Introduction

Cell therapy products (CTPs) derived from induced pluripotent and embryonic stem cells are associated with a risk of tumorigenicity; therefore, a highly sensitive cellular detection method should be applied in preclinical biodistribution studies to evaluate tumorigenicity risk in humans (1,2,3,4,5). Several cell quantification methods, including imaging, flow cytometry (FCM), and quantitative polymerase chain reaction (qPCR), have been established (6,7,8,9). While the imaging method requires pre-labeling of CTPs (10,11,12), cellular labeling may alter the biodistribution properties and cell function of CTPs. Additionally, leakage of labeling substances from CTPs can cause artifacts. Meanwhile, although the FCM-based cellular detection method is a powerful tool to quantify cells in blood samples without requiring pre-labeling (6), its application to cellular quantification in tissue samples is challenging as tissue dissociation impacts the cell viability and recovery rate of CTPs (13). The qPCR-based method is highly sensitive and enables detection of CTPs without pre-labeling. qPCR targeting Arthrobacter luteus (Alu) elements (Alu-qPCR) is a highly sensitive cell quantification method as Alu elements account for > 10% of the human genome, with ~ 1.1 million copies (14,15,16). In our previous study, we demonstrated that the cell number in biological samples could be determined from the calibration curve prepared by spiking the standard cell lysates into the blank matrix (14). However, Alu-qPCR is subject to a matrix effect and/or variability in DNA recovery among different matrices. Therefore, a matrix-matched calibration curve should be used for qPCR-based cellular quantification. Nevertheless, even when matrix-matched calibration curves were employed, Alu-qPCR exhibited inaccuracy of quality control (QC) samples in the lung and spleen, indicating a matrix effect and/or variability in DNA recovery among samples with the same matrices. Furthermore, a serious drawback of Alu-qPCR is non-specific amplification in non-template control samples.

The LINE1-qPCR method exhibits dose-dependent amplification of the human-specific sequence in not only rodent samples but also monkey samples without background noise, indicating its superior selectivity to that of Alu-qPCR (17). In addition, it enables the highly sensitive detection of human-specific genes owing to its high copy number. However, a common drawback of qPCR-based quantification methods is the requirement of matrix-matched calibration curves to minimize matrix effects and DNA recovery variability. This requirement impedes efficiency and increases cost as analysis of small tissues, e.g., bone marrow and genitalia, requires pooling of many animals to harvest blank samples. The addition of an external control gene (EC) to samples minimizes the variability in DNA recovery among biological samples. We previously reported that the inclusion of an EC improves the quantification accuracy of chimeric antigen receptor (CAR) transgenes in blood samples (18), with an excellent correlation between CAR-qPCR and FCM-based quantification.

Droplet digital (dd) PCR runs reactions to the endpoint in partitioned individual droplets (19,20,21). The positive and negative droplets are divided based on droplet fluorescence, and DNA copy number is determined based on Poisson statistics. The endpoint assay in ddPCR minimizes the matrix effect (22), facilitating preparation of a single surrogate calibration curve in combination with EC and DNA recovery normalization, which is superior to the qPCR-based method in terms of animal ethics and cost.

In this study, we aim to develop a novel method for quantifying human cells across various matrix samples using a single surrogate calibration curve to exclude bottlenecks in the CTP biodistribution. We aim to overcome three challenges: non-specific amplification in the non-template control sample by Alu-qPCR, variability in the DNA recovery rate, and matrix effect on PCR amplification. Hence, to avoid non-specific amplification, a human-specific LINE1 element was selected as the target sequence. Normalization by EC and ddPCR was applied to minimize variability in DNA recovery and matrix effects, respectively.

Materials and Methods

Cells and EC Gene

Human peripheral blood-derived CD8+ T cells, isolated from a 22-year-old male donor (Japanese), were used as model cells (Precision for Medicine, Frederick, MD, USA). Dog genomic DNA (gDNA) was obtained from Zyagen (San Diego, CA, USA) to use as the EC gene.

Primers and Probes

Primers and probes for the human LINE1 sequence and canine MC1R genes (EC) were prepared by Thermo Fisher Scientific. The sequences are as follows: primers for LINE1: 5′-CCATTACTGGGTATATACCCAAATGAG-3′ (forward) and 5′-GCGCTGCACCCACTAATGT-3′ (reverse), probe for LINE1: 5′-TGAGTTCATATCCTTTGTAGGGA-3′; primers for EC: 5′-CGCCCATGTATTACTTCATCTGTTGCC-3′ (forward) and 5′-CACGGCGATGGCGCCCAGGAA-3′ (reverse), probe for EC: 5′-GCCTTGGCTGCGCAGGCTGCTGTGGTGCAG-3′. The probe was conjugated with the fluorescent reporter dye FAM (LINE1) or VIC (EC) at the 5′ end and a non-fluorescent quencher with a minor groove binder at the 3′ end.

Development of a ddPCR Method for the Detection of the LINE1 Sequence

The ddPCR reaction mixture comprised 4 × ddPCR Multiplex Supermix, primers (900, 1,800, or 2700 nmol/L), probes (125, 250, or 500 nmol/L), dithiothreitol (DTT, 4 mmol/L), human gDNA (0.6 ng/reaction, Thermo Fisher Scientific), and restriction enzymes (ScaI, 10 U/μL; Takara-Bio, Shiga, Japan) in a final 20 μL volume. The ddPCR assay was carried out using the QX ONE Droplet Digital PCR system (Bio-Rad Laboratories). The initial denaturation step was performed at 95°C for 10 min. The LINE1 sequence was amplified in 45 cycles at 95°C for 0.5 min/56–66°C for 1.5 min. Following enzyme deactivation at 98°C for 10 min, the droplets were categorized as positive or negative fluorescent droplets and enumerated using the QX ONE Droplet Digital PCR system. The copy number of LINE1 sequence was calculated using the QX ONE Software (ver.1.1, Bio-Rad Laboratories).

Development of a Simultaneous Detection Method for LINE1 and EC Sequences

To develop a duplex assay for LINE1 sequence and EC, EC (20 ng/reaction) was added to the ddPCR reaction mixture. The primer and probe concentrations in the ddPCR reaction mixture were 2700 nmol/L and 250 nmol/L for LINE1 and 500 nmol/L and 250 nmol/L for EC, respectively. The annealing temperature was set at 58°C. The copy numbers were calculated as described in the “Development of a ddPCR Method for the Detection of the LINE1 Sequence” section.

Development of a Cellular Quantification Method Using a Single Surrogate Calibration Curve

Preparation of calibration standard samples and QC samples were performed as previously reported (6). Working cell lysates were prepared with ATL buffer/proteinase K solution (9:1, v/v; Qiagen) at final concentrations of 105, 3 × 104, 5 × 103, 103, 102, and 30 cells/70 μL. Mouse tissues and blood were obtained from severe combined immunodeficient (SCID) mice (CB17/Icr-Prkdcscid/CrlCrlj, male, 8 weeks old; Jackson Laboratory Japan, Inc., Kanagawa, Japan). Animal studies involving mice were approved by the Institutional Animal Care and Use Committee (IACUC) of Shonan Health Innovation Park (Kanagawa, Japan) in a facility accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC). The control tissues were homogenized in ATL buffer/proteinase K at a final concentration of 15 mg tissue/150 μL homogenate (liver, kidney, heart, and testis), and 7.5 mg tissue/150 μL homogenate (lung and spleen). Working cell lysates were added to the control tissue homogenate or blood to prepare standard samples for calibration (30 to 105 cells/15 or 7.5 mg tissue, 50 μL blood). QC samples were prepared at the concentrations of 30, 102, 103, and 10.4 cells/15 mg tissue, 7.5 mg tissue, or 50 μL blood. Blank samples were also prepared by adding ATL buffer/proteinase K to control samples. Dog gDNA was added to all samples as the EC. DNA was extracted from the calibration standard samples, QC samples, and blank samples using a DNeasy 96 Blood & Tissue Kit (Qiagen) as previously reported (6).

For the qPCR assay, the PCR mixture consisted of 2 × TaqPath ProAmp Master Mix (Thermo Fisher Scientific), primers (1000 nmol/L for LINE1 or 500 nmol for EC gene), probes (500 nmol/L), and water, in a final 20 μL volume. LINE1 and EC were measured using QuantStudio 7 Flex Real-Time PCR System with SDS software (Thermo Fisher Scientific) (not a duplex assay) as previously reported (17). PCR was performed in triplicate per each sample.

For the ddPCR assay, the ddPCR reaction mixture was prepared as described in the “Development of a ddPCR Method for the Detection of LINE1 and EC Sequences” (duplex assay) section. The droplet generation was carried out using a QX200™ Droplet Generator (Bio-Rad Laboratories). Target-specific genes were amplified using a T100™ Thermal Cycler. Thermal cycle conditions are described in the “Development of a ddPCR Method for the Detection of LINE1 Sequence” section. After thermal cycling, positive and negative droplets were counted using a QX100 Droplet Reader (Bio-Rad Laboratories). LINE1 and EC gene copy numbers were calculated using Quanta Soft (Bio-Rad Laboratories).

Optimization of DNA Extraction Methods

DNeasy Blood & Tissue Kits and the MagMAX™ DNA Multi-Sample Ultra 2.0 Kit (Thermo Fisher Scientific) were used for DNA extraction. For DNeasy Blood & Tissue Kits, preparation of calibration standard samples and DNA extraction were conducted as described in the “Development of a Quantification Method Using a Single Surrogate Calibration Curve” section. For the MagMAX™ DNA Multi-Sample Ultra 2.0 kit, cells were lysed at a concentration of 1429 cells/μL using DNA/RNA Shield (Zymo Research, Irvine, CA, USA)/proteinase K (9:1, v/v) at 65°C overnight. The cell lysate was serially diluted with DNA/RNA shield/proteinase K at final concentrations of 104, 3 × 103, 5 × 102, 102, 10, and 3 cells/70 μL (working cell lysates). The control tissue samples were homogenized in DNA/RNA shield to prepare the blank homogenates. Working cell lysates were spiked into blank homogenates or blood to prepare the calibration standards. The concentrations were 3, 10, 102, 5 × 102, 3 × 103, and 104 cells/20 mg tissue (liver, kidney, heart, and testis), 10 mg tissue (spleen and lung), or 50 μL of blood. EC was added to all samples. DNA extraction from the calibration standard samples was conducted according to the method reported by Sugimoto et al. (19), with minor modifications. DNA was extracted from 200 μL tissue homogenate or 50 μL blood using the KingFisher™ Flex system (MagMAX_Ultra2_200μL_Flex and MagMAXUltra 2TissueV). DNA elution from each kit was measured using ddPCR. The assay condition of ddPCR was described in the “Development of a Quantification Method Using a Single Surrogate Calibration Curve” section.

Method Validation for the Quantification of Human Cells in SCID Mouse Samples

Method validation was conducted on three separate days as noted in the “Optimization of DNA Extraction Methods” section, with minor modifications. The calibration standard samples were prepared using liver homogenates at concentrations of 3, 10, 102, 5 × 102, 3 × 103, and 104 cells/20 mg of tissue. Two aliquots of each standard sample concentration were prepared. QC samples were prepared using blank tissue homogenates and blood at concentrations of 3, 102, 103, and 104 cells/20 mg of tissue (liver, kidney, heart, and testis), 10 mg of tissue (spleen and lung), or 50 μL of blood (n = 4). Dog gDNA was added to all samples as the EC. DNA extraction was performed using a MagMAX DNA Multi-Sample Ultra 2.0 kit. The DNA in the eluate was measured using ddPCR. All PCRs were performed in triplicate for each sample.

In Vivo Study

Male 8-week-old SCID mice were housed in individually ventilated cages on a 12-h light–dark cycle with free access to water and laboratory chow (CE-2; CLEA Japan, Inc., Tokyo, Japan). The temperature in an experimental animal room was set at 23–26°C. Human CD8+ T cells were intravenously administered to mice at a dose of 1 × 106 cells/mouse (n = 3). At 10 min, 1 h, and 24 h after administration, blood and tissue samples were collected from three animals. Samples were frozen immediately after weighing and stored at − 80°C until analysis. All animal experiments were conducted with the approval of the IACUC. The copy numbers of LINE1 and EC gene were measured as described in the “Method Validation for the Quantification of Mouse Samples” section.

Application to Samples from Different Strains

Calibration curves were prepared with liver homogenates from SCID mice as described in the “Method Validation for the Quantification of Mouse Samples” section. QC samples (102 cells/20 mg tissue, 10 mg tissue, or 50 μL blood, n = 4) were prepared from blank tissues and blood from NOD.Cg-PrkdcscidIl2rgtm1Sug/ShiJic (NOG; In-Vivo Science International Inc., Tokyo, Japan) and NOD.Cg-PrkdcscidIl2rgtm1Wjl/SzJ (NSG; Jackson Laboratory Japan, Inc.) mice. Cellular quantification was performed as described in the “Method Validation for the Quantification of Mouse Samples” section.

Results

Development of a ddPCR Method for the Simultaneous Quantification of LINE1 Sequence and EC Genes

The ddPCR Multiplex Supermix (Bio-Rad Laboratories, Inc., Hercules, CA, USA) was used in the present study as positive droplets were not observed when the ddPCR Supermix for Probes (Bio-Rad Laboratories) was used to detect the LINE1 sequence in human gDNA. The intensity of positive droplets was increased as the annealing temperature decreased and plateaued at < 58°C (Fig. 1a). Considering that negative droplet intensity was also increased as the annealing temperature decreased, 58°C was identified as the optimal temperature for detection of the LINE1 sequence. Additionally, the distribution of positive droplets narrowed as the primer concentration increased from 900 to 2700 nmol/L (Fig. 1b). The intensity of the positive droplets at a primer concentration of 2700 nmol/L was slightly higher than that at lower concentrations, whereas the intensity of negative droplets did not change. Therefore, the optimal primer concentration was set at 2700 nmol/L. No clear improvement in the signal-to-noise ratio was observed after adjusting the probe concentration (Fig. 1c). Under optimal PCR conditions for detection of the LINE1 sequence, a duplex assay was performed to simultaneously quantify the LINE1 sequence and EC genes. The assay showed good separation of positive and negative droplets for the LINE1 and EC genes (Fig. 1d).

Fig. 1
figure 1

Optimization of ddPCR conditions for the detection of LINE1 sequence and EC genes. Effects of annealing temperature (a), primer concentration (b), and probe concentration (c) on positive–negative separation. To optimize the annealing temperature, the primer concentration was set to 900 nmol/L. After investigating the primer/probe concentration dependency, the annealing temperature was set to 58°C. Blue dots represent positive droplets and gray dots represent negative droplets. d The duplex assay was performed at a primer concentration of 2700 nmol/L for LINE1 and an annealing temperature of 58°C. Blue, green, orange, and gray dots show LINE1 positive, EC positive, double-positive, and double-negative droplets, respectively

Full size image

Development of a Cellular Quantification Method Using a Single Surrogate Calibration Curve

When the number of cells in the QC samples was calculated using a matrix-matched calibration curve without DNA recovery normalization by EC, the accuracies of ddPCR and qPCR assays varied with ranges of 56.0–306.4% and 60.5–247.7%, respectively (Fig. 2a, d). However, the accuracy was improved by normalization of DNA recovery in ddPCR (74.5 to 130.6%) and qPCR (54.2 to 169.8%) assays (Fig. 2b, e). When the cell number in the QC samples was calculated using the liver calibration curve, the accuracy of 11 samples exceeded ± 35% for qPCR, whereas that of all samples, excluding two, was within ± 35% for ddPCR (Fig. 2c, f). The average accuracy value was closer to 100% for ddPCR than for qPCR. In the qPCR assay, more than half of the accuracy values exceeded ± 35% for QC samples with 104 liver cells and 103 and 3 blood cells (Fig. 2f).

Fig. 2
figure 2

Effects of DNA recovery and matrix on the accuracy of quality control samples. Accuracy was calculated using matrix-matched calibration curves (a–e) and single surrogate calibration curves (c and f) with (b, c, e, and f) and without (a and d) the DNA recovery rate of ddPCR (a–c) or qPCR (d–f) assays. Circles represent the accuracy of each quality control (QC) sample and the black bars represent average values. QC samples were prepared at final concentrations of 30, 102, 103, and 104 cells/15 mg tissue, 7.5 mg tissue, or 50 μL blood (n = 4) Red dot lines represent 100% ± 35%

Full size image

Optimization of the DNA Extraction Methods

DNA was extracted from various matrices, including human cells, using the DNeasy 96 Blood & Tissue Kit (Qiagen, Hilden, Germany) silica column-type and the MagMAX™ DNA Multi-Sample Ultra 2.0 (Thermo Fisher Scientific, Waltham, MA, USA) magnetic bead-type extraction methods. For the silica column-type extraction kit, the liver and spleen showed the lowest LINE1 sequence copy numbers over the designated concentration ranges (Fig. 3a). Variability in LINE1 copy number among matrices was observed, indicating discrepancies in the DNA recovery rate among matrices. In contrast, LINE1 copy numbers were comparable across the matrices for the magnetic bead-type extraction kit (Fig. 3b). Furthermore, the magnetic bead-type method showed 3- to 9.6-fold higher copy numbers in the range of 100–10,000 cells than the silica column-type method without any change in input human cell number for DNA extraction and input DNA template for PCR, suggesting a higher recovery rate of the magnetic bead-type method. Hence, a magnetic bead-type extraction method was used in subsequent experiments owing to its low variability in the recovery rate and high sensitivity.

Fig. 3
figure 3

Comparison of DNA recovery between column-type and magnetic bead-type extraction methods. The copy number of the LINE1 sequence in the PCR reaction mixture was measured using ddPCR. DNA was extracted from various matrices, including human cells, by silica column-type (a) and magnetic bead-type (b) extraction methods. Orange, green, light blue, red, gray, yellow, and blue dots represent the copy numbers from the blood, liver, heart, kidney, spleen, lung, and testis samples, respectively

Full size image

Method Validation for the Quantification of Mouse Samples

To validate our method, QC samples of four different concentrations were prepared using blank matrices on three different days. The LINE1 sequence copy number was quantified using a single surrogate calibration curve with a blank liver sample. The relative error (RE) and coefficient of variation (CV) values were − 18.9%, 31.5%, and 0.7–33.1%, respectively (Table I). The upper limit of quantification was 10,000 cells. The lower limit of quantification was determined by a signal-to-noise ratio of more than 30-fold. The regression analysis showed good linearity (r2 = 0.9954–1.0000) and similar slopes of the calibration curves in the range of 3–10,000 cells between the validation batches. These results demonstrate the robustness of this quantitative method.

Table I Validation of LINE1-ddPCR Assay in SCID Mice Samples
Full size table

Application to In Vivo Samples

The applicability of LINE1-ddPCR to in vivo samples was evaluated using in vivo samples after a single intravenous (tail vein) injection of human CD8+ cells to SCID mice. The number of cells in the in vivo samples was quantified using LINE1-ddPCR (Fig. 4). Human CD8+ cells were mostly observed in the lungs at 10 min after administration. The cells showed short persistence in the lungs and subsequent distribution into other organs with a tendency to remain in the spleen. The total cell number 10 min after administration was ~ 7.0 × 105 cells, indicating a good mass balance.

Fig. 4
figure 4

Human CD8+ T cell concentration in the tissues and blood of SCID mice after a single intravenous administration of human CD8+ T cells. Each point represents observed values. The orange, light blue, green, blue, yellow, light blue, and gray circles represent concentrations of human CD8+ T cells in the blood, liver, heart, kidney, spleen, lung, and testis, respectively

Full size image

Application to Samples from Different Mouse Strains

To evaluate the applicability of our method to samples from different strains, QC samples were prepared using blank NSG and NOG mice samples, and the cell numbers in QC samples were quantified by surrogate calibration curves using blank liver samples from SCID mice. The positive droplet intensity was comparable to that for SCID mice; the same threshold separated positive and negative droplets (data not shown). The average RE value ranged from − 30.0–6.8% among matrices, and each RE value was within ± 35% for all QC samples except one from the testis of NOG mice. Moreover, the CV values ranged from 1.4 to 31.5% (Table II).

Table II Applicability of LINE1-ddPCR Assay to NSG and NOG Mice Samples
Full size table

Discussion

In the present study, we selected the LINE1 sequence as a human-specific target analyte. LINE1 is a retrotransposon with a copy number of ~ 500,000, accounting for ~ 17% of the human genome, thus enabling the highly sensitive detection of human-specific genes. However, it is technically challenging to detect retrotransposons using ddPCR as they include several subfamilies with different sequences (23), often resulting in rain droplets in the intermediate region between the positive and negative droplets. To overcome this issue, we used a primer/probe set designed to be identical to the Ta-1 family of LINE but with Ta-0 and non-Ta families as Ta-1 is a young family and contains human-specific sequences. Additionally, we optimized the PCR conditions, including master mix, annealing temperature, and primer and probe concentrations. LINE1-ddPCR required the ddPCR Multiplex Supermix, whereas it did not show any positive droplet when the ddPCR Supermix for Probes was used, indicating that the key ingredients of a PCR can be critical for the amplification of target subsequences. Although the intensity of positive droplets increased at low temperatures, that of negative droplets also increased (Fig. 1a), potentially due to the amplification of LINE1 subsequences other than the target sequence. Hence, an adjustment of the annealing temperature was necessary to obtain a clear separation in the ddPCR assay of multicopy genes. Nonetheless, the increase in primer concentration had a minimal effect on the intensity of positive droplets, and the increase in probe concentration did not improve the signal-to-noise ratio (Fig. 1b and c), which is consistent with a study by Newkirk et al. (24). The manufacturer (Bio-Rad Laboratories) suggested that primer/probe concentration (900/250 nmol/L) would be sufficient to reach a plateau of PCR amplification until the endpoint of the ddPCR assay.

The parallel amplification of multiple-copy genes, such as LINE1, and different target sequences are difficult with qPCR as PCR efficiency is affected by the amplification of different target sequences (25). In the present study, we were unable to achieve simultaneous detection of LINE1 and EC using qPCR (data not shown). In contrast, a duplex assay for the LINE1 and EC genes was performed using ddPCR (Fig. 1d). This could be owing to the results of the ddPCR assay being less affected by the matrix effect based on the endpoint feature of ddPCR. Given that the duplex assay reduces the effort, time, and cost of PCR reactions, while also avoiding experimental errors in sequential single-plex PCR assays, it is advantageous for CTP biodistribution studies.

To develop a PCR-based quantification method for human cells in various biological samples using a single surrogate calibration curve, it is critical to normalize the variability in DNA recovery. In this study, we used canine gDNA as an EC to correct DNA recovery and ddPCR to minimize the matrix effect. The variability in the QC sample accuracy was reduced by the normalization of DNA recovery, indicating that the DNA recovery rates were largely variable across samples. This was supported by our previous report on CAR transgene quantification in blood samples in CAR T-cell therapy (18), which also demonstrated the variability in DNA recovery across blood samples and the utility of EC to correct the variability. In the present study, we confirmed the robustness and applicability of the DNA recovery normalization concept to not only blood samples but also various tissue samples. Although ddPCR without DNA recovery normalization proved to be less accurate than the qPCR results, this may be because the matrix effect decreased PCR amplification in QC samples and inadvertently improved the accuracy of the qPCR assay (Fig. 2a and d). Indeed, when DNA recovery was normalized by EC, ddPCR showed superior accuracy to qPCR (Fig. 2b and e). In addition, 3/16 liver QC samples (2/3 for QC samples with 10,000 cells and 1/3 for QC samples with 1000 cells) showed unacceptable accuracy (> 35%) in qPCR, even with application of the matrix-matched calibration curve, indicating the presence of a matrix effect even in the same matrix samples. However, the accuracy of all liver samples was within ± 35% for the ddPCR assay. Furthermore, we found that ddPCR can quantify human cells in various tissue samples with a single surrogate calibration curve, whereas qPCR failed to quantify QC samples with 104 cells/20 mg for the liver and 103 and 3 cells/50 μL for the blood (Fig. 2c and f). These results indicate that DNA recovery normalization by EC and matrix effect avoidance by ddPCR are crucial for developing a PCR-based quantification method for human cells in various biological samples using a single surrogate calibration curve.

A previous study reported that NOG mice transplanted with 10 HeLa cells spiked with human somatic stem cells generated subcutaneous tumors with a probability of 1/6 (26). Therefore, the sensitivity of cellular quantification is highly critical to assess the tumorigenicity risk for biodistribution studies of CTPs. Moreover, Sugimoto et al. reported that the DNA extraction method using magnetic beads yielded a 2.65-fold higher DNA amount from NSG mouse blood samples than a silica-type method (19); therefore, we evaluated the utility of the magnetic bead-type method in various tissue samples. We detected three cells per 20 mg tissue (liver, kidney, heart, and testis), 10 mg tissue (spleen and lung), or 50 μL blood using the magnetic bead-type method, suggesting that this method has a higher sensitivity than the silica-type method (Fig. 3). Furthermore, the magnetic bead method reduced the variability in the recovery rate among matrices. As DNA recovery variability would limit application of the cellular quantification method to various biological samples using a single surrogate calibration curve, the magnetic bead method would be suitable for biodistribution studies.

The validation experiments demonstrated that our method had an acceptable quantitative power in terms of linearity, selectivity, accuracy, and precision (Table I). Furthermore, we confirmed that the developed method was applicable to human cell quantification of in vivo samples (Fig. 4); the results were consistent with those of a previous report (14). In addition, 51Cr-labeled mouse T cells injected into mice were reported to show similar biodistribution (27). Furthermore, the blood concentration–time profiles of human CD8+ T cells in the present study are comparable to those previously measured using FCM-based analysis (6). Hence, the LINE1-ddPCR assay can be utilized in biodistribution studies, as is the case with other methods, such as Alu-qPCR, radioactivity measurement, and FCM.

The LINE1-ddPCR method enables highly sensitive cell quantification with a single surrogate calibration curve. Moreover, it was applicable not only to samples from SCID mice but also to those from different strains, such as NSG and NOG mice (Table II), which are often used in CTP development owing to their severe immunodeficiency. With the conventional qPCR method, a large number of blank matrices are required to prepare matrix-matched calibration curves. Moreover, the large number of animals needed for cellular quantification in small tissues would cause bottlenecks in biodistribution studies owing to increased cost, ethical issues pertaining to the number of animals used, and the limited availability of animals with a supply shortage, such as NSG and NOG mice. The LINE1-ddPCR method would circumvent these issues. Biodistribution studies of CTPs are often conducted in animals other than mice, such as monkeys, owing to their physiological similarity to humans. However, it could be difficult to obtain blank matrices from monkeys using Alu-qPCR, which has been used to detect human cells in rodents. Considering that the LINE1 primer/probe is applicable to human cell quantification in preclinical biodistribution studies using monkeys (17), LINE1-ddPCR might facilitate the quantification of human cells in monkey samples using surrogate calibration curves.

The volume-based units such as “cells/mg tissue” and “cells/μL blood” are pharmacokinetically meaningful because they enable to compare concentrations between blood and tissue samples and build comprehensive and quantitative models such as the physiologically based pharmacokinetic (PBPK) model (18, 19). In the present study, we prepared the spiked-in calibration curve to obtain the cell number in biological samples. Although the ddPCR enables absolute quantification of target gene copy number in a reaction mixture, the DNA recovery rate should be considered to obtain the nominal cell number in biological samples. It has been reported that the absolute CAR level could be calculated via the normalization of the CAR copy number/μL reaction mixture generated in ddPCR assay without calibration curves using the sample volume, eluate volume, and DNA recovery rate (19). Although this calculation is based on the assumption that the DNA recovery is consistent among all samples, the recovery rate is variable. In addition, the DNA recovery should be determined by comparing the observed copy numbers between the pre-spike and post-spike QC samples. On the contrary, in our method, the combination of the spike-in calibration curve and the addition of EC normalized the variability in DNA recovery among samples and helped determine the nominal cell concentration in biological samples.

In the present study, we set the acceptance criteria for accuracy and precision to be within 35% RE and < 35% CV, respectively, based on our previous report (14). These criteria were in line with the recommendations from the Global CRO Council in Bioanalysis (GCC) (29). Although some studies have proposed different criteria (30,31,32,33,34), there is no one criterion for the spike-in method similar to our method. The discussion should be continued to provide recommendations or regulatory guidance/guidelines on assay qualification/validation using the spike-in method for the quantification of CTPs using ddPCR.

There are several limitations to our study. First, although we believe that our method is applicable to tissues and animal species other than mice, further research is needed to confirm this theory. In addition, it is necessary to determine whether our method can quantify cells in long-term biodistribution studies as most CTPs are typically designed to achieve long-acting therapeutic effects in the body. Cell population and tissue homeostasis can be altered during long-term treatment, which would cause time-dependent changes in matrix composition and raise concerns such as matrix effects and DNA recovery variability. We believe that these concerns can be minimized by our method, which avoids matrix effects and corrects DNA recovery variability. In terms of the drawback of ddPCR, this assay has a relatively narrow dynamic range owing to the limitation of the droplet number compared with qPCR (28), which can be critical as CTPs can show marked proliferation in vivo beyond the dynamic range of ddPCR. Nevertheless, as ddPCR technology is newly developed, its dynamic range should be expanded in the future.

Conclusions

In this study, we established a novel ddPCR method for the quantification of human cells using a single surrogate calibration curve and demonstrated its advantages over conventional qPCR. The normalization of DNA recovery and avoidance of matrix effects allows accurate quantification using a single calibration curve. The developed method has the potential to facilitate biodistribution studies of CTPs by enabling single surrogate calibration with minimized animal sacrifice. Our findings suggest a potentially standardized methodology for ddPCR-based biodistribution assays.