FXRs are essential for DMV clustering, related to Figs. 2 and 3. (A–H) Compared with NC-treated cells (A), the number of Nsp3/4⁺ puncta is slightly increased in siFXR1, siFXR2, siFMR1, siFXR1/2, or siFXR2/FMR1 cells (B–F). Cells with double KD of FXR1 and FMR1 show an obvious increase in the number of Nsp3/4⁺ puncta (G). Depleting all FXR genes causes dramatic DMV dispersion (H). Bars: 10 μm. (I) Numbers of DMVs are shown as mean ± SD (NC, n = 25; siFXR1, n = 35; siFXR2, n = 34; siFMR1, n = 35; siFXR1/2, n = 32; siFXR2/FMR1, n = 41; siFXR1/FMR1, n = 38; siFXRs, n = 39). *, P < 0.05; ***, P < 0.001; ****, P < 0.0001. (J) Fluorescence images reveal numerous small Nsp3/4⁺ puncta in cells coexpressing GFP-Nsp6, mCherry-Nsp3, and Nsp4-BFP after KD of all FXR genes, while a few big Nsp3/4⁺ puncta are formed in control cells. Images are representative of at least 10 cells. Bars: 5 μm; insets, 2 μm. (K) TEM images show that dispersed DMVs are observed in siFXRs cells coexpressing GFP-Nsp6, mCherry-Nsp3, and Nsp4-BFP, while clustered DMVs are present in NC-treated cells. Bars: 500 nm. (L) In a GFP-Trap assay, FXR1, FXR2, and FMR1 are immunoprecipitated by GFP-Nsp3, but not Nsp6. (M and N) FRAP results show that after photobleaching part of a punctum (arrows), most of the GFP-FXR1 fluorescent signal recovers, while the mCherry-Nsp3 signal fails to recover (M). Quantitative data are presented as mean ± SEM (n = 12) in N. Bars: 2 μm. (O) GFP-FXR1 with the L351P mutation, which disrupts the phase separation ability of FXR1, fails to accumulate at Nsp3/4⁺ sites. Images are representative of at least 10 cells. Bars: 5 μm; insets, 2 μm. (P) Most of the GFP-FXR1 signal is diffuse in cells treated with 1,6-hexanediol. Images are representative of at least 10 cells. Bars: 5 μm; insets, 2 μm. (Q) Nsp3N does not form liquid droplets in vitro. Bar: 5 μm. (R) In a GFP-Trap assay, mCherry-Nsp3 is precipitated by GFP-FXR1, but not GFP-FXR1(L351P). Source data are available for this figure: SourceData FS2.