Membranes arrest the coarsening of mitochondrial condensates in human cells

Abstract

Mitochondria contain double membranes that enclose their contents. Within their interior, the mitochondrial genome and its RNA products are condensed into ~100 nm sized (ribo)nucleoprotein complexes. How these endogenous condensates maintain their roughly uniform size and spatial distributions within mitochondria remains unclear. Here, we engineer optogenetic tools (mt-optoIDR) that enable controlled formation of synthetic condensates within live mitochondria upon light activation in HeLa cells. Using high-resolution microscopy, we visualize the nucleation of small, yet elongated condensates (mt-opto-condensates), which recapitulate the morphologies of endogenous mt-condensates. These narrow size distributions are independent of mt-optoIDR sequence features, suggesting the mitochondrial environment influences condensate formation. Consistently, mt-opto-condensates fluctuate within voids in between cristae in tubular mitochondria. To directly isolate the contribution of the mitochondrial membranes, we overexpress the dominant negative membrane fusion mutant (Drp1^K38A), which results in the formation of bulbous mitochondria with restructured cristae. Based on quantitative particle tracking, bulbous mitochondria support significantly increased dynamics and rapid coarsening of mt-opto-condensates into a single, prominent droplet–in contrast to the membrane confinement observed in tubular mitochondria. Together, these observations inform how membranes can constrain the growth and dynamics of the condensates they enclose, without the need for additional regulatory mechanisms.

Optogenetically controlled phase separation in live mitochondria reveals how membranes influence the morphology and dynamics of biomolecular condensates.

Introduction

Mitochondria are double-membrane-bound organelles that commonly form dynamic, tubular networks spanning the cytoplasm. Their membranes help contain multiple copies of their own genome (mtDNA), which encodes key subunits needed for oxidative phosphorylation. The mitochondrial genome is not freely diffuse within the matrix, but is packaged into discrete ~100 nm nucleoprotein complexes called mt-nucleoids¹. Under normal conditions, these structures tend to be distributed uniformly throughout the mitochondrial network², yet frequently assemble into large clusters in response to various forms of stress^3,4. Indeed, there are other examples of distinct nucleic acid and/or protein complexes within the mitochondrial matrix, including mtRNA granules^5–7, protein translation compartments^8,9, and mitochondrial stress bodies⁸. However, the mechanisms by which biomolecules organize into higher-order structures within mitochondria are still largely unclear.

Weak, multivalent interactions among biomolecules can promote the entry into a phase transition, resulting in the formation of condensed, liquid-like phases within cells¹⁰. Often these interactions involve intrinsically disordered regions (IDRs)^11–16, which can be sufficient to drive phase separation. Indeed, the molecular grammar driving condensation has been well described in the cytoplasm and nucleus^17,18. However, only recently has phase separation been observed for mt-nucleoids and RNA granules within mitochondria^3,19,20; importantly, the molecular driving forces are largely unexplored. Moreover, mitochondria represent a unique physicochemical environment in the cell, in part due to their highly membranous structure^21–23 and high metabolic activities^24–26, and it remains largely unclear how the mitochondrial physicochemical environment affects the ability of its contents to undergo phase separation.

One tool used to non-invasively study phase separation in live cells involves optogenetics^27–31. These engineered constructs typically contain an optogenetic protein that dimerizes or oligomerizes in response to light and a sequence capable of undergoing phase separation, such as that of an IDR^27,30–32. Light activation thus allows spatiotemporal control of condensation, facilitating detailed mapping of phase behavior within a given subcellular environment. Here, we develop an optogenetic construct, mt-optoIDR, to directly study phase separation in live mitochondria. With this optogenetic tool targeted to the mitochondria, we report how mitochondrial membrane architecture influences fundamental aspects of phase separation: droplet morphology, dynamics, nucleation, and dissolution.

Results

mt-optoIDR undergoes a concentration-dependent phase transition in live mitochondria

We adapted an optogenetic approach to induce the formation of condensates in live mitochondria. Our design of an optogenetic construct (mt-optoIDR, Fig. 1a) was based on the established optoDroplet constructs used in the nucleus and cytoplasm³¹. Specifically, our first prototype retained the optogenetic protein CRY2olig that undergoes reversible protein–protein clustering in response to blue light^31,33 and its fusion to the red fluorescent marker protein mCherry. Given the unique physicochemical environment of the mitochondrial matrix, we prioritized using an IDR from the mitochondrial proteome that would contain endogenous sequence features. After bioinformatic analysis of the mitochondrial proteome (Fig. S1), we identified the disordered N-terminal region of DDX28 (1–127 amino acids) that also contained a mitochondrial targeting sequence (MTS), ensuring our construct would be localized exclusively to the mitochondria. We confirmed the disordered nature of our identified IDR using D²P² disorder predictions³⁴ and Alphafold2 (Fig. S1b)³⁵.

Fig. 1. Light-induced phase separation of mt-optoIDR within the mitochondrial matrix of live cells.

a Schematic of the mt-optoIDR construct containing a mitochondrial targeting sequence (MTS, 1–23) and an intrinsically disordered region (IDR) from DDX28 (24–127 aa) fused to a fluorophore mCherry and an optogenetic protein CRY2olig. b Schematic diagram illustrating light activation experiment using 488 nm light to induce phase separation of mt-optoIDR within the mitochondrial matrix (see the “Methods” section). Pre-activation (blue light OFF, no droplets, c) and post-activation (blue light ON, droplets formed, d) of live HeLa cells expressing mt-optoIDR (green). Puncta in d are phase-separated mt-opto-condensates. Inset shows a zoomed-in version of the mt-opto-condensate, indicated by the arrow. Scale bars = 5 µm (main) and 0.5 µm (inset). A liquid-like fusion event of mt-opto-condensate droplets (e) and the corresponding line profiles (f). Arrowheads point to droplets undergoing coalescence. Scale bar = 1 µm. Intensities are normalized based on the first frame. 19 cells were obtained with similar results.

After transient co-expression of mt-optoIDR and a marker to visualize the mitochondrial network (Halo-MTS)³⁶, we observed that mt-optoIDR localized diffusely within the mitochondrial matrix (Figs. 1c and S2a). Upon five minutes of blue light activation to the entire cell (global activation), mt-optoIDR appeared to undergo phase separation, forming dozens of droplet-like structures (mt-opto-condensates) throughout the mitochondrial network of a HeLa cell (Figs. 1b, d and S2a–d, and Supplementary Video 1).

We next performed a series of experiments to confirm that mt-optoIDR exhibited features representative of a classic phase transition. First, we observed that neighboring mt-opto-condensates rapidly fused together, consistent with liquid-like behavior (Fig. 1e, f). Second, the mt-opto-condensates also represented de novo phases, as indicated by the lack of partitioning into existing endogenous condensates (Fig. S4). Third, we compared droplet formation as a function of expression level or concentration of our construct. We found that for low concentrations, mt-optoIDR remained soluble, but above a concentration threshold, formed a second, condensed phase (Fig. S3c–e). As a control, mt-opto-condensate formation occurred in the absence of our marker of the mitochondria (Halo-MTS) (Fig. S3a and b), but was sensitive to high levels of the mitochondrial marker (Fig. S3c). Together, these results show that our optogenetic construct can be used to induce a concentration-dependent phase transition into a condensed, liquid-like phase within live mitochondria.

mt-opto-condensates exhibit narrow size distributions irrespective of sequence features

One prominent feature of the mt-opto-condensates was that they appeared relatively small (Figs. 1d, S2, and Supplementary Video 1), particularly when compared to reports of micron-sized opto-droplets formed in the nucleus and cytoplasm^27,31. We performed quantitative image analysis to measure their size (see the “Methods” section) and observed several key features. First, the size distributions of the major and minor droplet axes were relatively narrow, ranging from a few hundred nanometers (~100 nm), suggesting these droplets were limited in their ability to grow into larger droplets. Second, we observed a significant difference between the major and minor axes (p = 1.03 × 10^-148): we estimated the droplet major axis diameter to be 270 ± 50 nm (mean ± s.d., n = 533) and droplet minor axis diameter to be 190 ± 20 nm (mean ± s.d., n = 533) (Fig. 2a), yielding an average aspect ratio (AR) of 1.5 ± 0.3 (mean ± s.d., n = 533) (Fig. 2b), indicating that these condensates were not perfectly spherical (AR = 1.0) as would be expected for a droplet at equilibrium. Intriguingly, these morphologies are roughly consistent with those of endogenous mitochondrial condensates^20,37,38.

Fig. 2. mt-opto-condensates exhibit narrow size distributions irrespective of sequence features.

a Quantitative analysis of major (purple) and minor (gray) axes of mt-opto-condensates (DDX28 IDR) formed in live HeLa cells (n = 533 droplets from 14 cells, p-value 1.03 × 10⁻¹⁴⁸, Mann–Whitney U test, two-sided). Inset includes an annotation of the major and minor axes. b Aspect ratio changes of mt-opto-condensates (DDX28 IDR) (n = 533 droplets from 14 cells). Inset includes schematic diagrams illustrating droplets with aspect ratios (AR = 1, 2, and 3) for reference. c mt-opto-condensates formed with different IDR sequences from DDX28, GRSF1, TWINKLE, and FUS followed by a negative control lacking the oligomerization domain ($\mathrm{\Delta }$CRY2olig). Representative images upon whole plate activation and fixation. Insets show a zoomed-in version of the mt-opto-condensate for each construct. Scale bars = 2 µm (main) and 0.5 µm (inset). d Quantification of droplet major or minor axis size for mt-opto-condensates for different IDRs in (c) (Kruskal–Wallis H-test p-value for major axis 3.45 × 10⁻⁷⁴, one-way ANOVA test for minor axis p-value 1.92 ×  10⁻⁶, overlap coefficient for all pairs of IDRs, >0.70 in major and >0.85 for minor axis). e Partition coefficients analysis for mt-opto-condensates from different IDRs in (c) (one-way ANOVA p-value 4.13 × 10⁻¹⁹⁹). For each condition in d and e, condensates were analyzed from 15 cells. n: DDX28 IDR = 700, GRSF1 IDR = 619, TWINKLE IDR = 531, FUS IDR = 855 mt-opto-condensates. For $\mathrm{\Delta }$CRY2olig, 15 cells were obtained with similar results.

We next tested if the observed morphology was due to the sequence-encoded features of our construct. We designed several variants in which we replaced the MTS-IDR sequence of DDX28 with those identified from our bioinformatic analysis (Fig. S1), specifically from the N-terminus of GRSF1 (1–150 aa) and Twinkle (1–170 aa). As a positive control, we also tested the IDR from the protein Fused in Sarcoma (FUS, 1–214 aa), which has been previously engineered to form optoDroplets in the nucleus and cytoplasm³¹. We included two negative controls, in which we removed the IDR entirely (ΔIDR) and the ability to oligomerize (ΔCRY2olig). We found that all constructs that contained IDRs (GRSF1, Twinkle, and FUS), regardless of sequence, formed condensates with similar morphologies as our prototype DDX28(IDR)-mCherry-CRY2olig (Fig. 2c). The exact mean droplet size and partition coefficient varied across the constructs (Fig. 2d and e, p < 0.001), yet the overall distributions were largely similar (overlap coefficient >0.7 for all pairs of IDRs for both major and minor axis). Moreover, the measured sizes were comparable to the size of structures formed upon oligomerization (ΔIDR) (Fig. S3f–h), highlighting an inherent limit to condensate growth. We confirmed that without the activation domain (ΔCRY2olig), droplet formation was abolished (Fig. 2c). These observations support that the size distribution of mt-opto-condensates is not strongly dependent on sequence features, but rather is likely determined by the mitochondrial environment.

mt-opto-condensates are elongated in the direction of tubular mitochondria and found in voids in between cristae

We thus reasoned that the tubular nature imposed by mitochondrial membranes may play a role in limiting the growth and relaxation of the droplets. If so, the membranes could compress or deform the droplets, resulting in the observed deviation from a spherical shape. To test this hypothesis, we defined a parameter (theta, θ), which is the angle between the mitochondrial axial axis and the droplet major axis (Fig. 3a and b). If the droplets were being deformed by the mt-membrane, we would expect the droplets to be aligned with the mt-membrane, with correspondingly small theta values. Indeed, the distribution of theta measured for elongated droplets resulted in small angles (θ = 18˚ ± 1˚, circular mean ± circular s.e.m., n = 230), reflecting that the droplets tended to be elongated in the direction of the mitochondrial axial axis of wild-type, tubular mitochondria (Fig. 3c).

Fig. 3. mt-opto-condensates are elongated along the mitochondria and fluctuate in voids in between cristae.

a Schematic diagram illustrating the angle (theta, θ) between the droplet major axis (dotted line) and mitochondrial axial axis (solid line). b An image of the mt-opto-condensate indicating the measured angle, θ. Scale bar = 0.25 µm. c Polar histogram showing the distribution of the measured angle (θ) (n = 230 droplets from 14 cells). d Airyscan 2.0 live imaging of mt-optoIDR (green) and cristae (magenta) in a HeLa cell. White arrowheads point to mt-opto-condensates. Scale bar = 2 µm. Dotted line e represents a mitochondrion without mt-opto-condensates and f represents an mt-opto-condensate positioned between cristae. e and f are corresponding line profiles. g Time-lapse showing the dynamics of mt-opto-condensates (green) and cristae (magenta). Scale bar = 1 µm. h line profiles of intensity distribution of mt-opto-condensate and cristae during the time-lapse in g. The line profiles were centered around the peak of mt-opto-condensate (x = 0) for each time point. At least 10 cells were imaged with similar results.

To confirm how mt-opto-condensates were organized relative to the mitochondrial membranes, we visualized mt-opto-condensates simultaneously with the dye Live Red mito that accumulated in the mitochondrial inner membrane, including cristae folds (Fig. 3d). First, we observed that the soluble fraction of mt-opto-condensate had inverse localization with cristae (Fig. 3e), consistent with the matrix localization of our construct. Second, the mt-opto-condensates were found positioned in prominent voids in between the cristae; these voids appeared greater than the typical cristae spacing (Fig. 3f). Next, we performed live imaging and found that both structures underwent dynamic fluctuations relative to each other within the mitochondria. In these experiments, we found the mt-opto-condensates rearranged within the voids demarcated by cristae, but remained elongated in the axial direction of the tubular mitochondria for the majority of the time (Supplementary Video 2). These results suggest that the cristae likely limit mt-opto-condensate mobility, but the outer membrane and/or inner boundary membranes are likely defining their overall structure. Together, these results point to the overall architecture of the mitochondria influencing the morphology and positioning of mitochondrial condensates.

To specifically isolate the role of the mitochondrial membrane in shaping condensate morphology, we overexpressed the dominant negative mutant of the mitochondrial fission protein Drp1 (Drp1^K38A). Inhibition of Drp1 is known to lead to hyperfused mitochondria and even more extreme phenotypes, including bulbous structures (Figs. S5 and S6a–c and Supplementary Video 6)^20,39–41. Consistent with previous reports^20,39, overexpression of Drp1^K38A mutant promoted the formation of swollen, bulbous mitochondria, allowing us to decouple the contribution of the double membranes from the matrix environment (Fig. S6a–c). We note that there were differences in the expression level of our mt-optoIDR construct in tubular and bulbous mitochondria. The bulbous mitochondria were associated with higher levels of the mt-optoIDR construct, roughly twice as concentrated (p = 0.002, Fig. S6d). We repeated the light activation experiments and found a single prominent mt-opto-condensate within each bulbous mitochondrion, indeed reaching considerably larger sizes up to ~500 nm (Fig. S6e). Interestingly, these enlarged condensates were highly dynamic and underwent significant shape fluctuations (Fig. S6k), frequently resulting in irregular morphologies (Fig. S6j–n). Together, these experiments indicate that mitochondrial membranes influence the morphology of condensates.

Mitochondrial membranes limit the diffusion of mt-opto-condensates

To explain why condensates were forming larger structures in the bulbous mitochondria, we performed time-lapse experiments to capture their dynamics. In wild-type tubular mitochondria, we noticed that mt-opto-condensates primarily moved along the direction of the mitochondrial axial axis (dimension X, Fig. S7a), while being confined to the diameter of the mitochondria (dimension Y) (Fig. 4a–c, Supplementary Videos 3 and 4). We observed significantly more dynamic behavior, irrespective of directionality, in the bulbous mitochondria (Fig. 4d–f, Supplementary Video 5). We note that we did not correct for any bulk motion of the organelle itself; thus, our trajectories of mt-opto-condensates likely reflect a combination of passive and/or active fluctuations of the condensates within the matrix as well as the directed motion of the organelle, which appeared to be more significant in the axial direction (X) for tubular mitochondria whereas the bulbous mitochondria were largely stationary.

Fig. 4. Diffusive motion of mitochondrial condensates is influenced by mitochondrial membrane structure.

2D tracking of the mt-opto-condensate in tubular, wildtype mitochondria (image (a), track (b)). Arrowhead shows the mt-opto-condensate. Flat-headed arrows indicate the mitochondrial diameter. c Absolute displacement (Δd) for a lag time, $\tau$, of 2 s, where ΔX (purple) and ΔY (gray) correspond to the mitochondrial axial axis and transverse axis, respectively. 2D tracking of mt-opto-condensate in a bulbous mitochondrion (Drp1^K38A) (image (d), track (e)). Arrowhead shows tracked mt-opto-condensate. Flat-headed arrows indicate the mitochondrial diameter. Scale bar = 1 µm in (a) and (d). f Absolute displacement (Δd) for a lag time, $\tau$, of 2 s. g Displacement probabilities of mt-opto-condensates movements in wild-type tubular (square) and bulbous Drp1^K38A mitochondria (circles). 2D displacement is decoupled into X (purple) and Y (gray) dimensions after individual tracks were aligned to X-axis (see the “Methods” section). h Mean squared displacement (MSD) of mt-opto-condensates in tubular wild-type and bulbous Drp1^K38A mitochondria. Diffusive exponents (α) are ${\alpha }_{{\mathrm{WT}}_X}$ = 0.67 ± 0.06, ${\alpha }_{{\mathrm{WT}}_Y}$ = 0.43 ± 0.05, ${\alpha }_{{{\mathrm{Drp}1}^{\mathrm{K}38\mathrm{A}}}_X}$ = 0.18 ± 0.04, and ${\alpha }_{{{\mathrm{Drp}1}^{\mathrm{K}38\mathrm{A}}}_Y}$ = 0.28 ± 0.08. The errors are presented as 95% confidence intervals of the standard error of the fit, where n = 1310 and n = 25 tracks were identified from 11 wild-type and 10 Drp1^K38A -transfected HeLa cells, respectively.

To further quantify the dynamics of mt-opto-condensates in wild-type mitochondria, we developed a particle tracking image analysis workflow (see the “Methods” section). We extracted the mt-opto-condensate tracks within a tubular mitochondrion and rotated individual tracks to be parallel to the X-axis in a 2D Cartesian coordinate (Fig. S7a–d), such that the new coordinate system had the mitochondrial axial axis as the X-axis and the transverse axis as the Y-axis (Fig. S7a). For example, when absolute displacements (|Δd|) of a mt-opto-condensate were calculated between consecutive frames in a tubular wild-type mitochondrion (Fig. 4a–c), mt-opto-condensates exhibited higher displacements ($\mathrm{\Delta }X$ > 100 nm) along the mitochondrial axial axis compared to the transverse axis ($\mathrm{\Delta }Y$ < 100 nm) (Fig. 4c), which is consistent with the dimensions of the voids from our imaging of cristae flanking mt-opto-condensates (Fig. 3f). In comparison, when we performed the same analysis for a mt-opto-condensate in bulbous mitochondria with disrupted membranes, displacements were significantly higher in both dimensions ($\mathrm{\Delta }X$, $\mathrm{\Delta }Y$ »100 nm) yet bound by the diameter of the bulbous mitochondria (Fig. 4d–f). Moreover, the displacements in both X and Y dimensions were indistinguishable, indicating the mt-opto-condensates were able to freely sample a majority of the matrix within a bulbous mitochondrion (Fig. 4d–f and Supplementary Videos 5 and 8).

We performed further analysis on the mt-opto-condensate tracks within tubular and bulbous mitochondria. We calculated the displacement probability distributions in both X and Y dimensions for a given lag time ($\tau$ = 20 s) (Fig. 4g). For Brownian motion in a simple liquid, the displacements in X and Y dimensions are expected to be independent of each other, exhibiting Gaussian distributions. Whereas, for tubular mitochondria, the displacements of mt-opto-condensates in X and Y were significantly different, where motion in Y was largely confined. However, displacements in bulbous mitochondria had broader distributions than for tubular mitochondria, and the distributions were roughly identical in both X and Y dimensions (Figs. 4g, and S7e).

Lastly, we computed the mean squared displacement (MSD) for both wild-type mitochondria and bulbous mitochondria in each dimension (X, Y) as a function of lag time, τ (Fig. 4h). We found power-law behavior across the lag times we experimentally sampled. For both wild-type and bulbous mitochondria, diffusion was highly sub-diffusive as characterized by a diffusive alpha exponent (α) that was less than 1 (α < 1). Specifically, MSD for wild-type tubular mitochondria in X dimension (along the mitochondrial axial axis) shows anomalous sub-diffusive behavior (α = 0.67 ± 0.06, slope and 95% confidence interval), whereas the diffusive exponent was smaller in the Y dimension (transverse axis, α = 0.43 ± 0.05, slope and 95% confidence interval). However, we observed significantly different behavior in bulbous mitochondria. The MSD values were an order of magnitude larger in bulbous mitochondria than in tubular mitochondria, reflecting that the mt-opto-condensates were much more mobile when the membranes were spaced apart. However, bulbous mitochondria had lower diffusive alpha exponents (α), roughly identical for each dimension (α in X = 0.18 ± 0.04, α in Y = 0.28 ± 0.08, slope and 95% confidence interval), which can be understood due to the mt-opto-condensate’s rapid diffusion within the confined bulb.

One key feature of Drp1^K38A cells was that they exhibited a mixed phenotype, containing hyperfused elongated mitochondria and bulbous mitochondria (Supplementary Video 6), which thus provided an internal control (Fig. S6f–h)). We first measured the size of mt-opto-condensates formed in hyperfused, elongated mitochondria relative to those in bulbous mitochondria from the same Drp1^K38A cell (Fig. S6i) and found analogous size differences as when we compared mt-opto-condensates in wild-type mitochondria to Drp1^K38A bulbous mitochondria (Fig. S6e). Next, we compared the dynamics of an mt-opto-condensate that sampled both tubular and bulbous morphologies within a club-shaped mitochondrion (Fig. 5a–d)³⁹. Indeed, we saw significantly larger displacements when the mt-opto-condensate was in the bulbous region relative to the tubular region (Fig. 5e, Supplementary Video 7). Together, the increased size and dynamics of mt-opto-condensates specifically in bulbous mitochondria within the same Drp1^K38A cell support the notion that local membrane structure directly influences condensate behavior.

Fig. 5. Differences in mt-opto-condensate dynamics are due to local membrane architecture.

mt-opto-condensate sampling bulbous (a, c) and tubular (b) morphologies in a club-shaped mitochondrion within the same Drp1^K38A overexpressing HeLa cell. Scale bar = 1 µm. d Trajectory of the mt-opto-condensate from a–c. e Absolute displacement (Δd) between frames (lag tme τ, of 2 s) with annotations of different morphologies sampled from a–c. f A representative image of cristae (magenta) and non-activated mt-optoIDR (green) in a Drp1^K38A overexpressing HeLa cell. Scale bar = 5 µm. g Diffusion of an mt-opto-condensate (green) inside a bulbous mitochondrion (magenta) induced by Drp1^K38A overexpression. Scale bar = 1 µm. Below, a schematic illustration shows a relative localization of mt-opto-condensate (green) and cristae (magenta); cristae markings serve only as a guide and are not the exact structure. At least 9 cells were activated and imaged with similar mt-opto-condensate dynamics relative to cristae.

Lastly, to confirm the increased dynamics in mitochondrial bulbs were due to changes in membrane architecture, we co-visualized the mt-optoIDR construct with the inner membrane marker Live Red mito. We found that bulbous mitochondria contained large regions of their matrix devoid of cristae (Fig. 5f), unlike tubular mitochondria. Moreover, within these bulbous mitochondria, mt-opto-condensates rapidly diffused around the restructured cristae in the expanded matrix (Fig. 5g, Supplemental Video 8), consistent with our biophysical measurements (Fig. 4). Together, these results show that the mobility of mt-opto-condensates is ultimately determined by membrane-imposed boundaries.

mt-opto-condensates undergo nucleation, growth, and dissolution, yet coarsen into prominent droplets only within bulbous mitochondria

From our global light-activation experiments on the entire cell, we observed that the bulbous mitochondria tended to contain a single, prominent mt-opto-condensate, whereas tubular mitochondria were filled with multiple, smaller condensates. We next sought to test how the membrane architecture contributed to the number of droplets formed. We performed a series of local light activation experiments in which we applied light locally to a small region of the cell and imaged concurrently with activation to capture the nucleation process in real time (see the “Methods” section).

Upon local activation in wild-type tubular mitochondria, mt-opto-condensates began to nucleate after ~100 s of activation and followed by a plateau in number (Figs. 6a, S8a, c, Supplementary Videos 9 and 10). Similar to the results from our five-minute global activation experiments (Fig. 1), we did not observe droplets growing beyond ~100 nm in size, even with longer periods of activation, up to ~15 min. To quantify the nucleation process, we calculated the droplet density (ρ) as a function of time⁴². Upon initiation of liquid-liquid phase separation in classic systems, the droplet density rapidly increases as nucleation occurs, followed by a subsequent decrease in cluster number as coarsening processes occur, such as coalescence or Ostwald ripening^31,42. However, in wild-type tubular mitochondria, the droplet density after ~15 min of activation saturated around 0.15 droplets µm⁻², likely due to the confined nature of mt-opto-condensates in the mitochondrial matrix. We measured the average nucleation rate (J, see the “Methods” section Eq. (1)) by taking the slope at t = 0 to be 0.0005 droplets µm⁻² s⁻¹, which was defined as the number of visible droplets per unit mitochondrial area per unit time. Moreover, when blue light was turned off, mt-opto-condensates dissolved within several minutes ($\tau$ = 940 s) (Figs. 6b, e, S8b,d, Supplementary Video 11), consistent with a reversible phase transition.

Fig. 6. Nucleation and dissolution of mt-opto-condensates.

a Local activation (blue light ON) and observation of nucleation of a single mt-opto-condensate within a HeLa cell (see the “Methods” section). Arrowheads point to the nucleating droplet. Scale bar = 2 µm. b Removal of blue light and observation of dissolution of a single mt-opto-condensate (blue light OFF). The arrowheads point to the dissolving droplet (same cell as in a). Scale bar = 2 µm. c Local activation and observation of nucleation and coarsening of mt-opto-condensates in a bulbous mitochondrion (Drp1^K38A). The first row shows the nucleation process, while the second row shows the coarsening of the nucleated droplets. Scale bar = 2 µm. d Droplet density (ρ) as a function of time for condensate nucleation in wild-type mitochondria (n = 6 cells). The shaded area shows the standard deviation. The dashed line is the fit to Eq. 2 (see the “Methods” section). e ρ as a function of time for condensate dissolution in wild-type mitochondria (n = 7 cells). The shaded area shows the standard deviation. The dashed line is the fit to Eq. (3) (see the “Methods” section). f Quantification of ρ as a function of time for the bulbous mitochondrion in (c).

We observed markedly different nucleation dynamics within bulbous mitochondria. We found that after 20 s of activation, numerous small nucleation clusters formed, which rapidly coarsened by fusing together within a particularly large bulbous mitochondrion, leading to a reduction in droplet density (Fig. 6c, f, Supplementary Video 12). The mt-opto-condensates also dissolved within bulbous mitochondria, albeit with slower kinetics than in tubular mitochondria (Fig. S9). Together, these experiments demonstrate that mitochondrial membranes act as physical barriers that limit the growth and coarsening of the condensates they contain.

Discussion

We developed an optogenetic construct, mt-optoIDR, to induce a phase transition in live mitochondria. Using mt-optoIDR, we formed de novo condensates (mt-opto-condensates) that captured the salient morphological features and dynamic behavior characteristic of endogenous mitochondrial condensates. The dominant negative mutant Drp1^K38A allowed us to isolate the contribution of the mitochondrial membrane architecture. Indeed, we found that mitochondrial membranes restricted the size and diffusion of condensates, preventing their ability to coarsen into prominent droplets.

Endogenous mitochondrial (ribo)nucleoprotein complexes—mt-nucleoids and mtRNA granules—persist as stable droplet-like structures that are known to disassemble only if their core nucleic acids are degraded^19,43,44. One advantage of our optogenetically inducible system is that it allows us to capture both the condensation and dissolution processes of condensates within mitochondria. Phase separation occurs through two mechanisms: nucleation and growth, where small clusters above a critical size begin to increase in size gradually, and spinodal decomposition, during which widespread fluctuations in concentration occur before forming larger domains that eventually lead to a stable, condensed phase^27,42. Indeed, we observed that our mt-optoIDR construct nucleates into small clusters that grow and arrest as ~100 nm sized, liquid-like droplets, consistent with nucleation and growth kinetics. We did not observe signs of spinodal decomposition in live mitochondria, potentially due to the limited expression levels we were able to sample with mt-optoIDR. Nonetheless, upon light-induced nucleation, the mt-opto-condensates were also capable of dissolving when light activation was removed, confirming that biomolecules can undergo reversible phase transitions within live, wild-type mitochondria.

Endogenous mitochondrial condensates are typically small, elongated structures, reaching only a few hundred nanometers^{2,20,37,38,45}. Strikingly, our mt-opto-condensates closely mirrored the reported size range of endogenous structures with correspondingly elongated aspect ratios. Interestingly, these morphologies differ from the condensates formed elsewhere in the cell, which can form micron-sized, spherical droplets^27,31, prompting that the mitochondrial environment may be contributing to the unique morphology of its condensates. In support, the same IDR sequence from the protein FUS previously tested in the nucleus and cytoplasm³¹ was significantly limited in size when activated in mitochondria, as were the induced condensates from the other mitochondrially-derived IDR sequences we tested. These results highlight the significant contributions of the mitochondrial environment to condensate morphology.

Indeed, we find that mt-opto-condensates are aligned and diffuse in the direction of tubular, wild-type mitochondria, supporting that the three-dimensional structure of mitochondrial membranes is a key determinant of condensate behavior. Mitochondria contain an outer membrane and an inner membrane, which is composed of the inner boundary membrane and inward folds called cristae. We found the mt-opto-condensates dynamically fluctuate within pronounced voids surrounded by cristae in tubular mitochondria, which is consistent with the confinement of mt-nucleoids within similar voids^46–48. In contrast, both membranes–including cristae–were restructured in bulbous mitochondria, significantly increasing the available space for mt-opto-condensates to diffuse and coarsen. The contrasting condensate morphology and dynamics within tubular and bulbous mitochondria led us to conclude that wild-type cristae restrict the mobility of the condensates, yet the average directionality is due to the inner boundary membrane and/or outer membrane. The overall organization likely arises from the interplay between the droplet material properties (e.g., surface tension) and membrane mechanics (e.g., geometry, bending energies)^49,50.

Importantly, mitochondrial condensates tend to also be evenly spatially distributed within the mitochondrial network. Indeed, there are multiple mechanisms reported for explaining how mt-nucleoids achieve their uniform spatial distributions, putatively arising from active processes during mitochondrial fission at ER contact sites^51–53, spontaneous pearling behavior⁵⁴, and/or cristae structure^46–48,55. Here, we find that nucleated mt-opto-condensates have limited diffusivities, preventing their growth beyond ~100 nm in diameter in tubular mitochondria. Thus, condensates appear to be largely confined in wild-type mitochondria, effectively undergoing 1D sub-diffusion along the axial direction. The resulting mt-condensate confinement has significant implications for function. For example, restricted mobility of endogenous mt-condensates would result in subsequent localization of transcription and translation: mitochondrially encoded OXPHOS subunits would likely be in close proximity to their mt-nucleoids and mtRNA granules. Given the heteroplasmic nature of mt-nucleoids^56,57, we predict this would contribute to the heterogeneous distribution of OXPHOS subunits throughout the mitochondrial network, leading to local variations in mitochondrial function within the cell. This is of particular importance in neuronal cells with long axonal and dendritic extensions^58,59.

Mitochondrial membrane architecture is highly regulated and cell-type specific. Many perturbations, including disruption of the mitochondrial organizing complex⁶⁰, fusion/fission machinery²³, or changes to membrane potential⁶¹, lead to significantly altered membrane structure. Moreover, particularly under pathological conditions, mitochondria can also lose their canonical tubular nature entirely, taking on more bulbous-like morphologies, similar to what we reported here. Such bulbous mitochondria have been associated with large clusters of mt-nucleoids and mtRNA granules^3,4,20. We showed that the clustering phenotype of our mt-optoIDR construct arises due to coarsening of condensates in membrane-disrupted mitochondria. Our synthetic condensates thus appear to recapitulate the morphological changes of endogenous mt-condensates, suggesting these features need not be actively regulated by the cell, but can arise spontaneously due to the biophysical nature of mitochondria. It is intriguing to speculate how other forms of disruption to mitochondrial membranes would influence the degree of coarsening—and resulting function—of endogenous mt-condensates.

Together, our results provide direct evidence that mitochondrial membranes act as physical barriers restricting the free diffusion of mt-condensates, and in this way, prevent their coarsening. Recent work shows that condensates are known to interact with membranes in other cellular contexts^{49,50,62–66}. For example, ER membranes influence the assembly and disassembly of associating condensates⁶², membrane-condensate wetting regulates autophagosome formation⁶⁷, and membranes nucleate signaling components to form productive signaling centers^68,69. Our findings show that membranes can also influence droplet morphology by sterically preventing their diffusion and growth. Overall, we find that the kinetics of phase transitions within mitochondria are limited by membrane architecture.

Methods

Bioinformatic analysis

To find an intrinsically disordered region (IDR) endogenous to the mitochondrial proteome, we screened the MitoCarta 3.0 database⁷⁰. We searched sequences from proteins that are known to localize to endogenous mt-condensates (i.e., mt-nucleoids or mtRNA granules) that contain an IDR region on the N-terminus immediately succeeding a mitochondrial targeting sequence (MTS). To identify IDR sequence propensity, D²P² disorder predictions³⁴ and AlphaFold structures³⁵ were used. The presence of the mitochondrial targeting sequence was validated with predictions from Mitofate⁷¹ and TargetP 2.0⁷². After the initial screening, as shown in Fig. S1, four sequences were identified and were cloned (see below). We found the best expression of the construct containing the MTS (1–23 aa) and IDR region (24–127 aa) of the mitochondrial dead box helicase 28 (DDX28, 1–127 amino acids), which we used in our prototypic mt-optoIDR construct. We confirmed the disordered nature of the identified IDR region (24–127 aa) using ColabFold (v1.5.5)⁷³. Inputs that deviated from default values include: num_relax = 5, model_type = alphafold2, and relax_max_iterations = 0. Protein structure was viewed using Mol* Viewer⁷⁴.

Construct design and cloning

Mitochondrial targeting sequence of DDX28 (MTS) (1–23 aa), and intrinsically disordered regions (IDR) of DDX28 (24–127 aa), and FUS (1–214 aa) were isolated from total RNA (wild-type primary human fibroblast, HGFDFN168, Progeria Research Foundation³) through reverse transcription first-strand cDNA synthesis. The manufacturer’s protocol from Thermo Scientific RevertAid Reverse Transcriptase (EP0441) was followed using Oligo(dT)₁₈ as a primer for the first-strand cDNA synthesis, RNase inhibitor (Invitrogen), and dNTP mixture (TaKaRa). From the first strand cDNA library, gene-specific primers were used to perform PCR to amplify the MTS-DDX28(IDR) fragment (forward primer: 5’-GAAACATGGCTCTAACGCGG-3’, reverse primer: 5’-GTGCTAGACTGCACGGTTGT-3’), and FUS (forward primer: 5’- AACTTCGTTGCTTGCTTGCC-3’, reverse primer: 5’- TGCTTGAAGTAATCAGCCACAG-3’) using the Primer-BLAST web server⁷⁵. Primers used for fragment amplification are reported in Supplementary Table 2. TWINKLE MTS and IDR (1–170 aa) were amplified using Addgene plasmid 129705⁷⁶, and GRSF1 MTS and IDR (1–150 aa) were amplified using GenScript Clone ID OHu107268. CRY2olig (Addgene plasmid 60032)³³ was cloned into the C-terminus of the mCherry in pmCherry-N1 (TaKaRa, 632523) (mCherry-CRY2olig). Then, each IDR was cloned to the N-terminus of plasmid mCherry-CRY2olig. A linker was placed in between each fragment of IDR, mCherry, and CRY2olig (Supplementary Table 1). For mitochondrial targeting, MTS from the DDX28 (1–28 aa) was used in the FUS construct. Fragment assembly was performed using the NEBuilder HiFi DNA Assembly kit. ΔCRY2olig construct was made using a site-directed mutagenesis kit (NEB E0552S) with primers (forward primer: 5’-TAACGGCCGCGACTCTAGATCATAATCAGC-3’ and reverse primer: 5’-CTTGTACAGCTCGTCCATGCCGCC-3’). ΔIDR construct was made by combining MTS from DDX28 (1–28 aa) to the N-terminus of the mCherry-CRY2olig vector. All plasmid sequences were verified with Whole Plasmid Sequencing. Briefly, MTS-DDX28(IDR)-mCh-CRY2olig, MTS-TWINKLE(IDR)-mCh-CRY2olig, and MTS-GRSF1(IDR)-mCh-CRY2olig were sequenced by Plasmidsaurus. MTS-FUS(IDR)-mCh-CRY2olig, MTS-mCh-CRY2olig, and MTS-DDX28(IDR)-mCh were sequenced by Oxford Nanopore long-read sequencing. Specifically, library preparation was done with a Rapid Barcoding Kit 96 V14 (SQK-RBK114.96, Oxford Nanopore Technologies) using an Oxford Nanopore MinION Mk1D with a R10.4.1 (FLO-MIN114) flow cell (Huck Institutes’ Genomics Core Facility, Penn State University, USA).

Mammalian cell culture

HeLa cells (ATCC, CCL-2, Lot #70046455) were cultured at 37 °C and 5% CO₂ and in Dulbecco’s modified Eagle medium (Gibco, 11960069) supplemented with 10% (v/v) fetal bovine serum (FBS) (Gibco, A5256701) and 1% Penicillin/Streptomycin/L-Glutamine (Gibco, 10378-016).

Transient transfection

Cells were first seeded in an 8-well imaging chamber (Nunc Lab-Tek) on Day 0. On Day 1, cells were transiently transfected at 70–90% confluency with Lipofectamine 3000 transfection kit following the manufacturer’s protocol (Invitrogen, L3000-001). Briefly, the Lipofectamine-mix had 10 µL Opti-MEM medium (Gibco, REF 31985-062) and 0.3 µL Lipofectamine 3000 (Invitrogen, 100022049). DNA-mix: 0.2 µg DNA (total amount), 10 µL Opti-MEM and 0.4 µL P3000 reagent (Invitrogen, 100022056). 10 µL of DNA-mix and 10 µL of lipofectamine-mix (1:1 ratio) were mixed and incubated for 10 min at room temperature. Then, 20 µL of the DNA–lipid complex (mixed solution) was added to each well of an eight-well imaging chamber. All volumes are presented as per well (0.7 cm²). Volumes were scaled up according to the number of wells transfected during each imaging session. Cells were imaged after 2 days of transfection (Day 4). All cells were transfected with both mt-optoIDR and Halo-MTS for wild-type cell experiments. Drp1^K38A (Addgene plasmid 45161)⁷⁷ overexpression was also done in combination with mt-optoIDR and Halo-MTS to induce the bulbous mitochondrial phenotype unless otherwise specified.

For different mt-optoIDRs and cristae labeling experiments, FuGene HD (Promega) was used to transfect mt-optoIDR plasmids. Briefly, cells were seeded in 12-well plates with cover slips for whole plate activation or four-well chambers (Nunc Lab-Tek) for live imaging, and mt-optoIDR plasmids or mt-optoIDR (DDX28(IDR)) and Drp1^K38A plasmid were transfected according to the manufacturer’s protocol. Cells were imaged after 2 days of transfection.

Labeling

Cells that were expressing Halo-MTS (Addgene plasmid 124315)³⁶ were labeled with abberior LIVE SiR HaloX (LVSIR-0146-10NMOL, Lot 30707LH-1). Before imaging, the cell culture media were removed, and cells were washed with the prewarmed (37 °C) live-cell, complete, phenol-red-free imaging medium (Gibco 31053036). After that, 200 µL of 0.1 µM staining solution (prepared using 0.1 mM stock solution in DMSO) of HaloX SiR in live-cell imaging media was added and incubated for 30 min at 37 °C, 5% CO₂ before imaging. After incubation, cells were imaged directly without any washing steps.

To visualize cristae, the cell culture media was first removed, and cells were washed with the prewarmed (37 °C) live-cell, complete, phenol-red-free imaging medium (Gibco 31053036). Then Live Red mito (Abberior LVRED-0146-30NMOL) staining solution prepared in phenol-red-free imaging medium (20 nM for WT live, 100 nM for bulbous live) was added to wells (500 µL per well for four-well chamber), then incubated for 1 h at 37 °C, 5% CO₂. Then, the cells were washed 3× with 15 min incubation for the final wash and subsequently imaged.

Live cell imaging

All live cell imaging was performed using a laser scanning confocal microscope equipped with an Airyscan 2.0 detector (Zeiss LSM 980 series). A Plan-Apochromat 63×/1.40 oil DIC M27 oil immersion objective was used to collect images. The microscope was equipped with a built-in incubation chamber maintaining conditions of 37 °C, 5% CO₂, and humidity. Raw images were processed through the Airyscan processing function (Zen Blue software) to achieve high-resolution images, which were used for all subsequent image analysis.

Single-cell global light activation

Once a single cell was identified, global light activation of CRY2olig was first performed on the entire field of view with 488 nm laser (1%) every 5 s (total 5 min) in confocal mode using the 63× oil objective at 4× zoom with an image size for the activation was 665 × 665 pixels (32.5 µm × 32.5 µm) and pixel dwell time: 4.1 µs. After the activation, Airyscan high-resolution modes were used to image droplets formed in an activated cell, using 561 nm laser (0.5%) and 639 nm laser (0.2%) every 2 s for dynamics analysis (total ~6.5 min).

Single-cell local light activation

Local light activation of CRY2olig was performed with bleaching settings in Airyscan line switch mode. 63× oil objective was used at 4× zoom. Bleaching was performed throughout the time-lapse, every 5 s after each image acquisition with a 488 nm laser (1%). Spot size diameter of the region of activation (ROA) = 1 µm circle, scan speed = 1. Image acquisition was done with 561 nm laser (0.5%) and  639 nm laser (0.2%) for a total of ~16.5 min. Light scattering led to diffuse activation surrounding the ROA.

For imaging mt-opto-condensates and cristae, first local activation was performed using the above-mentioned settings. To minimize Live Red mito photobleaching during the activation, only the mt-opto-condensate construct was imaged with 561 nm (0.5%). After droplet formation, ROA was imaged at 10× zoom with a pixel size of 20 nm and with an imaging interval of 2 s. 561 nm (0.5%) and 639 nm 1% were used to image mt-opto-condensates and cristae, respectively. Airyscan processing was done on acquired images.

Whole plate light activation

First, cells were seeded on coverslips and transfected in 12-well plates. 12-well plate activation was done using an LED Array system composed of Blue light Array (AMUZA LEDA-x) and LAD-1 LED Array Driver (AMUZA LAD-1). Before activation, cells were washed with the prewarmed (37 °C) CO₂- independent media (Gibco 18045088), and 500 µL of CO₂-independent media was added to each well. Then the whole 12-well plate was placed inside the pre-warmed incubator on top of the LED array at 5 cm height allowed to equilibrate at 37 °C for ~10 min. Next, the plate was activated with 470 nm for 30 min at constant light activation mode (7.5 V). After the activation, cells were immediately fixed in 4% PFA (Electron Microscopy Science Cat.15710) for 10 min at room temperature. Then cells were washed with PBS and stored in PBS. For imaging, slides were cured with Prolong Gold (Invitrogen P36930) at least 24 h prior to imaging.

Immunofluorescence imaging

Cells were fixed in 4% paraformaldehyde (PFA) (Electron Microscopy Sciences, 15710) by diluting 16% PFA 1:1 in phosphate-buffered saline (PBS) (MP-Biomedicals, 1860454) (8%) and 1:1 (final 4% PFA) directly in cell culture media for 10 min. Then, cells were washed with PBS before permeabilization with 0.5% Triton X-100 (Sigma-Aldrich, T8787) in PBS (v/v) for 10 min at room temperature (RT). Then the cells were washed with 0.05% Tween 20 (Sigma-Aldrich, P9416) in PBS (v/v) three times 2 min each. Blocking was done with 2% bovine serum albumin (BSA, Sigma-Aldrich A3294) in 0.05% Tween (w/v) for 30 min at RT. Blocking agent was removed, and specific primary antibodies (diluted in 5% BSA in 0.05% Tween (w/v)) (Supplementary Table 3) were incubated for one hour at RT in the dark, followed by three wash steps with 0.05% Tween in PBS, each for two minutes. Then, secondary antibodies diluted in the same buffer as primary antibodies were incubated for 1 h at RT in the dark, followed by three wash steps with 0.05% Tween in PBS for two min each and post fixation in 4% PFA for 10 min. Finally, cells were washed with PBS two times and stored in fresh PBS for imaging.

Image analysis

Droplet characterization

For quantitative image analysis, Python was used with in-house custom-built scripts⁷⁸. Droplets were located with trackpy (version 0.6.4)⁷⁹ package and segmented for further analysis. Characterization of isolated droplets was done by fitting the 2D (XY) intensity distribution to a bivariate Gaussian. Parameters such as size (full width at half max, FWHM) and orientation of the droplet were extracted from the fit. All the analysis scripts used are available through the GitHub webpage (https://github.com/fericlab/mt-optoIDR). For visualization purposes, Fiji (version 2.16.0/1.54p)⁸⁰ and Blender microscopy nodes (version 2.2.7)⁸¹ were used.

Droplet orientation analysis

To determine the orientation of the mitochondrial network, the network was binarized using an adaptive local thresholding. Then the network was skeletonized, and the droplet overlapping region of the network was used to estimate the mitochondrial orientation. A linear fit was performed on the skeletonized mitochondria to estimate the orientation. Then, the orientation of the mitochondria and the orientation of the droplet were used to calculate the angle theta ($\theta$).

MSD analysis

Droplet positions were tracked with trackpy in 2D (XY). To isolate independent motion in X and Y, each droplet track was first aligned to be parallel to the X-axis. A linear fit was performed on each track to define the major direction of motion. The angle between the X-axis and the fit was calculated, allowing us to apply a rotational matrix, which would align the track to the X-axis. With the aligned tracks, the mean squared displacement (MSD) was calculated with overlapping lag times, independently for X and Y dimensions. Only the first 20% of the MSD data was used from each track. To obtain the average MSD, squared displacements were logarithmically binned based on lag times $\left(\tau \right)$. The error for each data point is reported as the standard error of the mean (SEM). The diffusive exponent was obtained from linear fitting of the logarithmic data (log(MSD) versus log($\tau$)). Error of the diffusive exponent is presented as 95% confidence intervals of the slope from the linear fit.

Displacement probabilities

Displacement probabilities were calculated independently for X and Y dimensions on X-aligned tracks.

Nucleation and dissolution

Local activation was used to visualize nucleation of droplets live. For the nucleation and dissolution analysis, 400-pixels by 400-pixel image was isolated, centered on the activation region. Droplets were located with trackpy, and the mitochondrial network area was calculated from the background-subtracted image. Nucleation rate, J, was calculated at t = 0 s, after fitting to Eq.  (2)⁴²:

$$J=\frac{\mathrm{d}\rho }{\mathrm{d}t}$$ 1

$$\rho \left(t\right)={\rho }_0\left(1-{\mathrm{e}}^{-\frac{t-t_0}{\tau }}\right)$$ 2

where ρ is the droplet density, t is time, ρ₀ is the initial droplet density, t_o is the time off-set and τ is the characteristic time. For dissolution, the data were fit to Eq. (3) to obtain the characteristic time τ:

$$\rho \left(t\right)=A{\mathrm{e}}^{\frac{-t}{\tau }}$$ 3

where A is a constant.

Statistics and reproducibility

Each experiment had at least three technical replicates (n ≥ 3 cells) with each having multiple mt-opto-condensates, unless otherwise stated. The exact sample sizes in each experiment are reported in the figure legends. Each experiment in the main figures was repeated independently three times on different days. For expression levels and droplet size comparison, the non-parametric, two-tailed Mann–Whitney U test was used since the data do not follow a Gaussian distribution. The corresponding p-values were calculated from the Mann–Whitney U test and indicated in the figure legends. For experiments with different mt-optoIDRs, major axis size was compared with the Kruskal–Wallis H-test, and minor axis size and partition coefficient were compared with the ANOVA one-way test. To assess the extent of similarity of size distributions, the overlap coefficient for each pair of IDRs was calculated for both the major and minor axes.

Reporting summary

The reporting summary is provided in the Nature Reporting Summary linked to this article.

Author contributions

M.F. and S.A.G. designed the study, discussed the results, and wrote the paper. S.A.G. performed experiments and analyzed data. S.T.P. and S.A.G. wrote code. All authors reviewed the manuscript.

Peer review

Peer review information

Communications Biology thanks Nico Marx and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: Chao Zhou and David Favero.

Data availability

Source data and code used to generate figures in this paper are available through GitHub (https://github.com/fericlab/mt-optoIDR) and FigShare⁷⁸. Source data associated with Figs. 1–6 are provided in Supplementary Data 1, and source data related to Figs. S2–S8 are provided in Supplementary Data 2. Plasmids generated in this study have been deposited in Addgene under accession numbers: MTS-DDX28(IDR)-mCh-CRY2olig (Addgene 241146), MTS-FUS(IDR)-mCh-CRY2olig (Addgene 254612), MTS-TWINKLE(IDR)-mCh-CRY2olig (Addgene 254613), MTS-GRSF1(IDR)-mCh-CRY2olig (Addgene 254614), MTS-mCh-CRY2olig (Addgene 254615), and MTS-DDX28(IDR)-mCh (Addgene 254616). Whole Plasmid Sequencing data were deposited in the NCBI Sequence Read Archive (SRA) under BioProject accession number PRJNA1446749. All data supporting the findings of this study are available from the corresponding author upon reasonable request.

Code availability

All the analysis scripts used in this study are available through the GitHub webpage at https://github.com/fericlab/mt-optoIDR and FigShare⁷⁸.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-026-10085-3.