Intracellular dynamin elastin-like polypeptides assemble into rod-like, spherical, and reticular dynasomes

Abstract

Dynamin (DNM) is a family of large GTPases possessing a unique mechanical ability to ‘pinch’ off vesicles entering cells. DNM2 is the most ubiquitously-expressed of the DNM family. We developed a novel tool based on elastin-like polypeptide (ELP) technology to quickly, precisely, and reversibly modulate DNM2 structure. ELPs are temperature-sensitive biopolymers that self-assemble into microdomains above sharp transition temperatures (T_t). When linked together, DNM2 and a temperature-sensitive ELP fusion organize into a range of distinct temperature-dependent structures above a T_t, which were not observed with wild-type DNM2 or a temperature-insensitive ELP fusion control. The structures comprised three different morphologies, which were prevalent at different temperature ranges. The size of these structures was influenced by an inhibitor of DNM2 GTPase activity, dynasore; furthermore, they appear to entrap co-expressed cytosolic ELPs. Having demonstrated an unexpected diversity of morphologically distinct structures, DNM2-ELP fusions may have applications in exploration of dynamin-dependent biology.

Graphical Abstract

INTRODUCTION

Elastin-like polypeptides (ELPs) are a class of temperature-sensitive proteins derived from tropoelastin. ELPs are biocompatible and can quickly and reversibly phase separate above a tunable transition temperature (T_t) inside cells. They are composed of an amino acid repeat sequence (VPGXG)_n whereby the identity of X and n strongly determine T_t. When fused to effector proteins, their temperature-sensitive properties are conferred to the chimeric fusion. Previously, this approach has been used to generate molecular switches for clathrin light chain (CLC)¹, caveolin1 (CAV1)², and epidermal growth factor (EGFR)³. Genetically encoded ELP fusions were observed to reversibly form microdomains above their respective T_t. The appearance of condensed spherical morphologies was also shown modulate the biological function of their respective effector protein^1–3, which serves as a proof-of-concept platform technology for studying other effector proteins, like Dynamin (DNM).

DNM proteins are a large 100 kDa GTPase canonically known to regulate vesicular scission of endocytic vesicles from the inner cell-membrane⁴, and implicated in neurodegenerative disease Charcot Marie Tooth syndrome^{5, 6}. The DNM family of proteins are comprised of DNM1, DNM2, and DNM3. While DNM1 and DNM3 are highly expressed in neuronal and testes cells, DNM2 is ubiquitously expressed across all cell types⁷. Its role is well-established within clathrin-mediated endocytosis; however, much remains unknown about its role in other pathways, such as caveolin- and flotillin-mediated endocytosis^8–10. A DNM2-ELP fusion could modulate spatial organization of DNM2 by exploiting the ELP phase-transition. Understanding the characteristics of its temperature-mediated structural reorganization would serve as a useful tool to study the intricacies of DNM2’s heterogeneous constituents implicated in its biology, as well as their implication in diseases.

The current set of molecular tools available to study DNM2 biology are potentially limited by their mode of application. For example, overexpression of the dominant-negative mutant (DNM2-K44A) and or use of siRNA knockdown to DNM2 require an incubation period of 24-hrs. This time between treatment and effect may introduce difficulty in discriminating between the direct effects on internalization pathways by DNM2 versus indirect secondary effects. For instance, primary effects on clathrin-mediated endocytosis due to reduced DNM2 function can elicit compensatory upregulation of pinocytosis¹¹ or unintendedly affect cell morphology^{8, 12}. One solution to this challenge was the use of a faster-acting small molecule inhibitor, dynasore. However, this suffers from off-target binding to other GTPases and is attenuated by serum proteins and/or trace detergents commonly used in assays^{13, 14}. The Dyngo^™ series of DNM2 inhibitors are analogues of dynasore introduced as alternative solutions; however, they also share similar limitations that extend beyond inhibition of DNM¹³.

These challenges support the development of a novel tool that can modulate DNM biology. Here, we have engineered a probe to modulate DNM assembly using ELP technology. Above T_t, a temperature-sensitive fusion called DNM2-V96 can assemble a range of unique vesicular structures we term ‘dynasomes’ (Fig. 1). This manuscript characterizes the properties of these ‘dynasomes,’ which may eventually provide a tool for the entrapment of cytosolic proteins for the study of DNM biology.

Figure 1: Temperature-sensitive DNM2-V96 undergoes phase-transition above T_t.

The temperature sensitive properties of V96 drive the formation of vesicular dynasomes inside HE293T cells, ranging from elongated, large spherical, or small reticular morphologies.

MATERIALS AND METHODS

2.1. Construction of DNM2-ELPs.

The DNM2 DNA sequence was obtained by the National Center for Biotechnology Information (NCBI, Bethesda, MD) and was synthesized as a gBlock gene fragment (Table S1) by Integrated DNA Technologies (IDT, Coralville, IA). The sequence contained XbaI and NdeI restriction sites at terminal ends for digest and ligation into pET25b(+) A96 or V96 vectors previously generated by recursive directional ligation^{15, 16}. DNM2-ELP fragments were then cut with XbaI and EcoRI (#R0145S, #R3101S, New England Biolabs, Ipswich, MA) and inserted into an empty pcDNA3.1(+) mammalian vector downstream of HA and FLAG-tag sequences. DNA sequences of both DNM2-A96 and DNM2-V96 were verified by diagnostic restriction digest and Sanger sequencing (Genewiz, South Plainfield, NJ).

2.2. Cell culture and transfection

HEK 293T cells (#CRL-11268, ATCC, Manassas, VA) were maintained in Dulbecco’s Modified Eagle Medium (DMEM) (#11995065, ThermoFisher Scientific, Waltham, MA) supplemented with 10% fetal bovine serum (FBS) (#35-010-CV, Corning, Corning, NY) in a T-75 flask (#4616, Laguna Scientific, Laguna Niguel, CA) at 37°C in a 5% CO2 humidified Symphony 5.3A tissue culture incubator (#98000-366, VWR, Radnor, PA). Cells were sub-cultured to 80–90% confluence and washed with Dulbecco’s phosphate-buffered saline (dPBS) (#25-508, Genesee, San Diego, CA) before dissociating with 0.05% trypsin (#25300-120, ThermoFisher Scientific, Waltham, MA). Cells were resuspended in fresh media and subcultured in either 6- or 12-well plates for transfection. Transfection used Lipofectamine 3000 (#L3000015, ThermoFisher Scientific, Waltham, MA) following the manufacturer’s protocol with DNM2-ELP plasmid DNA, then incubated in an Isotemp tissue culture incubator (#FICO3500TABB, GS Laboratory Equipment, Asheville, NC) at 30°C for 72 hrs before assay.

2.3. Western blot analysis.

HEK 293T cells subcultured in 6-well plates (#25-105, Genesee, San Diego, CA) were grown to 70% confluency before transfection with either DNM2-V96 or DNM2-A96 plasmid DNA. After a 72 hr incubation, cells were lysed for immunoblotting following methods previously described³. Antibodies directed against ELPs (AK1)¹⁷ were diluted 1:1000 in 5% bovine serum albumin (BSA) (#A9647-50G, Sigma-Aldrich, St. Louis, MO), then tagged with 1:5000 dilution of α-mouse HRP-linked antibodies (#7076S, Cell Signaling Technologies, Danvers, MA). The PVDF membrane was visualized on an iBright FL1000 imaging system (ThermoFisher Scientific, Waltham, MA).

2.4. Fixed cell preparation.

HEK 293T cells were subcultured to 12-well plates with cover slips coated in poly-D-lysine. After transfection with DNM2-V96 or DNM2-A96, cells were fixed at different temperatures and labelled for indirect immunofluorescence². Briefly, 72 hrs post-transfection, cells were placed at 4°C for 1 hr to solubilize the ELPs. Cells were then incubated in DMEM containing 10% FBS at the indicated temperature using a heat block (#IC25, Torrey Pines Scientific, Carlsbad, CA). Temperature was monitored a thermal probe (#CP254765, Cole Parmer: Digi-Sense, Vernon Hills, IL) for 1 hr and quickly fixed with 4% paraformaldehyde (#43368-9M, Alfa Caesar, Haverhill, MA) for 15 mins. Coverslips were then incubated in 50 mM ammonium chloride for 5 mins, washed 3 times with dPBS, and permeabilized with 0.1% Triton X-100 (#9002-93-1, Sigma-Aldrich, St. Louis, MO) for 15 mins. Cells were washed another 3 times with dPBS for 3 mins, then blocked with 1% BSA for 1 hr at room temperature (RT). Cells were incubated with 1:100 AK1 anti-ELP antibody¹⁷ in 1% BSA at 4°C overnight, and washed 3 times with dPBS for 5 mins before incubation with 1:100 goat α-mouse AlexaFluor 488 (#A-11001, ThermoFisher Scientific, Waltham, MA) in 1% BSA at RT for 1-hr. The coverslips were incubated with DAPI (#D1306, ThermoFisher Scientific, Waltham, MA) with a final concentration of 5 μg/ml for 5 mins and washed with dPBS 5 times for 5 mins each. Coverslips were mounted with ProLong Gold Antifade (#P36934, ThermoFisher Scientific, Waltham, MA) mounting media on glass slides and cured overnight at RT.

2.5. Imaging and structural analysis.

HEK 293T cells expressing either DNM2-A96 or DNM2-V96 cells were fixed and labeled by indirect immunofluorescence as described above. For 2-dimensional analysis, at least three images were acquired at each temperature (4, 20, 30, 35, and 40°C) with a LSM880 confocal microscope (ZEISS Microscopy, Jena, Germany). Images were acquired in LSM Fast Mode with a 488 nm multiline argon laser and processed for Airyscan super-resolution quality. Individual cells were scored according to the appearance of one of 5 distinct morphologies at each temperature: ‘diffuse’, ‘puncta’, ‘Dynasome-Elongated (E)’, ‘Dynasome-Large (L)’, or ‘Dynasome-Small (S)’. ‘Diffuse’ morphology is defined by a lack of distinct boundaries of the expressed DNM2-ELP fluorescence with distribution along the plasma membrane (PM) or within the cytosol and can appear cloudy or amorphous. ‘Puncta’ morphology is defined as small bright points with distinct boundaries and no visible lumen. ‘Dynasome-E’ morphology is defined as thin rod-like structures with a lumen lined by DNM2-ELP. ‘Dynasome-L’ is defined by large round structures, appearing as large vacuole-like vesicles lined with DNM2-ELP. ‘Dynasome-S’ morphology is defined as round compartmental structures like ‘Dynasome-L’ that appear as smaller bundles with interconnected lumens.

The five possible states observed in either DNM2-A96 or DNM2-V96 groups were ranked from lower to higher rank according to the proportion at which they appear at each temperature. If a particular morphology appears at a higher temperature, it is ranked above one that appears more at a lower temperature. For example, the Dynasome-S morphology appears at a higher proportion at higher temperatures than the Dynasome-L morphology. As such, if a cell contains both the Dynasome-L and Dynasome-S morphologies, it would be characterized as Dynasome-S based on the hierarchy. Proportions and other statistical parameters were calculated and plotted in GraphPad Prism (La Jolla, CA, http://www.graphpad.com). Dynasome-L and Dynasome-S morphology diameters were quantified with the straight-line tool in FIJI¹⁸. Their diameters were then collected in GraphPad and plotted as a mean with standard deviation per temperature. ANOVA with Tukey multiple comparisons test was done on each temperature data set.

Fixed cells were visualized with 3-dimensional imaging. Cells were prepared as previously described, but at 4 and 37°C. Z-stacks were acquired with a LSM880 confocal microscope (ZEISS Microscopy, Jena, Germany) in LSM Fast Mode with Airyscan super-resolution using a 488 nm argon multiline laser. Cells were selected in 2-dimensions and their z-range was determined using DAPI signal. Optimal interval between slices were set to satisfy Nyquist over-sampling criteria. Orthogonal projections and 3-dimensional voxel reconstruction were also performed with Zen Black 2.3 (ZEISS Microscopy, Jena, Germany).

2.7. Dynasore inhibitor assay and analysis

Cells were transfected with DNM2-V96 as described above and then serum starved in DMEM for 24-hrs before being treated with either 0 μM or 80 μM Dynasore (#D7693-5MG, Sigma-Aldrich, St. Louis, MO) at 37°C for 30 mins. Cells were then fixed and stained as described above for imaging using Airyscan on the LSM880. The diameter of DNM2-V96 structures were measured in cells treated with 0 μM or 80 μM dynasore semi-automatically using FIJI. Images were converted to binary projections for automatic analysis of all structures via the particle analysis function. Data was collected and analyzed on Graphpad Prism version 7.0.

2.8. Live-cell imaging and analysis.

HEK 293T cells were seeded onto 35-mm glass bottomed dishes (P35G-0-10-C, MatTek Corporation, Ashland, MD) pre-coated with poly-D-lysine (P0899-10MG, Sigma, St. Louis, MO) at 10 mg/ml. Cells were transfected with either DNM2-pmCherryN1 (Addgene plasmid #27689; http://n2t.net/addgene:27689; RRID: Addgene_27689)¹⁹, GFP-V60, or GFP-V60 and DNM2-V96 as previously described^{2, 3}. After a 72-hr expression period, cells were incubated on ice for 30 mins. Plates were mounted in a heating insert PS1 (Zeiss Microscopy, Jena, Germany) and heated using a refrigerated circulator (#AP07R-40, PolyScience, Niles, Illinois). Temperature was maintained with an Incubator XLmulti S1 DARK LS (Zeiss Microscopy, Jena, Germany) and monitored using the unit’s internal probe. Media was replaced with ice cold imaging solution (#A14291DJ, Life Technologies, Carlsbad, CA) after ice cold dPBS wash. Cells were heated from a starting temperature range between 10 to 15°C and up to a maximum of 45°C within 20 mins. Superresolution images were acquired at least once per min on a LSM880 confocal microscope and focused with Zeiss Definite Focus 2. Temperatures were recorded manually and compared to the sudden appearance of bright structures. Individually observed T_t’s were analyzed in GraphPad using Student’s unpaired t-test between both conditions.

RESULTS AND DISCUSSION

The membrane trafficking community has developed several useful tools to study DNM2, but they are limited by their speed and specificity. Here, we present a potential new peptide-based approach based on a DNM2 fusion to a thermo-responsive polypeptide.

3.1. Generation of thermo-responsive DNM2-ELPs from a PCDNA3.1 plasmid capable of transfecting mammalian cells.

The first step in the development of this tool was to clone human DNM2 into an ELP backbone and insert it into a mammalian expression vector, PCDNA3.1 (Table 1). Forward and reverse Sanger sequencing at the N- and C-termini were used to validate sequences for expression (Supplementary Fig. S1). Diagnostic digests with EcoRI and XbaI flanking the DNM2-ELP sequence were used to confirm the approximate band size of the cloned construct (Supplemental Fig. S2A).

Table 1.

Summary of ELP fusion proteins evaluated in this work.

Nomenclature	Amino acid sequence1	Intracellular Tt (°C)2	Molecular weight (kDa)3
DNM2-V96	FLAG-HA-DNM2-G(VPGVG)96Y	23.13	140.5
DNM2-A96	FLAG-HA-DNM2-G(VPGAG)96Y	NA	137.8
GFP-V60	GFP-G(VPGVG)60Y	37.59	52.3

¹DNM2 and GFP sequences are available in supplementary table S2.

²Observed mean transition temperature by live cell imaging.

³Estimated molecular weight based on the amino acid sequence open reading frame using Snapgene Software 5.3.2.

3.2. Expression of DNM2-ELPs in HEK 293T cells is confirmed by Western blot and immunofluorescence.

After successful cloning of DNM2-ELPs into a mammalian plasmid, DNM2-ELP expression was verified in HEK 293T cells by Western blot and superresolution confocal microscopy using an anti-ELP antibody (AK1) (Fig. 2). The expected molecular weight of DNM2-A96 and DNM2-V96 are 137.8 kDa and 140.5 kDa, but they were measured within a reasonable difference at 150 kDa. The SDS gel was loaded with equivalent lysate protein, but sensitivity differences of AK1 to different ELPs has been reported (Supplemental Fig. S2B). To visually confirm expression of DNM2-V96, cells were fixed at 18 and 37°C and imaged using super-resolution confocal microscopy. DNM2-V96 was expected to form amorphous microdomains above T_t^1–3, but instead they formed highly organized vesicular structures (Fig. 2A). A heterogenous collection of vesicular structures lined with DNM2-ELP signal were observed at higher temperatures, thus defined as ‘transitioned’.

Figure 2: HA-DNM2-ELP expression is confirmed using α-HA and α-ELP antibodies with indirect immunofluorescent microscopy.

A) HEK293T cells expressing DNM2-V96 (yellow arrow) are shown to form temperature-dependent assembly of large intracellular vesicles at 37°C, but not at 18°C. At 18°C, fluorescent signal appears as small points distributed along the PM. Signals are highly overlapped at either temperature. B) Overlap between α-ELP (green) and α-HA (red) fluorescent signals were quantified by colocalization in FIJI, confirming a high MCC for DNM2-V96 at 18°C and 37°C (MCC = 0.93 and 0.93). Each point represents a single cell.

To confirm intact expression of the DNM2-ELP sequence, the HA tag at the N-terminus and ELP-chain at the C-terminus were identified by antibody and subsequent indirect immunofluorescence. Signal overlap from the HA-tag and ELP-chain at each temperature was quantified by quantifying colocalization MCC²⁰ (Fig. 2B). The overlap in signal between the HA tag and the ELP chain were nearly identical at both tested temperatures. The mean MCC was calculated at 0.94 (± 0.04) at 18°C and 0.94 (± 0.05) at 37°C, confirming successful full expression in cells. A Student’s t-test was used to show no changes to the MCC between temperatures (p > 0.9999), indicating no effects on fusion integrity above T_t. While the relative level of expression after transient transfection was higher for DNM2-A96, both the temperature sensitive DNM2-V96 and insensitive control were identified in cellular lysates at the expected MW (Supplemental Fig. S2B)

3.3. DNM2-ELPs form distinct structures as a function of temperature.

Next, DNM2-ELP (A96 and V96) expression was quantified at different temperatures (Fig. 3). From the structures observed in the trial experiment, morphologies were hypothesized to appear at different proportions in a temperature dependent manner. Additionally, it was unknown if additional structures would be observed between 18°C and 37°C, so cells were transfected and fixed at a broad range of temperatures (4, 20, 30, 35 or 40°C) to capture distinct morphologies. ‘Non-transitioned’ cells reflecting temperature-independent assemblies were defined based on the observed morphologies of DNM2-A96 at 4°C and used as a baseline. DNM2-A96 transfected cells revealed either bright ‘puncta’ or ‘diffuse’ staining at all observed temperatures. The combination of puncta and diffuse morphologies constituted 100% of all observations (Fig. 3B). On the other hand, cells expressing DNM2-V96 demonstrated evolving temperature-dependent morphologies that were classified into three differentiated structures above 20°C.

Figure 3: Temperature-sensitive DNM2-V96 forms distinct vesicular structures at different temperatures while temperature-insensitive DNM2-A96 does not.

HEK293T cells were transfected with either DNM2-A96 or DNM2-V96, fixed at 4, 20, 30, 35, or 40°C, and stained for indirect immunofluorescence. A) After Airyscan superresolution imaging, five distinct morphologies were observed between temperature-sensitive and -insensitive DNM2-ELPs. Diffuse and puncta morphologies are indistinguishable between DNM2-A96 and DNM2-V96, whereas Dynasome-E, Dynasome-L, and Dynasome-S morphologies only appear with DNM2-V96 at 20°C and above. B) DNM2-A96 does not appear to transition at any temperature between 4 to 40°C, while C) DNM2-V96 appears to transition at about 20°C. D) The morphology of DNM2-ELP can be characterized as one of five possibilities between 4 to 40°C, but only puncta and diffuse morphologies appear in cells expressing DNM2-A96 at any temperature – no instance of Dynasome-E, Dynasome-L, or Dynasome-S morphologies. E) In cells expressing DNM2-V96, the diffuse and puncta morphologies appear at or below 20°C. The Dynasome-E morphology only appears at 20°C. The Dynasome-L morphology begins to appear at 20°C and peaks at 30°C before falling off at higher temperatures. The Dynasome-S morphology begins to appear at 35°C and dominates by 40°C.

The emergence of long rod-like vesicles lined with DNM2-V96 at 20°C were noted and categorized as the elongated dynasome morphology, (Dynasome-Elongated (E)). Dynasome-E structures appeared at a proportion of 17.7 ± 4.9% of transfected cells only at 20°C. At this same temperature, large round vesicular morphologies (Dynasome-Large (L)) were observed in 15 ± 2.7% of cells. The remaining cell morphologies (67.3%) were non-transitioned diffuse or puncta morphologies. At 30°C, non-transitioned morphologies constituted a mean total of 3 ± 5.2% of DNM2-V96 transfected cells while Dynasome-L morphologies dominated observations by appearing in 90 ± 10.5% of cells at 30°C. The Dynasome-Small (S) structures with smaller, overlapping, and interconnected lumens constituted the remaining proportion of structures (7 ± 12.1%). By 35°C, the Dynasome-S morphology grew in proportion to a total of 68 ± 4.6% while the Dynasome-L morphology decreased to 32 ± 4.6% of cells. This trend continues to 40°C with the dominance of the Dynasome-S morphology, accounting for a proportion of 81.7 ± 6.5% of cells while the Dynasome-L morphology accounts for a proportion of 18.7 ± 6%. In summary, DNM2-V96 begins to assemble intracellular structures above 20°C, which is nearly complete by 30°C (Fig. 3E). No temperature dependent change in DNM2-A96 morphology was observed at any temperature.

3.4. Orthogonal and mesh reconstruction of dynasomes show distinct lumen contraction in Dynasome-S morphologies

Orthogonal projections and mesh reconstructions were used to better visualize 3-dimensional cellular structures formed by DNM2-A96 and DNM2-V96. Representative images were chosen to display expression and key transition characteristics of DNM2-ELPs. Orthogonal projection of DNM2-A96 at 37°C and DNM2-V96 at 4°C show diffuse or puncta spatial distribution. On the xy-axis, DNM2-ELP distribute along the plasma membrane or in the cytoplasm - corroborated on the xz- and xy- axis. (Fig. 4A and 4B). At 37°C, DNM2-V96 generally forms two distinct morphologies, the Dynasome-L and -S morphologies. The xy-, xz-, and yz-axis show the large lumen of a Dynasome-L structure above the nucleus and the Dynasome-S morphology below. The scale of their difference is displayed in the 3-dimensional mesh model, which also exhibits interconnected lumens. Differentiating between structures was achieved by differentiating the more spatially imposing Dynasome-L from the smaller Dynasome-S, which are clearly shown in both an orthogonal projection and a 3D mesh reconstruction (Fig. 4C).

Figure 4: 3-dimensional orthogonal projections of DNM2-ELP show their respective signal distribution on the -xyz axis using α-ELP secondary immunofluorescence.

A) An orthogonal projection (left) of DNM2-A96 at 37°C shows non-transitioned (diffuse and puncta) morphologies within the same cell. Signal is distributed along the plasma membrane and within the cytosol. A mesh reconstruction corroborates this interpretation (right). B) The orthogonal projection (left) of DNM2-V96 at 4°C shows non-transitioned morphology like DNM2-A96, which was corroborated with a mesh reconstruction (right). C) In contrast, the orthogonal projection (left) of DNM2-V96 incubated at 37°C shows both Dynasome-L and –S morphologies. The round morphology of Dynasome-L contrasts significantly with the smaller, interconnected Dynasome-S morphology. The -yx and –xz slice windows show the formation of a large spherical structure lined with DNM2-V96 and an empty lumen. The mesh reconstruction (left) conveys the interconnected lumens of the Dynasome-S.

3.5. Dynasome size depends on dynasore and temperature

Having characterized the temperature-dependent assembly of three distinct dynasome morphologies, we next assessed if their properties depend on GTPase activity by treatment with dyansore (Fig. 5). Compared to an untreated cell (Fig. 5A), there was a dramatic reduction in the assembly of dynasomes in cells treated with 80 μM dynasore (Fig. 5B), clearly indicating an effect to dynasome self-assembly. Manually quantifying this effect as in Fig. 3 required magnification of individual cells, which is impractical with a large dataset. To overcome this limitation, dynasome diameters were instead measured automatically from binary projections of structures. The average dynasome diameter in untreated cells at 37°C was 0.41 μm ±0.33, which is comparable to the average diameter quantified in untreated cells at 40°C at 0.40 ± 0.27 μm (Fig. 5C). After dynasore treatment, the average diameter of structures shrank to at 0.33 μm ±0.24 (Fig. 5D), suggesting that dynasome self-assembly depends on the activity of DNM2. This study demonstrates that dynasore clearly influences the self-assembly of dynasomes and implicates the GTPase function of DNM2 in the formation of dynasome structures.

Figure 5: The diameter of DNM2-V96 structures decreases with temperature and inhibition of Dynamin GTPase activity.

HEK393T cells were transiently transfected and incubated with or without dynasore or at different temperatures. Dynasomes were stained using secondary immunofluorescence (white) and imaged using confocal microscopy. The diameters for a large dataset of these structures were estimated using image analysis. Nuclei were stained with DAPI (Blue). A) At 37°C Dynasome-S morphologies are predominant. B) After incubation with 80 μM dynasore, the size of vesicular structures collapses. C) As temperature increases, the mean diameter of dynasome morphologies decreased significantly (****p<0.0001, ANOVA, Tukey post hoc). D) Having validated that this image analysis reflects the trend observed with temperature, this approach was used to compare the diameter of structures with and without dynasore. Dynasore treatment significantly decreased the diameter (p < 0.0001, two-tailed unpaired t-test).

While most cells displayed just a single dynasome phenotype, some cellular characteristics were difficult to distinguish from each other. For example, Dynasome-S cells have clusters of small vesicular structures, while Dynasome-L cells are defined by large singular vesicular structures. Given these defining characteristics, the diameters detected by image analysis reflect contributions from both Dynasome-S and Dynasome-L. As shown above, a qualitative rubric was used to categorically distinguish these morphologies (Fig. 3); however, that approach relies on arbitrary visual judgements. To complement that analysis and quality control this assessment of dynasore, we manually measured the diameters of each structure as a function of temperature (Fig. 5C). The diameter of dynasomes decreased sharply from 20 to 30°C, and then again from 30 to 35°C, but not between 35 and 40°C. This trend mirrors the observations and analysis described by their categorical variables where Dynasome-L species shifted to Dynasome-S as temperature increases (Fig. 3E).

3.6. Dynasomes appear morphologically independent from F-actin structures

While our experiments show that dynasome assembly is both temperature-sensitive and capable of forming at least three distinct intra-cellular morphologies, it is unclear how their assembly affects the overall cytoskeleton. For example, Dynasome-S morphologies share some similarities to circular dorsal ruffles (CDR). CDRs are unique plasma-membrane structures, which have been observed using F-actin tracking to form and close within 30-mins of stimulation²¹. CDR size and shape are qualitatively similar to dynasomes formed by DNM2-V96. CDR transient structures are defined by and dependent on F-actin and have previously been found colocalized with the actin ring²¹. While DNM and F-actin are involved in common cell biology²², the relationship between DNM2-ELPs and CDRs is unclear. To screen for overlapping effects on cellular biology, cells expressing DNM2-V96 were co-labelled with fluorescent phalloidin as well as antibody against ELP. If dynasome self-assembly relies on an F-actin template, F-actin might be expected to form rings similar to CDRs above T_t. This was not observed. Below T_t (4°C), DNM2-V96 overlaps with cytosolic F-actin near the periphery of the cell (Fig. 6A). Above T_t (37°C), DNM2-V96 formed dynasomes in the cytoplasm, while F-actin remained concentrated at the plasma membrane. Accordingly, the colocalization between F-actin and DNM2-V96 significantly decreased with dynasome assembly; furthermore, there was no obvious effect of dynasomes upon the morphology of F-actin. Both the decrease in colocalization and the absence of F-actin assembly peripheral to dynasomes suggests that dynasomes rely on an independent self-assembly process (Fig. 6B). Visually, dynasomes and F-actin are spatially averse above T_t, which is consistent with a decrease in colocalization indicated by the MCC (Fig. 6C).

Figure 6. Formation of dynasome morphologies do not appear to be driven by cytoskeletal dynamics.

HEK 293T cells were transiently transfected with DNM2-V96 and stained by secondary immunofluorescence for actin (red), ELP (green), or nuclei (blue) upon incubation above Tt (40 °C) or below Tt (4 °C). (A) Below Tt, cells expressing DNM2-V96 show an overlap in ELP and F-actin near the periphery of the cytosol. Diffuse and puncta morphologies distributed in the cytoplasm did not overlap as highly with F-actin. (B) Above Tt, F-actin remained at the PM whereas non-overlapping dynasome structures formed in the cytoplasm. (C) Comparing the whole cell colocalization (MCC) of both channels to quantify the overlap indicates that their localization decreases significantly upon dynasome assembly (*p = 0.02). This reflects the cytoplasmic localization of assembled dynasomes, which is clearly distinct from the strong PM association observed for F-actin structures. Dynasome assembly does not appear to significantly change the F-actin structural morphology.

3.7. Dynasome formation is visualized in live-cells with a GFP-based surrogate reporter

While fixed cells facilitate exquisite characterization of cells at specific temperatures, we cannot exploit the rapid thermo-responsiveness of ELPs to characterize their behavior. ELPs are visually sensitive to changes in temperature as small as 1°C, so live cell imaging would allow for dynamic visualization of transitional structures missed in fixed-cells incubated at selected temperatures. Our group has developed methodology for indirectly visualizing ELP assembly using a co-expressed fluorescent reporter. Co-expressing DNM2-V96 with a surrogate reporter, GFP-V60 enables visualization of DNM2-V96 microdomains in live-cells^{2, 3}. GFP-V60 was expressed in HEK 293T cells and cooled before heating at a steady rate to obtain images at consistent intervals (Fig. 7). By itself, GFP-V60 assembled microdomains above a mean temperature of 37.6 ± 4.0°C (n = 18) (Fig. 7A). When co-expressed with DNM2-V96, the assembly temperature of GFP-V60 significantly (p<0.0001, Student T-test) decreased to a mean T_t of 21.8 ± 4.2°C (n = 22) (Fig. 7B). Co-expressed DNM-V96 resulted a lower observed transition compared to the shorter-chained GFP-V60, which is consistent with ELP dependence on chain length (Fig. 7C). Unexpectedly, the vesicular morphologies observed by fixed-cell studies were not observed.

Figure 7. Live cells expressing either GFP-V60 (single) or GFP-V60 and DNM2-V96 (dual) were imaged by confocal microscopy to quantify the Tt of DNM2-V96.

(A) Two individual HEK 293T cells transfected with GFP-V60 were heated from 13.3 to 44.3 °C over 20 min to show Tts of 37.2 and 43.5 °C, respectively. This reflects normal cell-to-cell variability observed previously. (B) Cells transfected with GFP-V60 and DNM2-V96 were heated from 15.9 to 44.1 °C (showing 36.4 °C max) to show a Tt of 19.3 °C. (C) Multiple cells were evaluated using three independent temperature ramps, and the transition temperature in each evaluated visually. The mean Tt between single (n = 18) and dual transfected (n = 22) cells show a statistically significant difference (****p < 0.0001). Single transfected cells transitioned at a mean of 37.9 °C (α4.0 n = 18) with a 95% CI between 35.64 and 39.5 °C, whereas dual transfected cells transitioned at a mean of 21.8 °C (α4.2, n = 22) with a 95% CI between 21.3 and 25.0 °C. This is consistent with the assembly of transitioned cells in Figure 3C. (D) Fixed cell imaging was used to confirm the association of DNM2-V96 and GFP–V96 following incubation at 37 °C. Orthogonal projections were visualized by indirect immunofluorescence microscopy by α-HA (red) and α-GFP antibodies (green). Above Tt, they illuminate the formation of distinct GFP-V60 microdomains encapsulated within DNM2-V96 dynasome-S structures.

While the general globular shapes observed in live-cells resembles those in fixed-cell studies, it was unclear if the core of globular structures contained a mixture of miscible DNM2-V96 and GFP-V60. Other ELP temperature-sensitive fusions; CAV1-ELP and EGFR-ELP were similarly characterized and were observed to form miscible microdomains with GFP-V60. For example, CAV1-V96 formed very large amorphous structures localized at the plasma membrane, which were highly colocalized with GFP-V60. The morphology of the surrogate reporter in live-cell characterization was similar to the morphology observed in fixed-cells imaged by 3-dimensional super resolution confocal microscopy². Co-expression of EGFR-V96 and GFP-V60 or single expression of EGFR-GFP-V96 formed smaller round microdomains that are indistinguishable from each other in fixed and live-cells³. Given that context, the result of co-expression between DNM2-V96 and GFP-V60 was especially surprising.

The difference in T_t between singly and co-expressed DNM2-V96 suggests that DNM2-V96 affects the spatial organization of GFP-V60. Based on previous observations, dynasome structures were expected to be directly stained by GFP-V60 co-expression, but they instead appeared as two immiscible fractions. DNM2-V96 still forms vesicular structures above T_t, but they do not appear to be miscible with GFP-V60. Instead, GFP-V60 is encapsulated within the lumina of the formed structures (Fig. 7D). In a different study, miscibility in mixed ELP populations were recently characterized²³. The T_t of ELP mixtures in a water-in-oil emulsion revealed staggered phase-separation into immiscible ELP microdomains. ELPs with different characteristics were programmed to transition from miscible or immiscible mixtures with distinctly observed T_t’s corresponding to separate ELPs in the mixture. Our live-cell co-expression experiments are more nuanced because the reorganization of DNM2-V96 appears to entrap GFP-V60 without inducing colocalization. The behavior of this mixture appears to be influenced by dynasome formation because V96 fusions to other effector proteins like CAV1 or EGFR form a single structure where they colocalize with GFP-V60. This suggests that dynasome-based assembly with other ELP species is more complex than ELP-ELP coacervation. Rather, dynasomes and/or DNM2 biology aids in the selective entrapment of GFP-V60 through some interaction beyond the scope of this report.

3.8. Dynasome radius of curvature may depend on a balance between protein folding of DNM2 and ELP domains.

ELP fusions to intracellular CLC, CAV1, EGFR, and GFP self-assemble into condensed microdomains above T_t, and colocalize with GFP-V60 in live cells. In contrast, the striking formation of three distinct dynasome structures does not match the morphology observed for these other fusions. Instead, DNM2-V96 forms lamellar structures similar to those observed with diblock polypeptide fusions^24–28. For example, an amphiphilic diblock ELP comprised of a glutamic acid region (GEGVP_n) and a phenylalanine rich region (GFGVP_n) fused together (EF) promote spontaneous formation of a bilayer with a hydrophobic core and a hydrophilic exterior²⁹, analogous to amphiphilic DNM2-V96, driving dynasome self-assembly (Fig. 1).

To our knowledge, DNM2-V96 is the first polymer fusion to direct self-assembly of DNM2-mediated vesicular structures inside mammalian cells. While the ELP diblock fusions offer insight into self-assembly, the evolution of Dynasome-L to Dynasome-S is mysterious. The scission mechanism of DNM may offer insight into this nuanced behavior. Vacuolar protein sorting protein 1 (VPS1) is a DNM-like protein expressed in yeast that shares N-terminus GTPase and C-terminus GTPase effector domain homology with DNM. It is a key regulator of fusion and fission events³⁰. For example, a ΔVPS1 mutation results in a vacuolar fission defect that leads to an accumulation of multivesicular bodies (MVBs) similar in appearance to the Dynasome-L morphology. wt-VPS1 expression rescued the fission defect, suggesting that GTPase activity contracts vesicle size³¹. Thus, it would follow that dynasome curvature and diameter is regulated by the GTPase biology of DNM2; however, dynasomes are smaller after treatment with dynasore (Fig. 5), which contradicts a model suggesting a decrease in dynasome curvature is driven by GTPase contraction. Nevertheless, it is possible that a relaxed DNM2-V96 state may promote dynasome curvature and a smaller diameter.

The role that DNM2 plays in dynasome evolution might be explained by ELP phase transition. At the T_t, ELPs self-assemble after a conformational change in its backbone that favors hydrophobic intramolecular interactions between residue hydrogens and the backbone^{32, 33}. ELPs assemble an amorphous, optically-clear liquid phase; however, with additional heating many ELPs adopt a rigid, crystalline phase. This is commonly observed during purification of ELPs and their fusion proteins. As such, it is plausible that the temperature-dependent conversion from Dynasome-L to Dynasome-S reflects an additional phase change within the ELP-rich lamellar phase. If this were the case for DNM2-V96, then ELP lamellae below 35°C would be less rigid than above. An increased crystallinity of the ELP-phase reduces the radius of curvature, which collapses Dynasome-L into smaller Dynasome-S vesicles. Therefore, it is reasonable that changes in dynasome curvature might be due to a balancing act between DNM2 and V96 (Fig. 8).

Figure 8: Dynasome contraction potentially occurs as a result of a tighter radius of curvature.

A) A balance between temperature-dependent ELP self-assembly and DNM2 biochemistry drives the formation of dynasome structures. Dynasome-L self-assembles a lamellar phase that encloses a larger diameter, B) but as temperature increases, the average radius of curvature for DNM2-V96 decreases until the Dynasome-S morphology is dominant.

CONCLUSIONS

Functionalized intracellular ELP fusions have previously been reported to form coacervates above their respective T_t. In contrast, heated DNM2-V96 forms organized vesicular dynasomes with a much more significant hollow interior. Nonetheless, this reorganization shows that ELP dynamin fusions do self-sequester, similar to V96-CLC microdomains. The V96-CLC clathrin platform inhibits CME by sequestering endogenous clathrin heavy chain, which is necessary for internalization. In the case of clathrin, this prevented endocytosis of two GPCRs above the phase transition temperature¹. We present strong microstructural evidence that dynasome organizations forms similar, albeit more distinctive assemblies. Similarly, dynasomes have the potential to sequester DNM2-V96 further away from CME machinery, which may prevent a key scission step for vesicles undergoing either CME or CAVME^{34, 35} or secretion from the trans-golgi network³⁶. Additionally, DNM2-V96 may be of use in answering questions about DNM’s role in less-characterized pathways, such as flotillin mediated endocytosis^{37, 38}. As such, the rapid assembly of dynasomes has potential applications to study and modulate a variety of dynamin-dependent biological processes.

Here, we conclude that a novel temperature-sensitive DNM2-ELP fusion forms three distinct intracellular dynasomes at temperatures above 19.3°C. Unexpectedly, the dynasomes pass through a range of morphologies from elongated tubes, to large spherical vesicles, and then smaller interconnected vesicles. This process was inhibited by dynasore, a DNM2 GTPase inhibitor, which suggests the DNM domain retains biochemistry relevant to assembly. Despite a linkage between F-actin biology and dynamin, dynasomes exclude F-actin and had little effect on F-actin cellular morphology. Interestingly, dynasome assembly efficiently entraps cytosolic GFP-ELP within minutes. In addition to the temperature-dependent assembly of dynasomes, their rapid activation may have potential to modulate and study dynamin-dependent biological processes.

ABBREVIATIONS

AK1

mouse anti-ELP antibody

BSA

bovine serum albumin

CAV1

caveolin-1

CAVME

caveolin-mediated endocytosis

CDR

circular dorsal Dynasome-S

CHC

clathrin heavy chain

CIE

clathrin-independent endocytosis

CTB

cholera toxin subunit B

CLC

clathrin light-chain

CME

clathrin-mediated endocytosis

DNM

dynamin

DNM-K44A

dominant-negative dynamin without GTPase ability

dPBS

Dulbecco’s phosphate-buffered saline

ELP

elastin-like polypeptide

FBS

fetal bovine serum

FLOTs

flotillin-1 and flotillin-2

FME

flotillin-mediated endocytosis

MCC

Mander’s colocalization coefficient

PM

plasma membrane

RT

room temperature

TGN

trans-golgi network