Title:DropletOnTape:ProtocolAuthors:Franklin Fuller, Sheraz Gul, Ruchira Chatterjee, Jan Kern, Vittal Yachandra &JunkoYanoAffiliation:Yano/YachandraLabAbstract:A sample delivery instrument involving acoustic droplet ejection onto a conveyorbeltforXFELstudies,itsdesignanduse,isdescribedhere.Subjectterms:Spectroscopy,StructuralbiologyKeywords:Sample delivery, XFEL, acoustic droplet ejection, conveyor belt, photosynthesis,Metalloenzymes,X-rayemission,X-raydiffractionIntroduction:For biological studies themain draw of X-ray free electron lasers (XFELs) is thattheirintenseultrashortpulsesallowonetofeasiblyobtaintime-resolvedradiationdamage free structural and electronic information at physiological temperatures.Achievingthesamemeasurementsatsynchrotronsourcesaremuchmoredifficult,as a combination of low light flux and/or cryogenic temperatures are required toavoiddamage for thirdgenerationsources.Amyriadofapproaches forcouplingareaction initiation scheme to a particular sample delivery method exist in theliterature, each with various advantages and disadvantages. Here we describeDropletOnTape (DOT),a flexibleplatformtoapproachavarietyof timeresolvedreactioninitiationschemesonbiomoleculesparticularlywheresimultaneousX-raysignalsaredesired.DOThasbeenemployedforX-raydiffractionandX-rayemissionfrom protein micro-crystal slurry or protein solution as well as a number ofancillarysignals(liketransmissionorfluorescence).Asaplatform,thesystemhasalargenumberofconfigurableandoptionalcomponents,dependingontheneedsofthe user’s experiment. We describe a number of the operational details andmotivationsbehindthedesigndecisionsherethatarenotcoveredintheassociatedpublicationalongwithprotocolsforusingindividualcomponentsofthesystem.DOTiscomprisedoffivemainsub-systems:theADEinjector,controlssystems,theconveyorbelt, reaction initiation systems, and the sample environment.Reagents,materials,software,andprotocolsaregivenwhereappropriateunderthesesystemcomponentheadings.

Nature Methods doi:10.1038/nmeth.4195

Reagents:Thefollowingreagentsareforhydrophobiccoatingofthepolyimidebelts:fluoroalkylsilane (FAS), 1H, 1H, 2H,2H-Perfluorodecyltriethoxysilane(C10F17H4Si(OCH2CH3)3,97%fromSigma-Aldrich)isopropanoln-octaneMilli-QwaterEquipment:Agilent33602FunctionGeneratorAmplifierResearch500A250CRFAmplifierNewportXPS-8QMotorControllerLinux/Mac/PCcontrolcomputerLegato130SyringePumpPolymicro360micronO.D.,180-250micronI.D.100MHzorgreaterdigitalOscilloscopewithremoteaccesscapabilityPrecisionvacuumoven,model19LexiumM-DrivesteppermotorPulsedlasersystem(forphoto-activatedexperiments)Numerous custom machined components (see supplemental technical drawingsattachedtothisprotocol).Procedure:

Figure1:Attheleft:adrivingpulseissenttothetransducer.Theresultingacousticwave(center)isfocusedonthesample-airinterfacecreatingacolumnofsample.Adropletofsampleisejectedinthefinalstep(right)andanechoofthedrivingpulseisreturnedtothetransducer.Thesampleisconnectedtothetransducerviaacolumnofcouplingwater(thatisisolatedfromthesample).

Nature Methods doi:10.1038/nmeth.4195

Our Acoustic Droplet Ejection (ADE) setup is straightforward implementation ofprior work1,2. ADE involves focusing ultrasonic sound energy on a liquid-airinterfacetoejectasmallvolumeofliquidintotheair.Thesoundenergycomesfromanspot-focusedultrasonictransducerdevelopedforultrasonicimagingapplications(Olympus V319-SUwith 1’’ spherical focus). The transducer can eject droplets ofsample when driven by a tone-burst typically less than 100 µs long. We use awaveform generator (Agilent 33612A) to produce tone-burst at the resonancefrequencyofthetransducer(15MHzinmostcases)withcontrollableamplitudeandduration. The tone-burst is amplified in an Amplifier Research 500A250C RFamplifier to produce a final driving voltage of around 40 volts. This amplifier isover-engineered for our application, but was available to us and made theamplification process extremely simple. The amplified tone-burst is sent via BNCcable throughabi-directional coupler,which samples theamplified signal inbothforward and reverse directions with 30 dB coupling. The bulk of the tone-burstpoweristransmittedthroughthebi-directionalcouplertothetransducer.Afewmillisecondspriortotheejectiontoneburstwesenda“probe”spikethatisafewhundrednanosecondslongtodeterminethefluidlevelinthesamplereservoir.The operational principle here is similar to sonar: we are looking to seewhen areflectionofsoundreturnstothetransducer.Wemonitorthis“echo”bycollectingthe reverse signal coming back from the transducer through the bi-directionalcoupler. This probe signal is collected in an oscilloscope.We remotely access theoscilloscope(viaremotedesktop)tomonitorthereturntimeandamplitudeoftheechosignal.Whenthetransducerisfocusedonthesample-airinterface(seeFigure1),anechoshowsupataround35µsdelay(fora1inchfocallengthtransducer).Wejudge the focus quality bymaximizing the amplitude of the return echo and onceoptimized the transducer position is fixed and the sample influx rate into thereservoir is adjusted to maintain temporal position of the echo. The temporalpositionofourprobeechopulsegivestheabilitytomonitorthereservoirfluidlevelwithbetterthan50µmaccuracy.Inourpresentimplementation,thesampleinfluxrate is human controlled, but one can certainly envision amachine control of theflowratebasedontheechoamplitudeandposition.ControlSystemsTherearea fewpiecesofhardware in theDOTsetup that require remote controloperationinordertoensurereasonablesuccesswhenitisdeployedinarestrictedaccess “hutch”:SyringePump,WaveformGenerator,variousactuatormotors, tapedrivemotor,andvariouscameras. Ingeneral,weuseEPICS(ExperimentalPhysicsIndustrialControlSoftware)toprovidenetworkinterfacetoallofthesedevices.InthecaseofoursyringepumpandwaveformgeneratorwewrotecustomIOC(InputOutput Controller) applications to provide an EPICS interface and for the otherdevices existing IOC software, available from SLAC or APS IOC repositories, wasused.OurcustomIOCsarepublicallyavailableongithub:•Syringepumpcontrol:https://github.com/fullerf/LEGATO130•Waveformgeneratorcontrol:https://github.com/fullerf/Agilent33600A

Nature Methods doi:10.1038/nmeth.4195

All our cameras are Allied Vision Manta-style cameras and SLAC has an IOC forthem. For motor actuator control we used a Newport XPS-Q8 with an EPICSinterfaceprovidedbySLAC.Ourtapedrivemotor(LexiumM-Drive)isoperatedviaitsbuilt-inserialterminalinterfaceandcanbedirectlyconnectedthenetwork.Wemainly control the speed via adjustments of the motor’s slew rate (which areinfrequentlynecessaryoncethecorrectspeedisset;seeLaserTimingsection).ConveyorBeltTheconveyorbeltsystemisalmostentirelycustomengineered.Acomplete listingof all componentswould be overbearing, but the interested reader iswelcome tobrowse the technical drawings and 3D files in attached to this protocol as a .ziparchive. The general systems that are required are listed here with some of thedesignconsiderationsthatwentintothem.BeltMaterial&PropertiesWehaveusedseamlessbeltsmadeofKaptonwithdimensionsrangingfrom0.063-0.125’’ in width, 31-36’’ diameter, and 50 micron thickness, depending on ourimplementationofthetapedrive(whichchangesfromexperimenttoexperiment).Beltsofthesedimensionsareavailable,uponcustomrequest,fromacompanyBPBInc.,basedoutoftheBayArea,CA.WealsousedMylarfilmatanearlystageofDOTdevelopment, but foundKapton to similarphysicalpropertieswith superior x-rayscatteringproperties.Theseamlessnatureofthebeltisdeemednecessarybecausewithultrasonicallywelded filmbelts (themaincompeting technology)wenoticedthe belt would skip as the weld passed over rollers, which creates problems forsynchronizing the droplets to the XFEL. Belts are considered a consumablecomponent of this setup, but one belt can last several shifts if one is careful toavoidinghittingthebeltdirectlywiththeXFELbeam.SurfaceTreatmentofthePolyimideBeltSeamlesspolyimide(PI)belts(Kapton100H,C22H10O2N5,DuPont)wereobtainedfrom BPB Inc., USA. To clean the surface, PI belts were first sonicated in Milli-Qwaterfor10minutesfollowedbysonicationinisopropanolfor10minutes.PIbeltswerethendriedat50°Cinavacuumoven(Precisionvacuumoven,model19)foranhour.Next, thebeltswere treatedwith fluoroalkylsilane (FAS), 1H,1H,2H,2H-Perfluorodecyltriethoxysilane(C10F17H4Si(OCH2CH3)3,97%fromSigma-Aldrich)to achieve optimum hydrophobicity. Typically, the belts were refluxed in 0.25%(w/v)solutionofFAS inn-octane for2hours.ToremoveresidualFAS,beltsweredriedinvacuumovenat150°Cfor3hours.AfterFAStreatment,thewatercontactangle(WCA)increasedfrom~60°to100°±5°.

Nature Methods doi:10.1038/nmeth.4195

Tensioner

There are numerous ways one can apply tension to the belt, but we highlyrecommend that one employ a system that allows for repeatable and consistenttension.Toomuchtensionwill irreversiblydamagethebeltandtoolittlepreventsthebeltfromtrackingontherollerscorrectly.Overtensionishappeningifthebeltstartstocurlwhenthetensionisreleased.Oursystem,depictedinFigure2b,usesahangingweighttotorquethebeltovertworollersonarotatingarm.Thissystemiseasytomachineandthetensionmaybequicklymodifiedbyaddingsimplyaddingorsubtractingweights.CrownedRollersToreducefrictionasthebeltmovearound,wepassthebeltoverhighquality(ABEC5), low friction ball bearing rollers. In order to keep the belt centered on theserollers, a crown (slight outward radius of curvature) is addedbypress-fitting thebearing into a CNC machined cap. We used a 3 inch radius of curvature for thecrown.Crownedrollersarecommon innumerousbelt-drivenapplications for thissamepurpose.ElectrostaticControlWefindthatbothpassiveandactivecontrolofstaticbuildupisrequiredtoavoiddisturbingtheADEinjector.Wepositionsteelanti-staticbrushesnearthebeltdrivemechanismtomitigatestaticbuildup.WealsosprayionizedaironthebeltusingaTransformingTechnologiesIN3425staticgun.

Figure2:crownedbearings(a)andthebelttensioner(b).Thetensionercontactsthebelt(depictedinbrowncolor)viatwocrownedbearingsonarotatingturntable.

Nature Methods doi:10.1038/nmeth.4195

Mounting

OursystemismountedonaNewportUTS-100motorizedstageandconstructedofamodifiedThorlabsaluminumbreadboardsandmountingbrackets.(Figure3)DriveMechanismALexiumM-DriveLMDCE423micro-steppingsteppermotordirectlycoupledtoa1’’diameter stainless steel shaft drives the conveyor belt. Originally a gearedmechanism was attempted, but we find the gearing mechanism introducedunwantederrorinthevelocityandsoadirectdriveapproachisrecommended.

Figure3:AnoverallComputerAidedDesign(CAD)renderoftheconveyorbeltsystem,highlightingthemainstructuralcomponentsandmotors.

Nature Methods doi:10.1038/nmeth.4195

SampleEnvironment

A custom-made heliumenclosure made of 0.5’’acrylic was produced byProfessional Plastics Inc.(designed by the authors).Themainchallengehereistoaccommodate a largeinflatable door with an wideaperture near the X-rayinteraction region. Aschematic of the heliumchamber is shown in Figure4.Wherepossiblewehadtheacrylic joints solventwelded,otherwise rubber foamgaskets are used to preventHelium from leaking out ofthechamber.Acustom-madefeed-through panel was cutfrom 0.125’’ aluminum usinga water jet to accommodateall the electronic, gas, water,and fiber feed-throughs intothebox.ConveyorBeltCleaningCleaning thebelt so that it isfree of sample residue andwater is critical for thesuccessful operation of theDOT system. Accumulatedsample residue can bedetected quite sensitively byrunning the belt with nosamplebeing injectedonto itwhileprobingtheXESsignal.This test was performedperiodically through thecourseofourbeamtimes.Beltcleaning is accomplished bypassing the belt over fourhighpressure(200+psi)500µmI.D.waterjets.Thebeltis

Figure4:ACADrenderofthewide-apertureHeliumchamberusedforsimultaneousXRD/XESstudies.Panel(a)showsthechamberfromananglelookingupstreamoftheXFEL.Panel(b)showstheenclosurefromthetopwiththeXFELbeamrunningtoptobottom.Thefrontdooroftheenclosure(showningrey)isaremovablealuminumframeoverwhichaflexiblepolyethylenefilm“bag”(notshown)wassecured.Thebaginflateswhenthechamberisfilled,closingavariablegapbetweenthechamberandtheforwardscatteringdetector.AnotherflexiblepolyethylenebagfilledwithHelium(notshown)isplacedbetweentheXEScrystalspectrometerandtheHeliumchamber.

Nature Methods doi:10.1038/nmeth.4195

then immediately dried on top and bottom by two compressed Helium blow offnozzles. Three gas-tight diaphragmpumps (KNFNeubergerN145.1.2AN.18) drawheliumfromtheHeliumchamberandthenrecirculatethecompressedgasbackintothechambertopowertheblow-offnozzles.ReactionInitiationSystemsLaserexcitationsystem

Asdiscussed inmoredetaillater in this section, multi-step photoinitiatedreactions (like that ofPhotosystem II) require avery high precision spatialgrid of excitation points(Figure 5). To this end, weused a CNC machine toproduce a high precisiongrid of holes (tolerance to0.0005’’).Bareendedfibersarepositionedinthegridofholes and then our fiber-coupled laser focusingsystemisalignedtothegridof holes by maximizingthroughput into the bareended fibers. Each of ourthree fiber-coupledexcitations is given 5degrees of freedom tooptimize coupling into theend of the receiving barefiber.Tomonitorthearrivaltimeofdroplets in thegridwe use bare ended fibersthatemitIRlightthescatter

ofwhichiscollectedontoaphoto-diode.TheIRemittingfibersarealsomountedinthe high precision grid same as those that are used to align the excitation light.During operation,we collect the excitation light that passes through the belt andsampleforallfivefibers(3excitationand2IRgates)ontophotodiodesandmonitortheseonremote-accessibleoscilloscopes.

Figure5:ACADrenderofthecustomfiberexcitationmountingsystemusedtostudyPSII.Threelaserexcitationsaredeliveredatthetwosidesandinthecenter,eachpositionedwith5degreesoffreedom.Betweenthelaserexcitations,IRgatesarepositionedandtheirscatteriscollectedbyangled90°reflectivefibercouplers.

Nature Methods doi:10.1038/nmeth.4195

FreeSpaceTiming

For experiments requiring only a single laser excitation, we typically employ amovablelaserexcitationpointneartheX-rayinteractionregion.Thelaserismovedintothedesiredpositionbyadjustingamotorized1mfocallengthlensupstreamofthe laser interaction region. In this way we can target droplets up to 6 mm inadvanceoftheXFELinteraction,whichmeansfortypicalbeltspeeds(60mm/sto300mm/s)wecanachieveupto100msto20msdelaysrespectively.Todetermineatwhichpositionthedropletshouldbehitinordertoachieveaspecifictimedelay,we use a stroboscopic video system that observes the droplet orthogonal to thebeam(inaplaneapproximatelyparalleltotheground).This“side-view”cameraissynchronizedtotheXFELarrivalwithavariabledelay.WhenthecameraisatzerodelaywithrespecttotheXFEL,onecanobservetheXFELhittingdropletsbyseeingthedropletexplode.Thecameradelayischangedtoanegativedelayelectronicallyto match the desired laser delay and the position of the closest droplet to theinteraction region is noted. The laser spot ismoved to the noted position. In thiswayweareabletodirectlyconfirmspatialoverlapaccurateto100microns(witha

Figure1:Aspace-timediagramillustratingthetwodegreesoffreedominthemulti-excitationschemeweemployedforPSII:Δτdep(thedepositiondelay)andΔτlaser(thelaserdelay).Twodifferenttapevelocitiesareshowninred(slow)andblue(fast),evidencedbytheirdifferentslopes.

Nature Methods doi:10.1038/nmeth.4195

laserspotsizethatistypically400microns)andtemporaloverlaptoaccuratewithin200 microseconds (the camera exposure time). Laser timing (relative to XFELarrival)isactuallymuchmoreaccuratelydetermined(withinnanoseconds)timingbyrelativephotodiodemeasurements.FiberExcitationTimingAsmentioned in the Laser excitation system section, the sample is passed over agrid of laser excitation points. Assuming a constant velocity of the belt, we canvisualize the timingof thesystem through theuseofa space-timerepresentation.Trajectoriesofconstantvelocityappearaslineswithaslopecorrespondingtotheirvelocityandtheexcitationpositions,whicharefixedinspace,arethusfixedintime.Inordertomakelaserexcitationsmatchdropletarrivaltimeswerequirefirstthatthe deposition time of the droplet be such that the droplet arrives at the XFELposition at the right time. In order for the optical laser pumps to also strike thedropletswerequirethatthedelayof thefiber laserexcitationsbesetsothattheystrike the sample, which is a second degree of freedom. The desired depositiondelayisfoundbyadjustingituntiladropletappearsinthecrosshairsoftheXFELbeam (as seen on an in-line stroboscopic XFEL synchronized videomonitor) andthenfineadjusteduntilthedropletisobservedtoexplode.Additionalconfirmationthat the droplet is being hit comes from the observation of solvent scatter in theforward scatteringdetectoror emission signal in the emissiondetector.Once thisdelay is established,we change the belt velocity until the signal from the two IRgates overlap in time. This indicates that droplets passing over the IR gates areseparated by a time equal to the deposition period, i.e. that a drop is on both IRgatesatthesameexacttime.IRgatesareseparatedby60mmfromeachotherand30mmfromthenearestexcitationpoint.Providedthatthebeltspeedissuchthatthespacingbetweendropletsisanintegermultiple(ordivisor)ofboth30mmand60 mm, the time coincidence of IR gate signals indicates that droplets will besimultaneously present on all excitation points. Thus allowed speeds at 10 Hzdroplet rep rate are 30 mm/s, 60 mm/s, 150 mm/s, and 300 mm/s. The fiberexcitation delay needs, therefore, to be time coincident with the IR gate signals.Using this feedbackwe candial in the twodelays (deposition and laser delay) sothatwecanbesurebothopticalpumpsandtheXFELarehittingthedroplets.Figure6illustratesthesevariousdelaysinaspace-timerepresentation.GasActivatedReactionsTo generate reaction intermediates formed by interactionwith a reactive gaswepass the conveyor belt through a region of reactive gas with limited extent. Theregion limits of the reactive gas (oxygen in our case) are created by differentialpumping, i.e. separating the Helium partial pressure from reactive gas partialpressureviaaseriesofsmallorifices.Theconveyorbeltpassesthroughtheorifices.A schematic is shown in associated publication. Unlike laser excitations, the gasregion is not impulsive and so “timing” is less involved. One ismainly concernedwith the dead-time, i.e. the time the sample spends after it exits the reactionchamberbefore it isprobedby theXFELand the exposure time it receives in the

Nature Methods doi:10.1038/nmeth.4195

chamber.TheexposuretimeisgivenbyextentofthereactivegasLdividedbythetapevelocityvtape:L/vtape.Thedead-timeisgivenbyratioofthedistancefromtheendofthereactivegasextenttotheXFELinteractionLxfeltothetapevelocityto:Lxfel/vtape.Timing:Fabrication: The custom components in our latest setup took about a month,distributedoverseveralcontractors.Assembly:Assemblyofallcomponentsconsumedaboutamonthfortwopeople.Setup:Settingupthesystematthebeamlinetookfivedaysfor3people.Data collection: For each experiment we collected data for 4-5 days at thebeamline.AnticipatedResults:SeeassociatedpublicationReferences:1. Ellson, R. et al. Transfer of LowNanoliter Volumes betweenMicroplates UsingFocused Acoustics— Automation Considerations. J. Assoc. Lab. Autom. 8, 29–34(2003).2. Yin, X. et al. Hitting the target : fragment screening with acoustic in situ co-crystallization of proteins plus fragment libraries on pin-mounted data-collectionmicromeshesresearchpapers.ActaCrystallogr.Sect.DBiol.Crystallogr.70,1177–1189(2014).Acknowledgements:FDF,SG,RCwrotetheprotocolwithcontributionsfromJFK,VKY,andJY.AssociatedPublications:Thisprotocolisrelatedtothefollowingarticles:Fulleretal.,Drop-on-DemandSampleDeliveryforStudyingBiocatalystsinActionatXFELs.NatureMethods,2016.AffiliationsYano/YachandraLabFranklinFuller,SherazGul,RuchiraChatterjee, JanKern,VittalYachandra&JunkoYanoUnaffiliatedCompetingfinancialinterestsTheauthorsdeclarethattherearenocompetingfinancialinterests.CorrespondingauthorCorrespondenceto:JunkoYano(jyano@lbl.gov)

Nature Methods doi:10.1038/nmeth.4195