Enhanced and unified anatomical labeling for a common mouse brain atlas

Abstract

Anatomical atlases in standard coordinates are necessary for the interpretation and integration of research findings in a common spatial context. However, the two most-used mouse brain atlases, the Franklin-Paxinos (FP) and the common coordinate framework (CCF) from the Allen Institute for Brain Science, have accumulated inconsistencies in anatomical delineations and nomenclature, creating confusion among neuroscientists. To overcome these issues, we adopt here the FP labels into the CCF to merge the labels in the single atlas framework. We use cell type-specific transgenic mice and an MRI atlas to adjust and further segment our labels. Moreover, detailed segmentations are added to the dorsal striatum using cortico-striatal connectivity data. Lastly, we digitize our anatomical labels based on the Allen ontology, create a web-interface for visualization, and provide tools for comprehensive comparisons between the CCF and FP labels. Our open-source labels signify a key step towards a unified mouse brain atlas.

Fig. 1

Import and alignment of the FP labels onto the Allen CCF. a The Allen Common Coordinate Framework (CCF) that serves as base anatomical platform. A/P represent Bregma anterior/posterior coordinates. b Initial import of the Franklin-Paxinos (FP) vector labels into the Allen CCF. c Manual alignment based on anatomical features in the Allen CCF. Yellow arrows highlight distinct anatomical boundaries based on edges and white matter tracks. Red arrows indicate layer 4 in the somatosensory barrel cortex. d MRI images registered to the same CCF plane in (a). e Original FP based labels drawn in the MRI atlas registered to the Allen CCF. The lack of labels in the hypothalamic and amygdala regions are due to missing labels in the original MRI annotation. f Further adjustment of FP based anatomical delineation (white lines) based on the MRI labels. Scale bar = 2 mm

Fig. 2

Marker brains for further alignment of anatomical labels. a–o Examples of different marker brains registered to the Allen CCF that helped to align FP based labels in subregions as highlighted with yellow arrows. A/P represents Bregma anterior/posterior coordinates. a–c Chat-Cre:Ai75 brain to delineate (a) the basal forebrain structures including the nucleus of the horizontal limb of the diagonal band (arrow). It was also used to delineate (b) midbrain areas, including the laterodorsal tegmental nucleus (arrow1), the motor trigeminal nuclei (arrow 2), the lateral superior olive (arrow 3), and c the facial nucleus (arrow). d–f PV-Cre:H2B-GFP brain to delineate (d) the reticular nucleus (arrow), e the anterior pretectal nucleus (arrow 1), the substantia nigra, reticular part (arrow 2), and the retromamillary nucleus (arrow 3) as well as f the superficial gray layer superior colliculus (arrow 1), the ventral nucleus of the lateral lemniscus (arrow 2), and the reticulotegmental nucleus of the pons, pericentral part (arrow 3). g–i SST-Cre:H2B-GFP brain to delineate (g) the cerebral nuclei, such as the lateral septal nucleus, dorsal part (arrow 1) and the bed nuclei of the stria terminalis medial division posteromedial part (arrow 2), h the reticular nucleus (arrow 1) and the central amygdaloid nuclei (arrow 2), and i hypothalamic structures, such as the dorsomedial hypothalamic nuclei dorsal and ventral parts (arrow). j–l OTR:Ai14 brain to delineate (j) the dorsal endopiriform nucleus (arrow), k CA2 (arrow 1), the posteromedial cortical amygdala (arrow 2), and l the caudomedial entorhinal cortex (arrow1) as well as the postsubiculum (arrow 2). m–o Cortical layers defined by m Ntsr-Cre:Ai75 for layer 6, n Rbp4-Cre:Ai75 for layer 5, and o Cux2-Cre:Ai75 for layer 2/3. Scale bar = 2 mm

Fig. 3

Additional segmentations based on marker brains. a–t Examples of marker brains to further segment structures defined as a single structure in the FP label. Added segmentations are marked by yellow lines. a–d PV-Cre:H2B-GFP (a, b) and Cux2-Cre:Ai75 (c, d) marker brains were utilized to further segment ventral posteromedial nucleus of the thalamus (VPM) to dorsal and ventral parts (VPMd and VPMv, respectively). e–h OTR-Cre:Ai14 (e, f) and Ctgf-Cre:Ai75 (g, h) used to segment dorsal and ventral parts (PHnd and PHnv, respectively) of the posterior hypothalamic nucleus (PHn). i–l SST-Cre:H2B-GFP (i, j) and PV-Cre:H2B-GFP (k, l) used to segment laterodorsal tegmental nucleus, dorsal part (LDTg) into lateral and medial divisions (LDTg-dl and LDTg-dm, respectively). m–p Chat-Cre:Ai75 (m, n) and SST-Cre:H2B-GFP (o–p) used to segment Barrington nucleus (Bar) into dorsal and ventral parts (Bard and Barv, respectively). q–t SST-Cre:H2B-GFP (q, r) and PV-Cre:H2B-GFP (s, t) used to segment the medial vestibular nucleus, parvicellular part (MVp) to dorsal and ventral parts (MVpd and MVpv, respectively). See Supplementary Data 2 for full names of abbreviations. Scale bars in first and third columns = 2 mm, second and fourth columns = 300 µm

Fig. 4

Cortico-striatal projection based striatum segmentations. a–d Anterograde tracing datasets from different cortical domains registered into the Allen CCF. Scale bar = 1 mm. a for the anterior cingulate cortex (ACA), b for the primary somatosensory cortex (SSp), upper limb (ul), lower limb (ll), barrel field (bfd), and trunk (tr) area. c for the SSp, mouth (m) and secondary (s). d for the visceral (VISC) and the agranular insular cortex (AI). e 129 datasets clustered into 10 groups based on cortical input regions. Datasets in the same cluster have the same color. f Example of CP segmentation based on cortico-striatal projection patterns in CPi regions with 4 different community level segmentations. g–n Representative images of CP segmentations throughout several Bregma A/P (number in the left bottom of each figure) planes. imd: intermediate dorsal, m: medial, imv: intermediate ventral, vm: ventral medial, dt: dorsal tip, dl: dorsolateral, dm: dorsomedial, d: dorsal, vl: ventral lateral, v: ventral, cvl: central ventrolateral, cd: central dorsal, im: intermediate dorsal, GP: globus pallidus. Scale bar = 100 µm. Full name of abbreviations can be found in Supplementary Data 2

Fig. 5

Digitization of anatomical structures. a Example of our highly segmented FP based labels on the Allen CCF. Yellow arrows highlight the lateral subdivision of periaqueductal gray (PAG), Scale bar = 2 mm. b Exported delineation lines. c Digitization of labels with unique numerical ID for each anatomical structure. Different color of each structure pertains to different number. d SST-Cre:H2B-GFP showed distinct subregions in the PAG with different cell density level. Our labels (white font) divide the PAG into different subregions, as can be seen with the specific enrichment of SST neurons in the dorsolateral PAG (DLPAG) and the lateral PAG (LPAG, yellow arrow). Scale bar = 300 µm. e In contrast, the CCFv3 labels (color labels in the background) showed only 2 segmentations within the PAG (black font). f Hierarchical organization of anatomical labels based on the Allen ontology. Numerical IDs of individual structures assigned within parent structures for region-level and individual structure-level data analysis. For example, the PAG (shaded dark gray) is the parent structure of six subdivided structures (shaded light gray). Red font labels refer to structures further divided by the FP labels that are not present in the CCFv3 labels. Full name of abbreviations can be found in Supplementary Data 2

Fig. 6

Comparison between the Allen CCFv3, the ARA, and our labels. First column: Our highly segmented FP based labels on the Allen CCF. Scale bar = 2 mm, second column; our labels (white lines) with marker brain background, third column: comparison between our labels and the CCFv3 labels (colored background), fourth column: comparisons between our labels and the ARA labels (colored background). b–p Anatomical names in black and white are from the Allen CCF and our labels, respectively. b–d PV-Cre:H2B-GFP (b) to identify subregions in the zona incerta (ZI). Scale bar = 300 µm. Low in dorsal and high in ventral parts (ZID and ZIV, respectively) in our labels while the CCFv3 and the ARA labels have a single combined structure for ZI. f–h (f) Virtual overlay of Chat-Cre:Ai75 (red) and SST-Cre:H2B-GFP (green) to compare basal forebrain regions. Scale bar = 300 µm. g, h Our labels further segregate the single structure defined as the substantia innominate (SI, Allen) into the ventral pallidum (VP) and the extended amygdala (EA). Yellow arrow highlights the border between the basal forebrain and the hypothalamus. j–l Disagreeing borders between the somatosensory and the motor cortices. Yellow arrow highlights border between the somatosensory and motor cortices. j Virtual overlay of pseudo colored Cux2:Ai75 (L2/3, green), Rbp4:Ai75 (L5, magenta), Ntsr1:Ai75 (L6, yellow), and Ctgf:Ai75 (L6b, red). Scale bar = 200 µm. Note the lack of Cux2:Ai75 and Rbp4:Ai75 signal in layer 4 of the somatosensory cortex. n–p The BST is divided into several subregions in our labels compared to a single BST structure in the CCFv3 labels, despite the original ARA version with finer delineations for this structure. n SST-Cre:H2B-GFP used to identify subdivisions of BST. Scale bar = 300 µm. Note the higher degree of segmentations in our labels compared to the CCFv3 and the ARA (a–d, e–h, m–p), and inconsistencies in anatomical delineations of the same structures between the atlases (i–l). See Supplementary Data 2 for abbreviations