Mapping and targeting of C1ql1-expressing cells in the mouse

The C1Q complement protein C1QL1 is highly conserved in mammals where it is expressed in various tissues including the brain. This secreted protein interacts with Brain-specific Angiogenesis Inhibitor 3, BAI3/ADGRB3, and controls synapse formation and maintenance. C1ql1 is expressed in the inferior olivary neurons that send projections to cerebellar Purkinje cells, but its expression in the rest of the brain is less documented. To map C1ql1 expression and enable the specific targeting of C1ql1-expressing cells, we generated a knockin mouse model expressing the Cre recombinase under the control of C1ql1 regulatory sequences. We characterized the capacity for Cre-driven recombination in the brain and mapped Cre expression in various neuron types using reporter mouse lines. Using an intersectional strategy with viral particle injections, we show that this mouse line can be used to target specific afferents of Purkinje cells. As C1ql1 is also expressed in other regions of the brain, as well as in other tissues such as adrenal glands and colon, our mouse model is a useful tool to target C1ql1-expressing cells in a broad variety of tissues.

with genOway).A transgene was designed to contain two flippase recognition target (FRT) flanking a neomycin coding sequence, followed by the exon 2 of the C1ql1 gene where an ires-cre CDS cassette is inserted after the C1ql1 stop codon.This transgene was knocked in the corresponding region on chromosome 11 by homologous recombination (Fig. 1A).The neomycin sequence was removed later by breeding the transgenic mice with a flippase-expressing mouse (Fig. 1A).Identification of the C1ql1 cre F0 transgenic mice containing the ires-cre CDS was subsequently performed by PCR (Fig. 1B).Mice that are homozygous for the C1ql1 cre alleles are viable, fertile, normal in size and do not show any gross physical or behavioral deficits (observations on two different litters of five animals each, at birth, weaning and adult stages).
To visualize and characterize the driver capacity of the C1ql1 cre mouse line in the brain, homozygous C1ql-1 cre/cre mice were crossed with the Cre-dependent reporter mouse line R26 cas9,-GFP/cas9,-GFP8 (Fig. 1C).In this reporter mouse, Cre-mediated recombination allows the removal of the Stop cassette before the sequence coding cas9-P2A-EGFP.Due to the ubiquitous activity of the Rosa26 locus (R26) and the use of the CAG promoter, Cas9 and soluble GFP are expressed permanently in all the cells that have expressed Cre and their progeny.This expression history was mapped in C1ql1 cre/WT /R26 cas9,-GFP/WT heterozygotes at P30 using immunofluorescence on brain sagittal sections with an anti-GFP antibody to increase the signal to noise ratio (Fig. 1D).The GFP-positive cells were observed in different regions of the forebrain, midbrain, brainstem and cerebellum (Fig. 1D).At low magnification, two GFP-expressing regions were identified distinctly: the brainstem and the cerebellum (Fig. 1D, a).In the brainstem, GFP expression was strong in the inferior olive (Fig. 1D, a), in concordance with previous results showing expression of C1ql1 in inferior olivary neurons (IONs) from early development until adult age 3,4 .In the cerebellum, many granule cells (GCs) expressed GFP in a heterogeneous manner and in an antero-posterior gradient, with highest GFP signal in the anterior lobules.Based on in situ hybridization data 3 , only cells in the external granular layer, not the mature GCs, express C1ql1 during the first two postnatal weeks.This suggests that Cre-dependent recombination happens at the progenitor level in the granule cell lineage.Altogether, these results show that the C1ql1 cre mouse line is a powerful tool to genetically target two components of the olivocerebellar system, the IONs and the GCs.
At higher magnification, other GFP-expressing regions were identified in the brain of C1ql1 cre/WT /R26 cas9,-GFP/WT mice (Fig. 1D, a′-d′).According to the Allen Reference Brain atlas (https:// mouse.brain-map.org/ static/ atlas), we found GFP-expressing cells in several motor-related areas in the midbrain: ventral tegmental area, periaqueductal gray (Fig. 1D, a and a′), substantia nigra pars reticulata and red nucleus (Fig. 1D, b and b′).Adjacent to these regions, several GFP-positive cells were detected in two medial nuclei of the hypothalamus: supramammillary nucleus and medial mammillary nucleus (Fig. 1D, a and a′).In the pons, two nuclei were labelled with GFP: pontine reticular nucleus (motor related-area) (Fig. 1D, b and b′) and nucleus of the lateral lemniscus (sensoryrelated area) (Fig. 1D, d and d′).In addition to the IONs (Fig. 1D, a and a′′), two other motor-related nuclei were expressing GFP in the medulla: the paragigantocellular reticular nucleus (Fig. 1D, b and b′′) and vestibular nuclei (Fig. 1D, c and c′′).Expression of GFP is not restricted to the midbrain or hindbrain, but can be found also in the forebrain, in the accessory olfactory bulb and anterior olfactory nucleus (Fig. 1D, c and c′).The same pattern of expression was found when C1ql1 cre/cre mice were crossed with another reporter line, R26 R-EYFP/WT , where the enhanced YFP (EYFP) is expressed in cells following Cre-mediated excision of the Stop cassette preceding the EYFP sequence 9 (Fig. 2A and Supp Fig. 2), showing the reproducibility of neuronal targeting using the C1ql1 cre mouse line.Altogether, different nuclei and regions, especially motor-related areas in the midbrain, pons and brainstem, can be targeted genetically with the C1ql1 cre mouse model, in addition to the olivo-cerebellar network.
The expression of C1ql1 has been reported in several organs outside of the nervous system such as the adrenal glands and the colon using RT-PCR (Mouse ENCODE transcription data: https:// www.ncbi.nlm.nih.gov/ gene/ 23829).Using immunoblot analysis of protein extracts from different tissues of adult C1ql1 cre /R26 R-EYFP mice, the expression of the reporter EYFP was detected not only in the cerebellum, but also in the adrenal gland and the colon, two tissues that had C1ql1 expression reported in the Mouse ENCODE transcription database.In accordance with this database, no EYFP expression was found in extracts from the liver and the lung (Fig. 2A).These results show that the C1ql1 cre mouse model is not only useful for studies of the brain, but also in any other organ containing C1ql1-expressing cells.
Figure 1.Characterization of the driver capacity of the C1ql1 cre mouse line.(A) Illustration of the mouse C1ql1 locus at the 2 nd exon and flanking regions.Homologous recombination was used to insert a neomycin cassette upstream of the C1ql1 exon 2 and an ires-cre cassette downstream of the stop codon in the C1ql1 exon 2. After neomycin selection and mice generation, the neomycin cassette was removed using flippase to obtain C1ql1-ires-cre knockin mice (C1ql1 cre ).(B) Top panel: Table of the genotyping PCR amplicon bands and their respective sizes depending on the genotype.Bottom panel: Representative cropped image of a genotyping PCR run (original gel is presented in Supp Fig. 1).(C) Schematic illustration of the breeding strategy used to generate C1ql1 cre /R26 cas9,-GFP double heterozygous mice.(D) Medial to lateral sagittal sections from P30 C1ql1 cre /R26 cas9,-GFP mouse brain were immunostained for GFP at P30. Mosaic images were acquired using a Zeiss Axiozoom V16 macroscope (a-d).EGFP was highly expressed in the cerebellum (Cb), in particular in Granule Cells (GCs), and in the Brainstem (Bs), and present in other nuclei from the Midbrain (Mb) and Olfactory Bulb (OB).Scale bars = 1000 µm.(a′-d′) High magnification of different nuclei or neurons expressing EGFP in P30 C1ql1 cre /R26  90% 4,6 .We assessed by RT-qPCR whether the insertion of the transgene had any impact on the expression of the endogenous C1ql1 gene.At P14 in extracts from heterozygous animals, C1ql1 mRNA levels were decreased by 50-60% in all regions tested, compared to WT animals (Fig. 3A), showing that the insertion of the ires-cre CDS cassette in the 3'UTR of C1ql1 interferes with the proper expression of the C1ql1 allele.C1ql1 expression is thus expected to be reduced drastically in C1ql1 cre/cre homozygotes to levels that probably phenocopy the C1ql1 knockdown phenotype 4 (data not shown).To test whether the level of decrease in C1ql1 mRNA detected in C1ql-1 cre/WT heterozygotes leads to functional consequences, we performed morphological and functional analysis of CF/PC synapses comparing C1ql1 WT/WT and C1ql1 cre/WT heterozygotes.Morphological analyses were performed in two lobules of the cerebellum: lobules VI and IX, at P30 when the olivo-cerebellar circuit is established 10 , using immunofluorescence against vesicular glutamate transporter 2 (VGLUT2), to label the presynaptic boutons of CFs, and against calbindin (CaBP) to label PCs (Fig. 3B, left panel).Quantification showed no major difference in the thickness of the molecular layer of the cerebellar cortex and of the extension of CF innervation on the dendritic tree of PCs (Fig. 3B, right panel).While loss of function of C1ql1 at a level > 90% leads to decreased density of VGLUT2 clusters and decreased transmission 4,6 , our quantification did not detect any differences in the volume and density of VGLUT2 labelled presynaptic boutons between C1ql1 WT/WT and C1ql1 cre/WT heterozygotes (Fig. 3B, right panel).Patch-clamp recordings of CF/PC transmission during the fourth postnatal week showed similar amplitude of EPSCs in slices from C1ql1 WT/WT and C1ql1 cre/WT heterozygote mice (Fig. 3C).Altogether, the expression of cre and concomitant decrease of C1ql1 expression in C1ql1 cre/WT heterozygous animals do not impact morphological or electrophysiological properties of CF/PC synapses.The C1ql1 cre mouse line can thus be used to manipulate the olivo-cerebellar circuit when utilized as heterozygote without detrimental effect on normal development and function of the circuit.
Targeting different cell types in the olivo-cerebellar network using the C1ql1 cre mouse line The C1ql1 cre mouse line is expected to drive the expression of the Cre recombinase in one of the excitatory afferents of cerebellar PCs, the IONs, and in the precursors of the other excitatory afferent, the GCs.This targeting was indeed obtained using R26 cas9,-GFP (Fig. 1) and R26 R-EYFP (Fig. 2) reporter lines.The designer receptors exclusively activated by designer drugs (DREADD) are genetic tools that are widely used to probe neural circuits in the mouse brain.The hM4Di DREADD allows chemogenetic inhibition of activity in neurons that express it specifically when activated by the synthetic ligand Clozapine-N-Oxide 11 .In C1ql1 cre/WT /R26 hM4Di,-mCitrine/WT heterozygous mice, the hM4Di DREADD and soluble mCitrine are co-expressed in cells following the excision by the Cre of the Stop cassette preceding the hM4Di-P2A-mCitrine sequence 12 .Immunofluorescence labeling of brain sagittal sections at P30 using an anti-GFP antibody revealed expression of mCitrine in the inferior olive in the brainstem, with clear labeling of CFs in this region (Fig. 2B).mCitrine-positive GCs were also observed in the internal granular layer of the cerebellum (Fig. 2B).Of note, we observed in some instances mCitrine expression in Bergmann glia in the cerebellum when breeding a R26 hM4Di,-mCitrine/WT male with C1ql1 cre/cre females but not with the reverse cross (data not shown).This type of parental sex-related differences was not observed when breeding to the R26 cas9,-GFP reporter line, suggesting differences linked to the reporter mouse lines.
To specifically target the IONs using C1ql1 cre/WT heterozygous mice, we developed an intersectional strategy.A retrograde AAV driving the expression of a Cre-dependent VAMP2-GFP construct under the CamKII promoter was injected in the cerebellum of P0 mouse pups or P27 juvenile mice.VAMP2 is a vesicular SNARE abundantly present at presynaptic sites 13 .Thus, expression of VAMP2-GFP is expected to label presynaptic boutons made by IONs on their target neurons.As CaMKII is expressed in IONs but not in GCs, the conditional expression of VAMP2-GFP was specific to the inferior olive, both at P27 and P44 after P0 and P27 injections, respectively (Fig. 4A).While the breeding of C1ql1 cre with, for example, the reporter mouse line R26 cas9,-GFP led to labelling of the majority of ION population (Fig. 1D, a and a′′), only subpopulations of IONs are targeted using the intersectional strategy.Using immunofluorescent labeling of brain sagittal sections at P27 (Fig. 4B) or P44 (data not shown), with an anti-GFP antibody, GFP-positive CFs were observed in different lobules in the cerebellum, in particular in lobules V to X in the vermis or paravermis (Fig. 4B, top panel).At higher magnification, GFP-positive CFs were visualized along the proximal dendrites of PCs (Fig. 4B, bottom panel).No GFP-positive GCs were observed, showing the high specificity of expression in one of the two excitatory afferents of the PCs.Altogether, our results show that the C1ql1 cre mouse line can be used to either target both PC excitatory inputs or only one of them, depending on the strategy that is used.

Discussion
C1QL1 is a secreted protein involved in synapse formation and maintenance between inferior olivary neurons (IONs) and Purkinje cells (PCs) in the olivo-cerebellar circuit 4,6 .Since the expression of C1ql1 was identified in IONs and in the external granular layer during development, we developed a C1ql1 cre mouse model by inserting an ires-cre cassette after the stop codon of C1ql1, to enable the genetic manipulation of the two sources of excitatory inputs to PCs, IONs and GCs, in vivo.Using RT-qPCR, the expression of cre is detected in the cerebellum, brainstem, hippocampus and cortex.The specificity and capacity of the C1ql1 driver was characterized by crossing the C1ql1 cre mouse line with three Cre-dependent reporter mouse models: R26 cas9,-GFP , R26 R-EYFP and R26 hM4Di,-mCitrine .Expression of the reporter proteins is observed not only in IONs and GCs, but also in some other nuclei in the brainstem, midbrain and olfactory bulb, as well as in other organs than the brain such as the adrenal gland and colon.To specifically target IONs during postnatal development and adulthood, an intersectional strategy was developed via stereotaxic injection of a retrograde rAAV2 driving expression of a Cre-dependent reporter gene in the cortex of the cerebellum at P0 or P27, resulting in the transduction of numerous IONs and CFs 2-4 weeks after the injection.In general, the C1ql1 cre mouse model is a powerful genetic tool to target and  manipulate C1ql1-expressing cells in the brain, in particular IONs in the brainstem, and in other organs such as the adrenal gland.
Using the Cre-dependent reporter mouse models to map the location of C1ql1-expressing cells, we show the localization of the reporter gene in several nuclei in different regions of the brain.These nuclei are involved in different functions (e.g.olfaction, arousal and consciousness), but some of them found in the brainstem, cerebellum, pons and midbrain, such as inferior olive, red nucleus, substantia nigra pars reticulata and GCs, are motor-related structures.Several studies show the expression of C1ql1 outside of the brain, in particular in adipose tissue 14 , the adrenal glands, colon (https:// www.ncbi.nlm.nih.gov/ gene/ 23829), and ovaries 15 .Our results confirm the activity of Cre recombinase in several of these organs (the adrenal gland and colon) in the C1ql1 cre mouse line.
The function of C1QL1 in ovarian folliculogenesis has been investigated in mice using a full knockout model, showing its anti-apoptotic effect on granulosa cell apoptosis 15 .Loss of function of C1QL1 using a C1ql1 constitutive knockout mouse model results in abnormal hair cell innervation in the cochlea and loss of audition 16 .In addition to physiological conditions, several studies suggest that C1QL1 might be involved in some malignancies.C1QL1 is upregulated in colorectal cancer, one of the most common malignant cancers 17 , differentiated thyroid carcinoma 18 , as well as lung adenocarcinoma 19 .RNA-seq analyses have shown that C1ql1 expression is higher in glioblastoma compared to the normal brain 17 .All these studies show the diverse functions of C1QL1 in health and disease in different systems of the body.Thus, the C1ql1 cre mouse is a potentially useful model to study the role of C1ql1-expressing cells in different physiological and pathological conditions.

Methods and materials
All the methods were performed in accordance with the relevant guidelines and regulations.

Viral particles and Stereotaxic injections
Injections of rAAV particles were performed in the cerebellar vermis of ice-anesthetized P0 C1ql1 cre/WT heterozygous mice or isoflurane-anesthetized P27 C1ql1 cre/WT heterozygous mice, at 1 mm depth from the skull and 3.2 mm relative to Bregma or at 1.5 mm depth from drilled skull and 2.7 mm from Lambda respectively, to target Purkinje cell layer.rAAV2-retrograde serotype was used to target IONs.0.25 µl of AAV was injected per animal using pulled calibrated pipets.Retrograde rAAV-CamKII-DIO-mVAMP2/eGFP-WPRE viral particles allowing the Cre-dependent expression of mVAMP2-GFP under the control of the CamKII promoter were produced by Vector Biolabs: # AAV-275950 (RefSeq # BC055105), with a titer of 3.2 × 10 13 GC/mL.

Generation of C1ql1-ires-cre (C1ql1 cre ) knockin mouse
C1ql1 cre mice were generated using homologous recombination (genOway).A targeting vector containing Neomycin positive-selection cassette flanked with two FRT sites, and exon 2 of C1ql1 gene where an ires-cre cassette was inserted downstream of the stop codon in this exon, was electroporated into ES cells.The homologous recombination in the targeted ES cell clones was validated by PCR and DNA sequencing.The recombined ES cells with the correct sequence were injected into blastocysts.These blastocysts were then implanted in pseudopregnant females.Chimerism rate was then assessed in the progeny by coat color markers comparison.Highly chimeric males (chimerism rate above 50%) were generated and crossed with C57BL/6J flippase deleter female mice to excise the Neomycin selection cassette.Sixteen heterozygous mice were screened by PCR followed by sequencing to ensure germline transmission.

Gene expression analysis
The cerebellum, brainstem, hippocampus and a piece of cortex were extracted from C1ql1 WT/WT or C1ql1 cre/WT P14 mice and frozen immediately in liquid nitrogen, and stored at -80C°.Using the Qiagen RNeasy mini kit (Qiagen, Venlo, Netherlands, #74104), the total RNA was purified according to manufacturer's instruction.cDNA was synthetized using 100 ng of total RNA and SuperScript™ VILO™ cDNA Synthesis Kit (Life Technologies, California, USA, #11754050) according to manufacturer's instruction.Quantitative PCR on cDNA samples was performed using FAM-labeled C1ql1 (Taqman Gene Expression Assay, ThermoFisher, Assay ID: Mm00657289_ m1) or cre (Taqman Gene Expression Assay, ThermoFisher, Assay ID: Mr00635245_cn) probes and VIC-labeled Rpl13a probe as a reference (Taqman Gene Expression Assay, ThermoFisher, Assay ID: Mm01612986-gH) and the TaqMan Universal Master Mix II, with UNG (Applied Biosystems, #4440038) according to manufacturer's instructions.Bio-Rad CFX Manager was used for data analysis.The relative expression of C1ql1 to Rpl13a or cre to Rpl13a was calculated based on the following formula: Relative expression = (2 (Cq ofC1ql1orcre -Cq ofRpl13a) ) × 100.

Immunofluorescence
Mice tissues were fixed using intracardiac perfusion of 4% PFA in PBS solution.Brains were extracted, post-fixed with the same solution at 4 °C for 2-4 h, and transferred to 30% sucrose/PBS for 48 h at 4 °C for cryoprotection.30 µm-thick sections were obtained using a freezing microtome, and kept in 0.02% NaN 3 in PBS solution at 4 °C until use.

Image acquisition and analysis
The mosaic images for global brain morphology (Fig. 1D and Supp Fig. 2) were obtained using a Zeiss Axiozoom V16 macroscope, equipped with a digital camera (AxioCam HRm) using a 160× (pixel size: 0.4037 µm), and reconstructed using the Zeiss Zen software.Images for VGLUT2 quantifications (Fig. 3B) were acquired using a Zeiss spinning-disk confocal CSU-W1 microscope with 63× oil objective (Z-plane step size: 0.19 µm).Images for GFP analyses were acquired using the same spinning-disk confocal microscope with 63× or 25× oil objectives (single plane, Figs. 2 and 4).The mosaic images for GFP visualization were reconstructed using the stitching program from metamorph software (Fig. 4B top panel).
VGLUT2 quantification (cluster density and volume) were performed as described previously 20 .All the images were normalized using the quantile-based normalization plugin in Fiji.The intensity distribution of the images has been normalized using 256 quantiles for each staining.The synaptic boutons were extracted from the background using the 3D Weka Segmentation plugin (https:// imagej.net/ Train able_ Weka_ Segme ntati on) after manual selection of signal and background samples.The Fiji built-in plugin 3D object counter was then used in order to count and measure every object (cluster of VGLUT2 positive signal).CF territory and ML thickness were measured manually using Fiji.All the steps were performed in blind condition.

Figure 2 .Figure 3 .
Figure 2. The C1ql1 cre mouse line drives expression in different cell types of the olivo-cerebellar network in the brain as well as other organs.(A) Top left panel: Schematic illustration of the crosses used to generateC1ql1 cre /R26 R-EYFP heterozygous mice.SA, Splice Acceptor.Bottom left panel: Whole tissue protein lysates from the cerebellum, liver, adrenal gland, lung and colon of adult control (C1ql1 WT /R26 R-EYFP , wild-type: WT) and heterozygote (C1ql1 cre /R26 R-EYFP , Het) mice were subjected to denaturing electrophoresis and analyzed by immunoblot for EYFP (top blot) and β-ACTIN (bottom blot) expression.EYFP was detected in the cerebellum, adrenal gland and colon of a heterozygote C1ql1 cre /R26 R-EYFP mouse (original blot is presented in Supp Fig.3).Right panel: parasagittal cerebellar sections from P28 C1ql1 cre /R26 R -EYFP mice were immunostained for EYFP using anti-GFP antibody.Cerebellar granule cells (GCs, green arrowheads) and axons from inferior olivary neurons (IONs, cyan arrowheads), the climbing fibers (CFs, pink arrowheads), were EYFP labelled in the cerebellum and brainstem regions, respectively.(B) Left panel: Schematic illustration of the crosses used to generate C1ql1 cre /R26 hM4Di,-mCitrine heterozygous mice.Right panel: parasagittal brain sections from P30 C1ql1 cre /R26 hM4Di,-mCitrine mice were immunostained for mCitrine using anti-GFP antibody and images were acquired using a spinning disk confocal microscope.Cerebellar granule cells (GCs, green arrowheads) and axons from inferior olivary neurons (IONs), the climbing fibers (CFs, pink arrowheads), were positive for mCitrine in the cerebellum and brainstem regions, respectively.Scale bars = 20 µm for GCs and 125 µm for IONs and CFs.

Figure 4 .
Figure 4. Specific targeting of inferior olivary neurons using an intersectional strategy.(A) Schematic illustration of the strategy to specifically express the mouse GFP-tagged vesicle-associated membrane protein 2 (mVAMP2) in IONs.P0 or P27 stereotaxic injections of retrograde rAAVs expressing mVAMP2-GFP, under the CamKII promoter and in a Cre-dependent manner, were performed in the cerebellum of C1ql1 cre mice.Analyses were done at P27 for P0 injections or P44 for P27 injections, in parasagittal cerebellar sections.Subpopulations of IONs were labelled after anti-GFP immunostaining.Inferior Olive (IO) is indicated by dotted lines.Scale bars = 200 µm.(B) Top panel: Vermal (left) and paravermal (right) sections from C1ql1 cre/WT P27 mouse injected at P0 and immunostained for GFP revealed VAMP2-GFP expressing CFs in different lobules of the cerebellum.Images were acquired using a spinning disk confocal microscope.Lobules are numbered with roman numerals.Scale bars = 500 µm.Bottom panel: CFs of transduced IONs and PCs from P27 C1ql1 cre/WT mouse injected at P0 were immunostained for GFP (green) and calbindin (CaBP) (magenta), respectively.Molecular Layer (ML) and Purkinje cell layer (PCL) are indicated on the merge image.Scale bars = 20 µm.

Normal morphology and transmission of climbing fiber/Purkinje cell synapses in C1ql1 cre/WT heterozygotes
respectively).C1QL1 loss of function leads to detectable phenotypes in climbing fiber/Purkinje cell (CF/PC) synapse formation and function in the olivo-cerebellar network when C1ql1 expression is reduced by more than