Coordination of RNA and protein condensation by the P granule protein MEG-3

Germ granules are RNA-protein condensates in germ cells. The mechanisms that drive germ granule assembly are not fully understood. MEG-3 is an intrinsically-disordered protein required for germ (P) granule assembly in C. elegans. MEG-3 forms gel-like condensates on liquid condensates assembled by PGL proteins. MEG-3 is related to the GCNA family and contains an N-terminal disordered region (IDR) and a predicted ordered C-terminus featuring an HMG-like motif (HMGL). Using in vitro and in vivo experiments, we find the MEG-3 C-terminus is necessary and sufficient to build MEG-3/PGL co-condensates independent of RNA. The HMGL domain is required for high affinity MEG-3/PGL binding in vitro and for assembly of MEG-3/PGL co-condensates in vivo. The MEG-3 IDR binds RNA in vitro and is required but not sufficient to recruit RNA to P granules. Our findings suggest that P granule assembly depends in part on protein-protein interactions that drive condensation independent of RNA.


INTRODUCTION 24
In animals with germ plasm, specification of the germline depends on the 25 segregation of maternal RNAs and proteins (germline determinants) to the 26 primordial germ cells. Germline determinants assemble in germ granules, micron-27 sized dense assemblies that concentrate RNA and RNA-binding proteins (Jamieson-28 Lucy and Mullins, 2019; Marnik and Updike, 2019; Seydoux, 2018; Trcek and 29 Lehmann, 2019). Superficially, germ granules resemble RNA-rich condensates that form in the cytoplasm of somatic cells, including P bodies and stress granules. In 31 recent years, much progress has been made in our understanding of stress granule 32 assembly with the realization that stress granules resemble liquid condensates that 33 assemble by liquid-liquid phase separation (LLPS). LLPS is a thermodynamic 34 process that causes interacting molecules to dynamically partition between a dense 35 condensed phase and a more dilute phase (e.g. the cytoplasm) ( In this study, we examine the assembly of P granules, germ granule in C. 59 elegans. At the core of P granules are liquid condensates assembled by PGL proteins. Magenta bars indicate residues mutated to alanine in MEG-3 HMGL-.

MEG-3 698
supplement 1A). We conclude that both the IDR and C-terminus contribute to MEG-123 3 condensation in vitro. 124 125 Co-assembly of MEG-3/PGL-3 condensates in vitro is driven by the MEG-3 C-126 terminus and does not require RNA or the MEG-3 IDR 127 When combined in condensation assays, MEG-3 and PGL-3 form co-128 condensates that resemble the architecture of P granules in vivo, with the smaller 129 MEG-3 condensates (~100 nm) forming a dense layer on the surface of the larger 130 PGL-3 condensates (Putnam et al., 2019). MEG-3Cterm and MEG-3HMGL-formed co-131 condensates with PGL-3 that were indistinguishable from those formed by wild-type 132 MEG-3( Figure 2C). The MEG-3IDR, in contrast, failed to assemble condensates on the 133 surface of PGL-3, or away from PGL-3, and instead mixed homogenously with the 134 PGL-3 phase as previously reported (Putnam et al. 2019, Figure 2C). 135 We repeated the co-condensation assays in the absence of RNA using a 136 higher concentration of PGL to force PGL condensation in the absence of RNA. 137 MEG/PGL co-condensates assembled under those conditions were indistinguishable 138 from co-condensates assembled in the presence of RNA ( Figure 2C). Again, the C-139 terminus was necessary and sufficient for co-assembly. MEG-3IDR homogenously 140 mixed with the PGL-3 phase and did not form independent condensates, confirming 141 that MEG-3IDR is solubilized by PGL-3. We conclude that the MEG-3 C-terminus is the 142 primary driver of MEG-3 condensation and that condensation of MEG-3 on PGL 143 condensates depends on the MEG-3 C-terminus and does not require RNA in vitro. 144 145

The MEG-3 IDR is necessary and sufficient for RNA binding in vitro 146
Using fluorescence polarization and gel shift assays, we previously showed 147 that the MEG-3 IDR binds an RNA oligo (poly-U30) with near nanomolar affinity in 148 vitro (Smith et al., 2016). We repeated these observations using a filter binding 149 assay where proteins are immobilized on a filter to minimize possible interference 150 due to condensation of MEG-3 in solution (Methods, Figure 2D-G). Consistent with 151 previous observations (Smith et al., 2016), we found that the MEG-3IDR exhibits high 152 affinity for RNA (Kd=105 nM; Figure 2F). MEG-3698 also exhibited high affinity  Figure 2D), albeit at a lower affinity than MEG-3IDR and MEG-3698. In 155 contrast, MEG-3Cterm exhibits negligible RNA binding ( Figure 2E). HMG domains are 156 common in DNA-binding proteins and have been shown to mediate protein:nucleic 157 acid interactions in vivo (Genzor and Bortvin, 2015;Reeves, 2001;Thapar, 2015), 158 raising the possibility that the HMG-like domain in MEG-3 might contribute to RNA 159 binding. We found, however, that the MEG-3HMGL-bound to RNA with high affinity, 160 similar to wild-type MEG-3 ( Figure 2G). We conclude that the HMG-like domain does 161 not contribute to RNA binding, which is driven primarily by the IDR.  whereas MEG-3IDR and MEG-3HMGL-enriched as efficiently as wild-type ( Figure 3C). 210 Finally, none of MEG-3 derivatives enriched in condensates as efficiently as wild-211 type ( Figure 3D). 212 After the four-cell stage, the low levels of wild-type MEG-3 and MEG-3HMGL-213 inherited by somatic blastomeres were rapidly cleared. In contrast, MEG-3IDR and accumulate to similar levels, whereas MEG-3IDR and MEG-3Cterm were more abundant 217 in mixed-stage embryo lysates, consistent with slower turnover in somatic lineages 218 ( Figure 3 -figure supplement 1C). 219 The condensation, segregation, and turnover patterns of MEG-3, MEG-3IDR, 220 MEG-3Cterm MEG-3HMGL-are summarized in Figure 3E. From this analysis, we 221 conclude that: 1) the MEG-3 IDR is necessary and sufficient for enrichment of 222 cytoplasmic MEG-3 in germ plasm, 2) the MEG-3 C-terminus is necessary and 223 sufficient for condensation of MEG-3 in germ plasm starting in the zygote stage, 3) 224 the HMG-like motif is required for efficient MEG-3 condensation, and 4) both the C-225 terminus and the IDR are required for timely turn-over of MEG-3 in somatic 226 lineages. 227 228  Figure 4B). We found that PGL-3 condensates 255 co-localized with MEG-3Cterm condensates (37/37 PGL-3 condensates scored in P1; 256 Figure 4B) as in wild-type. In contrast, we observed no such colocalization with 257 MEG-3IDR or MEG-3HMGL-. The MEG-3IDR is mostly cytoplasmic and forms only rare 258 condensates in P2. We occasionally observed PGL condensates with an adjacent 259 MEG-3IDR condensate (5/19 PGL-3 condensates scored in P1, Figure 4B), but these 260

Co-assembly of MEG-3/PGL-3 condensates in vivo is driven by the MEG-3 C-229 terminus and requires the HMGL motif
were not co-localized. Unlike the MEG-3IDR, MEG-3HMGL-forms many condensates in 261 P2, although these tended to be smaller than wild-type ( Figure 3A Figure 5A, D). 305 Enrichment of mRNAs in P granules can also be detected using an oligo-dT 306 probe to detect polyadenylated mRNAs (Seydoux and Fire, 1994). In wild-type 28-307 cell stage embryos, strong poly-A signal is detected around the nucleus of the P4 blastomere ( Figure 5 -figure supplement 1C). This perinuclear signal was absent in 309 meg-3 meg-4 mutants as well as in embryos expressing MEG-3IDR, MEG-3Cterm and 310 MEG-3HMGL-. The lack of polyA signal was particularly striking in the case of MEG-311 3IDR and MEG-3HMGL-since those variants assemble robust perinuclear condensates 312 at this stage ( Figure 3A). terminus was more efficient at condensation than the MEG-3 IDR in vivo. This 345 difference could also be observed in vitro when MEG-3 condensation was assayed in 346 the presence of PGL-3, which solubilizes the MEG-3 IDR ( Figure 2C). When MEG-3 347 was assayed alone, however, a difference between the MEG-3 C-terminus and IDR 348 condensation could only be seen at the lowest concentration tested (Figure 2A conditions (SDS and high salt) than the condensation assays and may therefore be 355 better suited to reveal affinity differences sufficient to disrupt protein interactions 356 in the crowded cellular milieu. We conclude that, while in vitro experiments are 357 excellent tools to reveal self-assembly principles, condensation assays can lead to 358 conclusions (e.g. HMGL is dispensable for MEG-3/PGL co-assembly) that do not 359 necessarily hold in vivo. 360 361

MEG and PGL proteins and does not require RNA 363
The MEG-3 C-terminus is a 318 aa sequence with regions of predicted low 364 disorder including an HMG-like motif. We have found that the MEG-3 C-terminus is 365 sufficient to form condensates that dock on PGL droplets in vitro and in vivo. the future will be to understand the mechanisms that regulate the assembly and 449 disassembly of protein scaffolds at the core of germ granules.

452
Worm handling, maternal-effect sterility counts 453 C. elegans was cultured at 20˚ C according to standard methods (Brenner, 1974). To 454 measure maternal-effect sterility, ten gravid adults were picked to an OP50 plate 455 and allowed to lay eggs for ~2 hours, then removed. Adult progeny were scored for 456 empty uteri (white sterile phenotype) under a dissecting microscope. 457 458 Identification of MEG-3 HMG-like region 459 MEG-3 and MEG-4 protein sequences were aligned with HMG boxes from GCNA 460 proteins of Caenorhabditis and example vertebrates along with the canonical HMG 461 box of mouse SOX3 using MUSCLE (Edgar, 2004 Genome editing was performed in C. elegans using CRISPR/Cas9 as described in 467 (Paix et al., 2017). Strains used in this study along with guides and repair templates 468 are listed in Supplementary Table 1. Some strains were generated in two steps. For 469 example, MEG-3HMGL-was generated by deleting the entire HMGL-like motif in a first 470 step (JH3632), and inserting a modified HMG-like motif with the desired mutations 471 in a second step (JH3861). Genome alterations were confirmed by Sanger 472 sequencing and expression of tagged strains was verified by immunostaining and 473 western blotting (Sup. Figure 3B). 474 475 Statistical Analysis and plotting 476 On all scatterplots, central bars indicate the mean and error bars indicate one 477 standard deviation. Unless otherwise indicated, differences within three or more 478 groups were evaluated using a one-factor ANOVA and differences between two 479 groups using an unpaired Student's t-test. 480 481 Confocal Imaging 482 Fluorescence confocal microscopy for figure 2C Quantification of in situ hybridization images 537 All measurements were performed on a single confocal slice centered on the P cell 538 nucleus in ImageJ. For early embryos where there is distinct punctate signal (1 and 539 4 cell stage Figure 5B,C, Figure 5 -figure 1 B), a region of interest was drawn, the 540 Analyze Particles feature was used with a manual threshold to identify and measure 541 the integrated density of the puncta. The raw integrated density for all particles in 542 the region of interest was summed to give the total intensity of the mRNA in that 543 region. For 28 cell embryos (Fig, 5D)  His-tagged protein expression, purification and labeling 565 Expression and purification of MEG-3 His-tagged fusion proteins: MEG-3 full-length 566 (aa1-862), IDR (aa1-544), Cterm (aa545-862), and HMGL-proteins were fused to an 567 N-terminal 6XHis tag in pET28a and expressed and purified from inclusion bodies 568 using a denaturing protocol (Lee et al., 2020) 569 570 Purification of MBP-TEV-PGL-3 was expressed and purified as described (Putnam et  571 al., 2019) with the following modifications: MBP was cleaved using homemade TEV 572 protease instead of commercial. A plasmid expressing 8X-His-TEV-8X-Arg tag 573 protease was obtained from Addgene and purified according to the published 574 protocol (Tropea et al., 2009). Before loading cleaved PGL-3 protein on to a heparin 575 affinity matrix, cleaved MBP-6X-His and 6X-His-TEV protease were removed using a 576 HisTRAP column (GE Healthcare). 577 578 Protein labeling: Proteins were labeled with succinimidyl ester reactive 579 fluorophores from Molecular Probes (Alexa Fluor™ 647 or DyLight™ 488 NHS Ester) 580 following manufacturer instructions. Free fluorophore was eliminated by passage 581 through three Zeba™ Spin Desalting Columns (7K MWCO, 0.5 mL) into protein 582 storage buffer. The concentration of fluorophore-labeled protein was determined 583 using fluorophore extinction coefficients measured on a Nanodrop ND-1000 584 spectrophotometer. Labeling reactions resulted in ~ 0.25-1 label per protein.

585
Aliquots were snap frozen and stored. In phase separation experiments, 586 fluorophore-labeled protein was mixed with unlabeled protein for final reaction 587 concentrations of 25-100 nM of fluorophore labeled protein. 588 589 590 In vitro transcription and labeling of RNA 591 592 mRNAs were transcribed using T7 mMessageMachine (Thermofisher) using 593 manufacturer's recommendation as described (Lee et al., 2020). Template DNA for 594 transcription reactions was obtained by PCR amplification from plasmids. Free 595 NTPs and protein were removed by lithium chloride precipitation. RNAs were 596 resuspended in water and stored at −20°C. The integrity of RNA products was 597 verified by agarose gel electrophoresis. 598 599 600 In objective (Figure 2A). MEG-3 and PGL-3 co-condensate were imaged using thin 611 chambered glass slides (Erie Scientific Company 30-2066A) with a coverslip ( Figure  612 2C). Images are single planes acquired using a 40x oil objective over an area 613 spanning 171 x 171 μm. 614 615 To quantify the relative intensity of MEG-3 in condensates, a mask was created by 616 thresholding images, filtering out objects of less than 4 pixels to minimize noise, 617 applying a watershed filter to improve separation of objects close in proximity, and 618 converting to a binary image by the Otsu method using the nucleus counter 619 cookbook plugin. Minimum thresholds were set to the mean intensity of the 620 background signal of the image plus 1-2 standard deviations. The maximum 621 threshold was calculated by adding 3-4 times the standard deviation of the 622 background. Using generated masks, the integrated intensity within each object was 623 calculated. To remove non-specific background signal the mean intensity of an 624 image field in the absence of the labeled component was subtracted from each pixel 625 yielding the total intensity of each object.