Two familial ALS proteins function in prevention/repairof transcription-associated DNA damageSarah J. Hilla,b,c,1, Daniel A. Mordesd,e,f, Lisa A. Camerong, Donna S. Neubergh, Serena Landinib,c,i, Kevin Eggand,f,and David M. Livingstonb,c,i,1

aDepartment of Pathology, Brigham and Women’s Hospital, Boston, MA 02115; bDepartment of Genetics, Harvard Medical School, Boston, MA 02215;cDepartment of Cancer Biology, Dana-Farber Cancer Institute, Boston, MA 02215; dDepartment of Stem Cell and Regenerative Biology, The Harvard StemCell Institute, Harvard University, Cambridge, MA 02138; eDepartment of Pathology, Massachusetts General Hospital, Boston, MA 02114; fThe Stanley Center forPsychiatric Research, Broad Institute of Harvard and Massachusetts Institute of Technology, Cambridge, MA 02142; gConfocal and Light Microscopy Facility,Dana-Farber Cancer Institute, Boston, MA 02215; hBiostatistics and Computational Biology, Dana-Farber Cancer Institute, Boston, MA 02215; and iDepartment ofMedicine, Harvard Medical School, Boston, MA 02215

Contributed by David M. Livingston, October 1, 2016 (sent for review July 18, 2016; reviewed by Junjie Chen and Fred H. Gage)

Amyotrophic lateral sclerosis (ALS) is a progressive motor neurondysfunction disease that leads to paralysis and death. There iscurrently no established molecular pathogenesis pathway. Multipleproteins involved in RNA processing are linked to ALS, including FUSand TDP43, and we propose a disease mechanism in which loss offunction of at least one of these proteins leads to an accumulation oftranscription-associated DNA damage contributing to motor neuroncell death and progressive neurological symptoms. In support of thishypothesis, we find that FUS or TDP43 depletion leads to increasedsensitivity to a transcription-arresting agent due to increased DNAdamage. Thus, these proteins normally contribute to the preventionor repair of transcription-associated DNA damage. In addition, bothFUS and TDP43 colocalize with active RNA polymerase II at sites ofDNA damage along with the DNA damage repair protein, BRCA1,and FUS and TDP43 participate in the prevention or repair of R loop-associated DNA damage, a manifestation of aberrant transcriptionand/or RNA processing. Gaining a better understanding of the role(s)that FUS and TDP43 play in transcription-associated DNA damagecould shed light on the mechanisms underlying ALS pathogenesis.

DNA damage response | ALS | transcription | R loop

Amyotrophic lateral sclerosis (ALS) is a disease of both upperand lower motor neuron dysfunction that leads to progressive

paralysis and eventually death due to respiratory failure. No singlemodel of ALS disease pathogenesis has been revealed; however,multiple disease-associated genes are known (1). The variation infunction of these genes suggests that there may be multiple ALSmolecular subtypes. That said, the familial ALS gene product listis also enriched in protein groups with related functions.Mutations in multiple RNA processing genes, including FUS

(FUS RNA-binding protein), TDP43 (TAR DNA-binding pro-tein), SETX (senataxin), TAF15 (TATA box-binding protein-as-sociated factor 15), EWSR1 (EWS RNA-binding protein 1),HNRNPA1 (heterogeneous nuclear ribonucleoprotein A1), andHNRNPA2B1 (heterogeneous nuclear ribonucleoprotein A2/B1),among others, give rise to familial ALS (fALS) (1). FUS andTDP43 are currently the best studied, given that mutations inthese genes lead to classic histologic findings in ALS neural tissueand that similar histologic findings can be found in neural tissue ofsporadic ALS cases (2). At autopsy, cytoplasmic TDP43 inclusionsare found in ALS motor neurons from patients with (i) TDP43mutations and (ii) sporadic ALS (i.e., patients with no FUS,TDP43, or other known pathogenic mutation) (2). They are alsopresent in neurons in various parts of the brains of patients witheither sporadic or familial versions of the ALS-related disorder,fronto-temporal lobar dementia (FTLD) (2). At autopsy, FUSinclusions were detected in the cytoplasm of motor neurons ofFUS mutation-bearing fALS patients and in the cytoplasm ofneurons in various regions of the brain in some FTLD patients (2).FUS and TDP43 exhibit a wide range of functions, most of

which involve transcription and RNA processing. However, the

significance of their cytoplasmic inclusions in patients with sporadicand fALS is unclear with respect to disease pathogenesis. Conceiv-ably, nuclear TDP43 and/or FUS undergo a toxic loss of functionwhen the molecular pathways in which they participate are disrupted.By contrast, these cytoplasmic inclusions might also mirror a toxicgain of function, which, in turn, triggers motor neuron cell death.Animal models and cell-based assays have illuminated certain

functions of FUS and TDP43 in both cycling and terminally dif-ferentiated cells (3–5). Work on TDP43 has suggested roles for itin transcription regulation, mRNA splicing, mRNA transport, andstress granule formation among other functions (4). FUS also playsa role in multiple cellular processes including transcription regula-tion, mRNA splicing and transport, stress granule formation, ho-mologous DNA pairing through its DNA binding domain, andvarious forms of DNA damage repair (5). More specifically, FUSparticipates in both homologous recombination (HR) and non-homologous end joining (NHEJ)-directed DSB repair (6, 7). Theevidence suggesting a role for FUS in DNA repair is particularlyintriguing for ALS pathogenesis, given growing evidence that RNAbinding proteins are active in the prevention and repair of tran-scription-associated DNA damage (8–12). Other ALS-related RNA-binding proteins like SETX, EWSR1, and TAF15 have also beenlinked to DNA damage and repair, further suggesting the importanceof a breakdown in this process in ALS pathogenesis (11, 13–16).

Significance

Amyotrophic lateral sclerosis (ALS) is a neurodegenerative dis-ease in which progressive dysfunction of motor neurons leads toparalysis and death. There is currently no known pathway ofdisease pathogenesis. However, many genes with varyingfunctions have been linked to familial and sporadic forms ofALS suggesting multiple possible disease mechanisms. One setof gene products, including ALS-linked FUS and TDP43, func-tions in transcription and RNA processing. We find that defectsin transcription due to loss of function of FUS or TDP43 can leadto DNA damage, including in primary human neuronal cells,and hypothesize that dysfunction in their RNA processing rolesleads to DNA damage in motor neurons that, if incompletelyresolved, could contribute to motor neuron death and ALS.

Author contributions: S.J.H. designed research; S.J.H. performed research; D.A.M., L.A.C.,D.S.N., and K.E. contributed new reagents/analytic tools; S.J.H., D.S.N., and D.M.L. ana-lyzed data; S.L. aided S.J.H. in technical performance of experiments in Fig. 1; D.M.L.mentored S.J.H. in all aspects of the research; and S.J.H. and D.M.L. wrote the paperwith editing performed by all authors.

Reviewers: J.C., The University of Texas MD Anderson Cancer Center; and F.H.G., The SalkInstitute for Biological Studies.

The authors declare no conflict of interest.1To whom correspondence may be addressed. Email: sarah_hill@dfci.harvard.edu ordavid_livingston@dfci.harvard.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1611673113/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1611673113 PNAS | Published online November 14, 2016 | E7701–E7709

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 10

, 202

0

Depletion of various RNA binding/processing proteins triggersγH2AX (gamma H2A histone family, member X) focus forma-tion, a canonical marker of DNA damage (10). This result suggeststhat these proteins are important in preventing or repairing DNAdamage associated with transcription and RNA processing (10).Transcription-driven DNA damage can arise from (i) collapse oftranscription machinery and the subsequent evolution of double-strand DNA breaks (DSBs); (ii) collisions between transcriptionand replication machinery, again leading to DSBs; (iii) damage togenomic segments that have been opened up/prepared for tran-scription; and (iv) the formation and unnatural stabilization ofDNA-RNA hybrid R loop structures (8–10). R loops comprise asegment of a transcribed DNA strand annealed to the nascent RNAtranscript with the nontranscribed DNA strand looped out (8, 9).The latter can trigger the development of DNA damage (8, 9).Motor neurons are terminally differentiated and quiescent. Thus,

repair of DSBs cannot occur via HR and likely occurs, at least inpart, via the more error prone process, NHEJ, which operatesthrough much of the cell cycle (17). Accumulation of DNA damagein terminally differentiated motor neurons can eventually lead tomotor neuron cell death and potentially ALS. Thus, any mechanismthat leads to accumulating DNA damage, especially in G0 cells, is acandidate contributor to the pathogenesis of ALS.Given these possibilities, we attempted to study the role of FUS

and TDP43 in the prevention and/or repair of transcription-associ-ated DNA damage. A better understanding of the nature of DNAdamage associated with (i) a loss of FUS and TDP43 function and/or(ii) the role that these proteins play in preventing or repairing it, hasthe potential to generate insights into the pathogenesis of ALS.

ResultsDepletion of FUS and TDP43 Leads to Excessive Transcription Stalling.Arrest of transcription by the RNA Polymerase II (RNA Pol II)inhibitor, α-amanitin, triggers formation of γH2AX foci, a sign ofDNA damage. If depletion of a protein leads to increased sensitivityto α-amanitin–induced transcription arrest, the protein of interestmight be required for the prevention or repair of transcription ar-rest-driven DNA damage. Indeed, depletion of the tumor sup-pressor and DNA damage repair protein, BRCA1 (breast cancer 1),leads to increased α-amanitin sensitivity, and this effect is due toincreased DNA damage (18). We detected multiple roles forBRCA1 in the prevention and repair of transcription-associatedDNA damage, including in promoting the restart of transcriptionafter UV arrest and preventing/repairing R loop-driven DNAdamage (18). Thus, given their transcription/RNA processingassociation, we asked whether cells depleted of FUS or TDP43exhibit an increase in α-amanitin sensitivity.Specifically, we transfected cells with either a control siRNA

(siGL2) or with two different siRNAs that targeted each protein(Fig. 1 A and D). Then sensitivity to α-amanitin was assessed in acolony-forming assay (Fig. 1 B, C, E, and F). The results showthat depletion of FUS or TDP43 led to increased α-amanitinsensitivity, suggesting that loss of function of either protein leadsto defective prevention or repair of transcription arrest-associ-ated DNA damage.To test whether the increased α-amanitin sensitivity was due to

increased DNA damage, we performed alkaline comet assays onthe above-noted FUS- and TDP43-depleted cells (Fig. S1). Thisassay was chosen because it can recognize both single and DSBs,each a possible outcome of transcription arrest. Depletion ofeither FUS or TDP43 in α-amanitin–treated cells led to in-creased DNA breakage, again implying that these proteins sup-press or repair DNA damage associated with transcription arrest.

FUS Localizes at Sites of Transcription-Associated DNA Damage. Thefinding that FUS participates in the prevention or repair of tran-scription-associated DNA damage is consistent with the priorfinding that FUS localizes to DSBs induced by a UV laser (6, 7,

19). The latter observation suggests two functional possibilities:(i) that FUS participates in the repair of DSBs, per se, and (ii) thatit localizes to breaks located near sites of transcription-associatedDNA damage. In keeping with the latter, FUS functions in boththe soluble nuclear and chromatin fractions. In the chromatinfraction, it binds to the C-terminal domain of RNA Pol II andmodulates its phosphorylation on Serine 2, which helps to regulatethe transcription of certain genes (20, 21). FUS also colocalizeswith active (serine 2 phosphorylated) RNA Pol II in undamagedcells (22). There it likely aids in the transcription of a subset ofgenes, given that altered RNA Pol II Serine 2 phosphorylationoccurred when the protein bound to certain genes in fibroblastsfrom patients with FUS mutations (22). Given these results, itseems likely that FUS functions in the context of transcription-associated DNA damage.To determine whether FUS localizes at sites of transcription-

associated DNA damage, we searched for FUS at UV-inducedDNA damage sites that are associated with transcription arrest(18, 23). UV exposure induces the formation of active RNA PolII-containing nuclear foci that contain certain DNA damagemarkers such as phosphorylated-RPA (replication protein A),γH2AX, and BRCA1 (18). One possibility is that these foci areactive sites of transcription-associated DNA damage repair.Thus, we searched for FUS in post-UV RNA Pol II foci, using a

methanol–acetic acid fixation method that reveals concentratedchromatin-based complexes, e.g., those that contain RNA Pol II (23).The underlying hypothesis was that FUS localizes at these sites aspart of its role in transcription-associated DNA damage repair.Using FUS and RNA Pol II antibodies that are target specific

(Fig. S2), we were able to detect a significant increase in thenumber of UV-treated cells with colocalizing FUS and activeRNA Pol II foci compared with untreated controls (Fig. 2A). Thisobservation suggests that FUS localizes to sites of active tran-scription after UV damage (Fig. 2A).To test whether DNA damage was present at these sites, we

costained for FUS and the DNA damage-associated proteins,γH2AX (Fig. 2B), phosphorylated RPA (Fig. S3A), and BRCA1(Fig. S3B). A significant fraction of post-UV BRCA1-containingfoci contained active RNA Pol II, γH2AX, and phosphorylatedRPA, suggesting that a significant number of active RNA Pol IIfoci contain γH2AX and phosphorylated RPA and thus arose inassociation with transcription-associated DNA damage (18).Here we detected increased post-UV nuclear colocalization of

FUS with γH2AX (Fig. 2B), phosphorylated RPA (Fig. S3A), andBRCA1 (Fig. S3B), again suggesting that FUS localizes at tran-scription-associated DNA damage sites.These results imply that, after UV damage-induced transcrip-

tion arrest, FUS localizes at genomic sites where active tran-scription has been arrested by DNA damage, and it does sotogether with BRCA1 which is known to localize at such sites (18).Presumably, its role is to participate in the repair of this damagethrough its HR or NHEJ functions and/or to prevent the devel-opment of additional damage. FUS also colocalized with γH2AX(Fig. 2B), pRPA (Fig. S3A), and BRCA1 (Fig. S3B) in cells thatwere not ectopically damaged, suggesting that its baseline functionin cycling cells includes some form of endogenous DNA damageprevention or repair.

TDP43 Localizes at Sites of Transcription-Associated DNA DamageTogether with FUS. We also asked whether TDP43 concentratesat transcription-associated DNA damage sites. Using a specificTDP43 antibody (Fig. S4), we tested whether TDP43 alsocolocalizes with active RNA Pol II after UV damage. Indeed,more cells revealed colocalizing TDP43 and active RNA Pol IIfoci after UV damage than in an undamaged state (Fig. 3A),suggesting that TDP43, like FUS and BRCA1, also localizesto sites of transcription associated DNA damage.

E7702 | www.pnas.org/cgi/doi/10.1073/pnas.1611673113 Hill et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 10

, 202

0

TDP43 also localized at sites of transcription-associated DNAdamage, given its colocalization with γH2AX and phosphory-lated RPA in an increased percentage of UV-damaged cells (Fig.3B and Fig. S5A). Moreover, like FUS, TDP43 colocalized withγH2AX (Fig. 3B) and phosphorylated RPA (Fig. S5A) in un-

damaged cells, consistent with both proteins engaging in pre-venting or repairing spontaneously arising DNA damage.Finally, we tested whether FUS and TDP43 normally colocalize

and observed that, in the absence of ectopically induced damage,FUS and TDP43 foci overlapped significantly (Fig. S5B). This

A kDa

51392819

19197

64

14

51392819

1919764

14

FUS

Tubulin

WB=FUS

WB=Tubulin

U2O

S-siFUS #7

U2O

S-siGL2

U2O

S-siFU

S #8

kDa

51392819

1919764

14

51392819

1919764

14

WB=Tubulin

WB=TDP43 TDP43

TubulinU

2OS-siTD

P43 #2

U2O

S-siGL2

U2O

S-siTDP43 #3

U2O

S-siTDP43 #4

D

B

00.050.1

0.150.2

0.250.3

siGL2 siFUS #7 siFUS #8

uM

-Am

aniti

n

-Amanitin IC50s for FUS

p=0.007 p=0.02

C

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.351

10

100

uM -Amanitin

% V

iabl

e

siGL2siTDP43 #2siTDP43 #3siTDP43 #4

E

00.050.1

0.150.2

0.250.3

0.35

siGL2 siTDP43 #2 siTDP43 #3 siTDP43 #4

uM

-Am

aniti

n

-Amanitin IC50s for TDP43

p=0.024p=0.005

p=0.042

F

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.351

10

100

uM -Amanitin

% V

iabl

e

siGL2siFUS #7siFUS #8

Fig. 1. Depletion of fALS proteins leads to increased sensitivity to a transcription stalling agent. (A) Western blot demonstrating that FUS siRNAs deplete FUSprotein. (Top) Blot shows FUS protein levels in U2OS cells transfected with a control siRNA and the two FUS siRNAs that were also used in experiments depicted inB and C. (Bottom) This blot was stripped and stained with anti-tubulin antibody, which served as a loading control. (B) U2OS cells were transfected with the siRNAsused in A (the siGL2 control and two FUS-specific siRNAs), plated at a sufficient density for colony formation, and treated with varying doses of the transcriptioninhibiting agent α-amanitin for 24 h, at which point the media were removed and replaced with fresh non–drug-containing media. One week later, the cells werestained with crystal violet solution, and the colonies were counted. Dose–response curves and IC50s were generated from these counts. The experiment wasrepeated three separate times for each siRNA. The average IC50 from the three repetitions of the experiment for each siRNA is shown in the bar graph, with theerror bars representing the SD between the IC50s. The P values above the siFUS bars represent the significance of the difference between the siGL2 IC50 and eachsiFUS IC50 calculated using a paired, two-tailed t test in GraphPad Prism. (C) Average dose–response curves for the three repetitions of the colony-forming ex-periments described in B. The error bars at each dose represent the SEM between the three repetitions of the experiment for each siRNA. (D) Western blotdemonstrating that the three TDP43 siRNAs each deplete TDP43 protein compared with the siGL2 (firefly luciferase) control in U2OS cells (Top). This blot wasstripped and stained with tubulin as a loading control (Bottom). (E) U2OS cells were transfected with the siRNAs used in D and treated with α-amanitin as in B.Dose–response curves were generated from these results, and the data were analyzed as in experiments depicted in B and C. The P values above the siTDP43 barsrepresent the significance of the difference between the siGL2 IC50 and each siTDP43 IC50 calculated using a paired, two-tailed t test in GraphPad Prism.(F) Average dose–response curves from the three replicates of the colony formation experiments described in E. Error bars at each dose represent the SEMbetween the three repetitions of the experiment performed with each siRNA.

Hill et al. PNAS | Published online November 14, 2016 | E7703

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 10

, 202

0

result is consistent with the possibility that these structures are theproducts of certain common molecular events (Fig. S5B), as sug-gested by their known functional profiles (4, 5) and the resultsreported in Figs. 1–4. FUS and TDP43 colocalization also in-creased after UV treatment (Fig. S5B), which is consistent withthe notion that they both participate, possibly even together, indealing with transcription-associated DNA damage.

FUS and TDP43 Participate in Either the Prevention or Repair of R Loop-Associated DNA Damage. A particularly interesting form of tran-scription-associated DNA damage is that associated with R loops,as described above. Multiple proteins participate in their resolu-tion, including one protein linked to familial ALS, SETX (16, 24).When they are not resolved, e.g., due to loss of an R loop-resolvingprotein, unrepaired single-stranded breaks on the nontranscribedstrand can evolve into DSBs and lead to cell death (16, 24, 25).The notion that FUS and/or TDP43 is involved in the preventionor repair of R loop-associated DNA damage is of interest becauseof the possibility that these proteins associate with DNA damage

prevention or repair even in the setting of no exogenous damage(Figs. 2B and 3B). Moreover, other familial ALS proteins areinvolved in the prevention/repair of this type of damage (16, 24),and BRCA1, which colocalizes with FUS and likely TDP43at post-UV RNA Pol II DNA damage foci, also participates inthe prevention and/or repair of R loop-associated DNA damage(18, 26, 27).Thus, we asked whether FUS and/or TDP43 participate in the

prevention or repair of R loop-associated DNA damage bysearching for a decrease in DNA damage following FUS orTDP43 depletion in cells that express ectopic RNASEH1.RNASEH1 specifically digests DNA-RNA hybrids such as thosethat give rise to R loops (25). Assessing DNA damage markersin the setting of gene-specific depletion and RNASEH1 over-expression has previously been used to determine whether pro-teins are involved in R loop-associated DNA damage preventionor repair (18, 28).We assessed the results of RNASEH1 overexpression (Fig.

4A) together with FUS and TDP43 depletion or the transfection

0J4h

r pos

t 25J

0J

4hr p

ost 2

5J

A

B

DAPI FUS RPII Merge

DAPI FUS RPII Merge

4hr p

ost 2

5J

0

20

40

60

80

100

0J 4hr-25J

% c

ells

FUS-Pol II

p=0.0004DAPI FUS RPII Merge

DAPI γH2AX MergeFUS

DAPI FUS γH2AX Merge

DAPI FUS γH2AX Merge

4hr p

ost 2

5J

020406080

100

0J 4hr-25J

% c

ells

FUS-γH2AX

p=0.03

Fig. 2. FUS localizes to sites of post-UV transcription-associated DNA damage. (A and B) U2OS cells were treated with 0 J or 25 J, allowed to recover at 37 °Cfor 4 h, fixed with methanol:acetic acid, and costained for FUS and RNA Polymerase II phospho-S5 (RPII) (A) or FUS and γH2AX (B). The top row is the 0 J field,the middle row is the 4-h post-25 J field, and the bottom row is a magnified cell from the 4-h post-25 J field. Yellow arrows denote a cell in the 4-h post-25J field in which there is colocalization that has been cut out and magnified in the next row. Bar graphs representing the percentage of cells that containgreater than or equal to three FUS/RPII and FUS/γH2AX colocalizing foci are located next to the row of magnified single cell photographs. Bars represent theaverage of three separate experiments, and error bars represent the SD between experiments. P values denoting the significance of the difference betweenthe 0 J and 4-h post-25 J results are above the 4-h post-25 J bars. *Brightness was increased by 40% in every individual photograph in this figure usingMicrosoft PowerPoint. **These photographs are best viewed on a computer screen and not on a printed paper copy.

E7704 | www.pnas.org/cgi/doi/10.1073/pnas.1611673113 Hill et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 10

, 202

0

of an irrelevant siRNA, using an established immunofluorescence-associated fixation method (Fig. 4 B and C). We then costainedcells for γH2AX, as well as for RNASEH1 expression. The en-zyme was cotransfected with a control or FUS- or TDP43-specific siRNA.After control siRNA transfection, there was no significant

difference in the fraction of cells containing five or moreγH2AX foci between vector and RNASEH1 transfected cells(Fig. 4D and Fig. S6). However, in multiple replicate experi-ments, there was an incomplete but statistically significant de-crease in cells containing increased numbers of γH2AX fociin cells transfected with RNASEH1 compared with vector inboth siFUS and siTDP43 transfected cells (Fig. 4D and Fig. S6).This result implies that part of the DNA damage confronted byboth FUS and TDP43, like the other familial ALS proteinSETX (16), was R loop associated.

Motor Neurons Exhibit FUS Colocalization with UV-Induced RNAPolymerase Foci. All of the experiments in Figs. 1–4 were per-formed in U2OS cells, which are cycling cells derived from an

osteosarcoma. There are likely major differences in certain DNArepair responses (e.g., HR) between this rapidly cycling cancer cellline and quiescent motor neurons, the CNS cells affected in ALS.Therefore, it was important to test whether motor neurons behavesimilarly. This testing was accomplished using human cell line-derived motor neurons (29). Mature wild-type (WT) motor neu-rons were generated through directed differentiation as indicatedby many neurons being MAP2 and Islet-1 positive (Fig. S7).MAP2 is a neuronal marker, and Islet-1 is a motor neuron-specificmarker (30). Separate cell collections in wells derived from thesame individual differentiation event were left unirradiated orwere exposed to 25 J of UV-C and immunostained for FUS andeither active RNA Pol II or γH2AX.After immunostaining, observing red or green nuclear foci in

these experiments indicates that these structures are enriched forFUS (red) or RNA Pol II/γH2AX (green), respectively. Yellow fociindicate colocalization of FUS with either RNA Pol II or γH2AXand imply a shared function/response. FUS colocalized with UV-induced RNA Pol II pS5 and γH2AX foci 4 h after UV treatmentin some foci in some motor neuron nuclei in multiple experiments

0J

A

4hr p

ost 2

5J

B

0J4h

r pos

t 25J

020406080

100

0J 4hr-25J

% C

ells

TDP43-RPII

020406080

100

0J 4hr-25J

% C

ells

TDP43-γH2AX

p=0.002

p=0.002

DAPI RPIITDP43 Merge

DAPI RPIITDP43 MergeDAPI RPIITDP43 Merge

4hr p

ost 2

5J

DAPI γH2AXTDP43 Merge

DAPI γH2AXTDP43 MergeDAPI γH2AXTDP43 Merge

4hr p

ost 2

5J

Fig. 3. TDP43 localizes to sites of post-UV transcription-associated DNA damage. (A and B) U2OS cells were exposed to 0 J or 25 J, allowed to recover at 37 °Cfor 4 h, fixed with methanol:acetic acid, and costained for TDP43 and RNA Polymerase II phospho-S5 (RPII) (A) and TDP43 and γH2AX (B). The top row is the 0 Jfield; the middle row is the 4-h post-25 J field; and the bottom row is a magnified cell from the 4-h post-25 J field. Yellow arrows denote a cell in the 4-h post-25 J field in which there is colocalization that has been cut out and magnified. Bar graphs representing the percentage of cells with greater than or equal tothree colocalizing foci for each antibody pairing are next to the row of magnified single cell photos. Bars represent the average of three separate experi-ments, and error bars represent the SD between experiments. P values denoting the significance of the difference between the 0 J and 4-h post-25 J resultsare depicted above the 4-h post-25 J bars. *Brightness was increased by 40% in every individual photograph in Microsoft PowerPoint. **These photographsare best viewed on a computer screen and not on a printed paper copy.

Hill et al. PNAS | Published online November 14, 2016 | E7705

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 10

, 202

0

(Fig. 5 and Figs. S8–S11). Specifically, data from two individualexperiments (representative images from both experiments in Fig. 5and Figs. S8–S11) were pooled to reveal that exposure to UV lightincreased the percentage of foci with colocalization of RNA Pol IIpS5 and FUS (51.2% of 404 cells, 4-h post-25 J vs. 7.6% of 499cells, 0 J; P < 0.0001). Exposure to UV light also increased thepercentage of foci with colocalization of FUS and γH2AX (68.5%of 295 cells, 4-h post-25 J vs. 38.6% of 365 cells, 0 J; P < 0.0001).This result implies that elements of FUS operation at sites oftranscription-associated DNA damage are similar in G0 motorneurons, and, as shown above, in cycling cells. It also supports theview that this function occurs in motor neurons.Not every nuclear FUS focus colocalized with an RNA Pol II

or γH2AX focus after UV damage. FUS has many functions, not

all of which involve DNA damage (5), and FUS foci were de-tected in both UV-damaged and some untreated cells alike (Fig.5 and Figs. S8–S11). Conceivably, in UV damaged cells, the FUSfoci that did not colocalize with RNA Pol II or γH2AX mightreflect its operation during stages of DNA repair before γH2AXconcentration or at sites free of RNA Pol II or phospho-RNAPol II. Alternatively, they may represent FUS foci wherein FUSis operating in a non-DNA damage-associated manner.

DiscussionThe results of α-amanitin sensitivity, comet, RNASEH expression,and multiple DNA damage-linked immunofluorescence analysesrepeatedly suggest that both FUS and TDP43 are necessary forthe prevention or repair of transcription-associated DNA

siG

L2-V

ecto

r

DAPI RNASEH1

DAPI RNASEH1

siG

L2-R

NH

siG

L2-v

ecsi

FUS

#7-v

ecsi

FUS

#8-v

ec

DAPIFUS

DAPI FUS

DAPI

siTD

P43

#2-

RN

Hsi

TDP

43 #

3-R

NH

siG

L2-R

NH

DAPI TDP43

DAPI TDP43

DAPI TDP43

FUS

0

10

20

30

40

50

60

70

80

90

100

% C

ells

with

≥ 5

γH

2AX

foci GL2-vector

GL2-RNHFUS #7-vectorFUS #7-RNASEHFUS #8-vectorFUS #8-RNASEHTDP43 #2-vectorTDP43 #2-RNASEHTDP43 #3-vectorTDP43 #3-RNASEH

A

B C

D

Fig. 4. FUS and TDP43 are involved in the prevention or repair of R loop-associated DNA damage. (A) Representative images of U2OS cells cotransfected withsiGL2 siRNA and either an empty vector (vec) or an RNASEH1-encoding plasmid (RNH) and stained for RNASEH1. (B) Extra coverslips from one of the four replicatesof the RNASEH experiment not used for γH2AX assessment that were cotransfected with various siRNAs and empty vector (vec) were stained for FUS to assessdepletion of FUS protein. Representative images are shown with DAPI on the left and FUS on the right. (C) Extra coverslips from one of the four replicates of theRNASEH experiment not used for γH2AX assessment cotransfected with various siRNAs and RNASEH1 (RNH) were stained for TDP43 to assess depletion of TDP43protein. Representative images are shown with DAPI on the left and TDP43 on the right. (D) Dot plot representing the percentage of cell nuclei with greater thanor equal to five γH2AX foci after cotransfection with various siRNAs and with either empty vector or RNASEH for each of four individual experiments (a minimumof 200 nuclei were counted in each experiment for each condition/siRNA). Each dot corresponds to a single experiment. P values assessing the significance of thedifference between the vector and RNASEH1 value for each individual siRNA over the four replicates were calculated using a t test and are as follows: siGL2 =0.248, siFUS #7 = 0.018, siFUS #8 = 0.010, siTDP43 #2 = 0.002, siTDP43 #3 = 0.029. *Brightness was increased by 20% in every individual photo in this figure usingMicrosoft PowerPoint. **These photographs are best viewed on a computer screen and not on a printed paper copy.

E7706 | www.pnas.org/cgi/doi/10.1073/pnas.1611673113 Hill et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 10

, 202

0

damage, e.g., in the form of R loop-induced disorder (Figs. 1–5).These findings are compatible with the hypothesis of a role forDNA damage prevention and/or repair in suppressing both fa-milial and sporadic ALS. Consistent with this hypothesis, FUScolocalized with RNA Pol II pS5 at UV-induced DNA damagesites in human motor neurons (Fig. 5 and Figs. S8–S11).Human ALS tissue studies and mouse models have not yet

conclusively shown that markers of DNA damage coincide withsigns of impending cell death in motor neurons (6, 31). However,these results do not rule out the possibility that unrepaired tran-scription-associated DNA damage contributes to motor neurondeath in ALS. Based on work described here and that of others(5, 6), it is reasonable to propose a model of ALS pathogenesisin which various ALS proteins, such as FUS and TDP43, pre-vent or repair various forms of transcription-associated DNAdamage e.g., R loop-associated damage. Perhaps, through the

stable formation of hydrogel or fibrous states (32, 33), mutatedALS-related proteins with low complexity (LC) domains likeFUS, TDP43, EWSR1, or TAF15 lose an important nuclearfunction and thereby fail to respond to transcription-associatedDNA damage. This kind of damage is known to be associatedwith excessively stable R loops, stalled transcription machinery,and stable open DNA.In postmitotic neurons, DSBs are repaired by error-prone

NHEJ machinery and not by HR, which requires intact sisterchromatids and hence DNA replication (34). Over time, thesesites may accumulate mutations, genome rearrangements, and/orchromatid breaks. Enough genomic disorder might accumulateto result in or to accelerate cell death, the major feature of ALS.This genome disorder-induced cell death might be induced to-gether with the coparticipation of other as yet undetected neu-rotoxic events (35). Determining whether some forms or stages

0J4h

rpos

t25J

4hrp

ost2

5J

DAPI FUS RPII Merge

FUS MergeRPIIDAPI

DAPI MergeRPIIFUS

0J4h

rpos

t25J

4hrp

ost2

5J

DAPI

DAPI

DAPI

Merge

Merge

Merge

FUS

FUS

FUS γH2AX

γH2AX

γH2AX

0102030405060

0J 4hr-25J

%C

ells

FUS-RPIIp<0.0001

010203040506070

0J 4hr-25J

%C

ells

FUS-γH2AXp<0.0001

B

A

Fig. 5. FUS localizes to sites of post-UV transcription-associated DNA damage in WT motor neurons. (A and B) Motor neurons were exposed to 0 J or 25 J,allowed to recover at 37 °C for 4 h, fixed with methanol:acetic acid, and costained for FUS and RNA Pol II pS5 (RPII) (A) or FUS and γH2AX (B). Images wereobtained with a spinning disk confocal system at 100×. In each panel, the top row is the 0 J field, the middle row is the 4-h post-25 J field, and the bottom rowis a magnified cell from the 4-h post-25 J field. Yellow arrows denote a cell in the 4-h post-25 J field in which there is significant albeit not completecolocalization. A single cell, denoted by the yellow arrow, was excised from the relevant photographs and magnified in the row of photographs below them.Bar graphs representing the percentage of cells that contain greater than or equal to one FUS/RPII or greater than or equal to three FUS/γH2AX colocalizingfoci are located next to the row of magnified single cell photographs. Bars represent pooled data from two individual experiments. P values denoting thesignificance of the difference between the 0 J and 4-h post-25 J results are above the 4-h post-25 J bars. *Brightness has been uniformly increased in everyphotograph using Adobe Photoshop. **These photographs are best viewed on a computer screen and not in a printed paper copy.

Hill et al. PNAS | Published online November 14, 2016 | E7707

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 10

, 202

0

of ALS are, at least in part, driven by ongoing transcription-as-sociated DNA damage will require future work.Based on the results described here and the work of others, there

is already support for this view of ALS pathogenesis, especially withrespect to FUS-related cases. Our results fit with earlier observa-tions showing that FUS localizes to sites of UV-laser induced DSBsand supporting a role for FUS in DNA repair (6, 7, 19).FUS is recruited to DSBs in proliferating cells where it forms a

liquid phase compartment and where it is important in D loopformation in the initial phase of HR and in recruiting repairproteins to DSBs (6, 33, 36). In those studies, FUS localized tothese linear arrays of DSBs in a PARP-dependent manner beforethe appearance of markers of DNA damage or proteins involved inthe repair of DSBs (6, 7, 13). In addition, these repair phenomenadecreased in the setting of FUS depletion, suggesting that FUS isnecessary early on after damage has developed to stabilize dam-aged DNA, prepare it for repair, attract repair proteins, and/orinduce modifications needed to recruit other repair proteins (6).The observation that FUS localizes at sites of transcription-associated DNA damage, colocalizes with markers of DNA dam-age in the absence of exogenously generated DNA damage, andparticipates in the prevention/repair of R loop-induced DNAdamage fits with these prior findings. It is also consistent with thenotion that FUS functions, at least in part, in response to tran-scription-associated DNA damage.The finding that FUS localizes at sites of transcription-associ-

ated DNA damage after UV exposure suggests that it is recruitedto DNA damage structures and once there, participates in therepair process directly, and/or marks or stabilizes the damagedstructures for repair. Given existing observations from this workand others (6), FUS may arrive relatively early after damage hasoccurred at these sites, in part, to stabilize an “open” transcriptionstructure, because (i) FUS forms liquid compartments at sites ofDNA damage in cells (33) and (ii) without FUS recruitment, otherDNA damage repair proteins, like NBS1 or Ku70, operate in-efficiently (6). The fact that FUS did not colocalize with everyRNA Pol II pS5 or γH2AX focus in a single U2OS or motor neuronnucleus (Figs. 2 and 5 and Figs. S8–S11) is not surprising. FUSexerts multiple nuclear functions (5). Thus, not every FUS focus in acell need represent its participation in DNA repair. Alternatively, inUV-treated cells, where FUS was one of the earlier proteins toarrive, RNA Pol II pS5 or γH2AX foci may not yet have developed.In addition, it is conceivable that FUS marks sites of old, poorlyrepaired DNA damage that are no longer marked by γH2AX.Furthermore, instead of being recruited, perhaps FUS is al-

ready present at certain chromatin sites before damage develops,having processed along the DNA together with the transcriptionmachinery. Conceivably, such a process prepares it to deal withimpending chromatin damage. Whatever the case, FUS is knownto interact with active RNA Pol II and to regulate its phos-phorylation (20–22). Moreover, its nuclear staining pattern andthat of active RNA Pol II have been shown to overlap, in part, inunperturbed cell lines (20–22).We also found that TDP43, like FUS, plays a role in transcrip-

tion-associated DNA damage prevention and repair, in particular inR loop-associated DNA damage, and might share a role in stabi-lizing transcription-associated damage or aiding in its repair withFUS (Figs. 1, 3, and 4). In particular, TDP43 and FUS significantlycolocalized in foci before exogenously generated DNA damage andcolocalized even more so after DNA damage, suggesting the exis-tence of certain shared functions for these proteins (Fig. S5B).Some of these spontaneously arising foci in cells that were not

exposed to exogenous DNA damaging agents also colocalized withγH2AX and pRPA, again suggesting that one of the baselinefunctions of these structures may be in endogenous DNA repair. Itis possible that some of the other RNA binding proteins linked toALS through genetic studies, including EWSR1 and TAF15, alsoplay a role in one or more DNA repair processes and, when mu-

tated, contribute to ALS development via a similar DNA damage-driven process (1).A better understanding of the function of these RNA-binding

and transcription/RNA processing-associated ALS proteins couldalso aid in identifying and generating new biomarkers of the dis-ease that would help in making a definitive ALS diagnosis, intracking disease progression, and ideally, in directing the appli-cation of mechanism-based therapies in the future.

Materials and MethodsA more complete description is available in SI Materials and Methods.

α-Amanitin Sensitivity Assay. The assay was performed as described previously(18). Cells were plated on day 1, transfected on days 2 and 3, and plated atappropriate densities for colony formation on day 4. Enough cells to formbetween 100 and 300 colonies were plated in triplicate for each transfectionat each dose. Four hours after plating, the media were changed to mediacontaining various doses of α-amanitin (Sigma Cat. #A2263) for 24 h, atwhich point the cells were washed once with PBS, and fresh media con-taining no α-amanitin were added. The cells were grown at 37 °C until ap-propriate sized colonies had formed and then stained with crystal violet.Colonies were counted using a Microbiology International ProtoCOL colonycounter. Statistical analysis was performed, as described in SI Materials andMethods, using GraphPad Prism software.

Comet Assays. Cells were plated on day 1 and transfected with control orgene-specific siRNAs on days 2 and 3. On day 4, the medium was changed tomedium containing 0.35 μM α-amanitin or an equivalent volume of water,and the cells were incubated in it for 24 h at 37 °C. On day 5, alkaline cometassays were performed on these cells using a Trevigen CometAssay Kit (VWRCat. #95036-942) according to the manufacturer’s protocol. Analysis wasperformed as described in SI Materials and Methods.

Motor Neuron Differentiation and Immunofluorescence. Motor neurons werederived from the Hues63 cell line based on a previously described protocol(29) modified to use adherent cells, without producing embryoid bodyprecursors, and as follows. Stem cells were plated on matrigel (Corning) andcultured in mTeSR1 (Stemcell Technologies) until confluent. Then, the mediawas changed to neuron differentiation media: DMEM/F12:neurobasal 1:1(Life Technologies) supplemented with nonessential amino acids, glutamax,B27 (Life Technologies), and N2 (Stemcell Technologies). From day 0 to day 6,neuron differentiation media was supplemented with 1 μM retinoic acid(Sigma Aldrich), 1 μM smoothened agonist (DNSK), 0.1 μM LDN-193189(DNSK), and 10 μM SB-431542 (DNSK). From day 7 to day 14, neuron dif-ferentiation media was supplemented with 1 μM retinoic acid, 1 μMsmoothened agonist, 4 μM SU-5402 (DNSK), and 5 μM DAPT (DNSK). Motorneurons were plated on poly-D-lysine and laminin-coated chamber slides(Corning) in neurobasal media supplemented with nonessential amino acids,glutamax, N2, B27, and neurotrophic factors (BDNF, CNTF, GDNF). Afterbeing plated, they were allowed to mature at least 2 wk before UV treat-ment and immunofluorescence staining. For Islet-1 staining in which no UVtreatment was performed, motor neurons on chamber slides were fixed in4% paraformaldehyde/PBS and processed for immunofluorescent stainingwith an Islet-1 antibody (Developmental Studies Hybridoma Bank #40.2D6).

UV Immunofluorescence. For all experiments, cells were plated on coverslipson day 1. The fixation and staining method are briefly reviewed here (18, 23).UV treatment was performed as described previously (18). After varioustreatments or transfections, cells were washed with PBS, fixed in chilledmethanol:acetic acid (3:1) at 4 °C for 15 min and then washed with PBS. Cellswere exposed to blocking solution [1 mg/mL BSA, 5% (vol/vol) normal goatserum, 1% Triton X-100, and PBS] at 37 °C for 30 min and then washed inPBS. They were incubated in primary and then secondary antibodies dilutedin blocking solution. Finally, cells were mounted with mounting mediumcontaining DAPI (Vector Laboratories Cat. #H-1200) and then viewed andphotographed. U2OS cells were photographed using a Zeiss fluorescentmicroscope. Motor neurons were photographed with the 100× objective of aspinning disk confocal from Yokogawa attached to a Nikon Ti inverted mi-croscope. See SI Materials and Methods for details of statistical analysis.

RNASEH Assay. This assay was performed as described previously (18, 28), anddetailed methods are provided in SI Materials and Methods.

E7708 | www.pnas.org/cgi/doi/10.1073/pnas.1611673113 Hill et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 10

, 202

0

ACKNOWLEDGMENTS. We thank Drs. Merit Cudkowicz and MatthewFrosch for valuable discussions and help. Confocal microscopy images forthis study were acquired in the Confocal and Light Microscopy Core Fa-cility at the Dana-Farber Cancer Institute with the assistance of L.A.C.Funding was provided by Target ALS (K.E. and D.A.M.) and NIH TrainingGrant 5T32CA009216 (to D.A.M.). S.J.H. was funded initially by US De-

partment of Defense Breast Cancer Research Program FellowshipW81XWH-08-1-0748 and subsequently by National Cancer Institute Fel-lowship 1F30CA167895-01. S.J.H., S.L., and D.M.L. were funded by grantsfrom the ALS Therapy Alliance, Harvard NeuroDiscovery Center Neuro-degenerative Disease Pilot Study Grants Program, and the MurrayWinston Foundation.

1. Marangi G, Traynor BJ (2015) Genetic causes of amyotrophic lateral sclerosis: Newgenetic analysis methodologies entailing new opportunities and challenges. Brain Res1607:75–93.

2. Lagier-Tourenne C, Polymenidou M, Cleveland DW (2010) TDP-43 and FUS/TLS:Emerging roles in RNA processing and neurodegeneration. Hum Mol Genet 19(R1):R46–R64.

3. Ittner LM, et al. (2015) FTD and ALS: Translating mouse studies into clinical trials. NatRev Neurol 11(6):360–366.

4. Janssens J, Van Broeckhoven C (2013) Pathological mechanisms underlying TDP-43driven neurodegeneration in FTLD-ALS spectrum disorders. Hum Mol Genet 22(R1):R77–R87.

5. Sama RR, Ward CL, Bosco DA (2014) Functions of FUS/TLS from DNA repair to stressresponse: Implications for ALS. ASN Neuro 6(4):1759091414544472.

6. Wang WY, et al. (2013) Interaction of FUS and HDAC1 regulates DNA damage re-sponse and repair in neurons. Nat Neurosci 16(10):1383–1391.

7. Mastrocola AS, Kim SH, Trinh AT, Rodenkirch LA, Tibbetts RS (2013) The RNA-bindingprotein fused in sarcoma (FUS) functions downstream of poly(ADP-ribose) polymerase(PARP) in response to DNA damage. J Biol Chem 288(34):24731–24741.

8. Aguilera A, García-Muse T (2012) R loops: From transcription byproducts to threats togenome stability. Mol Cell 46(2):115–124.

9. Skourti-Stathaki K, Proudfoot NJ (2014) A double-edged sword: R loops as threats togenome integrity and powerful regulators of gene expression. Genes Dev 28(13):1384–1396.

10. Paulsen RD, et al. (2009) A genome-wide siRNA screen reveals diverse cellular pro-cesses and pathways that mediate genome stability. Mol Cell 35(2):228–239.

11. Izhar L, et al. (2015) A systematic analysis of factors localized to damaged chromatinreveals PARP-dependent recruitment of transcription factors. Cell Reports 11(9):1486–1500.

12. Adamson B, Smogorzewska A, Sigoillot FD, King RW, Elledge SJ (2012) A genome-wide homologous recombination screen identifies the RNA-binding protein RBMX asa component of the DNA-damage response. Nat Cell Biol 14(3):318–328.

13. Jungmichel S, et al. (2013) Proteome-wide identification of poly(ADP-Ribosyl)ationtargets in different genotoxic stress responses. Mol Cell 52(2):272–285.

14. Hurov KE, Cotta-Ramusino C, Elledge SJ (2010) A genetic screen identifies the TripleT complex required for DNA damage signaling and ATM and ATR stability. Genes Dev24(17):1939–1950.

15. Paronetto MP, Miñana B, Valcárcel J (2011) The Ewing sarcoma protein regulates DNAdamage-induced alternative splicing. Mol Cell 43(3):353–368.

16. Skourti-Stathaki K, Proudfoot NJ, Gromak N (2011) Human senataxin resolves RNA/DNA hybrids formed at transcriptional pause sites to promote Xrn2-dependent ter-mination. Mol Cell 42(6):794–805.

17. Pan L, Penney J, Tsai LH (2014) Chromatin regulation of DNA damage repair andgenome integrity in the central nervous system. J Mol Biol 426(20):3376–3388.

18. Hill SJ, et al. (2014) Systematic screening reveals a role for BRCA1 in the response totranscription-associated DNA damage. Genes Dev 28(17):1957–1975.

19. Rulten SL, et al. (2014) PARP-1 dependent recruitment of the amyotrophic lateralsclerosis-associated protein FUS/TLS to sites of oxidative DNA damage. Nucleic AcidsRes 42(1):307–314.

20. Schwartz JC, et al. (2012) FUS binds the CTD of RNA polymerase II and regulates itsphosphorylation at Ser2. Genes Dev 26(24):2690–2695.

21. Yang L, Gal J, Chen J, Zhu H (2014) Self-assembled FUS binds active chromatin andregulates gene transcription. Proc Natl Acad Sci USA 111(50):17809–17814.

22. Schwartz JC, et al. (2014) FUS is sequestered in nuclear aggregates in ALS patientfibroblasts. Mol Biol Cell 25(17):2571–2578.

23. Espinosa JM, Verdun RE, Emerson BM (2003) p53 functions through stress- and pro-moter-specific recruitment of transcription initiation components before and afterDNA damage. Mol Cell 12(4):1015–1027.

24. Mischo HE, et al. (2011) Yeast Sen1 helicase protects the genome from transcription-associated instability. Mol Cell 41(1):21–32.

25. Li X, Manley JL (2005) Inactivation of the SR protein splicing factor ASF/SF2 results ingenomic instability. Cell 122(3):365–378.

26. Bhatia V, et al. (2014) BRCA2 prevents R-loop accumulation and associates with TREX-2 mRNA export factor PCID2. Nature 511(7509):362–365.

27. Hatchi E, et al. (2015) BRCA1 recruitment to transcriptional pause sites is required forR-loop-driven DNA damage repair. Mol Cell 57(4):636–647.

28. Herrera-Moyano E, Mergui X, García-Rubio ML, Barroso S, Aguilera A (2014) The yeastand human FACT chromatin-reorganizing complexes solve R-loop-mediated tran-scription-replication conflicts. Genes Dev 28(7):735–748.

29. Amoroso MW, et al. (2013) Accelerated high-yield generation of limb-innervatingmotor neurons from human stem cells. J Neurosci 33(2):574–586.

30. Montoya G JV, et al. (2009) Muscle-conditioned media and cAMP promote survivaland neurite outgrowth of adult spinal cord motor neurons. Exp Neurol 220(2):303–315.

31. Qiu H, et al. (2014) ALS-associated mutation FUS-R521C causes DNA damage and RNAsplicing defects. J Clin Invest 124(3):981–999.

32. Murakami T, et al. (2015) ALS/FTD mutation-induced phase transition of FUS liquiddroplets and reversible hydrogels into irreversible hydrogels impairs RNP granulefunction. Neuron 88(4):678–690.

33. Patel A, et al. (2015) A liquid-to-solid phase transition of the ALS protein FUSaccelerated by disease mutation. Cell 162(5):1066–1077.

34. Sharma S (2007) Age-related nonhomologous end joining activity in rat neurons.Brain Res Bull 73(1-3):48–54.

35. Ojo JO, Mouzon BC, Crawford F (2016) Repetitive head trauma, chronic traumaticencephalopathy and tau: Challenges in translating from mice to men. Exp Neurol275(Pt 3):389–404.

36. Ceccaldi R, Rondinelli B, D’Andrea AD (2016) Repair pathway choices and conse-quences at the double-strand break. Trends Cell Biol 26(1):52–64.

37. Scully R, et al. (1996) Location of BRCA1 in human breast and ovarian cancer cells.Science 272(5258):123–126.

Hill et al. PNAS | Published online November 14, 2016 | E7709

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 10

, 202

0