Time-lapse imaging of disease progression in deep brain areas using fluorescence microendoscopy

Transcription

1 Time-lpse imging of disese progression in deep rin res using fluoresene miroendosopy Roert P J Brretto 1,6, Tony H Ko 1,6, Juergen C Jung 1,6, Tmmy J Wng 1, George Cpps 1, Allison C Wters 1, Yniv Ziv 1, Alessio Attrdo 1, Lwrene Reht 2,3 & Mrk J Shnitzer 1,4,5 211 Nture Ameri, In. All rights reserved. The omintion of intrvitl mirosopy nd niml models of disese hs propelled studies of disese mehnisms nd tretments. However, mny disorders fflit tissues inessile to light mirosopy in live sujets. Here we introdue ellulr-level time-lpse imging deep within the live mmmlin rin y one- nd two-photon fluoresene miroendosopy over multiple weeks. Bilterl imging sites llowed longitudinl omprisons within individul sujets, inluding of norml nd disesed tissues. Using this pproh, we trked CA1 hippompl pyrmidl neuron dendrites in dult mie, reveling these dendrites extreme stility nd rre exmples of their struturl ltertions. To illustrte disese studies, we trked deep lying glioms y oserving tumor growth, visulizing three-dimensionl vsulture struture nd determining miroirultory speeds. Averge erythroyte speeds in glioms delined mrkedly s the disese dvned, notwithstnding signifint inreses in pillry dimeters. Time-lpse miroendosopy will e pplile to studies of numerous disorders, inluding neurovsulr, neurologil, nerous nd trum-indued onditions. Chroni niml preprtions tht permit time-lpse intrvitl mirosopy hve llowed longitudinl imging studies of disese models nd hve yielded insights regrding disese mehnisms nd therpeuti strtegies 1. In neurosiene, time-lpse mirosopy studies 2 5 hve exmined disese nd injury in the peripherl nervous system 6,7 nd superfiil neoortex 1,8 11. Although mny pthologies fflit deeper rin strutures, limited penetrtion of light into tissue preludes intrvitl mirosopy in deep res 1 suh s the hippompus or stritum. To overome this, we developed time-lpse pilities for in vivo miroendosopy 12,13 llowing repeted oservtions over weeks in deep rin res. Miroendosopy enles ellulr imging eneth the penetrtion depth of onventionl light mirosopy 14, relying on mirolenses to provide miron-sle resolution during diret insertion into tissue 12,13,. Contrst modlities omptile with miroendosopy inlude widefield epifluoresene s well s lser-snning onfol, two-photon fluoresene, nd seond-hrmoni genertion fluoresene 14. The use of these modlities, inluding in humns 16,17, hs enled ellulr imging in the ohle, hippompus, thlmus, deep neoortex, digestive trt nd musles, leit only in ute studies 14. Here we introdue time-lpse miroendosopy nd illustrte its ppliility to studies of rin disese. We developed mouse preprtion permitting repeted one- nd two-photon fluoresene imging y insertion of miro-optil proes into surgilly implnted guide tues. These two modlities respetively enled high-speed (1 1,2 Hz) nd three-dimensionl imging t the sme tissue sites. We minly studied hippompus ut in some mie exmined stritum. In dult mie expressing fluoresent proteins in suset of CA1 hippompl neurons 18, time-lpse miroendosopy llowed us to test the hypothesis tht these neurons grdully hnge their dendriti rnhing ptterns. This hypothesis is ontrry to results from neoortil pyrmidl ells 5,19,2, ut CA1 reeives disynpti input from the dentte gyrus, where new neurons re ontinully dded throughout dulthood 21. It hs een unknown whether iruits downstrem from the dentte gyrus undergo onsequent remodeling, nd we expeted to see suh n effet. Insted, the dt reveled mrked stility of CA1 neurons dendriti struture, rising questions out how iruits downstrem of the dentte gyrus ommodte ontinul ddition of new inputs. To illustrte the ppliility to disese studies, we exmined mouse model of gliom, the most ommon primry mlignnt rin tumor. For poorly understood resons, gliom growth depends on ntomil lotion. Primry glioms rise preferentilly in deep rin strutures, nd lotion orreltes with tumor phenotype Studies in rts hve suggested speifi deep regions ssoite with the highest rtes of gliom growth 25. Thus, the lol miroenvironment, inluding onentrtions of ngiogeni ftors, is thought to e mjor influene on tumor enlrgement 26. Intrvitl mirosopy studies of gliom ngiogenesis hve involved tumor implnttion into the superfiil neoortex, outside norml sites of primry inidene 27,28. Time-lpse miroendosopy llowed us to oserve gliom ngiogenesis in n orthotopi, deep position in the rin nd to trk hllmrk fetures, inluding vessel sizes nd flow speeds. Our experimentl design, in whih eh mouse yields dt from oth norml tissue nd tumor t the ilterlly symmetri lotion, 1 Jmes H. Clrk Center for Biomedil Engineering & Sienes, Stnford University, Stnford, Cliforni, USA. 2 Deprtment of Neurology nd Neurologil Sienes, Stnford University, Stnford, Cliforni, USA. 3 Deprtment of Neurosurgery, Stnford University, Stnford, Cliforni, USA. 4 Howrd Hughes Medil Institute, Stnford University, Stnford, Cliforni, USA. 5 CNC Progrm, Stnford University, Stnford, Cliforni, USA. 6 These uthors ontriuted eqully to this work. Correspondene should e sent to M.J.S. (mshnitz@stnford.edu). Reeived 5 Novemer 29; epted 7 Otoer 21; pulished online 16 Jnury 211; doi:1.138/nm.2292 nture mediine dvne online pulition

2 Figure 1 Chroni mouse preprtion for repeted imging of deep rin tissues using miroendosopy. () The experiment egins with implnttion of imging guide tues, one into eh hemisphere of the mouse s rin. After the mouse reovers from surgery, miro-optil proes n e repetedly inserted into the guide tues to llow time-lpse imging. () Three miroendosope proes, 5-µm-dimeter singlet, 1-mm-dimeter singlet nd 1-mm-dimeter ompound doulet. Sle r, 1 mm. () Shemti of mirosope ojetive lens oupling illumintion into 5-µm-dimeter singlet miroendosope proe, suh s tht shown in. Mouse rin Miro-optil proe Experimentl tissue Imging guide tues Control tissue Ojetive 211 Nture Ameri, In. All rights reserved. seprtes puttive effets of the methodology from the disese. We performed >, determintions of vessel dimeters nd flow speeds. The dt show tht the si ngiogeni dynmis seen in superfiil tissues 26 re not speifi to those res nd revel the kinetis with whih vsulr morphology nd flow speeds evolve during gliom progression. This prompts future miroendosopy studies of how nd why tumor dynmis vry ross rin res. RESULTS Chroni preprtion for time-lpse miroendosopy We developed mouse preprtion for repeted imging of deep rin tissue (Fig. 1). Implnttion into the rin of one or two (ilterl) guide tues llowed repeted insertion of miroendosopes to the sme sites (Fig. 1). Eh guide tue omprised glss pillry with glss window t the tip, llowing optil ut not physil ess to tissue. In size, the miroendosopes (5- nd 1,-µm lens dimeters) resemle mirodilysis proes more thn miroeletrodes (Fig. 1), so the pillries hd to e ppropritely pled to minimize mehnil perturtion to the imged tissue. We hieved this y pling the guide tues tips just dorsl to, ut not within, the rin struture of interest. Neither the guide tues nor the miroendosopes entered the tissue eing studied, insted llowing imging from nery. Most of our studies were of hippompus, with guide tues just dorsl to the CA1 hippompl re, llowing our miroendosopes to inspet CA1 (Fig. 1). For imging stritum the guide tues resided just dorsl to this struture. After llowing the mie to reover from surgery, we performed miroendosopy repetedly over periods regulrly lsting 7 weeks, ut in few mie >1 yer. To initite imging, we inserted miro-optil proe into mouse s guide tue, nd we pled the mouse on the stge of n upright mirosope modified to ommodte miroendosopy 13,. The mirosope ojetive lens foused illumintion ner the top fe of the miro-optil proe, whih refoused the illumintion into the rin (Fig. 1). These proes were generlly ompound doulet grdient refrtive index (GRIN) mirolenses, omposed of n ojetive (.49 numeril perture) nd rely lens (.2 numeril perture), or single GRIN mirolenses ( numeril perture) (Fig. 1). A snug fit etween the proe nd guide tue filitted multiple returns to the sme tissue for one- nd two-photon miroendosopy. Fol djustment of the mirosope ojetive lens ltered the fous in tissue 12,13,. This enled three-dimensionl setioning y two-photon miroendosopy, in some ses up to ~65 µm from the tip of the guide tue (Supplementry Video 1). This optil penetrtion is muh further thn reported for ute miroendosopy 12,13 nd nerly omprle to depths (~6 7 µm) hievle y onventionl two-photon mirosopy 29. As with ny foreign entity implnted into the rin inluding eletrodes, mirodilysis proes or rnil windows we expeted glil tivtion surrounding the implnted guide tues. Post-mortem tissue exmintions ner our guide tues onfirmed this, reveling Dorsl Rostrl Miro-optil proe Skull thin (~25 4-µm) tissue lyer showing greter leling y ntiodies speifi for glil firillry idi protein s ompred to djent tissue (Supplementry Fig. 1 e). These speimens reveled no loss of neurons eneth the guide tues, s ompred to ontrlterl hippompl res in the sme mie (P < ; one-tiled Wiloxon signed-rnk test; n = 3 mie, 2,98 ells). Given the optil penetrtion of up to ~65 µm, the glil tivtion lyer did not ostrut the imging of tissue lying well eyond. This is onsistent with reports desriing mirosopy through rnil windows, whih use similr levels of gliosis 3. The min implition is tht miroendosopy studies of disese must ontrol for the potentil side effets of guide tue implnttion to seprte them from the effets of disese. Repeted imging of neurons nd mirovsulture Before studying disese, we ssessed our method s si pilities nd found it possile to trk individul neurons, dendrites nd ererl mirovsulture over weeks. In mie expressing yellow or green fluoresent protein (YFP or GFP) in sprse susets of CA1 pyrmidl neurons 18, we monitored neurons nd dendrites (Fig. 2). Three-dimensionl dt from twophoton miroendosopy filitted inspetion of dendriti morphologies up to ~53 µm from the dorsl CA1 surfe (Fig. 2,). This overs ll of hippompl lyers strtum oriens nd strtum pyrmidle, where pyrmidl neurons hve their sl dendrites nd somt, respetively, nd midwy into strtum rditum, where these ells pil dendrites lie. Whether dendriti rnhing ptterns of dult hippompl pyrmidl neurons re plsti or stle hs remined unnswered owing to the lk of suitle imging tehniques. Unlike in ute imging studies of hippompus 13,31, we repetedly inspeted dendriti rnhing ptterns over weeks (Fig. 2 ). We nlyzed imges of 4,257 dendrites, trked vi totl 33,596 dendrite oservtions over 7 weeks in ten mie (Supplementry Methods). The dt reveled only 16 instnes of dendrite turnover (Supplementry Fig. 2), defined s the dendrite either showing new segment or segment disppering from view. This yielded n estimte of <.13 ±.3 (men ± s.e.m.; n = 33,596 oservtions) for the proility per dy of n individul dendrite undergoing turnover event. This rte of hnge should e regrded s n upper ound nd, under simplifying ssumption of uniform rte, orresponds to n estimted lower ound of 8, ± 2, d for the men stility lifetime of dendrite, longer thn mouse s lifetime. Investigtions of how this stility might e ltered in the disesed hippompus represent importnt future studies. dvne online pulition nture mediine

3 Figure 2 Time-lpse two-photon miroendosopy of CA1 hippompl neurons. Time-lpse imge sequenes of CA1 pyrmidl neurons in three Thy1-GFP mie. (,) Two-dimensionl projetions of three-dimensionl stks ontining four imge slies quired t 4.2-µm xil sping over 16.8 µm in depth. shows enlrgements of the oxed re in. () Two-dimensionl projetions of three-dimensionl stks quired t 3-µm xil sping over pproximtely 54 µm in depth. (d) Enlrged, single-imge frmes reveling spiny dendrites. Sle rs in,, nd d re 1, 25, 5 nd 5 µm, respetively. Dy 41 Dy 4 d 211 Nture Ameri, In. All rights reserved. In wild-type mie, three-dimensionl imging enled exmintion of vsulr Dy 49 networks in hippompus (Fig. 3 nd Supplementry Video 1) nd stritum (Supplementry Fig. 3), whih were reveled up to over ~65 µm depths following intrvsulr injetion of fluoresein-dextrn dye. After struturl imging y two-photon miroendosopy, we ould swith immeditely to high-speed (1 1,2 Hz) imging of erythroyte flow y one-photon Dy 58 miroendosopy (Fig. 3). Single erythroytes were generlly pprent, ppering in relief ginst the dye-leled lood plsm (Supplementry Video 2). Computtionl nlysis extrted struturl prmeters, suh s vessel dimeters, nd physiologil prmeters, suh s flow speeds (Supplementry Dy 61 Fig. 4). Suh time-lpse monitoring of ererovsulr properties should lso e pplile to studies of stroke or neurodegenertive diseses. Time-lpse imging of gliom ngiogenesis To illustrte studies of disese, we exmined mouse model of gliom. We implnted two guide tues t ilterlly symmetri sites, dorsl to Dy 22 Dy 5 Dy 11 Dy 11 Dy 14 Dy 35 Dy 8 Dy 18 left nd right CA1 res (Fig. 1). We inoulted the right site with GL261 mouse gliom ells; the left site reeived no gliom ells nd served s ontrol. This enled ontrol nd experimentl dt to e quired from eh mouse, yielding mthed sets in regrd to sujets ges, sexes nd life histories. Using time-lpse miroendosopy, we were generlly le to trk imging sites over the entire disese time ourse, ~22 24 d in totl from gliom inoultion to sujet mortlity. We lso d Dy 13 e Speed (mm s 1 ) Dy 1 Dy Dy 13 Dy Figure 3 Time-lpse miroendosopy of CA1 mirovsulture shows norml lood vessel morphologies re stle over time. () Twodimensionl projetion of three-dimensionl stk of 22 imges of dye-leled vsulture quired t ~3-µm inrements over ~66 µm in depth. Supplementry Video 1 shows the entire imge stk. () Time-lpse imge sequene quired y one-photon miroendosopy. () Time-lpse sequene of two-photon imge stks, eh omposed of 4 5 imges quired pproximtely 3.7 µm prt in depth nd projeted to two dimensions. Figure 4 shows dye-leled tumor vessels from the opposing (experimentl) hippompus in the sme mouse. (d,e) One-photon imge of CA1 lood vessels (d) nd orresponding miroirultory speed mp (e), determined y high-speed (1-Hz) imging nd rossorreltion nlysis. Supplementry Video 2 shows lood flow from this sme field of view. Sle rs re 1 µm in nd 5 µm in d. nture mediine dvne online pulition

4 T e h n i l R e p o rt s Dy 3 d 1..5 Control (L) Quntittive nlysis of ngiogenesis We quntified hnges in vessel morphology nd flow speeds y nlyses of the three-dimensionl nd high-speed imging dt sets, respetively (Figs. 3e, 4d, 5 nd Supplementry Fig. 4; Supplementry Methods). Individul mie showed onsiderle dy-to-dy nd 1 Flow speed (mm s ) Control Tumor Vessel dimeter (µm) 1 Flow speed (mm s ).6 45 Vessel dimeter (µm) Dy 1 Dy 4 Dy 7 Dy 1 Dy 13 Dy 16 Dy 19 Dy Tumor Control Time (d) 5 1 Dy 11 Dy 4 Dy 13 Dy 12 Dy Dy 16 Dy 17 Dy 18 Dy 2 performed postmortem histopthologil nlyses to verify tumor invsiveness (Supplementry Fig. 1f,g). In some mie we performed dul-olor miroendosopy, llowing us to visulize oth tumor ells nd lood vessels y using GFP-trnsfeted GL261 ells nd intrvsulr injetion of rhodmine dye (Fig. 4). Quntittive studies of ngiogeni dynmis relied on untrnsfeted GL261 ells nd one fluoresene hnnel dedited to vsulr imging. Over the weeks of gliom growth, we sw progressive deformtions nd size inreses of tumor vessels (Fig. 4,). These onspiuous hnges (Fig. 4) did not our t ontrol sites in the sme mie (Fig. 3). In erly disese stges (dys 1 5), tumor vessels showed distortions from norml shpes (Fig. 4). At intermedite stges (dys 6 ), tumor vessels inresed in dimeter nd vsulr distortions worsened (Supplementry Video 3). By lte stges (dys 16 24), mny tumor vessels ppered thromoti, nd lood flow dropped notiely in tumors ut not norml tissue (Fig. 4 d). 2 Speed (mm s 1) 211 Nture Ameri, In. All rights reserved. Figure 4 Time-lpse imging of gliom ngiogenesis in mouse CA1 revels progressive distortions to vsulr geometry nd redued miroirultory speeds. () Dul-olor imge of GFP-expressing mouse gliom ells (green) nd rhodmine-dextrn leled mirovsulture (red), quired in live mouse y one-photon fluoresene miroendosopy on dy 3 fter gliom ell inoultion. (,) Time-lpse sequenes of two-photon miroendosopy imge stks, projeted to two dimensions, showing the progressive distortion of the mirovsulture due to gliom ngiogenesis. Eh stk ontined 4 5 imges quired 3.7 µm prt in depth. Figure 3 shows imges of the opposing (ontrol) hippompus from the sme mouse used in. (d) Mps of verge erythroyte speed, determined from 1-s videos quired t 1 Hz y one-photon miroendosopy, in left (ontrol) nd right (experimentl) hemispheres of live mouse on dy 2 fter gliom inoultion. Sle rs, 1 µm. Tumor (R) etween-sujet vritions ut generlly refleted findings mde from the ggregted popultion dt. Throughout gliom progression, men vessel dimeters in norml tissue remined onstnt (P =.23; Kruskl-Wllis nlysis of vrine) nd were 12.6 ± 1.7 µm (grnd men ± s.d.; n = 7 mie; 2,83 dimeter mesurements ggregted over ll dys) (Fig. 5,). Not surprisingly, erythroyte flow speeds in norml tissue (389 ± 191 µm s 1; n = 1 mie; 8,6 speed me surements over ll dys) flututed more dy to dy thn did vessel dimeters ut showed no signifint hnges ross the experiment s durtion (P =.6) (Fig. 5,). By omprison, progressive inreses in vessel dimeter nd dereses in flow speed t gliom sites were signifint (P =.9 nd P =.1, respetively; Kruskl-Wllis nlysis of vrine). Aggregted mesurements of dimeter (17. ± 5.3 µm; n = 7 mie; 1,695 mesurements over ll dys) nd speed (134 ± 112 µm s 1; n = 1 mie; 5,822 mesurements) from gliom lso differed signifintly Figure 5 Quntittive trking of gliom ngiogenesis in CA1 hippompus shows tumor vessels roden in dimeter ut undergo mrked delines in flow speed. Vessel dimeters nd erythroyte flow speeds were monitored s funtion of elpsed time fter n initil surgery nd gliom ell implnttion on dy. Not ll mie were imged t identil time points, so time vlues were inned into 3-d intervls. () Plots of flow speed versus vessel dimeter, for ontrol tissue (left) nd tumor sites (right), in whih eh dt point represents n individul vessel segment. Unlike t ontrol sites, t tumor sites the dt revel n overll progressive derese in flow speeds nd inrese in vessel dimeters s the disese dvned (n = 1 mie). () Popultion verges of erythroyte flow speed (left) nd vessel dimeter (right) re plotted (men ± s.e.m.) versus elpsed time. Men tumor vessel dimeters nd flow speeds differed signifintly from ontrol vlues (P < nd P < 1 1, respetively; Mnn-Whitney U test). dvne online pulition nture mediine

5 211 Nture Ameri, In. All rights reserved. from norml tissue (P < 1 6 nd P < 1 1, respetively; Mnn- Whitney U test). Signifint differenes in flow speed etween norml nd tumor sites ppered nd were mintined from s erly s dy 7 (P <.5; Mnn-Whitney U tests). Signifint differenes in vessel size were mintined from dy 16 onwrds (P <.5). Gliom lso ltered vsulr network struture, s rnhing rtios (the rtios of vessels totl lengths to the numer of rnhes) were lower t tumor sites (245 ± 13 µm per rnh; men ± s.d.; n = 1 mie) thn in ontrol tissue (43 ± 19 µm per rnh) from dy 13 onwrd (P <.5; Mnn-Whitney U tests). Overll, our nlyses showed tht, s the disese progressed, men flow speeds nd rnhing rtios delined mrkedly in gliom, ut not in norml tissue, despite n inrese in verge tumor vessel dimeters (Fig. 5). Plots of the vessels speeds versus their dimeters illustrted tht, unlike t norml sites, t tumor sites the reltionship etween these vriles showed onspiuous, progressive ltertions (Fig. 5). Thus, the si fetures of gliom ngiogenesis regrding redued flow speed nd distorted vsulture, previously identified in superfiil rin tissues 26 28, lso rise for glioms in n orthotopi, deep-rin lotion. DISCUSSION We hve introdued mens for ellulr-level, time-lpse imging of deep rin strutures. We foused on the hippompus, ut our methodology is not speifi to this re, s illustrted in stritum (Supplementry Fig. 3). Our pproh drws on the long experiene in neurosiene with implnttion of similrly sized guide tues for mirodilysis nd eletrophysiologil reordings nd should e rodly pplile. The imging depths ttined here, up to ~65 µm from the guide tue, will filitte studies of numerous disorders. Alterntively, one might implnt miro-optil proes diretly into the rin without guide tues. However, guide tues permit individul miro-optil proes to e used ross mny nimls nd llow inspetion of one tissue site with multiple proes of different optil designs. A key dvntge of our pproh over histologil pprohes onerns the experimentl designs tht n e hieved y longitudinl studies, s ompred to those in whih tissue ulture speimens re tken from different ohorts of nimls t distint time points 1. The enefits of longitudinl imging inlude sustntil redution in the numer of experimentl nimls needed, s eh niml idelly provides dt t ll time points. Moreover, longitudinl imging yields dynmi informtion from individul nimls, whih is often pivotl for disese studies y reveling how erly symptoms orrelte with lter disese outomes or might e reversed y intervention. As illustrted y our oservtions of hippompl dendrites, trking individul ells voids the onfound of ell-to-ell vritions in the nlysis of temporl dynmis. The ility to monitor dendrites should e prtiulrly useful for studying developmentl 32, neurodegenertive 8, ererovsulr 33 nd epilepti 34 disorders tht ffet dendriti strutures. The pity to monitor stritum should id the study of severl prevlent motor nd dditive disorders. Although time-lpse miroendosopy opens new experimentl possiilities, reserhers should lso ppreite the limittions. Guide tue implnttion neessrily perturs the rin. We minimized the impt y pling guide tues outside, not within, the tissue eing imged. Implnting miroendosope with miroprism for sidewys viewing of tissue djent to the insertion pth 35 or dvning miroendosope over dys, kin to how eletrodes re often implnted, my e vile lterntives. Our si pproh indued ~25 4-µm lyer of glil tivtion surrounding the implnt. As with implnted rnil windows for intrvitl mirosopy 3, glil tivtion generlly delined over time nd permitted imging well eyond the tivted lyer. Nevertheless, reserhers must design disese studies tht seprte the effets of experimentl proedures from those of the disese. A reent dete on the reltive merits of the thinned-skull versus rnil-window preprtions for intrvitl mirosopy studies rose in ontext tht did not esily permit omprisons etween ontrol nd experimentlly mnipulted sujets 3. By ontrst, oservle effets of mny rin diseses re profound, s shown here for gliom, nd imges from ontrol nd tumor sites were typilly distinguishle t glne. CA1 pyrmidl neurons support sptil nd episodi memory, ut the long-term stility of hippompl dendrites hd not previously een ssessed. Anlyses of dendrite stility in neoortil pyrmidl neurons hve een sed on dt from ~5 125 dendrites 5,19,2 nd imply ounds on dendrite turnover of <.5.8% over ~1 2 months. We trked >4,2 dendrites over 7 weeks to test the hypothesis tht CA1 pyrmidl neurons undergo grdul hnges in dendrite struture under norml onditions. Our dt provided little support for this ide. The 16 turnover events we sw imply n upper ound on turnover of <.4%, omprle to or less thn tht for neoortil neurons. This estlishes seline stility levels ginst whih studies of disese models or ged nimls n e ompred nd indites tht individul dendriti segments of CA1 pyrmidl neurons hve n estimted men lifetime of >8, ± 2, d. These neurons eh hve ~2 dendriti segments 36, suggesting individul neurons might undergo mjor dendriti hnge every >4 d. Under simplifying ssumption of uniform rte of ltertion, <4% of dendriti segments would hnge over yer. Thus, we nnot rule out dendriti turnover, ut the extremely few instnes we sw re just ove the estimted reliility limits of our soring proedures (Supplementry Methods). We studied helthy mie housed in stndrd onditions, nd it will e interesting to test whether mnipultions tht inrese hippompl plstiity, suh s environmentl enrihment, promote dendriti plstiity ove the rtes reported here. Gliom is the most ommon primry intrrnil tumor in humns 37. As gliom preferentilly strikes deep rin lotions 22 24, it hs een hllenging to study the ngiogeni dynmis in orthotopi models. Using time-lpse miroendosopy, we trked ngiogenesis in deep tumors over weeks. The GL261 gliom ells we used possess the min hrteristis of most gliom models, inluding invsive ut nonmetstti growth 38. By monitoring flow speeds, vessel dimeters, nd reltionships etween the two, we found tumor vessels show progressive dereses in flow speed despite more thn douling in men dimeter (Fig. 5). The si dynmil fetures of ngiogenesis seen y intrvitl mirosopy re thus not speifi to superfiil tissues. Miroendosopy should filitte omprisons of ngiogenesis etween different rin lotions nd tumor lines with distint ngiogeni hrteristis 11. Implnttion of multiple guide tues per niml should llow within-sujet omprisons etween different lolly pplied tretments. Overll, time-lpse miroendosopy should rodly enle studies of rin disorders suh s ererovsulr, neurodegenertive, epilepti nd trum-indued onditions tht hve eluded exmintion in deep tissues y light mirosopy 1. There re lredy lrge set of fluoresent mrkers of pertinene to disese studies, suh s lels for myloid plques 1, retive oxygen speies 9 nd gli nd immune ells 1, nd mny of these should e redily usle in omintion with miroendosopy. Miniturized mirosopes sed on miro-optis llow deep tissue imging in ehving mie 39, suggesting the fesiility of time-lpse studies in whih ellulr nd ehviorl effets of disese re visulized simultneously. nture mediine dvne online pulition

6 211 Nture Ameri, In. All rights reserved. Methods Methods nd ny ssoited referenes re ville in the online version of the pper t Note: Supplementry informtion is ville on the Nture Mediine wesite. Aknowledgments This work ws initited under Ntionl Institute on Drug Ause CEBRA DA17895 nd further supported y Ntionl Institute of Neurologil Disorders nd Stroke R1NS5533, Ntionl Cner Institute P5CA nd reserh ontrt with Mun Ke Tehnologies. We grtefully knowledge support from the Stnford US Ntionl Institutes of Helth Biophysis Progrm (R.P.J.B.) nd Mhih postdotorl fellowship (Y.Z.). We thnk M. Lim nd J. Weimnn for help with GL261 ells, S. Kim, T. Jng, A. Lui, J. Li nd M. Rmkumr for tehnil ssistne with histology nd imge proessing, E. Mukmel for helpful onverstions nd B. Colyer nd B. Wilt for help with grphi illustrtion. AUTHOR CONTRIBUTIONS R.P.J.B. designed experiments, developed trking of neuronl dendrites, performed the study on CA1 neuron stility, nlyzed the neuronl histology dt, vlidted the lgorithm for omputing erythroyte speeds nd omputed reltionships etween vessel dimeters nd speeds. T.H.K. designed experiments, performed the gliom experiments nd omputed flow speeds nd vessel sizes. J.C.J. designed experiments, developed the hroni preprtion nd tested it for imging neurons nd glioms. T.J.W. nd G.C. performed neuronl imging nd ontriuted to the gliom experiments. A.C.W. developed ilterl imging, performed neuronl imging nd ontriuted to the gliom experiments. Y.Z. developed nd performed stritl imging. A.A. performed histologil nlyses nd nlyzed vessel rnhing rtios. L.R. designed experiments nd supervised the gliom study. M.J.S. designed experiments, performed sttistil testing, initited nd supervised the projet nd wrote the pper. All uthors edited the pper. COMPETING FINANCIAL INTERESTS The uthors delre ompeting finnil interests: detils ompny the full-text HTML version of the pper t Pulished online t Reprints nd permissions informtion is ville online t reprintsndpermissions/. 1. Misgeld, T. & Kershensteiner, M. In vivo imging of the disesed nervous system. Nt. Rev. Neurosi. 7, (26). 2. Melder, R.J., Slehi, H.A. & Jin, R.K. Intertion of tivted nturl killer ells with norml nd tumor vessels in rnil windows in mie. Mirovs. Res. 5, (1995). 3. Lihtmn, J.W., Mgrssi, L. & Purves, D. Visuliztion of neuromusulr juntions over periods of severl months in living mie. J. Neurosi. 7, (1987). 4. Grutzendler, J., Ksthuri, N. & Gn, W.B. Long-term dendriti spine stility in the dult ortex. Nture 42, (22). 5. Trhtenerg, J.T. et l. Long-term in vivo imging of experiene-dependent synpti plstiity in dult ortex. Nture 42, (22). 6. Kershensteiner, M., Shw, M.E., Lihtmn, J.W. & Misgeld, T. In vivo imging of xonl degenertion nd regenertion in the injured spinl ord. Nt. Med. 11, (25). 7. Breyre, F.M., Kershensteiner, M., Misgeld, T. & Snes, J.R. Trnsgeni leling of the ortiospinl trt for monitoring xonl responses to spinl ord injury. Nt. Med. 11, (25). 8. Tsi, J., Grutzendler, J., Duff, K. & Gn, W.B. Firillr myloid deposition leds to lol synpti normlities nd rekge of neuronl rnhes. Nt. Neurosi. 7, (24). 9. MLelln, M.E., Kjdsz, S.T., Hymn, B.T. & Bski, B.J. In vivo imging of retive oxygen speies speifilly ssoited with thioflvine S positive myloid plques y multiphoton mirosopy. J. Neurosi. 23, (23). 1. Meyer-Luehmnn, M. et l. Rpid pperne nd lol toxiity of myloid-β plques in mouse model of Alzheimer s disese. Nture 451, (28). 11. Yun, F. et l. Vsulr permeility nd miroirultion of glioms nd mmmry rinoms trnsplnted in rt nd mouse rnil windows. Cner Res. 54, (1994). 12. Levene, M.J., Domek, D.A., Ksishke, K.A., Molloy, R.P. & We, W.W. In vivo multiphoton mirosopy of deep rin tissue. J. Neurophysiol. 91, (24). 13. Jung, J.C., Meht, A.D., Aksy, E., Stepnoski, R. & Shnitzer, M.J. In vivo mmmlin rin imging using one- nd two-photon fluoresene miroendosopy. J. Neurophysiol. 92, (24). 14. Fluserg, B.A. et l. Fier-opti fluoresene imging. Nt. Methods 2, (25).. Jung, J.C. & Shnitzer, M.J. Multiphoton endosopy. Opt. Lett. 28, (23). 16. Hsiung, P.L. et l. Detetion of oloni dysplsi in vivo using trgeted heptpeptide nd onfol miroendosopy. Nt. Med. 14, (28). 17. Llewellyn, M.E., Brretto, R.P., Delp, S.L. & Shnitzer, M.J. Minimlly invsive high-speed imging of sromere ontrtile dynmis in mie nd humns. Nture 454, (28). 18. Feng, G. et l. Imging neuronl susets in trnsgeni mie expressing multiple spetrl vrints of GFP. Neuron 28, (2). 19. Lee, W.-C.A. Dynmi remodeling of dendriti rors in GABAergi interneurons of dult visul ortex. PLoS Biol. 4, (26). 2. Chow, D.K. et l. Lminr nd omprtmentl regultion of dendriti growth in mture ortex. Nt. Neurosi. 12, (29). 21. Li, Y., Mu, Y. & Gge, F.H. Development of neurl iruits in the dult hippompus. Curr. Top. Dev. Biol. 87, (29). 22. Lim, D.A. et l. Reltionship of gliolstom multiforme to neurl stem ell regions predits invsive nd multifol tumor phenotype. Neuro-onol. 9, (27). 23. Lrjvr, S. et l. Inidene of glioms y ntomi lotion. Neuro-onol. 9, (27). 24. Rmnryn, R., Dodd, S., Ds, K., Heideke, V. & Rinov, N.G. Overll survivl in ptients with mlignnt gliom my e signifintly longer with tumors loted in deep grey mtter. J. Neurol. Si. 26, (27). 25. Jng, T. et l. A distint phenotypi hnge in glioms t the time of mgneti resonne imging detetion. J. Neurosurg. 18, (28). 26. Jin, R.K. et l. Angiogenesis in rin tumours. Nt. Rev. Neurosi. 8, (27). 27. Kshiwgi, S. et l. Perivsulr nitri oxide grdients normlize tumor vsulture. Nt. Med. 14, (28). 28. Grkvtsev, I. et l. The ndidte tumour suppressor protein ING4 regultes rin tumour growth nd ngiogenesis. Nture 428, (24). 29. Helmhen, F. & Denk, W. Deep tissue two-photon mirosopy. Nt. Methods 2, (25). 3. Xu, H.T., Pn, F., Yng, G. & Gn, W.B. Choie of rnil window type for in vivo imging ffets dendriti spine turnover in the ortex. Nt. Neurosi. 1, (27). 31. Mizrhi, A., Crowley, J.C., Shtoyermn, E. & Ktz, L.C. High-resolution in vivo imging of hippompl dendrites nd spines. J. Neurosi. 24, (24). 32. Dierssen, M. & Rmkers, G.J. Dendriti pthology in mentl retrdtion: from moleulr genetis to neuroiology. Genes Brin Behv. 5, Suppl 2, 48 6 (26). 33. Li, P. & Murphy, T.H. Two-photon imging during prolonged middle ererl rtery olusion in mie revels reovery of dendriti struture fter reperfusion. J. Neurosi. 28, (28). 34. Shpiro, L.A., Rik, C.E. & Jesserger, S. Struturl hnges for dult-orn dentte grnule ells fter sttus epileptius. Epilepsi 49, Suppl 5, (28). 35. Murym, M., Perez-Gri, E., Lusher, H.R. & Lrkum, M.E. Fieropti system for reording dendriti lium signls in lyer 5 neoortil pyrmidl ells in freely moving rts. J. Neurophysiol. 98, (27). 36. Pypli, G.K., Sik, A., Penttonen, M., Buzski, G. & Turner, D.A. Dendriti properties of hippompl CA1 pyrmidl neurons in the rt: intrellulr stining in vivo nd in vitro. J. Comp. Neurol. 391, (1998). 37. Louis, D.N. Moleulr pthology of mlignnt glioms. Annu. Rev. Pthol. 1, (26). 38. Sztmri, T. et l. Detiled hrteriztion of the mouse gliom 261 tumor model for experimentl gliolstom therpy. Cner Si. 97, (26). 39. Fluserg, B.A. et l. High-speed, miniturized fluoresene mirosopy in freely moving mie. Nt. Methods 5, (28). dvne online pulition nture mediine

7 211 Nture Ameri, In. All rights reserved. ONLINE METHODS Mie. We used mie expressing fluoresent proteins driven y the Thy1 promoter (YFP-H nd GFP-M lines; 2 12 months of ge) for neuronl imging 18, ut wild-type mie (BALB/ nd C57BL/6 mie; 2 6 months of ge) for gliom studies. All mouse experiments were pproved y the Stnford Administrtive Pnel on Lortory Animl Cre. Gliom ells. We used GL261 mouse gliom ells 38, mintined in Duleo s modified Egle s medium supplemented with 1% fetl lf serum. In some studies, to visulize tumor progression vi fluoresene we used polylonl mixture of GFP-trnsfeted GL261 lines tht were infeted with mphotropi retrovirus ontining ytomeglovirus hik β-tin promoter driving GFP (reted in house). We hrvested gliom ells to onentrtion of ~2, ells per µl. Mirolenses. For tumor imging we generlly used 1.-mm-dimeter GRIN lenses (.48 numeril perture,.45 pith, 3 µm working distne in wter) or.5-mm-dimeter GRIN lenses (.45 numeril perture,.92 pith, 3 µm working distne). For neuronl imging, we lso used 1-mm-dimeter doulet miroendosopes with.49 numeril perture, 25 µm working distne,.218 pith GRIN ojetive lens fused to 1.75 pith,.2 numeril perture GRIN rely. Guide tues. For 1-mm-dimeter miroendosopes, guide tue pillries were 1.6-mm-inner-dimeter, 1.8-mm-outer-dimeter nd 3 mm long; for.5-mm-dimeter miro-optis, pillries were.6-mm-inner-dimeter,.84-mm-outer-dimeter nd 3 mm long. We ut nd polished the pillries nd lened them y sonition. We then seled them t one end with over slip (Gold Sel #; ut to mth the pillries outer dimensions) with ultrviolet-uring epoxy (Norlnd Optil Adhesive 81) to llow optil ut not physil ess to tissue. We fshioned wshers (~5 µm thik) in silione Elstosil RTV 625, ured them etween two ryli pltes nd pled one wsher round eh pillry. Alterntively, we mde the wshers in Kwik-Sil nd died them to size. Surgery. We typilly implnted two guide tues t symmetri oordintes in opposing hemispheres (2. mm posterior to regm, 2. mm lterl for hippompus; 1. mm nterior to regm, 1.8 mm lterl for stritum). We positioned the guide tue windows just dorsl to, ut not within, CA1 hippompus or dorsl stritum. The Supplementry Methods ontin detils. Imging sessions. We generlly imged mie for multiple 3 6-min sessions under nesthesi with isoflurne ( % in O 2 ) or intrperitonel ketmine (75 mg per kg ody weight) nd xylzine ( mg per kg ody weight). We mounted the mie on stereotxi frme suh tht the pillry guide tue ws prllel to the optil xis. We lened the pillry s inside to remove deris. We then inserted miroendosopes, mthed to the inner pillry dimeter, up to the guide tue window. Osionlly, we used metl sheth fixed to the side of the miroendosope to id repetle positioning within the pillry. For gliom studies, we imged mie every 2 or every 3 d, from dys 1 to 24 fter surgery. For neurl imging, in erly studies we imged mie t irregulr intervls up to 396 d; in lter studies used for quntittive nlysis of dendrite turnover, we generlly exmined mie every 4 d over 7 weeks. Optil instrumenttion. Miroendosopy involved previously desried instrumenttion 13,. In rief, for one-photon miroendosopy we used ustom instrument with Hg-r lmp illumintion 13. To deliver illumintion into the miroendosopes, we used either 1,.25 numeril-perture mirosope ojetive (Zeiss, Ahrostigmt) or 2,.4 numeril-perture ojetive (Olympus, LMPlnFl) for the singlet mirolenses nd the 1,.25 numeril-perture ojetive (Zeiss, Ahrostigmt) for the doulet miroendosopes. High-speed eletron-multiplying hrge-oupled devie mers ptured videos of lood flow t frme rtes up to 1.2 khz, ut typilly 1 Hz. We used Csde 128 (Roper Sientifi) mer nd n ixon DU-897E (Andor) mer with the 1-mm- nd.5-mm-dimeter miroendosopes, respetively. A ooled hrge-oupled devie mer (Coolsnp HQ, Roper Sientifi) ptured higher-resolution imges. In dul-olor studies, dihroi module (Dul-View, Optil Insights) projeted two imges in distint olors onto one mer. For two-photon miroendosopy we used modified ommeril twophoton mirosope (Pririe Tehnologies) nd ustom two-photon mirosope 13,, oth dpted to ommodte the use of miroendosopy 17 nd equipped with tunle Ti:Spphire lser (Mi-Ti, Spetr-Physis). The lser ws tuned to 92 nm. We djusted verge power t the smple for onsisteny in signl strength ross imging sessions in eh niml. Typil powers t the speimen surfe were 5 18 mw, ut lwys <25 mw. To ouple illumintion into the miroendosopes, we used 1,.25 numeril-perture ojetive (Olympus, Pln N) in gliom studies or 2,.5 numeril-perture ojetive (Crl Zeiss, EC Epipln Neoflur) in stritl nd CA1 dendrite stility studies. In few sessions we used 1-mm-dimeter singlet GRIN miroendosope with n dditionl hemispheri lens t the tip 4, for inresed numeril perture nd enhned resolution over redued field of view (Fig. 2d). Fluoresene ngiogrphy. To lel the vsulture, we intrvenously injeted fluoresent dye (1 3 µl t 1 mg ml 1 onentrtion), either rhodmine-dextrn (Moleulr Proes; 7, moleulr weight, neutrl) or fluoresein-isothioynte dextrn (Sigm-Aldrih FD2S; 2,, moleulr weight), into the til vein. Anlysis softwre. We performed imge nlysis using Mtl (MthWorks) nd sttistil nlysis using Origin Pro (OriginL). Detils re in the Supplementry Methods. Additionl methods. Detiled methodology is desried in the Supplementry Methods. 4. Brretto, R.P., Messershmidt, B. & Shnitzer, M.J. In vivo fluoresene imging with high-resolution mirolenses. Nt. Methods 6, (29). doi:1.138/nm.2292 nture mediine