Calponin-3 is critical for coordinated contractility of actin stress fibers

Contractile actomyosin bundles, stress fibers, contribute to morphogenesis, migration, and mechanosensing of non-muscle cells. In addition to actin and non-muscle myosin II (NMII), stress fibers contain a large array of proteins that control their assembly, turnover, and contractility. Calponin-3 (Cnn3) is an actin-binding protein that associates with stress fibers. However, whether Cnn3 promotes stress fiber assembly, or serves as either a positive or negative regulator of their contractility has remained obscure. Here, we applied U2OS osteosarcoma cells as a model system to study the function of Cnn3. We show that Cnn3 localizes to both NMII-containing contractile ventral stress fibers and transverse arcs, as well as to non-contractile dorsal stress fibers that do not contain NMII. Fluorescence-recovery-after-photobleaching experiments revealed that Cnn3 is a dynamic component of stress fibers. Importantly, CRISPR/Cas9 knockout and RNAi knockdown studies demonstrated that Cnn3 is not essential for stress fiber assembly. However, Cnn3 depletion resulted in increased and uncoordinated contractility of stress fibers that often led to breakage of individual actomyosin bundles within the stress fiber network. Collectively these results provide evidence that Cnn3 is dispensable for the assembly of actomyosin bundles, but that it is required for controlling proper contractility of the stress fiber network.

features in this motif, and hence Cnn1 is also known as basic calponin, Cnn2 as neutral calponin, and Cnn3 as acidic calponin 6,7,9 . The best characterized calponin isoform is Cnn1, which is mainly restricted to smooth muscle cells. A wealth of biochemical and in vivo studies provided evidence that Cnn1 controls smooth muscle contractility by inhibiting actin-activated myosin ATPase activity without affecting phosphorylation of the myosin regulatory light chain (MLC) 5,[10][11][12] . Similarly, genetic studies on the C. elegans homologues of calponin demonstrated that they function as negative regulators of actomyosin contractility in vivo 13 . However, other studies provided evidence that Cnn1 may also affect the mechanical properties of actin filaments and regulate organization of the actin cytoskeleton in smooth muscle cells 14,15 .
Cnn2 and Cnn3 are widely expressed in non-muscle and smooth muscle cells [16][17][18][19][20][21][22][23] . Cnn2 knockout mice are viable, but display defects in the proliferation and migration of macrophages, neutrophils and monocytes 24,25 . In contrary, Cnn3 is essential for embryonic development and viability of mice 23,26 . The precise functions of Cnn2 and Cnn3 in non-muscle cells remain somewhat controversial. Both proteins localize to contractile actomyosin bundles, such as stress fibers 16,21,27 . Similar to Cnn1 in smooth muscle cells, Cnn2 depletion results in increased traction forces, suggesting that this protein is a negative regulator of actomyosin activity 18 . In contrast, knockdown of Cnn3 in primary fibroblasts was reported to reduce contractility, most likely due to impaired stress fiber assembly 27 . Cnn3 was also shown to regulate cell contractility through MEKK-1 signaling pathway in MLC independent manner 28 . Other studies provided evidence that Cnn2 and Cnn3 may stabilize actin filaments, and thus control the organization of the actin cytoskeleton 16,27,29 . Here, we applied the U2OS cell model system to reveal whether Cnn3 primarily controls organization of the actin cytoskeleton, or functions as a regulator of actomyosin contractility. We show that Cnn3 is a dynamic component of stress fibers, and that its depletion results in increased and uncontrolled contractility of the stress fiber network. These results provide evidence that, like Cnn1, also Cnn3 functions as a negative regulator of actomyosin contractility in cells.

Results
Calponin-3 is a dynamic component of contractile stress fibers. Calponin-3 (Cnn3) localizes to actin stress fibers ( Fig. 1A) 1,28,30 , but its precise localization pattern within the stress fiber network has not been reported. Therefore, we applied 3D structural-illumination microscopy (3D-SIM) imaging to study the localization of Cnn3 in human osteosarcoma (U2OS) cells. U2OS cells have been extensively used as a model system for studying the assembly and contractility of stress fibers (e.g. [31][32][33][34][35]. The conventional immunofluorescence microscopy and 3D-SIM experiments demonstrated that both endogenous Cnn3 (Fig. 1A,B) and exogenously expressed Cnn3-mCherry (Fig. 1C) localize to all three stress fibers categories; dorsal stress fibers, transverse arcs, and ventral stress fibers. 3D-SIM experiments further revealed that in U2OS cells Cnn3 is not uniformly distributed along stress fibers, but it is enriched in α-actinin-1 foci in transverse arcs and ventral stress fibers (Fig. 1C,D). Consistent with these data, Cnn3 is also present in non-contractile dorsal stress fibers that are enriched with α-actinin-1, but do no not contain myosin II (Suppl. Fig. 1A).
Actin-binding proteins display large differences in the dynamics of their stress fiber associations. For example, VASP and palladin display highly dynamic (half-time ~1 s) association with stress fibers, whereas α-actinin-1 displays intermediate dynamics (half-time ~2-5 s), and tropomyosins exhibit more stable (half-time ~5-10 s) association with stress fibers 32,36 . To reveal the dynamics of Cnn3 in stress fibers, we performed fluorescence-recovery-after-photobleaching (FRAP) experiments on cells expressing GFP-Cnn3, and monitored the recovery of GFP-Cnn3 fluorescence on mature, contractile ventral stress fibers. This analysis revealed that Cnn3 exhibits a half-life of approximately ~2-4 s on stress fibers ( Fig. 2A,B), and hence displays similar turnover on stress fibers with α-actinin-1 (Fig. 2B). Together these data demonstrate that Cnn3 is a dynamic stress fiber component, which co-localizes with α-actinin-1 in all three stress fiber categories of U2OS cells.
Calponin-3 is dispensable for stress fiber assembly. Previous RNAi studies suggested that Cnn3 is important for the assembly of stress fibers in dermal fibroblasts 27 . To examine the possible role of Cnn3 in stress fiber assembly in osteosarcoma cells, we generated Cnn3 knockout U2OS cells by CRISPR/Cas9. Sanger sequencing of the obtained Cnn3 knockout cell clone revealed disruption of the genomic region of CNN3 (Suppl. Fig. 2A). Moreover, MiSeq next generation sequencing of genomic DNA from CNN3 knockout cells confirmed that they were unable to synthetize functional Cnn3. We found three distinct CRISPR/Cas9-induced effects, all altering the reading frame, either by deletion of 4 nucleotides, or insertion of 1 and/or 2 nucleotides (Suppl. Fig. 2B). This indicates that U2OS osteosarcoma cells have three copies of CNN3 gene, most likely due to aneuploidy. Each variant in the Cnn3 knockout cell-line resulted in premature termination of translation, as stop codons were introduced in the exon 2 (Suppl. Fig. 2C). Finally, complete absence of Cnn3 protein was confirmed from the knockout cells by Western blotting using Cnn3-specific antibody (Suppl. Fig. 3B).
By visualizing actin filaments from control and Cnn3 knockout cells by phalloidin staining, we revealed that Cnn3 knockout U2OS cells were still able to generate all three categories of stress fibers (Fig. 3A). However, in many cells the morphology of the stress fiber network was altered. This is because actin bundles were often thinner and less prominent compared to the control cells, and the arrangement of stress fiber networks in Cnn3 knockout cells was somewhat less regular compared to the control cells. However, α-actinin-1, which co-localizes with Cnn3 in stress fibers and displays very similar dynamics with Cnn3 within the stress fiber network, still accumulated to stress fibers in Cnn3 knockout cells and displayed indistinguishable dynamics compared to the control cells ( Fig. 2A,C; Suppl. Fig. 1B,C). Thus, the defects in the organization of the stress fiber network in Cnn3 knockout cells do not arise from altered localization or dynamics of α-actinin-1 in the absence of Cnn3.
We also applied vinculin staining to uncover the possible effects of Cnn3 depletion on focal adhesions. By quantifying the numbers and sizes of focal adhesions in control and Cnn3 knockout cells, we revealed that the average number of focal adhesions was very similar in control and Cnn3 knockout cells, but that the average areas of focal adhesions in Cnn3 knockout cells were significantly smaller compared to the control cells (Fig. 3B). To confirm that the observed phenotypes do not result from CRISPR/Cas9 off-target effects, we also silenced Cnn3 from U2OS cells by RNAi (Suppl. Fig. 3A,B). Phalloidin staining and focal adhesion analysis revealed very similar effects of Cnn3 depletion by siRNA on the organization of the stress fiber network and average sizes of focal adhesions and number compared to the Cnn3 knockout cells (Fig. 3B,C; Suppl. Fig. 3C). Thus, Cnn3 is dispensable for stress fiber assembly, at least in U2OS cells. Loss of calponin-3 leads to uncontrolled contractility of the stress fiber network. Although cells lacking Cnn3 were able to generate stress fibers, the stress fiber networks in Cnn3 knockout and knockdown cells typically displayed abnormalities in their organizations. To gain insight into these stress fiber network organization defects, we performed live-imaging of control and Cnn3-deficient (both knockout and knockdown) U2OS cells, transiently transfected with a plasmid expressing GFP-actin. Importantly, these experiments revealed major disruption of stress fiber network contractility in Cnn3-deficient cells. Compared to control cells, which displayed 'smooth' retrograde flow of transverse arcs and balanced contractility of ventral stress fibers (Supplemental Movie 1), the Cnn3 knockout and knockdown cells typically exhibited abnormal, disorientated tug-of-war behavior of stress fibers (Supplemental Movies 2 and 3). This uncoordinated contractility was often accompanied by breakage of the stress fibers (Fig. 4A,B). By quantifying the numbers of stress fiber breakage events in control as well as Cnn3 knockdown and knockout cells from 30 min live-cell movies, we revealed that frequency of stress fiber breakage events was increased >10-fold in Cnn3 deficient cells compared to the control cells (Fig. 4C). We next applied pillar-based traction force measurements combined with live cell imaging to elucidate whether the abnormal behavior and breakage of stress fibers in Cnn3 knockout cells results from increased/ unregulated contractility of stress fibers (Fig. 5A,B). Pillar displacements were significantly greater in Cnn3 knockout cells compared to control cells, resulting in a larger traction stress per cell (Fig. 5C-E), which was due to a larger fraction of pillars under the cell that showed displacements above the noise level (30 nm). Breakage events were also observable on nanopillars. Because of the breakages occurring in a network of interlinked stress fibers, it was not possible to track the forces back to individual adhesions on pillars. Nevertheless, we frequently observed a buildup of traction forces prior to the breakage and a drop of forces on individual pillars close to the breakage point after the event (Suppl. Fig. 4). Therefore the results of the pillar experiments further underline the hypothesis that the breakages occur due to uncoordinated contractility, possibly enhanced through a structural instability of the thinner stress fibers in the Cnn3 knockout cells.

Discussion
Several studies have shown that Cnn1 and Cnn2 function as negative regulators of actomyosin ATPase activity in vitro and myosin II -driven contractility in cells (e.g. 10,18,37 ). In contrast, the ubiquitously expressed Cnn3 was proposed to regulate the assembly of actin stress fibers in fibroblasts, and therefore have a positive impact on myosin II-driven contractility of non-muscle cells 27 . Here, we examined the role of Cnn3 in the assembly and contractility of stress fibers in U2OS cells, which have been extensively used as a model system for determining the cellular functions of cytoskeletal proteins. Several actin-binding proteins display specific localization patterns within the stress fiber network. For example, palladin and vasodilator stimulated phosphoprotein (VASP) are enriched in non-contractile dorsal stress fibers, whereas non-muscle myosin II isoforms and certain tropomyosins (e.g. Tpm4.2) localize only to transverse arcs and ventral stress fibers, but are absent in dorsal stress fibers 32,36 . Importantly, our data revealed that Cnn3 is present in all three categories of stress fibers, where it co-localizes with α-actinin-1. Moreover, Cnn3 displays similar rapid dynamics on stress fibers as α-actinin-1. However, it does not appear to function as a regulator of α-actinin-1, because Cnn3 depletion did not affect the subcellular localization of α-actinin-1 or its dynamics on stress fibers.
Recent RNAi studies provided evidence that Cnn3 is critical for stress fiber assembly in dermal fibroblasts 27 . However, our CRISPR/Cas9 knockout and RNAi knockdown studies revealed that at least in U2OS cells Cnn3 is not required for stress fiber assembly, as all three categories of stress fibers were still present in Cnn3-depleted cells. Nevertheless, the morphology and especially the dynamic behavior of the stress fiber networks in Cnn3-deficient cells was abnormal. This is particularly evident from live-cell imaging experiments, which revealed peculiar tug-of-war behavior of stress fibers in Cnn3-depleted cells (Supplemental Movies 2 and 3), as well as increased physical breakage of individual stress fibers with the network (Fig. 4B,C). Traction-force microscopy experiments demonstrated that many stress fibers in Cnn3 knockout cells displayed higher contractility compared to the stress fibers of control cells, and that these increased tractions led to breakage of stress fibers. Although the contractility of stress fibers in Cnn3-depleted cells was increased, the average focal adhesions areas were smaller. In this context it is important to note that, although increased traction forces are linked to enlargement of focal adhesions during their assembly, it was reported that tension and adhesion size do not strictly correlate in mature focal adhesions 38 . Therefore, the effects of Cnn3 depletion on the average focal adhesion areas may result from differences in maturation stages of the adhesions or from indirect effects of Cnn3 depletion on the composition of focal adhesions. Together, these data demonstrate that Cnn3 functions as an indirect negative regulator of myosin II-driven contractility in U2OS cells.
What is the molecular mechanism by which Cnn3 controls the contractility of stress fibers? As Cnn3 localizes to α-actinin-1 -rich z-line regions of stress fibers, it most likely does not directly form a complex with non-muscle myosin II to control its activity. Thus, we propose that Cnn3 may restrict the movement of non-muscle myosin II molecules along the actin filament in stress fibers. Alternatively, Cnn3 may affect the organization or molecular composition of stress fibers in the way that absence of Cnn3 would lead to de-regulated activity of myosin II or its motility along actin filaments within stress fibers. The latter option is supported by the fact that Cnn3 also localizes to dorsal stress fibers that do not contain myosin II.
Cnn3 was earlier shown to form a complex with ERK1/2 and PKCα kinases on stress fibers. In this case, the ERK1/2-Cnn3-PKCα complex would be released from stress fibers upon a stimuli, translocated to the plasma membrane where MEK activates ERK1/2, which subsequently moves back to actin filaments to phosphorylate caldesmon and consequently increase the myosin II-driven contractility 21 . Although this scenario may be important upon specific stimuli in cells, it is unlikely the mechanism by which Cnn3 affects stress fiber contractility in U2OS cells. This is because Cnn3 displays constant, rapid turnover (t 1/2 = 2-4 s) in stress fibers. Thus, it is difficult to imagine how the fast association/dissociation kinetics of Cnn3 at stress fibers could be linked to such multi-step translocation/phosphorylation pathway, which is expected to be considerably slower 39 . Therefore, although the activity of Cnn3 can be regulated by various kinase pathways in cells 21,28 , we propose that Cnn3 regulates stress fiber organization and contractility in a more direct manner (e.g. by controlling the organization of actin filaments within stress fibers or by regulating the activities or localization of other stress fiber components).
Collectively, our work reveals important new aspects on the cellular function of Cnn3. We show that (1). Cnn3 is a dynamic component of stress fibers, (2). Cnn3 co-localizes with α-actinin-1 in all categories of stress fibers, including the non-contractile dorsal stress fibers, (3). Cnn3 is not required for stress fiber assembly, and (4). Lack of Cnn3 leads to uncontrolled contractility of the stress fiber network. In the future, it will be important to examine the precise molecular function of Cnn3 in stress fiber contractility and organization in the context of its reported interaction partners, for example tropomyosins.
Cell Imaging. All immunostainings (for wide-field and 3D-SIM) were performed as described previously 1 .
CRISPR/Cas9 knockout. GFP-Cnn3 CRISPR/Cas9 construct and knockout cell line were generated as previously described 41 , with target sgRNA sequence: TCAACAAGGGCCCTTCCTAT. knockout was confirmed with Western Blot, and by sequencing of the genomic DNA. For sequencing, genomic DNA was isolated from wild-type and Cnn3 knockout U2OS cells, the sgRNA targeted area was amplified with 5′-CTCTGTAGCACCCAGTTGGA and 5′-TTCTGCAGTCACCAAACGG primers and sent to Sanger sequencing reaction first. Due to non-interpretable overlap of the multiple signals we also performed MiSeq next generation sequencing of the same locus. The total of around 200 000 reads were analyzed from different PCR products by Geneious software.
Tractin-tdTomato transfected WT and CNN3-KO U2OS cells were plated for 1 hour on PDMS nano-pillar arrays and imaged every 30 seconds for ≥20 minutes on a Nikon Eclipse Ti Inverted spinning disk confocal with a Yokogawa CSU-1 disk head, Andor Neo sCMOS camera and equipped with a Solent Scientific chamber with temperature and CO2 control. After imaging cells were trypsinized to recover the original pillar positions. Pillar displacements were analysed with imageJ, using the PillarTracker plugin. Final deformations and forces were calculated with Matlab and statistics and graphs were generated with Graphpad Prism 7.

Statistical analyses.
Statistical analyses and all charts, except pillar assay analysis, were done with Sigma Plot 11.0 software. Statistical differences in focal adhesion and stress fiber breakage analysis are assessed by the Mann-Whitney-Wilcoxon rank-sum test (MWW), and in pillar assay analysis by the Student t-test. The datasets generated and analyzed during the current study are available from the corresponding author on a request.