Colocation of Tpm3.1 and myosin IIa heads defines a discrete subdomain in stress fibres

Stress fibres are typically contractile structures that allow cells to respond to mechanical force, to remodel connective tissue and apply force to their external environment (Pellegrin and Mellor, 2007; Tojkander et al., 2012; Burridge and Wittchen, 2013). Contractile stress fibres are made up of bundles of actin filaments, crosslinked by α-actinin and non-muscle myosin IIa and/or IIb (isoforms of myosin II with the heavy chain encoded by MYH9 and MYH10, respectively), organised in a repetitive, sarcomere-like fashion (Hu et al., 2017; Tojkander et al., 2012). Tropomyosins form co-polymers with actin and determine the function of an actin filament in an isoform-dependent manner by altering its stability, affinity for other actin-binding proteins, and activity/engagement of myosins (Gunning et al., 2015). Various tropomyosin isoforms have been observed to localise to stress fibres, and the loss of individual tropomyosin isoforms compromises the entire structure (Tojkander et al., 2011). Given that the majority of actin filaments in cultured cells associate with tropomyosin, it is important to investigate the detailed organisation of tropomyosins in actin structures (Meiring et al., 2018).

High molecular weight (HMW) tropomyosin isoforms of Tpm1.6, 1.7 and 2.1 (encoded by TPM1 and TPM2; hereafter, Tpm1.6/1.7/2.1), are not conducive to myosin II activity in vitro (Gateva et al., 2017); however, they are critical to stress fibre stability (Tojkander et al., 2011; Prasad et al., 1993). However, studies have suggested that tropomyosin isoforms Tpm3.1 and 3.2 (two nearly identical isoforms encoded by TPM3 that are not distinguishable via antibody staining; hereafter denoted Tpm3.1/3.2) and Tpm4.2 (encoded by TPM4) recruit (Bryce et al., 2003; Tojkander et al., 2011) and support the activity of non-muscle myosin II in vitro (Gateva et al., 2017; Pathan-Chhatbar et al., 2018; Hundt et al., 2016), therefore these isoforms were predicted to colocalise with non-muscle myosin II striations.

U2OS cells were chosen for this study because stress fibres have been well characterised in these cells (Tojkander et al., 2011; Hotulainen and Lappalainen, 2006). U2OS cells contain three different types of stress fibres: dorsal, ventral and transverse arcs (Hotulainen and Lappalainen, 2006). Dorsal stress fibres are not contractile and serve to anchor transverse arcs to focal adhesions. Transverse arcs are contractile and fuse with dorsal stress fibres in order to produce ventral stress fibres (Hotulainen and Lappalainen, 2006). Ventral stress fibres are typically generated from several transverse arcs, reducing the likelihood that the sarcomere-like structures within them will be aligned. Transverse arcs show the best alignment of the sarcomere-like structures, therefore making this class of stress fibre optimal for examining molecular organisation. Since different tropomyosin isoforms impart different properties on actin filaments, these results imply that stress fibres are the product of actin filaments with different functions. However, confocal imaging has not been able to discern detailed localisation of tropomyosin isoforms within the context of the semi-sarcomeric structures in stress fibres (Schevzov et al., 2011; Tojkander et al., 2011).

Here, we use a super-resolution microscopy approach to dissect the organisation of tropomyosin isoforms in stress fibres. We show that Tpm isoforms have discrete periodic localisations along actin filaments and that Tpm3.1/3.2 colocalises with myosin II heads, but not with either myosin II tails or α-actinin. To confirm the periodic localisation, we generated a Tpm3.1/3.2–APEX2 mouse via CRISPR to identify Tpm3.1/3.2 on actin filaments using electron microscopy. By tagging endogenous Tpm3.1/3.2 with APEX2 in vivo, we show for the first time, the periodic localisation of an endogenous tagged tropomyosin on actin filaments using electron microscopy and tomography. These results allow us to postulate a new model for the organisation of a stress fibre. The model predicts that Tpm3.1/3.2 influences the association of myosin II with an actin filament. We test this relationship by removing Tpm3.1/3.2 from actin filaments using the compound ATM3507. This results in reduced association of myosin IIa with stress fibres.

The anti-tropomyosin antibodies rabbit δ/9d (against Tpm4.1 and 4.2; hereafter, Tpm4.1/4.2) was paired with either mouse monoclonal CG3 (against Tpm3.1/3.2) or Tm311 (against Tpm1.6/1.7/2.1), to investigate the overlapping localisation of different tropomyosins within stress fibres. Two-colour STED microscopy of transverse arcs demonstrated that the subgroups of tropomyosin isoforms exhibited little overlap and appeared to be organised in segregated micro domains (Fig. 1A,B). From the STED images, discrete staining patterns are also evident for the different isoforms; a punctate pattern for Tpm4.1/4.2, a striated pattern for Tpm3.1/3.2 and largely continuous staining for the HMW tropomyosins Tpm1.6/1.7/2.1 (Fig. 1A,B). Line-scan analysis confirms these differences in organisation (Fig. 1A,B). These different staining patterns are in line with hypothesised functions for the different isoforms in stress fibres, and highlight the possibility that these different isoforms play spatially distinct roles in modulating stress fibre function. This result also favours the proposal that actin filaments are coated homogenously by either a single tropomyosin isoform or a subgroup of tropomyosin isoforms in cells (Bryce et al., 2003; Gunning et al., 2015; Gateva et al., 2017).

We predicted that if Tpm3.1/3.2 was regulating myosin II activity, then myosin II heads, but not tails would localise specifically with Tpm3.1/3.2. To test this, we compared the localisation of Tpm3.1/3.2 with non-muscle myosin IIa and IIb, the two myosin forms enriched in stress fibres. By using STED microscopy and antibodies that recognise the non-helical tailpiece of myosin IIa and IIb, we found that Tpm3.1/3.2 was in register with the tail domain of both myosin isoforms, but showed little overlap using line-scan analysis (Fig. 2A,B; Fig. S1A). This finding was consistent with our hypothesis, since the tail domain of myosin II does not interact with actin filaments, and hence would not be predicted to colocalise with Tpm3.1/3.2. Since there were no isoform-specific antibodies available that targeted the myosin II head region, we transfected cells with mEmerald–myosin IIa and mEmerald–myosin IIb which are N-terminally tagged, fluorescent non-muscle myosin II constructs that specifically label the head domain (Burnette et al., 2014). Since the mEmerald tag was incompatible with STED, we instead used Zeiss Airyscan semi-super-resolution imaging. Airyscan, while having a lower resolution than STED, has sufficiently high resolution to capture the myosin II head doublets in myosin II filaments (arrows, Fig. 2C,D). Tpm3.1/3.2 staining colocalised with both myosin IIa and IIb heads (Fig. 2C,D; Fig. S1A), as shown by line-scan analysis and supported by measurement of Manders’ coefficient of the Airyscan images (Fig. S1C,D). In order to confirm that the colocalisation of Tpm3.1 with myosin IIa and IIb heads was not a staining artefact, we also investigated the localisation of fluorescently labelled Tpm3.1 construct Ruby2–N-Tpm3.1, with respect to mEmerald–myosin IIa in live cells. Ruby2–N-Tpm3.1 is observed to colocalise with mEmerald–myosin IIa (arrows, Fig. S1B), consistent with our immunofluorescence data.

To gain further context for the localisation of Tpm3.1 in stress fibres relative to other stress fibre components, we also tested its localisation relative to the actin-bundling protein α-actinin, another major component of stress fibres. STED microscopy of co-immunostained samples revealed that Tpm3.1/3.2 did not overlap with α-actinin in transverse arcs (Fig. 2E). This was consistent with previous findings showing that tropomyosins are not compatible with α-actinin in in vitro actin-binding assays (Gateva et al., 2017).

The striated organisation of Tpm3.1/3.2 was confirmed in stress fibres by performing transmission electron microscopy (TEM). We produced an endogenously APEX2-tagged Tpm3.1 (Tpm3.1–APEX2) mouse (Fig. 3A), from which we harvested mouse embryonic fibroblasts (MEFs) that also produce actin stress fibres with distinct myosin IIa striations (Fig. 3B). Tpm3.1–APEX2 imaged via electron microscopy in heterozygous MEFs showed enrichment in bands across actin bundles (white arrowheads, Fig. 3C–E). Actin filaments that did not contain any Tpm3.1–APEX2 were also identified (black arrowheads, Fig. 3C). The average distance between the striations was 328 nm±123 nm (n=5 actin bundles) (Fig. 3F). Tomograms of actin bundles in homozygous MEFs (Movie 1) confirmed the presence of Tpm3.1–APEX2 striations and also the existence of intact actin filaments in the sections (Fig. 3G,H; Movie 2). Both TEM and tomography confirm the striated organisation of Tpm3.1/3.2 observed with fluorescence microscopy. Together, our results indicate that Tpm3.1 localises to discrete regions in stress fibres, overlapping with myosin IIa and IIb heads.

The simplest model we can derive from our microscopy data is one where bundles of actin filaments cross-linked by α-actinin are associated with Tpm3.1/3.2 in the regions where myosin IIa/IIb heads associate with actin filaments (Fig. 3I). This model is also supported by previous work showing that Tpm3.1 overexpression enhances non-muscle myosin IIa recruitment to stress fibres (Bryce et al., 2003), and in vitro studies demonstrating the ability of Tpm3.1 to stabilise non-muscle myosin II on actin filaments and enhance its activity (Gateva et al., 2017; Pathan-Chhatbar et al., 2018; Hundt et al., 2016).

Our colocalisation data and other previous studies (Bryce et al., 2003; Gateva et al., 2017) suggests that non-muscle myosin IIa and IIb associates with Tpm3.1/3.2-coated actin filaments. We therefore predicted that removal of Tpm3.1/3.2 from filaments may result in reduced non-muscle myosin II retention. While Tpm3.1/3.2 depletion disrupts the ability of cells to form stress fibres (Tojkander et al., 2011), recently characterised anti-Tpm3.1/3.2 drugs allow for more precise calibration of the magnitude of protein disruption. We therefore used an anti-Tpm3.1/3.2 drug, ATM3507 (Currier et al., 2017), at a concentration where Tpm3.1/3.2 localisation is disrupted without completely disassembling stress fibres, in order to assess the contribution of Tpm3.1/3.2 to myosin II colocalisation with stress fibres. The localisation of myosin IIa with stress fibres was diminished in drug-treated cells, although like Tpm3.1/3.2, some protein still colocalised with actin filaments (Fig. 4A). In contrast, the association of Tpm4.2, an isoform that is not impacted by ATM3507, with stress fibres was not disrupted with drug treatment (Fig. 4B). This is consistent with previous in vitro work indicating that Tpm3.1 enhances the activity of myosin II as well as its colocalisation with stress fibres (Pathan-Chhatbar et al., 2018; Gateva et al., 2017; Bryce et al., 2003). This is also consistent with the observation that Tpm3.1 recruitment to actin filaments precedes that of myosin II (Masedunskas et al., 2018).

In summary, we find that Tpm3.1/3.2 localise in discrete striations along stress fibres in both U2OS cells and MEFs. This is a clear demonstration that stress fibres do not consist of a homogeneous bundle of actin filaments but rather contain subdomains characterised by their tropomyosin isoform composition. We therefore suggest that stress fibres may be composed of different functional subdomains defined, at least in part, by their tropomyosin composition which provides an explanation of why stress fibres require five different tropomyosins for their assembly and/or maintenance (Tojkander et al., 2011). This is also in agreement with the recent demonstration that Tpm2.1- and Tpm1.6-containing actin filaments can be segregated within the same stress fibre in human foreskin fibroblasts and are engaged by myosin II motors to perform different functions (Sao et al., 2019).

The alignment of myosin IIa and IIb heads with Tpm3.1/3.2 has allowed us to construct a proposed organisation of a mini-contractile unit within stress fibres (Fig. 3I). The conservation of dimensions between the STED and electron microscopy in two cell types suggests that this may be a conserved structural unit.

U2OS cells (human, female, derived from an osteosarcoma; ATCC) were cultured in DMEM (Sigma) supplemented with 10% fetal bovine serum (FBS; Gibco) at 37°C under 5% CO₂. Cells were confirmed to be mycoplasma free with tests performed using the PCR mycoplasma test kit (AppliChem GmbH) as per the manufacturer's instructions. mEmerald–myosin IIa (Burnette et al., 2014) and mEmerald–myosin IIb constructs were Addgene plasmid #54190 and #54192, respectively (deposited by Michael Davidson); mRuby2-N-Tpm3.1 was synthesised by GeneArt (Thermo Fisher Scientific). Plasmids were and transfected into cells using Lipofectamine 3000 (Thermo Fisher Scientific) according to manufacturer's instructions and cells were used for experiments 48 h post transfection.

Animal experiments were performed in accordance with UNSW Australia Animal Care and Ethics Committee approval and Australian National Health and Medical Research Council (NHMRC) guidelines. The mouse line B6-Tpm3^{tm6(APEX2)Hrd} which expresses Tpm3.1 C-terminally tagged with APEX2 from the endogenous locus was generated by Cyagen using CRISPR. The strategy of targeting this exon via CRISPR was similar to that previously published (Masedunskas et al., 2018), with the APEX2 sequence (Ariotti et al., 2017) instead of the mNeon Green sequence inserted in frame with exon 9d (Fig. 3A). C57Bl/6 mouse zygote pronuclei were injected with Cas9 mRNA, CRISPR guide RNA and circular plasmid donor. Surviving zygotes were transferred to pseudo-pregnant female recipients and the resulting pups screened for homologous integration of the donor template by PCR using genomic DNA. The animals were genotyped by PCR using forward primer 5′-TCTTTGTCCTTGTCTGGAGTGGTC-3′ and reverse primer 5′-CAGCACTCACAGTTGGGTAAGAC-3′ resulting in a 478 bp band for the knock-in allele. The knock-in was confirmed via sequencing. Primary MEFs were isolated from day 13.5 B6-Tpm3^{tm6(APEX2)Hrd} mouse embryos and cultured as previously described (Schevzov et al., 2005).

Cells grown on #1.5 glass coverslips were fixed for 30 min in warmed 1% paraformaldehyde in DMEM with 20 mM HEPES. Coverslips were subsequently washed with PBS and permeabilised with ice-cold methanol for 5 min. Methanol was carefully diluted out with PBS so as to prevent drying out or damaging samples, and coverslips were blocked in 5% BSA in PBS for 1 h. Primary antibodies were diluted in blocking buffer and incubated on coverslips for 1 h. Coverslips were washed three times with PBS and incubated with secondary antibodies in blocking buffer for another hour. Coverslips were washed with PBS three times and mounted with Mowiol (Sigma) mounting medium. The primary antibodies used were as follows: STAR635p-conjugated mouse anti-Tpm3.1,3.2 (CG3, 1:50, Schevzov et al., 2011), mouse anti-Tpm1.6,1.7,2.1 (Tm311, 1:200, Sigma-Aldrich; cat. no. #T2780), mouse anti-α-actinin (BM-75.2, 1:500, Sigma- Aldrich cat. no. A5044); sheep anti-Tpm3.1,3.2 (γ/9d, 1:100, Percival et al., 2004), rabbit anti-Tpm4.2 [δ/9d (2009), 1:50, Schevzov et al., 2011], mouse anti-myosin IIa (1:250, Abcam cat. no. ab55456), rabbit anti-myosin IIa (1:250, Abcam cat. no. ab24762), mouse anti-myosin IIb (1:200, Abcam cat. no. ab684), and rabbit anti-myosin IIb (1:250, Covance cat. no. PRB-445P) antibodies. The secondary antibodies used were as follows: STAR635p-conjugated goat anti-mouse-IgG (1:500, Abberior cat. no. 2-0002-007-5), STAR580-conjugated goat anti-mouse-IgG (1:500, Abberior cat. no. 2-0002-005-1); STAR635p-conjugated goat anti-rabbit-IgG (1:500, Abberior cat. no. 2-0012-007-2); STAR580-conjugated goat anti-rabbit-IgG (1:500, Abberior cat. no. 2-0012-005-8) and STAR635p- conjugated donkey anti-sheep-IgG (1:250) antibodies.

STED images were acquired on a Leica TCS SP5 STED confocal microscope fitted with a 100×/1.4 oil STED objective and linked to a PicoQuant Microtime 200 with single-photon avalanche diode detectors and a PicoHarp 300 Time-Correlated Single Photon Counting System. 532 nm and 635 nm excitation pulsed Diode lasers were synchronised to a depletion multiphoton laser tuned to 750 nm. Data was acquired in Time Tagged Time-Resolved format using PH300 data acquisition software and images were extracted using a TTTR data analysis program, both written by and available from A.B., then deconvolved using Huygens Deconvolution software.

Airyscan confocal microscopy was performed on a Zeiss 880 equipped with a Plan-Apochromat 63×/1.4 NA Oil DIC M27 objective and an Airyscan detector. Samples were illuminated using 488 nm, 561 nm or 633 nm lasers, and all data was collected and processed using the Zen software (Zeiss).

Tpm3.1-APEX2 +/− and +/+ primary MEFs were seeded onto MatTek 35 mm no. 1.5 gridded coverslips and grown to 70–80% confluency. All sample processing was performed according to manufacturer's instructions in a PELCO BioWave microwave fitted with a PELCO ColdSpot Pro System for temperature control and using a PELCO EM Pro Vacuum Chamber. Cells were fixed in 2.5% glutaraldehyde (Electron Microscopy Sciences, Hatfield, PA) in 0.1 M sodium cacodylate buffer (pH 7.4; Sigma-Aldrich). After fixation, cells were washed in 0.1 M sodium cacodylate buffer. To prepare the 3′3-diaminobenzidine tetrahydrochloride (DAB; Sigma-Aldrich) mixture, DAB was dissolved in H₂O and vortexed for 3 min. The solution was diluted to a final concentration of 1 mg DAB/ml and 0.1 M sodium cacodylate, vortexed and syringe-filtrated using a 0.2 µm filter (Millipore). Cells were washed with DAB at room temperature for 2 min followed by a 30 min DAB wash supplemented with 5.88 mM H₂O₂ (Sigma-Aldrich) to generate the reaction product. After DAB treatment, cells were washed in 0.1 M sodium cacodylate buffer and post-fixed with 1% osmium tetroxide (OsO₄) in 0.1 M sodium cacodylate buffer to create an electron dense product surrounding Tpm3.1. Cells were washed in 0.1 M sodium cacodylate buffer and in H₂O. Serial dehydration was performed with 30%, 50%, 70%, 90% and 2×100% ethanol in double distilled H₂O. Serial infiltration was performed with 33%, 66% and 2×100% LX112 resin (Ladd, Williston, VT) in ethanol. Cells were flat-embedded using a resin bullet and resin was polymerised at 60°C for 16–24 h. After 30 min cooling at room temperature, polymerised cell-containing bullets were snapped off and the glass coverslip was cooled in liquid nitrogen and gently removed. Cells were visualised using a dissecting microscope (Leica EZ4, Leica Microsystems) and the resin was trimmed around them. Ultrathin 70 nm horizontal sections were cut using a 45° diamond knife (Diatome) on an ultramicrotome (Leica EM UC6, Leica Microsystems) and gently placed on formvar-coated mesh grids (ProSciTech).

Grids were imaged on a Tecnai G2 20 TEM transmission electron microscope at 200 kV at 7000× magnification. Cells were imaged with a 2k×2k BM Eagle Camera at 7000 times magnification with a binning of 1. Tomograms were acquired as described previously (Ariotti et al., 2015). Briefly, thick (200 nm) sections were cut on a Leica UC6 ultramicrotome and were coated with 10 nm fiducial gold markers. Dual axis tilt series were collected on a 200 kV Talos Arctica (Thermo Fisher) operated at room temperature. Unbinned 4k×4k images were acquired in two directions with 2° increments encompassing −60° to +60° on a Falcon 3 camera (Thermo Fisher) operated in linear mode under the control of Tomography software (Thermo Fisher Scientific). Tilt series were reconstructed using weighted back-projection in IMOD (Ariotti et al., 2015). Segmentation of tropomyosin-decorated actin filaments was performed using the Isosurface Render program in IMOD.

Images were analysed in ImageJ. Line-scan analysis was performed by drawing region of interest lines across stress fibres or cortical actin bundles in the case of the Tpm3.1–APEX2 TEM data. Manders’ co-occurrence analysis was performed using the JACoP plugin in ImageJ.

We would like to thank Tom Hall (IMB, University of Queensland) for advice on Apex sequences, the UNSW Biomedical Imaging Facility staff for technical assistance, and Pekka Lappalainen (University of Helsinki) for helpful discussions. The authors acknowledge the Victor Chang Innovation Centre, funded by the NSW Government, and the use of facilities at the Electron Microscope Unit and Biomedical Imaging Facility within the Mark Wainwright Analytical Centre at UNSW Sydney, funded in part by the NSW Government.