of labeled somata were found in PMBSF, GZo, S1FPo, S2, PV, PR, PPCm, CFAc, and IN. Labeled somata were rare in PPCl, RS, PirOl, CAA, and Vis, and were absent in PRh, Aud, To, and TP. Total labeled somata for each region and a ranking of somata densities normalized to the region area are shown in Supplemental Tables 1 and 2, respectively. Because of the sparse soma labeling using the anterograde tracer BDA10kDa, unbiased stereological analysis was not performed to quantify somata.

Quantitative description of BDA-labeled boutons

A Wilcoxon rank sum test was conducted for each pair of ROIs to determine whether the bouton density (estimated bouton count/ROI volume) of the 2 ROIs had similar distributions or not. Figure 6 displays the P values from pairwise Wilcoxon rank sum tests for each ROI. The smaller the P value, the more different the 2 distributions are. A greater P value indicates that the distributions are more similar, indicating a similar density of synaptic boutons between the 2 ROIs. For example, the rank test P value between Aud and CAA is 1, which means that their distributions are almost the same. This is a strong indicator that the 2 ROIs have a similar density of synaptic boutons and therefore similar connectivity with RFA (both have negligible connectivity). In contrast, the rank test P value between CFAr and INr is 0, which means that their distributions are quite different. This indicates the 2 ROIs have very different densities of synaptic boutons, and therefore different connectivity with RFA (CFAr has dense connectivity and INr has moderate connectivity). Based on this analysis, the ROIs were categorized as having dense, moderate, or sparse connectivity with RFA (Fig. 6, black dashed lines). FRo and CFAr had dense connectivity. INr, PR, DZo, CFAc, S2, PV, CFAov, S1FPo, INc, PMBSF, PRh, GZo, PPCm, and PPCl had moderate connectivity. TP, RS, To, PirOl, Vis, Aud, and CAA had sparse connectivity. The cluster of ROIs with moderate connectivity was further divided into 2 subclusters: moderate-1 (INr, PR, DZo, CFAc, S2, PV, CFAov, and S1FPo) and moderate-2 (INc, PMBSF, PRh, GZo, PPCm, and PPCl) (Fig. 6, blue dashed lines). The bouton densities of the ROIs within each of the moderate subclusters are more similar to each other than the bouton densities of ROIs from the other moderate subcluster.

Fig. 6.

Wilcoxon rank sum test of bouton density. A Wilcoxon rank sum test was conducted for each pair of ROIs to test if the bouton density of the two ROIs was similar or not. The rank test P value for each pair of ROIs was plotted with a color scale indicating its significance. The smaller the P value, the greater the difference between the two bouton density distributions. ROIs were categorized as having dense, moderate, or sparse connectivity with RFA (black dashed lines). The moderate cluster was further subdivided into moderate-1 and moderate-2 subclusters (blue dashed lines).

While the Wilcoxon rank sum test was used to indicate broad density categories, a few specific pair-wise comparisons are worth noting. For example, the bouton density of INr appeared qualitatively greater than INc. This was confirmed by the Wilcoxon test (P = 0.02) and their categorization of moderate-1 (INr) and moderate-2 (INc) connectivity. Similarly, the bouton density of DZo was greater than GZo, as confirmed by the Wilcoxon test (P = 0.03) and their categorization of moderate-1 and moderate-2 connectivity.

Bouton densities for each ROI (Table 2) were plotted on a log scale from highest to lowest mean density to depict non-overlapping groups of ROIs based on the Wilcoxon rank sum test (Fig. 7). Accordingly, each ROI was categorized as having dense, moderate, or sparse connectivity with RFA. The cluster of moderate connectivity ROIs was further subdivided into moderate-1 and moderate-2 subclusters. While the cluster of moderate connectivity was subdivided, the ROIs in this cluster are still more similar in bouton density distribution compared with the ROIs in the dense and sparse groups. Dense, moderate-1, moderate-2, and sparse connections are illustrated graphically in Fig. 8.

Table 2.

Bouton means and standard deviations.

	Volume (mm  3)		Total boutons		Bouton density Total boutons/Volume
Region	Mean	Std Dev	Mean	Std Dev	Mean	Std Dev
Aud	8.43	2.18	1102.74	1131.54	122.66	121.66
CAA	23.36	7.51	2089.43	1710.71	106.10	112.83
CFAc	0.41	0.20	10360.78	6130.08	25075.52	11748.59
CFAov	1.82	0.73	45572.05	48927.70	23185.12	15760.73
CFAr	3.18	0.48	367579.77	123307.09	120515.80	55447.59
DZo	5.24	1.17	96465.28	52427.39	19196.17	10241.52
FRo	17.29	2.87	2280748.21	682098.59	133952.44	45226.87
GZo	11.97	0.64	58666.76	43135.19	4913.41	3690.88
INc	7.00	1.82	71981.31	58062.57	10150.85	7440.99
INr	6.26	0.40	254126.84	114822.66	40147.90	16611.26
PiROl	33.24	11.38	3329.49	1595.84	108.80	67.04
PMBSF	5.63	1.06	39654.18	21912.17	7474.08	4455.35
PPCl	2.70	0.75	7787.62	4892.07	2838.33	1453.23
PPCm	3.60	0.50	29517.82	35626.19	7580.21	7681.39
PR	2.01	0.91	38119.10	20399.41	24129.68	16367.87
PRh	2.35	0.86	12688.78	7091.78	4938.50	2184.55
PV	4.28	0.35	120168.03	96856.04	27229.65	20638.13
RS	1.85	1.02	2906.63	6433.56	864.10	1666.17
S1FPo	0.88	0.52	10083.02	9619.63	11042.69	5658.95
S2	5.24	0.69	151651.75	125791.38	29006.56	23055.49
To	0.61	0.17	94.13	132.51	155.09	192.14
TP	2.15	0.60	339.07	249.58	156.95	101.93
Vis	6.75	1.44	277.55	366.56	41.80	49.94

Fig. 7.

Quantitative distribution of BDA-labeled boutons following BDA injection in RFA. Boxplots of the natural log of bouton density (boutons per ROI volume) were depicted for each of the 23 ROIs. It can be visualized that the distribution of bouton density values across the 23 ROIs can be clustered into 3 “nearly nonoverlapping” groups (black dashed lines) that were used to categorize ROIs with dense, moderate, or negligible connectivity with RFA. The moderate group was further subdivided into moderate-1 and moderate-2 clusters (blue dashed line).

Fig. 8.

Overview of RFA anterograde corticocortical connectivity in rat. The thickness of lines indicates whether the particular connection is dense (thick line), moderate-1 (medium line), moderate-2 (thin line), or sparse (*). Anterograde connectivity based on a Wilcoxon rank sum test of bouton density for each ROI.

Discussion

The purpose of this study was to provide a detailed, quantitative description of corticocortical connectivity of RFA in rats. These data are important due to the similarities in structure and function of RFA and premotor areas of primates, and the extensive use of both primate and rodent species in studies of sensorimotor function. Also, functional reorganization of RFA is frequently examined as a substrate for motor recovery after motor cortex lesions in rats. Using neuronal tract-tracing methods, we identified RFA connectivity with a number of distinct cortical areas. Using unbiased stereological analyses, we characterized RFA terminal projection patterns, quantified the RFA projections to each cortical area, and categorized RFA innervation to each area based on the relative density of RFA synaptic boutons (dense, moderate, or sparse connectivity). Using qualitative observations of retrograde somata labeling, we described patterns of reciprocal innervation to RFA from other cortical areas. The results of this study provide a more detailed understanding of the broader cortical network in which RFA participates and provide new insights into the functional role of RFA in motor behavior when considered in the context of the entire cortical network of sensorimotor communication.

Dense connectivity

RFA was found to have the densest terminal projections to 2 cortical regions, FRo and CFAr. FRo, as defined here, is the frontal cortex region rostral to CFA, but excluding RFA (the site of the tracer injection). While there is likely to be some bias for labeled terminals and somata in the area immediately surrounding RFA, simply due to the proximity to the injection site, clear projection patterns can be seen extending to territories quite medial and lateral to RFA (Fig. 5). RFA’s dense innervation and reciprocal connectivity (based on qualitative observations of retrograde somata labeling) with FRo suggest that, along with the posterior parietal cortex (PPCm and PPCl), RFA is a part of the cortical network, associated with directed attention in rats (Reep and Corwin 2009). CFA (divided in this study as CFAr, CFAov, and CFAc) is the area considered to be the forelimb representation of the rodent M1, in that forelimb movements can be evoked from this area with the lowest levels of ICMS current. CFAr, the rostral portion of CFA anterior to the overlap zone with S1FP, is a cytoarchitectonically distinct region that is coextensive with the lateral agranular field and is consistent with motor cortex with respect to large layer 5 pyramidal cells (Donoghue and Wise 1982). In this study, RFA had dense connectivity with CFAr. We also observed retrograde somata labeling indicating reciprocal connectivity. Previous studies examining CFA as a whole have reported similar results. Mohammed and Jain (2016) reported that the most dense ipsilateral inputs to RFA were from CFA followed by S1 and S2. The laminar distribution of the RFA–CFA connections differs substantially, however. Deeper layers in RFA project to CFA, while more superficial layers in CFA project to RFA (Rouiller et al. 1993). Based on these laminar connection patterns, the RFA projection to CFA has been described as a feedback connection, while the CFA projection to RFA has been described as a feedforward connection, consistent with 2 studies of RFA-CFA connections using photostimulation in mice (Hira et al. 2013) and voltage-sensitive dye in rats (Kunori and Takashima 2016). Additionally, a large number of corticospinal neurons originate from RFA, CFAr, and CFAov. (Neafsey and Sievert 1982; Nudo and Masterton 1990; Rouiller et al. 1993), reinforcing their role in motor behavior. In fact, a dense cluster of corticospinal neurons located in what is now called RFA was one of the original indications that a separate rostral motor representation exists in rat cortex (Hicks and D'Amato 1977). While RFA and CFA are strongly interconnected, contain large numbers of corticospinal neurons, and are responsive in similar ways to ICMS, suggestions have been made that they are specialized for different aspects of motor function. Reversible cooling experiments in RFA and CFA in rat have suggested that RFA is more involved in grasping and CFA in reaching (Hyland 1998; Brown and Teskey 2014). Also, while RFA neurons have similar basal spiking properties, time-course, amplitude, and direction preference during forelimb movements, RFA neurons are modulated to a greater extent based upon the behavioral context (Saiki et al. 2014). This is consistent with the notion that RFA is at a higher level of the motor hierarchy.

Moderate connectivity

RFA had moderate projections to 2 subdivisions of CFA: CFAov and CFAc. CFAov is the region of CFA that overlaps the granular somatosensory cortex (sometimes referred to as M1-S1 overlap zone), while CFAc, when present, overlaps the dysgranular cortex of S1. Additionally, retrograde somata labeling was observed in both motor areas indicating reciprocal connectivity with RFA.

RFA also had moderate projections to INr, S2, PV, PR, S1FPo, and DZo. Along with CFAov and CFAc, these areas comprised a subcluster of moderate connectivity (moderate subcluster-1) that was distinguishable from other areas with moderate connectivity. The vast majority of labeled terminals (and somata) in the insular cortex were located in the most rostral portion, INr, resulting in moderate innervation from RFA. Rouiller et al. (1993) also reported substantial connectivity between RFA and agranular insular cortex, i.e. the most rostral portion of IN. This connection with INr was one of the characteristics of RFA that distinguished it from CFA. Neurons in INr play a role in integrating somatosensory information with chemosensory information (Katz et al. 2001). It has further been suggested that such chemosensory–somatosensory processing is involved in perceptual valuation of food and reward (Baldo et al. 2016). Taken together with the present results, there is substantial evidence that RFA has direct access to the limbic system, deriving motivational aspects of environmental cues, and integrating them with motor commands. RFA displayed moderate projections to 2 multimodal somatosensory regions, S2 and PV, located in the lateral parietal cortex. In the rat, both S2 and PV are somatotopically organized, and each contains a full somatosensory representation of the body (Fabri and Burton 1991; Remple et al. 2003). We found 2 distinct clusters of labeled terminals, one in each region (Figs. 1I, 5A, 5C). While we did not explore the somatosensory representations in S2 and PV using neurophysiological mapping techniques, the densely connected zone appears to be located in more proximal body representations of both S2 and PV based the location of the labeled areas in their position relative to the auditory cortex (Koralek et al. 1990; Remple et al. 2003). This connectivity pattern may suggest the importance of modulation of the proximal somatosensory representations during forelimb movement. PR had moderate connectivity with RFA. Located lateral to S2 and PV, PR may relay visceral information based on its connections with brainstem and spinal cord (Cechetto and Saper 1987; Fabri and Burton 1991). Since this area is heavily interconnected to the S1 granular zones, it has been suggested that PR could play a role in fusion of internal and external body maps (Fabri and Burton 1991). The present results suggest that this network includes RFA as well. RFA innervation to the dysgranular cortex of S1, DZo, was also moderate. Other studies have reported that DZ also projects to the striatum (overlapping with GZ projections) and the thalamus (segregated from GZ inputs), while the caudal half of DZ receives deep proprioceptive information (Chapin et al. 1987), (Lee and Kim 2012). These results provide further evidence that RFA is involved in sensorimotor integration.

A second subcluster of areas with moderate terminal connectivity (moderate subcluster-2) included PMBSF, GZo, INc, PPCm, PPCl, and PRh. Bouton density in these areas was somewhat lower than in subcluster-1. As the granular cortex of S1 has been shown to receive cutaneous information (Chapin et al. 1987) and project to the thalamus and striatum (Lee and Kim 2012), this suggests RFA is less involved in sensorimotor integration in regard to cutaneous sensory information compared with proprioceptive information (DZo). RFA had moderate terminal connectivity with the posterior parietal cortex (PPCl and PPCm), an area thought to be part of a cortical network for directed attention in rats, while few retrogradely labeled somata were observed, consistent with results from a previous study (Reep and Corwin 2009). Similarly, RFA’s terminal connectivity with INc and PRh was moderate, while few retrogradely labeled somata were observed, consistent with previous studies (Rouiller et al. 1993; Burwell and Amaral 1998).

Sparse connectivity

The remaining areas (Aud, CAA, PirOl, RS, To, TP, and Vis) had sparse-to-negligible connectivity with RFA. Retrograde somata labeling was rare in CAA, PirOl, RS, and Vis. Labeled somata were absent in Aud, PRh, To, and TP.

Similarities between premotor connections in rodents and primates

The premotor cortex is classically defined as a region within the frontal cortex immediately anterior to M1, located within Brodmann’s area 6, and thought to represent the highest level of the motor hierarchy (Fulton 1935). Dum and Strick (2002) operationally defined premotor cortex as regions in the frontal lobe that project directly to motor cortex. Based on this definition, they demonstrated that in primate species, premotor cortex contains at least 6 (possibly up to 9) separate areas, including the SMA, the cingulate motor areas (CMA), dorsal premotor cortex (PMd), and ventral premotor cortex (PMv). Each of these premotor areas projects to the spinal cord via the corticospinal tract, and electrical stimulation of each area results in evoked movements. Thus, the RFA in rats meets these classical definitions of a premotor area. It is directly located anterior to M1 (i.e. CFA), it has dense reciprocal connections with M1, forelimb movements can be evoked with relatively low levels of ICMS current, and many of its layer 5 neurons project to the spinal cord.

From an evolutionary standpoint, a rodent homolog of the primate premotor cortex is unlikely. This would require that the areas were derived from a common ancestor. However, in a study of corticospinal neurons in 2 dozen mammalian species, Nudo and Masterton (1990) argued that neurologically more primitive (and extant) mammals that have more distant common ancestry with primates have only 2 cortical areas that originate corticospinal neurons, the combined M1/S1 region and the second somatosensory region. In fact, all mammals studied had this feature in common. A separate, dense cluster of corticospinal neurons anterior to M1 was found in all primate species (in the presumed PMv), all rodent species (in the presumed RFA), but no other mammalian Orders studied (except for rabbit, in the Order Lagomorpha, typically included with the Order Rodentia in the Superorder Glires; (Douzery and Huchon 2004)). Thus, Nudo and Masterton proposed that separate premotor areas emerged independently in Primates and Glires, i.e. after the divergence of the mammalian Orders. It is possible that similar selective pressure for higher order motor control existed in the 2 mammalian lineages, leading to parallel or convergent evolution of premotor cortex in primates and rodents.

A persistent question in motor neuroscience is whether RFA is most similar in structure and function to any particular primate premotor area (Nudo and Frost 2007). Alternatively, one could consider RFA an amalgam of primate motor areas, not yet evolutionarily differentiated into multiple subregions. Rouiller et al. (1993) noted the similarities between rat RFA and monkey SMA with respect to connections with insular cortex. However, SMA in primates contains a complete topographic map of the body, while the RFA representation in rats is dominated by the proximal and distal forelimb. Interestingly, while neural tract-tracing studies using the rabies virus have demonstrated hindlimb corticospinal projections from RFA (Strick et al. 2021), reports of hindlimb movements evoked from sites in this region are rare (Neafsey 1990).

Identical tract-tracing techniques and analytic approaches to those of this study have been used to quantitatively define the intracortical connections of PMv in squirrel monkeys (Dancause et al. 2006a, 2006b). Comparing rat RFA and squirrel monkey PMv connections reveals striking similarities (See fig. 11 in Dancause et al. 2006a). First, the cortical areas with the greatest connectivity with PMv are the rostro-lateral portion of M1 and frontal areas rostral to PMv; the connection pattern of rat RFA is similar. Cortical areas with moderate connectivity with PMv in squirrel monkey and RFA in rat include S2 and PV. Of course, there are areas connected to PMv that do not seem to have an equivalent counterpart in rats, including the additional premotor areas PMd, SMA, and CMA. However, it would seem that, in general, not only do RFA and PMv have similar connection patterns qualitatively, but the order of the hierarchy with respect to the strength of the connection is nearly identical.

The patterns of connectivity between RFA and PMv with S1 deserve special attention. In rats, RFA has moderate terminal connectivity with both the dysgranular (subcluster-1) and granular (subcluster-2) zones of S1. In contrast, in squirrel monkey, PMv has moderate terminal connectivity with area 3a (a dysgranular area), but negligible connectivity with area 3b (granular area). This difference may be due to the somewhat different arrangement of primary somatosensory areas in rodents, with granular and agranular regions more interdigitated.

One final similarity between RFA and PMv is the incomplete cortical representation of body movements. RFA is surrounded by face, neck, and trunk representations similar to PMv of primates. With rare exceptions, hindlimb movements cannot be evoked from RFA, PMv, or the immediately surrounding cortex.

One caveat regarding the similarities between RFA and PMv connectivity is that similar quantitative data do not yet exist for PMd, SMA, and CMA. Because all of the primate premotor areas are part of an interconnected network, similarities may also exist in connectivity patterns between RFA and these other primate premotor areas. Nonetheless, this study argues strongly for RFA as a premotor area with intracortical connections strikingly similar to those of premotor areas in primates.

Conclusion

This study provides the strongest evidence to date that RFA is a premotor area, much like premotor cortex of primates. Its connectivity patterns with primary motor cortex, primary somatosensory cortex, secondary somatosensory areas, and posterior parietal cortex are strikingly similar. The tangential sectioning approach reveals the distinct preference for RFA connections with dysgranular zones of S1, where proprioceptive information is processed. Along with its moderate connections with insular and entorhinal cortex, RFA appears to occupy a unique position in the motor hierarchy, integrating information regarding motivation and memory, as well as proprioceptive information from broad regions of the body, cutaneous information from the forelimb, and feedback from executive functions in the primary motor cortex to modulate motor behavior.

CRediT statement

Edward Urban, III (Data curation, Formal analysis, Investigation, Methodology, Validation, Writing—original draft), Heather M. Hudson (Formal analysis, Investigation, Methodology, Supervision, Validation, Writing—original draft, Writing—review & editing), Yanming Li (Data curation, Formal analysis, Validation, Visualization, Writing—review & editing), Mariko Nishibe (Investigation, Methodology, Visualization, Writing—review & editing), Scott Barbay (Formal analysis, Investigation, Supervision, Validation, Writing—review & editing), David Guggenmos (Formal analysis, Supervision, Validation, Writing—review & editing), Randolph Nudo (Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Supervision, Validation, Writing—original draft, Writing—review & editing).

Funding

This work was supported by the National Institutes of Health grant number R37NS030853 to RJN.

Conflict of interest statement: None declared.