A Simple Approach for Counting CD4 + T Cells Based on a Combination of Magnetic Activated Cell Sorting and Automated Cell Counting Methods

: Frequent tests for CD4 + T cell counting are important for the treatment of patients with immune deﬁciency; however, the routinely used ﬂuorescence-activated cell-sorting (FACS) gold standard is costly and the equipment is only available in central hospitals. In this study, we developed an alternative simple approach (shortly named as the MACS-Countess system) for CD4 + T cell counting by coupling magnetic activated cell sorting (MACS) to separate CD4 + T cells from blood, followed by counting the separated cells using Countess TM , an automated cell-counting system. Using the cell counting protocol, 25 µ L anti-CD4 conjugated magnetic nanoparticles (NP-CD4, BD Bioscience) were optimized for separating CD4 + T cells from 50 µ L of blood in PBS using a Dynamag TM -2 magnet, followed by the introduction of 10 µ L separated cells into a Countess TM chamber slide for automated counting of CD4 + T cells. To evaluate the reliability of the developed method, 48 blood samples with CD4 + T cell concentrations ranging from 105 to 980 cells/ µ L were analyzed using both MACS-Countess and FACS. Compared with FACS, MACS-Countess had a mean bias of 3.5% with a limit of agreement (LoA) ranging from − 36.4% to 43.3%, which is close to the reliability of the commercial product, PIMA analyzer (Alere), reported previously (mean bias 0.2%; LoA ranging from − 42% to 42%, FACS as reference). Further, the MACS-Countess system requires very simple instruments, including only a magnet and an automated cell counter, which are affordable for almost every lab located in a limited resource region.


Introduction
To support the treatment of HIV/AIDS and other deficient immune diseases, CD4 + T cell counts are frequently monitored using highly reliable equipment. HIV selectively infects and kills CD4 + T cells and causes the gradual death of this cell population [1]. Thus, CD4 + T cell concentrations reflect the immune status of patients; in HIV-infected individuals, CD4 + T cell counts are usually lower than 500 cells/µL of blood and patients who progress to AIDS have fewer than 200 cells/µL [2]. Most patients living with HIV/AIDS receive antiretroviral treatment, and initiation of antiretroviral therapy (ART) is recommended based on the CD4 + T cell count threshold [3]. The CD4 + T cell count also reduces in the case of severe acute respiratory syndrome coronavirus (SARS-CoV) infection, especially in severe and critically ill SARS-CoV-2 patients [4,5]. Proposed mechanisms excessive antibodies are removed after discarding the supernatant. Magnetic particles conjugated with streptavidin (MP-streptavidin) are then incubated with the sample to allow interaction of streptavidin and biotin. Finally, the MP-strep-bio-CD4 cell complexes are separated using a magnetic field. This format was used by Carinelli et al. for the semi-quantification of CD4 + T cells [21]. Magnetic-based cell separation systems have a simpler format compared with fluorescence-based methods, but the greatest obstacle preventing MACS to be applied in diagnostics is its fluctuating cell separation efficiency, followed by labor-extensive procedures. Despite being introduced as early as the nineteenth century [22], the application of MACS in diagnosis has still been limited. The main reason for this is that cells must be separated with stable efficiency so that the original cell counts can be calculated. In addition, the aforementioned procedure for cell separation remains complicated and/or requires a large sample input.
In an attempt to reduce the gap between MACS utilization and CD4 + T cell counting applications, we developed a CD4 + T cell counting system based on a simple MACSseparation for CD4 + T cells using a small blood input required, followed by automated cell counting. As CD4 + T cells are selectively isolated through the MACS process, the subsequent identification of cell types is not required, allowing the use of automated cell counting on Countess TM in the final step. Countess TM performs cell counting using image capture and analysis software. Cells within are defined using proper circular shape and diameter settings. As Countess TM only requires a digital camera for bright field recording, it is much less complicated than a fluorescent optic system that requires lasers, filters, and highly sensitive cameras.

Materials
Blood samples from healthy volunteers were collected by technicians at the National Hospital for Tropical Diseases under the protocol approved in ethics approval number IRB-VN01.001/IRB00003121/FWA 00004148 from the Ethics Committee at the Hanoi Medical University. The samples were collected between June and December 2020, stocked in K2 EDTA tubes, and processed within 24 h. Lymphocyte Separating Medium, Pancoll human (PAN Biotech, Aidenbach, Germany) was used for isolating PBMCs from total blood cells. Tritest reagent (BD Biosciences, Becton, San Jose, CA, USA) containing fluorescence-labeled antibodies including CD45-PerCP/CD4-FITC/CD3-PE, was used to label cells in the FACS method using the FACSCalibur system and Multiset software (BD Biosciences). FACS Lysis Solution 10X Concentrate (BD Biosciences) was diluted to 1X concentration by adding 9 mL of distilled water to 1 mL of 10X solution, and used to lyse red blood cells in the FACS method. Trucount tubes (BD Biosciences) with a predefined number of fluorescence beads were used as a reference to determine the absolute number of cells in FACS analysis. Anti CD4 Abs labeled with FITC and anti CD3 Abs labeled with PE (CD4-FITC, CD3-PE, Thermo Fisher Scientific, Waltham, MA, USA) were used for immunofluorescence staining. Anti-Human CD4 Particles-DM (NP-CD4) from BD Biosciences were used for cell capture and separation using a Dynamag™-2 magnet (Thermo Scientific), and were then counted using Countess TM , an automated cell-counting machine. A hemocytometer (Paul Marienfeld, Lauda-Königshofen, Germany) was used to estimate the cell concentration in the sample after PBMC isolation under a Primo Star (2) Microscope (Carl Zeiss, Oberkochen, Germany). Phosphate buffered saline (PBS, pH 7.4) supplemented with 0.5% BSA and 2 mM EDTA (PBS*) was used as a buffer for the incubating and washing steps.

Isolation of Peripheral Blood Mononuclear Cells (PBMC) from Whole Blood by Sucrose Gradient Centrifugation
PBMCs were isolated from blood by sucrose gradient centrifugation using the Pancoll Human Kit (PAN-Biotech). Briefly, approximately 4 mL of blood was collected in K2 EDTA tubes and diluted 2-fold with PBS. Diluted blood was then carefully layered on 8 mL Pancoll solution such that the two phases were separated. After centrifugation at 800× g on an Eppendorf 5810R for 20 min, brake = 0, and at room temperature, four fractions were separated from top to bottom: yellow serum, opaque white PBMCs, clear white Pancoll solution, and red blood cells. The PBMC phase was harvested into a new centrifuge tube using a pipette and was then filled with two-to three-fold volume of PBS, followed by centrifugation at 300× g for 10 min. After centrifugation, the cell pellet was precipitated at the bottom and the supernatant was discarded. The washing step was repeated once again for purified PBMCs.
2.3. Isolation of CD4 + T Cells from PBMCs and Whole Blood Using Anti-CD4 Antibodies Conjugated with Magnetic Nanoparticles As a conventional method, approximately 10 7 PBMCs previously isolated from whole blood were incubated in 1 mL PBS with 50 µL NP-CD4 for 30 min. The sample suspension was then placed on a Dynamag™-2 magnet for 10 min, following which the supernatant was discarded. The washing steps were performed twice by repeating washes with 1 mL PBS, placing the tube on the magnet for 7 min, and discarding the supernatant. The cell sample was finally resuspended in 1 mL PBS*. Separation efficiency (SE) was calculated as the ratio of the number of CD4 + T cells separated by the number of CD4 + T cells in the sample before separation.
MACS separation of CD4 + T cells from whole blood was performed using the same procedure as the conventional method, but with various ratios of blood input and NP-CD4. To find the optimal sample volume input, either 10 or 50 µL of blood with different cell concentrations were selected for CD4 + T cell separation using 10 or 50 µL of NP-CD4. Next, different NP-CD4 volumes were tested for cell separation, and the best working ratio was chosen. Separated cell samples in each experiment were analyzed by FACS to obtain the number of targeted cells after separation, and the separation efficiency was calculated. Parameters were selected based on separation of CD4 + T cells with high and stable efficiency using MACS.
To mimic the CD4 + T cell concentrations in the blood of HIV-infected individuals (200-600 cells/µL) or patients with AIDS (<200 cells/µL), and considering the limited biosafety levels for handling lethal infectious agents, whole blood was diluted 2-, 6-, and 12-fold, undiluted blood from healthy donors had CD4 + T cell concentrations ranging from to 600-1200 cells/µL.

Fluorescence-Activated Cell Sorting (FACS) and Flow Cytometry Analysis of Cell Population before and after Cell Separation
A total of 50 µL of cell samples (before or after MACS-separation) were incubated with 20 µL Tritest reagent in the Trucount tube for 15 min in the dark, followed by the addition of 450 µL lysis solution for fixing the cell-antibody complexes and lysing red blood cells in the samples. After 15 min, the samples were run on a FACSCalibur system, and the obtained data were further analyzed using Multiset software.

Immunofluorescence Labeling of Cell Population before and after Cell Separation
Twenty microliters of cell samples (before or after MACS-separation) were incubated with 5 µL CD4-FITC and 5 µL CD3-PE (Thermo Fisher Scientific) for 20 min in the dark, followed by the addition of formalin to the final concentration of 0.18% and incubation for another 10 min. Samples were centrifuged at 1500 rpm for 3 min then the supernatants were discarded to remove any excess antibodies that was not labeled to cells. Finally, fluorescence tagged cells fraction were re-suspended in 20 µL of PBS*. Fluorescence signals were observed under Axio Observer Co Apotome II (Carl Zeiss) fluorescence microscope at 20× magnification.

Cell Enumeration on Countess TM Automated Cell Counter
For cell counting on Countess TM , 10 µL of the sample was introduced into a Countess TM chamber slide, which was then inserted into the readout inlet of the Countess TM machine. In the preview mode, the focus was adjusted such that the cell margin appeared as clear as possible. The diameter was chosen to be 6-15 µm and the circularity was set at 75% to exclude any contaminants outside the range. Circularity was intentionally not set at an absolute 100% as the integrity of the cell's round shape may be affected during the MACS process or by image distortion.

Statistical Analysis
The bias of MACS-Countess against FACS was calculated as the ratio between the difference in cell count from MACS-Countess compared with that from FACS to the average cell count of the two methods as a percentage [23]. The correlation coefficient was used to assess the strength of the relationship between MACS-Countess and FACS, whereas the Bland-Altman plot was used to assess the agreement between the two methods; for each pair of cell count values, the average was calculated and then plotted on a graph as the mean against bias [23,24]. Mean bias and standard deviation (SD) were determined. The limit of agreement (LoA) (upper and lower) was referred by adding or subtracting the mean bias with 1.96 SD (mean bias ± 1.96 × SD).
Percentage similarity (PS) analysis was generated to directly compare both methods, and was calculated as the ratio between the average cell counts from the two methods and the reference method to percentages [25]. The mean percentage similarity (MPS) and SD were estimated, and the coefficient of variance (CV) was calculated as the ratio of SD to MPS in percentage.
The sensitivity, specificity, and misclassification rate [26] were also calculated at the threshold ≤350 cells/µL and ≤500 cells/µL, as at these cut-off values, initiation of ART is recommended for the treatment of HIV-infected patients [27]. The number of true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN) were determined. The TP at a particular cut-off, for instance, <350 (cells/µL), is defined as the case when FACS returns the cell count value of <350 (therefore eligible for ART treatment or "positive") whereas MACS-Countess also returned a cell count <350 (true prediction of positive case). Similarly (at cut-off 350), TN, FP, and FN are defined as the case when FACS and MACS-Countess return cell count values of >350 and >350, >350 and <350, and <350 and >350 (cells/µL), respectively. Sensitivity was then calculated as the ratio of TP to the sum of TP and FN, whereas specificity was the ratio between TN and the sum of TN and FP. The upward misclassification rate was calculated as 1 minus sensitivity, whereas the downward misclassification rate was calculated as 1 minus specificity. Positive Predictive Value (PPV) was calculated as the ratio between TP and the sum of TP and FP; Negative Predictive Value (NPV) was the ratio between TN and the sum of TN and FN. The same was carried out for the cut-off 500 cells/µL.

Separation of CD4 + T Cells from PBMCs Using NP-CD4
We aimed to establish a counting system that accurately determines CD4 + T cells in the whole blood. PBMC is known to be rich in CD4 + T cells (35%, [28]); therefore, to test the performance of NP-CD4 for the targeted cells, we first performed CD4 + T cell separation from PBMCs according to the recommended protocol.
As shown in Figure 1a,b, NP-CD4 had very high specificity for targeted cell capturing as CD4 + T cells accounted for >95% of the total lymphocyte population after the purification process, whereas only 32% of the population before separation was detected as CD4 + T cells. Further, we stained PBMCs and the separated cells with fluorescence-labeled CD4-FITC (green signal) and CD3-PE (red signal) antibodies and observed fluorescence emitting cells under a fluorescence microscope (Axio Observer Co Apotome II, Carl Zeiss). In this case, the monocytes emitted only green signals and CD4 + T cells emitted both green and red fluorescent signals, which were represented as green-and yellow-colored cells, respectively, in merged images under different light source illumination. Figure 1 indicated that the yellow-colored CD4 + T cells only form a minor part (Figure 1c), whereas after cell separation, they form the major population ( Figure 1d). The cell separation efficiency in five samples ranged from 46.2-85.2% (Figure 1e). purification process, whereas only 32% of the population before separation was detected as CD4 + T cells. Further, we stained PBMCs and the separated cells with fluorescencelabeled CD4-FITC (green signal) and CD3-PE (red signal) antibodies and observed fluorescence emitting cells under a fluorescence microscope (Axio Observer Co Apotome II, Carl Zeiss). In this case, the monocytes emitted only green signals and CD4 + T cells emitted both green and red fluorescent signals, which were represented as green-and yellow-colored cells, respectively, in merged images under different light source illumination. Figure 1 indicated that the yellow-colored CD4 + T cells only form a minor part (Figure 1c), whereas after cell separation, they form the major population ( Figure 1d). The cell separation efficiency in five samples ranged from 46.2-85.2% (Figure 1e). This result suggests that the procedure recommended by the manufacturer is of good quality for CD4 + T cell purification. However, this procedure requires a very large blood volume and complex instruments such as a swing-out centrifuge for operation. In the subsequent experiments, we tried to improve the cell separation conditions to be oriented for diagnostic purposes.

Optimizing Variation for CD4 + T Cell MACS Separation
To be applicable for diagnosis, the MACS procedure needs to separate CD4 + T cells directly from a small volume of blood (<100 µL) so that sample processing is both laborsaving, convenient, and fast for obtaining data. Therefore, we optimized the blood volume input and the NP-CD4 volume used. Similar to cell separation performance from PBMC, the isolated cells from blood were also primarily CD4 + T cells (>95% population) as confirmed by FACS analysis (Figure 2a,b), suggesting that non-specific binding of NP- This result suggests that the procedure recommended by the manufacturer is of good quality for CD4 + T cell purification. However, this procedure requires a very large blood volume and complex instruments such as a swing-out centrifuge for operation. In the subsequent experiments, we tried to improve the cell separation conditions to be oriented for diagnostic purposes.

Optimizing Variation for CD4 + T Cell MACS Separation
To be applicable for diagnosis, the MACS procedure needs to separate CD4 + T cells directly from a small volume of blood (<100 µL) so that sample processing is both laborsaving, convenient, and fast for obtaining data. Therefore, we optimized the blood volume input and the NP-CD4 volume used. Similar to cell separation performance from PBMC, the isolated cells from blood were also primarily CD4 + T cells (>95% population) as confirmed by FACS analysis (Figure 2a,b), suggesting that non-specific binding of NP-CD4 with monocytes and dendritic cells is negligible in a cell counting assay. Therefore, we continued to develop a CD4 + T cell separation assay without the need for a monocytedepletion step. cells in the sample covers the range of 100-600 cells/µL.
During the washing step, certain cell-NP-CD4 complexes were washed away as they did not bind strongly enough to the magnetic stand, and thus a small portion of complexes was lost, which interfered with the final separation efficiency (SE). As presented in Figure  2a, 50 µL of blood input had a tendency of higher separation efficiency and stability compared with 10 µL blood input, from all 3 dilution levels: X1 (67.9 ± 13.6% vs. 58.2 ± 12.3%), X2 (62.5 ± 22.2% vs. 40.4 ± 22.8%), and X6 (67.9 ± 16.9% vs. 48.9 ± 22.9%). We continued to test different amounts of NP-CD4 for separating CD4 + T cells from 50 µL of undiluted blood. Three different NP-CD4 volumes (50, 25, and 12.5 µL) were chosen. As shown in Figure 2b, the separation efficiency was not linear with respect to the amount of NP-CD4 in this range. While the use of 50 or 25 µL NP-CD4 for MACS separation was not significantly different, the use of 12.5 µL NP-CD4 showed a significantly lower efficiency (p < 0.05). In detail, 50 µL NP-CD4 was more than saturated for cell capture because 25 µL NP-CD4 provided a similar SE (72.4 ± 16.6%) to that of 50 As the CD4 + T cell concentration in a deficient immune system, such as in HIV-infected individuals, ranges from approximately 100 to less than 600 cells/µL, we used either 50 µL or 10 µL of blood with three dilution levels: undiluted (X1) and diluted twice (X2) or diluted sixth fold (X6) for MACS separation so that the concentration of CD4 + T cells in the sample covers the range of 100-600 cells/µL.
We continued to test different amounts of NP-CD4 for separating CD4 + T cells from 50 µL of undiluted blood. Three different NP-CD4 volumes (50, 25, and 12.5 µL) were chosen. As shown in Figure 2b, the separation efficiency was not linear with respect to the amount of NP-CD4 in this range. While the use of 50 or 25 µL NP-CD4 for MACS separation was not significantly different, the use of 12.5 µL NP-CD4 showed a significantly lower efficiency (p < 0.05). In detail, 50 µL NP-CD4 was more than saturated for cell capture because 25 µL NP-CD4 provided a similar SE (72.4 ± 16.6%) to that of 50 µL (59.7 ± 16.8%) (Figure 2b). The use of 12.5 µL NP-CD4 gained only 22.1 ± 19.1% SE (p < 0.05), indicating that the amount of NP-CD4 separated by this volume is insufficient for capturing the targeted cells in the samples. We thus conclude that a blood volume input of 50 µL and a NP-CD4 volume of 25 µL are optimal for good cell separation performance.

Comparison between MACS-Countess to FACS Methods
First, the linear regression plot between CD4 + T cells counted on Countess TM (Countess results) after MACS separation and the CD4 + T cells counted by FACS was generated. From that the experimental standard curves (STC) between two methods (n = 38) were yielded and correction factors for the MACS-Countess system were determined (Figure 3). After that, 15 samples with known CD4 + T cell concentration values from FACS analysis were analyzed on MACS-Countess, and CD4 + T cell counts were deduced using the correction factor of the STC (Table 1).
FACS system and the optimized MACS-separation procedure (50 µL blood input, 25 µL NP-CD4). The regression plot of CD4 + T cell counts from the two methods was generated using Microsoft Excel. We focused on the CD4 + T cell concentration from 600 to 100 cells/µL, as this is the reported cell concentration range in blood samples of patients with immunodeficiency. We also tested the MACS separation in blood samples with cell concentrations as low as <100 cells/µL to determine if MACS-Countess can return good results. Thus, we divided 38 samples into 4 groups based on the frequencies of CD4 + T cell counts found in HIV-infected patients: group I contained 4 samples with CD4 + T cell concentrations >600 cells/µL (undiluted blood), group II contained 14 samples with cell concentration 600-200 cells/µL (2-fold diluted blood), group III contained 15 samples with cell concentration 200-100 cells/µL (6-fold diluted blood), and group IV contained 5 samples with cell concentration <100 cells/µL (12-fold diluted blood). Thus, Figure 3a represents a linear regression equation for all samples from the four groups, which is y = 1.22x + 1.9, where x is the number of CD4 + T cell counts in samples after MACS separation and y is that with FACS. Upon further analysis for particular groups as shown in Figure  3b,c, two STCs were obtained: Groups I and II shared the same STC (I + II), which was y = 1.15x + 35.87 (Figure 3b), whereas the STC (III) for group III was y = 0.4x + 90.27 ( Figure  3c). Unfortunately, for group IV with extremely low CD4 + T cell counts, the error of MACS-Countess compared with FACS was very large because of the diversified plot  Table 1. Results of CD4 + T cell counts acquired using MACS-Countess and biases of MACS-Countess compared with FACS.

Generating STCs between Countess (MACS) and FACS
In all, 38 samples with different CD4 + T cell concentrations were analyzed using the FACS system and the optimized MACS-separation procedure (50 µL blood input, 25 µL NP-CD4). The regression plot of CD4 + T cell counts from the two methods was generated using Microsoft Excel. We focused on the CD4 + T cell concentration from 600 to 100 cells/µL, as this is the reported cell concentration range in blood samples of patients with immunodeficiency. We also tested the MACS separation in blood samples with cell concentrations as low as <100 cells/µL to determine if MACS-Countess can return good results. Thus, we divided 38 samples into 4 groups based on the frequencies of CD4 + T cell counts found in HIV-infected patients: group I contained 4 samples with CD4 + T cell concentrations >600 cells/µL (undiluted blood), group II contained 14 samples with cell concentration 600-200 cells/µL (2-fold diluted blood), group III contained 15 samples with cell concentration 200-100 cells/µL (6-fold diluted blood), and group IV contained 5 samples with cell concentration <100 cells/µL (12-fold diluted blood). Thus, Figure 3a represents a linear regression equation for all samples from the four groups, which is y = 1.22x + 1.9, where x is the number of CD4 + T cell counts in samples after MACS separation and y is that with FACS. Upon further analysis for particular groups as shown in Figure 3b,c, two STCs were obtained: Groups I and II shared the same STC (I + II), which was y = 1.15x + 35.87 (Figure 3b), whereas the STC (III) for group III was y = 0.4x + 90.27 ( Figure 3c).
Unfortunately, for group IV with extremely low CD4 + T cell counts, the error of MACS-Countess compared with FACS was very large because of the diversified plot (highest error of 72%, data not shown); therefore, it does not ensure the accuracy of cell counting at this threshold. Therefore, we concluded that the developed MACS-Countess system has a limit of detection (LOD) of 100 cells/µL.
Finally, considering all the data shown in Figure 2, we demonstrate the operation scheme of the MACS-Countess system, as illustrated in Figure 4.

MACS-Countess Compared with FACS
After determining the STCs and operation scheme of the MACS-Countess system, we performed CD4 + T cell counting using the MACS-Countess system on 15 samples, including 9 samples with CD4 + T of >200 cells/µL (belonging to groups I + II) and the other six samples of 100-200 cells/µL (belonging to group III). The original CD4 + T cell counts in the blood of each group were deduced from the experimental values using the corresponding STCs, as mentioned above. CD4 + T cell counts by MACS-Countess and their deduced values, CD4 + T cell counts by FACS, and the bias of the deduced values compared with the FACS values in each sample group are listed in Table 1. We found that the biases ranged from −34.8% to 32.7% in group I + II and from −25.1% to 29.3% in group III.
By adding 33 samples used for generating STCs (38 samples excluding 5 samples with blood concentrations of <100 cells/µL) to these 15 samples, a total of 48 samples was used for deducing the CD4 + T cell counts in blood based on the MACS-Countess method (in brief, deduced MACS-Countess) to evaluate the reliability of this method compared with FACS (in brief, FACS). Linear regression showed that the deduced MACS-Countess had a good correlation with FACS, with an R 2 equals to 0.86, slope of 0.96, and y-intercept of 22.3 (Figure 5a). We also compared the agreement of the deduced MACS-Countess with the PIMA analyzer, which is a commercially launched product for CD4 + T cell counting, using FACS as a reference. It has been reported that PIMA is comparable to conventional FACS with a mean bias of 0.2% and LoA from −42% to 42% in the field setting evaluation [13]. Meanwhile, the mean bias of the deduced MACS-Countess was 3.5% and LoA ranged from −36.4% to 43.3% (Figure 5b). The mean percentage similarity (MPS) of MACS-Countess was 102.9% with a coefficient of variance (CV) equaling 10.6% (Figure 5c), which is slightly different from that of PIMA with an MPS of 96.4% and CV of 13.1% [14]. The mean bias at cut-off <350 cells/µL of MACS-Countess was 6.4%, which was not significantly different from that at the cut-off of <500 cells/µL (5.4%, p < 0.05), suggesting that a good cell count performance is ensured at both low and high cell concentrations. The MPS of MACS-Countess at the two thresholds were 104.3% and 103.9%, respectively, and were not significantly different (p < 0.05).
Appl. Sci. 2021, 11, 9786 9 of 15 (highest error of 72%, data not shown); therefore, it does not ensure the accuracy of cell counting at this threshold. Therefore, we concluded that the developed MACS-Countess system has a limit of detection (LOD) of 100 cells/µL. Finally, considering all the data shown in Figure 2, we demonstrate the operation scheme of the MACS-Countess system, as illustrated in Figure 4.

MACS-Countess Compared with FACS
After determining the STCs and operation scheme of the MACS-Countess system, we performed CD4 + T cell counting using the MACS-Countess system on 15 samples, including 9 samples with CD4 + T of >200 cells/µL (belonging to groups I + II) and the other six samples of 100-200 cells/µL (belonging to group III). The original CD4 + T cell counts in the blood of each group were deduced from the experimental values using the corresponding STCs, as mentioned above. CD4 + T cell counts by MACS-Countess and their deduced values, CD4 + T cell counts by FACS, and the bias of the deduced values compared with the FACS values in each sample group are listed in Table 1. We found that the biases ranged from −34.8% to 32.7% in group I + II and from −25.1% to 29.3% in group III. Sensitivity,specificity, PPV, NPV, and misclassification rates were calculated to better understand the diagnostic results (Table 2). MACS-Countess deduced a good specificity of 94.7% at a cut-off of 350 cells/µL; however, the sensitivity was only 89.7%. This led to an upward misclassification rate of 10.3% and a downward misclassification rate of 5.3%. This means that among cases predicted to be eligible for ART treatment according to the reference method due to a CD4 + T cell count of <350 cells/µL, 10.3% of the cases would be falsely predicted (deduced by MACS-Countess) to be >350 and misclassified as ineligible Sensitivity, specificity, PPV, NPV, and misclassification rates were calculated to better understand the diagnostic results (Table 2). MACS-Countess deduced a good specificity of 94.7% at a cut-off of 350 cells/µL; however, the sensitivity was only 89.7%. This led to an upward misclassification rate of 10.3% and a downward misclassification rate of 5.3%. This means that among cases predicted to be eligible for ART treatment according to the reference method due to a CD4 + T cell count of <350 cells/µL, 10.3% of the cases would be falsely predicted (deduced by MACS-Countess) to be >350 and misclassified as ineligible until the next test. Additionally, at this threshold MACS-Countess had PPV and NPV of 96.3% and 85.7%, respectively. The PIMA system has either lower or similar upward misclassification values (0-10.4%) with a sensitivity of 89.6 to 100% [29,30], whereas it has a lower PPV at 91.2% and a higher NPV at 92.3% [31,32]. As the cut-off increases to 500 cells/µL, the specificity of MACS-Countess decreases to 88.9%, whereas the sensitivity remains 89.7%, which are not much different from those of PIMA, which are 95.5% and 84.2%, respectively [30].

Discussion
In this research, we used Anti-Human CD4 Particles-DM (NP-CD4, BD Biosciences) for capturing CD4 + T cells with a specificity higher than 95% from PBMCs sample (Figure 1a,b). Specificity of cell separation from whole blood sample as high as >95% was also confirmed (Figure 2a,b). This high selectivity could be attributed to a higher level of CD4 + biomarker expression on the surface of CD4 + T cells than on the surface of other cells, resulting in a stronger interaction affinity of the NP-CD4 with CD4 + T cells than with monocytes and dendritic cells. The optimal number of anti-CD4 molecules per nanoparticle of NP-CD4 could thus be a major factor in its high specificity.
Compared with the conventional method requiring 4 mL blood input, the optimized MACS procedure enables the separation of CD4 + T cells directly from a small blood volume of 50 µL, making the method friendly to patients. Further, the optimized procedure does not require a centrifugation process, resulting in a considerably shorter time and less labor for performing, while greatly reducing the risk of sample contamination to the environment.
The simplicity and small scale of the optimized MACS procedure are also available for further development for other applications, such as integration into microfluidic systems for automated operation. Microfluidic devices specifically for bio-separation have been developed for fractionation of CD34 + cells [33], rare cancer cells [34], and for capturing bacteria [35]. It is thus obvious that automation could reduce manual work and, therefore, decrease the errors originating from manual operation while increasing batch analysis. This study also provides a reproducible model of applying MACS to diagnosis with a simple cell isolation process, whereas a minimum of complicated instruments is required. Using a similar format, many other diagnostic tests can be developed by simply changing the antibody conjugated to the magnetic nanoparticles. The limit of detection of this assay is estimated to be 100 cells/µL and the assay accuracy deteriorates below this threshold. HIVinfected individuals with cell counts below 200 cells/µL have severe immunodeficiency and are usually susceptible to opportunistic infections. The treatment regimens for these patients in such circumstances are based on the type of opportunistic disease and/or CD4 + T cell counts mostly at a threshold of 100 cells/µL, as in yalaromycosis, Toxoplasma gondii encephalitis, cryptosporidiosis, and so on, a threshold of 150 cells/µL such as in histoplasmosis, or higher [36]. This LOD threshold is also sufficient enough to aid to assess the immune status of patient with severe SARS-CoV-2 infection because they usually have significantly lower CD4 + T cell counts (<200 cells/µL) than for mild-to-moderate group (<400 cells/µL) [4,5,37]. Lower CD4 + T cell counts also associated with higher mortality during disease treatment [5,38]. Furthermore, He et al. [37] has reported that among severe patients, the improved subgroup has a T lymphocyte count gradually back to a normal range within 15-25 days, while that of the dead subgroup keeps falling until death. These highlighted the need of monitoring CD4 + T cell counts in the treatment of COVID-19 patients having severe symptoms. Therefore, we consider that doctors can still refer to test results using the MACS-Countess method to make treatment decisions, or use this method for screening purposes.
In terms of accuracy, MACS-Countess showed high agreement with the gold standard FACS method with a mean bias of 3.5% and LoA from −36.4% to 43.3%. This is comparable to the commercially launched product, PIMA analyzer, which has a mean bias of 0.2% and LoA from −42% to 42% in the field setting evaluation (with FACS as the reference method).
Compared with the PIMA analyzer, MACS-Countess has slightly poorer sensitivity, resulting in a higher misclassification rate. This is reasonable as PIMA tends to underestimate the cell count with an MPS of 96.37% compared with 102.3% for MACS-Countess. Therefore, the upward misclassification rate of MACS-Countess can be reduced once the MPS is less than 100%. In practice, many other non-flow cytometric platforms tend to underestimate the CD4 + T cell counts. Underestimation would decrease the upward misclassification while increasing the downward misclassification, which will benefit the earlier initiation of ART. Lutwama et al. reported that Dynabeads (Dynal T4 Count, Dynal Biotech, Oslo, Norway) have a good sensitivity of 94% at a cut-off of <350 cells/µL. The specificity, however, is poor at 69% as Dynabeads highly underestimate the cell count, as illustrated by a low slope of 0.85 and the absolute mean difference (FACS-Dyna) of 8.8 cells/µL [39]. In the same report, the FDA-approved Cytospheres (CD4 Manual Count, Beckman Coulter, CA, USA) also repeat the same pattern with a high sensitivity yet low specificity, very low slope, and high absolute difference compared with the reference method at 99%, 32%, 0.58, and 58.6 cells/µL, respectively [39]. Conversely, a high positive bias leads to low sensitiity, typically illustrated by PointCare NOW (PointCare), also an FDA-approved platform [40], with a mean relative bias of 35% and a mean sensitivity of only 53% [41].
Based on these analyzed data, the MACS-Countess can be considered an alternative simple method for diagnostic purposes and is affordable for limited resource areas where expensive conventional FACS methods are unavailable.
To enable cell count by FACS, a center has to invest at least USD 27000 to buy a cytometer. In addition, the complexity of the fluidic system and optical system of the cytometer restricts its use to situations where the support of well-qualified technicians is present. In contrast, the MACS-Countess is very easy to access as the system requirements simply include: (i) a strong magnet for placing 1.5 mL microcentrifuge tubes that cost only several hundred US dollars, and (ii) an automated cell counter of only about USD 5000-6000, both of which are cost-effective and affordable for hospitals with limited resources. It is estimated that a CD4 + T cell counting test with MACS-Countess would cost less than USD 10 (the cost of 25 µL NP-CD4 accounts for USD 5). Meanwhile, the cost of a test based on the FACS method includes fluorescent-labeled antibodies, Trucount tubes, the fluid needed to run the cytometer machine (shutdown solution, clean solution, etc.) and cytometer maintenance and calibration, which is estimated to range roughly from USD 18-on the FACSCount system-to USD 30-40 on the FACSCalibur/FACSCanto system (these costs were estimated based on the import prices of reagents and consumables into Vietnam, which may not be universal).

Conclusions
In this study, we developed a simple and low-cost MACS-Countess system for the CD4 + T cell counting method. Using FACS as the reference method, MACS-Countess performance in blood covering the entire clinical range of CD4 + T cell concentration (105-980 cells/µL) has good reliability with a mean bias of 3.5% and LoA from −36.4% to 43.3%) along with good diagnostic performance with a sensitivity of 89.7% and specificity of >88.9% (n = 48), comparable to that of the commercially available point-of-care PIMA analyzer. With relatively good accuracy and affordable instrument requirements including a magnet and automated cell counter, the MACS-Countess offers a simple CD4 + T cell counting method that is suitable for resource-limited areas, where the gold standard FACS method is neither available nor affordable. This model is not only limited to CD4 + T counting, but can also be applied for counting other types of immune cells or cancer cells.  Institutional Review Board Statement: All subjects gave their informed consent for inclusion before they participated in the study. The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of Hanoi Medical University, No. IRB-VN01.001/IRB00003121/FWA 00004148.
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data used to support the findings of this study are available from the corresponding author upon request.