Visualizing an convergence of three mitochondrial molecule mRNAs for DRP1/MFN2/UCP4 on soma of cerebellar Purkinje cells by RNAscope

We know little about how mitochondrial dynamic are regulated in the Purkinje cells. To explore it, we rst set up a transgenic mice in which Purkinje cells expressed tdTomato in the cerebellum of Gad2-cre;ZsGreen-tdTomato/ mice. Secondly Double stainings veried tdTomato cells were Calbinin (CB)-positive Purkinje cells, but not colocalized either with astrocyte marker GFAP or with microglia marker Iba1. Thirdly, application of RNAscope in situ hybridization with the identication of mRNAs of mitochondrial fusion (Mfn2), calcium transporter (Mcu and Nclx) and uncoupling proteins (Ucp2 and Ucp4) were used onto Purkinje cells for specic spatial analysis. The RNAscope assay used a semi ‐ quantitative H scoring guideline to evaluate the staining results. The number of bins ranges from 0– 4 according to the ACD scoring system. Moreover, ACD scoring system was used to calculate the overall H scores of Dendritic Weighted Formula (DWF) and Soma Weighted Formula (SWF). Our data for the rst time demonstrated that Mfn2 mRNAs expression was evident in Prukinje cells with the H scores of DWF and SWF as 60 and 139, respectively. And few Ucp4 mRNAs expression was present in dendritic shafts of Prukinje cells with the H scores of DWF and SWF as 14 and 103, respectively. It should be noted that Mcu mRNAs, Nclx mRNAs, as well as Ucp2 mRNAs expression were only scattered on both soma and dendrites in Prukinje cells with the low H scores of DWF and SWF. Double RNAscope proling of mitochondrial molecules showed The data showed Mfn1 mRNAs are present only in the soma of the Purkinje cells, instead of processes. Double RNAscope showed almost none of dots of Drp1 mRNAs was co-localized with dots of Mcu mRNAs, as well as almost none of dots of Ucp2 mRNAs was co-localized with dots of Mfn2 mRNAs. All of these results show the mitochondrial Drp1/Mfn2/UCP4 convergence on the Purkinje cells. Finally, a major focus of present research will be to develop new and more specic molecules that tune the activity of the Purkinje cells activate or inactivate and to address diseases for which such druglike molecules may open therapeutic windows for Purkinje cells-related manipulation in the clinic. The molecular identication of drug targets, mechanism of action, and structural basis of their activity will crucially enable preclinical development.


Introduction
Purkinje cells are the sole output of the computational circuitry of the cerebellar cortex and provide the signals required for motor planning, execution and coordination in their rate of ring and pattern of activity (Louis ED and Faust PL, 2020). Malfunction of these neurons often results in cerebellar ataxia: a disorder characterized by poor balance, loss of posture and di culties in performing rapid coordinated movements (Pan MK et al., 2020;Kostadinov D et al., 2019). Purkinje cells re spontaneously with highly regular pacemaking and their relayed neurotransmission depend on more than 150,000 excitatory and inhibitory synapses ( Kostadinov D et al., 2019). So Purkinje cells have a particular dependence on precise control of mitochondrial dynamics.
Mitochondria are dynamic double-membrane-bound organelles that are associated with ATP generation, calcium regulation, and the biosynthesis of aminoacids, lipids, and nucleotides (Patron M et al., 2018).
Progress in the eld of mitochondrial biology in the past few years has shown that mitochondrial

Animals and mouse breeding
To examine the distribution pattern of mitochondrial moleculars in purkinje cells, we used cre-loxP system to generate a conditional knock-in mice with speci c uorescence in GABAergic purkinje cell.
The C57BL/6-Gad2 EM1(IRES-ICRE-PA)SMOC (named as Gad2-ires-Cre) mice were purchased from Institute of Model Zoology, Nanjing University with the Product Number of D000668, previously reported (Taniguchi H et al., 2011). It is well-known that GABA (g-aminobutyric acid) in mammals was synthesized by two glutamic acid decarboxylase (GAD) GAD67 and GAD65 (Fish KN et al., 2021). These two proteins were encoded by Gad1 and Gad2 genes. The Gad2-ires-Cre mice inserted an internal ribosome entry site (ires) and a Cre recombinase sequence at the 3' end of the Gad2 allele, so that all of GABAergic neurons contained Cre recombinase.
After crossing of these two mice, all GABAergic neurons in the newly generated transgenic  ZsGreen-tdTomato / mice would transfer from expressing green uorescence to expressing red uorescence protein due to Cre recombinase cutting loxP sites. Therefore, Purkinje cells, a typical GABAergic subneurons in the brain of Gad2-cre; ZsGreen-tdTomato / mice, could be easily identi ed under confocal microscopy. These mice were then used for the following immuno uorescence labeling and single RNAscope in situ hybridization.
It should be indicated that the normal adult C57BL/6 mice were used for double RNAscope in situ hybridizations.
All of the mice are kept in a barrier facility, and all animal experiments were conducted in accordance with the procedures approved by the Ethical Committee of the Air Force Medical University and followed the institutional guidelines for the use of laboratory animals..
Animals were housed at a constant 23 °C in a 12 h light/dark cycle (lights off at 20:00), with food and water available ad libitum. nts. The day of birth was considered as postnatal day 0 (P0).

Genotyping
Genotype was identi ed by PCR with genomic DNA obtained from the tails. The primers sequences and PCR programs were listed in Table 1.

Immuno uorescence labeling
To verify the reliability of red tdTomato uorescence within Purkinje cell in the generated Gad2cre;ZsGreen-tdTomato / mice, immuno uorescence labeling was conducted according to the methods described previously (Huang J et al., 2008) with minor modi cations.
Six eight-week male Gad2-cre;ZsGreen-tdTomato / mice were perfused transcardially with 0.1 M phosphate buffer (PB; pH 7.4) containing 4% paraformaldehyde. The whole cerebellum were obtained and post xed with the same xative for 4 h, placed in 30% (w/v) sucrose reliability solution in 0.05 M PB solution (PBS; pH 7.4) overnight at 4 °C, and cut into sagittally 30 mm thick sections on a freezing microtome. Then the cerebellum sections were mounted onto the slides and incubated in 0.01 M PBS supplemented with 3% hydrogen peroxide for 10 min to block endogenous peroxidase and then in a blocking buffer containing 5% BSA/10% normal goat serum/0.25% Triton X-100 for 60 min at room temperature to prevent nonspeci c staining. Following this, the sections were incubated in the blocking buffer for 60 min at room temperature and then in a solution containing primary antibodies of the marker of Purkinje cells (Calbindin, CB), the marker of granular cells (NeuN), the marker of astrocytes (GFAP), the marker of microglia (Iba1) ( Table 2) from different species simultaneously for 18 h at 4 °C. After washing, sections were incubated with appropriate secondary antibodies (Table 2) for 2 h at room temperature.
It should be indicated that in the cerebellum of Gad2-cre;ZsGreen-tdTomato / mice, Purkinje cells expressed bright tdTomato uorescence, while non-GABAergic expressed bright ZsGreen uorescence. So we observed the sections under a confocal laser scanning microscope (FV-1000, Olympus, Tokyo, Japan) with a confocal depth of 1.0 mm. The mode of triple immuno uorescence labelings was used. The laser beams and lters for ZsGreen were the 488 nm of excitation and 525 nm of emision, the parameters for tdTomato were 550 nm of excitation and 570 nm of emision, and for the antibody stained were 649 nm of excitation and 670 nm of emision. Around 30 slices obtained from 6 mice (5 slices per mouse) were randomly chosen. Images were carried out by individuals blinded towards the experimental groups.

Imaris le format description of immuno uorescence labeling
In order to more accurately imaging three-dimensional (3D) structure of GABAergic Purkinje cells expressing red uorescence in the cerebellum of Gad2-cre;ZsGreen-tdTomato / mice, Imaris × 64 image analysis software (version 9.6.0, Oxford Instruments, England) was used. The Imaris le had high performance rendering and processing of laser confocal images. The Path Attribute Description was /DataSet/ResolutionLevel 0/TimePoint 0/Channel 0 with informations concerning resolution 0, time point 0 and channel 0. ImageSizeX = 285, indicating the size in X in pixel for Resolution Level 0; ImageSizeY = 218, indicating the size in Y in pixel; ImageSizeZ = 64, indicating the size in Z in pixe.

Single and double RNAscope in situ hybridization
To provide Purkinje cells-speci c spatial analysis about the RNAs pro ling of important mitochondrial moleculars, the single RNAscope technology (Wang F et al., 2012) was conducted on the cerebellum sections from the Gad2-cre;ZsGreen-tdTomato / mice. Moreover, in order to verify the characterization of the molecules, we conducted double RNAscope assays to simultaneously detect two targets on the normal mice section.

Tissue preparation
RNAscope® Multiplex Fluorescent Reagent Kit manual was performed as instructed by Advanced Cell Diagnostics (ACD). Newly prepared 1× phosphate buffered saline (PBS, pH 7.4) and 4% paraformaldehyde (PFA, pH 7.4) were used to perfuse the heart of mice. Tissues of interest were then removed in 4% PFA at 4˚C for 24 hours, and then treated with sucrose solutions for dehydration. The tissues were cryostat-sectioned at 15μm onto SuperFrost Plus charged slides. Sections were only brie y thawed to adhere to the slide but were immediately returned to the -20˚C cryostat chamber until completion of sectioning. Before used for following histology, The slides were baked in chamber at 37˚C for 3 hours. Slides were removed from baking chamber and immediately transferred to 1× PBS at room temperature for 5 minutes. Each slide was incubated with Hydrogen peroxide at room temperature for 10 minutes. After washed by distilled water, the slides were treated by the boiled target retrieval solutions at 96˚C for 10 minutes. The tissues were placed in distilled water immediately at room temperature, and then dehydrated in 100% ethanol for 3 minutes. The slides were air-dried brie y and then boundaries were drawn around each section using a hydrophobic pen (CIRISC PAP pen, I.S. CIRCLE WRITER, Japan). When hydrophobic boundaries had dried, protease III reagent was added to each section until fully covered, incubation at 40˚C for 30 minutes. During this period, slides were placed in a prewarmed humidity control tray containing dampened lter paper in the HybEZ oven (ACD).

Hybridization
Each slide was then added only one probe or a mixture of two probes. Channel 1, Channel 2, and Channel 3 probes (50: 1: 1 dilution, as directed by ACD due to stock concentrations), which were pipetted onto each section until fully submerged. This was performed one slide at a time to avoid liquid evaporation and section drying. The humidity control tray was placed in a HybEZ oven (ACD) for 2 hours at 40˚C. A table of all the probes used is shown in Table 3. After probe incubation, the slides were washed twice in 1× RNAscope® wash buffer and returned to the oven for 30 minutes after submersion in AMP-1 reagent. Washes and ampli cation were repeated using AMP-2, and AMP-3 reagents with a 30-, and 15-minute incubation period, respectively. HRP-C1 signal, HRP-C2 signal, and HRP-C3 signal were developed respectively. Different dyes were needed to differentiate probes deriving from three kinds of channels. We employed Opal 520 (FP1487001KT, PerkinElmer) to mark channel 1 probes, Opal 570 (FP1488001KT, PerkinElmer) to mark channel 2 probes, and Opal 690 (FP1497001KT, PerkinElmer) to mark channel 3 probes. Slides were incubated with DAPI for 5 minutes before being washed, air-dried, and coverslipped with Prolong Gold Antifade mounting medium.

RNAscope images acquisition and semi-quantitative analysis
Anatomical structures were analyzed in sagittal sections and mapped according to Paxinos and Franklin atlas (Paxinos G and Franklin KBJ., 2001) and Allen map (http://mouse.brain-map.org/static/atlas). Fluorescent signals of mRNA hybridization for mitochondrial molecules were imaged with a 10×, 20×, 40× and 60× objective lens on a a confocal laser scanning microscope (FV-1000, Olympus, Tokyo, Japan). All microscope and camera settings were identical for all images. We used the ImageJ64 software (National Institute of Health. Bethesda, MD, USA), as previously described (Fe Lanfranco M et al., 2017).
The RNAscope assay used a semi-quantitative H scoring guideline to evaluate the staining results. We scored the number of dots per Purkinje cell rather than the signal intensity to interpret RNAscope staining, because the number of dots correlates to the number of RNA copy numbers, whereas dot intensity re ects the number of probe pairs bound to each molecule (Gebara E et al., 2021). In the present study, H score was to visualize the dynamic expression level by binning the percentage of cells with a certain expression level or number of dots in one bin. The number of bins ranges from 0 -4 according to the ACD scoring system. The overall H score can range from 0 -400 and is calculated as shown: Negative control (dapB) stained slides were always imaged at the settings used for target probe imaging and did not result in appreciable signal. Positive control (Polr2a/PPIB/UBC) should be indicated as successful staining which have a score ≥2.

Statistical analysis
Figures for RNAscope and Immuno uorescence labeling are representative of sections obtained from three animals. Data for Figure 4, expressed as the mean ± SEM (sections obtained from ve animals), were analyzed using one-way analysis of variance (ANOVA) with Bonferroni's multiple comparison post hoc test using GraphPad Prism software v.5.0 (Graphpad). A p-value < 0.05 was considered statistically signi cant.

Characteristic of tdTomato cells in GAD2-cre;ZsGreen-tdTomato / mice
Firstly, we used a Cre-loxP strategy to generate a conditional knock-in mice with speci c uorescence in GABAergic Purkinje cell (Fig. 1A). The Gad2-ires-Cre mice inserted a Cre recombinase sequence at Gad2 allele to make all GABAergic neurons containing Cre recombinase.
So in the ZsGreen-tdTomato /f mice, ZsGreen and tdTomato were knocked in and the loxP sites were buried on both sides of ZsGreen. After crossing of these two mice, all GABAergic neurons would transfer from expressing ZsGreen to expressing tdTomato due to Cre recombinase cutting loxP sites. The genotyping strategy used four sets of primers to produce four bands of 1465 bp (for ZsGreen-tdTomato with loxP site), 285 bp (for wild type site), 352 bp (for Cre recombinase), and 250 bp (for wild type site) (Fig. 1B).
Secondly, average body appearances were compared between control Gad2-cre mice and the Gad2cre;ZsGreen-tdTomato / mice (Fig. 1C). Interestingly and apparently, Gad2-cre;ZsGreen-tdTomato / mice showed dazzling green light visible to the naked eye in external auricle skin (arrow), plantar skin (arrow), as well as perianal skin (arrow), although the body weights of both types of mice had no signi cant difference. Moreover, the brains of Gad2-cre;ZsGreen-tdTomato / mice showed green uorescence, while the brains of control Gad2-cre mice showed normal pink, which made it very easy to distinguish the GABAergic knock-in mice.
Thirdly, the confocal microscope images were taken to show that Purkinje cells expressed tdTomato in the cerebellum of Gad2-cre;ZsGreen-tdTomato / mice, while non-GABAergic cells expressed ZsGreen (Figs. 1D and 1E). We could nd that the neuronal bodies of Prukinje cells with red uorescence (asterisk marked in Fig. 1E) were aligned like dominos stacked throughout the PCL (purkinje cell layer), also they had large dendritic arbors (Fig. 1E) like within the molecular layer (ML) and sent axons like bright ame out off cerebellar cortex (Fig. 1E), therefore they left the dark area of granular layer (GCL) which were occupied by green non-GABAergic cells (Fig. 1E).
Fourthly, the Imaris le was designed to allow better visualization of Purkinje cells speci cally expressed tdTomato uorescence in GAD2-cre;ZsGreen-tdTomato / mice (shown in Videos. 1-3). We could see that the neuronal bodies of Prukinje cells with red uorescence were aligned like dominos stacked throughout the PCL, while the dark area of granular layer (GCL) were occupied by green non-GABAergic cells.

Double stainings veri ed tdTomato cells were Purkinje cells
Previous studies have shown that Cerebellar Purkinje cells could be marked by Calbinin (CB) (Singh-Bains MK et al., 2019), while the granular cells be marked by NeuN (Yawno T et al., 2019). In addition, the astrocytes (as GFAP as marker) (Levite M et al., 2020) and microglia (as Iba1 as marker) (Janakiraman U et al., 2020) were believed to be scattered throughout the cerebellar cortex. So, in order to verify the cells expressing tdTomato uorescence in cerebellum of Gad2-cre;ZsGreen-tdTomato / mice were exclusively Purkinje cells, the present immunostainings with the four antibodies mentioned above were observed under the psudo blue color. The percentage of blue cells in the distinct cell subgroups was determined.
It could be indicated that dominos stacked -like tdTomato-positive cells (red) were 95.0% colocalized with CB (green) ( Fig. 2A), so that they appeared the color of purple. On the contrary, immunostaing data showed few of tdTomato-positive cells were colocalized with NeuN, the marker of granular cells (Fig. 2B). Moreover, almost none of tdTomato-positive cells were colocalized either with astrocyte marker GFAP (Fig. 2C) or with microglia marker Iba1 (Fig. 2D).
At the same time, ZsGreen-positive cells were densely distributed cross GCL, scattered though PCL, and few in ML (Fig. 2). Double stainings showed almost none of these green cells were colocalized with Purkinje cell marker CB ( Fig. 2A). On the contrary, these green cells were mainly consist of NeuN-positive granular cells (Fig. 2B). Our results also con rmed that ZsGreen-positive cells were not astrocyte (Fig.  2C) or microglia (Fig. 2D).
These data suggested that tdTomato-positive cells were primarily expressed in GABAergic Purkinje cells within cerebellar cortex. Then, RNAscope in situ hybridization was performed to examine the presences of seven mitochondrial proteins within the Purkinje cells of Gad2-cre;ZsGreen-tdTomato / mice.
3.3 Application of RNAscope in situ hybridization onto GAD2-cre;ZsGreen-tdTomato / mice for Purkinje cell -speci c spatial analysis Identi cation of mRNAs of mitochondrial fusion (Mfn2), calcium transporter (Mcu and Nclx) and uncoupling proteins (Ucp2 and Ucp4) in Purkinje cells in the cerebellar cortex of GAD2-cre;ZsGreen-tdTomato / mice with RNAscope probes. Red cells were Purkinje cells; green cells were non-GABAergic cells; blue dots were RNA uorescence. Moreover, ACD scoring system was used to calculate the overall H scores of Dendritic Weighted Formula (DWF) and Soma Weighted Formula (SWF).

Mfn2 mRNA
Mitofusin 2 (Mfn2) control the fusion of the outer mitochondrial membrane (OMM), but the physiological function of Mfn2 in Purkinje cells remains unclear. Our data rstly demonstrated that Mfn2 mRNAs expression was evident in Prukinje cells in cerebellum of Gad2-cre;ZsGreen-tdTomato / mice (Fig. 3A). Moreover, 76.5% dendrite shafts were ranked as Bin 1 because they only had 1-3 dots per shaft (Fig. 3B); On the contrary, 71.8% soma were ranked as Bin 4 because they had more than 15 dots per cell body (Fig.   3C). The overall H scores of DWF and SWF were calculated as 60 and 139, respectively (Table 4).

Mcu mRNA
Mitochondrial Ca 2+ uptake is mediated by the Mitochondrial Calcium Uniporter (MCU) complex, located on the inner mitochondrial membrane (IMM). Our data con rmed the previous report (Fecher C et al., 2019) that few Mcu mRNAs expressions were present on Prukinje cells in cerebellum of Gad2cre;ZsGreen-tdTomato / mice (Fig. 3D). About Moreover, 57.1% dendrite shafts were ranked as Bin 0 because they had no dot per shaft (Fig. 3E); Similarly, 58.1% soma were also ranked as Bin 1 (Fig. 3F). The overall H scores of DWF and SWF were high as 70 and 22, respectively (Table 4).

Nclx mRNA
Conversely to MCU, Ca 2+ release is under the control of the Na + /Ca 2+ exchanger, encoded by the NCLX gene, located on IMM. Our data rstly demonstrated that Nclx mRNAs expression were only scattered on both soma and dendrites in Prukinje cells in cerebellum of Gad2-cre;ZsGreen-tdTomato / mice (Fig. 3G).
About 91.7% dendrite shafts were ranked as Bin 0 because they had no dot per shaft (Fig. 3H); Similarly, 58.3% soma were ranked as Bin 1 because they had only no more than 3 dots per cell body (Fig. 3I). The overall H scores of DWF and SWF were 4 and 31, respectively (Table 4).

Ucp2 mRNA
UCP2 are IMM proteins that may regulate mitochondrial energy metabolism and ROS generation. Our data rstly demonstrated that unexpectedly, few Ucp2 mRNAs expression was present in Prukinje cells in cerebellum of Gad2-cre;ZsGreen-tdTomato / mice (Fig. 3J). Moreover, 75.0% dendrite shafts and 41.2% soma were ranked as Bin 0 because they had no dot (Figs. 3K and 3L). The overall H scores of DWF and SWF were only 15 and 27, respectively (Table 4).
3.3.5 Ucp4 mRNA UCP4, another IMM protein for regulating mitochondrial energy metabolism and ROS generation, has been looked like a twin of UCP2. Our data con rmed that few Ucp4 mRNAs expression was present in dendritic shafts of Prukinje cells in cerebellum of Gad2-cre;ZsGreen-tdTomato / mice (Fig. 3M). However, 67.6% soma were ranked as Bin 3 because they had 10-15 dots per cell body (Figs. 3N and 3O). The overall H scores of DWF and SWF were only 14 and 103, respectively (Table 4).

Double RNAscope pro ling of mitochondrial molecules in cerebellar cortex of normal mice
In order to verify the characterization of the molecules, we conducted double RNAscope assays to simultaneously detect two targets on the normal mice section.
3.4.1 Mfn1 (green) and Pcp2 (red) Because our present Mfn1 probe was not suitable for the detection in the samples from GAD2cre;ZsGreen-tdTomato / mice, we did double RNAscope pro ling of Mfn1 mRNA (green) in Purkinje cells (red) which were distinguished by the red uorescence with the mRNA probe of Purkinje cell protein-2 (pcp2) gene (Fig. 4A). Higher magni cation images showed that within the PCL, some dots could be found in the soma of Purkinje cells (red) (Fig. 4B). However, within both the ML and the GCL, few green dots could be found in the dendrites or axons of Purkinje cells, respectively (Fig. 4B). ACD quanti cation con rmed the middle-level expression level of Mfn1 mRNAS in soma (Fig. 4C). The calculation of the percent of double Pcp2-Mfn1 stainings on single Pcp2-positive expressions con rmed the very low colocalization both in the ML and in the GCL (Fig. 4D). The data suggested Mfn1 mRNAs are present only in the soma of the Purkinje cells, instead of processes.

Drp1 (green) and Mcu (red)
Because our previous report had presented the detailed distribution of Drp1 mRNA in the cerebellar cortex of normal mice by in situ hybridization method, the present double RNAscope pro ling was to verify the distribution of Mcu in Purkinje cells because these cells could be distinguished by Drp1 uorescence (Fig.  5A). Higher magni cation images con rmed that green Drp1-positive dots could outline the soma of Purkinje cells (Fig. 5B). Our results of Gad2-cre;ZsGreen-tdTomato / mice by in situ hybridization had showed that a number of dots of Mcu mRNAs were present in processes and soma of Prukinje cells (Fig.  3D). Here, the double RNAscope pro ling was consistent with the previous distribution pattern. Unexpectedly and interestingly, almost none of dots of Drp1 mRNAs was co-localized with dots of Mcu mRNAs, wherever at ML, PCL or GCL (Fig. 5C). The data suggested the separation of Drp1 on OMM and Mcu on IMM even they were all abundant in the Purkinje cells.

Ucp2 (green) and Mfn2 (red)
Our results of Gad2-cre;ZsGreen-tdTomato / mice by RNAscope in situ hybridization had showed that a number of dots of Mfn2 mRNAs were present in soma of Prukinje cells, although this kind of high level did not happen in dendrites of Purkinje cells (Figs. 3A-3C). On the contrary, the dots of Ucp2 mRNAs were only scattered not only on the soma but also on the dendrites of Prukinje cells (Figs. 3J-3L). The present double RNAscope pro ling con rmed the dense expressions of Mfn2 in soma of the Purkinje cells (Fig.  6A), so that in the higher magni cation images these cells could be distinguished by Mfn2 red uorescence (Fig. 6B). Moreover, unsurprisingly, almost none of dots of Ucp2 mRNAs was co-localized with dots of Mfn2 mRNAs (Fig. 6C). The data veri ed the presence of Mfn2 but the nonexistence of Ucp2 on the Purkinje cells.

Discussion
To explore the mitochondrial mechanisms of Purkinje cells, we rst set up a transgenic mice in which Purkinje cells could be distinguished easily by expressing red uorescence. We then detected spatial RNA pro ling of seven mitochondrial molecules, including dynamin-related protein-1 (Drp1), mitochondrial calcium uniporter (MCU); mitofusion 1 and 2 (Mfn1/2), sodium/lithium/calcium exchanger (NCLX), and uncoupling protein 2 and 4 (UCP2/4), by RNAscope combined with ACD quanti cation. We nally proposed a mitochondrial Drp1/Mfn2/UCP4 convergence on the Purkinje cells, which would make up a mitochondrial quality control system. Our results establish a framework for understanding the pathogenic mechanism underlying cerebellum-related neurological diseases.

Drp1/Mfn1/Mfn2
Mitochondrial ssion and fusion play critical roles in creating new mitochondria and removing damaged mitochondria. In mammalian cells, ssion/fusion events are mainly mediated by several large dynaminrelated GTPase proteins, including conserved dynamin-related GTPase (Drp1), conserved dynamin-related GTPase mitofusion 1 and 2 (Mfn1 and 2), and optic dominant atrophy 1 (Opa1). Our lab have published the data previously about the expression of Drp1 at high level on the soma of cerebellum Purkinje cells by the combination of immunohistochemistry and in situ hybridization on GAD67 (glutamic acid decarboxylase 67) -GFP (green uorescent protein) transgenic mice (Luo TT et al., 2020). The present ndings have con rmed the distribution pattern of Drp1 on Purkinje cells. These data suggest that the mitochondrial ssion in Purkinje cells may be dependent on the mitochondrial ssion mediator, Drp1. In fact, mitochondrial ssion mediated by the GTPase Drp1 is an attractive drug target in neurodegenerative disorders (Chandra R et al., 20107;Bordt EA et al., 2017). Basal Drp1-dependent mitochondrial ssion is required for mitochondrial tra cking to synapses, mitochondrial quality control, and brain development (Ishihara N et al., 2009;Wakabayashi J et al., 2009;Kageyama Y et al., 2014). Drp1 is highly conserved and contains many critical functional features that correspond to speci c target structures within the enzyme, such as GTP binding, GTP hydrolysis, self-assembly and protein interactions with key functions in mitochondrial division (Frank S, et al., 2001;Lackner LL and Nunnari J, 2010). While the ssion defects may limit mitochondrial motility, decrease energy production, promote oxidative stress and lead to accumulating of mtDNA defects, thereby promoting neuronal dysfunction and cell death (Grohm J et al., 2012;Oettinghaus B, et al., 2012). Thus, regardless of the different upstream stress stimulus, Drp1mediated mitochondrial fragmentation and downstream mitochondrial pathways play a major role for the fate of Purkinje cells. Signi cantly, Drp1 must be an e cient strategy for the neuroprotection against multiple cerebellar damage.
The nding of dense distribution of Mfn2 on Purkinje cells deserves high attention for four reasons. First, the mutations in Mfn2 have been found to cause a human neurodegeneration disease, Charcot-Marie-Tooth neuropathy type 2A (Chen H et al., 2007;Züchner S et al., 2004;Chung KW et al., 2006;Ishikawa K et al., 2019). Second, our data have supported previous paradigm that Mfn2 is expressed at signi cantly greater levels in Purkinje cells than is Mfn1 (Chen H et al., 2007). Third, it has been reported that Purkinje cells require Mfn2 but not Mfn1 for cell survival and dendritic outgrowth (Chen H et al., 2007;Züchner S et al., 2004;Chung KW et al., 2006). Fourth, Mfn2-de cient Purkinje cells have shown impaired respiratory complex activity and defects in inner membrane structure characteristic of respiratory dysfunction (Chen H et al., 2007). Our RNAscope studies provide insight into the dependence of the fusion of outer mitochondrial membrane (OMM) of soma of Purkinje cells on the molecular Mfn2.
In fact, neurodegeneration in neurodegenerative diseases has been related to several mitochondrial dynamics imbalances such as excessive fragmentation of mitochondria, impaired mitophagy, and blocked mitochondrial transport in axons. Our ndings raise the intriguing possibility that a convergent pathway underlies the pathogenesis of neurodegenerative disorders. Nonetheless, the exact role of Drp1/Mfn2-dependent mitochondrial dynamics in Purkinje cells requires further investigation.

MCU and NCLX
Mitochondrial Ca 2+ homeostasis plays a central role in nervous system. Previous studies have suggested that dysfunction of Ca homeostasis is associated with oxidative stress and many neurological diseases (Rizzuto R et al., 2012;Giorgi C et al., 2018;Fan M et al., 2020;Baughman JM et al., 2011;De Stefani D et al., 2015;Peng TI and Jou MJ, 2015;Owens K et al., 2013;Di Lisa F and Bernardi P, 2009;Zhang L et al., 2019). In the last 5 years, most of the molecules that control mitochondrial Ca 2+ homeostasis have been nally identi ed. Mitochondrial Ca 2+ uptake is mediated by the Mitochondrial Calcium Uniporter (MCU) complex, a macromolecular structure that guarantees Ca 2+ accumulation inside mitochondrial matrix upon increases in cytosolic Ca 2+ . Conversely, Ca 2+ release is under the control of the Na + /Ca 2+ exchanger, encoded by the NCLX gene, and of a H + /Ca 2+ antiporter, whose identity is still debated (De Stefani D et al., 2016). Audrey Bonnan et al. have found that in Purkinje cells the dendritic Ca 2+ transients are su cient, potent triggers of plasticity induction that instruct the acquisition of cerebellar learning, by using optogenetics and animal behavioral tests (Bonnan A et al., 2021). In the present study, as expected, our results have con rmed the previous important conclusion (Fecher C et al., 2019) that very low level of MCU expressions were on Prukinje cells. At the same time, we have at the rst time reported that the numbers of Dendritic Weighted Formula (DWF) and Soma Weighted Formula (SWF) were even lower as 4 and 31, respectively. If it were true, it would be reasonable to speculate that there is other mitochondrial calcium regulators, such as of a H + /Ca 2+ , antiporter which provides exibility to the cerebellum. Our data pave a way to appreciate mitochondria in Purkinje cells as a highly cell-typespeci c biology.

UCP2 and UCP4
Mitochondria take up Ca 2+ through the mitochondrial calcium uniporter complex to regulate energy production, cytosolic Ca 2+ signaling, and cell death (Rizzuto R et al., 2012;Giorgi C et al., 2018;Fan M et al., 2020). In mammals, the uniporter complex (uniplex) contains four core components: the pore-forming MCU, gatekeeper MICU1 and MICU2, and an auxiliary EMRE subunit essential for Ca 2+ transport. Previous studies have suggested that MCU-regulated Ca homeostasis is associated with oxidative stress and many neurological diseases (Baughman JM et al., 2011;De Stefani D et al., 2015;Peng TI and Jou MJ, 2015;Owens K et al., 2013;Di Lisa F and Bernardi P, 2009;Zhang L et al., 2019). It has been found that in Purkinje cells the dendritic Ca 2+ transients are su cient, potent triggers of plasticity induction that instruct the acquisition of cerebellar learning, by using optogenetics and animal behavioral tests (Bonnan A et al., 2021). In the present study, we have found the high level of Mcu expression dots on dendrites (Fig. 4D), soma (Fig. 4D), even axons (Fig. 5) of Prukinje cells. In view of this, it is reasonable to speculate it is due to Mcu which provides exibility to the cerebellum in its role in producing appropriate behavioral responses to different adaptive stimuli. Thus MCU has been con rmed to be a potential therapeutic target in neurological diseases in the future.

Conclusion
When applied to the purkinye cells of the cerebellum, our approach yielded a number of insights. First, our veri ed the convergence of mitochondrial proteins Drp1/Mfn2/MCU/UCP4 on the purkinje cells in situ. Second, we generated GAD2-cre;ZsGreen-tdTomato / mice to allow resolving of purkinje cell-speci c mitochondrial changes under multiple pathological conditions because of their high-resolution imaging for purkinje cells, both by light microscopy and electron microscopy. Third, mitochondrial proteins topographical analysis of our cerebellar RNAscope resulted in clear predictions for differentially regulated mitochondrial dynamic mechanisms, which could be studied in the further research. Fourth, the lack of Mfn1/Nclx/Ucp2 indicated the unique intracellular mitochondrial mechanisms in Purkinje cells related to the cell distinct morphology, ring pattern and synaptic plasticity.    with a certain number of dots in one bin in cerebellum of Gad2-cre; ZsGreen-tdTomato fl/fl mice.
The number of bins ranges from 0 -4 according to the ACD scoring system. The overall H score can range from 0 -400 and was calculated as methods described in the Methods. Purkinje cells speci cally expressed tdTomato in GAD2-cre;ZsGreen-tdTomato / mice. (A) Strategy to generate a conditional knock-in mice with speci c uorescence in GABAergic Purkinje cell by using cre-loxP system. The Gad2-ires-Cre mice inserted an ires and a Cre recombinase sequence at the 3' end of the Gad2 allele, so that all of GABAergic neurons contained Cre recombinase. In the ZsGreen-tdTomato /f mice, ZsGreen and tdTomato were knocked in and the loxP sites were buried on both sides of ZsGreen.