r 1999 Wiley-Liss, Inc.
Cytometry 37:156–159 (1999)
Plug Flow Cytometry: An Automated Coupling Device for Rapid Sequential Flow Cytometric Sample Analysis Bruce S. Edwards, Frederick Kuckuck, and Larry A. Sklar* Cancer Research and Treatment Center, Departments of Cytometry and Pathology, University of New Mexico Health Sciences Center, Albuquerque, New Mexico Received 18 February 1999; Revision Received 23 April 1999; Accepted 10 June 1999
Background: The tools for high throughput flow cytometry have been limited in part because of the requirement that the samples must flow under pressure. We describe a simple system for sampling repetitively from an open vessel. Methods: Under computer control, the sample is loaded into a sample loop in a reciprocating eight-way valve by the action of a syringe. When the valve position is switched, the plug of sample in the sample loop is transported to the flow cytometer by a pressure-driven fluid line. By coupling the plug-forming capability to a second multi-port valve, samples can be delivered sequentially from separate vessels.
Results: The valve is able to deliver samples at rates ranging up to about 9 samples per minute. Each plug of sample has uniform delivery characteristics with a reproducible coefficient of variation (CV). Even at the highest sampling rate, carryover between samples is limited. Conclusions: Plug-flow flow cytometry has the potential to automate the delivery of small samples from unpressurized sources at rates compatible with many screening and assay applications. Cytometry 37:156–159, 1999.
Flow cytometry is often used in routine assays, but automation is not normally considered to be part of the flow cytometry repertoire. Drug discovery efforts are currently focused on screening large combinatorial libraries of compounds that potentially alter or mimic receptor– ligand interactions. Flow cytometry has the potential of being a sensitive means of measuring such receptor–ligand interactions (1). If samples can be handled rapidly, and suitable receptor binding assays developed, flow cytometry could be competitive with other techniques for high throughput screening of such libraries and for many diagnostic applications. Recently, flow cytometers have been successfully interfaced with computer-controlled syringes and valves to enable rapid sample handling for subsecond kinetics measurements of molecular interactions (2–5). Most flow cytometers require samples to be pressurized for delivery to the point of analysis in the laser beam, a feature that introduces important constraints upon the design of an automated system. For example, the samplecontaining vessel is limited to a shape and size compatible with the dimensions of the flow cytometer sampling port and amenable to formation of an airtight seal. When the sample vessel is pressurized, access to the sample is restricted, which is a constraint upon reagent addition in automated kinetic analyses. Commercial automated flow cytometry systems presently exist in which individual samples are sequentially drawn from racks of tubes de-
signed to fit the sample port. The sampling apparatus in these devices is cleared between samples. The present work describes an alternative approach to automated sample handling. The concept is that there is a continuously flowing stream of fluid into which individual samples are sequentially inserted as a bolus or ‘‘plug’’ of precisely defined volume. The stream delivers the sample plugs, separated by empty volumes of fluid (which consists of the same buffer of which the sample stream is composed), to the point of analysis in the laser beam. The present work demonstrates analysis of multiple 5-µl sample plugs at rates of up to 9 plugs per minute with distinct resolution of the individual samples and highly reproducible distributions of sample fluorescence. Because an automated syringe was used for sample uptake, pressurization of the sample-containing vessel was unnecessary. In principle, faster sampling rates should be possible with improved microfluidic control and smaller sample volumes.
r 1999 Wiley-Liss, Inc. Key terms: flow cytometry; automation; sample handling
MATERIALS AND METHODS Cells and Microspheres Chinese hamster ovary (CHO) cells transfected with the green fluorescent protein (GFP) encoding expression Grant sponsor: University of New Mexico; Grant sponsor: National Institutes of Health; Grant numbers: RR11830, HL56384, RR01315. *Correspondence to: Larry A. Sklar, CRTC, 2325 Camino de Salud, University of New Mexico, Albuquerque, NM 87131. E-mail:
[email protected]
PLUG-FLOW CYTOMETRY
157
vector pTracer (Invitrogen, San Diego, CA) were obtained from Dr. Richard Larson (University of New Mexico, Albuquerque, NM). Parental untransfected CHO cells were labeled with hydroethidine (Molecular Probes, Eugene, OR) by incubating 2 ⫻ 106 CHO cells for 15 min at 37°C in 0.5 ml RPMI medium supplemented with 10% fetal bovine serum and 80 g/ml hydroethidine. CHO cells were washed by centrifugation in 1 ml HHB buffer (110 mM NaCl, 10 mM KCl, 1 mM MgCl2, 1.5 mM CaCl2, 30 mM HEPES, 10 mM glucose, and 0.1% human serum albumin), then resuspended in HHB and stored on ice until analyzed in the flow cytometer. Immuno-Check fluorescent microspheres were obtained from Coulter Corporation (Hialeah, FL). Sample Handling and Control Components Samples were aspirated with a Cavro XL3000 modular digital pump equipped with a 500 µl syringe barrel (Cavro Scientific Instruments, Sunnyvale, CA) and coupled to a VICI C22Z two-position eight-port switching and sampling valve with micro-electric actuator (Valco Instruments, Houston, TX). The syringe pump and valve were controlled by computer via a bi-directional RS232 port interface in which signals were routed through a four-port Mini Smart Switch (B&B Electronics, Ottawa, IL). C program software to control the RS232 interface and associated hardware was developed in this laboratory. Inserted into the valve were two sample loops, which consisted of 10-cm lengths of 0.01-inch inner diameter (ID) Teflon tubing with an internal volume of approximately 5 µl (Upchurch Scientific, Oak Harbor, WA). A standard Falcon tube (25 ⫻ 100 mm) filled with distilled water or phosphate-buffered saline (GIBCO, Grand Island, NY) was inserted into the sample port of a Coulter Elite flow cytometer as a reservoir of pressurized fluid. The fluid was continuously driven under pressure to the valve, through one of the sample loops, and on to the flow nozzle. Teflon tubing of 0.01-inch ID was used to transport samples from the sample vessel to one of the valve sample loops, and to connect the loop to the aspirating syringe. In some experiments, the sample input line from the two-position valve was joined to the common port of a six-position valve (VICI C25Z, Valco) for sequential sampling from separate sample vessels as described below. Suspensions of cells or microspheres were continuously stirred with a magnetic bar during sampling. RESULTS AND DISCUSSION Illustrated in Figure 1 is a schematic representation of the device by which ‘‘plugs’’ of sample are delivered to the flow cytometer. The plugs of sample are created by the integrated actions of a syringe, a reciprocating twoposition eight-port valve, and a pressure-driven fluid line. The sampling syringe first draws a sample into a sample loop (Fig. 1, top, loop A); when the valve position switches, the sample loop becomes connected to the fluid line that then moves the sample out of the loop and into the flow cytometer (Fig. 1, bottom). As the first sample flows to the point of analysis, the syringe draws a second sample
FIG. 1. Schematic of the automated delivery of sample plugs to the flow cytometer. Plugs of sample are created by the action of a syringe, a reciprocating two-state eight-port valve, and a pressure driven fluid line. Top panel: The syringe draws sample from an unpressurized vessel into sample loop ‘‘A’’ while transport fluid is driven under pressure through sample loop ‘‘B’’. Bottom panel: The valve switches to its alternate state so that sample loop ‘‘A’’ is connected to the transport fluid line. The sample plug is cleared from loop ‘‘A’’ and delivered to the flow cytometer by the continuously flowing transport fluid. Concurrently, the syringe draws a sample from the vessel into sample loop ‘‘B.’’ A subsequent switch of the valve to its original state promotes delivery of the sample plug in loop ‘‘B’’ to the flow cytometer by the transport fluid while sample loop ‘‘A’’ is reloaded, and so on. The rate of sample plug analysis is determined by the frequency of valve switching and the transport fluid flow rate.
aliquot into sample loop B (Fig. 1, bottom). After another valve switch, this sample is pushed to the flow cytometer concomitantly with the drawing of a third sample into loop A (not shown). Multiple samples are analyzed by repeating this alternating cycle at a frequency determined by the controlling software. When a stirred suspension of fluorescent microspheres was repetitively sampled from a nonpressurized sample reservoir (Fig. 2A), the multiple sample plugs were clearly resolved as discrete clusters over the sampling time domain. The number of microspheres in each sample plug (represented by the series of peaks at the bottom of the panel) was 502 ⫾ 35 (mean ⫾ SD). This indicated a concentration of 100,400 microspheres/ml in the sample vessel with a 7% coefficient of variation (CV) for this estimate. As expected of repetitive sampling from a homogeneous stirred suspension, the fluorescence intensity profiles of the microspheres (represented by the series
158
EDWARDS ET AL.
FIG. 2. Characterization of sample plug fluorescence profiles and cross-contamination. A: Sample plugs of fluorescent microspheres were repetitively aspirated from a sample vessel and analyzed at a frequency of 7 sample plugs/min. Illustrated are the log fluorescence intensity profiles of microspheres (dot plots at top) and the temporal distribution of microsphere numbers (peaks at bottom) detected over an 80 s time interval. B: To quantify cross-contamination between sample plugs, alternating samples were taken from two separate vessels containing distinctly red and green fluorescently labeled Chinese hamster ovary (CHO) cell preparations. CHO cells were sampled at 7 sample plugs/min and analyzed for green fluorescence of cells from one vessel (black dot plots at top), red fluorescence of cells from the other vessel (gray dot plots in the middle), and the number of cells of each color (black and gray peaks at the bottom, respectively) in each sample plug. A total of 9 sample plugs was analyzed, of which the last 8 (4 of each color) were evaluated to determine the mean and SD percent of total cells represented by inappropriately fluorescent cells from preceding sample plugs. Contaminating cells represented 2.2 ⫾ 1.6% of total cells in each sample plug peak.
of dot plots at the top of panel A) were also highly reproducible over the entire sampling interval. An important issue in the analysis of multiple sequential sample plugs is to be able to measure and minimize contamination of one sample plug with particles from the preceding sample plug. To assess this parameter, an additional multi-port valve was incorporated in the sampling apparatus. The new valve permitted the sample acquisition port of the original valve to be connected separately to each of six discrete sample ports under computer control. Two of these ports were used to acquire sample plugs from two separate sample vessels containing CHO cells with distinct fluorescent labels: green fluorescence from green fluorescent protein and red fluorescence from hydroethidine. Sampling at 7 sample plugs per minute resulted in spillover between successive samples that corresponded to 2.2 ⫾ 1.6% (mean ⫾ SD) of total cell counts in each sample (Fig. 2B). At 9 sample plugs per minute, the individual sample plugs were still clearly resolved, but the extent of cross-contamination between sample plugs increased to 4.4 ⫾ 0.8% (data not shown).
In order to distinguish samples as discrete populations at high sampling rates, it was necessary to rapidly clear sample plugs from the sample loops en route to the flow cytometer. This was accomplished by maintaining a relatively high fluid flow rate in the transport fluid line (⬇3.6 µl/s). As a consequence of such a high flow rate, we observed a broadening of light scatter and fluorescence intensity profiles that reflected a proportion of particles that were off the optical alignment axis when passing through the laser beam. For well-defined mono-sized particles, such as microspheres or purified cells and cell lines, light scatter gating is an option by which off-axis passages may be eliminated from analysis. In an analysis of multiple sample plugs of fluorescent microspheres flowing at 3.6 µl/s, we determined that ⬇30% of the microspheres within each sample plug fell within light scatter gates characteristic of optimally aligned microspheres. When fluorescence intensity was measured on a linear scale, the average fluorescence CV was 5.2% for the light-scatter-gated microspheres under plug flow conditions, as compared to 3.3% when the microspheres were analyzed under optimal alignment conditions (data not shown). The automated sampling system described here has several important features that promise to increase the usefulness of flow cytometry for repetitive screening assays. First, sample pressurization is not required, a feature which, for example, makes feasible the use of multi-well microplates for sample containment and presentation. Second, because sample plugs are of a precisely defined volume (5 µl in the present system), particle concentrations are directly determined from the total particle counts in each sample. Third, discrete sample populations are successfully resolved and analyzed at rates of up to 9 samples per min (⬇7 s per sample) in the present system. There are several important criteria for the usefulness of this system in screening or other repetitive sampling applications. These include several interdependent variables: (i) the fluorescence CV requirements necessary for proper sample evaluation; (ii) the minimal acceptable levels of cross-contamination between sample plugs; and (iii) the minimal required rate of sample throughput. Fluorescence CVs are dependent upon sample alignment and can be improved by slowing the flow velocity in the fluid line so that the bulk of the sample particles flow along the alignment axis. This approach assures accurate analysis of every particle in the sample at the expense of reduced sample throughput. Alternatively, at high fluid flow rates, light scatter gates may be implemented to eliminate particles that fall off the alignment axis. This latter approach permits maximal sample throughput under circumstances in which the particles are known to be of a single uniform size, and it is not essential to analyze every particle in the sample. Cross-contamination between sample plugs reflects the overlap of the leading edge of one sample plug with the trailing tail of the preceding sample plug. Another source of contamination is particles from the preceding sample
PLUG-FLOW CYTOMETRY
plug that are trapped in the dead volume of valve junctions that join sample loops to the sampling and fluid flow lines. For assays in which precise quantitative accuracy is not required, the observed levels of sample cross-contamination (2–4%) are likely to be acceptable. If reduction of sample cross-contamination to lower levels is absolutely necessary, this could be accomplished by increasing the sampling interval and/or reciprocating the sampling valve multiple times for each sample to completely clear the valve junction dead volume. Such approaches would be expected to decrease the maximal sample acquisition rate. An alternative approach, compatible with high throughput sampling, would be to fluorescently encode the particles in alternating adjacent sample wells with two distinctive fluorescent tags so that each successive sample has a tag distinct from that of the preceding sample. Such tags might include fluorochromes with distinctive fluorescence spectra (as in Fig. 2B) or a single fluorochrome with distinctively discrete levels of fluorescence intensity. The parameter to be assayed would then be measured with a detection reagent with fluorescence properties distinct from the encoding tag(s). By establishing fluorescence analysis gates that exclude particles with the inappropriate tag, it should be possible to virtually eliminate contaminating particles from the analysis of each sample. Such an approach would be implemented at the cost of losing a potential fluorescence measurement parameter for each distinctive fluorochrome used in the encoding scheme. It should be apparent that plug flow sample handling has a potential in flow cytometry in a variety of applica-
159
tions. Automation of sampling will be compatible with many screening procedures. There are a number of ways in which this prototypic system could be improved to promote increased multiple-sample screening efficiency. For example, a multiplex analysis approach has been established in which multiple assays can be performed simultaneously in a single sample by performing the different assays on the surfaces of microspheres with distinctive fluorescence properties (5,6). By the use of multiplex analysis in combination with the present automated sample handling system, increases of one or more orders of magnitude in sample throughput rates seem feasible. Reducing sample plug volumes and using multiwell plate compatible microfluidics holds the promise for even further amplification of sample throughput. LITERATURE CITED 1. Nolan JP, Sklar LA. The emergence of flow cytometry for sensitive, real-time measurements of molecular interactions. Nature Biotech 1998;16:633–638. 2. Nolan JP, Posner RG, Martin JC, Habersett RC, Sklar LA. A rapid kinetic flow cytometer with subsecond resolution. Cytometry 1995;21:223– 229. 3. Sklar LA, Seamer LC, Kuckuck F, Posner RG, Prossnitz E, Edwards B, Nolan JP. Sample handling for molecular assembly in flow cytometry. Proc SPIE 1998;3256:144–153. 4. Seamer LC, Kuckuck F, Sklar LA. Sheath fluid control to permit stable flow in rapid mix flow cytometry. Cytometry 1999;35:75–79. 5. Fulton RJ, McDade RL, Smith PL, Kienker LJ, Kettman JR. Advanced multiplex analysis with the FlowMetrixTM system. Clin Chem 1997;43: 1749–1756. 6. Chandler VS, Denton D, Pempsell P. Biomolecular multiplexing of up to 512 assays on a new solid state 4 color flow analyzer. Cytometry 1998;(Suppl 9):40.
