A passive-flow microfluidic device for imaging latent HIV activation dynamics in single T cells

Ramesh Ramji, Victor C. Wong, Arvind K. Chavali, Larisa M. Gearhart, and Kathryn Miller-Jensen*

Electronic Supplementary Information Figure S1. Scanning electron microscope (SEM) images of S and V traps. Figure S2. Effect of cell loading density and volume on cell trapping efficiency. Figure S3. Drug-stimulated latent HIV activation in the passive-flow device is similar to behavior observed in plate-based flow cytometry assays. Figure S4.  Trends in the CV for ton are consistent at longer stimulation times. Figure S5. CVs for ton, FmaxO, and slope are consistent between cell cycle synchronized and unsynchronized cells after PMA treatment. Figure S6. Three metrics that describe HIV expression are strongly correlated. Figure S7. Trends between different treatments are similar with and without the quantification of non-activators. Table S1. Analysis of distributions for onset of activation time (ton), maximum HIV expression (FmaxO), and rate of HIV production (slope) demonstrates statistically significant differences in reactivation dynamics. Movie S1. Comparison of docking of Jurkat cells treated with PMA in the microfluidic device versus poly-lysine-coated plate. Time stamp in hours:minutes (00:00). Movie S2. Solution delivery within the microfluidic device. Time stamp in seconds. Movie S3. Cell loading in the microfluidic device. Time stamp in minutes:seconds (00:00).

Electronic Supplementary Material (ESI) for Integrative Biology.This journal is © The Royal Society of Chemistry 2015

S trap V trap

Fig. S1. Scanning electron microscope (SEM) images of square pillars (S traps) and rectangular pillars (V traps).

10 μm 10 μm

1 2.5 50

50

100

1 2.5 50

50

100

50 1000

50

100

H d-S S V

50 1000

50

100

1 2.5 50

50

100

1 2.5 50

20

40

60

80

100

50 1000

50

100

50 1000

20

40

60

80

100

Pre Wash Post Wash

A B C D

E F G H

% T

rapp

ing

effic

ienc

y (o

vera

ll)%

Tra

ppin

g ef

ficie

ncy

(ove

rall)

% T

rapp

ing

effic

ienc

y (s

ingl

e ce

ll)%

Tra

ppin

g ef

ficie

ncy

(sin

gle

cell)

Flow volume (μl) Flow volume (μl) Flow volume (μl) Flow volume (μl)

% T

rapp

ing

effic

ienc

y (o

vera

ll)%

Tra

ppin

g ef

ficie

ncy

(ove

rall)

% T

rapp

ing

effic

ienc

y (s

ingl

e ce

ll)%

Tra

ppin

g ef

ficie

ncy

(sin

gle

cell)

Cell density (million cells/ml) Cell density (million cells/ml) Cell density (million cells/ml) Cell density (million cells/ml)

Fig. S2. Effect of cell loading density and volume on cell trapping efficiency. (A-B) Effect of varying cell density on (A) overall and (B) single-cell trapping efficiency for Hd-S, S and V traps. (C-D) Final trapping efficiency after a wash step to remove undocked cells. (E-F) Effect of inlet cell flow volume on (E) overall and (F) single-cell trapping efficiency for Hd-S, S and V traps. (G-H) Final trapping efficiency after a wash step to remove undocked cells. Data are presented as the mean ± standard deviation of independent experiments (SD; n=2).

PMA

TNF

Pros

tratin

TSA

SAH

A

TNF

TSA

Pros

tratin

SAH

A

0

50

100

% H

IV-G

FP p

ositi

ve

6 hours 12 hours

PMA

TNF

Pros

tratin

TSA

SAH

A

TNF

TSA

Pros

tratin

SAH

A

0

50

100

Passive-flow Flow cytometryA B

TF activator HDACInhibitor

Combination

22 hours

TF activator HDACInhibitor

Combination

6 hours 12 hours 24 hours

% H

IV-G

FP p

ositi

ve

Fig. S3. Drug-stimulated latent HIV activation in the passive-flow device is similar to behavior observed in plate-based flow cytometry assays. (A) Bar graph of % HIV-GFP positive cells in response to indicated treatments in the passive-flow device. Fluorescence images for GFP were taken every 5 minutes after stimulation, and the number of GFP positive cells were counted at 6, 12 and 22 hours after stimulation. GFP positive cells calculated at 22 hours included cells that were locally displaced from the trap (up to 500 µm) in order to record enough cells. (B) Bar graph of % HIV-GFP positive cells in response to indicated treatments in a cell-culture dish. Fluorescence was measured at 6, 12, and 24 hours after stimulation with flow cytometry. Data for 0 hours not shown because % HIV-GFP positive cells was less than 1% for both assays.

PM

A

TNF

Pro

stra

tin

TSA

SA

HA

TNF

TSA

Pro

stra

tinS

AH

A

TF activator HDACInhibitor

Combination

0.0

0.5

1.0

Coe

ffici

ent o

f var

iatio

n

Fig S4. Trends in the CV for ton are consistent at longer stimulation times. Jurkat cells were stimulated and observed for 22 hours in the passive-flow device. All cells that remained in the field of view for 22 hours were tracked, including cells that were locally displaced from the trap (up to 500 µm).

PM

A

CC

SP

MA

0.0

0.2

0.4

0.6

0.8

1.0

Coe

ffici

ent o

f var

iatio

n

PM

A

CC

SP

MA

0.0

0.5

1.0

1.5

2.0

Coe

ffici

ent o

f var

iatio

n

PM

A

CC

SP

MA

0.0

0.5

1.0

1.5

2.0

Coe

ffici

ent o

f var

iatio

nton FmaxO Slope

A B C

Fig. S5. CVs for ton, FmaxO, and slope are consistent between cell cycle synchronized and unsynchronized cells after 12 hours of PMA treatment. (A) Bar graph of coefficient of varia-tion (CV) for onset of activation time (ton) for cell cycle synchronized cells versus unsyn-chronized cells after PMA treatment. (B) CVs for observed maximum expression (FmaxO). (C) CVs for slope. Cells were synchronized for 24 hours via serum starvation prior to treat-ment with PMA. Error bars represent 95% confidence interval obtained by bootstrapping.

PM

A

TNF

Pro

stra

tin

TSA

SA

HA

TNF

TSA

Pro

stra

tinS

AH

A

0

2

4

6

8

Coe

ffici

ent o

f var

iatio

n

PM

A

TNF

Pro

stra

tin

TSA

SA

HA

TNF

TSA

Pro

stra

tinS

AH

A

0

2

4

6

Coe

ffici

ent o

f var

iatio

n

0

10

20

F final (A

U)

PM

A

TNF

Pro

stra

tin

TSA

SA

HA

TNF

TSA

Pro

stra

tinS

AH

A

A

C Area under curve (AUC) D

TF activator HDACInhibitor

Combination

0

40

80

120

AU

C (A

U)

PM

A

TNF

Pro

stra

tin

TSA

SA

HA

TNF

TSA

Pro

stra

tinS

AH

A

Final HIV expression (Ffinal) B

0 10 20 300

10

20

30

FmaxO (AU)

F fin

al (A

U)

R = 0.99

TF activator HDACInhibitor

Combination

PM

A

TNF

Pro

stra

tin

TSA

SA

HA

TNF

TSA

Pro

stra

tinS

AH

A0.0

0.5

1.0

Cor

rela

tion

(R)

FmaxO vs Ffinal FmaxO vs AUC Ffinal vs AUC

E FPMA

Fig. S6. Three metrics that describe HIV expression are strongly correlated. (A) Distributions of HIV expression at 12 hours of treatment (Ffinal) for all cells (n ≥ 230). White dot indicates mean of distribution. (B) CV of final HIV expression. Error bars represent 95% confidence interval obtained by bootstrapping. (C) Distributions for AUC for HIV expression over time for all cells. (D) CV for AUC. (E) Example of scatterplot for correlation between FmaxO and Ffinal in cells activated by PMA treatment. R indicates Pearson correlation coefficient. (F) Bar graph for correlation between indicated metrics for each treatment.

PM

A

TNF

Pro

stra

tin

TSA

SA

HA

TNF

TSA

Pro

stra

tinS

AH

A

0.0

0.5

1.0

Coe

ffici

ent o

f var

iatio

n OFF+ONON only

Onset of activation time (t0)

PM

A

TNF

Pro

stra

tin

TSA

SA

HA

TNF

TSA

Pro

stra

tinS

AH

A

0

2

4

6

Coe

ffici

ent o

f var

iatio

n

Max HIV expression (FmaxO)

ON onlyOFF+ON

PM

A

TNF

Pro

stra

tin

TSA

SA

HA

TNF

TSA

Pro

stra

tinS

AH

A

0

2

4

6

8

Coe

ffici

ent o

f var

iatio

n

Area under curve (AUC)

ON onlyOFF+ON

PM

A

TNF

Pro

stra

tin

TSA

SA

HA

TNF

TSA

Pro

stra

tinS

AH

A

0

2

4

6

Coe

ffici

ent o

f var

iatio

n

Rate of HIV production (slope)

ON onlyOFF+ON

PM

A

TNF

Pro

stra

tin

TSA

SA

HA

TNF

TSA

Pro

stra

tinS

AH

A

0

2

4

6

Coe

ffici

ent o

f var

iatio

n

Final HIV expression (Ffinal)

ON onlyOFF+ON

A B

C D

E

NF-ĸB activator

HDACInhibitor

Combination

Fig. S7. Trends between different treatments are similar with and without the quantification of non-activators. (A) Bar graph for coefficient of variation (CV) for onset time for each treatment without (blue) and with (red) non-activators. (B-E) Bar graphs for CVs for (B) FmaxO, (C) Ffinal, (D) AUC, and (E) slope. For each metric, CVs for each perturbation were normalized by setting SAHA OFF+ON as the greatest CV, scaling SAHA ON to the same value, and applying the same scaling constant to the ON CV for each perturbation. CVs were normalized to demon-strate similarity in trends. Error bars represent 95% confidence interval obtained by bootstrap-ping.

Onset Time (ton) PMA TNF TSA TNF+TSA Pro SAHA Pro+SAHA

PMA 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001

TNF 0.0546 0.0001 0.0224 0.0074 0.0038

TSA 0.0001 0.0001 0.15 0.0001

TNF+TSA 0.1 0.0001 0.0482

Prostratin 0.0001 0.93

SAHA 0.0001

Pro+SAHA

A

Max (FmaxO) PMA TNF TSA TNF+TSA Pro SAHA Pro+SAHA

PMA 0.0001 0.0001 0.0269 0.0001 0.0001 0.0034

TNF 0.95 0.0001 0.0004 0.0001 0.0001

TSA 0.0001 0.0001 0.0001 0.0001

TNF+TSA 0.0001 0.0001 0.5

Prostratin 0.0001 0.0001

SAHA 0.0001

Pro+SAHA

B

C

Slope (S) PMA TNF TSA TNF+TSA Pro SAHA Pro+SAHA

PMA 0.0001 0.0001 0.22 0.0001 0.0001 0.0071

TNF 0.77 0.0001 0.0016 0.0001 0.0001

TSA 0.0001 0.0001 0.0001 0.0001

TNF+TSA 0.0001 0.0001 0.1269

Prostratin 0.0001 0.0001

SAHA 0.0001

Pro+SAHA

α = 0.0023

Table S1. Analysis of distributions for (A) onset of activation time (ton), (B) maximum HIV expression (FmaxO), and (C) rate of HIV production (slope) demonstrates that LRAs activate latent HIV with significantly different dynamics. Pairwise comparisons evaluated by Wilcox-on-Mann-Whitney test with Bonferroni correction.