Carbonic anhydrase VII regulates dendritic spine morphology and density via actin filament bundling

Intracellular pH is a potent modulator of neuronal functions. By catalyzing (de)hydration of CO2, intracellular carbonic anhydrase (CAi) isoforms CAII and CAVII contribute to neuronal pH buffering and dynamics. The presence of two highly active isoforms suggests that they form spatially distinct CAi pools enabling subcellular modulation of pH. Here we show that CAVII, unlike CAII, is localized to the filamentous actin network, and its overexpression induces formation of thick actin bundles and membrane protrusions in fibroblasts. In neurons, CAVII is enriched in dendritic spines, and its over-expression causes aberrant spine morphology. We identified amino acids unique to CAVII that are required for direct actin interactions, promoting actin filament bundling and spine targeting. Lack of CAVII in neocortical neurons leads to reduced spine density and increased proportion of small spines. Thus, our work demonstrates highly distinct subcellular expression patterns of CAII and CAVII, and a novel, structural role of CAVII.

2 Abstract 24 25 Intracellular pH is a potent modulator of neuronal functions. By catalyzing (de)hydration of CO2, 26 intracellular carbonic anhydrase (CAi) isoforms CAII and CAVII contribute to neuronal pH buffering and 27 dynamics. The presence of two highly active isoforms suggests that they form spatially distinct CA i pools 28 enabling subcellular modulation of pH. Here we show that CAVII, unlike CAII, is localized to the 29 filamentous actin network, and its overexpression induces formation of thick actin bundles and 30 membrane protrusions in fibroblasts. In neurons, CAVII is enriched in dendritic spines, and its over-31 expression causes aberrant spine morphology. We identified amino acids unique to CAVII that are 32 required for direct actin interactions, promoting actin filament bundling and spine targeting. Lack of 33 CAVII in neocortical neurons leads to reduced spine density and increased proportion of small spines. 34 Thus, our work demonstrates highly distinct subcellular expression patterns of CAII and CAVII, and a 35 novel, structural role of CAVII.

7
As CAVII-expressing fibroblasts often showed abnormally thick actin structures, we set out to test if CAVII 119 modulates actin bundling. To this end, we used an in vitro actin bundling assay in the absence and presence of 120 1.12 µM mCAVII ( Figure 2B and Figure 2, Videos 1 and 2). Both actin filament length and fluorescence intensity 121 of their cross section increased over time significantly in the presence of mCAVII when compared to the vehicle 122 control (P < 0.001 for both length and intensity) ( Figure 2C) indicating that CAVII had a profound effect on three-123 dimensional actin structures. 124 The ability of CAVII to increase bundling by crosslinking actin filaments implies a bivalent binding mechanism that 125 conventional actin cross-linkers, such as a-actinin and fimbrin, most often achieve through homodimerization 126 These data demonstrate that CAVII directly binds to and bundles F-actin, and that CAVII over-expression stabilizes 139 existing actin filaments. that, despite these similarities, CAVII has unique characteristics, profiling it as a novel, multifunctional protein 240 within the CNS. We demonstrate a prominent difference in the subcellular localization of the isoforms, which is 241 based on a pH-dependent interaction of CAVII and actin. In neurons, CAVII is enriched in the actin-dense dendritic 242 spines and affects the actin cytoskeleton thus altering both dendritic spine morphology and density. 243 244

The subcellular distribution of CAVII is dictated by its interaction with F-actin 245
In cultured fibroblasts, expression of dsRed-and EGFP-fusion constructs of CAVII and CAII revealed a mutually 246 exclusive subcellular localization. CAVII is present in the immediate vicinity of actin filaments, in contrast to the 247 diffuse cytoplasmic localization of CAII. In hippocampal neuronal cultures, as well as in cortical neurons 248 transfected in vivo, CAVII shows a preferential spine-targeted localization, whereas CAII distributed evenly in the 249 cytosol over the somato-dendritic axis. The diffuse cytoplasmic distribution of CAII is well in line with previous 250 localization results from non-erythroid cells (Wang et al., 2002;Stridh et al., 2012;Al-Samir et al., 2013) and fits 251 with the idea that this ubiquitous and high-activity isoform serves a housekeeping role in cytosolic pH buffering. 252 Homogenously distributed, soluble CA can efficiently dissipate cytosolic acid-base gradients (Voipio, 1998;253 Stewart et al., 1999;Boron, 2010). Interestingly, a recent study on cardiac myocytes shows that the majority of 254 nuclear pH buffering is sourced from the cytoplasm in the form of mobile buffers (Hulikova and Swietach, 2015), 255 motivating further work on the localization of CAII and its possible role in pH-dependent regulation of 256 transcription (Bumke et al., 2003;Neri and Supuran, 2011). 257 13 Our main finding is that the subcellular distribution of CAVII is dictated by its binding to actin, in particular to F-259 actin, which is a novel property with regard to all cytosolic CAs studied to date. Furthermore, CAVII interacts only 260 with a specific subset of actin filaments. EGFP-CAVII strongly co-localized with F-actin in fibroblast stress-fibers 261 and neuronal dendritic spines, but the edges of the highly dynamic fibroblast lamellipodia, consisting mainly of 262 branched actin, were largely devoid of CAVII. Given the key role of CAs in the modulation of pH, a particularly 263 interesting finding is that the CAVII-actin interaction is pH-sensitive and enhanced at acidic pH (6.5 vs. 7.4). 264 Compared with the actions of the previously known pH-sensitive actin binding proteins (see below), CAVII could 265 thus counteract gelsolin-dependent severing of F-actin, which takes place upon acidification (Lagarrigue et al., 266 2003). Together with our latrunculin data these results indicate that CAVII could stabilize actin structures when 267 the intracellular compartment is acidified, which occurs under various pathophysiological conditions such as 268 stroke (Pavlov et al., 2013) and epilepsy (Siesjö et al., 1993). Furthermore, an increase in neuronal activity 269 subjects brain cells to surges of lactate produced by the glycolytic pathway (see Yellen, 2018 2). The rapid assembly/disassembly of actin filaments seen under control conditions was suppressed in the 285 presence of CAVII and the generated actin-bundles were thicker and more stable. Based on the SEC-MALLS 286 results, CAVII-dependent bundling may be achieved, at least partly, through CAVII homodimerization. In the 287 intracellular milieu, even a small proportion of free dimers might be enough if actin acts as a sink for the CAVII 288 dimers thereby shifting the dimerization process to the right. 289 290

CAVII-actin interaction induces morphological changes 291
The changes detected in the actin cytoskeleton of CAVII-expressing cells are consistent with the biochemical 292 assay data and show that CAVII modulates higher-order actin structures in the cytoplasm. Fibroblasts transfected 293 with CAVII generate numerous filopodia-like protrusions projecting from the cell surface, and have thick, 294 sometimes curving, cytosolic stress fibers. In neurons overexpressing CAVII, both in cell cultures and in vivo, 295 dendritic spines eventually lose their morphological diversity (categorized as thin, stubby, and mushroom spines 296 according to (Bourne and Harris, 2008; Hotulainen and Hoogenraad, 2010)) and turn into thick protrusions, which 297 lack a clear spine head. Similar filopodia structures sprout even from the neuronal cell body. These phenotypic 298 characteristics closely resemble those seen with the brain-specific actin-bundling protein drebrin-A in fibroblasts 299 (Shirao et al., 1994) and in cultured neurons (Hayashi and Shirao, 1999;Mizui et al., 2005). The disturbed spine 300 morphology shows that the actin network normally forming these structures is modified to more rigid actin 301 bundles. Together with the biochemical bundling assay results, the CAVII overexpression phenotype suggests 302 that CAVII has a stabilizing effect on F-actin. We tested this hypothesis in experiments where fibroblasts were 303 exposed to the actin polymerization inhibitor latrunculin B. Compared to control cells, fibroblasts expressing 304 CAVII maintained F-actin structures significantly longer, confirming a direct stabilizing effect of CAVII. 305 306

Identification of the CAVII -actin interaction site 307
When Montgomery et al. (1991) first identified the human CAVII, they recognized several poorly conserved 308 regions that were predicted to "be located towards the surface of the protein". Our work shows that one of these 309 regions, residues 232 -248 encoded by exon 7, is critically involved in CAVII-actin interaction. When we replaced 310 amino acids 237 -242 (DDERIH) with corresponding amino acids from CAII either alone (EGFP-CAVII-mutant3) 311 or together with additional mutations (amino acids 101-105, 113 and 115; EGFP-CAVII-mutant1), co-localization 312 of F-actin and the mutated EGFP-CAVII proteins decreased compared to EGFP-CAVII. Replacing the corresponding ). The present study shows that CAVII is optimally localized not only to separately 360 modulate, but also to provide a link between F-actin dynamics and activity-dependent pH transients, within 361 spines, thereby identifying a novel mechanism of morphofunctional plasticity. respectively. Constructs containing full-length CAII and CAVII (human isoform 1) coding sequences were obtained 385 from ImaGenes (human CAII and human CAVII including start and stop codon; OCAAo5051H1054 and 386 OCAAo5051E0588, respectively) and GeneCopoeia (human CAII without stop codon and mouse CAVII including 387 start and stop). Constructs encoding CAVII-R223E and CAVII-H96/98C were generated using site-directed 388 mutagenesis (Phusion high fidelity PCR, ThermoFisher) and the correct sequence of PCR amplified sequences 389 was confirmed by full-length sequencing of both strands (DNA Sequencing and Genomics Laboratory, Institute 390 of Biotechnology, Helsinki). More complex mutants containing multiple nucleotide exchanges were commercially 391 synthesized (GenScript). All coding sequences were either available as Gateway entry vectors or subcloned into 392 pDONR or pENTR vectors using the Gateway technology (LifeTechnologies). Expression constructs encoding N-393 or C-terminal fusion proteins of CAII or CAVII and various reporter proteins (EGFP, dsRed, mCherry) were 394 generated using the Gateway technology and appropriate destination vectors. To allow for stable expression in phalloidin-488. We made three independent replicates of such experiments. A hundred transfected cells from 447 each time point from each experiment were categorized either as "normal", "some shape/F-actin left" or "round" 448 (example cells for the three categories are depicted in Figure 3A,B, upper panels). 449 450 CAVII biochemistry: pull down assay, measure of enzymatic activity, bundling (In vitro TIRF) assay Actin co-451 sedimentation assay was carried out in 20 mM Hepes pH 7.4/6.5, in the presence of 0.2 mM DTT. Mouse CAVII 452 (produced as a secreted protein with a C-terminal His-tag in the CHOEBNALT85 cell line) was stored in PBS but 453 the buffer was changed to Hepes (pH 7.4/6.5) before the experiment. Lyophilized powder of CAII (Sigma) was 454 reconstituted in MilliQ and diluted in Hepes-buffer pH 7.4/6.5 to 33 μM. ZnCl2 (1 μM) was added to CAVII/CAII 455 one hour before incubation with actin. β/γ-G-actin (0, 1, 5, 10 and 15 μM) was pre-polymerized in Hepes-buffer 456 pH 7.4/6.5 by addition of 1/10 of 10x-initiation mixture (1 M KCl, 10 mM EGTA, 50 mM MgCl2, 2.5 mM ATP and 457 20 mM Hepes pH 7.4/6.5) for 30 min at room temperature. CAVII or CAII (1 μM) was added to polymerized actin, 458 gently mixed and incubated for another 30 min at room temperature. Actin filaments were sedimented by 459 22 centrifugation for 30 minutes at 20S o C in a Beckman Optima MAX Ultracentrifuge at 353,160 × g in a TLA100 460 rotor. Equal proportions of supernatants and pellets were run on 13.5 % SDS-polyacrylamide gels, which were 461 stained with Coomassie Blue. The intensities of β/γ-actin and CAVII/CAII bands were quantified with QuantityOne 462 program (Bio-Rad), analyzed and plotted as CAVII/CAII bound to actin (μM, CAVII/CAII in pellet) against actin. 463 The mCAVII-actin co-sedimentation assay was repeated three times for each pH value and averaged curves were 464 presented (± SEM). 465 In vitro TIRF imaging was performed as previously described (Suarez et  Slices were then rinsed in PBS solution and incubated for an additional 24 h with Alexa conjugated secondary 503 antibodies (Invitrogen; 1:1,000). After mounting the slices were coverslipped using Immumount (Thermo 504 Scientific, Pittsburgh, PA), and stored at +4 °C until analysis. 505

Confocal Laser Scanning Microscopy and Image Analysis
Second order dendrites were imaged for spine analysis 506 using LSM700 confocal microscope and 63× oil-immersion objective. Spine analysis was performed on acquired 507 24 stacks of images using a homemade plug-in written for OsiriX software (Pixmeo, Geneva, Switzerland). This plug-508 in allows precise spine quantification, individual tagging, and measurement in 3D by scrolling through the z-axis. 509 We defined spines as structures emerging from the dendrites that were longer than 0.4 µm and for which we 510 could distinguish an enlargement at the tip (spine head). Spines head diameters were measured at their largest 511 width in xy-axis on the z-image corresponding to the central axis of the spine head. The difference in spine head 512 width distribution between WT and CAVII KO mice was analyzed using a Wilcoxon rank sum test with continuity 513 correction. Note that for illustration purposes, images presented in figures are maximum intensity projections of 514 z stacks with volume rendering, further treated with a Gaussian blur filter.

A B
C A V I I C A I I C A V I I -m u t a n t 2 C A V I I -R 2 2 3 E C A V I I -H 9 6 / 9 8 C C A V I I -m u t a n t 3 C A V I I -m u t a n t 1