Pannexin 1 Regulates Dendritic Protrusion Dynamics in Immature Cortical Neurons

Abstract The integration of neurons into networks relies on the formation of dendritic spines. These specialized structures arise from dynamic filopodia-like dendritic protrusions. It was recently reported that cortical neurons lacking the channel protein pannexin 1 (PANX1) exhibited higher dendritic spine densities. Here, we expanded on those findings to investigate, at an earlier developmental time point (with more abundant dendritic protrusions), whether differences in the properties of dendritic protrusion dynamics could contribute to this previously discovered phenomenon. Using a fluorescent membrane tag (mCherry-CD9-10) to visualize dendritic protrusions in developing neurons [at 10 d in vitro (DIV10)], we confirmed that lack of PANX1 led to higher protrusion density, while transient transfection of Panx1 led to decreased protrusion density. To quantify the impact of PANX1 expression on protrusion formation, elimination, and motility, we used live cell imaging in DIV10 neurons (one frame every 5 s for 10 min). We discovered that at DIV10, loss of PANX1 stabilized protrusions. Notably, re-expression of PANX1 in Panx1 knock-out (KO) neurons resulted in a significant increase in protrusion motility and turnover. In summary, these new data revealed that PANX1 could regulate the development of dendritic spines, in part, by controlling dendritic protrusion dynamics.

It was unclear from the previous study (Sanchez-Arias et al., 2019) whether the increase in dendritic spine density associated with Panx1 knock-out (KO) was because of an increase in the formation and/or an increase in the stability of the processes that develop into dendritic spines. Moreover, the previous study did not address whether re-expression of PANX1 could "rescue" the impact of Panx1 KO. Dendritic spines develop from long, thin, highly dynamic processes called "dendritic protrusions" that are motile (i.e., grow and retract), and either disappear relatively rapidly (on the order of tens of seconds to minutes) or "survive" and stabilize. The goals of the current study were therefore 2-fold: (1) to investigate the impact of PANX1 on dendritic protrusion density and dynamics in developing cortical neurons; and (2) to determine whether the impact of Panx1 KO could be rescued by re-expression of PANX1.
Because dendritic protrusion development is more readily accessible to study in vitro, and these systems (e.g., primary neuronal cultures) recapitulate many features of spine development observed in vivo (Ziv and Smith, 1996;Dunaevsky et al., 1999;Prange and Murphy, 2001;Zuo et al., 2005), we established an approach to study dendritic protrusions in immature cortical neuron cultures. Dendritic protrusions are long, thin, and contain relatively little cytoplasm. Therefore, standard cytoplasmic fluorescent proteins are not ideal for studying dendritic protrusions in immature, developing neurons. To this end, we devised a novel approach using the transmembrane tetra-spanin CD9-10 (Tspan29) protein fused to the monomeric red fluorescent protein mCherry (mCherry-CD9-10) and imaged dendritic protrusion dynamics in living neurons over a 10-min time interval. To determine the impact of Panx1 KO on dendritic protrusion density and dynamics, and also to determine whether any observed effects could be rescued by re-expression of PANX1, we transiently transfected wild-type (WT) and Panx1 KO neuronal cultures with control EGFP or PANX1-EGFP (as well as mCherry-CD9-10) and analyzed dendritic protrusions in fixed and living neurons at 10 d in vitro (DIV10). We confirmed that loss of PANX1 led to higher dendritic protrusion density. As anticipated, rescue of PANX1 in Panx1 KO cultures, and transient expression of PANX1 in WT cultures led to decreased dendritic protrusion densities. Moreover, Panx1 KO was associated with increased dendritic protrusion stability. Finally, transient PANX1 expression in Panx1 KO cultures significantly increased motility and turnover of dendritic protrusions. In summary, these new data revealed an inverse relationship between PANX1 expression levels and dendritic protrusion stability, suggesting that the loss of PANX1, either through genetic deletion or during developmental downregulation, results in increased cortical dendritic spine density, in part by stabilizing dendritic protrusions.

Experimental animals
All animal procedures were approved by the University of Victoria Animal Care Committee and performed in accordance with the guidelines set by the Canadian Council on Animal Care. Male and female postnatal day (P)0-P1 mice were used in this study. C57BL/6J mice were obtained from The Jackson Laboratory (#000664, Table 1). The global Panx1 KO strain in this study was derived from the Panx1 Vsh null mice, originally generated by Valery Shestopalov . Note that this strain was developed using targeted mutagenesis on 129embryonic stem cells (129-ESCs, 129 Â 1/SvJ x 129S1/Sv) with C57BL/6J as the recipient strain and backcrossed for five generations by the original authors. Recent comparative genomic analysis revealed the original Panx1 Vsh null mice contained a loss of function passenger mutation in Casp4 (Casp11), common to five 129-substrains (129 Â 1/ SvJ, 129S1/SvImJ, 129S2/SvPas, 129S6/SvEvTac and 129P3/J) and are predicted to have passenger mutations in these additional genes: Mmp1a, Olfr832, Fbxl12, ENSMUSG00000095186, and ENSMUSG00000095891 (Vanden Berghe et al., 2015). These mice were further backcrossed in-house onto a C57BL/6J for at least six generations (Sanchez-Arias et al., 2019). Mice were housed under a 12/12 h light/dark cycle starting at 8 A. M., with food and water ad libitum; temperature was maintained between 20°C and 25°C and humidity at 40-65%.

Primary cortical neuron cultures and transfections
Primary cortical neuron cultures were prepared as previously described (Sanchez-Arias et al., 2019). Briefly, cortices from male and female P0 pups from timed-pregnant WT and Panx1 KO breeding pairs were microdissected and incubated with papain, dispase-1, and DNase-1 for 40 min in HBSS, followed by mechanical dissociation in DMEM/F12 medium supplemented with Neurocult SM1, GlutaMAX, and penicillin/streptomycin (P/S). A total of 125,000 cells were plated in a Nun Lab-Tek eight-well chambered coverglasses coated with PDL. One to two hours after plating, the medium was replaced with Neurocult supplemented with Neurocult SM1, GlutaMAX, P/S, and gentamicin. From DIV4 onwards, partial (half) medium changes were done with BrainPhys maturation medium (Bardy et al., 2015); to limit proliferation of glial cells, ara-C was added to the medium at DIV4. Transfections were performed at DIV6 using Lipofectamine2000. DNA/ lipid complexes were diluted in OptiMEM-I at a ratio of 2 mg DNA:1 ml lipofectamine and incubated at room temperature for 30 min. Then, these DNA/lipid complexes were added to cells in BrainPhys medium without antibiotics and incubated for 1-1.5 h. Neurons were transfected with either pEGFP-N1 (250 ng) or Panx1-EGFP (250 ng, gift from Silvia Penuela and Dale Laird;Penuela et al., 2007). All transfections contained mCherry-CD9-10 (250 ng, a gift from Michael Davidson; Addgene plasmid #55013; http://n2t. net/addgene:55013; RRID:Addgene_55013) to visualize neurons and dendritic protrusions. All neurons in this study were used at DIV10. Neurons used for fixed quantifications were plated on PDL-coated coverslips.

Genotyping
A set of primers targeting upstream exon 3, exon 3, and downstream exon 4 of Panx1 (CTTTGGCATTTTCCCA GTGT, CGCGGTTGTAGACTTTGTCA, and GTCCCTACA GGAGGCACTGA) were used to genotype mice as previously described (Sanchez-Arias et al., 2019). Genomic DNA was extracted from tail-clips using MyTaq Extract PCR kit. DNA from WT mice amplifies a single 585-bp band, whereas DNA amplified from global Panx1 KO mice have a single 900-bp band.

Imaging and analysis of dendritic protrusions in fixed cortical neurons
Dendritic protrusions (including filopodia) were defined as any membranous protrusions between 0.4 and 10 mm. Neurons were fixed on coverslips with 4% paraformaldehyde and 4% sucrose for 10 min and mounted on microscope slides with VectaShield antifade mounting medium. High-resolution images (3320 Â 3320, pixel size: 0.088 mm, z-step size: 0.4 mm) were acquired using a Leica TSC SP8 microscope using a 40Â immersion oil objective (1.30 NA) and exported to FIJI for analysis (Schindelin et al., 2012). Two experimenters were involved in dendritic protrusion density analysis. One experimenter quantified dendritic protrusions according to the inclusion criteria [neurons that expressed both mCherry-CD9-10 and EGFP/PANX1-EGFP, did not show signs of stress (blebbing, large vacuoles), and where dendritic protrusions were originating from the longest primary neurite] from all experimental groups. The other experimenter quantified dendritic protrusions only from mCherry-CD9-10 and EGFP-expressing neurons to obtain quantitative analysis of the improvement in dendritic protrusion detection realized with mCherry-CD9-10 versus EGFP signal and record interobserver variance in protrusion analysis. For quantitative assessment of dendritic protrusion detection realized with mCherry-CD9-10 versus EGFP signal, neurons were randomly selected by a second experimenter for analysis. Using the mCherry-CD9-10 signal, the experimenter traced individual dendritic protrusions emanating from the shaft of the longest neurite (primary neurite). The corresponding EGFP signal for each dendritic protrusion (initially traced using the mCherry-CD9-10 signal) was inspected and the presence or absence of a corresponding dendritic protrusion in the EGFP channel was recorded. The interobserver variability was determined by Pearson's correlation and reported as R 2 . Dendritic protrusions from all experimental groups were traced in the same manner, and densities were calculated by dividing the total number of dendritic protrusions by the segment length and multiplying by 10 (dendritic protrusions per 10 mm). Representative images were processed uniformly with a Gaussian blur of 0.5 pixels, and uniform adjustments to levels and contrast were made using Photoshop CS6 Extended suite (Adobe Systems).

Imaging and analysis of dendritic protrusions in live cortical neurons
Cortical neurons plated on chambered coverglasses and cultured in BrainPhys medium were placed in an incubation chamber attached to a Leica TSC SP8 microscope and held at 37°C and 5% CO 2 for the entire duration of the experiment. After 15 min of acclimatization, primary and secondary dendrite segments within 100-150 mm from the cell body were imaged. We selected segments from the longest neurite (dendrite) that were clearly discernible from surrounding structures (other dendrites and cells). These segments ranged from ;65 to ;75 mm long. Time lapses of z-stacks (0.7-mm z-step size) were acquired every 5 s for 10 min at a resolution of 1024 Â 256 (pixel size: 0.06 mm) using Leica's TCS SP8 resonant mode (8000 Hz) and a 63Â water immersion objective (1.20 NA). Images were exported to FIJI for analysis. First, the four-dimensionality (x, y, z, t) was reduced by creating maximum z projections before additional image processing, x-y drift was corrected with MultiStackReg v1.45 (developed by Brad Busse; http://bradbusse.net/ MultiStackReg1.45_.jar) when required. Then, images were subjected to a low-pass filter using a Gaussian blur (kernel size 2) and thresholded using the triangle method (Zack et al., 1977;Polanco et al., 2018). From these binary images, outlines for each time frame were created and temporally color coded. Dendritic protrusions were manually counted, and four basic characteristics were recorded: formation, elimination, lability, and motility. We defined formation as any de novo appearance of a dendritic protrusion within the timelapse recording; elimination was defined as the complete disappearance of a dendritic protrusion. Lability was defined as dendritic protrusions that were formed and eliminated within the duration of the time lapse, typically short-lived and lasting 1-3 min. To assess dendritic protrusion motility, we annotated partial extensions and partial retractions of individual dendritic protrusions. The survival fraction of dendritic protrusions was calculated by dividing the number of dendritic protrusions at the end of each time lapse (10-min mark) by the number of dendritic protrusions at the start (0-min mark). The overall turnover rate was calculated as the net percent gain and loss (sum of formation, elimination, and lability) of dendritic protrusions divided by the number of dendritic protrusions at the start of the time lapse. Lastly, the overall movement change of dendritic protrusions (Dmovement) was calculated by adding the fundamental characteristics of dynamics (formation, elimination, lability, and motility) divided by the number of dendritic protrusions at 0 min. Neurons that displayed signs of blebbing or for which focal planes were lost during the acquisition of images were excluded from analysis. Representative images were processed uniformly with a Gaussian blur of 0.5 pixels, and uniform adjustments to levels and contrast were made using Photoshop CS6 Extended suite (Adobe Systems Inc.).

Experimental design and statistical analysis
Note that no additional experiments were completed after the initial submission because of the COVID-19 pandemic; these data are also included in the PhD thesis of J.C.S.-A (Sanchez-Arias, 2020). For experiments with fixed neurons, two coverslips from three independent cultures were used; for live cell imaging experiments, two to three eight-well chambered coverglasses from three independent cultures were used. Neurons from either WT or Panx1 KO mice were plated in PDLcoated eight-well chambered cover glass or on PDLcoated coverslips in a randomized order by one experimenter, and all transfections, imaging, and image analysis were performed without knowledge of the genotype of the cultures by another experimenter. Neurons that displayed signs of blebbing or for which the focal plane was lost during image acquisition were excluded from analysis. Relevant details are described in Results, figure legends, and where appropriate, illustrated on the figures themselves. Data are presented as mean 6 SD. Data analysis using bootstrap estimation (5000 bootstrap resamples), determination of effect size, bias-corrected confidence intervals, and Cumming estimation plots were generated using the dabestr package for R (Bernard, 2019;Calin-Jageman and Cumming, 2019;Ho et al., 2019). Null-hypothesis significance testing was performed using R (version 3.6.2), and p , 0.05 was used as the significant threshold for these tests. Normality was tested using the Shapiro-Wilk test (McDonald, 2014). Group analyses for normally distributed data were performed with a two-way ANOVA coupled to multiple comparisons with Bonferroni's correction. For non-normally distributed data, Kruskal-Wallis pairwise comparisons with Bonferroni's correction were used.

Results
Detection of dendritic protrusions in cortical neurons was improved with a membrane-bound fluorescent marker Dendritic spines develop from highly dynamic dendritic protrusions. Genetic methods used to detect and measure these structures have mainly consisted of expression of cytoplasmic fluorescent proteins, such as EGFP. Transient transfection of EGFP (or one of its many variants) represent a convenient way to sparsely label neurons. There is a reasonable concern, however, that these cytoplasmic markers could fail to detect a substantial proportion of dendritic protrusions, given that these structures are relatively thin (and therefore contain a relatively small amount of cytoplasm; Fig. 1A). Membrane-bound lipophilic dyes (DiI, DiO, etc.), on the other hand, label thin membrane processes quite well (Mancuso et al., 2013). In order to combine the convenience of transient transfection of fluorescent proteins with the effectiveness of membranebound labels in thin process detection, we transfected cortical neurons with a fluorescent protein, mCherry, fused to a transmembrane protein, tetraspanin CD9-10 (mCherry-CD9- Figure 1. Detection of dendritic protrusions in cortical neurons was improved with a membrane-bound fluorescent marker. A, Representative maximum intensity projection of a dendritic segment from a neuron transfected with mCherry-CD9-10 and EGFP at DIV6 and fixed at DIV10. Thin and long dendritic protrusions are more clearly visualized with mCherry-CD9-10 (mid) than the cytoplasmic volume marker EGFP (bottom). Structures not clearly labeled with EGFP are denoted by p, and those missed entirely are denoted with arrowheads. B, Slopegraph showing the quantification of dendritic protrusions detected using the mCherry-CD9-10 signal compared with the EGFP signal. On average, 34% of dendritic protrusions detected with mCherry-CD9-10 were missed in the EGFP channel (EGFP: 72 6 5.9 dendritic protrusion; mCherry-CD9-10: 110 6 9.7 dendritic protrusions; p = 0.00384, Student's t test; N = 12 neurons, 6 WT and 6 Panx1 KO a1 ). C, The interobserver variability was evaluated with Pearson's correlation and found to be R 2 = 0.95 (95CI 0.84-0.98, p = 5.1e À08 ) a2 . d.p., dendritic protrusion; pp,0.01. Effect sizes are reported in Table 2. Scale bar: 5 mm. These data are included in the PhD thesis of J.C.S.-A, University of Victoria (Sanchez-Arias, 2020), found at http://hdl.handle.net/1828/11714.
Research Article: New Research 10, from Michael Davidson, Addgene plasmid #55013). As expected, the fluorescence signal from mCherry-CD9-10 delineated the plasma membrane (in addition to detecting somatic puncta most likely representing endosomes; Polanco et al., 2018). Qualitatively, it appeared that mCherry-CD9-10 improved the detection of dendritic protrusions. To quantify the improvement with mCherry-CD9-10, we traced dendritic protrusions using the mCherry-CD9-10 signal and compared this with the EGFP signal to confirm whether or not the same dendritic protrusions could be detected in that channel. We found that ;34% a1 of dendritic protrusions detected using the mCherry-CD9-10 signal were not readily detected with the EGFP signal (Fig. 1B). We compared the number of dendritic protrusions detected between two observers using Pearson's correlation and found them to be R 2 = 0.95 a2 , suggesting mCherry-CD9-10 facilitated dendritic protrusion detection independent of the observer (Fig. 1C). These results suggest that mCherryCD9-10 represents an improved method for measuring dendritic protrusions in neurons.

Novel methods for measurement of dendritic protrusions dynamics in living neurons
To examine the mechanisms contributing to differences in dendritic protrusion densities between groups, we acquired 10-min time lapses (one frame every 5 s) of primary and secondary dendrites from cortical neurons at DIV10. These cultures were transfected with mCherry-CD9-10 and either EGFP or Panx1-EGFP at DIV6. At DIV10, dendrites harbor highly dynamic, thin, and long dendritic protrusions that are the precursors for dendritic spines (Ziv and Smith, 1996;Fiala et al., 1998). We reduced the dimensionality of the time lapses by creating maximum zprojections, and then images were passed through a lowpass filter and thresholded to create outlines (Fig. 3A). The dendritic silhouettes (Fig. 3B) were then temporally color coded to facilitate the detection of formation, elimination, lability, retraction, and growth of dendritic protrusions (Fig. 3C).

PANX1 increases dendritic protrusion movement
Previous work showed that Panx1 KO was associated with an increase in dendritic spine density in situ and in vitro (Sanchez-Arias et al., 2019). However, it was unclear whether the increased dendritic spine density arose from additional formation and/or enhanced stability of spines. In order to investigate how PANX1 regulates the stability of dendritic protrusions, the precursors to dendritic spines, we developed a novel approach consisting of sparse expression of a membrane bound fluorescent protein, mCherry-CD9-10. This method significantly improved detection of dendritic protrusions over commonly used cytoplasmic fluorescent protein expression approaches (Mancuso et al., 2013). This novel methodology enabled us to discover a reciprocal relationship between PANX1 expression levels and dendritic protrusion density and stability. Taken together with the previous findings, this work suggests that PANX1 regulation of dendritic protrusion dynamics contributed to the increase in cortical dendritic spine density previously observed with Panx1 KO. Moreover, these findings suggest that developmental downregulation of PANX1 likely contributes to increased spine stability underlying spine formation in vivo during the first four postnatal weeks of brain development.
Transient expression of PANX1-EGFP significantly affected dendritic protrusion dynamics in Panx1 KO Figure 3. Novel methods for measurement of dendritic protrusion dynamics in living neurons. Ten-minute time lapses were acquired by imaging dendrite segments from cortical neurons every 5 s. Note that this a DIV10 WT cortical neuron transfected with mCherry-CD9-10 and PANX1-EGFP; only mCherry-CD9-10 is shown. The dimensionality of these recordings was reduced by creating maximum z-projections. Images were thresholded to create outlines (A; scale bar: 10 mm), which were temporally color coded (B; scale bar: 10 mm), allowing the visualization of various events, such as the percentage of dendritic protrusion (relative to time 0) undergoing formation (de novo appearance), elimination (complete disappearance by the end of the time lapse), lability (appearance and disappearance by the end of the time lapse), and motility (incomplete shrinkage or growth to an existing protrusion) shown in C (scale bar: 2 mm). Note that examples in C (cropped to highlight the event in question with (p) denoting the protrusion events) come from different cultures and different genotypes all at DIV10 transfected with mCherry-CD9-10 and either EGFP or Panx1-EGFP at DIV6. Note that the example provided here for dendritic protrusion elimination ( †) in the box in part C comes from the larger neurite depicted in panel B. Also note that the data shown in cultures but had less of an impact in WT cultures. The dampened impact of PANX1-EGFP in WT neurons implied the presence of a ceiling effect. Saturation of PANX1mediated increases in dendritic protrusion movement may have resulted from limited availability of additional machinery for proper PANX1 trafficking and/or auto-regulation via ATP-dependent internalization (i.e., increase in PANX1 leading to increase in ATP release feeding back to internalize PANX1; Boyce et al., 2015;Boyce and Swayne, 2017). Alternatively, effects of supplementary PANX1 were perhaps constrained by limited availability of interacting partners or saturation of downstream . These examples are cropped from the full regions of analysis from primary neurites. B, Effect of PANX1 expression on dendritic protrusion formation, elimination, lability, and motility in WT and Panx1 KO using Cumming estimation plots. Bi, Dendritic protrusion formation was significantly higher in Panx1 KO neurons transiently expressing PANX1-EGFP compared with those expressing EGFP (KO-EGFP: 0.2 6 0.1%, KO-PANX1-EGFP: 4.6 6 1.3%, p = 0.0028, Kruskal-Wallis test c1 ). No significant differences were observed between genotypes in EGFP-expressing neurons (WT-EGFP: 1.7 6 0.7%; Panx1 KO-EGFP: 0.2 6 0.1%, p = 0.2267, Kruskal-Wallis test c1 ). Bii, Similarly, only transient expression of PANX1-EGFP in Panx1 KO neurons increased dendritic protrusion elimination (KO-EGFP: 0.3 6 0.15%; KO-PANX1-EGFP: 4.6 6 1.28%, p = 0.00024, Kruskal-Wallis test c2 ). No significant differences were found between EGFP and PANX1-EGFP expressing WT cells (p = 0.62307 c2 ). Biii, Dendritic protrusion lability was higher in Panx1 KO neurons transfected with PANX1-EGFP (KO-EGFP: 2.1 6 0.5%; KO-PANX1-EGFP: 9.4 6 1.7%, p = 0.0034, Kruskal-Wallis test c3 ), beyond that observed in WT expressing EGFP control (p = 0.0291, Kruskal-Wallis test c3 ). Transient expression of PANX1-EGFP in WT neurons had no significant effects (p . 0.9999 c3 ). Biv, Dendritic protrusion motility was significantly reduced in Panx1 KO neuron expressing EGFP control (WT-EGFP: 20.5 6 2.3%; KO-EGFP: 7.2 6 1.3%, p = 0.00016, Kruskal-Wallis test c4 ). Transient PANX1-EGFP expression increased dendritic protrusion motility in Panx1 KO neurons only (KO-PANX1-EGFP: 17.4 6 2.8%, p = 0.03582, Kruskal-Wallis test c4 ). N = cells, all analyzed cells were obtained from three independent cultures. Effect sizes are reported in the main text and Table 2. Red arrowheads on the y-axis on the bottom panel of Cumming estimation plots represent WT-EGFP means. d.p., dendritic protrusion; ppp,0.001, pp,0.01, p,0.05. This figure was modified from the PhD thesis of J.C.S.-A, University of Victoria (Sanchez-Arias, 2020), found at http://hdl.handle.net/1828/11714. purinergic signaling pathways (Abbracchio et al., 2009;Bhalla-Gehi et al., 2010;Wicki-Stordeur and Swayne, 2013;Dahl, 2015;Yang et al., 2015;Xu et al., 2018). PANX1 regulation of dendritic protrusion stability could underlie neuronal network and synaptic plasticity changes associated with PANX1 disruption. For example, Rafael et al. (2020) recently reported that Panx1 KO hippocampal neurons treated with tetrodotoxin (TTX) for 36 h (in a model of "chronic inactivity") did not scale up their synapses accordingly, hinting that loss of PANX1 altered homeostatic plasticity mechanisms. Interestingly, the lack of synaptic scaling was seen both in Panx1 KO neuron-glial co-cultures as well as Panx1 KO neurons co-cultured with WT glia. On the other hand, WT neuron-glial co-cultures and WT neurons co-cultured with Panx1 KO glia scaled up their synapses in the presence of chronic inactivity induced by TTX treatment. These findings suggest that neuronal, but not glial, PANX1 plays a role in the regulation of synaptic homeostatic plasticity and raises the possibility that increased stability of dendritic protrusions in Panx1 KO could be a contributing factor to these altered synaptic plasticity adaptations (Holtmaat and Svoboda, 2009;O'Donnell et al., 2011;Yin and Yuan, 2015). Future studies should aim to elucidate the molecular mechanism underlying PANX1 regulatory role(s) in the maturation and maintenance of excitatory synapses throughout development and in the context of neuron-neuron (excitatory-inhibitory balance) and neuron-glial interactions (astrocytes, oligodendrocytes, microglia).

PANX1 regulation of dendritic spines is consistent with its role in synaptic plasticity and neurodevelopment
Our findings provide important new insight into recent studies demonstrating that PANX1 regulates brain and cognitive development. Panx1 KO has been associated with altered hippocampal plasticity (LTP facilitation and LTD deficits), impaired memory flexibility, and larger cortical neuron network ensembles and spine densities (Prochnow et al., 2012;Ardiles et al., 2014;Gajardo et al., 2018;Sanchez-Arias et al., 2019). Dendritic spine abnormalities have been observed in several neurodevelopmental disorders (Forrest et al., 2018;Lima-Caldeira et al., 2019). Brain enriched, autism spectrum disorders (ASDs)associated single nucleotide polymorphisms (SNPs) have been associated with changes in PANX1 expression levels (Davis et al., 2012). A germline PANX1 SNP was associated with intellectual disability (Shao et al., 2016). PANX1 is also linked to neurodevelopmental disorders indirectly through its interacting partner CRMP2 and its downstream signaling effectors and interactors, purinergic receptors (Boyce et al., 2015;Boyce and Swayne, 2017; for review, see Swayne and Boyce, 2017). CRMP2 auto-antibodies have been implicated in ASD (Braunschweig et al., 2013), and brain-specific deletion of the gene encoding for CRMP2 resulted in decreased dendritic spine density and reduced LTP (Zhang et al., 2016), while suramin treatment corrected synaptic and behavioral phenotypes in the Fmr1 KO mouse (Naviaux et al., 2013(Naviaux et al., , 2015(Naviaux et al., , 2017.

Limitations
Immature neurons contain mainly filopodia-like structures and immature spines that lack fully developed postsynaptic densities (PSDs) (Berry and Nedivi, 2017). In order to facilitate observation of a substantial number of dendritic protrusions simultaneously, we selected a single immature time point (DIV10). We used this strategy because dendritic protrusions are the precursors to mature spines, although it should be noted that not all protrusions develop into spines. With this in mind, an important extension of the current study will be to assess synaptic markers within the same cells in a longitudinal approach (i.e., over multiple days), as this would enable tracking of dendritic protrusions that eventually develop into spines at a more mature DIV, as well as, retrospective grouping of our dendritic protrusion dynamics data based on the "fate" of the dendritic protrusions (i.e., attributes of those that become spines vs those that do not). Somewhat surprisingly, Panx1 KO cortical neurons already exhibited increased dendritic protrusion density at this relatively immature analysis time point, suggesting that PANX1 begins to regulate dendritic protrusions, the precursors to dendritic spines, early in neuronal development. Previous work revealed a proportional increase in PSD-95positive spines associated with Panx1 KO at a more mature DIV (DIV12), suggesting the increase in dendritic protrusion density associated with Panx1 KO indeed leads to a corresponding increase in dendritic spine density. In addition to facilitating tracking the fate of the dendritic protrusions, assessment of PSD-95 trafficking into spines, as a readout of maturation, would also permit simultaneous investigation of the mechanistic underpinnings of PANX1 regulation of dendritic spine formation. Spine formation relies on microtubule dynamics (Hu et al., 2011), and CRMP2, a PANX1 interacting protein, is a well-known regulator of microtubule dynamics as well as dendritic spine formation (Jin et al., 2016;Zhang et al., 2016). Additionally, it will be important to examine functionally similar proteins or mechanisms that could compensate for the loss of PANX1, such as changes in expression levels of other pannexins, functionally related connexins or purinergic receptors. Future pharmacological and genetic experiments disrupting PANX1 and its interactome will continue to enhance our understanding of PANX1 regulation of neuronal development.
There are some additional limitations associated with our KO model that are important to note. The original Panx1 f/f strain was developed on a 129 Â 1/SvJ background, and Panx1 KO was generated via a cross with a CMV-Cre line , which were then backcrossed over 10 generations onto the C57Bl/6J background. The original strain contained a loss of function mutation in Casp4 (Casp11) and was predicted to have variants in Mmp1a, Olfr832, Fbxl12, ENSMUSG00000095186, and ENSMUSG00000095891 (Vanden Berghe et al., 2015). We cannot rule out the possibility that these passenger mutations could have been present in the global Panx1 KO and not their control littermates; however, we also compared a conditional Panx1 KO (Panx1 f/f ;Emx1 IRES-Cre ) strain with control Panx1 f/f littermates (i.e., same genetic background in conditional KO and control littermates), which yielded the same result in terms of dendritic spine densities, suggesting that this phenotype can be attributed to deletion of Panx1. Furthermore, transient expression of PANX1-EGFP rescued the change in dendritic protrusion density associated with Panx1 KO, further supporting our findings from Panx1 KO mouse that PANX1 inversely correlates with regulation of dendritic spine formation.

Conclusions
In summary, this work has significantly advanced our understanding of the role of PANX1 in dendritic spine development, identifying it as a regulator of dendritic protrusion dynamics. PANX1 promotes dendritic protrusion movement and is associated with greater dendritic protrusion instability. The role of PANX1 in the regulation of this process that relies inherently on complex cytoskeletal dynamics is consistent with its previously identified interactions with ARP2/3 and CRMP2, known to control both cytoskeletal dynamics and dendritic spine development. Moreover, these novel findings presented here also align with the marked downregulation of cortical PANX1 expression levels in mice that coincides with spine formation during the first four postnatal weeks. Together, these findings reveal an important role for PANX1 in regulating dendritic protrusion dynamics in developing cortical neurons.