To select a marker gene for Class-A or -B cells that could be used to construct a Cre-driver mouse line, we queried published single-nucleus RNA sequencing data of NeuN+ neurons in the mouse cerebellar nuclei [4] as well as of non-neuronal cerebellar cells in the Allen Brain Cell (ABC) Atlas [12]. Class-A and -B cells in the medial, interposed, and lateral cerebellar nuclei differentially express several genes, some specific to each nucleus, others specific to each class across all nuclei. Among these genes, we selected Acan as a promising marker for Class-B neurons. It is expressed at moderate levels in Class-B cells across all cerebellar nuclei, with no detectable expression in Class-A or inhibitory neurons and minimal expression in non-neuronal cells (Fig. 1B, C). Acan is thus a promising gene to drive recombinase expression specifically in Class-B cells.

The protein encoded by Acan, Aggrecan, is a major structural proteoglycan in cartilage, where it plays a critical role in skeletal development and resistance to mechanical load [13,14,15]. In the central nervous system, Aggrecan is a major component of perineuronal nets, extracellular matrix structures that regulate neuronal plasticity [16]. Manipulation of Acan expression can profoundly alter circuit dynamics, including reinstating juvenile-like plasticity in the adult visual cortex [16] and regulating synaptic plasticity in excitatory neurons of the hippocampal CA2 area [17]. While conditional Acan loss-of-function models have been developed, no faithful Cre driver line is available to selectively target Acan-expressing neuronal populations [18]. An Acan-creERT2 mouse line already exists (Jax stock #019148) [19] but unfortunately does not label cells in the cerebellar nuclei (data not shown), likely because of an unintended deletion in the Acan gene 3’UTR in this specific line [18]. We therefore set out to develop a new Acan-Cre driver line.

To drive Cre expression mimicking Acan expression levels without disrupting Acan expression or any regulatory elements, we chose to replace the endogenous Acan stop codon with a P2A-Cre cassette. To do so, we designed a CRISPR/Cas9 guide RNA targeting a sequence close to the endogenous stop codon of the Acan gene and an HDR donor including the P2A-Cre cassette flanked by 5′ and 3′ homology arms corresponding to the targeted locus (Fig. 1D, Supplemental Data S1). After injection into B6SJL oocytes and embryo transfer, we identified founder animals that successfully transmitted the modified allele to F1 progeny and backcrossed these animals onto a C57Bl6/J background for at least 5 generations. We confirmed correct integration of the knock-in cassette without inadvertent deletions into the endogenous locus by PCR genotyping and long-read sequencing (Fig. 1E).

Aggrecan is a principal cartilage extracellular-matrix proteoglycan, whose reduced function causes chondrodysplasia and short stature. Hemizygous Acan-P2A-Cre animals are overtly healthy and Acan expression levels in the cerebellar nuclei, including in Class-B neurons, are not statistically different from those in wild type animals (Supplemental Fig. 1). This indicates that in hemizygous animals, sufficient Aggrecan is expressed and this mouse line can be used to accurately recapitulate Acan expression in the mouse brain. However, homozygous mice exhibited a dwarf phenotype, suggesting that the Acan-P2A-Cre allele is slightly hypomorphic. We therefore exclusively maintained and used the Acan-Cre line as hemizygous.

To determine whether our new Acan-Cre line provides selective access to Class-B neurons, we performed stereotaxic injections of a Cre-dependent AAV (AAV2-CAG-FLEx-eGFP) into the cerebellar nuclei of hemizygous Acan-Cre adult mice and wild type control mice (Fig. 1F). As expected, we observed strong viral labeling of large neurons in all three cerebellar nuclei only in Acan-Cre animals (Fig. 1F). We then used BARseq3 spatial transcriptomics to determine the molecular identities of the Gfp-labeled cells [11]. In both wild type and Acan-Cre animals, we probed for Gfp mRNA and a panel of marker genes selected to distinguish major neuronal and non-neuronal populations of the cerebellar nuclei: excitatory neurons were identified based on expression of Slc17a6 (Vglut2), a pan-glutamatergic marker labeling both Class-A and Class-B populations. Within this population, Acan and Sv2c expression were used to identify Class-B neurons, while Fnbp1l served as a marker enriched in Class-A neurons. Inhibitory neurons were identified using Gad1, Slc6a1 (GABAergic marker), and Slc6a5 (glycinergic marker). The special population of rhombic-lip derived, large glycinergic cells [20] found in the medial nucleus was counted as Class-B neurons in accordance with [4]. Non-neuronal populations were identified using Gfap as an astrocytic marker, Tmem119 as a microglial marker and Pdgfra as an oligodendrocyte marker. This marker combination enabled robust classification of cerebellar nuclei cell types (Fig. 1G, H) and allowed us to determine which cell types were labeled in the viral injections.

As expected in the dense neuropil of the brain and from high viral expression levels, Gfp signal was detected not only within neuronal somata but also in axons and dendrites in virus injected samples, which may contaminate gene calling in spatial data. Therefore, we evaluated the robustness of viral labeling of specific cell types across different Gfp expression thresholds (Fig. 1I). Across all thresholds, Gfp detection remained strongly enriched in Class-B neurons, whereas Class-A, inhibitory, and non-neuronal populations showed minimal labeling. Unassigned cells, which have low quality gene expression data, also displayed low levels of Gfp signal and likely reflect low quality cells (Fig. 1I, J). Taken together, these results demonstrate that Acan-Cre provides selective genetic access to the Class-B population in the adult cerebellar nuclei.

With the Acan-Cre mouse line in hand, we then sought to contrast the brain-wide projection patterns of Class-A and -B neurons of the lateral nucleus. To this end, we injected Cre-dependent AAV (AAV2-CAG-FLEx-eGFP) into the lateral nucleus of either Acan-Cre or Slc17a6-Cre mice (Fig. 2A). As demonstrated above, viral injections into Acan-Cre animals label Class-B cells, whereas injections into Slc17a6-Cre mice label all excitatory cells and thus both Class-A and -B neurons. Injections were confined to the lateral nucleus, with no spill into extracerebellar regions and only very minor labeling of the lateral part of the interposed nucleus (Supplemental Fig. 2). Projections found in Slc17a6-Cre animals, therefore, represent the total excitatory output of the lateral nucleus, and projections found in Acan-Cre animals should be a subset of these projections. Conversely, any brain regions innervated only in Slc17a6-Cre animals are likely exclusively innervated by A-class cells. Indeed, we found that projections in Acan-Cre animals are a subset of regions innervated in Slc17a6-Cre animals (Fig. 2C-E). Several regions, including the superior colliculus and the anterior pretectal nucleus (APT), were robustly innervated in Slc17a6-Cre mice but showed minimal or no innervation in Acan-Cre mice (Fig. 2C, D). Similarly, projections to the posterior and lateral posterior thalamic nuclei (PO and LPMR) were prominent in Slc17a6-Cre mice but largely absent in Acan-Cre mice (Fig. 2C, E). In contrast, thalamic nuclei such as VM, VL, PC, and CM were comparably targeted in both lines, indicating partial overlap in projection targets. ZI was targeted by both Slc17a6 and Acan projections but was more prominent for the Slc17a6 projection. Interestingly, we often found that in brain regions labeled in both mouse lines, including VP, VL, and VM, Acan-Cre projections formed spatially restricted subdomains embedded within the broader projection fields observed in Slc17a6-Cre mice (Fig. 2C, leftmost column), suggesting functional segregation within these target regions.

Together, these results demonstrate that Class-B neurons of the lateral nucleus exhibit selective and spatially restricted projection patterns within the broader excitatory cerebellar output system and validate the usefulness of our new Acan-P2A-cre line in understanding cerebellar output.