a cell-based screen for inhibitors of flagella-driven ...chuang/pdf/20504_ftp.pdfa cell-based screen...

A Cell-Based Screen for Inhibitors of Flagella-DrivenMotility in Chlamydomonas Reveals a Novel Modulator ofCiliary Length and Retrograde Actin Flow

Benjamin D. Engel,1* Hiroaki Ishikawa,1 Jessica L. Feldman,1 Christopher W. Wilson,2

Pao-Tien Chuang,2 June Snedecor,3 Janice Williams,3 Zhaoxia Sun,4 and Wallace F. Marshall11Department of Biochemistry & Biophysics, University of California, San Francisco2Cardiovascular Research Institute, University of California, San Francisco3Small Molecule Discovery Center, University of California, San Francisco4Department of Genetics, Yale University School of Medicine, New Haven, Connecticut

Received 2 December 2010; Accepted 27 January 2011Monitoring Editor: Ritsu Kamiya

Cilia are motile and sensory organelles with criticalroles in physiology. Ciliary defects can cause numeroushuman disease symptoms including polycystic kidneys,hydrocephalus, and retinal degeneration. Despite theimportance of these organelles, their assembly andfunction is not fully understood. The unicellular greenalga Chlamydomonas reinhardtii has many advantagesas a model system for studies of ciliary assembly andfunction. Here we describe our initial efforts to build achemical-biology toolkit to augment the genetic toolsavailable for studying cilia in this organism, with thegoal of being able to reversibly perturb ciliary functionon a rapid time-scale compared to that available withtraditional genetic methods. We screened a set of 5520compounds from which we identified four candidatecompounds with reproducible effects on flagella atnontoxic doses. Three of these compounds resulted inflagellar paralysis and one induced flagellar shortening ina reversible and dose-dependent fashion, accompanied bya reduction in the speed of intraflagellar transport. Thislatter compound also reduced the length of cilia in mam-malian cells, hence we named the compound ‘‘ciliabrevin’’due to its ability to shorten cilia. This compound alsorobustly and reversibly inhibited microtubule movementand retrograde actin flow in Drosophila S2 cells. Ciliabre-vin may prove especially useful for the study of retro-grade actin flow at the leading edge of cells, as it slowsthe retrograde flow in a tunable dose-dependent fashion

until flow completely stops at high concentrations, andthese effects are quickly reversed upon washout of thedrug. VC 2011 Wiley-Liss, Inc.

KeyWords: ciliogenesis, flagellar length, IFT, actin,mouse, zebrafish, Chlamydomonas, IMCD3, high throughput

screen, HTS

Introduction

Cilia and flagella are motile and sensory organellesfound in most cells of the body [Sloboda and Rose-

nbaum, 2007; Satir et al., 2010]. Defects in cilia underliea diverse set of human diseases known collectively as theciliopathies, including Polycystic Kidney Disease, Immo-tile Cilia Syndrome, Joubert Syndrome, Meckel Syn-drome, and Bardet-Biedl Syndrome [Afzelius, 2004;Badano et al., 2006; Fliegauf et al., 2007; Marshall 2008].There is currently an outstanding need for pharmacologi-cal treatments for these diseases. Cilia and flagella, namesthat can be used interchangeably, are composed of ninemicrotubule doublets surrounded by a specialized mem-brane. Although the protein components of cilia havebeen enumerated in recent years [Pazour et al., 2005;Gherman et al., 2006; Inglis et al., 2006], the molecularpathways that coordinate the assembly and function ofthis complex organelle remain poorly understood and arecurrently a topic of intense investigation [Pedersen et al.,2008; Santos and Reiter, 2008]. Most analyses of ciliathus far have relied on genetics, imaging, or biochemicalapproaches. A chemical biology approach, using smallmolecule modulators of ciliary assembly or function,would provide an orthogonal set of tools for probing cili-ary biology. As the use of small molecule inhibitors is awell established strategy for studying the cytoskeleton[Peterson and Mitchison, 2002], we decided to extend

Additional Supporting Information may be found in the onlineversion of this article.

*Address correspondence to: Benjamin D. Engel, Wallace F.Marshall, Department of Biochemistry and Biophysics,GH-N372F, Genentech Hall, Box 2200, UCSF, 600 16th St.,San Francisco, CA 94158. E-mail: [email protected]

Published online 18 February 2011 in Wiley Online Library(wileyonlinelibrary.com).

RESEARCH ARTICLECytoskeleton, March 2011 68:188–203 (doi: 10.1002/cm.20504)VC 2011 Wiley-Liss, Inc.

n 188

this strategy to cilia. One major advantage of a chemicalapproach compared to genetics is that a much higherdegree of temporal control can be attained than with con-ditional mutations. Such an approach, however, requires atoolbox of small molecules affecting cilia. While a fewsuch compounds have been found in tests of functionallypreselected molecules [Nakamura et al., 1987; Tuxhornet al., 1998; Wilson and Lefebvre, 2004; Ou et al., 2009;Besschetnova et al., 2010], and in an assay for modulatorsof hedgehog signaling [Hyman, 2009], a systematic searchfor compounds targeting cilia using direct assays for ciliaryassembly or function has not been reported.Molecules targeting cilia would not only provide useful

tools for basic studies of ciliary biology, but could alsoserve as starting points for pharmacological treatments ofcilia-related diseases [Afzelius, 2004]. Moreover, severaltypes of tumors rely on hedgehog signaling and can betreated by hedgehog signaling inhibitors [Low and deSauvage, 2010]. Since cilia are required for hedgehog sig-naling [Huangfu and Anderson, 2005], inhibitors of cilio-genesis may be useful as chemotherapeutic agents forhedgehog-dependent tumors. We expect that cilia wouldmake good drug targets, since they protrude into theextracellular environment where they are directly exposedto dissolved compounds, they lack export pumps anddetoxifying enzymes found in the cell body [Pazour et al.,2005], and they contain numerous druggable targetsincluding kinases and channels [Beck and Uhl, 1994; Ber-man et al., 2003; Pan and Snell, 2004; Qin et al., 2005;Wang et al., 2006; Qin et al., 2007]. Moreover, many cili-ated tissues in the human body are physically accessible todirect forms of drug delivery, for example via injectioninto the retina [Gaudana et al., 2010] or aerosol-mediateddelivery to the airway [Kleinstreuer et al., 2008]. Hence,the pharmacodyamic barriers to treatment of ciliopathiesaffecting these tissues may be lower than many otherhuman diseases that involve deep internal tissues ororgans. To begin exploring chemical strategies, we need toimplement high throughput screens for compounds target-ing cilia. One approach is to use mammalian culture cellsand employ imaging-based screening to directly measurecilia. However, this approach may be complicated by thehigh sensitivity of the degree of ciliation of tissue culturecells to variation in proliferation state, confluence, andmetabolic conditions. An alternative approach is to takeadvantage of the highly consistent levels of ciliation typi-cally seen in free-living unicellular model organisms.We have previously reported a simple cell-based assay

for ciliary function suitable for small molecule screeningin a 96 well plate format [Marshall, 2009]. This assayexploits the unicellular green alga Chlamydomonas rein-hardtii [Merchant et al., 2007], which swims through liq-uid media at a speed of 100–200 lm/sec using its twoflagella, which are structurally and molecularly equivalentto the cilia of animal cells. Much of our present molecular

understanding of ciliary assembly and motility has comefrom experiments first performed in Chlamydomonas.Examples include the discovery of intraflagellar transport[Kozminski et al., 1993; Cole et al., 1998] and the identi-fication of the radial spoke complex [Yang et al., 2006].Although Chlamydomonas is routinely used for geneticscreens [Silflow and Lefebvre, 2001], chemical screens forflagella-affecting compounds in this organism have notbeen reported. Although the Chlamydomonas cell is sur-rounded by a strong cell wall that might impede drugentry into the cytoplasm, the flagella themselves protrudefrom the cell wall and should therefore be fully accessibleto any membrane-permeable small molecules in thegrowth media.Our plate imaging-based motility assay exploits the

tendency of certain wild-type strains of Chlamydomonas todive rapidly to the bottom of a container. This movementrequires active cell motility [Marshall, 2009]. It is not cur-rently known if this active diving to the bottom of thecontainer is a form of gravitaxis driven by downwarddirected swimming, a form of negative phototaxis awayfrom light that is more intense above versus below theplate, or some type of bioconvection [Pedley et al., 1988].For the purposes of our assay, we are less concerned withthe mechanism of diving than with the fact that it ishighly robust and reproducible. This is a strain-dependentbehavior such that some wild-type strains will dive to thebottom of a container while other strains do not. Cellsfrom diving strains grown in liquid culture will dive tothe center of a U-bottom well, which is the well’s lowestpoint. Importantly, cells that cannot swim are not able tomove to the bottom and remain dispersed throughout theentire well [Marshall, 2009]. Cellular distribution in thewells is easily visualized thanks to the green color pro-duced by chlorophyll in the chloroplasts. Motile cells pro-duce a dark green spot in the center of the well, whilenonmotile cells form a larger, light green circle with diam-eter equal to the well diameter (Fig. 1A).In this report, we employed this high throughput motil-

ity assay to screen a diverse set of drug-like small moleculesand identified four that affect Chlamydomonas flagella. Wenamed one of these compounds ‘‘ciliabrevin’’ because italso shortens the length of cilia in mammalian cells. Inaddition to modulating ciliary length and intraflagellartransport speed and frequency, ciliabrevin also inhibits ret-rograde actin flow in Drosophila S2 cells, hinting at a possi-ble link between actin dynamics and ciliogenesis.

Materials and Methods

Plate-Well Image Analysis Assay

We used wild-type strain cc-124 obtained from the Chla-mydomonas Genetic Center (www.chlamy.org), which wechose because it shows robust diving. Another wild-type

CYTOSKELETON Inhibitors of Chlamydomonas Flagella 189 n

strain of the opposite mating type, cc-125, does not ex-hibit this diving behavior, so in setting up this screen it iscritical to make sure the starting strain reliably dives.In our screen, cells were grown in TAP media to an

OD600 of 1.0 under continuous illumination in 200 mlflasks on a floor shaker, and then diluted to an OD600 of0.03 with fresh TAP media. One-hundred and ninetymicroliters of diluted cells were dispensed into wells ofCorning Costar 96-well U-bottom clear polystyrene platesusing a Wellmate dispenser system. Before adding cells,wells were preloaded with compounds from a diverse col-lection of small molecules whose structures predict drug-like properties (ChemDiv, San Diego, CA), to yield a finalconcentration of 5 lM in 0.5% DMSO after addition ofthe cell media. Cell-based chemical screens are typicallyperformed in the 1–10 lM range. We chose a concentra-tion in the middle of this range to balance the desire toobtain compounds with low effective doses with potentialconcerns that the hit rate might become too low if thescreening concentration was further reduced. We did notperform any systematic tests to determine whether thisparticular concentration is optimal. The first and last col-umns of each plate were used for controls, the first col-

umn being a DMSO-only control and the last columnbeing loaded with a culture of a nonmotile mutant, bld1,that lacks flagella [Brazelton et al., 2001]. Plates wereallowed to grow under continuous light for 1 day andthen scanned with a flat-bed scanner (HP Scanjet 4070).Custom software was used to analyze the plate images

in a fully automated image processing pipeline, similar tothat previously described [Marshall, 2009]. Scanned RGBimages of each plate were processed to extract a series of96 subimages corresponding to each well based on posi-tion within the image. The green channel was convertedto black and white and the contrast inverted, so thatdarker pixels, indicating high density of green cells, wereassigned larger numerical values. The software then esti-mated the distribution of cells as an intensity weightedfirst moment:

M ¼

Px;y

Ix;y

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffix � xð Þ2þ y � yð Þ2

q

Px;y

Ix;y

Where x and y are the coordinates of each pixel, ‘‘I’’ isthe intensity at each pixel, and summations are taken over

Fig. 1. Screen characterization and results. (A) Image of a plate from the assay. In all plates, untreated wild-type cells were loaded onthe left column of eight wells in media containing 0.5% DMSO as a negative control and bld1 mutant cells lacking flagella were loadedon the right column of eight wells as a positive control. (B) Heat-map scoring of a typical plate. Color scale shown on inset color-bar,ranging from 4 (dark blue, indicating a small assay score characteristic of untreated cells) to 18 (red, indicating a high assay score charac-teristic of nonmotile cells). (C) Screening statistics. Red and green marks signify positive (bld1 mutants) and negative (untreated wild-typecells) controls, respectively. Gray circles signify unknown compounds. Red and green lines indicate three standard deviations above andbelow the mean assay scores for positive and negative controls, respectively.

n 190 Engel et al. CYTOSKELETON

all pixels in the well subimage. The resulting first momentserves as the assay score for that well, as depicted in theheat map of Fig. 1B. Potential hits can be detected aswells with a high value of ‘‘M.’’Because the calculation of the assay score ‘‘M’’ involves

normalization by the total intensity over the whole well,toxic compounds can yield high assay scores since cell deathcan lead to a low uniform greenish-yellow colorationthroughout the well. Hence, we implemented a built-incounter screen for toxicity by measuring the average celldensity in each well based on the average intensity of thewell images. Outliers with average density more than threestandard deviations below that measured in control wellswere scored as toxic and discarded from further analysis.

Secondary Assays in Chlamydomonas

For secondary assays, resupplied compounds were storedas stock compounds in 100% DMSO at a concentrationof 1 mg/ml, and then used at dilutions indicated in thetext and tables.Visual assessment of the presence or absence of flagella

(Table I) was performed by mounting 10 ll of live cellculture under a coverslip with a Vaseline ring and imagingon a Zeiss Axioskop using a 40� air lens and DIC optics(Carl Zeiss MicroImaging, LLC, Thornwood, NY). Swim-ming speeds (Table II) were calculated by adjusting the

condenser on a Zeiss Axioskop to give a dark-field likeimage, then acquiring 0.5 second exposures with a SPOTcamera (Diagnostic Instruments, Sterling Heights, MI).This produced streak-like images for swimming cells, andspeed was estimated by tracing the length of each streakusing the SPOT analysis software and dividing length bythe exposure time. Flagellar regeneration (Fig. 3B) wasperformed by transiently lowering the pH of liquid cul-tures as previously described [Lefebvre, 1995]. Detailedflagellar length measurements in fixed Chlamydomonascells (Figs. 3A and 3B) were attained by capturing imageson a Zeiss Axioskop equipped with a SPOT camera andmeasuring flagellar lengths with NIS-Elements AR soft-ware (version 3.2, Nikon).For TIRF measurement of IFT, cells expressing GFP-

tagged kinesin-2 motor subunit KAP [Mueller et al.,2005] were grown in TAP media under continuous illu-mination in test tubes on a roller drum. Cells weremounted and imaged as previously described [Engel et al.,2009a], and IFT speeds were measured by kymographanalysis as detailed previously [Engel et al., 2009b].

Mouse IMCD3 and MTEC Cell Culture Methods

Mouse IMCD3 cells were cultured in Dulbecco’s modifiedEagle’s medium and Ham’s F-12 medium mixture con-taining 10% heat-inactivated fetal bovine serum (FBS) at37�C in 5% CO2. Cells were cultured on acid-washedcoverslips, fixed with methanol for 5 min at �20�C, andwashed three times with phosphate-buffered saline (PBS).Cells were blocked in 1% bovine serum albumin in PBSfor 10 min at room temperature and incubated with pri-mary antibodies for 1 hour in a humid chamber. Cellswere then washed with PBS and incubated with FITC,TRITC and Cy5 conjugated secondary antibodies (Jack-son ImmunoResearch laboratories, West Grove, PA) for30 min. Samples were washed with PBS and mountedwith Vectashield (Vector Laboratory, Burlingame, CA),then observed using a DeltaVision microscope (AppliedPrecision, Issaquah, WA) with a 60� objective (Olympus,

Table I. Dose Response for CandidateCompounds

CompoundConcentration

(lM)Normal(%)

Paralyzed(%)

Lacking(%)

B 3 83 17 0

6 75 25 0

15 80 5 15

30 24 15 61

E 0.6 0 100 0

1.5 0 100 0

3 0 76 24

6 0 51 49

15 3 20 77

30 0 7 93

F 0.6 38 57 5

1.5 11 86 1

3 1 98 1

6 1 75 24

P 0.6 65 35 0

1.5 15 85 0

3 4 96 0

Cells were grown for 4 hours with the indicated dose of compoundand then examined by differential interference contrast (DIC)microscopy and scored for percent of cells having normal beatingflagella, paralyzed flagella, or lacking flagella.

Table II. Swimming Speed VersusConcentration of Compounds E and F

CompoundConcentration

(lM)Swimming

Speed (lm/sec)

E 0.15 83.3 6 21.3

0.3 56.2 6 29.7

0.6 23.2 6 26.3

1.5 11.9 6 5.2

F 0.15 108.2 6 14.3

0.3 106.3 6 12.7

0.6 90.2 6 25.3

1.5 46.7 6 32.7


Tokyo, Japan). Z-stack images were obtained at 0.2-lmintervals, then deconvolved and projected with DeltaVi-sion software (Applied Precision).MTECs were derived and cultured essentially as described

[You et al., 2002]. Tracheae were excised from adult mice,cleaned of surrounding tissue, and digested with 1.5 mg/mlPronase (Roche) overnight at 4�C. The next day, FBS (Cell-gro) was added to neutralize the enzyme, and the suspendedtracheae were inverted to release cells. The supernatant wascentrifuged at 200 x g and the cell pellet was resuspendedin MTEC/Basic [You et al., 2002] media and plated for 4hours at 37�C to remove contaminating fibroblasts. The su-pernatant was again removed, spun, and resuspended inMTEC/Plus [You et al., 2002] media. 2.5 � 104 MTECswere plated on 0.4 lm collagen I (Roche) coated Transwellinserts (Corning) and media was added to the upper andlower chambers. After confluence was reached, media wasremoved from the upper chamber, and MTEC/NS [Youet al., 2002] was added to the lower chamber to create anair-liquid interface (ALI) and induce the ciliary differentia-tion program. After 7–10 days of culture, MTECs weretreated with compounds for 24 h, and were then imagedusing established procedures [Wilson et al., 2009].

Analysis of Cytoskeletal Dynamics in DrosophilaS2 Cells

Cells were grown at room temperature in Schneider’s Dro-sophila medium supplemented with 10% FBS. Before imag-ing, cells were plated on 14 mm glass bottom microwelldishes (P35G-1.5-14-C, MatTek, Ashland, MA) coated with10% Conconavilin A, and allowed to adhere for 1–2 hours.The time allowed for cell adhesion was important to ensurethat cells were well spread but still had roughly uniform circu-lar morphology. Drug addition was performed without inter-rupting movie acquisition by carefully adding 10 ll of a 10�dilution of drug in cell media to 90 ll cell media already cov-ering the cells. Washout was also performed without inter-rupting movie acquisition by carefully pipetting out thedrugged media and replacing it with 150 ll fresh media with-out touching the cell chamber. Towards the end of eachmovie, signal intensity decreased due to GFP photobleaching.Cells were imaged on a Nikon Eclipse Ti-E motorizedinverted spinning disc confocal microscope (Nikon, Tokyo,Japan), equipped with an integrated Nikon Perfect Focus Sys-tem, a 100�/1.40na oil Plan Apo VC objective, and a Photo-metrics Evolve EMCCD camera (Photometrics, Tucson, AZ).Four kymographs were generated from each movie (verti-

cal, horizontal, and two diagonal cross-sections) using a cus-tom-made ImageJ plugin (version 1.42 g, NationalInstitutes of Health) that plotted the maximum intensityalong each point of a straight 50-pixel wide line for eachmovie frame. The speed of retrograde microtubule and actinflux was calculated from the angle of traces moving inwardfrom the cell periphery (measured with ImageJ) and average

speeds for each movie were plotted in Figs. 5G and 6E.The plots of microtubule and actin position in Figs. 5Hand 6F were generated with a custom-made Matlab script(version R2007a, Mathworks), which plotted peak pixelintensities from a single kymograph trace and fitted this dis-tribution with a simple exponential decay trend line.

Results

Screening Collection of Diverse CompoundsYields Candidate Hits

Before screening compounds, we conducted control experi-ments using mutant cells, comparing assay scores for wild-type cells versus bld1 mutant cells [Brazelton et al., 2001] thatlack flagella (Fig. 1). We found that the coefficients of varia-tion of the assay scores were in the range 6–8% for wild-typecells and 2–4% for mutant cells. We characterized assay selec-tivity using the Z’-factor [Zhang et al., 1999], a measure ofthe ability of the assay to discriminate positive from negativecontrols. The Z’-factor ranges from 0 to 1, with 1 being a per-fect assay, 0 being a completely uninformative assay. A scoreover 0.5 is considered an ‘‘excellent assay’’ for cell-based highthroughput screening [An and Tolliday, 2010] on the groundsthat the assay is sufficiently selective to yield a high fraction oftrue hits without the assay becoming swamped with false posi-tives. For our assay, we found that Z’ was 0.72, using wild-type and bld1 mutants as the two control groups. This scorefalls into the ‘‘excellent’’ range of Z’ especially for cell-basedassays, which tend naturally to have higher variability thanassays with purified enzymes, and indicates the assay should beeffective at discriminating motile from nonmotile cells.We screened a diverse set of 5520 compounds (results

shown in Fig. 1C). The average assay scores (‘‘M’’ values,see materials and methods) were 4.5 6 0.7, 17.6 6 0.8,and 4.6 6 0.8 for the negative control wells, positive con-trol wells, and the 5520 unknown sample wells, respectively.To evaluate the efficacy of the screen, we calculated the Z-factor, another indicator of assay performance [Zhang et al.,1999], which differs from the Z’-factor in that rather thancomparing a positive control set to a negative control set, itcompares a positive control set to the results of a set ofdiverse compounds. Assays that are overly sensitive to per-turbation by small molecules would give a low Z-factor,indicating that the false hit rate is unacceptably high for usein a screen. Like Z’, Z factors range from 0 to 1, with 0meaning the screen cannot work and 1 indicating an idealscreen. We calculated a Z-factor of 0.63, demonstratingthat our assay is not overly sensitive to random small mole-cules. We also evaluated plate to plate variability by screen-ing the first 2000 compounds in duplicate. The correlationcoefficient for corresponding wells between the two sets ofplates was 0.98, indicating a low plate to plate variability.Out of the 5520 compounds screened, a total of 20

compounds were selected as potential hits because they


gave high assay scores and also gave high scores for celldensity, indicating a lack of toxicity (see Materials andMethods). In all cases, the original well images were visu-ally inspected and it was confirmed that each of these can-didate compounds caused cells to uniformly fill the wells,mimicking the pattern seen in motility-defective mutants(for example, see well E2 of the plate image in Fig. 1A).All 20 compounds continued to yield positive scores whenreassayed in fresh plates, confirming the reproducibility ofthe screening assay. Of these 20 compounds, sixteen wereavailable for resupply from the vendor, and all of thesecontinued to produce a positive assay score when theresupplied compounds were retested using the same assay.

Secondary Screening for Effects on Flagella

In principle there are many reasons why a Chlamydomonascell might not undergo the diving behavior characteristicof the cc-124 strain, some of which might be unrelated toflagella, such as changes in cell density, viability, or mem-brane excitability [Yoshimura et al., 2003]. We thereforetested the sixteen candidate compounds in a set of simplesecondary assays. In our initial screen we had excludedovertly toxic compounds where cells failed to grow withinone day. To test more carefully for subtler toxic effects, wegrew lower density cultures in the presence of the com-pounds for 4 days, providing a more sensitive assay foreffects on growth. Based on visual inspection of these cul-tures, we ruled out five compounds as having a noticeableeffect on growth rates.Next, we grew cultures for 3 days in the presence of

each compound at doses sufficient to completely blockdiving behavior, mounted live cells under coverslips, andchecked for presence of flagella using DIC microscopy.Live imaging allowed us to determine if the cells weremotile or not. From these rough visual analyses, we foundthat eight of the sixteen compounds showed a clear-cuteffect on flagella as judged by a majority of cells in theculture either lacking flagella or having paralyzed flagellathat failed to beat at concentrations that produced no vis-ually observable toxicity.One potential nonspecific effect that posed a possible

concern is the pH shock response. Following a sudden dropin pH from neutral down to pH 4-5, Chlamydomonas cellsundergo ‘‘flagellar autotomy,’’ in which cilia are severed atthe base and released into the media [Quarmby, 2004]. Torule out the possibility that some of our compounds weresufficiently acidic that they triggered the pH shock responseand therefore induced cilia loss for a trivial reason, we meas-ured the pH of solutions of each compound diluted in cellgrowth media. We found that none of the compoundsresulted in a detectable reduction in pH, with all culturesincluding untreated controls having pH 7.3, the usual pHof the TAP media used to grow the cells.

We next examined the homogeneity of the effect on apopulation of cells by microscopy-based measurement offlagellar length and motility at a range of doses. We foundthat four of the eight compounds had nonhomogenouseffects, that is, at any nontoxic dose assayed, only a subsetof cells showed a defect in the length or motility of theflagella, with others looking normal. Such compounds,which only affected a fraction of the total cell population,were deemed undesirable for follow-up because the hetero-geneous response would not only greatly complicate fur-ther efforts to understand the nature of the effect, itwould also limit the utility of the compounds as experi-mental tools for studying flagella. These four compoundswere thus excluded from further consideration. Such inho-mogeneous effects had not been noticed during the initialscreen, apparently because the nonmotile cells were suffi-ciently numerous to provide a uniform green pattern fill-ing the well, thus leading to a high assay score.These secondary screening assays thus narrowed the

potential hits to four compounds that showed homoge-nous effects on flagella at nontoxic doses (Fig. 2). Of thesefour, our preliminary visual survey indicated that one(compound B) affected the length of the flagella, whilethe other three (compounds E, F, and P) produced paraly-sis of flagellar motility. We named compound B ‘‘ciliabre-vin,’’ reflecting its ability to shorten cilia and flagella.The four candidate compounds were tested over a range

of concentrations and scored for the fraction of cells thathad motile flagella, paralyzed flagella, or no flagella after 4hours of growth in the presence of compound. As indi-cated in Table I, all four compounds showed activity inthe micromolar range of concentrations.

Ciliabrevin Induces Flagellar Shortening andLoss

We noted that treatment with 30 lM ciliabrevin (com-pound B) predominantly caused loss of flagella in theoriginal assay, which involved 1–4 days of growth in thepresence of the compound. If ciliabrevin really requiredmultiple days to act, it would be far less useful as a toolfor studies of flagellar dynamics than if its speed of actionwere relatively fast compared to cellular processes such asthe cell cycle. Therefore, we monitored the flagellar phe-notype as a function of time. We added ciliabrevin at arange of concentrations and examined cells by DIC mi-croscopy at regular intervals (Fig. 3A). No immediateeffect was seen after adding the compound, but by 30–60minutes all cells that retained flagella had shortened theirflagella depending on the dose of ciliabrevin added. Theseshorter steady-state flagellar lengths were then stably main-tained for 12 hours when cells were left in the presence ofthe compound. While 25 lM ciliabrevin produced cellswith half-length flagella, higher concentrations of thecompound did not shorten flagella further, but rather lead


to flagellar loss and visible signs of toxicity, including celllysis. In separate experiments, flagella returned to normallength within 2 hours following ciliabrevin wash out (datanot shown). The time-scale of ciliabrevin action (� 1hour) is thus substantially shorter than the time scale ofthe Chlamydomonas cell division cycle (� 8 hours). Thefact that the flagella eventually stop shortening and remainconstant at approximately half-length and no shortercould indicate a difference between the proximal and dis-tal halves of the flagellum. Dentler reported that colchi-cine induces shortening of the distal but not proximalportions of the flagellum [Dentler and Adams, 1992],while Piperno has shown that the inner dynein arm pro-tein composition of the proximal and distal halves of theflagellum are distinct [Piperno and Ramanis, 1991], andYagi demonstrated that the proximal portion of the flagel-lum possesses novel low-abundance dynein heavy chains[Yagi et al., 2009]. It is thus possible that whatever thetarget of ciliabrevin is, it has a more direct influence ofdynamics of the distal half of the flagellum.Ciliabrevin also affected the kinetics of flagellar regener-

ation following flagellar excision by pH shock [Lefebvre,1995]. We found that 25 lM ciliabrevin caused flagella togrow more slowly and reach a shorter steady-state lengthof �8 lm (Fig. 3B, red line), similar to the stable lengththat flagella shorten to when treated with the same con-centration of the compound (Fig. 3A, red line). Thiseffect on the initial growth rate is distinct from the effectof protein synthesis inhibitors, which cause flagella to

regenerate to a shorter length but have no effect on theinitial growth rate [Rosenbaum et al., 1969].These results show that ciliabrevin induces a dynamic

dose-dependent shortening of flagella, which is accompa-nied by flagellar loss in a fraction of cells (see Table I).Furthermore, flagellar regeneration in the presence of cil-iabrevin proceeds more slowly and yields shorter flagella.Based on these observations, the balance-point model offlagellar length control [Marshall et al., 2005; Engel et al.,2009b] predicts that ciliabrevin either deceases the assem-bly rate or increases the disassembly rate at the flagellartip, as both shortening and regenerating flagella approachthe same shorter steady-state length.

Effect of Ciliabrevin on Intraflagellar Transport

The combined shortening and loss of flagella upon treat-ment with ciliabrevin is reminiscent of the phenotypeobserved in a specific class of Chlamydomonas mutants.These mutants are defective in the process of anterogradeintraflagellar transport (IFT), the kinesin-powered trans-port of flagellar proteins from the cell body to the site ofassembly at the flagellar tip [Cole, 2003; Scholey, 2003;Pedersen and Rosenbaum, 2008]. Anterograde IFT is alsorequired to maintain flagella after assembly, and whenconditional mutants in this process are shifted to the re-strictive temperature, their flagella undergo shortening[Kozminski et al., 1995]. The reported shortening ratesseen in IFT conditional mutants following shift to

Fig. 2. Structures of candidate compounds obtained in the initial screen and validated in secondary assays to confirm homog-enous effects at nontoxic doses.


restrictive temperature [Kozminski et al., 1995] are similarto the shortening rates that we observed here.When conditional mutants in IFT are shifted to restric-

tive temperature, not only do the flagella shorten they alsobegin to detach from the cell [Adams et al., 1982; Parkerand Quarmby, 2003]. This flagellar detachment utilizesthe same flagellar autotomy mechanism that causes flagel-lar detachment in response to pH shock [Parker andQuarmby, 2003], which is mediated by a physical severingof the flagella away from the cell at the base of the flagel-lum. The fa1 mutation inhibits flagellar autotomy [Finstet al., 1998] and prevents conditional IFT mutants from

detaching their flagella when IFT is turned off [Parkerand Quarmby, 2003]. We therefore asked whether theshedding of flagella induced by ciliabrevin would likewiserequire the flagellar autotomy system. We treated fa1 mu-tant cells with 30 lM ciliabrevin for 24 hours. While thisconcentration was sufficient to cause complete flagellarloss in wild-type cells, 82% of fa1 mutant retained flagellaafter 24 hours of treatment. In this experiment we didnot measure flagellar lengths, but only scored their pres-ence or absence. We also tested adf1 mutants [Finst et al.,1998], which block flagellar autotomy at an upstream sig-naling point but, unlike fa1, do not inhibit flagellar

Fig. 3. Effect of ciliabrevin (compound B) on Chlamydomonas flagellar length and intraflagellar transport. (A) Flagellar lengthversus time after addition of ciliabrevin to wild-type cc-125 cells at a variety of concentrations. N 5 3698 flagella for all concentra-tions and timepoints combined. (B) Regeneration kinetics of wild-type cc-125 flagella following pH shock in the presence (red) orabsence (blue, adapted from Engel et al., 2009b) of 25 lM ciliabrevin. N ¼ 2464 flagella with ciliabrevin and 2178 flagella withoutdrug. (C) Kymograph of a KAP-GFP cell imaged by TIRF microscopy, showing anterograde IFT traces 1 hour after addition of 34lM ciliabrevin. Horizontal scale bar ¼ 2 lm, vertical scale bar ¼ 2 seconds. (D) Speed and (E) frequency of anterograde KAP-GFPtraffic during pH shock regeneration (blue, adapted from Engel et al., 2009b) and 34 lM ciliabrevin-induced shortening (red),binned by flagellar length. Error bars in all panels indicate 99% confidence intervals. N ¼ 94 flagella with ciliabrevin and 101 flag-ella without drug.


detachment following cessation of IFT (Jeremy Parker andLynne Quarmby, personal communication). Unlike therescue seen with fa1, 97% of adf1 mutant cells still losttheir flagella following treatment with 30 lM ciliabrevin.Thus, in terms of the induced flagellar shortening rate,the shedding of flagella, and interactions with the fa1 andadf1 mutants, ciliabrevin treatment appears to mimic theinhibition of IFT.Based on the apparent similarities between ciliabrevin

treatment and IFT loss, we imaged and quantified IFT inthe flagella of living cells expressing a GFP-tagged subunitof the kinesin-2 anterograde motor using total internalreflection fluorescence (TIRF) microscopy (Fig. 3C; Sup-porting Information Movie S1; method described in Engelet al., 2009a,b). As plotted in Figs. 3D and 3E, we foundthat IFT was not blocked in cells treated with 34 lM cil-iabrevin. However, the frequency and speed of anterogradeIFT experienced a moderate decrease in full-length flagellathat had begun to shorten following drug addition. Con-sidering just full-length flagella, the average frequency ofIFT particles passing any given point of the flagellumdropped from 1.26 6 0.18 per second (N ¼ 16 flagella)to 1.08 6 0.20 per second (N ¼ 28 flagella), and thespeed of the particles dropped from 2.36 6 0.18 lm/secto 1.87 6 0.23 lm/sec (6 denotes standard deviation).These differences are statistically significant (P < 0.003, t¼ 3.14 and P < 0.0001, t ¼ 8.55), indicating that34 lm ciliabrevin decreases both the frequency and speedof intraflagellar transport in full-length flagella. Becausewe used the KAP subunit as the marker for IFT, whichdoes not clearly reveal retrograde IFT traces [Engel et al.,2009b], we were not able to determine whether or not cil-iabrevin affects retrograde IFT or only anterograde. Wehave previously shown that anterograde IFT speed and fre-quency vary moderately as a function of flagellar length inregenerating untreated cells [Engel et al., 2009b], so onepossible explanation is that the reduction in IFT speedand frequency are not a direct effect of ciliabrevin butrather an indirect effect of the flagellar shortening inducedby the compound. However this observed correlationbetween IFT and flagellar length would not explain thefact that IFT speed and frequency were reduced in full-length ciliabrevin-treated flagella before they had under-gone shortening.It is interesting to note that the anterograde IFT speed

in full-length ciliabrevin-treated flagella was reduced toexactly the speed that would normally be seen in regener-ating flagella that are 6–8 lm in length (Fig. 3D). Thiscorresponds to the length that flagella would ultimatelyattain following ciliabrevin treatment, suggesting that thecompound may alter the molecular regulatory pathwaysthat control IFT as a function of flagellar length, adjustingthe system to a new length set-point. However, becausethe exact role of IFT in the maintenance of flagellarlength is still not fully understood [Marshall et al., 2005;

Wemmer and Marshall, 2007], at this point we are notable to conclude whether the shortening and loss of flag-ella caused by ciliabrevin can be accounted for solely bythe observed reduction in IFT speed and frequency.

Characterization of Flagella-ParalyzingCompounds

In contrast to ciliabrevin, the other three remaining com-pounds obtained from our screen do not affect flagellarlength, but rather seem to act by paralyzing flagellar mo-tility. To quantify this effect, we treated cells for 1 hourwith compounds E and F at a range of concentrationslisted in Table II and measured swimming speeds asdescribed in the Materials and Methods. As shown in Ta-ble II, both compounds show a dose-dependent reductionin swimming speed in the micromolar to submicromolarconcentration range, confirming that these compounds doindeed result in paralysis of flagella-driven motility.

Effects of Compounds in Other Cell Types

We next tested whether any of the four compounds thatwe characterized in Chlamydomonas would have efficacy incell types from other species. The ciliate Paramecium aure-lia is a unicellular protozoan that swims using hundredsof cilia on its cell surface. We added the compounds at aconcentration of 34 lM to cultures of Paramecium andexamined the cells by microscopy after 1 hour and after 1day. After 1 hour, ciliabrevin induced lysis of all cells inthe culture. Compounds E, F, and P had no effect after 1hour, nor did a DMSO control. In contrast, after 1 dayof incubation with the compounds, the fraction of cellsthat were swimming dropped from 96% in controls to37% in cells treated with compound E and 30% in cul-tures treated with compound F. In contrast, the fractionof swimming cells was 95% in cultures treated with com-pound P. We conclude that compounds E and F paralyzecilia-driven swimming in Paramecium while ciliabrevincauses rapid cell lysis and compound P produces no dis-cernable effect.We next tested the compounds on cultured mammalian

cells. We added ciliabrevin to cultured mouse IMCD3cells at a concentration of 14 lM for 3 days and thenvisualized cilia by immunofluorescence (Fig. 4). We foundthat ciliabrevin caused cilia to become roughly 25%shorter, comparable to the effect seen in Chlamydomonaswith a similar concentration of the compound.Because the cilia of IMCD3 cells are normally nonmo-

tile we did not test the other compounds in this cell type.As an alternative cell culture system with motile cilia, wetested the effects of the compounds on cultured mousetracheal epithelial cells (MTECs), which are covered witharrays of beating cilia. Ciliabrevin reduced the fraction ofcells with motile cilia from 45% down to 22% after 16hours of treatment, while compounds E, F, and P had


comparatively less effect, with fraction of cells with motilecilia after 16 hours measured as 28, 29, and 46%, respec-tively. After 40 hours of treatment, the fraction of cellswith motile cilia dropped from 54% down to 20% afterciliabrevin treatment, while the fraction only dropped to45% after 40 hours treatment with compound E, and incells treated with compounds F, or P, the fraction of cellswith motile cilia actually increased slightly. We concludethat ciliabrevin inhibits cilia formation in cultured mam-malian cells, but that compounds E, F, and P have onlyminor effects, if any.To determine if any of these compounds could be used

to study cilia function in animal development, we treatedzebrafish embryos by adding the compounds at a finalconcentration of 34 lM to the embryo media. We foundthat ciliabrevin as well as compounds E and F causedembryos to die before hatching, while compound P hadno effect on development (data not shown). No obviousphenotypes attributable to cilia defects (e.g., kidney cysts)were detected in ciliabrevin treated fish other than an

overall body curvature sometimes associated with ciliarydefects (data not shown). At higher dosage, ciliabrevintreatment led to shortened body axis and embryoniclethality yet direct examination of embryos treated withciliabrevin at various stages of development showed thatcilia were still present in the embryos. We conclude thatciliabrevin has a broader impact on zebrafish developmentthan affecting cilia alone, although we did not directlymeasure ciliary motility, leaving open the possibility thatmotility could have been affected.

Ciliabrevin Reversibly Arrests CytoskeletalDynamics in Cultured Cells

Because ciliabrevin shortens cilia, which are microtubule-based organelles, we decided to examine whether the com-pound affects the dynamics of cytoplasmic microtubulesby direct microscopic observation of live Drosophila S2cells expressing tubulin-GFP. Treatment with 34 lM cil-iabrevin caused a dramatic cessation of microtubule

Fig. 4. Effect of ciliabrevin (compound B) on mammalian cell ciliary length. (A) Control mouse IMCD3 cells and (B) IMCD3cells treated with 14 lM ciliabrevin, showing reduced cilia length. Cells were stained for the cilia marker acetylated tubulin (green),the cell junction marker ZO-1 (red), and the centrosome marker pericentrin (magenta). Scale bar ¼ 10 lm. (C) Average length ofcilia in treated and control IMCD3 cells. Error bars indicate 99% confidence intervals. (D) Ciliary length distribution in treated(red) and control (blue) cells. N ¼ 931 control cilia and 562 cilia in presence of drug.


movements, abrogating both microtubule bending in thecell interior and the normal centripetal motion of freemicrotubule plus ends away from the cell periphery (Fig.5A; Supporting Information MovieS2). Microtubuleslengthened for several minutes immediately after drugaddition, then shortened to stable lengths (Fig. 5A kymo-graph). The effects of ciliabrevin were reversible, as micro-tubule movements were restored within 10 minutesfollowing washout with fresh media (data not shown).In contrast, 1 lM taxol caused free tubulin to form

small aggregates and microtubules to stiffen and break asthey were pushed towards the center of the cell byinwardly-directed forces (Fig. 5B; Supporting InformationMovie S3). Pretreatment of cells with increasing concen-trations of ciliabrevin slowed the centripetal movement ofmicrotubules and tubulin aggregates following taxol addi-tion (Figs. 5C–5E; Supporting Information Movies S4–S6; quantified in Fig. 5G). When 34 lM ciliabrevin wasadded before taxol, microtubules were not pushed inwardsat all, and instead appeared to thicken and acquire bul-bous ends (Fig. 5E; Supporting Information Movie S6).The retrograde motion of microtubules in these cells wasrestored following washout with one change of freshmedia (Fig. 5F; Supporting Information Movie S7), albeitat slower speeds due either to incomplete restoration ofinwardly-directed forces or to strengthening of the thickbulbous microtubules. When 1 lM taxol was addedbefore 34 lM ciliabrevin had stopped microtubulemotion, the velocity of the inward movement of microtu-bule plus-ends from the cell periphery decayed over timeas a simple exponential (Fig. 5H; Supporting InformationMovie S8), suggesting that ciliabrevin may reduce micro-tubule speed in a series of random independent eventsthat irreversibly inactivate individual force-generating unitsin a stochastic manner.One of the primary mechanical forces that moves

microtubules in S2 cells is the retrograde flow of corticalactin from the cell periphery [Rodriguez et al., 2003]. Toaddress whether ciliabrevin affects actin dynamics, weimaged S2 cells expressing actin-GFP and Arp2-mCherrywhile adding increasing concentrations of the drug (Fig.6). Even at lower 7 lM and 17 lM concentrations, ciliab-revin arrested retrograde actin flow (Figs. 6A and 6B; Sup-porting Information Movies S9 and S10). At 34 lM,ciliabrevin not only inhibited retrograde flow, but alsocaused round actin-GFP aggregates to accumulate at thecenter of the cell (Fig. 6C; Supporting Information MovieS11). Ciliabrevin began to act immediately after additionand partially or completely inhibited retrograde actin flowin a dose-dependent fashion, reaching full effect within 5–10 minutes (Fig. 5E). These effects were reversible, as 10minutes following washout of 34 lM ciliabrevin with onechange of fresh media, actin-GFP aggregates disappearedand retrograde flow was restored (Fig. 6D; Supporting In-formation Movie S12). Like the cessation of microtubule

movements in Fig. 5H, the retrograde flow of Actin-GFPalso slowed exponentially after drug addition, suggesting amechanism of ciliabrevin action involving multiple inde-pendent random events (Fig. 6F; Supporting InformationMovie S11). Since retrograde actin flow involves a com-plex interplay between myosin motors and pathways regu-lating actin polymerization and depolymerization[reviewed in Welch et al., 1997], it is not possible to guessfrom these results what molecular element of the retro-grade flow machinery might be affected by ciliabrevin.The phenotype of ciliabrevin treatment shared some simi-

larities with the actin-interfering drugs phalloidin, latruncu-lin B, and jasplakinolide, yet had unique characteristics of itsown (Supporting Information Fig. S1). Like ciliabrevin,phalloidin removed actin-GFP from the cell periphery andproduced actin-GFP aggregates within the cytoplasm (Sup-porting Information Fig. S1A and Movie S14). However,the phalloidin actin-GFP aggregates were more filamentousand less centrally located compared to ciliabrevin actin-GFPaggregates. Furthermore, phalloidin did not stop retrogradeflux, as small particles were still seen moving inward fromthe cell periphery. The effects of latrunculin B and jasplaki-nolide were even more distinct from ciliabrevin. LatrunculinB did stop retrograde actin flow, but left small actin-GFPaggregates throughout the cytoplasm and around the cell pe-riphery (Supporting Information Fig. S1C and Movie S16).Jasplakinolide produced a rapid increase in Actin-GFP fluo-rescence throughout the cytoplasm, which was pushedtowards the cell center, where much of the fluorescence dis-appeared and the rest remained as bright Actin-GFP aggre-gates (Supporting Information Fig. S1E and Movie S18).Drug washout with one change of fresh media produced atransient increase of cytoplasmic Arp2-mCherry followed bythe restoration of actin-GFP dynamics in cells treated withciliabrevin (Fig. 6D, Supporting Information Movie S12and S13), phalloidin (Supporting Information Figure S1Band Movie S15), and latrunculin B (Supporting InformationFigure S1D and Movie S17), but had no restorative effecton jasplakinolide-treated cells (Supporting Information Fig-ure S1F and Movie S19). Ciliabrevin and phalloidin onlyrobustly impaired actin dynamics at concentrations over 10lM, while latrunculin B and jasplakinolide were both potentat submicromolar concentrations. Considering the similar-ities in phenotypes, working concentrations, and washout,ciliabrevin most closely resembles phalloidin, though it dif-fers in the cessation of retrograde flow and the productionof round actin aggregates at the center of the cell.

Discussion

Efficacy of Chlamydomonas Plate-Well ImagingBased Screen

This study demonstrates that our plate-well imaging basedscreen is able to identify compounds that affect the


Fig. 5. Effects of ciliabrevin (compound B) on tubulin dynamics in Drosophila S2 cells. (A–F) Cells expressing tubulin-GFPwere treated with ciliabrevin, taxol, both drugs sequentially, or cells treated with both drugs and then washed with fresh media asindicated. Frames from movies are shown in left panels, kymographs generated from cross-sections of movies are shown in right pan-els. The time following drug addition or washout is indicated on the movie frame panels, Pre ¼ before treatment. On kymographs,black arrowheads ¼ ciliabrevin addition, white arrowheads ¼ taxol addition, asterisks ¼ drug washout. Horizontal scale bar ¼ 10minutes, vertical scale bar ¼ 10 lm. (G) Quantification of the mean microtubule retrograde flux speed for the experiments in panelsB–F, before (blue) and after (red) drug addition or washout. Error bars indicate 99% confidence intervals. (H) Plot of the maximumpixel intensity position vs. time (blue line) and exponential decay fit (red line) of a kymograph trace from a cell where 1 lM taxolwas added before 34 lM ciliabrevin had stopped the retrograde flux of microtubule plus ends.


assembly and motility of cilia. This suggests that if wecontinue this screen on a much larger scale it could be aproductive source of new chemical tools for perturbingcilia.The screen and secondary assays reported above were all

performed in the unicellular green alga Chlamydomonasreinhardtii. This is one of the major model systems cur-rently in use for studying cilia and flagella, and we expect

that these compounds will provide useful tools to augmentcurrent genetic approaches in this organism. Ciliabrevinin particular, with its ability to reduce IFT speed and sta-bly shorten flagella to specific lengths, may be applicableto studies of flagellar length regulation.Our preliminary studies in different cell types suggest

that of the four compounds we identified, only ciliabrevinhas efficacy in vertebrate cells. However, this was a limited

Fig. 6. Effects of ciliabrevin (compound B) on actin dynamics in Drosophila S2 cells. (A–D) Cells expressing actin-GFP (green)and Arp2-mCherry (red) were treated with increasing concentrations of ciliabrevin or washed with fresh media as indicated. Two-color frames from movies are shown in left panels, Actin-GFP kymographs generated from cross-sections of movies are shown inright panels. The time following drug addition or washout is indicated on the movie frame panels, Pre ¼ before treatment. Onkymographs, black arrowheads ¼ ciliabrevin addition, asterisks ¼ drug washout. Horizontal scale bar ¼ 10 minutes, vertical scale bar¼ 10 lm. (E) Measurement of the mean actin retrograde flow speed over time following ciliabrevin addition for the experiments inpanels A–D. Error bars indicate 99% confidence intervals. (F) Plot of the maximum pixel intensity position vs. time (blue line) andexponential decay fit (red line) of a kymograph trace from a cell where actin flow was stopped by adding 34 lM ciliabrevin.


screen using a small number of compounds, and the factthat one out of four hits showed efficacy in vertebrate cellsis actually encouraging for future prospects. If a greaternumber of compounds were screened in Chlamydomonas,it should be possible to obtain additional compounds thatare applicable to vertebrate studies.

Joint inhibitor of Ciliogenesis and RetrogradeActin Flow

There are several conceivable explanations for the sharedactivity of ciliabrevin on both cilia and retrograde actinflow. One possibility is that there are simply two differenttargets hit by this compound, one affecting cilia and oneaffecting actin flow. We have no data at present to ruleout this possibility. However, we note that components ofthe actin cytoskeleton, as well as regulators of actin dy-namics, are known to play important roles in variousaspects of ciliogenesis [Tamm and Tamm 1988; Bois-vieux-Ulrich et al., 1990; Pan et al., 2007; Dawe et al.,2009; Bershtyn et al., 2010; Kim et al., 2010]. Thus onepossibility is that the compound hits one target moleculethat is involved in both actin flow and ciliogenesis. Athird, more interesting possibility is that the compoundprimarily targets actin and the effect on cilia reflects therole of actin itself in ciliogenesis. One published studyindicates that cytochalasin treatment can result in flagellarshortening in Chlamydomonas [Dentler and Adams,1992]. A recent investigation combining inhibitors ofactin regulation with micropatterned substrates to modu-late cell stretching has further implicated actin networksin the regulation of ciliogenesis [Pitaval et al., 2010]. It isalso worth noting that actin is known to interact withcytoplasmic microtubules [Rodriguez et al., 2003], so thatif the primary target of the drug is involved in the actincytoskeleton, this could affect cytoplasmic microtubuleswhich then might alter trafficking to the cilium or otheraspects of ciliogenesis. Overall, our studies support the ideathat actin perturbations might lead to effects on cilia, butfurther work will be required to test the role of actin pertur-bation on the cilia-related effects of ciliabrevin. In the meantime, ciliabrevin will provide a potentially useful tool toexplore two important fundamental cell biological processes– ciliogenesis and actin flow.

Prospects for Target Identification

Identification of candidate compounds that affect cilia isclearly just the first step, and the more difficult task ofidentifying molecular targets still lies ahead. Indeed, iden-tification of the relevant target is currently the major chal-lenge in cell-based high throughput screening [An andTolliday, 2010]. Actin and components of the actin regu-latory system are obvious candidate targets for ciliabrevin,however the chemical structure of ciliabrevin does not

bear any obvious similarity to that of the standard naturalproduct inhibitors of actin (latrunculin and cytochalasin).The fact that our screen was performed in a genetically

tractable unicellular model system means that resistancescreens can in principle be used to identify potential mo-lecular targets. Resistance screens have been conducted inChlamydomonas using a number of microtubule-targetingherbicides and successfully identified mutations in alphatubulin [James et al., 1993] confirming the feasibility ofsuch an approach in this organism.In the case of the paralyzing compounds E, F, and P, we

performed a proof of concept suppression study using aseries of mutants (sup-pf-1, sup-pf-2, and sup-pf-3) that areknown to suppress mutations in the central pair of the flag-ella, restoring limited motility to mutants that are otherwisecompletely paralyzed [Porter et al., 1994]. We treated eachof these three suppressor strains with our compounds andfound that the sup-pf-1 mutation conferred partial resistanceto compound P (data not shown). Since the sup-pf-1 geneencodes a component of the dynein regulatory complex, thisresult suggests that this complex, or an upstream regulatoryfactor, may be the target of compound P.In the case of ciliabrevin, we note that this same com-

pound was reported in a screen for inhibitors of calciumactivated potassium channels [Sorensen et al., 2008]. Wedo not know how this activity might relate to the effectsof this compound on ciliogenesis, IFT, or actin flow, butthis result does at least provide one possible starting pointfor target identification.

Acknowledgments

The authors thank Ethan Scott and Herwig Baier for pro-viding Paramecium cultures; Elena Ingerman, Lauren Goins,and Dyche Mullins for providing Drosophila S2 cellsexpressing actin-GFP and ARP2-mCherry; Sarah Goodwinand Ron Vale for providing S2 cells expressing tubulin-GFP; Chao Zhang for advice, and Prachee Avasthi Croftsfor careful reading of the manuscript. They are also gratefulto Kurt Thorn for microscopy assistance and the NikonImaging Center at UCSF for access to the confocal andTIRF microscopes used in the study. Susanne Rafelskidesigned the ImageJ kymograph plugin. This work wasfunded by the Sandler Program in Basic Science and theW.M. Keck Foundation.

References

Adams GM, Huang B, Luck DJ. 1982. Temperature-sensitive, as-sembly defective flagellar mutants of Chlamydomonas reinhardtii.Genetics 100: 579–586.

Afzelius BA. 2004. Cilia-related diseases. J Pathol 204: 470–477.

An WF, Tolliday N. 2010. Cell-based assays for high-throughputscreening. Mol Biotechnol 45: 180–186.

Badano JL, Mitsume N, Beales PL, Katsanis N. 2006. The ciliopa-thies: an emerging class of human genetic disorders. Annu RevGenomics Hum Genet 7: 125–148.


Beck C, Uhl R. 1994. On the localization of voltage-sensitive cal-cium channels in the flagella of Chlamydomonas reinhardtii. J CellBiol 125: 1119–1125.

Berman SA, Wilson NF, Haas NA, Lefebvre PA. 2003. A novelMAP kinase regulates flagellar length in Chlamydomonas. Curr Biol13: 1145–1149.

Bershtyn M, Atwood SX, Woo WM, Li M, Oro AE. 2010. MIMand cortactin antagonism regulates ciliogenesis and hedgehog signal-ing. Dev Cell 19: 270–283.

Besschetnova TY, Kolpakova-Hart E, Guan Y, Zhou J, Olsen BR,Shah JV. 2010. Identification of signaling pathways regulating pri-mary cilium length and flow-mediated adaptation. Curr Biol 20:182–187.

Boisvieux-Ulrich E, Laine MC, Sandoz D. 1990. Cytochalasin Dinhibits basal body migration and ciliary elongation in quail oviductepithelium. Cell Tissue Res 259: 443–454.

Brazelton WJ, Amundsen CD, Silflow CD, Lefebvre PA. 2001. Thebld1 mutation identifies the Chlamydomonas osm-6 homolog as agene required for flagellar assembly. Curr Biol 11: 1591–1594.

Cole DG, Diener DR, Himelblau AL, Beech PL, Fuster JC, Rose-nbaum JL. 1998. Chlamydomonas kinesin-II-dependent intraflagellartransport (IFT): IFT particles contain proteins required for ciliaryassembly in Caenorhabditis elegans sensory neurons. J Cell Biol 141:993–1008.

Cole DG. 2003. The intraflagellar transport machinery of Chlamy-domonas reinhardtii. Traffic 4: 435–442.

Dawe HR, Adams M, Wheway G, Szymanska K, Logan CV, Noe-gel AA, Gull K, Johnson CA. 2009. Nesprin-2 interacts with meck-elin and mediates ciliogenesis via remodeling of the actincytoskeleton. J Cell Sci 112: 2716–2726.

Dentler WL, Adams C. 1992. Flagellar microtubule dynamics inChlamydomonas: cytochalasin D induces periods of microtubule short-ening and elongation; and colchicine induces disassembly of the distal,but not proximal, half of the flagellum. J Cell Biol 117: 1289–1298.

Engel BD, Lechtreck KF, Sakai T, Ikebe M, Witman GB, MarshallWF. 2009a. Total internal reflection fluorescence (TIRF) micros-copy of Chlamydomonas flagella. Meth Cell Biol 93: 157–177.

Engel BD, Ludington WB, Marshall WF. 2009b. Intraflagellartransport particle size scales inversely with flagellar length: revisitingthe balance-point length control model. J Cell Biol 187: 81–89.

Finst RJ, Kim PJ, Quarmby LM. 1998. Genetics of the deflagella-tion pathway in Chlamydomonas. Genetics 149: 927–936.

Fliegauf M, Benzing T, Omran H. 2007. When cilia go bad: ciliadefects and ciliopathies. Nat Rev Mol Cell Biol 8: 880–893.

Gaudana R, Ananthula HK, Parenky A, MItra AK. 2010. Oculardrug delivery. AAPS J 12: 348–60.

Gherman A, Davis EE, Katsanis N. 2006. The ciliary proteomedatabase: an integrated community resource for the genetic andfunctional dissection of cilia. Nat Genet 38: 961–962.

Huangfu D, Anderson KV. 2005. Cilia and hedgehog responsive-ness in the mouse. Proc Natl Acad Sci USA 102: 11325–11330.

Hyman JM, Firestone AJ, Heine VM, Zhao Y, Ocasi CA, Han K,Sun M, Rack PG, Sinha S, Wu JJ, Solow-Cordero DE, Jiang J,Rowitch DH, Chen JK. 2009. Small-molecule inhibitors revealmultiple strategies for Hedgehog pathway blockade. Proc Natl AcadSci USA 106: 14132–14137.

Inglis PN, Boroevich KALeroux MR. 2006. Piecing together a cil-iome. Trends Genet 22: 491–500.

James SW, Silflow CD, Stroom P, Lefebvre PA. 1993. A mutationin the alpha 1-tubulin gene of Chlamydomonas reinhardtii confersresistance to anti-microtubule herbicides. J Cell Sci 106: 209–218.

Kim J, Lee JE, Heynen-Genel S, Suyama E, Ono K, Lee K, IdekarT, Aza-Blanc P, Gleeson JG. 2010. Functional genomic screen formodulators of ciliogenesis and cilium length. Nature 464:1048–1051.

Kleinstreuer C, Zhang C, Donohue JF. 2008. Targeted drug-aerosoldelivery in the human respiratory system. Annu Rev Biomed Eng10: 195–220.

Kozminski KG, Johnson KA, Forscher P, Rosenbaum JL. 1993. Amotility in the eukaryotic flagellum unrelated to flagellar beating.Proc Natl Acad Sci USA 90: 5519–5523.

Kozminski KG, Beech PL, Rosenbaum JL. 1995. The Chlamydomo-nas kinesin-like protein FLA10 is involved in motility associatedwith the flagellar membrane. J Cell Biol 131: 1517–1527.

Lefebvre PA. 1995. Flagellar amputation and regeneration in Chla-mydomonas. Meth Cell Biol 47: 3–7.

Marshall WF, Qin H, Rodrigo Brenni M, Rosenbaum JL. 2005.Flagellar length control system: testing a simple model based onintraflagellar transport and turnover. Mol Biol Cell 16: 270–278.

Marshall WF. 2008. The cell biological basis of ciliary disease. JCell Biol 180: 17–21.

Marshall WF. 2009. Quantitative high-throughput assays for flag-ella-based motility in Chlamydomonas using plate-well image analy-sis and transmission correlation spectroscopy. J Biomol Screening14: 133–141.

Merchant S, et al. 2007. The Chlamydomonas genome reveals theevolution of key animal and plant functions. Science 318: 245–250.

Mueller J, Perrone CA, Bower R, Cole DG, Porter ME. 2005. TheFLA3 KAP subunit is required for localization of kinesin-2 to thesite of flagellar assembly and processive anterograde intraflagellartransport. Mol Biol Cell 16: 1341–1354.

Nakamura S, Takino H, Kojima MK. 1987. Effect of lithium onflagellar length in Chlamydomonas reinhardtii. Cell Struct Funct 12:369–374.

Ou Y, Ruan Y, Cheng M, Moser JJ, Rattner JB, van der HoornFA. 2009. Adenylate cyclase regulates elongation of mammalian pri-mary cilia. Exp Cell Res 315: 2802–2817.

Pan J, Wang Q, Snell WJ. 2004. An aurora kinase is essential forflagellar disassembly in Chlamydomonas. Dev Cell 6: 445–451.

Pan J, You Y, Huang T, Brody SL. 2007. RhoA-mediated apicalactin enrichment is required for ciliogenesis and promoted byFoxj1. J Cell Sci 120: 1868–1876.

Parker JD, Quarmby LM. 2003. Chlamydomonas fla mutants reveala link between deflagellation and intraflagellar transport. BMC CellBiol 4: 11.

Pazour GJ, Agrin N, Leszyk J, Witman GB. 2005. Proteomic analy-sis of a eukaryotic cilium. J Cell Biol 170: 103–113.

Pedley TJ, Hill NA, Kessler JO. 1988. The growth of bioconvec-tion patterns in a uniform suspension of gyrotactic micro-organ-isms. J Fluid Mech 195: 223–237.

Pedersen LB, Veland IR, Schroder J, Christensen ST. 2008. Assem-bly of primary cilia. Dev Dyn 237: 1993–2006.

Pedersen LB, Rosenbaum JL. 2008. Intraflagellar transport (IFT)role in ciliary assembly, resorbtion and signaling. Curr Top DevBiol 85: 23–61.

Peterson JR, Mitchison TJ. 2002. Small molecules, big impact: ahistory of chemical inhibitors and the cytoskeleton. Chem Biol 9:1275–1285.

Piperno G, Ramanis Z. 1991. The proximal portion of Chlamydo-monas flagella contains a distinct set of inner dynein arms. J CellBiol 112: 701–709.


Pitaval A, Tseng Q, Bornens M, Thery M. 2010. Cell shape andcontractility regulate ciliogenesis in cell cycle-arrested cells. J CellBiol 191: 303–312.

Porter ME, Knott JA, Gardner LC, Mitchell DR, Dutcher SK.1994. Mutations in the SUP-PF-1 locus of Chlamydomonas rein-hardtii identify a regulatory domain in the beta-dynein heavy chain.J Cell Biol 126: 1495–1507.

Qin H, Wang Z, Diener D, Rosenbaum J. 2007. Intraflagellartransport protein 27 is a small G protein involved in cell-cycle con-trol. Curr Biol 17: 193–202.

Qin H, Burnette DT, Bae YK, Forscher P, Barr MM, RosenbaumJL. 2005. Intraflagellar transport is required for the vectorial move-ment of TRPV channels in the ciliary membrane. Curr Biol 15:1695–1699.

Quarmby LM. 2004. Cellular deflagellation. Int Rev Cytol 233:47–91.

Rodriguez OC, Schaefer AW, Mandato CA, Forscher P, BementWM, Waterman-Storer CM. 2003. Conserved microtubule-actininteractions in cell movement and morphogenesis. Nat Cell Biol 5:599–609.

Rosenbaum JL, Moulder JE, Ringo DL. 1969. Flagellar elongationand shortening in Chlamydomonas. The use of cycloheximide andcolchicine to study the synthesis and assembly of flagellar proteins.J Cell Biol 41: 600–619.

Santos N, Reiter JF. 2008. Building it up and taking it down: theregulation of vertebrate ciliogenesis. Dev Dyn 237: 1972–1981.

Satir P, Pedersen LB, Christensen ST. 2010. The primary cilium ata glance. J Cell Sci 123: 499–503.

Scholey JM. 2003. Intraflagellar transport. Ann Rev Cell Dev Biol19: 423–443.

Sloboda RD, Rosenbaum JL. 2007. Making sense of cilia and flag-ella. J Cell Biol 179: 575–582.

Silflow CD, Lefebvre PA. 2001. Assembly and motility of eukaryo-tic cilia and flagella. Lessons from Chlamydomonas reinhardtii. PlantPhysiol 127: 1500–1507.

Sorensen US, Strobaek D, Christophersen P, Hougaard C, JensenML, Nielsen, EO, Peters D, Teuber L. 2008. Synthesis and struc-ture-activity relationship studies of 2-(N-substituted)-aminobenzimi-

dazoles as potent negative gating modulators of small conductanceCa2þ-activated Kþ channels. J Med Chem 51: 7625–7634.

Tamm S, Tamm SL. 1988. Development of macrociliary cells inBeroe. I. Actin bundles and centriole migration. J Cell Sci 89:67–80.

Tuxhorn J, Daise T, Dentler WL. 1998. Regulation of flagellarlength in Chlamydomonas. Cell Motil Cytoskeleton 40: 133–146.

Wang Q, Pan J, Snell WJ. 2006. Intraflagellar transport particlesparticipate directly in cilium-generated signaling in Chlamydomonas.Cell 125: 549–562.

Welch MD, Mallavarapu A, Rosenblatt J, Mitchison TJ. 1997.Actin dynamics in vivo. Curr Opin Cell Biol 9: 54–61.

Wemmer KA, Marshall WF. 2007. Flagellar length control in Chla-mydomonas—a paradigm for organelle size regulation. Int Rev Cytol260: 175–212.

Wilson CW, Nguyen CT, Chen MH, Yang JH, Gacayan R, HuangJ, Chen JN, Chuang PT. 2009. Fused has evolved divergent roles inHedgehog signaling and ciliogenesis. Nature 459: 98–102.

Wilson NF, Lefebvre PA. 2004. Regulation of flagellar assembly byglycogen synthase kinase 3 in Chlamydomonas reinhardtii. Eukaryo-tic Cell 3: 1307–1319.

Yagi T, Uematsu K, Liu Z, Kamiya R. 2009. Identification ofdyneins that localize exclusively to the proximal portion of Chlamy-domonas flagella. J Cell Sci 122: 1306–1314.

Yang P, Diener DR, Yang C, Kohno T, Pazour GJ, Deines JM,Agrin NS, King SM, Sale WS, Kamiya R, Rosenbaum JL, WitmanGB. 2006. Radial spoke proteins of Chlamydomonas flagella. J CellSci 119: 1165–1174.

Yoshimura K, Matsuo Y, Kamiya R. 2003. Gravitaxis in Chlamydo-monas reinhardtii studied with novel mutants. Plant Cell Physiol44: 1112–1118.

You Y, Richer EJ, Huang T, Brody SL. 2002. Growth and differen-tiation of mouse tracheal epithelial cells: selection of a proliferativepopulation. Am J Physiol Lung Cell Mol Physiol 283:L1315–L1321.

Zhang JH, Chung TD, Oldenburg KR. 1999. A simple statisticalparameter for use in evaluation and validation of high throughputscreening assays. J Biomol Screen 4: 67–73.


a cell-based screen for inhibitors of flagella-driven ...chuang/pdf/20504_ftp.pdfa cell-based screen...

Documents