Dual Color Photoactivation Localization Microscopy of Cardiomyopathy-associated Desmin Mutants*

Background: Heterozygous DES mutations affect filament formation leading to skeletal and cardiomyopathies. Results: Our results reveal different extent of filament formation defects by various desmin mutants under heterozygous conditions. Conclusion: Analysis of interaction and co-localization of mutant and wild-type desmin proves the co-existence of heterogeneous filaments in living cells. Significance: These results might be of relevance for the understanding of filament formation defects. Mutations in the DES gene coding for the intermediate filament protein desmin may cause skeletal and cardiac myopathies, which are frequently characterized by cytoplasmic aggregates of desmin and associated proteins at the cellular level. By atomic force microscopy, we demonstrated filament formation defects of desmin mutants, associated with arrhythmogenic right ventricular cardiomyopathy. To understand the pathogenesis of this disease, it is essential to analyze desmin filament structures under conditions in which both healthy and mutant desmin are expressed at equimolar levels mimicking an in vivo situation. Here, we applied dual color photoactivation localization microscopy using photoactivatable fluorescent proteins genetically fused to desmin and characterized the heterozygous status in living cells lacking endogenous desmin. In addition, we applied fluorescence resonance energy transfer to unravel short distance structural patterns of desmin mutants in filaments. For the first time, we present consistent high resolution data on the structural effects of five heterozygous desmin mutations on filament formation in vitro and in living cells. Our results may contribute to the molecular understanding of the pathological filament formation defects of heterozygous DES mutations in cardiomyopathies.

Mutations in the DES gene coding for the intermediate filament protein desmin may cause skeletal and cardiac myopathies, which are frequently characterized by cytoplasmic aggregates of desmin and associated proteins at the cellular level. By atomic force microscopy, we demonstrated filament formation defects of desmin mutants, associated with arrhythmogenic right ventricular cardiomyopathy. To understand the pathogenesis of this disease, it is essential to analyze desmin filament structures under conditions in which both healthy and mutant desmin are expressed at equimolar levels mimicking an in vivo situation. Here, we applied dual color photoactivation localization microscopy using photoactivatable fluorescent proteins genetically fused to desmin and characterized the heterozygous status in living cells lacking endogenous desmin. In addition, we applied fluorescence resonance energy transfer to unravel short distance structural patterns of desmin mutants in filaments. For the first time, we present consistent high resolution data on the structural effects of five heterozygous desmin mutations on filament formation in vitro and in living cells. Our results may contribute to the molecular understanding of the pathological filament formation defects of heterozygous DES mutations in cardiomyopathies.
Desmin is a 53-kDa highly conserved muscle-specific intermediate filament (IF) 4 protein. Missense mutations and small deletions in the desmin (DES) gene have been linked to skeletal and cardiac myopathies (1)(2)(3)(4). In many cases, these mutations result in a loss of the filamentous network caused by abnormal aggregation of desmin and desmin-associated proteins (5). For clinical diagnosis of desmin-related myopathy or cardiomyopathy, histological examination of aggregate formation is often used. In patients suffering from arrhythmogenic right ventricular cardiomyopathy (ARVC), several groups recently identified different mutations in the DES gene (6 -9). ARVC is a heritable cardiomyopathy, which is pathologically characterized by predominant dilatation of the right ventricle and arrhythmias often leading to heart failure or even sudden cardiac death. However, little is known about the molecular and cellular pathomechanisms of ARVC-associated desmin mutations. In healthy cardiomyocytes, desmin forms a three-dimensional scaffold around myofibrillar z-bands and connects desmosomes as well as costamers to the contractile apparatus, thus providing cellular stability (2,10,11). Like other IF proteins, desmin consists of a central ␣-helical rod domain of 310 amino acids flanked by nonhelical head and tail domains. Previously, it was proposed that the rod domain is divided into four fragments (1A, 1B, 2A, and 2B) by three nonhelical linker regions (12). However, recent crystallographic studies of rod domainlocalized fragments of the homologous IF protein vimentin demonstrated that coil 1 and coil 2 are not separated by nonhelical linker regions (13,14). A generally accepted model for the in vitro assembly process was developed by numerous investigations (see Ref. 15 for a review). Presumably, desmin assembly can be divided into three major steps. In the first step, unitlength filaments are built by lateral association of eight tetramers, each consisting of two coiled-coil dimers arranged in an anti-parallel alignment. In the second step, short filaments are formed by longitudinal annealing of unit-length filaments. In the third assembly step, these filaments are extended into wide filaments with a typical diameter of ϳ10 -15 nm (12).
In most studies, desmin mutants were investigated in homozygously transfected cells by immunohistochemistry (4,8,16,17). However, in the clinical setting, DES mutations are usually heterozygous and lead to autosomal dominant inheritance. Previous in vitro investigations are of limited value, because wild-type and mutant desmin molecules could not be differentiated. Therefore, it is currently unknown whether mutant desmin molecules disrupt filament formation of wildtype desmin or whether both forms are incorporated into the same filament when co-expressed. Furthermore, previous studies on desmin filament formation were compromised by the limited resolution of conventional far field microscopy.
In this study, we have focused on the heterozygous character of ARVC-associated DES mutations. First, we compared filament and aggregate formation of five ARVC-associated desmin mutations in homozygously transfected SW-13, H9c2, HL-1, and C2C12 cells and in cardiomyocytes derived from human induced pluripotent stem cells (hiPS-CMs). We further assessed filament formation of different mutants at the molecular level by atomic force microscopy (AFM) to understand which step of filament formation is impaired by the mutants. Furthermore, we investigated whether co-expression of mutant and wild-type desmin disturbs filament assembly using superresolution dual color photoactivation localization microscopy (PALM) and FRET in combination with confocal microscopy.

EXPERIMENTAL PROCEDURES
Cell Culture, Transfection, and Fluorescence Microscopy-SW-13, H9c2, and C2C12 cells (LGC Standards, Middlesex, UK) were grown in culture dish assemblies on glass cover slides in DMEM (Invitrogen) supplemented with 10% FBS and penicillin/ streptomycin (Invitrogen). HL-1 cardiomyocytes were kindly provided by W. C. Claycomb (18) and cultured in Claycomb medium (Sigma-Aldrich) supplemented with 10% FBS, penicillin/streptomycin, 2 mM L-glutamine, and 0.1 mM norepinephrine. Cardiac differentiation of hiPS cell line derived from foreskin fibroblasts, clone 1 (19), was carried out on the murine visceral endoderm-like cell line END2 in knock-out DMEM containing 1 mM L-glutamine, 1% nonessential amino acids, 0.1 mM ␤-mercaptoethanol, and penicillin/streptomycin as described earlier (20,21). The co-culture was left undisturbed at 37°C for 4 days. The first medium change was performed on day 5 and later on days 9, 12, and 15 of differentiation. Spontaneously beating areas from hiPS-CM cultures were microdissected on day 25 of differentiation and dissociated into single cells by trypsinization. Single hiPS-CM were plated on fibronectin-coated (2.5 g/ml) dishes in Iscove's modified Dulbecco's medium supplemented with 10% FBS containing 1 mM L-glutamine, 1% nonessential amino acids, 0.1 mM ␤-mercaptoethanol, and penicillin/streptomycin. The cells were transfected using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol using 800 ng of plasmid DNA for a single transfection and 400 ng of each plasmid for co-transfections. For details on the plasmids, see supplemental Table SI. The medium was replaced 8 h after transfection. 24 h later, the cells were fixed in 4% paraformaldehyde, permeabilized with 0.1% Triton X-100, and blocked with 5% FBS and then stained overnight at 4°C with primary antibodies against sarcomeric ␣-actinin (Sigma-Aldrich) and later with Alexa Fluor 555 conjugated anti-mouse IgG1 secondary antibody (Molecular Probes). The nuclei were stained with Hoechst 33342. The samples were embedded in ProLong Gold antifade reagent (Invitrogen). Fluorescence images were recorded with an Eclipse TE2000-U microscope (Nikon, Tokyo, Japan) equipped with a digital sight DS-2MV CCD camera (Nikon); YFP, Cy3, and DAPI filter sets (AHF, Tübingen, Germany); and an oil immersion objective (Plan Apochromat 60ϫ/1.40 oil; Nikon).
Immunoblotting-24 h after transfection, the cells were harvested using 200 l of ProFound lysis buffer (Thermo Scientific, Waltham, MA) containing 1 mM proteinase inhibitor mixture (Sigma-Aldrich). After homogenization and centrifugation, the concentrations of the solubilized proteins were measured by the BCA method. Equal amounts of total lysates were separated by SDS-PAGE and electrotransferred (Trans-blot SD; Bio-Rad) to nitrocellulose membranes (Bio-Rad). After blocking for 1 h in 5% fat free milk powder in TBS, the membranes were incubated overnight at 4°C with antibodies against desmin (200 ng/ml; R & D Systems, Minneapolis, MN) and GAPDH (20 ng/ml; Abcam, Cambridge, UK). After rigorous washing, the membranes were incubated with HRP-conjugated secondary antibodies against goat (100 ng/ml; R & D Systems) or mouse IgG (400 ng/ml; BD Pharmingen, San Diego, CA) for 1 h at room temperature followed by washing the membranes with TBST. Chemiglow (Cell Biosciences, Santa Clara, CA) was used as a substrate for HRP. Chemiluminescence signals were recorded with a CCD camera system (Cell Biosciences) and analyzed with the Alpha View FluorChem FC2 software (Cell Biosciences).
Immunostaining-SW-13 cells were transfected with desmin expression constructs without any fluorescent tag. 24 h after transfection, the cells were washed three times with PBS, fixed in 4% paraformaldehyde, permeabilized with 0.1% Triton X-100 (20 min at room temperature), blocked with 5% FBS, and then stained overnight at 4°C with primary antibodies against desmin (R & D Systems) and later with Cy3-conjugated secondary antibody against goat IgG (Jackson ImmunoResearch, Suffolk, UK).
Desmin Expression, Purification, and Folding-Bacterial cells (Escherichia coli BL21Star DE3), transformed with the corresponding plasmids, were grown in LB medium at 37°C to an A 600 nm of 0.5-0.8 in a shaking incubator. Protein expression was induced with isopropyl ␤-D-thiogalactopyranoside (1 mM). After 4 h, the bacteria were harvested by centrifugation, and inclusion bodies were prepared as previously described (8). Inclusion bodies containing recombinant desmin molecules were finally dissolved in buffer composed of 8 M urea, 10 mM Tris-HCl, 100 mM NaH 2 PO 4 , pH 8.0. After centrifugation (45 min at 4°C at 45,000 g), the supernatants were supplied to a HiTrap DEAE-Sepharose Fast Flow column (GE Healthcare) Filament Formation Defects of Heterozygous Desmin Mutants using a FPLC system (GE Healthcare). Desmin molecules were eluted by a linear salt gradient (0 -0.35 M NaCl), and purification of recombinant desmin was analyzed by SDS-PAGE and Coomassie Blue R-250 staining. Fractions containing desmin were pooled and purified with a Ni 2ϩ -NTA-Sepharose column (Qiagen). The column was washed with buffer (8 M urea, 20 mM Tris-HCl, 10 mM imidazole, pH 8.0) until the absorbance at 280 nm decreased to a constant value below 0.01. Afterward, recombinant desmin molecules were eluted with imidazolecontaining buffer (8 M urea, 20 mM Tris-HCl, 300 mM imidazole, pH 6.9). Fractions containing more than 95% recombinant desmin were pooled and stored at Ϫ80°C. Purified recombinant desmin was dialyzed against buffer (20 mM Tris-HCl, 1 mM DTT, 1 mM EDTA, 0.1 mM EGTA, and 10 mM NH 4 Cl, pH 8.4) as previously described (8). Protein concentrations were determined by the BCA assay according to the manufacturer's instructions (Sigma-Aldrich) using BSA as a standard.
Atomic Force Microscopy-Filament assembly of recombinant desmin was initiated immediately after dialysis by adding an equal volume of assembly buffer (100 mM NaCl, 45 mM Tris-HCl, pH 7.0) and subsequent heating to 37°C for 1 h (22). The assembled desmin filaments (12.5 g/ml) were applied to freshly cleaved mica substrates (Plano, Wetzlar, Germany), rinsed with deionized water to remove unbound protein, and dried under a gentle flow of nitrogen. Topographic AFM imaging was done in ambient conditions with a Mutimode AFM and Nanoscope IIIa controller (Bruker, Santa Barbara, CA) in tapping mode using Tap300AI-G (Budget Sensors, Sofia, Bulgaria) or PPP-NCH (Nanosensors, Neuchatel, Switzerland) probes.
Photoactivation Localization Microscopy-After cell transfection, PALM images were acquired on a modified inverted microscope (Axiovert 200; Zeiss, Jena, Germany). Three diodepumped solid state lasers supplied the laser lines 561 nm (GCL-150-561; CrystaLaser, Reno, NV), 473 nm (LSR473-200-T00; Laserlight, Berlin, Germany), and 405 nm (CLASII 405-50; Blue Sky Research, Milpitas, CA) for excitation and photoactivation of the fluorophores. All of the laser sources were combined via dichroic mirrors (AHF) and coupled into a single mode fiber (OZ Optics, Ottawa, Canada). During the experiment, laser intensities were controlled via an acousto-optic tunable filter (AOTFnC-400.650, A-A; Opto-Electronic, Orsay, France). Before and after PALM imaging, we recorded a bright field image of each cell to monitor cell viability. PALM images of single transfected cells expressing mEosFPthermo fusion proteins were acquired using an electron multiplying charge coupled device camera (Ixon DV887ECS-BV; Andor, Belfast, UK). A total of 10,000 -20,000 frames using a camera exposure time of 50 ms were recorded for each high resolution image. Wide field excitation of the red form of mEosFPthermo was performed with 561-nm laser light at 0.3-0.5 kW/cm 2 ; green-tored conversion was induced by 405-nm laser light at 0 -0.02 kW/cm 2 . Fluorescence light was filtered by a 610/75-nm bandpass (Chroma HQ 610/75; AHF) after passing the excitation dichroic mirror (z 405/473/561; AHF). PALM images of cotransfected cells expressing mEosFPthermo as well as mIris-GFP1 fusion proteins were recorded using a different electron multiplying charge coupled device camera (Ixon EMϩ DU-860; Andor, Belfast, UK), which allowed faster imaging with an exposure time of 3 ms/frame, each of which consisted of 128 ϫ 128 pixels. For each PALM image, 5,000 -20,000 frames were recorded depending on the protein expression of the individual cells. Wide field excitation of the red form of mEosFPthermo was performed with 561-nm laser light at 2.5-4.5 kW/cm 2 ; green-to-red conversion was induced by 405-nm laser light at 0 -0.2 kW/cm 2 . For excitation and off switching of mIrisGFP1, 473-nm light at 2.8 -3.6 kW/cm 2 was used, whereas on switching was induced by 405-nm light at 0 -0.2 kW/cm 2 . Fluorescence light was separated by the excitation dichroic mirror and filtered by a 610/75-nm bandpass for detection of red fluorescence and by a 535/70-nm bandpass (Chroma HQ 535/70; AHF) for detection of green fluorescence. All of the PALM images shown were analyzed by fitting a two-dimensional Gaussian distribution to the single molecule signals detected using a homemade algorithm written in Matlab (Matlab R2010b; The MathWorks, Natick, MA) (23,24). According to the strategy of Shroff et al. (25), data acquisition was stopped in the red and started in the green channel when no more molecules were detectable in the red channel under photoconverting 405-nm irradiation. In this way, we ensured that all of the molecules had been photoconverted and subsequently bleached. In controls we confirmed that all mEosFPthermo-molecules ("green") were indeed converted to their red-emitting forms using singly transfected cells under identical experimental conditions as compared with the dual color experiments. After photo-conversion with 405-nm light, no mEosFPthermo molecules were detectable in the green channel (supplemental Fig.  S2, A and B). For control purposes, we also checked whether mIrisGFP1 was not converted to a red emitting form. We imaged cells transfected with mIrisGFP1 with the same illumination protocol as for the dual color experiments. No mIris-GFP1 molecules were detectable in the red channel under these conditions (supplemental Fig. S2, C and D).
FRET Imaging-FRET measurements were performed on a confocal laser scanning microscope LSM710 (Zeiss) equipped with an oil immersion objective (Plan-Apochromat 63ϫ/1.40 oil; Zeiss). In each FRET experiment, three different images were recorded. First, the donor fluorophore enhanced cyan fluorescent protein was excited at 458 nm, and the fluorescence emission was detected in the spectral range of 462-515 nm. Second, the fluorescence signal of the acceptor fluorophore eYFP was detected in the spectral range of 554 -651 nm (458 nm, excitation; apparent FRET signal). Third, the acceptor fluorophore eYFP was excited at 514 nm, and the fluorescence emission was detected in the spectral range of 554 -651 nm. A corrected FRET signal was obtained by correcting for leakage of the donor fluorescence emission into the acceptor detection range and for direct excitation of acceptor fluorophore (26). This correction requires two correction factors (␣ ϭ 0.599 Ϯ 0.147; ␤ ϭ 0.488 Ϯ 0.051) for donor leakage and direct excitation of the acceptor (26), which were determined from 10 individual transfected cells expressing only the donor or the acceptor fusion protein.

Cytoplasmic Aggregate Formation by Desmin Mutants in
Transfected Cells-To compare filament or aggregate formation of different ARVC-associated desmin mutants, we generated expression constructs of desmin wild type, p.S13F, p.E114del, p.N116S, p.N342D, and p.R454W fused at the C terminus to eYFP and analyzed filament and aggregate formation in homozygously transiently transfected SW-13, H9c2, HL-1, C2C12 cells, and hiPS-CM. Because we used tagged desmin constructs carrying fluorescent proteins at the C terminus, we and others (28) excluded an influence of the tag on filament or aggregate formation by immunolabeling (see Fig. 2C).
In agreement with other studies (8,9,29), we detected cytoplasmic aggregates with p.E114del, p.N116S, and p.N342D independent of the transfected cell type (Fig. 1, rows 3-5). In contrast, the mutant p.S13F and p.R454W assembled into filamentous networks (Fig. 1, rows 2 and 6) similar to wild-type desmin (Fig. 1, row 1), and the proportion of cells with aggregates reached no significant level in comparison with wild-type transfected cells (Fig. 2A). These results were also verified by high resolution PALM on SW-13 cells (Fig. 3). Because differ-ences in protein expression levels between individual desmin mutants might contribute to their different aggregation propensities in cells, we compared the expression levels of the different desmin mutants with the wild-type desmin using immunoblotting (Fig. 2B). In all of the experiments, the expression levels of transfected desmin forms were comparable and therefore were not causative for aggregate formation (Fig. 2B and  supplemental Fig. S3). Furthermore, we analyzed in H9c2, C2C12, and HL-1 cells the ratio of the endogenously and exogenously expressed desmin by semiquantitative immunoblotting (supplemental Fig. S3). These data indicate that 2-4% of the desmin can be accounted to the endogenous desmin in all transfected cells and is therefore negligible.
Analysis of Filament Assembly of Mutant and Wild-type Desmin by AFM-To investigate filament formation of the different desmin mutants at the molecular level, we expressed mutant and wild-type desmin in E. coli and investigated the structures of the purified proteins by AFM. The purity of the recombinantly expressed molecules was assessed by SDS-PAGE and Coomassie Blue staining (supplemental Fig. S1). As expected, wild-type desmin formed filaments of variable length. In high resolution AFM images, a coiled structure with a righthanded twist was observed (Fig. 4A). In contrast, all mutants with the exception of p.R454W did not assemble into regular filaments (Fig. 4, B-E). The mutants p.S13F and p.N116S (Figs. 2C and 4B) showed filamentous aggregates of different size. In the case of p.N342D (Fig. 4E), the samples were sparsely covered with irregularly shaped aggregates and small fibrils, presumably desmin tetramers or unit-length filament precursors. Interestingly, the deletion mutant p.E114del differed remarkably from the other observed filament assembly defects. This mutant did not assemble into regular filaments at all but formed fibrils that were like beads on a string (Fig.  4D) with a length of 114 Ϯ 55 nm (n ϭ 88), which is significantly less compared with wild-type desmin filaments (500 Ϯ 166 nm, n ϭ 54; t test p Ͻ 0.001). Apparently, the observed structures represent short filament precursors. The desmin filaments formed by the p.R454W mutant did not differ significantly from wild-type desmin filaments in the in vitro assembly experiments (Fig. 4F).

Dual Color PALM Analysis of Filament Structure of ARVCassociated Desmin Mutants Co-expressed with Wild-type
Desmin-Because all ARVC-associated desmin mutations presently known are heterozygous, we analyzed filament or aggregate formation in mutant and wild-type co-transfected cells. H9c2, C2C12, and HL-1 cells and hiPS-CM express endogenous cytoplasmic IF proteins. Therefore, we focused our analysis on transfected SW-13 cells, lacking endogenous cytoplasmic IF proteins (30), to characterize filament formation in a controlled heterozygous constellation.
To this end, we investigated whether wild-type and mutant desmin molecules co-exist in the same IF and whether filament assembly was disrupted by mutant desmin. Mutant and wildtype desmin were fused either with a thermostable variant of the green to red convertible fluorescent protein mEosFP termed mEosFPthermo (31)(32)(33) or with the green photoswitchable mIrisGFP1, a variant of IrisFP (24,34,35). After co-transfection with desmin expression constructs under the same constitutive promoter, filament formation was ana-  in living cells by dual color PALM. This technique provides a roughly 10-fold improvement in resolution over conventional wide field microscopy (compare Fig. 3) (24, 25, 36).
As a result, co-expression of p.S13F (Fig. 5, D-F) or p.N116S (Fig. 5, J-L) with wild-type desmin led to filaments as well as aggregates in the same cell. In contrast, the co-expression of p.E114del in combination with wild-type desmin blocked filament formation completely because only cytoplasmic aggregates could be detected in transfected cells (Fig. 5, G-I). The co-expression of p.N342D (Fig. 5, M-O) or p.R454W (Fig. 5, P-R) with wild-type desmin led to an incorporation of mutant molecules into IF without aggregates similar to the controls (Fig. 5, A-C).
FRET Imaging of Co-expressed Mutant and Wild-type Desmin-To confirm the results of the PALM experiments with a second approach providing nearly molecular resolution on filament formation mimicking a heterozygous genotype, we  performed FRET microscopy by transfecting cells with enhanced cyan fluorescent protein-and eYFP-tagged desmins. FRET experiments provide additional information on the molecular assembly of desmin filaments and aggregates, because a strong FRET efficiency signal can only be detected if two proteins are in close proximity. We first performed a control experiment in filamentous structures with wild-type desmin fused to both fluorescence tags. We detected a FRET signal in all filaments (Fig. 6A), which indicates a suitable interprotein distance for this approach. We found that mutant desmin molecules p.S13F (Fig. 6B), p.N342D (Fig. 6E), and p.R454W (Fig. 6F) were incorporated together with wild-type desmin into the same filaments. No aggregates were detected in these co-transfected cells. In cells co-expressing p.N116S mutant and wild-type desmin, we observed a FRET signal in both small filaments as well as aggregates (Fig. 6C). In cells co-expressing the desmin mutant p.E114del in combination with the wild-type desmin FRET signal could be detected only in cytoplasmic aggregates (Fig. 6D), further revealing severe impairment of desmin fiber formation in this constellation. We conclude that mutant and wild-type desmin were co-assembled in both structures and that mutant desmin partially (p.N116S)  MAY

DISCUSSION
Up to now, more than 50 disease-causing DES mutations distributed over the entire amino acid sequence were identified (4). The disease phenotypes of these mutations are clinically heterogeneous, ranging from isolated myopathy, dilated cardiomyopathy with or without skeletal muscle disorder, and hypertrophic, restrictive, or arrhythmogenic right ventricular cardiomyopathy (37).
In this study, we have systematically compared the effects of five recently identified ARVC-associated DES mutations on filament formation to obtain further molecular insights into the assembly defects of this subgroup of mutants. We analyzed desmin filament formation in five different cell types ranging from human adenocarcinoma cells (SW-13) to human cardiomyocytes derived from hiPS to exclude putative influences of the cell environment on filament assembly defects. As a result, we did not find differences between the cell types used for the homozygous transfection. However, in these cell lines it was only possible to differentiate between desmin aggregates or filament formation, respectively (Fig. 7). Moreover, the DES mutations analyzed in our study are found in patients as a heterozygous genotype. Thus, this cell system is not directly comparable with the in vivo situation.
To analyze whether mutant desmin molecules are incorporated into common filaments of the wild-type form, we cotransfected SW-13 cells with mutant and wild-type desmin fused to different photoconvertible fluorescent proteins to be able to perform PALM imaging (38). As an additional approach, we analyzed filament formation in the same cell system by application of FRET.
We hypothesized that the heterozygous desmin genotype will result in remarkably different forms of desmin filament and aggregate formation compared with the homozygous cell culture model, because filament formation might be influenced by co-expression of mutant desmin according to the following types (compare Fig. 7): type A, only filaments are formed: mutant and wild-type molecules are incorporated within the same filament; type B, only aggregates are formed composed of a mixture of wild-type and mutant molecules; Type C, aggregates and filaments are found within the same cell: the aggregates and filaments are formed by mutant and wild-type desmin; and type D, aggregates and filaments are found within the same cell: wild-type desmin is found in filaments, whereas the mutant forms aggregates. To identify which of these types might be found in desminopathies, we analyzed cardiomyopathy-associated mutations in cell culture.
Sharma et al. (17) and Pica et al. (41) found equivocal results on the desmin mutant p.S13F, because it formed small fila-

Filament Formation Defects of Heterozygous Desmin Mutants
ments and aggregates after homozygous transfection in different cell lines. In our experiments, the homozygous transfection of p.S13F reveals exclusively filaments in the majority of cells, whereas in the heterozygously transfected cells, we found mixed filaments of mutant and wild type, as well as cells with mixed filaments and aggregates. Thus, p.S13F forms filaments according to types A and C, respectively (Fig. 7). Serine 13 is localized within the conserved nonapeptide in the head domain, which is essential for filament assembly (39). The heterogeneous in vitro effects found with the mutant p.S13F might therefore be due to the degree of phosphorylation at serine 13 (40). It is remarkable that the ambiguous classification of the in vitro findings are paralleled by the heterogeneous clinical phenotype ranging from isolated myopathy to different forms of cardiomyopathy even within the same family (6,7,17,41).
The mutant p.E114del revealed a severe filament formation defect in vitro, because we found exclusively cells with aggregates, which is consistent with previous studies on homozygously transfected SW-13, 3T3, and C2C12 cells (9). In the heterozygously transfected cells, mutant p.E114del showed a dominant effect even on wild-type desmin and led to aggregates, which also incorporated wild-type desmin according to type B (Fig. 7). This deletion mutant also impairs protein interactions with the IF protein synemin, resulting in a complete co-aggregation of both proteins (42). p.E114del presumably induces a twist of the hydrophobic seam, which is essential for formation of the coiled-coil dimer (9). We therefore conclude from our data that this mutant provides a dominant loss of the ability of wild-type desmin molecules to form filaments. Furthermore, the AFM experiments revealed that the longitudinal extension step is presumably impaired similar to the previously characterized mutations p.L385P and p.R406W (43), because this mutant forms significantly shorter filaments compared with the wild type.
Previously, we showed that the mutant p.N116S leads to aggregate formation when homozygously expressed in SW-13 cells. However, in the cardiac tissue of the patient, both filaments and aggregates of desmin were detectable (8). In the heterozygous cell culture experiments, mutant p.N116S forms also filaments and aggregates within the same cell composed of both desmin forms according to type C (Fig. 7).
In our study, homozygous expression of p.N342D desmin led to a phenotype comparable with the one previously described (29). However, when extending these investigations to heterozygous co-expression, we exclude a dominant inhibiting effect on filament assembly, because mutant and wild-type desmin molecules were incorporated into the same filamentous structures without aggregate formation according to type A (Fig. 7).
Different groups investigated the homozygous expression of p.R454W desmin in different cell lines, leading to conflicting results concerning the degree of aggregate formation (28,44). In our hands, homozygously transfected p.R454W desmin formed filament networks independently of the cell type investigated. These results were confirmed by high resolution fluorescence microscopy and AFM experiments with purified desmin. Interestingly, our AFM data revealed right-handed helical filaments of wild-type and p.R454W desmin (Fig. 4), which are also known from the IF protein vimentin (45). Furthermore, we found that the heterozygously expressed p.R454W desmin forms mixed filaments in transfected cells according to type A (Fig. 7). Co-expression experiments were also reported by Bär et al. (46) leading to similar results. However, in that study, co-localization of wild-type and mutant desmin could not be analyzed for technical reasons. The amount of dimers, tetramers, and oligomers of p.R454W in transfected cells was previously determined by single particle fluorescence spectroscopy, demonstrating that p.R454W differs only marginally from the wild type (28). Consequently, the pathogenic role of this variant is not clear. Further studies might clarify the clinical impact of this sequence variant.
Our data clearly reveal that all of the mutants form mixed filaments or aggregates (Fig. 7, types A-C). Interestingly, in any of the mutations analyzed in this study, we did not observe a separation of mutant and wild-type desmin into filaments or aggregates within the same cell, respectively (Fig. 7, type D). Despite this systematic absence of segregation, it has to be noted that slight variations in the local concentration of mutant and wild-type desmin can be observed in the PALM images, as well as in the FRET images. Also, because of the sequential nature of the PALM image acquisition procedure, both color channels are affected by cell movement, resulting in a slight shift of the structures in the red and green channels, respectively. This is in agreement with previous reports describing IFs as dynamic structures that move in living cells (for a review see Ref. 47).
The detection of filaments composed of wild-type and mutant desmin raises the question of whether the nanomechanical properties of mixed filaments are impaired by the incorporation of the mutant forms. However, currently we cannot exclude that some other desmin mutants might lead to a type D filament formation defect.
The methodological approach of our study includes two complementary techniques for the analysis of desmin filament formation in living cells. Whereas the dual color PALM experiments provide insight into the structural composition of mutant and wild-type desmin, the FRET data reflect molecular interactions of these IF forms within the range below 10 nm. Because of the dual color PALM experiments, we could exclude for the investigated mutants the hypothetical type D filament formation defect, which is not detectable by FRET analysis.
In summary, our methodological approach using high resolution dual color PALM and FRET measurements substantially improves the understanding of the structures and the molecular assembly of heterozygously expressed mutants in living cells. In addition, the experiments led to a more sophisticated scheme of filament formation defects caused by DES mutations.