Label-Free Method for Cell Counting in Crude Biological Samples viaParamagnetic Bead AggregationJingyi Li,†,∥,■ Qian Liu,†,∥ Li Xiao,† Doris M. Haverstick,‡ Alison Dewald,†,▲ Linda Columbus,†

Kimberly Kelly,§ and James P. Landers*,†,‡,⊥,∥

†Department of Chemistry, ‡Department of Pathology, §Department of Biomedical Engineering, ⊥Department of Mechanical andAerospace Engineering, ∥Center for Microsystems for the Life Sciences, University of Virginia, Charlottesville, VA 22904

*S Supporting Information

ABSTRACT: Under chaotropic conditions, DNA released from lysed cellscauses the aggregation of paramagnetic beads in a rotating magnetic field in amanner that is independent of the presence of other cellular components. Theextent of aggregation correlates with the mass of DNA in a quantitative manner(Leslie, D. C. et al., J. Am. Chem. Soc. 2012, 134, 5689−96), and from this, thenumber of DNA-containing cells in the sample can be enumerated. Microbialgrowth testing is demonstrated by monitoring bead aggregation with E. coli inthe presence of ampicillin. Without the need for fluorescent labeling or Coultercounting, the white blood cell count can be defined directly from a microliter ofcrude whole blood. Specificity is brought to the process by coupling bead-basedimmunocapture with DNA−bead aggregation allowing for the enumeration ofCD4+ T cells from human blood samples. The results of DNA-induced beadaggregation had a 95% correlation with those generated by flow cytometry. With the process requiring only inexpensive, widelyavailable benchtop laboratory hardware, a digital camera, and a simple algorithm, this provided a highly accessible alternative tomore expensive cell-counting techniques.

For decades, cell counting has been the modern workhorseof biomedical science and molecular biological research.

Flow cytometry is regarded as the gold standard for countingcells, primarily due to its speed, throughput and accuracy.However, its cost, large footprint, maintenance, cost persample, and complexity of operation limit its widespread use.Microscopy is a low-cost alternative for counting cells, but it is alaborious procedure, subjective in nature, and often does notmeet the throughput needs of laboratories. Coulter countingprovides an intermediate technology that bridges the complexbut highly automated flow cytometry and the low-tech, manuallight microscopy. Efforts to miniaturize cell counting have beenexplored by a number of groups. Chung et al. reviewedminiaturized flow cytometry, including cell transport anddetection in a microfluidic system, either optical orimpedometric.1 Rodriguez et al. reported the use of amembrane in miniaturized flow cell to capture cells of interest,which can be enumerated with microscopic imaging.2 Cheng etal. developed a similar strategy, where cells were isolated withimmunoseparation in a microfluidic chamber before enumer-ation under a microscope,3,4 and later reported the use of celllysate impedance spectroscopy as the detection module.5 Moonet al. integrated microfluidic chips with a lensless imagingsystem to shorten the time required for the whole process.6

Wang et al. labeled the captured cells with chemiluminescencein a microfluidic chip and correlated the intensity with cellnumber.7 Despite the promising progress, high cost andcomplexity remain as issues. This is due not only to the use

of fluorescent labels that are inherently unstable but also to theneed for complex detection systems.In clinical settings, Coulter counting suffices for generic

white blood cell (WBC) and red blood cell (RBC) counting,and it is reasonably fast and cost-effective. For more specific cellquantification, flow cytometry is the dominating technology.For example, the level of CD4+ T cells is often quantified inwhole blood from patients infected with the humanimmunodeficiency virus (HIV) or suspected of HIV infectionbased on symptomology. In addition to providing a snapshot ofthe status of the immune system, specific quantification ofCD4+ T cells is a critical metric for evaluating the response toantiretroviral therapy (ART). In fact, a 2007 UNAIDS studyestimated that only ∼30% of eligible HIV patients had access toantiretroviral therapy (ART), with one of the main reasonsbeing the lack of accessible cell counting facilities in resource-limited areas. Interestingly, Redd and Quinn suggest thatsimple, accurate, and cost-effective technologies for quantifyingwhole blood CD4+ cell levels is a critical component inexpanding the access to ART.8 This single example brings tolight the deficit that exists in cell-counting chemistries andinstrumentation that are portable, simple, and cost-effective foruse at the point-of-care.

Received: May 9, 2013Accepted: November 4, 2013Published: November 4, 2013

Article

pubs.acs.org/ac

© 2013 American Chemical Society 11233 dx.doi.org/10.1021/ac401402h | Anal. Chem. 2013, 85, 11233−11239

Paramagnetic beads have been extensively studied for theirunique magnetic properties9 and application to a wide array ofbioanalytical processes; these include immunoseparation,10

transportation,11 mixing,12 and the detection13,14 of selectanalytes of interest. Recently, we reported the “pinwheeleffect”,8 which describes chaotropic-driven adsorption of DNAonto paramagnetic silica beads in a quantitative manner.Captured images of the induced bead−DNA aggregatesfollowed by image analysis (using a simple algorithm) allowsfor direct quantification of DNA in crude biological samples.The simplicity of the chemistry (guanidine + beads + sample),the instrumentation (magnetic field + digital camera), and themethod (pipet + magnetize + capture image) make this a cost-effective, low-tech method for quantifying DNA in microlitersamples with a picogram limit of detection. We showedpreliminary evidence for the potential to count WBCs in wholeblood, and this spawned a more in-depth exploration of thecapabilities of the pinwheel effect for cell counting.In this report, we show that the same assay can be used to

monitor the growth of cultured cells using only a fewmicroliters of sample. We also demonstrate that leukocytescan be directly enumerated in whole blood without labeling orpurification with an accuracy and precision that comparesfavorably with Coulter counting. Finally, we make quantifica-tion cell-specific by coupling pinwheel with immunomagneticcell capture, demonstrating that CD4+ T cells can beenumerated with sufficient sensitivity and accuracy to monitorHIV/AIDS patients. As such, the pinwheel assay represents anew, label-free method for cell counting where the simpleprotocol and cost-effective instrumentation provides a highlyaccessible alternative to more expensive counting techniques.

■ METHODS AND MATERIALSReagents. MagneSil paramagnetic particles were purchased

from Promega (Madison, WI). GdnHCl was bought from MPBiomedicals (Solon, OH). MES was bought from AcrosOrganics. Tris base was bought from Fisher Scientific. EDTAwas bought from Sigma-Aldrich. Human SH-SY5Y cells andmouse RAW 264.7 cells were obtained from ATCC. E. coliBL21(DE3) and D29 bacteriophage were obtained from NewEngland Biolabs (Ipswich, MA). Whole blood samples weredonated by consenting donors. CD4 isolation kit was boughtfrom Life Technologies. All solutions were prepared inNanopure water (Barnstead/Thermolyne, Dubuque, IA).Reagent Preparation. Thirty microliters of stock Magnesil

beads was washed once with deionized, distilled water(Nanopure) followed by one wash with GdnHCl solution (8M, 1× TE, adjusted to pH 6.1 with 100 mM MES) andresuspended in 1 mL of GdnHCl solution to make thesuspension.Microwell Fabrication, Assay Instrumentation, and

Magnetic Field Application. Details have been describedpreviously. The system, with a rotating magnetic field (RMF),of the pinwheel assay for DNA quantification was employed forcell counting without any modifications.15

Image Processing. A gray level threshold was set in theimages of magnetic beads and aggregates by an isodataalgorithm written in Mathematica software, which identifiesthe pixels representing the beads and aggregates. The totalnumber of these pixels in the photograph without DNA isdefined as 100%, and as more DNA molecules are releasedfrom cells, tighter aggregation evolves, corresponding to asmaller percentage. The change of the percentage is defined as

degree of aggregation (DA). More DNA molecules result intighter bead aggregation and higher DA values.

Establishment of Calibration Curves. Samples were firstincubated with GdnHCl solution in 1:100 volume ratio at roomtemperature for 30 min. The samples were then serially dilutedin GdnHCl solution to appropriate concentrations as shown inthe figures. Five microliters of diluted sample, 3 μL of beadsuspension, and 12 μL of GdnHCl were mixed and exposed tothe RMF for 5 min, after which three photographs of the beadsand aggregates were acquired for image analysis. Thecalibration curve was generated by correlating the mean DAvalue of three images for each diluted sample with thecorresponding final cell concentration in the pinwheel assay.

Microbial Growth Testing. E. coli strain BL21(DE3) wascultured in 1000 mL of liquid LB medium and incubated at 37°C. Samples (1 mL) were collected at hour intervals for 5 h.The culture was split into two equal volumes at t = 3 h, and onevolume was treated with 0.1 mg/mL ampicillin. Cell countingbased on absorbance measurements at 600 nm was obtainedwith a UV/Visible spectrophotomer (Beckman Coulter,DU800) in a 1 mL plastic cuvette at room temperature.

White Blood Cell Counting for Unknown Samples.Raw blood samples were first incubated with GdnHCl solutionin 1:100 volume ratio at room temperature for 30 min. Thesamples were diluted by 2000 fold. Five μL of diluted sample, 3μL of bead suspension, and 12 μL of GdnHCl were mixed andexposed to the RMF for 5 min, after which photographs of thebeads and aggregates were acquired for image analysis.Considering mixing sample and beads results in another 4-fold dilution, the samples were diluted by 8000 fold from rawblood to the pinwheel assay. The concentration of white bloodcells in blood was acquired by comparing DA value with thecalibration curve (Figure 4). For the samples with WBC countlower than 4 × 103 per μL, a final 4000-fold dilution wasperformed in the pinwheel assay, and more accurate resultswere acquired because of an optimized dynamic range.

Immunomagnetic Separation of CD4+ T Cells fromBlood. CD4+ T cells were isolated by a Dynal T4 Quant kitfollowing its protocols. Briefly, whole blood was incubated androtated with CD14-coated beads for 10 min, followed byanother 10 min incubation and rotation with CD4-coatedbeads. Samples were then washed three times with PBS buffer.

Enumeration of CD4+ T Cells with Hemocytometer.Isolated CD4+ T cells were lysed by the lysis solution providedin the Dynal T4 Quant kit. The lysed cells were stained withSternheimer−Malbin solution and loaded on a hemocytometer,and the nuclei were counted using a conventional white lightmicroscope.

Enumeration of CD4+ T Cells with the PinwheelAssay. After immunomagnetic separation, the isolated CD4+T cells were incubated in 100 μL of 8 M GdnHCl solutions for30 min, and the solution was further diluted 20, 40, 80, and160-fold with 8 M GdnHCl. These four aliquots were thenloaded into the pinwheel microwell following the sameprocedure as white blood cell counting.

■ RESULTS AND DISCUSSIONParamagnetic Bead Aggregation Enables Cell Enu-

meration. With the pinwheel effect, DNA adsorbs to silica-coated paramagnetic beads under chaotropic conditions, anddriven by a rotating magnetic field (RMF), the beads areintertwined by DNA and rapidly aggregate (in seconds tominutes) (Figure 1A). The degree of aggregation correlates

Analytical Chemistry Article

dx.doi.org/10.1021/ac401402h | Anal. Chem. 2013, 85, 11233−1123911234

well with the concentration of DNA, allowing for fast DNAquantification without the need of fluorescent labels.15 Owingto the specificity of the chaotrope-driven DNA adsorption onsilica,16 bead aggregation induced by DNA released from thecell lysate is unaffected by protein, lipid, and other cellularcomponents present at concentrations orders of magnitudehigher; this is the case for even a few cells (Figure 1B). Giventhat each DNA-containing cell in a blood sample releases thesame mass of DNA (6.25 pg), the extent of bead aggregationdefines the number of cells in the sample. Figure 1 panels B2−B4 show images of beads that have been aggregated by 13, 52,and 195 pg of DNA, respectively. While the protein releasedfrom 30 cells is ∼15 ng, it is noteworthy that augmenting thesolution with additional protein (BSA; to a final concentrationof 1 μg/μL) does not adversely affect aggregation.17

An inherent advantage of the chemistry that induces thepinwheel effect (i.e., guanidine) is that it also lyses cellseffectively; thus, no additional reagents are required for directcell counting. A cell solution added to 8 M GdnHCl at a 1:100volume ratio induces bead aggregation immediately followingvortexing, which is key to the pinwheel assay enabling theenumeration of cells. The rapidity with which cell lysis occurs issupported by the fact that aggregate formation (visuallydetectable aggregates in seconds) does not improve withlonger incubation times (e.g., up to 60 min) (see SupportingInformation Figure S1). This yields a simple protocol for cellenumeration: (a) lyse cells in 8 M GdnHCl to release DNA,(b) bring the sample DNA concentration within the dynamicrange of the assay through sample dilution, (c) mix the samplewith silica-coated magnetic beads in a RMF to induceaggregation, (d) capture digital images and process the imagesthrough an algorithm to generate a % DA and, finally, (e)compare resultant value to a standard curve to calculate cellconcentration (details are provided in the experimentalsection). As DNA begins to entangle with the beads,15 smallbead aggregates evolve from the initial dispersed state, growing

through nucleation with other beads, allowing smalleraggregates to coalesce into larger ones. The higher the DNAconcentration, the more aggressive the aggregate formation andsubsequent coalescing become. Overall, the beads experience atransition from the dispersed state in the negative control (i.e.,without DNA) to a condensed aggregation state with sufficientDNA. Even with only a few cells, aggregation induced byreleased DNA leads to a decrease of the total number of pixelsrepresenting the beads and/or bead aggregates; this we refer toas dark area.15 A threshold is set by an isodata algorithm15 toseparate the beads and/or aggregates from image background,shown as the black line in the histogram (Figure 1C). The peakarea below the threshold stands for the dark area, and thedecrease of dark area represents the loss of dispersed beads dueto aggregation. Therefore, we define the decrease of dark areathat is normalized against the negative control as DA.Correlating DNA concentration and DA generates a calibrationcurve for DNA quantification. With each somatic cellcontaining 6.25 pg DNA (discussed earlier), a simplecalculation allows for plotting DA against the number of cellsin a particular volume (typically 1 μL). Regardless of the sourceof DNA, the higher the DNA concentration, the larger the DA.This provides a quantitative correlation between the extent ofaggregation and the number of cells in a specific volume ofsample, for example, as in Figure 2.Pinwheel assay standard curves for the enumeration of: (a)

eukaryotic cells (human WBCs and mouse cells), (b)prokaryotic cells (E. coli), and (c) bacteriophage (D29) aregiven in Figure 2. While not a “cell”, phage are virus-likemacromolecular assemblies of protein and nucleic acid, with anintact genome of length of 49 × 103 base pairs. The DNAextracted from this phage is double-stranded, and with a 49 kbplength, a rough estimate of the length is ∼16 μm.18 As shown inFigure 2, it is still capable of inducing aggregates, but it does sowith an effectiveness that is roughly an order of magnitude lessthan E. coli, whose genome is 4.6 Mb. There is a 3 order of

Figure 1. Schematic of the pinwheel effect induced by lysed cells. (A) GdnHCl lyses the cells and drives the released DNA to adsorb on to silica-coated paramagnetic beads, which form aggregates in a rotating magnetic field (RMF). Images of the aggregation are digitized, from which the initialconcentration of cells can be measured. (B) Photographs of the same amount of beads with different amounts of DNA that are equivalent to 0, 2, 8,and 32 cells, respectively. Magnetic beads, which remain dispersed in RMF, form visually detectable aggregates within minutes upon mixture withlysed human cells. (C) Gray level histogram of the four photographs. Three regions are identified as aggregates, dispersed beads, and background,respectively, on the basis of the gray level value. The change of the histogram clearly illustrates the loss of dispersed beads and the formation ofaggregates.

Analytical Chemistry Article

dx.doi.org/10.1021/ac401402h | Anal. Chem. 2013, 85, 11233−1123911235

magnitude differential in the size of the E. coli DNA versusmouse or human DNA. As the curve in Figure 2 (a semilogplot) shifts from left to right, the sensitivity decreases over 3orders of magnitude; not surprisingly, this roughly correlateswith the decrease in genome size with each cell type. Forhuman cells, ∼8 cells in 20 μL corresponds to 0.4 cells/μL anda DNA concentration of 2.5 pg/μL. Despite this minisculemass, the induced aggregation is still visually detectable (Figure1B), with the limit of detection (through image analysis)approaching 0.1 cell/μL, that is, 2 cells in a total of 20 μL(Figure 2). The concept that sensitivity is proportional to DNAlength is consistent with our previous observations15 and isrationalized as longer strands of DNA induce bead aggregationmore effectively than shorter ones. As a result, each cell typeshould be calibrated individually in order to acquire accurateresults. To further study the size dependence, the actual DNAsize distribution should be measured by other techniques suchas electrophoresis and correlated more quantitatively with thesensitivity of the pinwheel assay.Microbial Growth Testing with the Pinwheel Assay.

Accurate and quick determination of bacteria concentration isessential in microbiology studies. To test the effectiveness ofmeasuring microbial growth with the pinwheel assay, E. coli wascultured in media, and the time evolution of growth wasmonitored by two methods: light scattering (spectrophotom-etry at 600 nm) and the pinwheel assay. From the liquidculture, 1 mL was sampled every hour, and after 3 h, the cellswere split, where half were dosed with buffer and the other halfwith ampicillin (0.1 mg/mL). Monitoring the bacterial growthconventionally by absorbance at 600 nm (Figure 3A), bacterialcell doubling is apparent (from a 1 mL sample) by the increaseof cell number up to the 3 h point. However, addition ofampicillin retards E. coli growth, and not only levels the growthcurve (i.e., no further doubling), but the addition of ampicillinactually leads to a reduction in the bacterial concentration.19

Similar evaluation was carried out by pipetting 5 μL aliquotsfrom the growth medium into microwells containing GdnHCland paramagnetic beads. As given in Figure 3B, the resultsacquired from the pinwheel assay also illustrate a comparable

trend of bacterial growth. For the sake of simplicity, DA isdirectly compared to illustrate the change of cell concentration.Because DA does not increase linearly with cell count (Figure2), the difference of DA between “with ampicillin” and “withoutampicillin” does not appear to be as substantial as that of theabsorbance measurement. A closer match between theabsorbance measurement and the pinwheel assay can beobtained by converting the signals in both methods to cellcounts.

Enumeration of White Blood Cells in Whole Blood.WBCs (leukocytes) are critical components in our defenseagainst foreign bodies and infectious agents. The white bloodcell count can be a generic indicator of immune system statusand is useful for clinical testing, in particular with moleculardiagnostics. The pinwheel assay has the potential to provide asimple and rapid method for determining of WBC count.To generate the calibration curve for WBC counting, blood

samples were drawn from four normal patients on the day ofthe analysis, and the WBC counts were determined by Coultercounter. The samples were lysed with 8 M GdnHCl and seriallydiluted for pinwheel analysis. Figure 4 illustrates the correlationbetween degree of aggregation and WBC count after dilution.Curves derived from 4.1, 10.1, and 20.3 (× 103 WBCs/μLpredilution) samples overlap well with each other. The 2.3 ×103 WBCs/μL sample, however, reproducibly yields a higher

Figure 2. Calibration curve of the pinwheel assay for cell counting. DAis correlated with cell (or phage) concentration and fitted into anexponential model. The sensitivity decreases from human and mousecells (eukaryote) to phage D29 (although not a cell), which can beattributed to the decrease of DNA length and genome size. The x-axisis plotted in log scale in order to illustrate the difference of calibrationcurves between human and mouse cells, which have similar genomesizes. The solid line represents an exponential calibration functionfitted to the data at logarithmic scale; the shaded area and gray linesshow the 95% and 99% confidence bands (the same for all curvefitting, unless otherwise noted).

Figure 3. Microbial growth testing of E. coli in the presence ofampicillin. At the 3 h mark, half the cells were dosed with buffer, whilethe other half were dosed with ampicillin, as growth monitoringcontinued.

Figure 4. Calibration curve of the pinwheel assay for WBC countingdirectly in raw blood samples. Four blood samples, with their WBCcounts determined with Coulter counter, were analyzed with thepinwheel assay, as represented by ▽ = 2.3, Δ = 4.1, □ = 10.1, and • =20.3 (× 103 per μL) with associated error bar denoting the standarddeviation of three experiments on each sample. The symbol • and thecorresponding error bar denote the mean and standard deviation ofthe four samples.

Analytical Chemistry Article

dx.doi.org/10.1021/ac401402h | Anal. Chem. 2013, 85, 11233−1123911236

DA at the same [WBC]; this is because it falls outside thedynamic range of the assay and, thus, results in a larger standarddeviation for the mean values above 40% DA. This may becaused by components in the blood other than WBCs (e.g.,proteins such as albumin, or a combination of protein anddivalent cations like Ca2+), of which the concentration remainsrelatively high after dilution due to the low initial WBCconcentration and low dilution factor. Elucidation of themechanism is currently ongoing and will be the topic of aseparate report when completed. As a result, only the regionabove a DA of 60% is regarded as the effective calibration rangefor WBC quantification. Therefore, DA values in that range(>60%) must be obtained for accurate cell counting.To evaluate the ability of the pinwheel process to enumerate

the WBCs in unknown samples, dilution is needed so that theDA falls in the effective calibration range. By comparing themeasured value with the calibration curve, the WBC count canbe estimated following dilution ([WBC]dil), which is thenconverted to the initial WBC count ([WBC]init) by multiplyingby the dilution factor. Given that the calibration curve coversfrom 0.5 to 5 WBCs/μL, a 4000-fold dilution enablesquantification of blood samples with WBC count between2000 and 20 000 per μL. Since the WBC count of adultstypically varies between 4000 and 11 000 per μL, a 4000-fold or

8000-fold dilution would be sufficient to cover the normalrange and to identify samples with abnormal WBC count.To test accuracy of the pinwheel-derived WBC counting,

nine raw blood samples with unknown WBC counts wereacquired, lysed, and diluted appropriately. Three replicates wereperformed on each sample, and the mean DA was compared tothe standard curve to determine initial WBC count. Thesevalues were compared to those independently obtained by aCoulter counter, and the results are given in Figure 5A. Thecounts from both compare favorably for the samples with[WBC]init ranging from 2 × 103 per μL to 14 × 103 per μL; thiscovers the normal range for WBC count and suffices forsamples where there is a need to identify counts outside thenormal range (bracketed by the two dashed lines). Figure 5Bprovides graphical representation of the excellent correlationbetween the pinwheel results and those from Coulter counter(R2 = 0.97).

Counting CD4+ T Cells with the Pinwheel Assay.Whileeffective for WBC counting, the pinwheel effect is genericbecause the adsorption of DNA on silica is not specific togenomic sequence; hence, subtyping of cells in a complexmixture is not possible without invoking the use of a “selective”step. We chose CD4+ T cells, a WBC subtype, as a model todemonstrate that selectivity and pinwheel sensitivity could becoupled in a powerful way. CD4+ T-cell levels are critical in

Figure 5. WBC counts of nine raw blood samples determined by the pinwheel assay are shown in (A) and compared with the results from Coultercounter. (B) The WBC counts of the nine samples determined with the pinwheel assay correlate well with the results from Coulter counter. Thedata points represent the mean results of three experiments on each sample, and the error bars denote the standard deviation. The solid blue linerepresents a linear calibration function fitted to the data with R2 = 0.97, and the black dashed line shows y = x.

Figure 6. Immunomagnetic separation of CD4+ T cells from blood and the calibration curve for quantification. (A) Schematic of immunomagneticseparation of CD4+ T cells from crude blood samples. Monocytes, which also express the CD4 antigen, are depleted prior to CD4+ T cells. (B)Purified samples are serially diluted after lysis, and the calibration curve is established by correlating degree of aggregation with the CD4 countdetermined by microscopy.

Analytical Chemistry Article

dx.doi.org/10.1021/ac401402h | Anal. Chem. 2013, 85, 11233−1123911237

assessing the progress of AIDS and also in monitoring anti-HIVtreatment. CD4+ T-cell quantitation is currently accomplishedusing flow cytometry, where the obvious drawbacks are highcost, high maintenance, and poor portability. A rapid,substantially less expensive and less complex methodology forCD4+ T-cell counting could be valuable, especially if point-of-care applications are to be considered. One solution to thisinvolves a commercial immunomagnetic separation kit. This kitexploits antiCD4+-coated magnetic beads for isolating CD4+ Tcells from blood, but also captures monocytes (they expressCD4). However, monocytes also express CD14 on theirsurface,4 allowing for antiCD14-coated beads to deplete thesample of monoctyes, prior to the capture of CD4+ T cells(Figure 6A). In the protocol described by the manufacturer, thecaptured CD4+ T cells are partially lysed so that their nucleican be released and enumerated with a hemocytometer under amicroscope after staining. While labor-intensive and, in someways, subjective, immunocapture with microscope-basedcounting certainly provides a cost-effective alternative to flowcytometry. Studies of normal and HIV-positive individuals haveshown a correlation coefficient of 0.97 for CD4+ cell levelsrelative to values from hemacytometer.To circumvent the microscopic counting of stained cells, we

substituted this step with the pinwheel assay. With the pinwheelprocess, the isolated cells are completely lysed, and pinwheelaggregation is induced by the released DNA, allowing forquantification of cells. The correlation between DA and cellconcentration is illustrated in Figure 6B, which provides thecalibration curve to determine the CD4+ T-cell count in bloodsamples. Since the enumeration of CD4+ T cells involvespreassay isolation of the cells, the effectiveness of this processrelies on the efficient and consistent CD4+ T-cell isolation. Themean isolation yield that was observed (calculated by dividingthe microscopic cell count with the flow cytometry count of theoriginal sample) was found to be 91.5 ± 4.3% for four blood

samples when the blood samples were withdrawn within 24 h.To evaluate the accuracy of the pinwheel assay, two aliquots ofpurified cells were acquired from each unknown sample, withone partially lysed for microscopy quantification and the othercompletely lysed for the pinwheel assay. Comparison shows agood correlation between the two methods with 13 patientblood samples (see Supporting Information Figure S2), whichsuggests that the pinwheel assay can, indeed, enumerate theisolated cells accurately.In the clinical setting, CD4+ T-cell count can be indicative of

the immune status of a patient. When the CD4+ T-cell count islower, along with other factors, it can represent the progress ofHIV infection; when the count is higher, it can indicaterecovery from immunodeficiency through decreased viral load.The pinwheel CD4+ levels in 21 blood samples followingimmunomagnetic separation were compared with the resultsfrom flow cytometry (Figure 7A,C). The collective data is fromtwo sample groups from two sources (8 samples analyzed sameday as blood draw; 13 samples analyzed >24 h after blooddraw). A good correlation was observed between theimmunocapture−pinwheel assay and flow cytometry, whenthe samples were analyzed within 24 h of the blood draw(Figure 7A,B). Dramatically poorer correlation between thepinwheel and flow cytometry methods was observed withsamples analyzed more than 24 h postdraw (Figure 7C). Thisobservation is consistent with the observation by Diagbouga etal. that 31%−69% of blood samples exhibit more than a 20%decrease of CD4 count after a 24 h delay,20 which directlyaffects the accuracy of the pinwheel assay.For HIV-infected patients, initiation of ART is determined by

a CD4+ T-cell count that drops below 350 cells/μL. What wehave demonstrated here is that, with the 21 whole bloodsamples evaluated from HIV patients, the pinwheel assaycorrectly categorized those determined by flow cytometry to beabove or below the 350 cells/μL cut off. Only one sample (one

Figure 7. Quantification of CD4+ T cells. The pinwheel result correlates well with flow cytometry when the post-drawn time of blood is less than 24h, as shown in (A) and (B). For the samples with post-drawn time over 24 h, the accuracy of the pinwheel assay is compromised due to thedeterioration of sample quality, as shown in (C) and (D). The solid blue lines in (B) and (D) represent linear calibration functions, and the blackdashed lines show y = x. The gray dashed lines represent the threshold (350 cells per μL) that defines the starting time of ARV for HIV patients (for(B), R2 = 0.997; for (D), R2 = 0.775). The arrows in (C) and (D) point to the only sample that generated a false result in the pinwheel assay.

Analytical Chemistry Article

dx.doi.org/10.1021/ac401402h | Anal. Chem. 2013, 85, 11233−1123911238

from the >24 h storage time group) was categorized as falsepositive, and this point is labeled with a gray arrow in Figure7D. Despite the asynchrony in the storage history of thesesamples, this still represents a >95% correlation between theimmunocapture−pinwheel assay and flow cytometry. This isremarkable and points to the obvious merit of a method thatdoes not require fluorescent reagents, microscopy, orcytometric counting.

■ CONCLUSIONS

In conclusion, we have extended the application of thepinwheel assay from nucleic acid quantification to cell counting,and demonstrated (1) microbial growth testing, (2) cellcounting in a complex medium (WBC quantification in rawblood), as well as (3) isolation and enumeration of a specificsubtype of cells from a mixture (CD4+ T cells). While we havea simple methodology for generic cell counting, specificity wasbrought to the method by coupling with immunocapture.These examples show the versatility of the pinwheel assay (i.e.,that the assay has the potential to be applied in various fields),although we do not anticipate that the pinwheel assay willreplace all the conventional methods. As the aggregation ofmagnetic particles only requires microliter volume of samplesand a digital camera as the detection modality, the pinwheelassay could serve as a portable and cost-effective alternative toconventional technologies for point-of-care applications.One of the main advantages is that the pinwheel assay does

not require fluorescent labels and complex optics or expensiveinstrumentation to enumerate cells. Fluorescence providesaccurate and sensitive quantification of analytes in variousapplications, but the labels are often subject to photobleaching,resulting in unstable optical properties, which increases the costfor reagent storage and induces the need for repetitivecalibration. By avoiding fluorescent labels, the pinwheel assayoffers enhanced simplicity and cost-effectiveness compared toconventional techniques, which are always desired whenapplying a new technique in resource-limited regions. One ofthe major drawbacks of the pinwheel assay is that the cells arelysed after counting, and thus will not be available to other cellanalyses, which may be problematic if sample availability islimited.Day-to-day variability with the pinwheel method was low, a

likely result of the simplicity of the method. User-to-uservariability is more likely to be significant, but this can beovercome by building a larger database for calibration of theassay independent of the user. We are currently exploringdesigns for integrating all of these processes on a singlemicrofluidic device to minimize hands-on operation and user-to-user variabilitythis is essential for progressing toward a cellcounting device with sample-in-answer-out capability. Progresson device development has been made in terms of enhancingassay throughput21 and process automation.22 Successfulemployment of the pinwheel assay in an integrated microfluidicsystem will not only benefit basic biomedical research but alsooffer capability to a wide range of clinical applications.

■ ASSOCIATED CONTENT

*S Supporting InformationAdditional information as noted in the text. This material isavailable free of charge via the Internet at http://pubs.acs.org.

■ AUTHOR INFORMATIONCorresponding Author*E-mail: jpl5e@virginia.edu.Present Addresses■Microlab Horizon, LLC, 705-D Dale Ave., Charlottesville, VA22903 (J.L.)▲Salisbury University, 1101 Camden Ave., Salisbury, MD21801 (A.D.)NotesThe authors declare no competing financial interest.

■ REFERENCES(1) Chung, T. D.; Kim, H. C. Electrophoresis 2007, 28 (24), 4511−4520.(2) Rodriguez, W. R.; Christodoulides, N.; Floriano, P. N.; Graham,S.; Mohanty, S.; Dixon, M.; Hsiang, M.; Peter, T.; Zavahir, S.; Thior, I.;Romanovicz, D.; Bernard, B.; Goodey, A. P.; Walker, B. D.; McDevitt.,J. T. PLoS Med. 2005, 2 (7), 663−672.(3) Cheng, X.; Irimia, D.; Dixon, M.; Sekine, K.; Demirci, U.; Zamirb,L.; Tompkins, R. G.; Rodriguez, W.; Toner, M. Lab Chip 2007, 7 (2),170−178.(4) Cheng, X.; Gupta, A.; Chen, C.; Tompkins, R. G.; Rodriguez, W.;Toner, M. Lab Chip 2009, 9 (10), 1357−1364.(5) Cheng, X.; Liub, Y.; Irimia, D.; Demirci, U.; Yang, L.; Zamir, L.;Rodríguez, W. R.; Toner, M.; Bashir, R. Lab Chip 2007, 7 (6), 746−755.(6) Moon, S.; Keles, H. O.; Ozcan, A.; Khademhosseini, A.;Hæggstrom, E.; Kuritzkes, D.; Demirci, U. Biosens. Bioelectron. 2009,24 (11), 3208−3214.(7) Wang, Z.; Chin, S.; Chin, C. D.; Sarik, J.; Harper, M.; Justman, J.;Sia, S. K. Anal. Chem. 2010, 82 (1), 36−40.(8) Redd, A. D.; Quinn, T. C. Nat. Rev. Microbiol. 2008, 6, S1.(9) Gijs, M. A. Microfluid. Nanofluid. 2004, 1 (1), 22−40.(10) Chen, G. D.; Alberts, C. J.; Rodriguez, W.; Toner, M. Anal.Chem. 2010, 82 (2), 723−728.(11) Sista, R. S.; Eckhardt, A. E.; Srinivasan, V.; Pollack, M. G.;Palanki, S.; Pamula, V. K. Lab Chip 2008, 8 (12), 2188−2196.(12) Grumann, M.; Geipel, A.; Riegger, L.; Zengerle, R.; Ducree, J.Lab Chip 2005, 5 (5), 560−565.(13) Afshar, R.; Moser, Y.; Lehnert, T.; Gijs, M. A. Anal. Chem. 2011,83 (3), 1022−1029.(14) Moser, T.; Lehnert, T.; Gijs, M. A. Lab Chip 2009, 9 (22),3261−3267.(15) Leslie, D. C.; Li, J.; Strachan, B. C.; Begley, M. R.; Finkler, D.;Bazydlo, L. A. L.; Barker, N. S.; Haverstick, D. M.; Utz, M.; Landers, J.P. J. Am. Chem. Soc. 2012, 134 (12), 5689−5696.(16) Melzak, K. A.; Sherwood, C. S.; Turner, R. F. B.; Haynes, C. A.J. Colloid Interface Sci. 1996, 181 (2), 635−644.(17) Leslie, D. C.; Li, J.; Finkler, D. M.; Price, C. W.; Landers, J. P.Proc. 13th Int. Conf. Miniaturized Systems Chem. Life Sci., Jeju, Korea,November 1−5, 2009; vol. 2.(18) Voet, D.; Voet, J. G. Biochemistry, 2nd ed.; John Wiley and Sons:New York, 1995; p 1361.(19) Rogers, H. J.; Mandelstam, J. Biochem. J. 1962, 84, 299−303.(20) Diagbouga, S.; Chazallon, C.; Kazatchkine, M. D.; Van de Perre,P.; Inwoley, A.; M’Boup, S.; David, M. P.; Tenin, A. T.; Soudre, R.;Aboulker, J.; Weiss, L. AIDS 2003, 17 (15), 2201−2208.(21) Li, J.; Landers, J. P. Proc. 15th Int. Conf. Miniaturized SystemsChem. Life Sci.Seattle, WA, October 2−6, 2011; pp 1956−1958(22) Ouyang, Y.; Li, J.; Landers, J. P. Proc. 17th Int. Conf. MiniaturizedSystems Chem. Life Sci., Freiburg, Germany, October 27−31, 2013;

Analytical Chemistry Article

dx.doi.org/10.1021/ac401402h | Anal. Chem. 2013, 85, 11233−1123911239