1

An Integrated Microfluidic Chip System for Single-Cell Secretion Profiling of Rare

Circulating Tumor Cells

Yuliang Deng, Yu Zhang, Shuai Sun, Zhihua Wang, Minjiao Wang, Beiqin Yu, Daniel M. Czajkowsky,

Bingya Liu, Yan Li, Wei Wei, Qihui Shi

Supplementary Methods

1. Herringbone microfluidic chip fabrication for rare circulating tumor cell capture

The herringbone microfluidic chips employed for rare cell capture were fabricated based on previous

studies. (1,2) The patterned silicon mold was produced by a standard two-step photolithographic

process. In the first step, a thick layer (50 μm) of negative photoresist (SU8-2050, MicroChem Corp.,

USA) was spin-coated onto a 4-inch silicon wafer. After UV exposure and development, a serpentine

fluidic channel with rectangular shape in the cross section was obtained (length 200 mm and width 1.0

mm). In the second step, another layer (50 μm) of SU8-2050 was spin-coated on the same wafer. Prior

to the UV exposure the second mask was aligned to ensure a good alignment between the previous

pattern and the pattern to be fabricated. To facilitate the separation between the replicate and molded

poly(dimethylsiloxane) (PDMS) component, the replicate was treated by trimethylchlorosilane (TMCS,

Sigma-Aldrich, USA). The PDMS pre-polymer (Sylgard 184, Dow Corning, USA) was mixed in a

ratio of 10:1, and subsequently casted on this lithographically patterned replicate. After curing at 80°C

for 2 h, the PDMS component was separated from the replicate, followed by trimming and hole

introduction.

To chemically modify the PDMS device, the microfluidic channels were treated with O2 plasma

(PDC-32G, Harrick Plasma, USA) to create silanol groups, followed by treatment of 4% (v/v) solution

of (3-Aminopropyl)trimethoxysilane (Sigma-Aldrich, USA) in ethanol for 1 h at room temperature.

Finally, the amoni-group-modified PDMS chip was thermally bonded to a poly-L-Lysine (PLL) coated

glass slide (Thermo Fisher Scientific Inc., USA). A holder composed of two pieces of polyacrylate

plates was home-machined to sandwich the PDMS chip and the glass slide substrate. Then the holder

sandwiched these two parts by screws at the four corners of plastic plates. Two through-holes were

drilled on the plastic plate (top) with a good alignment to the inlet and outlet holes of the PDMS chip,

thus allowing convenience connection of tubing to the microfluidic channel of the chip. The complete

assembly was capable of enduring back pressure of about 10 psi on the tested fluid samples without

leaking.

2

To immobilize DNA on the internal wall of the device (including glass substrate and PDMS slab), a

mixture (1:1 v/v) of amine-modified DNA solution (300 μM, in DMSO) and 2mM BS3 solution (in

PBS, Thermo Scientific, USA) was filled into the device and flowed for 1.5 hours. The device was then

stored in a sealed Petri dish with damp cotton for 1.5 hours, followed by washing with 0.02% SDS (in

water) and deionized Millipore water. BS3 crosslinks amine-modified DNA with glass substrate and

PDMS that are also modified with amine groups.

2. Spiral microfluidic chip fabrication for depletion of red blood cells

The spiral chip was fabricated in PDMS using standard microfabrication soft-lithographic techniques

described previously for depletion of red blood cell.(3,4) Briefly, a replicate for molding of PDMS was

obtained by pattering a silicon wafer using photoresist SU-8 2100. The chip design consisted of a

single inlet and triple-outlet spiral microchannel with 500 μm in width and 150 μm in height. The

replicate was treated by TMCS to facilitate separation between the replicate and molded PDMS

component. The PDMS pre-polymer was mixed in a ratio of 10:1, and subsequently casted on this

lithographically patterned replicate. After curing at 80 °C for 2 h, the PDMS component was separated

from the replicate, followed by trimming and hole introduction. The finished PDMS component was

treated with oxygen plasma (PDC-32G, Harrick Plasma, USA) for bonding onto a glass slide. Finally,

the microfluidic chip was treated at 70 °C for one week to make PDMS surface hydrophobic.

3. Fabrication of single-cell barcode chip (SCBC)

The SCBC was assembled from a DNA barcode microarray glass slide and a PDMS slab containing a

microfluidic circuit. (5,6) The PDMS microfluidic chip for the single-cell experiment was fabricated

using a two-layer soft lithography approach. A push-down valve configuration was utilized with a thick

control layer bonded to a thin flow layer underneath. The control layer was molded from a SU8 2025

negative photoresist (~40 μm in thickness) silicon master using a mixture of PDMS (RTV615, GE

Toshiba Silicones Co. Ltd., USA) pre-polymer part A and part B (10:1). The flow layer was fabricated

by spin-casting the pre-polymer of PDMS part A and part B (20:1) onto a SPR220-7 positive

photoresist (Rohm Haas, USA) master at ~2000 rpm for 1 minute. The SPR 220 mold was ~10 μm in

height, and these patterns were rounded via thermal treatment to allow complete valve closure. The

control layer PDMS chip was then carefully aligned and placed onto the flow layer, which was still

situated on its silicon master mold, and an additional 60 min thermal treatment at 80°C was performed

to enable bonding. Afterward, this two-layer PDMS chip was cut off and access holes drilled. Finally,

the microfluidic-containing PDMS slab was thermally bonded onto the barcode-patterned glass slide to

3

give a fully assembled microchip. The barcode-patterned glass slide is composed of DNA barcode for

detection of secreted proteins and polylysine patterns for on-chip cell capture.

4. Microfluidic chip fabrication for barcode patterning

The barcode patterns for single-cell secretion profiling and on-chip cell capture were created with

microchannel-guided flow patterning. The barcode patterning chips were fabricated based on previous

studies.(5,6) The mixture of PDMS pre-polymer and curing agent in 10:1 ratio (w/w) was poured onto

the SU8-2015 negative photoresist mold with a 20-microchannels array (20 m in width and 30 m in

height) and cured at 80°C for 1 hour. The cured PDMS slab was released from the mold, inlet/outlet

holes punched, and thermally bonded onto a PLL-coated glass slide to form enclosed channels.

5. Fabrication of DNA and PLL barcodes

The DNA and PLL barcodes pattern were created with microchannel-guided flow patterning. After

bonding of PDMS device to the PLL-coated glass substrate, a library of DNA solutions (C, D, and M),

diluted in a mixture of DMSO and deionized water (v/v=1:2) with a final concentration of 267 μM, as

well as 0.2 w/v% PLL solution (Mw≥300,000, Sigma-Aldrich) were flowed into each of the

microfluidic channels. The solution-filled chip was placed in a desiccator to allow solvent (DMSO and

water) to evaporate completely through the gas-permeable PDMS, leaving the DNA and PLL

molecules behind. This evaporation process took 2~3 days to complete. Lastly, the PDMS elastomer

was removed from the glass slide, and the barcode-patterned DNA was cross-linked to the glass surface

by thermal treatment at 80°C for 4 hours. Residual crystals were readily removed by rapidly dipping

the slide in deionized water, followed by blow-dry with nitrogen gun. This glass substrate with DNA

and PLL barcodes were then ready for bonding with a SCBC.

6. DNA barcode validation

Each DNA barcode chip needs validation before bonding to the single-cell assay chip. A small area

close to the edge was validated to check the DNA loading and uniformity. To do this, fluorescent Cy3-

labeled complementary DNAs in 1% Bovine Serum Albumin (BSA, Sigma-Alrich)/ phosphate-

buffered saline (PBS) were hybridized for 1 hr. After washing three times with 1% BSA/PBS and PBS,

the slide was dried by nitrogen gun and scanned by GenePix 4400A (Molecular Device, USA, see Fig.

S2). Under laser power of 10% and gain of 400, fluorescence intensity above 15,000 was acceptable for

CTC capture in herringbone chip and detection of secreted proteins in SCBC.

4

7. Antibody-ssDNA conjugation and validation

As-received antibodies (see Table S1) were desalted, buffer exchanged to pH 7.4 PBS and

concentrated to 1 mg/mL using Zeba protein desalting spin columns (Thermo Fisher Scientific Inc.,

USA). Succinimidyl 4-hydrazinonicotinate acetone hydrazine (SANH, Solulink, USA) in N,N-

dimethylformamide (DMF) was added to the antibodies at variable molar excess of (200:1) of SANH

to antibody. Separately, succinimidyl 4-formylbenzoate (SFB, Solulink, USA) in DMF was added at a

16-fold molar excess to 5‘aminated 30 mer oligomers in PBS. After incubation for 4 hours at room

temperature, excessive SANH and SFB were removed by buffer exchange of both samples to pH 6.0

citrate buffer using protein desalting spin columns. A 30-fold excess of derivative DNA was then

combined with the antibody and allowed to react 2 hours at room temperature followed by overnight

incubation at 4oC. Non-coupled DNAs were removed by AKTA Purifier 100 FPLC (GE, USA) with

Pharmacia Superdex 200 gel filtration column (GE, USA) at 0.5 mL/min isocratic flow of PBS. The

conjugates were then concentrated to 0.5 mg/mL by Amicon Ultra-4 Centrifugal Filter Unit with

Ultracel-10 membrane (10kDa, Millipore, USA) and stored at 4°C.

Validation antibody-ssDNA conjugates and check of cross-reactivity among antibodies were

performed on DNA spot microarrays (see Fig. S4) printed in-house using the same DNA oligos as

those in microfluidic DNA barcode patterning (see Fig. S3). See Supplementary Ref. 6 for details of

the method.

8. Cell staining

Immediately before experiments, HCT116 cells were detached from the culture dish, rinsed with

PBS and re-suspended in PBS. By following the manufacturer’s instructions, the HCT116 cells were

stained with Vybrant DiI cell-labeling solutions (Life Technologies, USA) and re-suspended in

McCoy's 5A medium. Labeled HCT116 cells were stored on ice and further diluted or spiked into

blood to the desired concentrations before experiments.

Captured tumor cells on the herringbone chips were fixed with 4% paraformaldehyde and

subsequently permeabilized with 0.2% Triton X-100 in 1% BSA, all in PBS. Cells were

immunostained with phycoerythrin (PE)-conjugated anti-pan cytokeratin (CK) and FITC-conjugated

anti-CD45 (BD Biosciences, USA) overnight at 4oC and washed twice with PBS before imaging.

Nuclei were stained with DAPI (Sigma-Aldrich, USA). Microscopic tumor cell enumeration was

performed to detect CK+/CD45-/DAPI+ stain criteria. Cells staining positively for CD45 and DAPI

and negatively for pan-CK are identified as leukocytes.

5

Analysis of cell viability after on-chip release assisted by brief UV irradiation was conducted by

staining on a cell settling chamber described by Ozkumur et al. (7) A simple PDMS chamber was

fabricated and adhered to the PLL-coated glass slide. The released cells were loaded into the chamber

and settled onto the glass surface via gentle centrifugation to generate a cell monolayer. Live/dead

viability/cytotoxicity kit (Life Technologies, USA) was applied on the cells to check the cell viability.

Live and dead cells were visualized and counted with a Nikon Eclipse Ti scope.

9. Protocol for isolation and purification of CTCs spiked into blood samples

In this study, anti-coagulant-containing human whole blood from healthy donors was obtained from

Ruijin hospital affiliated to School of Medicine, Shanghai Jiao Tong University (Shanghai, China). The

healthy donor blood samples were obtained with patient-informed consent under Ruijin hospital review

board approval. According to a standard protocol, all blood samples were collected into EDTA-

contained vacutainer tubes and were processed within 8 h. To demonstrate the performance of the

integrated microfluidics system, cultured EpCAM-positive tumor cells (HCT116) were spiked into

whole blood from healthy donors as the model system. For experiments using lysed blood, RBC lysis

buffer (Beyotime, China) was added to whole blood in 10:1 v/v ratio and mixed for 5 min at RT. After

5-min centrifugation at 500 g, the supernatant was discarded and the cell pellet was washed twice with

PBS and then re-suspended in an equivalent volume of buffer for capture experiments.

In demonstrating performance, ~200 cells were spiked into 1 mL of whole blood in triplicate. Cell

spiking was performed by first diluting the pre-stained HCT116 cells into their corresponding medium

and counting the number of cells in a series of 10 μl spots. Based on the number of cells counted, the

average was calculated for the number of cells per microliter and then additional calculations were

made to obtain the volume required for the number of cells needed to spike.

1 mL of whole blood sample was spiked with HCT116 cells (~200 cells), actively mixed with 1 mL

of anti-EpCAM-PC-M’ conjugate (4 μg/ml), BSA (0.1%), human FcR blocking reagent (20 μL,

Miltenyl Biotec), in PBS for 25 minutes at RT. Whole blood was then centrifuged for 3 minutes at 500

g at 4oC. Supernatant containing excessive anti-EpCAM-PC-M’ conjugate and plasma proteins were

carefully removed by aspiration with use of a vacuum wand. The cell pallet was then re-suspended in

PBS supplemented with 0.1% BSA to generate a 2 mL cell suspension. Cell suspensions were

processed through herringbone chips pre-blocked with 1% BSA in PBS (1 h) at the flow rate of 1 ml /h.

After a wash with 1% BSA/PBS (5 mL/h, 5 min) to remove nonspecifically bound blood cells, a

mixture of magnetic beads coated with CD 45 and CD 15 (R&D Systems, USA) was flowed through

the herringbone chips for binding with non-specific bound WBCs and incubated for 15 min (7),

6

followed by a quick wash to remove excessive magnetic beads. Herringbone chips were then exposed

to UV irradiation for photochemical cleavage of linker between ssDNA and antibody by a near-UV,

low intensity lamp (~365 nm, 1.2 mW/cm2, Cole-Parmer, USA) for 10 min. A fast wash with culture

medium (12 mL/h) was conducted in the last minute of UV irradiation to retrieve all released cells

(CTCs and blood cells) and transport them to and flow through spiral chip for depletion of RBCs.

CTCs and WBCs were collected at the outlet 2 and 3 (see Fig. S13) and flowed through a microfluidic

chip and an array of magnets were used to deplete WBCs.(8) The remaining cells were finally

transported to a PLL-patterned SCBC for on-chip secretion profiling. Importantly, microchannels in

spiral chip and WBC depletion chip, as well as tubing and pins were pre-blocked with 1% BSA to

prevent non-specific adsorption.

10. On-chip selection of specific phenotypic subsets of rare tumor cells from whole blood by a

combination of surface markers

As shown in Fig. S16, HCT116 cell-spiked whole blood was firstly incubated with anti-EpCAM-PC-

M’ conjugates. After centrifuging and resuspension, EpCAM positive HCT166 cells were captured by

the herringbone chip with immobilization of M, complementary to the ssDNA in the conjugate. Upon

brief UV irradiation, EpCAM positive HCT166 cells were photochemically released, followed by

depletion of RBCs and WBCs. The remaining cells were then processed with herringbone chips with

pre-coating of anti-CD44-PC-L’ conjugates to capture EpCAM+CD44+ HCT116 cells, while

EpCAM+CD44- cells were in the solution. In colorectal cancer, EpCAMhighCD44+ cell were found as

robust cancer stem cell phenotype. Thus, this strategy enables selection of specific phenotypic

subpopulations of rare target cells with a combination of surface markers.

11. Selection of specific phenotypic subsets of cultured tumor cells by magnetic beads

For selection of EpCAM+CD44+ and EpCAM+CD44- phenotype from cultured HCT116 cells, a two-

round magnetic bead-based selection process was employed. First, anti-EpCAM antibody (R&D

Systems) was labeled with a modified biotin through DSB-X Biotin Protein Labeling Kit (Invitrogen)

and added to cultured HCT116 cells. After incubation, FlowComp Dynabeads containing a modified

strepavidin were added to labeled cells, thus DSB-X biotinylated antibodies bound to EpCAM positive

cells were captured by modified strepavidin-coupled Dynabeads. Then, a magnet was applied to

separate bead-bound EpCAM positive cells and the supernatant was discarded. Then, a FlowComp

Release Buffer was added to gently detach the Dynabeads from the EpCAM positive cells. In the

second round magnetic bead-based selection, anti-CD44 antibody-labeled magnetic beads (R&D

7

Systems) were applied to the bead-free EpCAM positive cells to separate them into two subpopulations:

EpCAM+CD44+ and EpCAM+CD44- cells. Obviously, this strategy is not suitable for selection of rare

cells from complex biological samples.

12. CTC isolation and identification in lung cancer patient samples

The CTC capture and enumeration from cancer patient samples was performed in a similar

procedure to that described above. Blood samples were obtained from healthy donors and metastatic

lung cancer patients with informed consent, which has been approved by the Ethics and Scientific

Committee of our Institution. A total of 10 blood samples (4 from lung cancer patients and 6 from

healthy donors) were obtained at the middle of vein puncture after first 3 mL of blood was discarded.

According to a standard protocol, all blood samples (4 mL taken from patients, and 2 mL taken from

healthy donors) were collected into EDTA-contained vacutainer tubes and were processed within 6 h.

A volume of 4 mL peripheral blood from patients was equally divided into two parts for CTC

enumeration and single-cell secretion profiling. Briefly, 4 mL of blood was centrifuged to remove the

plasma and re-suspended in 0.1% BSA/PBS to generate a 4 mL cell suspension. It is worth to note that

removal of plasma from patient blood samples before mixing with conjugates is necessary to increase

precision of measurements, although we used whole blood from healthy donors in the proof-of-concept

experiments. Cell suspensions were then actively incubated with a cocktail of photocleavable ssDNA-

antibody conjugates (PC-M’-EpCAM, PC-M’-EGFR, PC-M’-HER2 and PC-M’-CD44). Excessive

conjugates in the supernatant were then removed by centrifugation, and cells were re-suspened in in

0.1% BSA/PBS to generate a new 4 mL cell suspension. The 4 mL cell suspension was equally split

into two parts. A herringbone chip with a 400 mm-long microvotex-generating channel was employed

to capture CTCs from cell suspensions. For the first 2 mL of cell suspension, CTCs were captured on

the chip followed by enumeration based on traditional CK+/CD45-/DAPI+ stain criteria. For another 2

mL of cell suspension, CTCs were captured, released and transported to SCBC for secretome profiling.

All steps were as same as described in proof-of-concept experiments (see Supplementary Methods 9).

For control experiments, 2 mL blood samples from healthy donors were centrifuged to remove plasma,

followed by the same procedure described above. CTC enumeration was based on CK+/CD45-/DAPI+

stain criteria.

13. Protocol for detection of secreted proteins from rare tumor cells using SCBC with a PLL

barcode pattern

8

a) Prior to cell loading, all microfluidic channels were blocked with a blocking buffer (3% BSA in

PBS) for 60 minutes to prevent non-specific binding.

b) After blocking, a 100 l cocktail containing all DNA-antibody conjugates at 10 μg/ml in working

buffer (1% BSA in PBS) was flowed through the micro-channels for 1 hour. Unbound DNA-antibody

conjugates were washed away with 20 l of working buffer.

c) Rare tumor cells collected from WBC depletion chip were flowed into the channels from outlets

of the SCBC 2.0. Tumor cells were then captured by polylysine stripes localized in the middle of

micro-chambers of the SCBC. A control valve was then activated at 20 psi to partition the channel into

multiple discrete chambers. Images of each chamber were recorded using a CCD camera, and used

later for cell counting.

d) The device was then placed in an incubator in humidified atmosphere of 5% CO2 and 95% air at

37°C for 6 hours to allow cellular secretions to accumulate.

e) Afterwards, the control valve was released and a rapid flush was conducted to wash off cells and

unbound secreted molecules with working buffer. A 100 l cocktail containing biotin-labeled

detection antibodies in working buffer at a concentration specified in the insert of that ELISA kit (See

Table S1 for reagents. The concentration varies from lot to lot) was then flowed through the

microchannels. Unbound detection antibodies were then flushed out by flowing the working buffer for 10

minutes.

f) A 100 l mixture of 2 g/ml Cy5-labeled straptavidin (eBioscience, USA) and 50 nM Cy3-labeled

M’ ssDNA in working buffer was flowed through the microchannels.to complete the immune sandwich

assay.

g) Finally, the microchannels were rinsed with working buffer and PBST. The PDMS chip was

removed, and the bare microarray slide was rinsed sequentially with 1PBS, 0.5PBS, and deionized

Millipore water before spin-drying.

h) Optical readout: The slide was scanned by a GenePix 4400A (Molecular Devices) at laser power

80% (635nm) and 10% (532nm), and at 2.5 μm/pixel resolution. Signals from two color channels (the

green Cy3 and the red Cy5) were collected and the average fluorescence signals from each barcode

were extracted by a custom MATLAB (The Mathworks) code for further analysis.

14. Data analysis

The microarray slides (barcode microarray and spot microarray) were scanned with a GenePix

4400A (Molecular Devices, USA) to obtain a fluorescence image for both Cy3 and Cy5 dyes.

9

Complementary DNA validation scans were performed in the 532 nm channel with 10% laser power

and 400 optical gain. Secretion study scans were performed with 80% (635 nm) and 15% (532 nm)

laser power, 600 (635 nm) and 400 (532 nm) optical gain. Spot intensities were quantified with the

software program GenePix Pro 6.0 using the fixed circle method. The averaged fluorescence intensities

for all barcodes were obtained by custom-developed Matlab and Excel code. The SCBC results from the

microchambers with 0, 1, and 2 cells were collected to form a data table. Each row of the table corresponds

to a measurement and each column contains digitized fluorescence intensities that provide the readout of the

levels of each of the assayed proteins.

15. Cell lines and reagents.

HCT116 cells were routinely maintained in McCoy's 5A Modified Medium containing 10% FBS in

humidified atmosphere of 5% CO2 and 95% air at 37oC. ssDNA and ssDNA with photocleavable

linker were synthesized by Sangon Biotech (China). Anti-VEGF, anti-IL-8, anti-EpCAM, anti-CD44,

anti-EGFR, anti-HER2, biotinylated anti-CD45, biotinylated anti-CD15 and PE-conjugated anti-

EpCAM antibodies, as well as Human CD44+ Cancer Stem Cells PlusCellect and streptavidin

ferrofluid were purchased from R&D Systems. PE-conjugated anti-pan cytokerain (CK) and FITC-

conjugated anti-CD45 antibodies were purchased from BD Biosciences. Succinimidyl 4-

hydrazinonicotinate acetone hydrazine (SANH) and succinimidyl 4-formylbenzoate (SFB) were

purchased from Solulink. Vybrant DiI and DiO cell-labeling solutions, as well as the Live/Dead

Viability/Cytotoxicity kit were purchased from Life Technologies. Dynabeads FlowComp Flexi,

including beads and DSB-X Biotin Protein Labeling kit, was obtained from Invitrogen. Negative

photoresist SU8-2100 and positive photoresist SPR220-7 were purchased from MicroChem Corp and

Rohm Hass, respectively. Poly(dimethylsiloxane) (PDMS) RTV615 pre-polymer was purchased from

GE Toshiba Silicones Co. Ltd. Trimethylchlorosilane (TMCS) and DAPI were purchased from Sigma-

Aldrich. Poly-L-Lysine-coated glass slides and Zeba protein desalting spin columns were obtained

from Thermo Fisher Scientific Inc. Amicon Ultra-4 Centrifugal Filter Units were purchased from

Millipore. Red blood cell lysis buffer was obtained from Beyotime (China). UV bench lamp (6 watts,

365 nm, Intensity 1200 μW/cm2) was purchased from Cole-Parmer.

10

Table S1. ssDNAs and antibodies used in this study. The upper part of the table provides the sequences

of the oligonucleotides used in the protein immunoassays. All oligonucleotides were synthesized by

Sangon Biotech (Shanghai, China) and purified via high performance liquid chromatography (HPLC).

The DNA coding oligomers were pre-tested for orthogonality to ensure that cross-hybridization

between non-complementary oligomer strands was negligible (<1% in photon counts). Below the

oligonucleotides is a list of the antibodies and standard proteins used for the multiplex protein assay.

Name DNA sequence Source

C 5'- AAA AAA AAA AAA AGC ACT CGT CTA CTA TCG CTA -3'

C’ 5' NH3-AAA AAA AAA ATA GCG ATA GTA GAC GAG TGC -3'

D 5'-AAA AAA AAA AAA AAT GGT CGA GAT GTC AGA GTA -3'

D’ 5' NH3-AAA AAA AAA ATA CTC TGA CAT CTC GAC CAT -3'

L 5’-AAA AAA AAA AGT GAT TAA GTC TGC TTC GGC -3’

L’ 5' NH3-AAA AAA AAA AGC CGA AGC AGA CTT AAT CAC -3’

L’-PC 5’NH3-AAA AA-PC spacer-A AAA AGC CGA AGC AGA CTT

AAT CAC -3’

M 5'-AAA AAA AAA AGT CGA GGA TTC TGA ACC TGT-3'

M’ 5' NH3-AAA AAA AAA AAC AGG TTC AGA ATC CTC GAC-3'

M’-Cy3 5' Cy3-AAA AAA AAA AAC AGG TTC AGA ATC CTC GAC-3'

M’-PC 5’NH3-AAA AA-PC spacer-A AAA AAC AGG TTC AGA ATC

CTC GAC-3’

A-EcoRI 5' AAA AAA AAA AAA GAG CTA AGT CCG TAG AAT TCA AAA

AAA AAA GAG CTA AGT CCG TAG AAT TCA AAA AAA AAA

AAA-3’

A-EcoRI' 5’NH3-AAA AAA AAA AGA ATT CTA CGG ACT TAG CTC CAG

GAT-3’

B-BamHI 5’-AAA AAA AAA AAA TTG AAT CAT GCC TAG GAT CCA AAA

AAA AAA TTG AAT CAT GCC TAG GAT CCA AAA AAA AAA

AAA-3’

B-BamHI’ 5’NH3-AAA AAA AAA AGG ATC CTA GGC ATG ATT CAA TGA

GGC-3’

Sangon

Biotech

（Shanghai,

China）

11

DNA

label

Antibody Source

C’ Human VEGF DuoSet (DY293B)

D’ Human CXCL8/IL-8 DuoSet (DY208)

M’-PC Human EpCAM/TROP1 DuoSet (DY960)

Human EGF R MAb (Clone 102618)

Human ErbB2 MAb (Clone 191924)

L’-PC,

M’-PC

Human CD44s Pan Specific MAb (Clone 2C5)

R&D Systems

(USA)

Table S2. List of healthy samples as controls

Healthy Sample No. CTCs / 2 mL

1 0

2 1

3 0

4 0

5 0

6 1

12

Supplementary Figures

Figure S1. The overall strategy for on-chip CTC capture, photochemical release, purification and

single-cell secretomic analysis. In brief, EpCAM-positive cells in the whole blood are bound to anti-

EpCAM-PC-ssDNA conjugates (PC: photocleavable linker), and then captured by microvotex-

generating herringbone microfluidic chips with immobilization of complementary ssDNA. A mixture

of CD 45 and CD 15 immunomagnetic beads are then introduced in the microchannels and incubated

with non-specifically bound WBCs. After photochemical release upon UV irradiation, a two-step

purification is conducted on the released cells to deplete contaminated RBCs and WBCs and generate

high-purity CTC population. Finally, the purified CTCs are transported to a PLL-patterned SCBC for

on-chip cell culture and secretion profiling. The enhanced PLL patterning integrated into the SCBC

enables adhering of very low number of target cells at the specific position of the microchambers on

SCBC.

13

Figure S2. Experimental set-up and microfluidic chips. Digital photographs of the integrated

microfluidic system (A), microvotex-generating herringbone chip (B), spiral chip (E) and single-cell

barcode chip (F) are shown, as well as a magnified herringbone structure (C) and immobilized ssDNA

pattern (D) in the herringbone chips. Immobilized ssDNA pattern is detected by complementary

ssDNA functionalized with fluorescent molecules. The single-cell barcode chip is comprised of a two-

layer microfluidic network, the flow layer (red) and the control layer (blue), as visualized in (F).

14

Figure S3. Single-cell barcode chip (SCBC) for quantitative and multiplexed detection of secreted

proteins at the single cell level. The SCBC platform is relatively simple in concept: a single cell, or a

defined number of cells, is isolated within a microchamber. That microchamber also contains an

antibody microarray, patterned as a barcode, that allows for the capture and detection of proteins

secreted from cells (A). The volume of the microchamber is sufficiently small (~1 nanoliters) that

many proteins, when released, are present at detectable levels of antibodies. We can roughly estimate

the sensitivity of the chip. The chamber size is 1 nL. If the molecular weight of the protein is 40 kDa,

and the antibody sensitivity is 10 pg/mL, the sensitivity of the chip is 150 molecules per chamber. This

sensitivity means if a cell secretes more than 150 molecules, that protein can be detected on SCBC. The

15

antibody barcodes are created from DNA barcodes using a DNA-encoded antibody library (DEAL)

technique developed before (ref). For this conversion, capture antibodies are first labeled with distinct

ssDNA oligomers, followed by DNA hybridization to assemble the antibodies onto an array of

(complementary) ssDNA oiligomers (D). In the assays, the created antibody arrays allow for capture of

multiple target proteins in biological samples (B). The signals are then developed by applying a

cocktail of biotinylated detection antibodies and fluorophore-labeled streptavidin (C). The DNA

barcode patterning chemistry leads to highly uniform and intensive barcodes across the entire surface

of the glass slide (E,F). DNA barcodes are stable for fabrication of microfluidic devices.The barcode is

20 micrometers wide, enabling an antibody microarray with a high spatial density, and allowing

multiplexed protein measurement in a very small volume of biological sample.

16

Figure S4. Cross-reactivity check and validation of DNA and antibodies were performed on DNA spot

microarrays printed in house, as described before.(6) The sandwich immunoassays are identical to the

DEAL-based assays used within the SCBC. For device assembly, a 12-well PDMS slab was bonded to

the glass slide with DNA spot microarrays. Each well contains repeated 3 DNA (C, D, M) spot

microarrays with a diameter of each spot 150 μm. (A) Cross-reactivity check of DNA. (B) Cross-

reactivity check of antibodies used in secretion profiling. (C) Validation of anti-EpCAM-M’ and anti-

17

EpCAM-PC-M’ conjugates. The result shows that incorporation of a photocleavable linker in the

conjugate doesn’t cause loss of binding activity. (D) Layout of DNA spot microarrays. The inset shows

a basic unit consisting of 3 DNA spots (C, D, M, comprised of the same ssDNA oligomers used for the

DNA barcodes). Spot diameters are 150 μm. (E) Photograph of a spot-based chip for multiplexed

protein measurement from large numbers of cells, and for antibody cross-reactivity studies. This 12-

well PDMS slab is bonded to the glass slide with the pre-printed DNA spot microarrays. Each well has

a volume of 50 μl and contains a complete set of 3 DNA spots for simultaneously assaying the entire

panel of proteins.

Figure S5. Quality assessments of DNA patterning in herringbone chips. At left is the fluorescence

image of a small area close to the edge of the DNA patterning, with an intensity profile shown at right.

A GenePix 4400A (Molecular Devices) array scanner was used to obtain this image in the 532nm

channel with a laser power set at 10% and an optical gain of 400. This data reflects the uniform DNA

loading (~20,000) across the whole slide. Each DNA-immobilized glass substrates and PDMS

herringbone chips were validated before bonding. The DNA barcode microarrays for SCBC assays

were validated in a similar way, which was described before (6).

18

Figure S6. Representative micrograph image of herringbone-chip-based capture of HCT116 cells that

were spiked in lysed blood.

Figure S7. Pre-stained HCT116 cells (DiI+) were counted on the chip under fluorescence microscope.

19

Figure S8. Counting uncaptured HCT116 cells (DiI+) in the waste based on fluorescence and

morphological features, (A) bright field, (B) fluorescence image. After on-chip capture of HCT116

cells from whole blood, cell suspensions in the waste were collected and lysed to remove RBCs for

manually counting uncaptured HCT116 cells in the entire waste area under microscope.

Figure S9. A putative CTC was detected in a healthy donor based on DAPI+(blue)/CK+(red)/CD45-

(green) criteria.

20

Figure S10. Viability of released cells is calculated based on counting viable cells and dead cells under

microscope (top, fluorescence image; bottom, bright field). Pre-stained tumor cells (DiI+, red) are

stained with calcein AM (green) for counting viable cells.

21

Figure S11. Fluorescent microscopy images of on-chip captured cells under UV irradiation (~365 nm)

and high-rate flowing (12 mL/h), demonstrating cell release upon light irradiation in intact cells. 24

cells (A), 13 cells (B), 7 cells (C) and 1 cell (D) were found under microscope at 0 min, 3 min, 6 min

and 10 min of photcleavable reaction with high-rate flowing, respectively.

Figure S12. 5 μm-sized polystyrene spheres as the model of RBCs were used to determine the

optimized flow rate of spiral chip for RBC depletion. Our study shows all 5 μm-sized polystyrene

spheres leave at the top outlet when flow rate is higher than 12 mL/h.

22

Figure S13. Distribution of spiked tumor cells in different outlets at different flow rates. 100,000 pre-

stained HCT116 cells were spiked in 1 mL of whole blood, and diluted with PBS containing 2 mM

EDTA to generate a cell suspension of 2 mL. When the flow rate is higher than 10 mL/h, nearly all pre-

stained HCT116 cells were found in the middle and bottom outlets (outlet 2 and 3, C, D) with minimal

in the top outlet (outlet 1, B). E shows the bright field and fluorescence image of the outlet 3,

demonstrating small number of pre-stained tumor cells in a huge number of surrounding WBCs.

23

Figure S14. Captured WBCs in the WBC depletion chip closely above arrayed magnets. CD45- and

CD15-coated magnetic beads are found bound on the cells.

Figure S15. Released target cells after blood cell depletion were validated as EpCAM positive cells by

staining with PE-conjugated anti-EpCAM antibody (R&D Systems, USA).

24

Figure S16. The strategy for selection of EpCAM+CD44+ and EpCAM+CD44- HCT116 cells by

double labeling. See Supplementary Methods for details.

Figure S17. The top two sets of channels are loaded with EpCAM+CD44+ cells (cell loading direction:

from right to left), and the bottom three sets of microchannels are loaded with EpCAM+CD44- cells.

There are 80 microchannels on the SCBC, and upon compartmentalization by valve, a total of 2720

microchambers are generated.

25

Figure S18. Representative fluorescence images of single-cell secretion assays of isolated

EpCAM+CD44- HCT116 cells on PLL-patterned SCBC.

26

Figure S19. IL-8 and VEGF measurements on spot-based assays from bulk culture supernatant. To

mimic the concentration of cells (1 cell in a 1 nL microchamber) in SCBC microchamber, 700,000

HCT116 cells (EpCAM+CD44+, EpCAM+CD44- and unsorted cells) were cultured in 2 mL of medium

for 20 h. Positive and negative controls were included. (Top) Representative scanned images of spot

signals from cells. (Bottom) Average intensities of IL-8 and VEGF levels measured from culture

supernatant of EpCAM+CD44+, EpCAM+CD44- and unsorted HCT116 cells

27

Figure S20. Coefficients of variation (CVs) of IL-8 and VEGF levels from 1-cell and 2-cell

microchambers of EpCAM+CD44+ and EpCAM+CD44- cells.

Figure S21. Scatter plots of IL-8 versus VEGF derived from SCBC measurements on CTCs that were

isolated and purified from patients’ 2 mL blood samples.

28

Supplementary References

1. N. G. Zhang, Y. L. Deng, Q. D. Tai, B. R. Cheng, et al. “Electrospun TiO2 nanofiber-based cell

capture assay for detecting circulating tumor cells from colorectal and gastric cancer patients”, Adv.

Mater. 2012, 24, 2756-2760.

2. S. T. Wang, K. Liu, J. Liu, Z. T. -F. Yu, et al. “Highly Efficient Capture of Circulating Tumor Cells

Using Nanostructured Silicon Substrates with Integrated Chaotic Micromixers” Angew. Chem. Int. Ed.

2011, 50, 3084–3088.

3. J. S. Sun, M. M. Li, C. Liu, Y. Zhang, D. B. Liu, W. W. Liu, G. Q. Hu, X. Y. Jiang, "Double spiral

microchannel for label-free tumor cell separtaion and enrichment" Lab Chip 2012, 12, 3952-3960.

4. H. W. Hou, M. E. Warkiani, B. L. Khoo, Z. R. Li, R. A. Soo, D. S. –W. Tan, W. –T. Lim, J. Y. Han,

A. A. S. Bhagat, C. T. Lim, “Isolation and retrieval of circulating tumor cells using centrifugal forces”

Sci. Rep. 2013, 3, 1259.

5. C. Ma, R. Fan, H. Ahmad, Q. H. Shi, B. Comin-Anduix, T. Chodon, R. C. Koya, C. C. Liu, G. A.

Kwong, C. G. Radu, A. Ribas, J. R. Heath, “A clinical microchip for evaluation of single immune cells

reveals high functional heterogeneity in phenotypically similar T cells” Nat. Med. 2011, 17, 738-743.

6. Q. H. Shi, L. D. Qin, W. Wei, F. Geng, R. Fan, Y. S. Shin, D. L. Guo, L. Hood, P. S. Mischel, J. R.

Heath, “Single-cell proteomic chips for profiling intracellular signaling pathways in single tumor cells”

Proc. Natl. Acad. Sci. USA 2012, 109, 419-424.

7. E. Ozkumur, A. M. Shah, J. C. Ciciliano, B L. Emmink, et al. “Inertial focusing for tumor antigen-

dependent and –independent sorting of rare circulating tmor cells”, Sci. Trans. Med. 2013, 5. 179ra47.

8. K. Hoshino, Y. –Y. Huang, N. Lane, M. Huebschman, J. W. Uhr, E. P. Frenkel, X. J. Zhang,

“Microchip-based immunomagnetic detection of circulating tumor cells“ Lab Chip 2011, 11, 3449-

3457.