DNA damage checkpoint triggers autophagyto regulate the initiation of anaphaseFarokh Dotiwala1,2, Vinay V. Eapen2, Jacob C. Harrison2,3, Ayelet Arbel-Eden4, Vikram Ranade5, Satoshi Yoshida,and James E. Haber6

Rosenstiel Center and Department of Biology, Brandeis University, Waltham, MA 02445

Contributed by James E. Haber, October 18, 2012 (sent for review September 24, 2012)

Budding yeast cells suffering a single unrepaired double-strandbreak (DSB) trigger the Mec1 (ATR)-dependent DNA damage re-sponse that causes them to arrest before anaphase for 12–15 h.Here we find that hyperactivation of the cytoplasm-to-vacuole(CVT) autophagy pathway causes the permanent G2/M arrest ofcells with a single DSB that is reflected in the nuclear exclusion ofboth Esp1 and Pds1. Transient relocalization of Pds1 is also seen inwild-type cells lacking vacuolar protease activity after induction ofa DSB. Arrest persists even as the DNA damage-dependent phos-phorylation of Rad53 diminishes. Permanent arrest can be over-come by blocking autophagy, by deleting the vacuolar proteasePrb1, or by driving Esp1 into the nucleus with a SV40 nuclearlocalization signal. Autophagy in response to DNA damage can beinduced in three different ways: by deleting the Golgi-associatedretrograde protein complex (GARP), by adding rapamycin, or byoverexpression of a dominant ATG13-8SA mutation.

adaptation | separase | securin | cell cycle arrest

In the presence of a single unrepairable DNA double-strandbreak (DSB), Saccharomyces cerevisiae arrest is due to Mec1

checkpoint kinase-dependent cell cycle arrest, but cells eventu-ally escape from the arrest and reenter the cell cycle in a processtermed adaptation (1–6). Arrest before adaptation typically lasts12–15 h, a time equivalent to five to six normal cell cycles. Adap-tation is accompanied by the loss of checkpoint-induced hyper-phosphorylation of checkpoint kinases Rad53 and Chk1 and theloss of association with damaged DNA of the Mec1 (ATR) kinase-associated Ddc2 (ATRIP), despite the persistence of DNA dam-age (7, 8). Several proteins are required for adaptation, includingthose with known roles in DNA repair (Yku70, Yku80, Rdh54/Tid1, Rad51, Srs2, Sae2, Fun30, and Sgs1), as well as proteinsthat are required to turn the checkpoint off (the PP2C phos-phatases Ptc2 and Ptc3 and casein kinase II) and the Polo kinaseCdc5, which plays several roles in regulating mitosis (5, 6, 9–14).When DSB repair is relatively rapid, as during HO endonu-

clease-induced mating-type (MAT) switching, there is no cellcycle arrest or Rad53 phosphorylation (7), although Mec1 andTel1 kinases rapidly phosphorylate histone H2A (15, 16). How-ever, when DSB repair is slower, e.g., during ectopic homologousrecombination or long-distance single-strand annealing, the DNAdamage checkpoint is fully activated and G2/M arrest is main-tained for several hours until repair is effected and cells reenterthe cell cycle in a process termed recovery (12, 17). Recovery isan active process that requires the functions of the Srs2 helicaseand also the casein kinase II and PP2C phosphatases to dephos-phorylate checkpoint kinases (11–13). None of the other adap-tation mutants tested (yku70Δ, rdh54Δ, rad51Δ, sae2Δ, cdc5-ad,and fun30Δ) is deficient in recovery (12, 14) although a number ofdouble mutants are recovery defective (3). Analysis of checkpointrecovery in human cells has shown a requirement for the Plk1polo kinase homologous to yeast Cdc5 and the mitotic phos-phatase Cdc25 (17).Although the DNA damage response leading to cell cycle

arrest appears to be nuclear autonomous (18), recent evidencehas suggested that the checkpoint response involves alterations

of cytoplasmic processes (19–24). Most notably, inhibition ofseveral yeast histone deacetylases in cells suffering DNA damageresults in the destruction of Sae2, a key DNA repair protein, byautophagy (25). Very little is understood about how DNA damagetriggers cytoplasmic responses and how autophagy might play arole in the regulation of cell cycle progression in the face of DNAdamage. Here we report that mutations in the Golgi-associatedretrograde protein (GARP) complex unexpectedly block check-point adaptation and recovery. We show that this defect is as-sociated with an increase in autophagy—specifically in the selectivecytoplasm-to-vacuole targeting (CVT) pathway. Increased auto-phagy after DNA damage causes the partial destruction of thechaperone/inhibitor securin (Pds1) that controls entry of sepa-rase (Esp1) into the nucleus to facilitate anaphase. The exclusionof Pds1 from the nucleus after DSB induction can be observedtransiently in cells lacking vacuolar protease activity, but thisstate is maintained by increasing autophagy either by rapamycininhibition of TOR or by overexpression of a dominant ATG13mutation. The inhibition of permanent DNA damage-inducedcell cycle arrest can be overcome by inhibiting autophagy, byinhibiting vacuolar proteolysis, or by driving Esp1 into the nu-cleus with a heterologous nuclear localization sequence (NLS).

ResultsIn a screen for new checkpoint recovery-defective mutations(Materials and Methods), we identified a transposon insertion inthe VPS51 gene, a component of the GARP complex. Subse-quently we found that complete deletions of VPS51 or otherGARP genes (VPS52, VPS53, VPS54, and YPT6) prevent adap-tation (Fig. 1), similar to the phenotype of a previously identifiedadaptation-defective mutation tid1Δ (hereafter referred to asrdh54Δ) (9). Adaptation was assayed by inducing a HO endonu-clease-mediated unrepairable DSB at the MAT locus and ob-serving cells under the microscope for arrest at G2/M (dumbbell-shaped cells) and subsequent budding (6). GARP mutants allmaintain arrest with the characteristic large-budded phenotypewith a single DAPI-staining nucleus, usually at, or stretchedacross, the bud neck. This phenotype eliminates the possibilitythat mutants have completed mitosis but fail to bud in the next

Author contributions: F.D., V.V.E., J.C.H., A.A.-E., S.Y., and J.E.H. designed research; F.D.,V.V.E., J.C.H., A.A.-E., V.R., and S.Y. performed research; F.D., J.C.H., A.A.-E., and S.Y.contributed new reagents/analytic tools; F.D., V.V.E., J.C.H., A.A.-E., V.R., S.Y., and J.E.H.analyzed data; and F.D., V.V.E., J.C.H., S.Y., and J.E.H. wrote the paper.

The authors declare no conflict of interest.1Present address: Harvard Medical School, Boston, MA 02115.2F.D., V.V.E., and J.C.H. contributed equally to this work.3Present address: Samsung Advanced Institute of Technology, Cambridge, MA 02142.4Present address: Hadassah Academic College, Jerusalem 91010, Israel.5Present address: Department of Genetics and Development, Columbia University, NewYork, NY 10032.

6To whom correspondence should be addressed. E-mail: haber@brandeis.edu.

See Author Summary on page 15 (volume 110, number 1).

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1218065109/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1218065109 PNAS | Published online November 19, 2012 | E41–E49

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Oct

ober

29,

202

0

cell cycle. We note that not all vacuolar protein-sorting (vps)mutants, all of which cause carboxypeptidase Y to be secretedrather than retained in the vacuole (26), exhibit an adaptation de-fect; for example, vps25Δ and vps35Δ all display normal adap-tation (Fig. 1).

To establish whether the defects in GARP mutants werespecific to cell cycle arrest provoked by DNA damage, we askedif blocking cells with nocodazole before anaphase would alsolead to a recovery defect. Log-phase cells grown in YEPD liquidmedium were arrested for 12 h with 30 μg/mL nocodazole suchthat >95% were arrested as large dumbbells. These cells werethen spread on YEPD without nocodazole to assess their abilityto resume cell cycle progression and form colonies. vps51Δ cellsexhibit a modest delay in cell cycle progression 8 h after releasefrom nocodazole, but by 24 h 100% of both WT and vps51Δ cellsprogressed beyond the dumbbell stage (Fig. S1). Hence, noco-dazole arrest, which triggers the spindle checkpoint activation,does not cause the same defect, as does a single unrepaired DSB.

GARP Mutants Are Suppressed by Inactivating Vacuolar Protease Prb1.It has been observed that GARP mutants, like other vps mutants,cause mislocalization of vacuolar proteases (26); hence the ad-aptation defect of GARP mutants could be due to the aberrantproteolytic degradation of a key mitotic activator. We thereforeasked whether vps51Δ mutants are suppressed by deleting Pep4,which activates several vacuolar proteases (27). We found astriking suppression of vps51Δ by pep4Δ (Fig. 2A). In contrast,there was no significant suppression of three adaptation-defectivemutations involved in DNA metabolism or dephosphorylation ofthe checkpoint kinase cascade (rdh54Δ, srs2Δ, and ptc2Δ ptc3Δ).Pep4 activates a number of vacuolar proteases. We established

that deleting the gene encoding one of these, proteinase B (PRB1),showed the same suppression of the adaptation defects of vps51Δas pep4Δ, again without affecting the other adaptation-defectivemutations (Fig. 2B). These results lead to the suggestion that cellswith defects in retrograde vesicle trafficking fail to adapt becauseof the Prb1-dependent degradation of a necessary mitotic activator.

Nuclear Accumulation of Pds1 and Esp1 Is Defective in GARP Mutants.The checkpoint-mediated G2/M arrest is enforced by preventing

n=100

8 hr

24 hr

48 hr

% c

ells

ada

pted

20

40

60

80

100

0

Fig. 1. Vps51 is required for adaptation to the DNA damage checkpoint. WT(JKM179) and vesicular transport mutants were tested in the checkpoint ad-aptation assay (Materials and Methods). Of these, WT cells adapt whereasvps51Δ, vps53Δ, and ypt6Δ were found to be adaptation defective. vps52Δ hasa similar effect but was measured only at 24 h and is not shown for clarity.Retromer mutant (vps35Δ) and ESCRT II mutant (vps25Δ) were not adaptationdefective. The rdh54Δ cells serve as positive control for the adaptation defect.Cells were pregrown in YEP-Lactate (YEPL) media overnight and then ma-nipulated onto minimal complete media supplemented with 1% yeast extractand containing 2% galactose (CYG). The percentage of cells progressing be-yond the “dumbbell” arrest stage (“percentage of adaptation”) is shown.

A

B

8 hr

24 hr

48 hr

% c

ells

ada

pted

20

40

60

80

100

0

% c

ells

ada

pted

20

40

60

80

100

0

8 hr

24 hr

48 hr

Fig. 2. (A and B) pep4Δ and prb1Δ can rescue the adaptation defect in vps51Δ cells but not in rdh54Δ, srs2Δ, and ptc2Δ ptc3Δ cells.

E42 | www.pnas.org/cgi/doi/10.1073/pnas.1218065109 Dotiwala et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

29,

202

0

separase, Esp1, from releasing sister-chromatid cohesion (28).There are several constraints on Esp1. First, it is inhibited bysecurin, Pds1, which is stabilized against APC/C-mediated deg-radation by Chk1- and Rad53-dependent phosphorylation (29).Second, Esp1 appears to be sequestered in the cytoplasm untilthe proper time in the cell cycle (30). Nuclear import of Esp1 isat least in part mediated by Pds1 acting as a chaperone (31)although other chaperones may also play important roles (32).To examine localization of Pds1 and Esp1, we integrated a GFPtag in the genomic copy of PDS1 or ESP1. Both PDS1-GFP andESP1-GFP are functional, as the cells expressing these GFP fu-sion proteins in the absence of wild-type PDS1 or ESP1 did notshow any detectable mitotic defects even at 37 °C, whereas esp1Δis lethal at all temperatures and pds1Δ is inviable at 37 °C. Thelocalization of Pds1-GFP and Esp1-GFP in rapidly growing cellcultures is consistent with previous reports (28, 30, 33). Bothtagged proteins accumulated in the nucleus in medium-to-largebudded cells (Fig. S2).When the cells were checkpoint arrested in G2/M phase by

induction of a DSB and examined 6 h later, Pds1-GFP accu-mulated in the nucleus both in wild-type cells and in the adap-

tation-defective rdh54Δ strain (Fig. 3A). However, in vps51Δ,Pds1-GFP is not concentrated in the nucleus and is apparentlymuch less abundant, as measured by total fluorescence intensityof the cells (Fig. 3A). This delocalization of Pds1-GFP in vps51Δis specific to DNA damage-induced G2/M arrest; in large-bud-ded growing cells and in nocodazole-arrested cells Pds1-GFP ismostly found in the nucleus (Figs. S2 and S3). Like Pds1-GFP,Esp1-GFP is strongly concentrated in the nucleus of DSB-arrestedwild-type and rdh54Δ cells 6 h after DSB induction; however, invps51Δ cells, nuclear accumulation of Esp1-GFP was not seen(Fig. 3A). Remarkably, this defect in Esp1-GFP and Pds1-GFPlocalization is rescued by deleting PEP4 (Fig.3A). These datasupport the idea that—in the GARP mutants—the DNA damagecheckpoint response prevents nuclear accumulation of both Pds1and Esp1, resulting in prolonged checkpoint-mediated arrest.Pds1-GFP and Esp1-GFP were often found in endosome-like

structures in the cytoplasm in vps51Δ pep4Δ cells (Fig. 3A). Ourattempts to visualize both Pds1-GFP and endosome/vacuolemembranes were unsuccessful because vps51Δ has a fragmentedvacuole and is defective in the endosomal marker FM4-64 up-take (34). However, the altered localization and degradation of

A

B

WT rdh54 vps51 vps51pep4

Esp

1-G

FPP

ds1-

GFP

Esp1-myc

Loading

no ta

g

WT

vp

s5

1

pe

p4

vp

s5

1 p

ep

4

Pds1-GFP

Loading

no ta

g

WT

vp

s5

1

pe

p4

vp

s5

1 p

ep

4

0

20

40

60

80

100Pds1-GFPEsp1-GFP

C

% n

ucle

ar G

FP s

igna

l

Fig. 3. Pds1 and Esp1 are mislocalized in the cells arrested by DNA damage. (A) Localization of Esp1 and Pds1 during DSB-induced G2/M arrest. The cells werearrested in G2/M phase for 6 h after galactose induction of HO endonuclease. Images were taken without fixation. (B) Quantitation of the images shown in A.The percentage of cells displaying nuclear GFP signal was calculated in each case by counting 100 G2/M-arrested cells per sample. (C) Western blotting of Pds1-GFP and Esp1-myc in the cells arrested by DNA damage. Yeast lysates were prepared 12 h after galactose induction of HO endonuclease. Western blotting ofRho1 is shown as a loading control. Esp1 protein level was not affected 6 h after the DNA DSB. The anti-PSTAIR antibody was used as a reference control.

Dotiwala et al. PNAS | Published online November 19, 2012 | E43

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Oct

ober

29,

202

0

Pds1 are not only seen in these vps51Δ mutant cells, but also canbe seen in pep4Δ cells after induction of an unrepaired DSB (Fig.4). To further examine whether Pds1 is targeted to the vacuolefor degradation during DNA damage-induced G2 arrest, wecarefully quantified Pds1-GFP localization in the cell, as describedin Materials and Methods. In WT cells, 6 h after DSB induction,Pds1-GFP accumulates in the nucleus and is absent from thevacuole (less signal in the vacuole than in the cytoplasm, as seenin Fig. 4). In contrast, in pep4Δ cells, Pds1-GFP signal in thevacuolar area was much higher than in the cytoplasm. Vacuolaraccumulation of Pds1-GFP in pep4Δ was further confirmed bystaining the vacuole with FM4-64 (Fig. S4).

Degradation of Pds1 and Nuclear Localization Defects of Pds1 andEsp1 in GARP Mutants Are Suppressed by PEP4 Deletion. Reducednuclear Pds1 and Esp1 could be due to their mislocalization,degradation, or both.By Western blot analysis we found that Pds1-GFP levels were

significantly lower in vps51Δ afterDNAdamage compared withWTand that the level of Pds1 in vps51Δ strains was restored by deletionof PEP4 (Fig. 3C). Importantly, defective nuclear accumulationof Pds1-GFP in the vps51Δ strain was also restored by deletion ofPEP4 (Fig. 3A). The importance of Pds1 in nuclear localization ofEsp1 has already been established (33). These results are inagreement with the idea that the adaptation defect seen in vps51Δ iscaused at least in part by the exclusion of Esp1 from the nucleuswhen Pds1 is partially degraded and apparently mislocalized in re-sponse to the DNA damage checkpoint in these mutant cells. Thus,stabilization of Pds1 is sufficient to drive Esp1 back into the nucleus.Consistent with the Western blotting data, Pds1-GFP is stabilizedin pep4Δ and the GFP signal in the nucleus is much stronger,possibly explaining how pep4Δ could rescue the Pds1 localizationdefect in vps51Δ. These results lead to the surprising conclusionthat cells with defects in retrograde vesicle trafficking fail to adaptbecause of the Pep4- and Prb1-dependent degradation of one ormore necessary mitotic activators, including Pds1.

GARP Mutants Are Suppressed by Driving Separase into the Nucleus.Because Esp1 seems to be excluded from the nucleus after DNAdamage in vps51Δ cells, we hypothesized that we could suppressthe mislocalization of Esp1 by fusing it to the NLS of SV40 (30).In our experiments the fusion protein gene was transcribed un-der the control of the normal chromosomal ESP1 promoter at its

normal chromosomal location (Materials and Methods). Indeed,Esp1-NLSSV40 efficiently suppressed the adaptation defects ofvps51Δ (Fig. 5). Esp1-NLSSV40 does not affect the permanentarrest of rdh54Δ or ptc2Δ ptc3Δ, which presumably affects eventsentirely within the arrested nucleus. Esp1-NLSSV40 did partiallysuppress srs2Δ; in addition, wild-type cells adapt more completelywhen Esp1-NLS SV40 (superscripted as in other instances) isexpressed. As we do not know the mechanism by which eithersrs2Δ or rdh54Δ blocks the process of adaptation, we cannotaccount for the difference. These data argue strongly that thedisruption of normal vesicle traffic results in the retention orinactivation of Esp1 in a cytoplasmic compartment and this canbe overcome when Esp1 is driven into the nucleus

Inhibition of Adaptation in vps51Δ Requires Autophagy via the CVTPathway. If Pds1 and/or other factors that are needed to facilitateEsp1 entry into the nucleus are degraded in vps51Δ strains, thenthere should be a pathway to direct these proteins to the vacuoleor some other vesicular compartment harboring Prb1. Two ma-jor ways in which this might occur should involve either thetargeted (CVT) or general autophagy pathways (35, 36).We first examined Atg1 and Atg5 mutations that affect both

autophagy pathways (37). In otherwise WT cells, a defect inautophagy did not affect adaptation, as seen by examining atg1Δand atg5Δ mutations (Fig. 6 A and B). However, when we con-structed atg5Δ vps51Δ and atg1Δ vps51Δ double mutants, wefound that these mutations suppressed the adaptation defect ofvps51Δ (Fig. 6 A and B). Furthermore, deletion of atg5Δ did notsuppress adaptation-defective mutants rdh54Δ, srs2Δ, and ptc2Δptc3Δ. We then examined atg17Δ and atg11Δ that are specificallyrequired for the bulk and CVT pathways, respectively (37). Asseen in Fig. 6B, atg11Δ suppressed the permanent arrest ofvps51Δ cells whereas atg17Δ did not. These results support theidea that the DNA damage-induced response involves the tar-geted degradation of mitotic regulators via the CVT pathway andthat this phenotype is exacerbated in GARP mutants. We notethat vps51Δ cells with knockouts in ESCRT2 (vps25Δ) or theretromer complex (vps35Δ) are still adaptation defective (Fig. S5).Deletion of VPS64 is synthetically lethal with pds1Δ at normallypermissive temperatures (31), which could suggest that it playsan alternative role in transport of Esp1. However, a deletion ofvps64Δ did not have an adaptation defect, nor did it suppress thevps51Δ adaptation defect. (Fig. S5).

Induction of Increased Autophagy Causes an Adaptation Defect inWild-Type Cells. Increased autophagy, in response to nutrientdeprivation or after alkylation-mediated DNA damage, can betriggered by inactivating the TOR complex with rapamycin(20, 38). We asked whether the normal adaptation response ofwild-type cells could be blocked by inhibiting TORC1 with rapa-mycin (Materials and Methods). Because rapamycin-inhibited

WT pep4

Pds

1-G

FP

Fig. 4. Prevention of vacuolar degradation by pep4Δ reveals that Pds1-GFPcan be targeted to vacuole/endosomes. The cells were arrested in G2/Mphase for 6 h after galactose induction of HO endonuclease. Pds1-GFP signalintensity was pseudocolored to reflect their signal intensity. These sampleswere prepared, taken at the same time in the same condition, and theimages were equally processed. Arrows point to the position of vacuoles.

% c

ells

ada

pted

20

40

60

80

100

0

8 hr24 hr48 hr

Fig. 5. Artificial nuclear targeting of Esp1 rescues the adaptation defect ofvps51Δ. Adding the nuclear localization signal (NLS) from SV40 to the Cterminus of Esp1 suppresses the adaptation defect of vps51Δ and partially ofsrs2Δ but has no effect on two other adaptation-defective mutations.

E44 | www.pnas.org/cgi/doi/10.1073/pnas.1218065109 Dotiwala et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

29,

202

0

cells complete mitosis but fail to complete cytokinesis (39), wecould not use our standard adaptation assay. Instead we moni-tored the proportion of cells suffering a single DSB that arrestedwith a single nucleus (thus adaptation defective) or could completemitosis and become binucleate. As shown in Fig. S6, rapamycinprovoked a failure of adaptation. When this assay was repeatedin an atg5Δ strain, we found that nuclear division could proceed,supporting the idea that hyperactivation of autophagy by rapa-mycin results in permanent G2/M arrest following induction of asingle unrepaired DSB (Fig. S6).Because rapamycin has a number of pleiotropic effects (39),

we searched for another method to specifically induce autophagyfollowing DNA damage. We took advantage of the discoverythat overexpression of a nonphosphorylatable form of ATG13(ATG13-8SA) induces autophagy without affecting normal cellcycle progression (40). Here, too, we found that WT cells sufferinga DSB were unable to adapt when ATG13-8SA was overexpressed;moreover, deleting either ATG1 or PEP4 suppressed the actionof ATG13-8SA overexpression and allowed adaptation (Fig. 7A).This adaptation defect is dependent on a functional DNA damageresponse as a deletion of the MEC1 checkpoint kinase completelysuppressed the effect of ATG13-8SA overexpression, as did astrain lacking the HO endonuclease cut site (Fig. 7A). Further-more, we confirmed that the failure to adapt was correlated withan inability to complete nuclear division by assessing the numberof mononucleate (hence arrested) cells. As seen in Fig. S7, ∼80%of ATG13-8SA overexpressing cells remain arrested with a singlenucleus at 24 h.We then examined the localization of both Pds1-GFP and

Esp1-GFP following DNA damage while overexpressing ATG13-8SA. In these cells both Pds1-GFP and Esp1-GFP are unable tolocalize properly following DNA damage, mirroring what weobserve in the GARP mutants (Fig. 8 and Fig. S8). Finally,we found that atg11Δ was also able to suppress the effect of

ATG13-8SA overexpression whereas atg17Δ did not (Fig. 7A).These results mimic what was seen in the GARP mutations andsuggest that increasing autophagy in wild-type cells is sufficientto block adaptation via the CVT pathway.

Rad53 Checkpoint Kinase Is Dephosphorylated in GARP Mutants AfterDNA Damage.We had previously shown that during the process ofadaptation the checkpoint kinase RAD53 is robustly phosphor-ylated in response to DNA damage and subsequently is dephos-phorylated at the time of adaptation. In some adaptation-defectivemutants, such as rdh54Δ, ptc2Δ, ptc3Δ, or cdc5-ad, Rad53 remainshyperphosphorylated for at least 24 h (7). In contrast to otheradaptation-defective mutations, vps51Δ cells also dephosphory-lated Rad53 with kinetics similar to those of WT cells (Fig. 7B),suggesting that hyperphosphorylation of Rad53 is not responsiblefor the permanent arrest defect of GARP mutants. Moreover,Rad53 is dephosphorylated after 12 h in ATG13-8SA overex-pressed cells (Fig. 7B), suggesting that these cells closely mimicthe enhanced autophagy conditions seen in vps51Δ cells. Theseresults suggest that, whereas the DNA damage response is re-quired to establish cell cycle arrest and designate Pds1’s degra-dation by the CVT pathway, when autophagy is hyperactivated,

A

0

20

40

60

80

100

WT vps51 atg11 vps51

atg11

atg1 vps51

atg1

atg17 vps51

atg17

% a

dapt

ed a

t 24h

B

****

8 hr24 hr48 hr%

ada

pted

0

20

40

60

80100

Fig. 6. Inhibition of autophagy rescues the adaptation defect of vps51Δ. (A)Deletion of ATG5 suppresses vps51Δ but has little effect on other adapta-tion-defective mutants. (B) Deletion of ATG1 and ATG11 but not ATG17suppresses the adaptation defect of vps51Δ. ** denotes statistically signifi-cant difference compared with vps51Δ cells. Error bars reflect SEM of threeindependent experiments.

% a

dapt

ed a

t24h

A

B

0

20

40

60

80

100

WT WT atg1 pep4 atg11 atg17 WT(-cs) mec1

sml1

** **

**

+ pGAL ATG13-8SA

0 3 6 9 12 24 27

0 3 6 9 12 15 18 24

Rad53Rad53-P

vps51Δ

WT+ pGAL ATG13-8SA

WT

Rad53Rad53-P

Rad53Rad53-P

Fig. 7. (A) Induction of autophagy prevents adaptation in WT cells. Adap-tation assay was performed as described in Materials and Methods exceptthat strains harboring plasmids were grown in selective media containingraffinose overnight to maintain the plasmid and then unbudded cells werespread onto YEP-Galactose plates and monitored for adaptation at 24 h.** denotes a statistically significant difference compared with the WT strainoverexpressing ATG13-8SA. Error bars reflect SEM of three independentexperiments. (B) The Rad53 checkpoint kinase is dephosphorylated in GARPmutants with WT kinetics. The indicated strains were grown in YEP-Lactateand DNA damage was induced by adding galactose to the cultures. Sampleswere taken at various time points denoted and Western blot was performedas described in Materials and Methods.

Dotiwala et al. PNAS | Published online November 19, 2012 | E45

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Oct

ober

29,

202

0

this sequence of events makes it impossible for Esp1 to accu-mulate in the nucleus and execute the release of anaphase.

Recovery Defects of vps51Δ Involve Prb1-Mediated Degradation andAutophagy. Recovery is defined as the process by which cells turnoff the DNA damage checkpoint and resume the cell cycle afterrepairing the DNA DSB. To assess recovery we used strainYMV80, which repairs a HO-induced DSB through single-strandannealing, in which the flanking homologous sequences areseparated by 25 kb and repair requires 6 h (12). To assay recovery,strains in the YMV80 background were grown in preinductionmedia (YEP-Raffinose or YEP-Lactate) and the cells were thenplated on YEPD and YEP-Gal. For each strain viability wasmeasured as a ratio of the number of colonies on YEP-Gal tothose on YEPD, as well as the percentage of cells arrested asdumbbells at 24 h. Among previously studied adaptation-defectivemutants only srs2Δ, ptc2Δ ptc3Δ, and sae2Δ are recovery defective(12). Approximately 35% of vps51Δ cells fail to resume cell cycleprogression (Fig. 9). We also found that overexpression of ATG13-8SA, which prevented adaptation in WT cells, also prevents re-covery as measured by colony formation on galactose plates (Fig. 9).As in adaptation assays, we found that atg11Δ and atg1Δ suppressedthe recovery defect of overexpressing ATG13-8SA (Fig. 9B).

DiscussionAfter DNA damage provokes G2/M arrest, there is a transientactivation of the CVT-mediated degradation of Pds1 and thefailure of Esp1 to enter the nucleus to allow resumption of mitosis.This process is exacerbated when the CVT pathway of autophagyis hyperactivated, in GARPmutants, by overexpression of ATG13-8SA or by rapamycin treatment; under these conditions cellscannot adapt and show reduced recovery even when damage isrepaired. This arrest can be bypassed in at least four ways: (i) byinactivating the DNA damage checkpoint; (ii) by driving Esp1into the nucleus without its normal chaperone; (iii) by blockingautophagy, specifically with mutations that appear to block CVTpathway; or (iv) by preventing vacuolar proteolysis.

It is striking that deletion of PRB1 or PEP4 can suppressvps51Δ, not only where the vacuole is fragmented, but also whenATG13-8SA is overexpressed and vacuolar morphology is ap-parently normal. One might have imagined that once cargo suchas Pds1 was delivered to the vacuole, it would remain sequesteredand could not escape, undegraded, and thus could not enableEsp1 to enter the nucleus and initiate anaphase; but preventingproteolysis does not trap sufficient Pds1 (or some other compo-nent) to block mitosis. We suggest that there is some sort of“clogging” of the CVT pathway so that enough Pds1 remains atliberty to carry out its function or that the sequestration of cargo isin some way reversible in the absence of proteolysis. We note thatthe abundance of Pds1 is increased in a pep4Δ strain, back to wild-type levels (Fig. 3); moreover, there is strong localization not onlyin the vacuole but also in the nucleus (Fig. 4).We note that previous studies have suggested that Vps51,

apart from its role in endosome-to-Golgi retrograde transport, isessential for Cvt targeting (34), whereas our results suggest thatvps51Δ and other GARP mutants actually exacerbate autophagyvia this pathway. One possible explanation for this apparent

% n

ucle

arlo

caliz

atio

nof

GFP

Pds1-GFP Esp1-GFP

Fig. 8. Overexpression of ATG13-8SA causes mislocalization of Pds1-GFPand Esp1-GFP. The images were taken as described in Fig. 3A. Representativeimages are shown in Fig. S6. The graph denotes the percentage of G2/M-arrested cells that showed nuclear localization of GFP 6 h after galactoseinduction. At least 100 G2/M-arrested cells were counted per sample.

YMV80 cdc5-ad

% viability% dumbbellarrested cells at 24 h

0

10

20

30

40

50

60

70

80

90

100P

erce

nt

pGAL- ATG13*

B

A

0102030405060708090

100

6h

24h

% G

2/M

arre

sted

cel

ls

Fig. 9. Overexpression of ATG13-8SA (here designated ATG13*) andvps51Δ prevents recovery from DNA damage. WT (YMV80), vps51Δ, cdc5-ad,and YMV80+pGAL ATG13-8SA were pregrown in YEP-Lactate and equalnumbers of cells were spread onto YEPD plates or YEP-Gal plates. The ratioof the number of viable colonies on YEP-Gal vs. YEPD is presented as per-centage of recovery (solid bars). Cells were pregrown in YEP-Lactate liquidmedia and HO expression was induced with galactose for 6 h. Cells werethen spread onto YEP-Gal plates and the cell cycle progression of individualcells was monitored at 24 h after induction. The percentage of cells thathave not yet recovered and are still arrested at the two-cell body “dumb-bell” stage is presented as percentage of dumbbells (shaded bars). (B) Theblock in adaptation imposed by overexpression of ATG13* is suppressed bydeleting ATG1 or ATG11.

E46 | www.pnas.org/cgi/doi/10.1073/pnas.1218065109 Dotiwala et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

29,

202

0

difference is the effects we see are coupled to the induction ofthe DNA damage checkpoint, which may alter several cytoplas-mic targets in addition to the well-documented effects on septins(19) and dynein-dependent nuclear positioning (41). Thus, theidea that DNA damage should manifest its response by an al-teration of cytoplasmic function is not without precedent.We interpret the defects of GARP mutants in resuming cell

cycle progression after DNA damage as a result of an alterationin the localization and stability of a key mitotic activator. Mostlikely this activator is Esp1 and/or an associated chaperone orcytoplasmic receptor, because we can suppress the adaptationand recovery defects of vesicle transport mutants by driving Esp1into the nucleus with an SV40 NLS. Previous studies have sug-gested that the transport of Esp1 into the nucleus is highly reg-ulated, so that even when securin is absent, mitosis is notexecuted prematurely (32). After DNA damage there appear tobe additional constraints that prevent Esp1 entry or activity. Forexample, even in a pds1Δ strain, there is a substantial cell cycledelay in response to a single DSB, a delay that is not seen in astrain lacking the ATR homolog, Mec1 (41). The basis of theinhibition of Esp1 activity is not known; we suggest that separasecould be restrained in the cytoplasm by association with a trans-membrane-bound receptor, e.g., in the way that the cyclin Cln3 isbound at the endoplasmic reticulum (ER) surface and releasedby the chaperone Ydj1 (42). Here the release and/or transport ofEsp1 appear to depend on Pds1. Recently, Sarin et al. (31)identified several possible chaperones whose deletions are syn-thetically lethal with the absence of Pds1. These proteins arecandidates to be cochaperones with Pds1 in coordinating thelocalization of Esp1, but deleting VPS64 did not mimic vps51Δ.Still, there are important questions remaining. For example, isadaptation or recovery accompanied by a change in APC-Cdc20activity? We do know that APC activity is not required for ad-aptation as it is reflected in the loss of Rad53 hyperphosphorylation(7, 8) but it should continue to be important for the nuclear deg-radation of Pds1. It is also intriguing that maintenance of check-point-mediated arrest of yeast cells to DNA damage (i.e., a partialloss of telomere protection) depends on APC-Cdh1 (43).Defects in GARP mutants have revealed that the DNA dam-

age checkpoint prevents nuclear accumulation of Esp1. Our datasuggest that Pds1 is the key factor that is targeted for degradationin the vacuole. It is likely that transporting excess Pds1 to thevacuole is ensuring stable cell cycle arrest in G2 phase and thatthis transport is likely mediated via the CVT pathway. Consistentwith this observation we find that atg1Δ cells, which are defectivefor both autophagy pathways, show a reduced duration of check-point arrest and adapt more extensively. Thus, the sequestrationand/or degradation of key mitogenic factors by autophagy couldbe one way by which cells can maintain a robust checkpoint arrestin response to DNA damage. We note that other factors may alsobe sequestered or degraded, so that the degradation of Pds1might not be the only, or even most important, alteration. It isimportant to note that we also see evidence of vacuolar locali-zation of Pds1-GFP in normal cells after the induction of anunrepaired DSB when vacuolar degradation is blocked by PEP4deletion. The GARP mutants exacerbate trafficking of Pds1 tothe vacuole, but these mutants are exacerbating a normal, under-lying phenomenon.We note also that inmammalian cells depletingRab6, the homolog of Ypt6, results in mitotic defects that suggesta second role for this G protein, much as we suggest here (44).Recently there have been other reports of the autophagic

degradation of proteins critical for DNA repair. The Sae2 pro-tein is lost from the nucleus and degraded via autophagy in cellssuffering DNA damage when treated with an inhibitor of histonedeacetylase (25). Similarly, the Rnr1 protein of the ribonucleotidereductase complex is degraded after alkylation damage; more-over, as we report here, inhibition of the TOR complex exacer-bates this response (20). The degradation of Sae2 upon valproic

acid treatment is partially rescued by deletion of ATG19, a genespecifically required for the CVT pathway (25). However, theseexperiments did not rule out the possibility that some amount ofthe degradation could be mediated via the bulk autophagicpathway as well. In this paper, we show that deletion of eitherATG1 or ATG11 but not ATG17 suppresses the permanent check-point arrest of both vps51Δ and ATG13-8SA overexpressing cells.We suggest, on the basis of our results with the fate of Pds1, thatthese nuclear proteins are likely targeted for degradation via theCVT pathway.Another recent report suggests that DNA damage induced by

sublethal doses of camptothecin or etoposide results in autoph-agy-induced cell death in HeLa cells (45), and treatment ofMCF7 breast cancer cell lines with rapamycin appears to inhibitboth homologous recombination and nonhomologous end-join-ing pathways (46). However, in the latter study it is not clearwhether the defect is due to autophagic induction or some othereffect of rapamycin treatment. We note that the specific induc-tion of autophagy in strains that robustly activate the DNA dam-age checkpoint, but that eventually manage to repair the lesion,resulted in a dramatic drop in viability. This could be due to aninability to resume mitosis after DNA damage, defects in homol-ogous recombination, or a combination of both factors. Furtherwork will address the contribution of autophagy in these diversepathways of homology-driven repair.As we noted before, the GARP proteins may play roles in

several steps in vesicle transport, but are characterized as beingdefective in Golgi-to-ER retrograde transport because that is thestep most clearly affected. However, GARP mutants are just oneof many subsets of vesicle transport gene mutations that com-monly affect vacuolar protein sorting; but other vps mutationssuch as those affecting the Retromer complex or the GET com-plex have no effect in adaptation after DNA damage. We have,however, found deletion mutations in other vesicle transportfunctions that exhibit similar adaptation defects, including theARF-GAP Gcs1 and the TRAPP complex protein Gsg1.

Materials and MethodsYeast Media, GAL-HO Induction, Adaptation, and Recovery Assays. Yeast cellswere grown in YEPD [1% yeast extract, 2% peptone, 2% dextrose (wt/vol)]media for normal growth and before spot assays. YEP plus 3% lactic acid(YEP-Lactate) or YEP plus 2% raffinose (YEP-Raffinose) was used as a pre-induction medium for most adaptation and recovery assays. Strains carryingplasmids were either pregrown in synthetic media with raffinose or grownin synthetic media with dextrose and then inoculated into YEP-Lactate forpreinduction growth of 12–16 h. We have not observed any differenceresulting from these preinduction regimes. For adaptation assays unbuddedG1 cells in the JKM179 background were micromanipulated into grids onplates of synthetic complete media with 2% galactose supplemented with1% yeast extract (CYG) or on YEP-Galactose (YEP-Gal) plates and cell cycleprogression was monitored at 8, 24, and 48 h (6). The percentage of cellswith three or more cells or buds is plotted as a measurement of adaptation.To assay recovery, YMV80 cells (12) were grown overnight in YEP-Lactateand equal numbers of cells were spread onto YEPD (HO off) and YEP-Galplates (HO on). The percentage of recovery is the ratio of the number ofviable colonies (YEP-Gal/YEPD) after 3 d growth at 30 °C. To assess a delay inrecovery we induced HO expression by adding galactose to YEP-Lactateliquid for 6 h and then plated the cells on YEP-Gal plates. Plates were in-cubated overnight at 30 °C and the number of cells in individual micro-colonies was determined 24 h after the initial HO induction.

Strains, Plasmids, and PCR Manipulations. All yeast strains are described inTable S1. Strain JKM179 and its MATa derivative, JKM139, used for adap-tation assays, and strain YMV80, used for recovery assays, have been de-scribed in refs. 6 and 12. All other strains were created by transformation ofeither PCR-generated or plasmid-derived deletion fragments or by standardgenetic crosses. Standard PCR conditions and Taq polymerase were used in allcases. Oligo sequences are available on request. The double-deletion strainswere created by crossing the appropriate single mutants and dissectingtetrads after sporulating the diploid. We also used rdh54Δ::URA3 (YSL301),srs2Δ::LEU2 (YSL302), and ptc2Δ::URA3 ptc3Δ::KAN (MCM298). Strains

Dotiwala et al. PNAS | Published online November 19, 2012 | E47

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Oct

ober

29,

202

0

without an HO cut site were created by picking rare colonies that grew ongalactose plates. Such colonies have cut and then religated MAT by impre-cise end joining, generating a sequence that is no longer recognized by theHO endonuclease. The JKM179 strain expressing Esp1-NLSSV40, under theendogenous Esp1 promoter, was created by transforming with the URA3-marked plasmid pFD026 after cutting with BstAPI.

pFD026 was constructed by amplifying 1.5 kb of 3′ Esp1 ORF, using Esp1forward and Esp1 reverse primers (NLS sequence included in the primers),and inserting it into pRS306 using BamHI and EcoRI. Single-copy Esp1 NLSSV40 canbe introduced into yeast by transforming after cutting pFD026 with BstAPI.

PDS1-GFP and ESP1-GFP strains were constructed using a single-step PCRmethod (47). A 6-aa flexible linker (Gly-Gly-Ser-Gly-Gly-Ser) was insertedbetween the ORFs and GFP to secure functionality of the fusion proteins.

Screen for Novel Recovery Mutants. To identify novel checkpoint recoverymutants, the wild-type recovery strain YMV80 (12) was mutagenized withthe mTn-LacZ-LEU2 library (48), and colonies were assayed for recovery ina replica-plating regime. A total of 17,500 independent colonies were ana-lyzed, and the two strongest secondary positive mutants were analyzedfurther. One mutant with delayed colony formation (T11) was found to havethe transposon inserted at nucleotide 132 of the 495-nt VPS51 ORF. Thesecond mutant (T20) was found to have an insertion at nucleotide 1,435 ofthe 2,880-nt MSH1 ORF. Cells lacking MSH1 grow poorly on plates contain-ing ethanol and glycerol as the carbon source, which is the second step inour screening regime; hence the apparent recovery defect in this strain ismost likely a false positive. This screen was not extended to identify allpossible candidates.

Microscopy. Cells were imaged using a Zeiss 200M inverted or a Zeiss AX10microsope (Carl Zeiss) equipped with an MS-2000 stage (Applied ScientificInstrumentation), a Lambda LS 175-W Xenon light source (Sutter Instru-ments), and 63× Plan-Fluar objectives. Image acquisition was performedwith a CoolSnap HQ camera (Roper Scientific). The microscope and camerawere controlled by SlideBook software (Intelligent Imaging Innovations).

All image manipulations and fluorescence intensity measurements werecarried out using SlideBook software and figures were prepared withPhotoshop software.

Western Blots.Western blotting of Rad53 was performed as described in ref. 7with anti-RAD53 antibodies (provided by M. Foaini). Briefly, indicated strainswere grown in YEP-Lacate, galactose was added to induce the DSBs, andthey were collected at various time points indicated. Cells were collected bycentrifugation and washed with 20% TCA. All purification steps were per-formed on ice with prechilled solutions. Cell pellets were resuspended in500 μL 20% TCA and subjected to glass bead lysis. The suspension was col-lected, 1 mL of 5% TCA was added to wash the beads, and wash minus theglass beads was added to the suspension. The precipitated proteins werecollected by centrifugation. Pellets were washed with 1 mL 100% ethanoland proteins solubilized in 200 μL 1 M Tris (pH8.0)/300 μL 3× SDS/PAGEloading buffer [60 mM Tris (pH 6.8), 2% SDS, 10% glycerol, 100 mM DTT,0.2% bromophenol blue]. After 20 min at 95 °C, insoluble material was re-moved by centrifugation and the supernatant analyzed further (7). Westernblotting of Pds1-GFP and Esp1-GFP was performed as described in ref. 49,using mouse monoclonal anti-GFP (Roche).

ACKNOWLEDGMENTS. We thank Scott Emr, Richard Kahn, Anita Corbett,Gerald Johnston, Alan Tartakoff, Vytas Bankaitis, Ilia Ouspenski, Marie-ClaudeMarsolier-Kergoat, Nevan Krogan, John McCusker, David Schild, Mike Gustin,Mike Snyder, Yoshiaki Kamada, and David Stillman for plasmids, advice, anddiscussion. Scott Emr suggested the possibility that vacuolar proteolysis couldbe involved, and Jonathan Weissman, Sean Collins, Maya Schuldiner, VladDinec, and other members of the Weissman laboratory provided an invalu-able sounding board when J.E.H. spent part of a sabbatical at University ofCalifornia, San Francisco (UCSF). We thank David Pellman for the use of hismicroscope. We thank Kathryn Patterson for help in early stages of thisproject. This work was supported by National Institutes of Health GrantGM61766 (to J.E.H.). J.C.H. was a Fellow of the Leukemia and LymphomaSociety. J.E.H. benefited from a sabbatical visit to UCSF and support fromUCSF. S.Y. is supported by a Massachusetts Life Sciences Consortium grant.

1. Clémenson C, Marsolier-Kergoat MC (2009) DNA damage checkpoint inactivation:Adaptation and recovery. DNA Repair 8(9):1101–1109.

2. Putnam CD, Jaehnig EJ, Kolodner RD (2009) Perspectives on the DNA damage andreplication checkpoint responses in Saccharomyces cerevisiae. DNA Repair 8(9):974–982.

3. Harrison JC, Haber JE (2006) Surviving the breakup: The DNA damage checkpoint.Annu Rev Genet 40:209–235.

4. Sandell LL, Zakian VA (1993) Loss of a yeast telomere: Arrest, recovery, and chromosomeloss. Cell 75(4):729–739.

5. Toczyski DP, Galgoczy DJ, Hartwell LH (1997) CDC5 and CKII control adaptation to theyeast DNA damage checkpoint. Cell 90(6):1097–1106.

6. Lee SE, et al. (1998) Saccharomyces Ku70, mre11/rad50 and RPA proteins regulateadaptation to G2/M arrest after DNA damage. Cell 94(3):399–409.

7. Pellicioli A, Lee SE, Lucca C, Foiani M, Haber JE (2001) Regulation of SaccharomycesRad53 checkpoint kinase during adaptation from DNA damage-induced G2/M arrest.Mol Cell 7(2):293–300.

8. Melo JA, Cohen J, Toczyski DP (2001) Two checkpoint complexes are independentlyrecruited to sites of DNA damage in vivo. Genes Dev 15(21):2809–2821.

9. Lee SE, Pellicioli A, Malkova A, Foiani M, Haber JE (2001) The Saccharomycesrecombination protein Tid1p is required for adaptation from G2/M arrest induced bya double-strand break. Curr Biol 11(13):1053–1057.

10. Lee SE, et al. (2003) Yeast Rad52 and Rad51 recombination proteins define a secondpathway of DNA damage assessment in response to a single double-strand break.MolCell Biol 23(23):8913–8923.

11. Leroy C, et al. (2003) PP2C phosphatases Ptc2 and Ptc3 are required for DNAcheckpoint inactivation after a double-strand break. Mol Cell 11(3):827–835.

12. Vaze MB, et al. (2002) Recovery from checkpoint-mediated arrest after repair ofa double-strand break requires Srs2 helicase. Mol Cell 10(2):373–385.

13. Guillemain G, et al. (2007) Mechanisms of checkpoint kinase Rad53 inactivation aftera double-strand break in Saccharomyces cerevisiae. Mol Cell Biol 27(9):3378–3389.

14. Eapen VV, Sugawara N, Tsabar M, Wu W-H, Haber JE (2012) The Saccharomycescerevisiae chromatin remodeler Fun30 regulates DNA end-resection and checkpointdeactivation. Mol Cell Biol 32(22):4727–4740.

15. Unal E, et al. (2004) DNA damage response pathway uses histone modification toassemble a double-strand break-specific cohesin domain. Mol Cell 16(6):991–1002.

16. Kim JA, Kruhlak M, Dotiwala F, Nussenzweig A, Haber JE (2007) Heterochromatin isrefractory to gamma-H2AX modification in yeast and mammals. J Cell Biol 178(2):209–218.

17. van Vugt MA, et al. (2004) Polo-like kinase-1 is required for bipolar spindle formationbut is dispensable for anaphase promoting complex/Cdc20 activation and initiation ofcytokinesis. J Biol Chem 279(35):36841–36854.

18. Demeter J, Lee SE, Haber JE, Stearns T (2000) The DNA damage checkpoint signal inbudding yeast is nuclear limited. Mol Cell 6(2):487–492.

19. Enserink JM, Smolka MB, Zhou H, Kolodner RD (2006) Checkpoint proteins controlmorphogenetic events during DNA replication stress in Saccharomyces cerevisiae.J Cell Biol 175(5):729–741.

20. Dyavaiah M, Rooney JP, Chittur SV, Lin Q, Begley TJ (2011) Autophagy-dependentregulation of the DNA damage response protein ribonucleotide reductase 1. MolCancer Res 9(4):462–475.

21. Bae H, Guan JL (2011) Suppression of autophagy by FIP200 deletion impairs DNAdamage repair and increases cell death upon treatments with anticancer agents. MolCancer Res 9(9):1232–1241.

22. Potenski CJ, Klein HL (2011) Molecular biology: The expanding arena of DNA repair.Nature 471(7336):48–49.

23. Rodriguez-Rocha H, Garcia-Garcia A, Panayiotidis MI, Franco R (2011) DNA damageand autophagy. Mutat Res 711(1–2):158–166.

24. Kroemer G, White E (2010) Autophagy for the avoidance of degenerative,inflammatory, infectious, and neoplastic disease. Curr Opin Cell Biol 22(2):121–123.

25. Robert T, et al. (2011) HDACs link the DNA damage response, processing of double-strand breaks and autophagy. Nature 471(7336):74–79.

26. Raymond CK, Howald-Stevenson I, Vater CA, Stevens TH (1992) Morphologicalclassification of the yeast vacuolar protein sorting mutants: Evidence for a prevacuolarcompartment in class E vps mutants. Mol Biol Cell 3(12):1389–1402.

27. Woolford CA, et al. (1993) Phenotypic analysis of proteinase A mutants. Implicationsfor autoactivation and the maturation pathway of the vacuolar hydrolases ofSaccharomyces cerevisiae. J Biol Chem 268(12):8990–8998.

28. Ciosk R, et al. (1998) An ESP1/PDS1 complex regulates loss of sister chromatid cohesionat the metaphase to anaphase transition in yeast. Cell 93(6):1067–1076.

29. Cohen-Fix O, Peters JM, Kirschner MW, Koshland D (1996) Anaphase initiation inSaccharomyces cerevisiae is controlled by the APC-dependent degradation of theanaphase inhibitor Pds1p. Genes Dev 10(24):3081–3093.

30. Jensen S, Segal M, Clarke DJ, Reed SI (2001) A novel role of the budding yeast separinEsp1 in anaphase spindle elongation: Evidence that proper spindle association of Esp1is regulated by Pds1. J Cell Biol 152(1):27–40.

31. Sarin S, et al. (2004) Uncovering novel cell cycle players through the inactivation ofsecurin in budding yeast. Genetics 168(3):1763–1771.

32. Agarwal R, Cohen-Fix O (2002) Mitotic regulation: The fine tuning of separaseactivity. Cell Cycle 1(4):255–257.

33. Agarwal R, Cohen-Fix O (2002) Phosphorylation of the mitotic regulator Pds1/securinby Cdc28 is required for efficient nuclear localization of Esp1/separase. Genes Dev 16(11):1371–1382.

34. Reggiori F, Wang CW, Stromhaug PE, Shintani T, Klionsky DJ (2003) Vps51 is part ofthe yeast Vps fifty-three tethering complex essential for retrograde traffic from theearly endosome and Cvt vesicle completion. J Biol Chem 278(7):5009–5020.

35. Gill DJ, et al. (2007) Structural insight into the ESCRT-I/-II link and its role in MVBtrafficking. EMBO J 26(2):600–612.

E48 | www.pnas.org/cgi/doi/10.1073/pnas.1218065109 Dotiwala et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

29,

202

0

36. Klionsky DJ (2007) Autophagy: From phenomenology to molecular understanding inless than a decade. Nat Rev Mol Cell Biol 8(11):931–937.

37. Kamada Y (2010) Prime-numbered Atg proteins act at the primary step in autophagy:Unphosphorylatable Atg13 can induce autophagy without TOR inactivation. Autophagy6(3):415–416.

38. Kamada Y, Sekito T, Ohsumi Y (2004) Autophagy in yeast: A TOR-mediated responseto nutrient starvation. Curr Top Microbiol Immunol 279:73–84.

39. Loewith R, Hall MN (2011) Target of rapamycin (TOR) in nutrient signaling andgrowth control. Genetics 189(4):1177–1201.

40. Kamada Y, et al. (2010) Tor directly controls the Atg1 kinase complex to regulateautophagy. Mol Cell Biol 30(4):1049–1058.

41. Dotiwala F, Haase J, Arbel-Eden A, Bloom K, Haber JE (2007) The yeast DNA damagecheckpoint proteins control a cytoplasmic response to DNA damage. Proc Natl AcadSci USA 104(27):11358–11363.

42. Vergés E, Colomina N, Garí E, Gallego C, Aldea M (2007) Cyclin Cln3 is retained at theER and released by the J chaperone Ydj1 in late G1 to trigger cell cycle entry. Mol Cell26(5):649–662.

43. Zhang T, Nirantar S, Lim HH, Sinha I, Surana U (2009) DNA damage checkpointmaintains CDH1 in an active state to inhibit anaphase progression. Dev Cell 17(4):541–551.

44. Hill E, Clarke M, Barr FA (2000) The Rab6-binding kinesin, Rab6-KIFL, is required forcytokinesis. EMBO J 19(21):5711–5719.

45. Gao W, Shen Z, Shang L, Wang X (2011) Upregulation of human autophagy-initiationkinase ULK1 by tumor suppressor p53 contributes to DNA-damage-induced cell death.Cell Death Differ 18(10):1598–1607.

46. Chen H, et al. (2011) The mTOR inhibitor rapamycin suppresses DNA double-strandbreak repair. Radiat Res 175(2):214–224.

47. Longtine MS, et al. (1998) Additional modules for versatile and economical PCR-basedgene deletion and modification in Saccharomyces cerevisiae. Yeast 14(10):953–961.

48. Ross-Macdonald P, Sheehan A, Roeder GS, Snyder M (1997) A multipurposetransposon system for analyzing protein production, localization, and function inSaccharomyces cerevisiae. Proc Natl Acad Sci USA 94(1):190–195.

49. Yoshida S, Bartolini S, Pellman D (2009) Mechanisms for concentrating Rho1 duringcytokinesis. Genes Dev 23(7):810–823.

Dotiwala et al. PNAS | Published online November 19, 2012 | E49

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Oct

ober

29,

202

0