Cell-surface sensors for real-time probing of cellular environments.

Engineering aptamer sensors onto the MSC surface MSCs (~1 M after trypsinization) (see Supplementary Information for routine MSC culture) were dispersed in biotin–NHS solution (1 mM in PBS–/–, 1 ml), and the solution was allowed to incubate for 10 min at room temperature. After washing, streptavidin solution (50 μg ml –1 in PBS–/–, 1 ml) was then used to treat the cells for 5 min. Finally, biotin-modified sensor solution (two DNA strands were first annealed at 5 μM each in PBS +/+ at 90 °C for 3 min and cooled at room temperature for 30 min, 200 μl) was added, and the suspension was incubated for 5 min at room temperature. The cells were then washed once by PBS–/– and subsequently used for experimentation. The fluorescent nature of the sensors allowed the conjugation step to be easily followed by flow cytometry (BD FACS Calibur flow cytometer) then analysed using Cell Quest software. Microneedle experiment Microneedle experiments were performed using a microinjector (FemtoJet, Eppendorf) with Eppendorf Femtotips and an Eppendorf micromanipulator (InjectMan NI 2, Eppendorf) ( Supplementary Fig. S5 ). Glass microneedles with inner tip diameters of ~3 μm were made using a micropipette puller (P-97 Sutter Instrument Company). Microneedles were backfilled with the PDGF (2 μM in PBS–/–) using Eppendorf Femtotips Capillary Pipet Tips Microloaders. The microneedle, controlled by a micromanipulator, was lowered onto a dish with sensor-engineered MSCs settled on the surface in PBS–/–, positioned at a defined lateral distance (~40 μm) from the settled cells and at a height of ~30 μm from the underlying substrate. PDGF was released from the micropipette by applying a defined pressure (26 hPa). Simultaneously, phase contrast and fluorescence images of the cells were collected sequentially with a 1 s interval exposure time. Between 10 and 12 cells were measured in each experiment. Computational model for PDGF transport in microneedle experiment See Supplementary Information .

Cell–cell communication microwell assay and analysis

Time-course microwell assays were performed following previously reported protocols with some modifications 38 . The fabrication of the microwell is presented in the Supplementary Information . A sensor MSC suspension (1 × 10 5 cells/ml) was first placed on the surface of the array and cells were permitted to settle into the microwells by gravity. After 2 min, excess cells were washed away with serum-free media. Next, PDGF-producing MDA-MB-231 cells (1 × 10 5 cells/ml) in medium containing 15% FBS were loaded into the wells as described above. After a brief incubation at 37 °C with 5% CO 2 , the array was delivered to the microscope for imaging. All images were acquired on an automated inverted fluorescence microscope (Zeiss Observer Z-1, Carl Zeiss Inc.) equipped with a stage incubator (PM S1) and incubation chamber for live-cell imaging (37 °C, 5% CO 2 ). Phase and fluorescence (GFP and Cy5) micrographs were collected every 3 min for 6 h. A total of 3000 microwells were imaged at each time point. A custom-written image analysis program was used to identify the location and fluorescence intensity of each cell in the microwell array, as described previously 38 . A MATLAB script was written to track the fluorescence signal intensity of each sensor MSC over the 6 h time course. The signal intensity of each sensor MSC was normalized to the signal intensity at t = 0 min to account for the baseline cell-to-cell variation in sensor MSC intensity. Sensor MSCs were divided into groups based on the number of PDGF-producing MDA-MB-231 cells residing in the same microwell (0, 1, 2 or 3+). More than 100 MSCs (see Fig. 5 for exact numbers) from each group were tracked. The fraction of sensor MSCs in each group with signal intensity less than 50% of the initial signal intensity was calculated at each time point.

Supplementary Material 1 Corresponding Author Jeffrey M. Karp Co-Director of the Center for Regenerative Therapeutics Brigham and Women's Hospital, Room 313, PRB 65 Landsdowne Street Cambridge, MA 02139, USA Tel: 617-817 9174 Fax: 617-768-8338 jkarp@rics.bwh.harvard.edu Materials All DNA (names and sequences in Table S1) were obtained from Integrated DNA Technologies (IDT). Recombinant PDGF-BB and PDGF-BB ELISA kit were purchased from R&D systems (Minneapolis, MN). Primary human MSCs were obtained from the Center for Gene Therapy at Texas A&M which has a grant from NCRR of the NIH, Grant # P40RR017447. PDGF-BB producing human MDA-MB-231 cells, provided by Prof. Jain at Massachusetts General Hospital, were genetically engineered using retrovirus transduction following a previously established protocol. 1 α-MEM, Fetal Bovine Serum (FBS), LGlutamine and Penn-Strep were purchased from Invitrogen. Sulfonated biotinyl-N-hydroxy-succinimide (NHS-Biotin), streptavidin, trypsin/EDTA solution, PBS-/- (NaCl 137mM, KCl 26.8mM, Na 2 HPO 4 8.1mM, KH 2 PO 4 1.5mM) and PBS+/+ (PBS-/- supplemented with 0.9 mM CaCl 2 and 0.5 mM MgCl 2 ) were purchased from Sigma.

Mesenchymal Stem Cell Culture and Characterization

Primary human MSCs were isolated from human marrow of healthy consenting donors and thoroughly characterized as previously described before sending to the researchers 2, 3 . Before use, we further confirmed the basic characteristics (i.e. CD90+, CD29+, CD106+, CD34-, CD45-) of MSCs using flow cytometry. MSCs were adherent on tissue culture plate and were maintained in MSC medium that consisted of 15% FBS, 1% (v/v) L-Glutamine, 1% (v/v) Penn-Strep, and α-MEM. To detach cells from plates for passaging or experiments, cells were incubated in 1 × trypsin/EDTA solution at 37 °C for 3 min. MSCs at passage number 4-6 with confluency of ~ 80% were used for all experiments.

Computational model for PDGF transport in microneedle experiment

We built a computational model to estimate the local concentration of the PDGF near the cell. The following simplifying assumptions were made: 1. The flow is determined primarily by the direction and magnitude of injection velocity, and pipette body has minimal effect on the flow profile. This allows us to model the pipette as a thin vertical tube (Figure S6) and the direction of injection (30°) and magnitude of velocity (100 μm/s) are similar to those used in the experiment. 2. The flow is assumed to be symmetric about the pipette (one vertical plane of symmetry) allowing us to model only half of the computational domain. 3. It is assumed that the cells do not alter the flow pattern appreciably. Thus, we model only one cell (the cell of interest, which was photographed in the micro needle experiment), as a hemispherical cap, in our computational domain. 4. The cell surface concentration of aptamer was assumed to be low such that binding of PDGF on the surface does not appreciable alter the local PDGF concentration. The computational domain was created and meshed in the commercial software GAMBIT (preprocessor of FLUENT, Ansys Inc.) using tetrahedral elements with edge lengths graded from 1 μm (boundary elements) to 3 μm (elements in the bulk fluid) (Figure S6). The meshed volume was exported into the computational software FLUENT (Ansys, Inc.) and the appropriate boundary conditions were applied (Figure S6). An unsteady incompressible laminar fluid flow model along with non-reacting species transport was chosen. This model uses finite volume method to discretize the continuity, Navier-Stokes and the mass transport equations shown below (gravity was neglected): ∇ • ( ρ ϑ → ) = 0 ∂ ϑ t ( ρ ϑ → ) + ϑ → • ∇ ( ρ ϑ → ) = − ∇ P + μ ∇ 2 ϑ → ∂ C ∂ t + ϑ → • ∇ C = D ∇ 2 C where ρ is the fluid density, ϑ is the fluid velocity vector in cartesian coordinates, P is the static pressure, μ is fluid viscosity, C is the concentration of the transported species and D is the coefficient of diffusion of the species in the medium. The material properties used in our simulations were: ρ=998.2 kg/m 3 , μ= 0.001003 kg/m-s, molecular weight of water = 18.01 Da, molecular weight of PDGF-BB=24.3 kDa, D=1×10 -10 m 2 /s 4 . A segregated solver along with 1 st order implicit time stepping method was used. The pressure was discretized using PRESTO scheme while the momentum and species equation used a 2 nd order upwind scheme (both are inbuilt options in the software). The injected stream is assumed to have a PDGF mass fraction of 1 (accordingly, the calculated mass fraction of PDGF is interpreted as the concentration relative to the injected value). Unsteady simulations were performed with time step of 0.1 s with a maximum of 50 iterations per time step. The solution was terminated when all the residuals were below 10 -4 . The initial condition was no flow, and no PDGF present in the geometry (i.e. mass fraction of water is 1). The simulation was run in double precision mode. PDGF-BB production by MDA-MB-231 cells determined by ELISA The amount of PDGF-BB produced by genetically engineered or wild-type MDA-MB-231 cells was quantified using an ELISA kit following a protocol provided by R&D Systems. It was determined that PDGF is continuously secreted to a total concentration of ~1.75 nM in 6 h, which falls in the detection range of our cell surface sensors. Fabrication of microwell Microwell arrays were prepared by injecting a silicone elastomer mixture (polydimethylsiloxane (PDMS), Dow Corning Inc.) into a mold and curing at 70°C for 2h. The prepared arrays were 1 mm thick and bound to a glass slide. Each array consisted of ~85,000 microwells (each 50 μm × 50 μm × 50 μm) arranged in 7 × 7 blocks. Arrays were treated for 30 s in an oxygen plasma chamber (Harrick PDC-32G) to render the surface sterile and hydrophilic.

Animals and intravital confocal microscopy

All mice were housed according to IACUC guidelines and used for experiments when 6-10 weeks old. Aptamer-modified MSCs (labeled with DiD) and native MSC (labeled with DiO) or FRET probe-modified MSCs (10 6 cells/100 μL PBS) were injected via retro-orbital injection into anesthetized female Balb/C mice. 24 h after the injection, the mice were anaesthetized by an intraperitoneal injection of a Ketamine/Xylazine cocktail at 100 mg/kg / 15 mg/kg body weight and prepared for in vivo imaging by making a small incision in the scalp to expose the underlying dorsal surface of the skull. High-resolution images were acquired through the intact mouse skull of a live mouse using a non-commercial confocal/multi-photon microscope specifically designed for live animal imaging. 7, 8 Circulating blood was stained through delivery of rhodamine-dextran conjugate (Invitrogen, 70kDa) at 10 mg/kg just before imaging. DiO was imaged using 491nm excitation (Dual Calypso, Cobolt AB, Stockholm, Sweden) and 509-547 nm emission (Semrock, Rochester, NY), DiD was imaged using 635nm excitation (Radius, Coherent, Santa Clara, CA) and 667-722 nm emission (Omega Optical, Brattleboro, VT). Rhodamine was imaged using 561 nm excitation (Jive, Cobolt AB, Stockholm, Sweden) and 573-613 nm emission (Semrock, Rochester, NY) whereas collagen in bone was imaged via second harmonic generation using an 880nm incident two-photon laser source (Mai Tai HP, Spectra-Physics, Irvine, CA) and a 432-482 nm bandpass filter (Semrock, Rochester, NY). Z-sections were acquired using 5 um steps. To measure extravasation, fluorescent images from different channels must remain aligned during image acquisition, therefore rhodamine, DiO, and DiD images were acquired simultaneously during z-stack acquisition. Image analysis was performed using ImageJ (NIH, Bethesda, MD).

Statistical differences between

MSC and aptamer-MSC transmigration were determined using the Fisher exact test. For experiments where MSCs were labeled with FRET probe, fluorescence images were acquired before and after photobleaching of the FRET acceptor (Cy5). Images were acquired using 532 nm excitation from a solid-state laser (Dual Calypso, Cobolt AB, Stockholm, Sweden) and the Cy3 and Cy5 fluorescence were detected through a 570-620 nm bandpass filter (Chroma Technologies, Rockingham, VT) and a 650-760 nm bandpass filter (Chroma Technologies, Rockingham, VT), respectively. The frame rate of the microscope was fifteen frames per second. Images were captured, after averaging fifteen frames, using a Macintosh computer equipped with an Active Silicon snapper card (Active Silicon Chelmsford, MA). Each channel was acquired individually, but simultaneously, in 8-bit grayscale and merged to form an RGB image, with Cy3 fluorescence in the green channel and Cy5 fluorescence in the red channel, using custom developed software (iPhoton). The FRET acceptor, Cy5, was photobleached by illuminating the region of interest (330 × 330 μm) with approximately 3.6 mW of power (measured after the microscope objective) for 3 minutes using a 635 nm helium-neon laser (Radius, Coherent Inc., Santa Clara, CA). Table S1. Names and sequences of DNA molecules used in this study. Figure S1. Performance of original and engineered PDGF sensor in solution (PBS+/+). The sensor was engineered by changing a C-G base pair to A- -G non base pair. Y axis represents the signal that is defined as the ratio of fluorescence, obtained by flow cytometry, before and after addition of PDGF. 10 nM sensor molecules were used in this experiment and spectra were recorded immediately at room temperature. Figure S2. Binding specificity of the PDGF aptamer sensor on the MSC surface. Y axis is the ratio of fluorescence signal (at 520 nm) before and after adding tested molecule (20 nM in PBS). bFGF: basic fibroblast growth factor; IGF-1: insulin-like growth factor-1. Figure S3. Calibration curve of the FRET sensor on the MSC surface for quantifying PDGF concentration. Signal is defined as the fluorescence decrease of donor dye (Cy3) × fluorescence increase of acceptor dye (Cy5) on MSC at a particular PDGF concentration. Each signal value (y-axis) represents the average fluorescence intensity from ~ 60 individual cells imaged by fluorescent microscopy at 37°C in PBS-/-. FRET sensor modified MSCs were excited by a 568 nm laser and emissions were measured at 607 nm and 670 nm for Cy3 and Cy5, respectively. Figure S4 . (a) Performance of the quench sensor and (b) FRET sensor immobilized on the MSC surface in different solution. Medium contains 15% FBS, 1% (v/v) L-Glutamine, 1% (v/v) Penn-Strep, and α-MEM. Signal for the quench sensor is defined as the ratio of geometric means of the flow cytometry histogram before and after addition of PDGF. Signal for the FRET sensor is defined as the fluorescence decrease of donor dye (Cy3) × fluorescence increase of acceptor dye (Cy5) based on the geometric means in the flow cytometry histogram. 20 nM PDGF was used and measurement was performed immediately at room temperature. Figure S5. A representative example how PDGF was injected by microneedle. Figure S6: Modeling of PDGF concentration around the sensor-cell. a) Computational domain used for modeling the PDGF transport in the microneedle experiment. The boundary conditions are shown (blue text) along with their values (red text). The dimensions are also shown. Note that the flow is determined primarily by the direction and magnitude of injection velocity, and pipette body has minimal effect on the flow profile. This allows us to model the pipette as a thin vertical tube (a) and the direction of injection (30°) and magnitude of velocity (100 μm/s) are similar to those used in the experiment. b) Discretized computational domain showing the tetrahedral elements used for meshing. Figure S7 . No significant signal difference on sensor-MSCs in the presence and absence of non-genetically engineered MDA-MB-231. The signal is defined as the percentage of MSCs that have fluorescence intensity less than 50% of their initial value at the indicated time. Note that the fluorescence intensity of the quench sensor on the cell surface may be influenced by external factors including the sensor site density per cell and medium conditions including pH, temperature, and composition, which may explain the difference observed in the overall fluorescence quenching between this experiment and the experiment reported in Figure 5 . Figure S8. Representative in vivo IVM images of FRET-probe-modified MSCs that homed to the skull bone marrow in the live mice. Images were taken 24 h after systemic infusion of cells via retro-orbital injection. a, c and b, d are images acquired before and after in situ photobleaching of Cy5, respectively. Cy3 fluorescence was acquired in the green channel and Cy5 fluorescence was acquired in the red channel. Scale bar is 50 μm. Figure S9. The performance of the quench sensor using Cy5/Iowa Black RQ before and after addition of PDGF (10 nM) in PBS-/-. References 1. Au, P. et al. Paradoxical effects of PDGF-BB overexpression in endothelial cells on engineered blood vessels in vivo. The American Journal of Pathology 175 , 294-302 (2009). 2. Sekiya, I. et al. Expansion of human adult stem cells from bone marrow stroma: Conditions that maximize the yields of early progenitors and evaluate their quality. Stem Cells 20 , 530-541 (2002). 3. Sarkar, D. et al.

Chemical Engineering of Mesenchymal Stem

Cells to Induce a Cell Rolling Response. Bioconjugate Chemistry 19 , 2105-2109 (2008). 4. Haugh, J.M. Deterministic model of dermal wound invasion incorporating receptor-mediated signal transduction and spatial gradient sensing. Biophysical Journal 90 , 2297-2308 (2006). 5. Ogunniyi, A.O., Story, C.M., Papa, E., Guillen, E. & Love, J.C. Screening individual hybridomas by microengraving to discover monoclonal antibodies. Nature Protocol 4 , 767-782 (2009). 6. Han, Q., Bradshaw, E.M., Nilsson, B., Hafler, D.A. & Love, J.C. Multidimensional analysis of the frequencies and rates of cytokine secretion from single cells by quantitative microengraving. Lab Chip 10 , 1391-1400 (2010). 7. Runnels, J.M. et al. Imaging molecular expression on vascular endothelial cells by in vivo immunofluorescence microscopy. Mol Imaging 5 , 31-40 (2006). 8. Celso, C. et al. Live-animal tracking of individual haematopoietic stem/progenitor cells in their niche. Nature 457 , 92-97 (2009).