DNA damage and : target gene selection and signaling dynamics

Jacob Stewart-Ornstein Dept. Computational and Systems Biology University of Pittsburgh & Hillman Cancer Center Stress is Universal

Environmental Fluctuations Poisons = Infection Radiation Stress Response

Molecular Pathways that respond to cellular damage and:

(1) Adjust the cell to its new environment

(2) Repair Any Damage Stress responses are ubiquitous and conserved

Conserved from yeast, Human

Nutrient (m)TOR, RAS, GCN2/4 Heat HSF1 Oxidative NRF2 Hypoxia HIF1 ER IRE1/XBP1, PERK, ATF6 DNA damage ATM/ATR, tp53

Vihervaara… Lis, Nat. Rev. Gen., 2018 Growth Rate Response Stress responses act to restore homeostasis Time (~ Time hrs ) UPR response in Yeast Budding Pincus et al., 2010, PLOS 2010, Bio et al., Pincus 2mM DTT 2mM DTT 1mM Response Response In multicellular systems stressresponses specify cell fate Cell death Cell of growthof Resumption Damage Cell death Cell Dose of DNA damage determines cell fate

damage

γH2AX Dapi

Cell cycle arrest + Repair Or Senescence How can the cell determine its DNA damage status?

damage

γH2AX Dapi

Apoptosis arrest Or + Repair Senescence

Appropriate # breaks Transcriptional response Tissue specificity of DNA damage signaling

# breaks Cell fate

Small Intestine Spleen (5Gy IR) (5Gy Tunel , DAPI Outline

(1) How does the cell respond proportionately to DNA damage?

(2) How universal is the transcriptional response to DNA damage? When cells are treated with radiation

IR dose

# Breaks (~dose2)

γH2AX

Viability

DAPI What is the transcriptional response to IR?

IR dose

# Breaks (~dose2)

γH2AX

Transcriptional Response?

Linear Saturating Response Response

dose dose What signaling Pathways are activated by DNA damage?

Human Cell Lines

x12

RNAseq + Ionizing Radiation What pathways are enriched? Number of of Number Genes

Z-Score ΔGene Expression The transcriptional response to radiation is dominated by p53

Human Cell Lines

x12

p53 Number of of Number Genes p53 (e-21) Highly conserved DNA damage responsive factor

Z-Score ΔGene Expression The transcriptional response to radiation is dominated by p53

Human Cell Lines Mouse Tissues

x12 x5 Number of of Number Genes p53 (e-21) p53 (e-35)

Z-Score ΔGene Expression Z-Score ΔGene Expression Knockdown of p53 largely eliminates the short term transcriptional response to radiation in epithelial cells

Significantly induced transcripts

6

5

4

3

2

1

Number of of Number Genes 0

p53 (e-21) Log2 Fold Change p53sh 90% (43/48) -1 p53 dependent -2 1 2 3 4 5 6

Log2 Fold Change WT Z-Score ΔGene Expression p53 Determines DNA damage Sensitivity

thymocytes

p53-/-

p53+/-

p53+/+

Lowe… Jacks, 1993, nature p53 Determines DNA damage Sensitivity

thymocytes 2 p53

1.5 -/- p53 1

0.5

0

-0.5

p53+/- -1 Treated Fitness -1.5 +/+ p53 -2 -2 -1 0 1 2 Untreated Fitness

CAS9 Screen A549

Lowe… Jacks, 1993, nature p53 is the key DNA damage responsive transcription factor

DNA

ATM Cell Cycle Arrest Oxidative Stress Apoptosis p53 Hypoxia

DNA Repair Translational Stress DeLeo et al., 1979 Crawford et al., 1979 Metabolism Finlay et al., 1989 Farmer et al., 1992 El-Deiry et al., 1993 Harper et al., 1993 … How does the cell respond transcriptionally to DNA damage?

IR dose

# Breaks (~dose2)

γH2AX p53? p53?

dose dose Acute response to DNA damage is not dose sensitive

+DNA damage p53 Log2 Dose p53

dose Longer Term response to DNA damage is dose dependent p53 Log2 Dose dose Across 12 human cell lines the acute p53 response is dose invariant, but the longer term response depends on DNA damage Models of p53 signaling Duration of signaling is dose sensitive Response Response

Need real time reporters of p53 signaling! A live cell reporter of p53 activity reveals complex dynamics

MCF7 (breast cancer) p53-YFP In response to DNA damage a negative feedback loop drives oscillations of p53

DNA

ATM

p53

p53-YFP In response to DNA damage a negative feedback loop drives oscillations of p53

60 p53

50

DNA YFP(AU) 40

30 P53 - 0 5 10 15 20 25 30 35 250 MDM2 ATM 200

150

100 p53 0 5 10 15 20 25 30 35

p53-YFP MDM2 In response to DNA damage a negative feedback loop drives oscillations of p53

60 p53

50

DNA YFP(AU) 40

30 P53 - 0 5 10 15 20 25 30 35 250 MDM2 ATM 200

150

100 p53 0 5 10 15 20 25 30 35

P53->MDM2 P53 period ~5hr MDM2 1 0.6

0.4 0.5

0.2

0 0 Autocorrelation Crosscorrelation

-0.5 -0.2 0 2 4 6 8 0 1 2 3 4 Time (hrs) Time (hrs) Recapitulates results from Lahav, 2004 Given oscillations, how is dose encoded?

DNA DNA

ATM ATM p53 vs p53

MDM2 MDM2 In MCF7 number of ‘pulses’ of p53 is controlled by dose Signaling is duration encoded

4

3.5 Low IR (1Gy)

3

2.5

2

1.5 P53 (AU)

1

0.5

0

-0.5 0 5 10 15 20 25 Dose proportional # Pulses 4 Dose is not proportional to Amplitude 3.5 High IR (8Gy) 3

2.5

2

1.5

1 P53 (AU)

0.5

0

-0.5 0 5 10 15 20 25 Time after IR (hrs) p53-YFP (AU) Dose ( Analysis of p53 dynamics across 12 cell lines cell 12 across dynamics p53 of Analysis Gy ) Low IR (1Gy) IR Low Time High IR (8Gy) IR High Time p53-YFP (AU) Other cell lines show show both duration cell lines controlOther amplitude and Dose ( Gy ) Low IR (1Gy) IR Low Time High IR (8Gy) IR High Time In A549s amplitude of p53 is regulated by dose

6 Low IR (1Gy) Low IR (1Gy) 5

4

3 P53 (AU) 2

1

0

0 5 10 15 20 25

10 High IR (8Gy) 8 High IR (8Gy)

6

4 P53 (AU) 2

0

0 5 10 15 20 25 Time after IR (hrs) Why do some cell lines show amplitude vs duration response to DNA damage dose?

Low IR (1Gy) High IR (8Gy)

Duration

MCF7

Time

A549 Amplitude Chemical Screen for p53 dose response behavior

High IR (8Gy)

Kinase inhibitor screen for dynamics P53 (AU)

Drug screen P53 (AU)

Time In maximum A549 amplitude of p53 is regulated by dose

High IR (8Gy) P53 (AU)

Drug screen P53 (AU)

Time Cell types with low ATM levels show reduced dynamic range In p53 signaling

DNA MCF7 ATM act ATM p53

ATM pATM DNA damage (AU)

p53 A549 ATM act ATM p53 MDM2 DNA damage (AU) Cell types with low ATM levels show reduced dynamic range In p53 signaling

8Gy IR

A549 Time UO31 UACC62 6

5

4 MCF7 3 HCT116

2 frequency

1

0 -4 -3 -2 -1 0 1 2

Log ATMact (AU) A general model reconciling p53 response and ATM activity

DNA DNA*

ATM p53

MDM2

Stewart-Ornstein and Lahav., Science Signaling, 2017 Stewart-Ornstein and Lahav., Current Opinion is Systems Biology, 2017 Outline

(1) How does the cell respond proportionately to DNA damage?

-duration or amplitude of p53 signaling are proportional to DNA damage -balance between amplitude and duration signaling is cell lines specific -ATM is a regulator of the dynamic range of p53

(2) Do all cells respond to DNA damage in the same way? Is the p53 transcriptional response uniform? A diverse set of cell lines to model variation in DNA damage response

12 Human Cell Lines x12

All express WT p53

Range of Radiosensitivites EC50: 1-3Gy Before and After DNA damage p53 levels are comparable Across 12 cell lines

Fold change after damage 1.5-3.5

2

1.5

1

0.5 p53 Fold Change p53 Fold Change log2 Change p53 Fold 0 What is p53 doing in each cell line?

What is p53 binding to? ChIP-seq to measure p53 binding in the genome p53 binding in the genome is quite stereotyped

2hrs after IR treatment (4Gy) Different cell lines show quantitatively similar binding

Pearson R in dataset ranged from 0.5-0.75 p53 binding is stereotyped, with some exceptions Some universal, some cell type specific p53 targets

IL1A GREB1 ~5% of p53 binding sites show variability across cell lines

IL1A GREB1 Why would p53 bind different locations in different cell lines?

Cofactors or Chromatin Access? Variation in p53 binding can be explained by chromatin accessibility (measured with Mu-Transposase ATACseq)

MCF7

LOXIMVI variation in p53 binding can be explained by chromatin accessibility

MCF7

LOXIMVI

GREB1 IL1A variation in p53 binding can be explained by chromatin accessibility

More accessible in LOXIMVI

Differential atac seq signaling explains 22% of the variance in p53 binding

More accessible in MCF7 Does chromatin regulate p53 binding or does p53 regulate chromatin?

Regulated by Chromatin “pioneer factor” No. For core targets of p53 chromatin does not need to be open

p21 PUMA MDM2

Consistent with Genome Res. 2015 Sammons... Berger. Treatment with DNA methylase inhibitor to test effect of increased chromatin accessibility

Decitabine Treatment with DNA methylase inhibitor alters chromatin state in MCF7 cells and opens up new p53 binding sites

Decitabine p53 is both a pioneer and a follower

Tight Regulation by Pioneering activity at chromatin ‘poised’ sites

Cell type specific ‘Universal’ Binding sites DNA binding sites Are there physiological cell states that alter p53 binding? Mesenchymal and Estrogen+ Cell States are associated with novel p53 binding

IL1A GREB1

EMT A549 cells when treated with TGFB undergo EMT Does p53 binding change?

+TGFB

Axl dapi A549 cells forced to undergo EMT show p53 binding at the IL1 locus

+TGFB

Axl dapi

rep1 ChIPseq p53

+TGFB rep2 IL1 expression in EMTed A549 cells depends on p53

+TGFB

Axl

dapi siCTR Log2 siTP53/ p53 IL1A IL1B ChIPseq p53 +TGFB What happens to p53 binding when a cell becomes cancerous?

57% of cancers have WT p53 Conclusions

Dose of DNA damage alters the duration and amplitude of p53 signaling -balance is controlled by ATM levels

Chromatin tunes p53 binding imparting cell type specificity to the DNA damage response -- Some binding sites are ‘poised’ for p53 binding -- Other binding sites require exogenous stimuli How to think about p53?

Master regulator Linear Transducer / Signal integrator

Damage

Cell Fate 1 Cell Fate 2

Modified from Bieging and Attardi - Trends in Cell Bio, 2015. Acknowledgements

Stewart-Ornstein Lab Jacob Stewart-Ornstein Pinakin Pandya Lyubov Kublo LAB Shayla Goller

Current Funding NCI R00 (JSO) Hillman Cancer Center Novartis Lahav Past Funding Bill Forrester Adrian Granada Caroline Mock Ludwig Cancer Center Weissleder NIGMS Tonia Hafner Ralph Weissleder Galit Lahav Novartis Yoshi Iwamoto