Custom purified mouse anti-human CD4 antibody, clone SK3, was bulk ordered from Becton Dickinson (Franklin Lakes, NJ). The purified antibody was custom conjugated with Alkaline Phosphatase (AP) by MOSS Inc. (Pasadena, MD) and supplied at a concentration of 1 mg/ml. Leukoreduction filter LRF10S was purchased from Haemonetics Corporation (Braintree, MA). DynabeadsCD3, DynabeadsCD14, and a magnetic separator rack were purchased from Invitrogen/Thermo Fisher Scientific (Waltham, MA).AP substrate (BCIP/NBT) solution and stabilizing diluent for ALP conjugates from MOSS Inc.; Levamisole solution from Vector Laboratories (Burlingame, CA); Muse^® Human CD4 T Cell Kit from EMD Millipore Corporation (Billerica, MA); PE mouse anti-human CD14 from BD Biosciences (San Jose, CA) and absorbent pads were purchased from Bio-Rad Laboratories (Hercules, CA). All other reagents were purchased from Millipore Sigma (St. Louis, MO).

Donor blood samples drawn in K₃EDTA tubes were purchased from Biological Specialty Corporation (Colmar, PA) and used within 24 - 48 hours from the collection time. The blood samples were processed immediately upon receipt overnight. The total lymphocytes, CD4 T-lymphocytes, and monocytes were counted using a Guava EasyCyte flow cytometer (EMD Millipore). Samples with lower CD4 counts than normal were prepared either by (a) dilution of the whole blood with WBC-depleted autologous blood (Section 2.2.1.) or by (b) CD3 depletion of the whole blood (Section 2.2.2.). Overall~100 blood samples, including whole original blood and processed blood, were used in this study covering a wide range (17 - 1540 cells/μL) of CD4 T-cell concentrations.

Written consent for collecting blood samples was obtained by Biological Specialty Corporation from which such samples were acquired. Clinical samples were coded and unlinked to maintain patient privacy.

As described earlier [36, 37], the FTCA method is based on measurement of the concentration of CD4 T-cells retained on a leukoreduction filter membrane after unbound RBCs have been washed away. In short, the FTCA method comprises of four steps: (1) mixing of the whole blood sample with CD4 antibody-AP conjugate, (2) controlled application of the reaction mixture on the membrane, (3) membrane washing to remove unbound reagents and cells, and (4) addition of AP enzyme-substrate and color development step on the membrane. A more detailed protocol is described below.

All reagents were equilibrated to room temperature and used according to manufacturers’ recommendations. The AP-conjugated anti-human CD4 antibody was diluted 250-fold (i.e.4μg/mL final concentration) with the stabilizing diluent for ALP conjugates. Nine μL of the diluted antibody-AP conjugate was mixed with 42 μL of blood sample and9 µL of Dynabeads CD14 magnetic beads in a 0.65 ml Eppendorf tube. The reaction mixture was incubated on a rotator for 12 min at room temperature and was then placed on a magnetic separator rack. After 3 min, with the tube on the magnetic rack, 50 μL of the supernatant reaction mixture was removed and applied onto the retainer membrane (pre-wet with 100 μL of 10 mM Tris, pH 7.4) through a 3.2 mm orifice of the insert in the FTCA cartridge (described in Section 2.4). Once the mixture was completely absorbed in the membrane (~15-20 s), the insert and the enclosure were removed and the membrane was washed twice with 1 mL of wash buffer (10 mM Tris, pH 9.5) added through the 10 mm orifice on the top of the FTCA cartridge. All reagents delivered on the membrane were allowed to soak completely before applying the next reagent. After washing, 100 μL of the AP substrate (BCIP/NBT solution) was applied over the sample spot on the membrane, and the FTCA cartridge was placed in the dark for color development. After 10 min, 200 µL of 0.1 N H₂SO₄ solution was added to stop the color development reaction, followed by 1 mL water to remove excess acid from the membrane. Test results were recorded and scored using a custom-built imaging system (described in Section 2.5).

Previously, we described a laboratory setup of the basic FTCA cassette comprising of a sample delivery tube held manually or by mechanical means over a WBC retainer filter membrane on an absorbent pad [36]. The use of a sample delivery tube/cylinder with a defined orifice size (4 mm diameter) and gently pressed over the membrane allowed a controlled sample application and an effective flow of the sample through a confined area of the membrane. These parameters were found to be crucial in our system for highly efficient retention of WBCs on the LRF10S membrane to obtain a more uniform and stronger signal compared to open/direct sample application on the membrane. However, the delivery of subsequent solutions (wash buffer, development solution, etc.) through the same tube resulted in higher background signal compared to when these solutions were delivered directly on the membrane, thus requiring the removal of the delivery tube after sample application. Based on these findings, we developed a disposable cartridge for performing FTCA. The current prototype of the FTCA cartridge shown in Fig. (1) is a multi-unit assembly comprising of four parts, (i) a removable insert or sample application port/device; (ii) a cartridge top; (iii) a cartridge bottom; and (iv) an enclosure, as shown in Fig. (1A). All parts were designed using SolidWorks software and 3D printed (stereolithography) using Accura Xtreme White high-resolution material build in 0.002” layers (Proto Labs, Inc., Maple Plain, MN). The characteristics of cartridge parts are described below:

The insert or sample application port/device (labeled 1) is a rounded funnel-shaped unit for sample delivery through a 3.2 mm diameter orifice (1 or 2 orifices) at the bottom of a flat stem. It is positioned on the cartridge top (labeled 2) in two shallow groves at opposite ends of the top. The insert is removed after the sample application step in the FTCA protocol. The cartridge top has a 10 mm diameter orifice with peripheral indentation underneath to hold a 12 mm diameter membrane disk (0.18 mm thick) in place. The membrane (labeled 3) is sandwiched between the cartridge top and an absorbent pad (labeled 4) contained in the cartridge bottom (labeled 5). In the current design (Fig. 1A), the upper and lower parts of the cartridge are held with mini-screws. The insert can be held manually over the cartridge top or secured firmly by enclosing the entire cartridge assembly in an enclosure (labeled 6) with an opening to access the insert for the sample application. The overall dimensions of a fully assembled cartridge with the enclosure are 45 mm x 40 mm x 40 mm. After sample application, the enclosure and insert are removed, and subsequent solutions are delivered through the 10 mm orifice in the cartridge top. The use of a removable sample application device (i.e., insert) over a fixed cartridge top allows sample delivery through a 3.2 mm orifice in the insert and limits sample application to a defined area (sample spot) on the membrane as mentioned earlier. After removing the insert, washing in “open” through the 10 mm opening of the cartridge top results in minimal background signal. It should be emphasized that the use of mini-screws and enclosure was necessary because all parts were 3D printed and were not suitable for assembly using a snapping mechanism. In the final design of the cartridge, however, a snapping mechanism will be used for holding all parts together, eliminating the need for mini-screws and enclosure (Fig. 1B). This assembly mechanism will be feasible because injection molding will be employed for the manufacturing of all cartridge parts. In our study, we used two types of inserts, (i) 1-hole insert, and (ii) 2-hole insert, which both works via the same principle and provided practically the same results as described in Section 3.1.

A prototype of a low-cost and portable imaging device was developed for signal measurement from the wet membrane within the cartridge immediately after the experiment and without requiring any manipulation or handling of the membrane. The general views of the imaging system are shown in Fig. (2A-C). At the end of the FTCA test procedure, the cartridge is simply placed into the optical imaging device from underneath (Fig. 2C) for a numerical readout (image score) of the signal via analysis of the spot intensity on the captured image (Figs. 2D and 2E). The can-shaped imaging system is enclosed in a 1/16” thick aluminum tube with dimensions of 3” (diameter) x 6” (height). The aluminum shell also serves as a heat sink and encloses a computer USB powered 2.0megapixel camera (labeled 2), LED ring (labeled 3), and internal light and dark signal intensity references on a plate (labeled 4). Inside the shell (Fig. 2A) is a manifold of upper and lower/base plate rings supported by long bolts. The power unit/voltage regulator (4 - 12 V) is enclosed within the shell and mounted on the upper plate (labeled 1). The camera with a 2.8 - 12 mm varifocal lens is mounted under the upper plate and a 12 LED ring is mounted at a lower level closer to the bottom plate (labeled 5). The internal light and dark references are mounted on a plate (labeled 4) between the LED ring and the bottom plate. The bottom plate has a slot underneath for the cartridge (Fig. 2C). The internal references and the test sample spot on the membrane lie at the same focal distance from the camera lens.

During signal measurement, the imaging system is placed over the cartridge (Fig. 2C) and an image is acquired in the Portable Network Graphics (PNG) file format (Figs. 2D and 2E). A custom-built scoring algorithm computes test scores from image analysis involving a two-stage calibration: (1) correction of measured sample intensity to compensate for changes in lighting and camera parameters (brightness, contrast, etc.), and (2) conversion of corrected sample intensity into an image score. The first correction is based on using the measured intensity of internal light and dark reference marks and is performed simultaneously with the measurement of the sample intensity. The second correction is performed only once for multiple experiments (i.e., for dozens or hundreds of samples) and can be repeated if necessary. In this case, the value of the dark internal reference is assigned to the darkest signal by measuring an external reference spot on the cartridge (from a sample with the highest CD4 concentration).

Fig. (1). Prototype of the FTCA cartridge: Parts labeled are, 1 – insert (2-hole); 2 - cartridge top; 3 – membrane; 4 - absorbent pad; 5 - cartridge bottom; and 6 - enclosure. (A) General schematic of a disposable cartridge with a snap-fit mechanism (marked S); (B) Actual photograph of the 3D printed cartridge parts (1, 2, 5 and 6) and assembly using screws and enclosure.

Fig. (2). Prototype of the optical imaging device: Panel (A): Superimposed images representing a cross-sectional view of the device showing the top plate, 1; camera, 2: LED ring, 3; plate, 4 for internal references; and bottom plate, 5. The top plate, 1 has a power on/off switch and USB connection. Panel (B): View from underside showing one of the internal references (arrow) on a clear plate, 4. Panel (C): Cartridge, 6 positioning in the slot of bottom plate, 5 during measurement. Panels (D,E): Representative images from one sample (using 1-hole insert) and two-sample (using 2-hole insert) application procedures.

The intensity correction function that converts measured intensity to normalized (initial, standard) intensity is assumed to be linear. The conversion of measured sample intensity into an image score is computed assuming a linear scoring scale over the entire range of the normalized intensities of the light and dark references. The process of signal measurement from one cartridge to capture an image and generate readout as an image score takes less than a minute. The imaging system is a robust yet versatile system and can be used for extended periods of time without overheating or instability issues. The imaging system was used for obtaining image scores during FTCA assay optimization in the cartridge and for additional testing of approximately 100 blood samples in a pilot study, the results of which are described below.

In an earlier paper [36], we presented the CD4 calibration curve by the FTCA method using a biotin anti-CD4 antibody-streptavidin AP conjugate system. We reported that the LRF10S membrane was efficient in providing a signal specific for CD4 lymphocytes only and that there was a minimal contribution of monocytes. In the present study, using a crosslinked anti-CD4 antibody-AP conjugate, a noticeable difference in signal was observed for samples containing the same concentration of CD4 lymphocytes but different concentrations of monocytes. To check if this was due to the use of a different conjugate than used previously, the samples were also tested with biotin anti-CD4 antibody-streptavidin AP conjugate as described previously [36]. We also performed experiments using a modified FTCA protocol with samples after the monocyte depletion step (Section 2.3), and representative results with both antibody-AP conjugate systems are shown in Table 1.

For both anti-CD4 AP conjugate systems, the image scores decreased with monocyte depleted samples, and therefore, the signal was specific to CD4 T-cell concentration. Using the modified FTCA protocol, the FTCA cartridge system was initially tested using a 1-hole insert, where one sample was applied at a time, and an image score was measured at the end. However, up to four cartridges were used in parallel for testing multiple samples simultaneously. Results from 47 samples using a 1-hole insert cartridge system are shown in Fig. (3A).

Similarly, 50 samples were tested using the 2-hole insert cartridge system. The calibration curve for the 2-hole system gave similar results (y = 0.0006x + 0.24, R² = 0.89, data not shown) to the 1-hole system (y = 0.0006x + 0.23, R² = 0.90). Since the 1-hole and 2-hole systems share the same operation principle and are identical in terms of their orifice diameter (3.2 mm), we combined the 1-hole insert and 2-hole insert cartridge data (a total of 97 samples) to get the calibration curve as shown in Fig. (3B). The combined calibration curve represents an overall distribution of samples covering a wide range (17 - 1540 cells/μL) of CD4 T-cell concentrations. All further results and discussion, including statistical analysis and correlation studies with a reference method, are based on the calibration curve from combined data.

Fig. (3). CD4 calibration curves prepared by the FTCA method and cell count by flow cytometry. (A) using a 1-hole insert cartridge with 47 samples and (B) combined data from 1-hole and 2-hole insert cartridges with a total of 97 samples.

In order to estimate the clinical sensitivity and specificity of the FTCA method, a cut-off value, i.e., an FTCA signal at and below or above which samples will be considered positive or negative, respectively, needs to be established. The FTCA method sensitivity and specificity for the CD4 count threshold ([CD4]) at100, 200, 350 and 500 cells/μLwere evaluated using the ROC method described below.

3.2.1. ROC Analysis of Several Signal Cut-off Values

In this method, several signal cut-off(C_O) values were selected for a specific threshold CD4 count ([CD4] = 100, 200, 350 or 500 cells/μL), and anoptimal C_O value for clinical analysis was determined by receiver operating characteristic (ROC) analysis. The clinical specificity and sensitivity were determined based on the optimal C_O value. For all samples, the FTCA signal (i.e., image score) was plotted vs. CD4 count measured by flow cytometry. Then, the FTCA calibration curve as shown in Fig. (3B) was prepared by using linear regression analysis,

(1)

where Y is the image score, and x is the sample CD4 count from flow cytometry. The C_O value for the selected [CD4] was calculated from the calibration curve as C_{O [CD4]}. Then, several C_O values were selected within ± 20% of C_{O [CD4]}, and the most optimal C_O value was determined by plotting the clinical sensitivity and specificity for various C_O values by means of ROC analysis to achieve the best relationship between the clinical sensitivity and specificity.

The clinical sensitivity and specificity of the FTCA method were further calculated as,

(2)

(3)

where, True positive (TP), True negative (TN), False positive (FP) and False negative (FN) have usual meanings and are defined here as follows:

TP = sample CD4 count ≤[CD4]and image score≤C_{O [CD4]}.

TN = sample CD4 count>[CD4] and image score>C_{O [CD4]}.

FP = sample CD4 count>[CD4] and image score≤C_{O [CD4]}.

FN = sample CD4 count≤[CD4] and image score>C_{O [CD4]}.

A typical ROC analysis for [CD4] = 500 cells/μL is shown in Table 2. Similarly, ROC analysis was performed for [CD4] = 100, 200 and 350 cells/μL and the results are summarized in Table 3.

As shown in Table 2 and 3, the FTCA system, with a small number of samples, provides acceptable levels of sensitivity for all CD4 count thresholds, acceptable levels of specificity at [CD4] of 100 and 500 cells/μL, and moderate levels of specificity at [CD4] of 200 and 350 cells/μL. With further development of the technology and the use of a much larger clinical sample size, the clinical specificity of the FTCA method may be improved.

3.2.2. Correlation Studies

CD4 count determination by the FTCA method is based on measurement of the color intensity on the membrane rather than direct cell counting as in microscopy or flow cytometry. The color intensity is directly proportional to the concentration of CD4 T-cells in the sample. Flow cytometry, the current gold standard for CD4 count measurement, was initially used in order to obtain a preliminary understanding of the correlation of the FTCA method. The CD4 cell concentration calculated from the FTCA calibration curve (Fig. 3B) was plotted against the CD4 count measured by flow cytometry, as shown in Fig. (4). The correlation curve was linear over the entire range of CD4 concentrations studied, with a slope = 1.0, intercept = 0.006 and R² = 0.9.

We also used Bland-Altman analysis to more accurately quantify the agreement between FTCA and flow cytometry, as shown in Fig. (5). The Bland-Altman comparison analysis predicts a bias of -7 cells/μL (95% confidence interval, -250 and 236 cells/μL), and all 97sample data are within 95% limits of agreement. These results indicate that FTCA has great potential to be a fully quantitative method.

Fig. (4). Correlation analysis of CD4cell counts measured by FTCA method and by Flow cytometry. The CD4 counts by the FTCA method were calculated from the calibration curve shown in Fig. (3B).

Table 2. Clinical sensitivity and specificity for a threshold CD4 count of 500 cells/μLfrom ROC analysis at various signal cut-off (C_O) values. The most optimal parameters are indicated in bold.

CO	TP	TN	FP	FN	Sensitivity %	Specificity %
0.52	63	28	0	6	91	100
0.55	65	28	0	4	94	100
0.58	65	28	0	4	94	100
0.61	68	26	2	1	99	93
0.64	69	24	4	0	100	86

Table 3. Summary of clinical sensitivities and specificities for threshold CD4 count ([CD4]) of 100, 200, 350 and 500 cells/μL at respective optimal signal cut-off (C_O) values.

[CD4]	Optimal CO	Sensitivity %	Specificity %
100	0.29	95	95
200	0.40	94	81
350	0.50	95	83
500	0.61	99	93

Fig. (5). Bland-Altman analysis of an agreement between the FTCA and flow cytometry methods for CD4 count measurement from 97 samples. The CD4 counts using FTCA method were calculated using raw image.