a| Schematic of the device. There are three inlet ports (labelled I_left, I_middle and I_right), which feed down through a diffusion array to the imaging chamber. b–d| Depending on the make-up of the three inlet ports, various spatial gradients can be achieved. Figure is reproduced, with permission, from REF. 56 © (2001) American Chemical Society.

a| A schematic of the device used by Balaban et al. in their study of bacterial persistence63. The device consists of four layers. On top of the glass base resides a grooved layer of polydimethylsiloxane (PDMS) for growing linear chains of bacteria. A permeable membrane rests on top of the grooved PDMS to constrain the cells to the focal plane and allow media to pass into the growth chamber. Above the membrane lies another layer of PDMS that houses the media flow channel. b–d| Representative time-lapse images of E. coli expressing yellow fluorescent protein taken approximately 45 minutes apart from each other. Figure is reproduced, with permission, from REF. 63 © (2004) American Association for the Advancement of Science.

a| The device designed by Ryley et al. has an array of single-cell ‘jails’ that can trap yeast cells52. b| As cells grow and divide in the jail, the geometry forces the daughter cell outside the jail, where it is flushed away by media flow. Di Carlo et al. designed a similar trap for mammalian cells54. c| Cells (green circles) are loaded through the array, in which they are trapped by low-hanging areas of polydimethylsiloxane (PDMS; light blue). d| Fluid flow helps to keep the cell trapped against the PDMS, and other cells are diverted around the PDMS. e| A microscope image of an array of these traps, showing individual cells. Part a is reproduced and part b is modified, with permission, from REF. 52 © (2006) Wiley. Parts c and d are modified and part e is reproduced, with permission, from REF. 54 © (2006) American Chemical Society.

a| Cells are loaded through reservoirs leading into port C and trapped in the imaging region (which is enlarged in part b). Once enough cells are trapped, the flow is reversed by increasing the pressure in the media inlet, port M. Excess media is routed through port W (waste). Heated water is passed through two channels (labelled T1 and T2) on either side of the imaging chamber to control the temperature on the chip. b| A close-up diagram of the imaging region. Once the flow has reversed, the trapping region (grey) is isolated from the flow and only receives nutrients through diffusion. c| A typical fluorescence image of Saccharomyces cerevisiae expressing GFP. The height of the imaging chamber is 4 μm, which forces the cells to grow in a monolayer and creates a uniform focal plane. Outside the chamber, the channels are twice as tall, which allows cells to be easily washed away by the flow. Figure is reproduced, with permission, from REF. 51 © (2005) Macmillan Publishers Ltd. All rights reserved.

a| The overall design of the dial-a-wave device is similar to that of the Tesla microchemostat51. Cell input channels (dark blue) feed into an imaging chamber (grey), in which cells are trapped for long periods of time. The growth medium for the imaging chamber is supplied by a switching channel system (switch; green channels) that can provide programmable, time-dependent concentration waves. b| As the growth medium (in the green channel) flows near the imaging chamber, small perfusion channels (light blue) allow the medium to diffuse into the cell culture. c| The switch region determines the make-up of the medium supplied to the imaging chamber. Two inputs, labelled input 1 and input 2 in a and c, feed into the switch region. Owing to the small flow rates and the geometry of the channels, the interface between the two inputs is laminar, which allows for precise control over the mixing ratio. d| By differentially pressurizing the two input reservoirs, the laminar interface can be moved from one side of the switching region to the other. In this way, the relative concentrations of the two media can be programmed to any desired waveform. Figure is modified, with permission, from Nature REF. 60 © (2008) Macmillan Publishers Ltd. All rights reserved.

a| The concentrations of key components of the growth media are controlled, or ‘driven’, by a fluidic waveform generator. In the chip, cells are continually measured with fluorescence microscopy as they react to the changing media conditions. Setups similar to this have been used by several groups to explore signalling pathways in Saccharomyces cerevisiae. b| Mettetal et al.58 drove the high-osmolarity glycerol (HOG) pathway in yeast and discovered key signalling characteristics for wild-type and network-deficient strains. c| The dynamic data obtained were used to validate and refine the mathematical model. The graph compares wild-type S. cerevisiae with a strain expressing a reduced amount of Pbs2 (polymyxin b sensitivity 2). d| Bennett et al.60 drove the galactose utilization network in yeast with periodic carbon source changes and found new network connections. e| The dynamic data obtained from the experiment were used to validate and refine the mathematical simulation. The data show how the network created a low-pass filter. YPH499 is an S. cerevisiae strain that exhibits reduced galactose sensitivity in static environments. a.u., arbitrary units; Gal, galactose; Gal3^★, Gal3 bound to galactose; Glc, glucose; Gpd1, glycerol-3-phosphate dehydrogenase 1; Gpp2, (DL)-glycerol-3-phosphatase 2; P, pressure. Parts b and c are reproduced, with permission, from REF. 58 © (2008) American Association for the Advancement of Science. Parts c–e are reproduced, with permission, from Nature REF. 60 © (2008) Macmillan Publishers Ltd. All rights reserved.