figure s1 - massachusetts institute of technology · 2019. 3. 6. · figure s1: a- b) phlip...

0 100 200 3000

1

2

3

4

5

% 3

+ C

AIX

Ctrl

Trea

ted

0

200

400

600

800

pH

LIP

in

ten

sit

y a

.u

***

Ctrl Treated

Heart Kidney Lung

Tumors MDAMB231 XenograftB B’ B”

Ctrl

Trea

ted

Ctrl Treated

Trea

ted

Ctrl

Heart Kidney Tumors PYMT Ctrl Treated A A’ A”

pHLIP

CA9

Dap

i

D

D’

PyMT Lung Mets

7 26 4

Human IDC F

CA9 D

api

% 3+

CA9

+cell

s

Distance from tumor edge

E

0 50 100 150 2000.0

0.5

1.0

1.5

2.0

Distance from Tumor Edge

CA

9


% 3+

CA9

+cell

s

Ctrl

Trea

ted

0

5

10

15

20

pH

LIP

Mean

In

tesn

tiy a.u

Ctrl Treated

***

pHLI

P Me

an in

tens

ity a.

u

pHLI

P Me

an in

tens

ity a.

u

CA9 D

api

PyMT Lung Mets

Figure S1

3+ 2+ 1+

0-1+

H and E pHLIP Dapi pHLIP pHLIP intensity heat map

Segm

ented

imag

e

C C’

%cells singles doubles OverlapP value

Odds ratio

Pearson r

pHLIP 21.2 78.8 7.03E-13 2.46 0.75*** CA9 13.3 86.7

Figure S1: A- B) pHLIP localization to the acidic tissues in vivo is pH-dependent. A) Intensity heat map for Cy7-pHLIP localization in tumor and organs collectedfrom MMTV-PyMT subject to tumor neutralization experiment. Cy7-pHLIP signal intensity is lower in tumors collected from the bicarbonate water-treated group that have neutralized tumor microenvironment, n=7 mice per group. A’-A”) pHLIP mean intensity per area measured from tumor cross sections collected from mice in the bicarbonate treated and control groups shows significant reduction. B) Same as A, but for NSCID Xenograft MDAMB231 mice. n=4 per group. C) pHLIP membrane localization in cross sections of the tumor, H and E serial section shows tumor and adjacent stroma. C’) Cell-based segmentation and intensity profiling shows degree of positivity, cells with highest pHLIP retention at the membrane (3+ population) are located closer to the tumor-stroma interface. D-D’) pHLIP membrane retention and CA9 expression correlate in metastatic lesions from mice. E) Spatial distribution of CA9-positive cells in lung metastases from PyMT mice indicates lower extracellular pH closer to the rim of the metastatic lesion, a pattern similar to the primary tumor. F) CA9 expression is highest at the tumor-stroma interface in human Invasive ductal carcinoma. CA9 expression, similar to pHLIP labeling, marks cells within the areas of the tumor that are acidic; these areas are frequently at the tumor-stroma interface. Scale bars C-F=100um, A-B=4mm.

Figure S2 pH

LIP C

D31D

api

Vascularized Poorly Vascularized E E’ E”

CD31

CA9 L

AMP2

Dap

i

174 343 105 %cells singles doubles Overlap

P value Odds ratio

Pearson r

PM-LAMP2 66.3 33.7 2.76E-17 3.19 0.507***

CA9 37.6 62.4

0 100 200 3000

1

2

3

Distance from edge pixel

Lam

p2%

3+ P

M - L

AMP2

+cell

s


C D

C’

Human IDC

PyMT

LAMP2 Dapi

A’ 652 412 125 %cells singles doubles Overlap

P value Odds ratio

Pearson r

pHLIP 38.8 61.2 3.48E-137 12.3 0.47*** PM-

LAMP2 16.1 83.9

0 50 100 150 200 250 3000

2

4

6

Distance from Tumor Edge

% 3

+ P

M L

AM

P2


% 3+

PM

-LAM

P2+c

ells

pHLIP

LAMP

2 Dap

i PyMT A B

Figure S2: A) PM-LAMP2 overlaps significantly with cells with retained pHLIP at their membrane. A’) Cell based co-localization and overlap analysis number of single and double positivity and percentages summarized in venn diagram and table respectively. B-D) The spatial distribution of PM-LAMP2 in mouse B and human D. Most cells with highest membrane positivity for LAMP2 (3+) localize closer to tumor-stroma interface. C) Most CA9 positive cells in human invasive ductal carcinoma tumor sections overlap plasma membrane LAMP2. C’) cell based co-localization analysis indicate significant overlap between CA9 expression and LAMP2 membrane signal. E) Immunofluorescence labeling for CD31 relative to pHLIP membrane pHLIP membrane localization is not restricted to vascularized areas localization. pHLIP peptide is well diffused in poorly vascularized areas. E’) Highly vascularized area of the tumor as indicated with CD31 labeling, cells located distant from the vasculature have retained pHLIP at the membrane. E”) Poorly vascularized –perfused area of the tumor, pHLIP retained in cells in areas with little or no proximity to CD31 labeled endothelial cells. Images from at least 4 mice (5 areas each) were subjected to cell-based segmentation and intensity profiling in all experiments. Pearson p values were quantified using two-tailed Student’s t test. (Statistical analysis Student’s t test: ***=< 0.001, *=<0.05). Scale bars=100um.

0 100 200 3000

2

4

6

8


%3+

MMP1

4+ ce

lls


%3+

Ki67

+cell

s

CA 9

Lami

nin D

api

CA9 M

MP14

Dapi

A

382 494 213

88 20 12

CA9 K

i67Da

pi

49 62 84

MMP

14Da

pi Ki

67Da

pi

D

A’

C’

B

C

E

0 100 200 3000

5

10

15

D’

Figure S3

Degraded laminin Intact laminin

%cells singles doubles Overlap P value

Odds ratio

Pearson r

CA9 30.1 69.9 1.36E-27 3.10 0.48*** MMP14 43.6 56.4


Odds ratio

Pearson r

CA9 81.5 18.5 1.76E-02 2.63 0.2ns Laminin 37.5 62.5


Odds ratio

Pearson r

CA9 63.2 36.8 8.9E-13 5.26 0.53*** Ki67 55.9 44.1

3

Figure S3: Acidic front is invasive and proliferative in human IDC. A-A’) The expression of MMP14 (antibody specific to active domain) relative to CA9 in human IDC. A’) Cell based overlap and Pearson correlation coefficient for MMP14 and CA9 intensity per cell indicates significant correlation. Most MMP14 expressing cells overlap CA9 positive cells. B) MMP14 distribution relative to tumor dimension, cells expressing highest levels of MMP14 frequently located at the tumor-stroma interface. C) Most CA9 positive and laminin positive cells are frequently excluded. Laminin is degraded on cells that are positive for CA9 however cellular areas that are negative for CA9 have maintained linear laminin structure (magnified insets). C’) Overlap and correlation analysis, Pearson correlation coefficient indicates absence of correlation between CA9 expression and Laminin. Cell based segmentation analysis shows only a small degree of overlap. D-D’) Human tumor staining shows Ki67 and CA9 partially overlap. 44% of Ki67 positive cells also express CA9. E) Ki67 positive cells were enriched at the tumor-stroma interface a pattern reminiscent of low pH areas marked by CA9. Image segmentation and analysis for human samples was performed on 3 tumors from 3 different patients and at least 5 random field each. Pearson p values were quantified using two-tailed Student’s t test. (Statistical analysis Student’s t test: ***p< 0.001, *p<0.05)

4T1 PH6

PyMT PH6

-10

-5

0

5

Log2 Fold change relative to Ctrl

MenaINV

Dock7 E23

DLG1E6

MDAMB 231 PH6

LM2 ph6

SUM159 PH6

PDX-pH6

-5

0

5

10


Mena INV

CD44E19

DOCK7E23

DLG1E6

MDAMB 231 PH6

LM2 ph6

SUM159 PH6

PDX-pH6

-5

0

5

10


Mena INV

CD44E19

DOCK7E23

DLG1E6

C

D

Gene Name

Exon Usage Functional consequence

Protein class Species

CD44 Exon19 - short tail

Drug resistance Transmembrane glycoprotein

Human

ENAH-Mena

Exon 4-INV Invasion-metastasis

Actin binding protein

Mouse-Human

DOCK7 Exon-23 Not known Guanyl nucleotide exchange factor

Mouse-Human

DLG1 Exon-6 Not known Trans membrane adaptor protein

Mouse-Human

**

**

**

**

**

**

** **

**

**

**

**

**

**

**

**

D’

**

**

*

**

**

**

B

Figure S4

DOCK7: ENSG00000116641.11

DLG1: ENSG00000075711.16

CD44: ENSG00000026508.12

ENAH: ENSG00000154380.12

0.0

0.1

0.2

0.3

0.4

pH6 pH7

PSI

chr3:196888510:196888609:−

0.00

0.05

0.10

0.15

pH6 pH7

PSI

chr1:225723208:225723265:−

0.00

0.05

0.10

0.15

0.20

pH6 pH7

PSI

chr1:63010617:63010710:−

0.00

0.05

0.10

0.15

0.20

pH6 pH7

PSI

CD44

A3SS

A5SS

MXE

RI

SE

−0.5 0.0 0.5

PSI

event type

A3SS

A5SS

MXE

RI

SE

−1.0 −0.5 0.0 0.5 1.0

PSI

event type

A

ΔPSI =Acidic-normal

(4T1)

(MDA-MB-231)

pH7.4

pH6.4

4

Figure S4: Validation of alternatively spliced candidates associated with cytoskeletal regulation. A) Comparison of the distributions of splicing changes by events type. Δ PSI values (PSI Acidic-PSI normal) for different types of splicing events in normal versus low pH conditions in human and mouse. Retained introns (RI) show highest magnitude of changes in low pH conditions. B) Table summarizes functional relevance of the included and excluded exons for selected genes; CD44 among the candidates is human specific. C) Integrative genomic viewer (IGV) view of the sequencing read counts over exon of interest for selected genes in low pH and control conditions. Bar graph indicates PSI values (percent spliced) in low pH (pH6.4) and normal pH (pH7.4). D) qPCR validation of the candidate splicing events in a panel of human and mouse breast mammary carcinoma cell lines. Log2 fold change of inclusion/exclusion ratio for the tested exons in a panel of human carcinoma cell lines. D’) Log2 fold change of candidate pH-induced splicing events in mouse carcinoma cell lines. N=4 experiments 3 technical replicates, Students' t-test p<0.05 *, p<0.01 **.

ctrl

ctrl+

hepes

hypoxia

10D

hypoxia

10D +

hepes

0.0

0.5

1.0

1.5

2.0

2.5

L/G

rat

io

ns

** **

Ctrl Metformin Metformin +Hepes

Norm. Norm. +Hepes

Hypoxia Hypoxia +Hepes

C D

Figure S5

ctrl

Metfo

rmin

5mM D

8

met

form

in 5m

M D8+

hepes

0

2

4

6

8

10

L/G

rat

io

ns

**

**

pH6.4-4H

R

pH6.4-12

HR

pH6.4-48

HR

pH6.4-4D

ays

-10

-5

0

5

10

Log2

Fol

d ch

ange

rela

tive

to C

trl

Mena INV

Cd44E19

DOCK7E23DLG1E6

pH6.4 4Hrs

pH6.4 12Hrs

pH6.4 48Hrs

pH6.4 96Hrs

A

**

**

****

****

****

**

ns

nsns

ns * ns

ns

pH6.4

pH6.4+T

SA-6

-4

-2

0

2

4

6

Log2

Fol

d ch

ange

rel

ativ

e to

Ctr

l

Mena INV

CD44E19

DOCK7E23

DLG1E6

pH6.4 pH6.4 +TSA

** *

ns ns

E F

TSA Ctrl 6.4 7.4 6.4 7.4

H3 Acetyl K27

Tubulin

1 1 169 17.1

0hr

4hrs

12hrs

48hrs

96hrs

0

2

4

6

8

10

Nor

mal

ized

Cel

l Gro

wth pH6.4

ph7.4

4 12 48 96 0 Time (hours)

**

B

5

Figure S5: A) The expression of selected candidate splicing events relative to time cultured in low pH. CD44E19 is significantly included at 4hrs and increases in a time dependent manner whereas, MenaINV inclusion and DOCk7 E23 , DLG1E6 exclusions only occurs at 48 hour onwards. Buffering of medium by HEPES does not alter the lactate to glucose ratio. B) Cell growth measured over time in pH7.4 and pH6.4 cell number in each time point is normalized to time 0.C-D) Medium from cells under (C) long-term hypoxia (10days) or (D) metformin addition has significantly higher lactate to glucose ratio. Neutralization of the pH of the medium by addition of HEPES does not affect this ratio. E-F) The inclusion of MenaINV and CD44E19 is sensitive to histone acetylation levels. Addition of 500nm Trichostatin A (TSA) to culture media under low pH significantly reduced the inclusion of MenaINV and CD44E19 exons but does not affect DLG1E6 and DOCK7E23 skipping. F) Western blot indicating increases in H3 acetylation in response to TSA under pH7.4 and pH6.4 conditions. Values indicate normalized fold change in acetylated H3 signal.

A

log2 fold change

CX3CR

CXCL1

IL6

MMP11

DDIT3

CLK1

CLK3

DOCK10

ITGA1

-1

0

1

In vivo

Log2 Fold Change (Cy7+/Cy7-)

4t1

CX3CR

CXCL1

IL6

MMP11

DDIT3

CLK1

CLK3

DOCK10

ITGA1

0

5

In vitro QPCR

log2fold change

CX3CR

CXCL1

IL6

MMP11

DDIT3

CLK1

CLK3

DOCK10

ITGA1-5

0

5

In vitro RNAseq

Log2 Fold Change (pH6.4/pH7.4)

Figure S6

B

Mena

INV C

A9 D

api

84 121 86

Mena

INV D

api

C

B’

20 40 60 80 1000

1

2

3

4


%3+

INV

%3+INV


%3+

Men

aINV +

cells


Odds ratio

Pearson r

CA9 41.5 58.5 1.1E-22 5.94 0.60*** Mena INV 41.0 59.0

D

CD4

4 E19

-RNA

CA9

Dap

i CD

44 E

19 R

NA D

api

E

107 318 674

0 50 100 150 200 250 3000.0

0.5

1.0

1.5

2.0


%3+

CD

44 E

19


%3+

CD4

4E19

+ ce

lls

D’ %cells singles doubles Overlap

P value Odds ratio

Pearson r

CA9 67.9 32.1 9.19E-12 2.28 0.18*** CD44E19 25.2 74.8 Hum

an ID

C

6

Figure S6: Validation of pH-responsive transcriptomic changes in vivo. A) Heat map comparing log2 fold changes of candidate genes from pH signature from the RNAseq dataset that were validated by qPCR in vitro in 4T1 cells and in vivo in cells collected by FACS sorting approach. B) CA9 expression overlaps the expression of the MenaINV isoform in human invasive ductal carcinoma (IDC). B’) Cell based co-localization and Pearson correlation coefficient indicates high degree of overlap. C) MenaINV distribution in human tumor samples relative to tumor dimensions shows the highest expressing cells located proximal to the tumor-stroma interface. D) RNAscope exon-specific mRNA detection relative to CA9 expression in human IDC tumors. D’) Cell based co-localization and Pearson correlation indicates significant overlap. Most CD44E19 RNA-positive cells are also expressing CA9 protein. E) CD44E19 expressing cell distribution in human IDC tissue samples. Highest expressing cells are located closer to the tumor-stroma interface.

1

Supplementary Methods: Illumina sequencing: RNA samples from mouse and human cell lines in duplicate were subject to RNA-sequencing. KAPA Hyper Prep Kit (Kapa Biosystems) was used for library preparation. Short read sequencing was performed on NextSeq500 platform to a depth of more than 40 million reads per sample using 75bp paired-end method and Nextseq150nt kit. Raw and processed RNA-seq data: Gene expression ominibus, NCBI, accession number: GSE119646. Bioinformatics analysis: We confirmed RNA-seq data quality by running an in-house computational pipeline that estimated sequencing error rate, read complexity, contamination, read alignment percentages to exons, UTRs, introns, and intergenic regions, rRNA contamination, and sense to antisense read ratios. Hierarchical clustering and heatmap creation were carried out using TIBCO Spotfire 7.6.1.17 based on log2 (fpkm+1) values of the expressed coding genes. We mapped sequencing reads to either the GRCh38/hg38 reference genome for human samples or the GRCm38/mm10 reference genome for mouse samples using STAR version 2.5.3a(1). Gene-level read counts were calculated using RSEM version 1.3.0 in a strand specific manner and exon level counting for alternative splicing analyses was performed using HTSeq packaged within DEXSeq (2). We performed differential gene expression analysis of sample read counts between the low and normal pH conditions with DESeq2 version 1.10.1(3). For alternative splicing analysis, we used the DEXSeq package (version 1.22.0) (4). Filtering criteria to define significance was set to base mean greater than 10, log2FoldChange greater than 0.5 to identify up-regulation and less than -0.5 to identify down-regulation, and padj smaller than 0.05. Overlapping between human and mice samples was based on gene symbols after matching mouse genes symbols to human gene symbols. The significance of overlap was tested using hypergeometric statistics. Event-centric splicing analyses were performed using rMATS. Events were filtered for having >= 20 junction-spanning reads supporting inclusion and/or exclusion. Exons with abs (ΔPSI) >= 0.05 and FDR < 0.05 were considered significantly changing and exons with abs (ΔPSI) <= 0.02 and FDR > 0.5 were used as control exons. Cell culture MDA-MB231 cell line was purchased from ATCC in 2016, LM2, the highly metastatic derivative of MDA-MB231, was obtained from J.Massague (5) in 2016 and cultured in DMEM and 10% FBS, 1% penicillin-streptomycin-glutamine. SUM159 were obtained from Joan Brugge’s lab in 2016 at Harvard Medical

2

School and cultured according to ATCC guidelines. The patient-derived xenograft (PDX) cell line- EL12-58 was established from a previously untreated triple-negative breast cancer sample; the cells were a gift from Dr. Daniel G Stover in 2016. The culture was established using a protocol adapted/optimized from (6). (In short: 2/3 DMEM, 1/3 HAMS F12, 0.5ug Hydrocortisone/0.5ug EGF, 5mg Insulin, Cholera Toxin 0.5uM, Fungizone, 5uM ROCK inhibitor Y-27632). 4T1 cell lines were purchased from ATCC in 2015 and cultured in DMEM, 5% FBS+5% FCS+1% non-essential amino acids, 1%penicillin streptomycin glutamine. John Lamar at MIT generated PyMT cell lines. The cells were derived from MMTV-PyMT tumors and metastatic lesions, and selected to enrich for highly metastatic variants by repeated in vivo sorting for metastatic potential. These were cultured in 50% DMEM and 50% F12, 2% calf serum, 0.01% BSA, 10 ng/ml EGF, 10ug/ml insulin ,1%penn/strep ,1% glutamine. Growth rate Cell count was determined over time and up to 96 hours for MDAMB 231 cells expressing nuclear RFP that are cultured in pH7.4 and pH6.4 in IncuCyte®. Growth rate was measured using normalized cell number (cell number at t=t1/ cell number at t=t0). References: 1. Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, et al.

STAR: Ultrafast universal RNA-seq aligner. Bioinformatics. 2013;29:15–21. 2. Li B, Dewey CN. RSEM: Accurate transcript quantification from RNA-Seq

data with or without a reference genome. BMC Bioinformatics. 2011;12. 3. Love MI, Huber W, Anders S. Moderated estimation of fold change and

dispersion for RNA-seq data with DESeq2. Genome Biol. 2014;15:1–21. 4. Anders S, Reyes A, Huber W. Detecting differential usage of exons from

RNA-seq data. Genome Res. 2012;22:2008–17. 5. Minn AJ, Gupta GP, Siegel PM, Bos PD, Shu W, Giri DD, et al. Genes that

mediate breast cancer metastasis to lung. Nature. Nature Publishing Group; 2005;436:518.

6. Palechor-Ceron N, Suprynowicz FA, Upadhyay G, Dakic A, Minas T, Simic V, et al. Radiation Induces Diffusible Feeder Cell Factor(s) That Cooperate with ROCK Inhibitor to Conditionally Reprogram and Immortalize Epithelial Cells. Am J Pathol. American Society for Investigative Pathology; 2013;183:1862–70.

3

Supplementary Tables: Supplementary Table 1: Table S1: List of all antibodies. Antibody Cat# Company Laminin Ab11575 Abcam MMP9 Ab38898 Abcam MMP14 (active domain) MAB3328 Millipore FitC-Cyanine NB-1207628 Novus Biologicals Piminidazole MAb1 Hpi, hypoxyrpobe LDHA NBP1-48336 Novus Biologicals Ki 67 (SP6) Ab16667 Abcam CA9 Ab15086 Abcam HDAC1 Ab7028 Abcam Anti-Histone H3 (acetyl K27) ab4729 Abcam LAMP2 Ab203224 Abcam

4

Supplementary Table 2: Table S2: q-PCR primer sequences.

Gene Forward Reverse Species

CD44 TGAATATAACCTGCCGCTTTG GCTTTCTCCATCTGGGCCAT H

CD44-E19 GCAGTCAACAGTCGAAGAAGTTGAAG CCATGTGTCAGTTCTAGCGAGG H

Mena GAA CAA AAA GAG GAC AAA GGT GA TGC CAT TCA TTG TAT TTG TTC TTT H

Mena GGG CAG AAA GAT TCA AGA CC GCG AAG ACA TTG GCA TCC M

Mena INV AGA GGA TGC CAA TGT CTT CG TTA GTG CTG TCC TGG GTA GC H

Mena INV AGA GGA TGC CAA TGT CTT CG TTA GTG CTG TCC TGC GTA GC M

DOCK7 GTGGCAGCCGAAGTTAGGAAG GCACTGTGGTGTGATGGGATA H

Dock7 E23 GGGACTCCCACGTCACCAG GACTGGGGTTGGATCCCCAT H

DLG1 GCAGGAGGTACGGACAACC ATTGACCCGCAATCTTCCATC H

DLG1-E6 ATGAAGTGATAGGTCCAGAATTGGTTCAT ACAGGGATCACAGGGACAGTG H

DLG1 AGTGACGAAGTCGGAGTGATT GTCAGGGATCTCCCCTTTATCT M

DLG1 E6 ATGAGGTGCTAGGTCCAGAACT GGACGACAGTACTCTCAGCAGG M

DOCK7 CCATCTGGAAGCGCCTTTG ACGATGATCTCTAGCGTGTCT M

DOCK7E23 CTCCCACATCACCAGATGATGAAG ACTTGGACTGGGGTTGGATCC M

CLK1 ACAAGACATTATAGAGCACCGGA GTGGTCCAAGAATCCTTTCCATC M

CLK3 TGTCACAAACGCCGTACCAG ACGATCTCATATCGCTCTTGGA M

DDIT3 CTGGAAGCCTGGTATGAGGAT CAGGGTCAAGAGTAGTGAAGGT M

ITGa1 CCTTCCCTCGGATGTGAGTCA AAGTTCTCCCCGTATGGTAAGA M

DOCK10 TTTTGTGAGCGTGGCACTTTA CTTCCGATTGTCTTGGTGTACC M

CX3CR1 GAGTATGACGATTCTGCTGAGG CAGACCGAACGTGAAGACGAG M

IL6 TAGTCCTTCCTACCCCAATTTCC TTGGTCCTTAGCCACTCCTTC M

MMP11 CCGGAGAGTCACCGTCATC GCAGGACTAGGGACCCAATG M

CXCL1 CTGGGATTCACCTCAAGAACATC CAGGGTCAAGGCAAGCCTC M

GAPDH CAT GTT CCA GTA TGA CTC CAC TC GGC CTC ACC CCA TTT GAT GT M

GAPDH CAT GAG AAG TAT GAC AAC AGC CT AGT CCT TCC ACG ATA CCA AAG T H

5

Supplementary Table 3: MDAMB231- SPMA motif hits results:

motif ID RBPs adjusted R squared

polynomial degree

Consistency Score (CS) p-value

783_7972035 ELAVL2 0.76 1 0.000001

M112_0.6 ELAVL1, ELAVL3 0.709 1 0.000001

M012_0.6 CPEB3, CPEB2 0.74 1 0.000001 M124_0.6 ELAVL3 0.733 1 0.000001

M031_0.6 ELAVL1, ELAVL3 0.749 1 0.000001

951_12324455 ZFP36 0.646 1 0.000001 M120_0.6 CPEB3 0.589 1 0.000006

M227_0.6 PTBP1, PTBP2, ROD1 0.673 1 0.00025

LC1 ELAVL1 0.771 1 0.000001 M075_0.6 TIA1 0.694 1 0.000002

784_7972035 ELAVL2 0.777 1 0.000014 M149_0.6 CPEB3, CPEB4 0.764 1 0.000002 M158_0.6 HNRNPCL1 0.775 1 0.000001 M025_0.6 HNRNPC 0.775 1 0.000001 M150_0.6 RALY 0.717 1 0.000001 M201_0.6 SF3B4 0.762 1 0.000001

782_8497264 ELAVL2 0.743 1 0.000021

M232_0.6 ELAVL1, ELAVL3 0.758 1 0.000001

M156_0.6 TIA1 0.763 1 0.000001

M127_0.6 ELAVL1, ELAVL3 0.703 1 0.000001

M077_0.6 U2AF2 0.73 1 0.000002 147_19457263 KHDRBS3 0.636 1 0.000017

M176_0.6 KHDRBS2 0.768 1 0.000001 M001_0.6 A1CF 0.475 1 0.002261 M229_0.6 CELF3 0.475 1 0.004264

M108_0.6 ELAVL1, ELAVL3 0.698 1 0.000001

M160_0.6 KHDRBS1 0.75 1 0.000001

M157_0.6 CELF3, BRUNOL5 0.395 1 0.105509

M079_0.6 CELF3 0.503 1 0.000272 M051_0.6 RBM41 0.378 1 0.0007

1171_19561594 PTBP1 0.649 1 0.001153

6

4t1- SPMA motif hits results:

Table S3: Spectrum motif analysis (SPMA) identified motif hits for MDA-MB-231 and 4t1 cell lines differential gene expression dataset sorted according to Log2 fold change in expression. Linear relationship between fold change in transcript expression and enrichment of putative RBP binding sites, are measured by adjusted R2 values and local consistency scores (CS) p-values.

motif ID RBPs adjusted R squared

polynomial degree

Consistency Score (CS) p-

value LC1 ELAVL1 0.827 1 0.000017

M031_0.6 ELAVL1, ELAVL3 0.73 1 0.000079 M176_0.6 KHDRBS2 0.662 1 0.000189 M112_0.6 ELAVL1, ELAVL3 0.714 1 0.002543

783_7972035 ELAVL2 0.692 1 0.005896

784_7972035 ELAVL2 0.631 1 0.008048

M236_0.6 HNRNPR, SYNCRIP 0 0 0.008048

M232_0.6 ELAVL1, ELAVL3 0.462 1 0.008048

782_8497264 ELAVL2 0.708 1 0.011619

M156_0.6 TIA1 0.446 1 0.016212 M075_0.6 TIA1 0.42 1 0.016212 M149_0.6 CPEB3, CPEB4 0.487 1 0.030619 M025_0.6 HNRNPC 0.467 1 0.017497 M158_0.6 HNRNPCL1 0.467 1 0.019996 M201_0.6 SF3B4 0.438 1 0.040825 M150_0.6 RALY 0.381 1 0.034993 M077_0.6 U2AF2 0.373 1 0.049727

figure s1 - massachusetts institute of technology · 2019. 3. 6. · figure s1: a- b) phlip...

Documents