Mechanisms for Active Regulation of Biomolecul.pdf

OpinionMechanisms for Active Regulation ofBiomolecular CondensatesJohannes So ding,1,*David Zwicker,2Salma Sohrabi-Jahromi,1Marc Boehning,3and Jan Kirschbaum2Liquidliquid phase separation is a key organizational principle in eukaryotic cells,on par withintracellular membranes.It allows cells to concentrate specific proteins into condensates,increasing reaction rates and achieving switch-like regulation.We propose two active mecha-nisms that can explain how cells regulate condensate formation and size.In both,the cell regu-lates the activity of an enzyme,often a kinase,that adds post-translational modifications tocondensate proteins.In enrichment inhibition,the enzyme enriches in the condensate andweakens interactions,as seen in stress granules(SGs),Cajal bodies,and P granules.In localiza-tion-induction,condensates form around immobilized enzymes that strengthen interactions,as observed in DNA repair,transmembrane signaling,and microtubule assembly.These modelscan guide studies into the many emerging roles of biomolecular condensates.Biomolecular Condensates Can Be Formed and Dissolved in the Blink of anEnzymeEukaryotic cells possess numerous types of membraneless organelles(see Glossary).Each containsbetween tens and several thousands of protein and RNA species that are highly enriched comparedwith the surrounding nucleoplasm or cytoplasm.These biomolecular condensates are held togetherby weak,multivalent,and highly collaborative interactions,often between intrinsically disordered re-gions of their constituent proteins 1,2.In contrast to membrane-bound organelles,biomolecular condensates can easily be formed or dis-solved by merely changing the activity of an enzyme,such as a kinase,that post-translationally mod-ifies key condensate proteins 35,71.The modifications usually lie in intrinsically disordered regionsand modulate the strength of attractive interactions with other condensate components 6,7.Due tothe highly cooperative nature of phase transitions,small changes in interaction strengths can result inthe formation or dissolution of condensates,and this switch-like nature makes them ideal for dynamicregulation.Forinstance,SGsformoncellularstressandaredissolvedwhenthestress ceases3.Also,P-bodiesinthe cytoplasm and Cajal bodies,nuclear speckles,paraspeckles,and PML bodies in the nucleus havetobedissolvedduringmitosisandreformedafterwardstoensureabalanceddistributionoftheircon-tents to daughter cells 4,8.These droplet organelles are large enough to be visible using simple light microscopy techniquesand have long been known.Recently,liquidliquid phase separation has been implicated in multifar-iousprocesses in which often submicrometer-sized condensatesare formed at particular locationsin the cell:at sites of DNA repair 9,Polycomb-mediated chromatin silencing 10,transmembranesignaling 11,12,microtubule formation 1315,actin polymerization 16,endocytosis 17,18,pre-synaptic active zones 19,20,and ribonucleoprotein(RNP)transport 2123.Such localized conden-sates form on a local stimulus to recruit the required set of proteins and are dissolved once their job isdone.Cells do not only need to regulate the formation and dissolution of each type of condensate.Theysometimes also need to regulate their size and with it their numbers,to allow many condensatesto form in different locations;for instance,to activate many genes at the same time 2427.Here,we propose two active mechanisms used by cells for these purposes.1Quantitative Biology and Bioinformatics,Max Planck Institute for BiophysicalChemistry,Am Fassberg 11,37077Go ttingen,Germany2Max Planck Institute for Dynamics andSelf-Organization,Am Fassberg 17,37077Go ttingen,GermanyHighlightsBiomolecularcondensatesarephase-separated domains in thenucleoplasm or cytoplasm formedby weak and highly multivalent in-teractions between their proteinand RNA components.They allowcellstocreatecompartmentsstrongly enriched or depleted inspecific proteins or RNAs,withoutthe need for membranes.Cells can actively regulate the for-mation,dissolution,and localiza-tion of condensates by phosphor-ylation and other post-translationalmodificationsofcondensatecomponents.We propose two unifying,genericmechanisms of active regulation.Inthe first,the protein concentrationis above saturation and the modi-fications limit the condensate sizeby inhibiting intracondensate in-teractions.We believe this mecha-nism regulates most membranelessorganelles.In the second mechanism,the pro-tein concentration is below satura-tion,but localized enzymes attachmodificationsthatpromotein-teractions and induce condensateformation in the volume aroundthem.This mechanism can explainhow a single kinase(e.g.,recruitedto a double-strand DNA break)canrecruithundredsofspecificproteins.We argue that many types of con-densates are actively regulated byone of these two mechanisms.Trends in Cell Biology,-,Vol.-,No.-https:/doi.org/10.1016/j.tcb.2019.10.006 2019 Elsevier Ltd.All rights reserved.1Please cite this article in press as:So ding et al.,Mechanisms for Active Regulation of Biomolecular Condensates,Trends in Cell Biology(2019),https:/doi.org/10.1016/j.tcb.2019.10.006Trends in Cell BiologyPhase Separation and Condensate Size BehaviorTo keep the models simple,we consider only one type of condensate protein.In the dilute regimebelow the saturation protein concentration cout,no condensate can form(Figure 1A).Above cout,in the phase separation regime,condensates can be stable.However,in a passive system more than one condensate cannot exist in equilibrium because largercondensates will grow at the expense of smaller ones(Figure 1B).The reason is that proteins on thesurfaceofsmalldropletshavefewerfavorableinteractionsamongthemselvesthanproteinsonthesur-face of larger droplets,due to the difference in surface curvature.They are therefore more easily lost,resulting in a higher equilibrium concentration outside the droplet(section 1.1 in the supplementalinformation online).Due to this size dependence,the protein concentration decreases from smallto large condensates,and the decrease generates a diffusive flux in the direction of steepest descent.Consequently,there exists a critical radius Rcritbelow which condensates will shrink,while conden-sates aboveRcritwill grow(Figure 1C andsection 1.2in the supplemental information online).The crit-ical radius increases until a single large condensate survives,a phenomenon called coarsening 28.The timescale for droplets to change their size by coarsening scales roughly with their radius cubed(derivedinsection1.3inthesupplementalinformationonline).Therefore,smalldropletscangroworshrinkfast,onascale ofminutesfor Rz100 nm,whereasdropletsofamicrometerradiusalreadytakedays.Thisexplains why,in in vitro experiments,droplets of micrometer size and above can coexist for long periods.We now show that,to actively regulate the formation and size of liquid droplet condensates,twogeneric mechanisms exist.A protein concentration maintained above saturation leads to the enrich-ment-inhibition model,in which a regulating enzyme such as a protein kinase inhibits favorable inter-actions and is enriched in condensates.A concentration maintained below saturation leads to thelocalization-induction model,in which the enzyme is localized or attached and induces favorable in-teractions.Although simplified,these models might capture two essential mechanisms for activeregulation of cellular condensates.Both mechanisms modulate interaction strengths of key condensate proteins by an enzyme that addsor removes post-translational modifications.Often,the regulating enzyme will be a kinase that at-taches phosphoryl groups to disordered regions of condensate proteins.However,other post-trans-lational modifications could take this role:poly-ADP-ribosylation in DNA repair 9,SUMOylation(e.g.,in PML bodies 29),arginine demethylation of proteins in RNA granules 30,31,lysine acetyla-tion and methylation 15,32,ubiquitination 33,and RNA modifications 34,35.Even RNA helicaseactivity can take over this role in RNA-containing condensates 36.The EnrichmentInhibition ModelAbove the saturation concentration,a mechanism must exist that limits the size of larger condensatesto allow the coexistence of multiple condensates.This can be achieved if the loss of proteins from thecondensate increases faster with condensate radius R than the gain by net diffusive influx.The influx isproportional to R?Rcrit(Figure 1C).A loss that scales with the condensate volume(4p/3)R3wouldgrow faster than R?Rcrit.Above a certain radius,the loss would surpass the influx,shrinking conden-sates that are too large and thereby resulting in a stable condensate size.We propose the loss mechanism to be the modification of condensate proteins(or RNA)by an antag-onistic regulating enzyme(orange)that is itself enriched in the condensate(Figure 2,Key Figure).Weuse phosphorylation as an example,but the mechanisms work the same for other modifications.Because the concentration of unphosphorylated proteins(blue)is approximately constant in thecondensate,the phosphorylation rate scales with the condensate volume.In this model,unphos-phorylated proteins as well as the kinases attract each other,while the phosphorylation weakensthe interactions with other condensate proteins.It might seem counterintuitive that the droplet-dis-solving kinase enriches in the droplet,yet it is this feature that allows the droplet growth to be self-limiting.3Department of Molecular Biology,MaxPlanck Institute for Biophysical Chemistry,Am Fassberg 11,37077 Go ttingen,Germany*Correspondence:soedingmpibpc.mpg.de2Trends in Cell Biology,-,Vol.-,No.-Please cite this article in press as:So ding et al.,Mechanisms for Active Regulation of Biomolecular Condensates,Trends in Cell Biology(2019),https:/doi.org/10.1016/j.tcb.2019.10.006Trends in Cell BiologySince the concentration of the unphosphorylated proteins is above saturation,the concentration de-creases towards the condensate,leading to a net influx of unphosphorylated proteins(Figure 2A).This influx is compensated by the loss of proteins that become phosphorylated inside the conden-sate,which diffuse out along the negative concentration gradient.Outside,they are dephosphory-lated by phosphatases(green),closing the circle of protein flux.To avoid wasting energy by ashort-circuited phosphorylationdephosphorylation reaction,the phos-phatase and kinase would best be concentrated in different phases.Therefore,we expect the phos-phatase to be strongly depleted in the condensates.For phosphorylation rates k below a certain threshold kthr,all condensates will grow or shrink to thesame stable radius R,which is determined by k(Figure 2B and section 2 in the supplemental informa-tiononline).ThedependenceofRonthephosphorylationratekhasaswitch-likebehavior(Figure2C):above kthr,no condensates can exist.Evidence Supporting Enrichment-InhibitionWe give five examples of biomolecular condensates that behave as expected from the enrichment-inhibition model:(i)their key condensate proteins are phosphorylated by a kinase,(ii)increased ki-nase activity dissolves the condensates,and(iii)the kinase is enriched in the condensates.The modelalso predicts the main phosphatase to be depleted in the condensates.This information appears tobe mostly unavailable.P granules are condensates of RNAs and proteins in the one-cell embryo of the worm Caenorhab-ditis elegans.These localize to the posterior end of the cell and after cell division end up in the onecell that will give rise to the germ line.P granules are highly enriched for the intrinsically disorderedMEG proteins.(i)They are phosphorylated by MBK-2 and dephosphorylated by the PPTR-1 phospha-tase.(ii)Phosphorylation of MEGs promotes granule disassembly and dephosphorylation promotesassembly.Furthermore,(iii)MBK-2 localizes to P granules 37.ThevertebrateorthologofMBK-2,DYRK3,playsacentralroleasdissolvaseofseveraltypesofmembrane-less organelles during mitosis.Rai et al.suggested that,as for P granules,DYRK3 is involved in the sizecontrol ofmanyothertypesofcondensates4,aswewouldexpect fromthe enrichment-inhibitionmodel.SGs are another example.(i)They are regulated by DYRK3 3.However,since DYRK family kinases areconstitutively active,it is unclear how the stress signal could be quickly relayed via DYRK3.Wurtz andLee proposed a plausible mechanism 38:on stress,ATP levels can fall by 50%,within the same timescaleas SG formation.Also,ATP depletion alone is sufficient to induce SG formation.The reduction in DYRK3activity(kinFigure2C)byATPdepletionmightreducethelevelofphosphorylationofitstargets,(i)severalof which arekey SGproteins.(ii)The concomitant increase infavorable interactions thenwould trigger SGformation.(iii)In accord with the enrichment-inhibition model,DYRK3 localizes to SGs 3.Nuclear speckles concentrate proteins involved in pre-mRNA splicing.These factors possess a terminallow-complexity RS region enriched for arginine and serine,which is required for the multivalent interac-tions within the speckles 39.(i)The CLK kinase phosphorylates the RS domains of splicing factors and(ii)phosphorylation by CLK promotes the disassembly of nuclear speckles 40.Finally,(iii)CLK itself pos-sesses an RS domain that is required and sufficient for its enrichment in the speckles 41.Cajal bodies are nuclear condensates defined by the key architectural self-oligomerizing proteincoilin.(i,ii)Hyperphosphorylation of coilin by Cdk2/cyclin E dissolves them.Also,(iii)Cdk2/cyclin isstrongly enriched in Cajal bodies 42.Synaptic vesicles(SVs)containing neurotransmitters form dense clusters at synapses.Synapsin,themajor constituent ofthematrixaroundSVs,formscondensatesunder physiological conditionsinvitro43.The condensatesenrich small lipid vesicles,explaining SV clustering at synapses.As expected,(i)GlossaryCondensate:the protein-richliquid,gel-like,or solid phase inthe liquid cytoplasm or nucleo-plasm.Liquid condensates areusually spherical to minimize theenergetically unfavorable inter-face with the dilute phase.How-ever,their shape can be influ-enced by scaffolding structures asin the case of chromatin.Theircontent is exchanged rapidly withthe surroundings by diffusion.Themean retention time can be on theorder of seconds or below,evenfor proteins highly enriched in thecondensate.Liquidliquid phase separation:inmixtures of two or more compo-nents(e.g.,proteins and watermolecules),it may be energeti-cally favorable for the compo-nents to separate into two liquidphases of different relative con-centrations.For example,if pro-teins attract each other but haveless favorable interactions withwater,they can condense intoliquid droplets with high proteinconcentration in a dilute phase oflow protein concentration.Membraneless organelles:bio-molecular condensates thatorganize cytoplasmic and nucleo-plasmic space,such as nucleoli,nuclear speckles,PML bodies,Cajal bodies,paraspeckles,SGs,and P-bodies.Phase transition:transitions be-tween states of matter deter-min