Abstract
Chimerism analysis by next-generation sequencing (NGS) is an emerging method for engraftment monitoring after allogeneic hematopoietic cell transplantation. A high-sensitivity method is required for the detection of microchimerism (<1% chimerism), which may have clinical utility in early relapse detection, allograft monitoring in organ transplantation, and other allogeneic cellular therapies (such as microtransplantations). As more clinical laboratories adopt this method, a thorough assessment of performance is needed. This study evaluated one such NGS-based assay that uses both single-nucleotide polymorphisms and insertions/deletions as genetic markers. An assessment of accuracy, linearity, sensitivity, and reproducibility was performed. Analytical sensitivity was 0.2% donor for single donor and 0.5% donors for double donors. The assay showed a high degree of reproducibility over a full range of chimerism. Comparison to short-tandem-repeat (STR) PCR showed high concordance; yet <5% chimerism was consistently detected by NGS, but not by STR-PCR. Comparison to real-time quantitative PCR showed high concordance, but with lower correlation in the midrange (40% to 60% chimerism). Overall, the assay showed consistent performance with high sensitivity and accuracy compared with STR-PCR and real-time quantitative PCR across a full range of chimerism in the setting of single-donor and multidonor transplantations. In addition, criteria for quality metrics were established for sequencing performance and data analysis and considerations made for clinical laboratory validation of NGS-based chimerism assay and analysis software.
Allogeneic hematopoietic cellular transplant (alloHCT) is an essential curative therapy for many hematologic and genetic diseases, with indications for cellular therapies increasing in recent years.1 Successful alloHCT achieves engraftment of donor hematopoiesis for restoring hematopoietic functions and achieves graft-versus-leukemia effect in cases of malignant diseases. Engraftment is assessed through chimerism analysis, whereby polymorphisms distinguishing the donor are detected and quantitated, relative to the recipient, and can be reported as percentage donor. Currently, short-tandem-repeats (STRs) analysis is the most used molecular assay and gold standard for chimerism analysis.2 STR is based on differences in the length of STR between donor and recipient using distinct fluorescent primers, in multiplexed reactions detected by capillary electrophoresis. Although robust, STR analysis has limited analytical sensitivity, typically between 1% and 5% donor. Increasing recognition of the potential prognostic value of microchimerism (defined as chimerism levels <1%) in early relapse detection highlights a need for clinical laboratory assays with increased analytical sensitivity and precision at the limit of detection.3 More recently developed methods, including real-time quantitative PCR (qPCR) and droplet digital PCR, have greater analytical sensitivity than STR; however, they have not yet been widely adopted, each with its own advantages and limitations.2 Chimerism testing by qPCR relies on detection of single-nucleotide polymorphisms (SNPs) or insertion/deletion (InDel) polymorphisms for the donor-recipient pair and quantitation based on calculation of differences in threshold cycles (ΔΔCT), normalized against a reference gene CT. Although qPCR exhibits greater sensitivity than STR (detecting at least 0.1% chimerism), it has reduced accuracy and precision at increased proportion of the minor genome.4,5 Additionally, chimerism assessment involving more than two genomes is extremely challenging, and in some cases, not feasible by qPCR. In the case of droplet digital PCR, the technology allows for higher accuracy and precision, attributable to individual reaction partitioning and amplification end point data measurements, permitting direct absolute quantitation, not requiring a reference.6, 7, 8 Next-generation sequencing (NGS) offers unparalleled multiplexing and high-throughput interrogation of DNA sequences, allowing the development of highly sensitive chimerism assays, which target many biallelic SNPs and/or InDels in multiplexed reactions. With reported sensitivity of 0.1%, improved accuracy, and higher throughput, compared with qPCR and droplet digital PCR,9, 10, 11, 12 our laboratory validated and implemented one such assay, the ScisGo Chimerism Multi-Donor Assay, for clinical testing and monitoring of patients after alloHCT, including patients with multiple donors.
Materials and Methods
Sample Source
A total of 321 samples were tested in this validation, consisting of 80 pretransplant and 241 chimeric samples. The chimeric samples included 84 clinical samples, collected between 2019 and 2021 from patients after alloHCT. These included samples from patients undergoing related or unrelated donor transplants. Some of the transplants were double donor or retransplantation (ie, these post-alloHCT samples contained three genomic contributors). The remaining chimeric samples were artificial chimeras, prepared by mixing DNA from two unrelated individuals (one as donor, another as recipient), simulating a single donor transplantation. Multidonor transplantations (or previous transplantations) were simulated by mixing three unrelated individuals (two as donors, the third as recipient) or four unrelated individuals (three of them as donors, the fourth as recipient) in variable ratios. Method comparison study (NGS versus STR-PCR) and reproducibility study used the 84 clinical samples, which were tested by a reference laboratory using STR-PCR. The STR-PCR testing was performed using the GlobalFiler PCR Amplification Kit (ThermoFisher, Waltham, MA; catalog number 4476135), a six-dye STR multiplex kit, amplifying 21 autosomal STR loci, amelogenin, 1 Y STR locus, and 1 Y InDel locus. PCR products were processed on an ABI PRISM 3500xl genetic analyzer (ThermoFisher) and analyzed by GeneMapper ID-X v1.5 (ThermoFisher). Method comparison study (NGS versus qPCR) used 85 of the artificial chimeric samples. Chimerism testing by qPCR was performed in two steps, initial genotyping using a commercially available 39 markers KMRtype Genotyping kit (GenDx Products Inc., Utrecht, the Netherlands) to identify informative markers (IMs) that can distinguish the donor from the recipient, followed by second step to track the chimeric status for quantifying the identified IMs using the commercially available KMRtrack kits (GenDx Products Inc.). These reagents are available both as RUO (Research Use Only) and CE-IVD (Conformité Européenne In Vitro Diagnostic). Results of testing were analyzed by a companion software KMRengine version 2.0 (GenDx Products Inc.). For accuracy study, limit of detection (LOD) study, and linearity study, 148 of the artificial chimeric samples were used.
The use of remnant samples for development and validation of new clinical laboratory tests was approved by the Memorial Sloan Kettering Cancer (New York, NY) Center Institutional Review Board, under the protocol Institutional Review Board 13-037, and determined to present minimal risk to patients and patient privacy. Patient consent and Health Insurance Portability and Accountability Act authorization for this study were waived.
DNA Extraction and Quantification
DNA was isolated from either pretransplant blood or buccal swab samples (recipient and donor) or from post-transplant samples (blood and cell subpopulation) using EZ-1 DNA blood kit or using EZ-1 DNA Tissue kit on Qiagen (Hilden, Germany) EZ1 Advanced XL automated instrument, according to manufacturer's instructions. The DNA concentration was quantified by dsDNA HS kit or dsDNA BR Assay kit on QuBit 4.0 fluorometer (ThermoFisher), according to manufacturer's instructions.
Next-Generation Sequencing Chimerism Library Preparation
The NGS-based chimerism assay, ScisGo Chimerism Multi-Donor Assay (Scisco Genetics Inc., Seattle, WA), targets >200 SNPs and InDels spanning all chromosomes. Markers were chosen from the dbSNP database to identify SNPs and InDels that spanned all chromosomes and for which allele frequency was approximately 50% for both reference and alteration (average heterozygosity, approximately 0.5) in all populations comprising the dbSNP database. Samples were tested according to manufacturer's protocol. Briefly, each sample was amplified using four multiplex PCRs to generate a target amplicon library (stage 1 PCR). In a second PCR (stage 2 PCR), sequencing adapters including unique sample barcode were introduced into each amplicon, enabling pooling of up to 48 samples in each sequencing run. The stage 2 PCR pools were then purified using Select-A size DNA Clean & Concentrator (Zymo Research, Irvine, CA), according to manufacturer's instructions. The purified and denatured library was loaded onto the MiSeq reagent cartridge v3 kit (150 cycles) and sequenced 2 × 50 bp cycles on an Illumina MiSeq instrument (Illumina, San Diego, CA).
Sequencing Data and Chimerism Analysis
The generated FASTQ files were analyzed with a cloud-based bioinformatics pipeline provided by the manufacturer (ScisCloud version 1.9.14), for automated data analyses, including NGS quality control (QC) data analysis and chimerism data analysis. The following parameters were included in NGS QC data analysis: NGS library concentration, cluster density, cluster passing filter rate, and percentage >Q30 (quality score of 30). Briefly, chimerism analysis involves identification of donor- and recipient-specific IMs and calculation of percentage of both donor and recipient contribution along with a 95% CI in the post-transplant samples involving two or more genomes. The minimum coverage depth of markers in the computation was defined as 1000× for both pre and post samples. Markers are classified as heterozygous (HET) or homozygous (HOM) based on read count per allele (heterozygous if min/sum >0.45; or homozygous if min/sum <0.01; where min = minor allele, or the allele with lower read depth). Informative marker selection is based on genotyping the patient (pretransplant) and donor samples. The three types of informative markers are as follows: i) opposite HOM, the patient and donor are homozygous for different alleles, ii) HET-HOM, the patient is heterozygous and the donor is homozygous, and iii) HOM-HET, the patient is homozygous and the donor is heterozygous. All IMs (ie, opposite HOM, HOM-HET, and HET-HOM) are automatically selected, and dominant heterozygous markers are automatically excluded, but also can be deselected by manual review for outliers. The 95% CI is a measure of variation among the informative markers used in the computation. LOD approximates the noise level of the markers and is computed using homozygous (noninformative) markers that are above threshold. An LOD >1% may indicate either contamination or misassigned donor-recipient samples. The percentage donor or recipient in the post-transplant sample is calculated as the average of results of the informative markers.
Results
Sequencing Performance
Quality Metrics for Each NGS Run
A total of 26 NGS sequencing runs were performed. Average cluster density, clusters passing filter, and percentage of reads with >Q30 are summarized in Figure 1. All quality metrics were within ±3 SDs. Average concentrations of four amplicon pools libraries were 13.5 ± 3.0, 13.9 ± 3.2, 15.2 ± 2.7, and 17.3 ± 2.8 ng/μL, and the lowest library was 8.5 ng/μL but normalized into 2 ng/μL as per manufacturer's instructions (Figure 2A), which is equivalent to 4 nmol/L. The average concentration of pooled libraries was 32.0 ± 5.4 nmol/L, and the lowest library was 23.9 nmol/L, well above the minimum requirement of 4 nmol/L, per manufacturer's instructions (Figure 2B). For these QC parameters, variation across runs was low, and range of data was narrow. Because of this and the limited number of runs in this study (n = 26), using SDs was not the most appropriate statistical method to establish acceptability criteria or minimum QC metrics. The acceptable range or minimum for each QC parameter was originally based on sequencing platform performance specifications provided by the manufacturer, or instructions provided by the chimerism assay vendor. With additional data following these initial 26 runs, acceptability criteria have been confirmed. The criteria for QC parameters are summarized in Supplemental Table S1. Establishing these criteria is crucial for ensuring library preparation consistency and sequencing performance consistency over time.
Figure 1.
Sequencing performance using the chimerism multidonor kit. Quality metrics for each next-generation sequencing (NGS) run. For 26 NGS runs, cluster density, frequency of clusters passing filter, and percentage of reads >Q30 (quality score of 30) were determined using Sequence Analysis Viewer version 2.4.7 (Illumina), after each NGS run. Dashed lines indicate 3 SDs for each parameter. A: Cluster density. B: Clusters passing filter. C: Percentage of positions ≥Q30.
Figure 2.
Library performance using the chimerism multidonor kit. Summary of four amplicon (Amp) pools and final pooled library for total 26 next-generation sequencing (NGS) runs. Dashed line indicates minimum library concentration. A: Concentration of four amplicon pools. B: Concentration of final pooled library.
Quality Metrics for Each Sample
The cloud-based bioinformatics pipeline software user interface provides visual representation of quality metrics, including heat maps of read bias and divergent coverage (Supplemental Figure S1) for each sample in each NGS run as indicators of contamination and other DNA library preparation problems. Read bias indicates orientation bias in the DNA library preparation, which may result from contamination. Divergent coverage indicates that a marker has amplified at a significantly higher level (relative to the other markers) in a sample when compared with the other samples on the run. This imbalance may indicate poor PCR performance or contamination. Additionally, the Amp Coverage bar chart (Supplemental Figure S1) indicates total read depth for each amplicon pool as well as dimer artifacts.
In this study, average coverages per amplicon pool across 80 pretransplant and 155 post-transplant samples were 4121× (SD, 1575×) and 4347× (SD, 1678×), respectively, and average coverage in each amplicon was similar (Figure 3). In the software, a threshold for minimum coverage is set at 1000× per amplicon as a default for analysis. Additionally, a threshold of 100× was used in cases of samples with <1000× coverage per amplicon and generated accurate results (data not shown). Overall, 1 of the 80 pretransplant validation samples (Figure 3A) showed <1000× average coverage but was sufficient for chimerism identity analysis by applying 100× as threshold. Meanwhile, 1 of the 155 post-transplant validation samples showed <100× average coverage (Figure 3B), which was not analyzable because of insufficient reads. In addition, 2 of the 155 post-transplant samples showed <1000× but >100× average coverage (Figure 3B), which were sufficient for chimerism analysis using 100× as threshold. All remaining pretransplant and post-transplant samples showed sufficient average coverage, which successfully analyzed chimerism results.
Figure 3.
Average coverage per amplicon across total 235 validation samples was summarized among four amplicon pools. A: Pretransplant samples. B: Post-transplant samples. Green dashed line (1000) indicates vendor-recommended default threshold; red dashed line (100) indicates minimum threshold based on internal validation; red asterisks indicate post-transplant samples with average coverage <1000× but >100×. Red arrow indicates pretransplant sample with <1000× average coverage but was sufficient for chimerism identity analysis by applying 100× as threshold (A) and post-transplant sample with <100× average coverage and unable to analyze due to insufficient reads (B). n = 80 (A). n = 155 (B).
Quality Metrics for Chimerism Data Analysis
LOD (percentage) is a measure of the sequencing signal noise and is used in the analysis software to monitor for PCR contamination or misassigned sample(s). In the latter case, for example, the LOD will be well above an acceptable level if any of the pretransplant or donor samples are not associated with the post-transplant sample. LOD is computed from the average ratio of minor allele reads/total reads of markers for which donor and recipient are homozygous and homologous. Ninety-eight percent (152 of 155) of post-transplant validation samples had an LOD of <0.2% (Figure 4A). Three samples had an LOD >0.2%, but <1.0% (Figure 4A), the level of sequencing noise at which flags alert the user of the possibility of PCR contamination or use of misassigned donor/recipient sample. One of the three samples with an LOD of 0.23% had sufficient average coverage (Figure 3B), yet one of the LOD markers (total, 26) was an outlier at 2.53%, which may indicate slight PCR contamination. Further review of IMs showed that 4 of total 33 IMs were outliers and were flagged for either read bias or divergent coverage. Therefore, these four IMs were excluded from the calculation of percentage donor/recipient. Another two samples with LODs >0.2% (0.69% and 0.91%) had low overall coverage (Figure 3B), requiring use of a 100× threshold for chimerism analysis, which allows for higher background noise, conceding decreased sensitivity. For these samples, the analytical sensitivity level was adjusted on the basis of the high LODs. A 95% CI for each percentage donor chimerism result was calculated from all IMs. In the 155 post-transplant samples, it ranged from 0.01% to 2.48% (average, 0.46%; and median, 0.27%), which covered a range of percentage donor (0% to 100%) (Figure 4B). In the analysis software, the level of chimerism influences informative marker selection. In the high (>80) and low (<20) ranges of chimerism, dominant heterozygous markers are excluded to remove noise introduced by allele bias. In the range of 20% to 80% chimerism, all informative markers (homozygous and heterozygous markers) are used.
Figure 4.
Quality metrics for chimerism data analysis. A: Limit of detection (LOD; %) in each sample across total 155 post-transplant samples. Gray dashed line (0.2%) indicating sensitivity level for the chimerism test. Black dashed line (1.0%) indicating alert level when PCR contamination or wrong samples occurred. B: The 95% CI in each donor across total 155 post-transplant samples. Black dashed line represents alert level when percentage donor <20% or >80%; gray dashed line represents alert level when percentage donor is within 20% to 80%. C: Number of informative markers in each donor across total 155 post-transplant samples.
Number of Informative Markers
In the total 155 post-transplant samples, three types of transplantation were included: 59.4% (92) single donor, 25.8% (40) double donors, and 14.8% (23) triple donors (haplo-double cord blood transplants, for example). The number of IMs for each donor across the 155 post-transplant validation samples was variable, depending on number of donors, relationship of donor to recipient (related versus unrelated), and depth coverage. IMs per donor ranged from 2 to 120 (average, 32; median, 22) across all donors (Table 1 and Figure 4C), with more IMs per donor in single-donor transplants (range, 15 to 120; average, 59; median, 60) compared with double-donor (range, 4 to 45; average, 21; median, 19), and triple-donor transplants (range, 2 to 22; average, 8; median, 6). On the basis of relationship to patient, related donors had fewer IMs per donor (range, 2 to 66; average, 21; median, 21) than unrelated donors (range, 3 to 120; average, 34; median, 22). Two samples had less than three IMs, which was insufficient for analysis. These samples were from a haplo-double cord transplant (triple donor, 4 genomes). A total of 21.4% of all donors had 3 to 10 IMs and were seen in transplants with multiple donors. A total of 43.6% of donors had between 11 and 30 IMs from single or multiple donor transplants. A total of 11.1% of donors had between 31 and 50 IMs; and 22.6% of donors had >50 IMs, all from single-donor transplants. Number of IMs was also dependent on depth coverage of the sample, which can be directly related to DNA input, such as post-transplant sample from subsets with low cell frequency and lower depth had fewer informative markers.
Table 1.
Number of IMs per Donor
| Donors per transplant type by relationship to patient | IMs per donor, average (range) |
|---|---|
| Single (n = 92) | 59 (15–120) |
| Related (n = 23) | 32 (15–66) |
| Unrelated (n = 69) | 69 (27–120) |
| Double (n = 82) | 21 (4–45) |
| Related (n = 11) | 16 (4–21) |
| Unrelated (n = 71) | 22 (5–45) |
| Triple (n = 69) | 8 (2–22) |
| Related (n = 11) | 3 (2–7) |
| Unrelated (n = 58) | 9 (3–22) |
| All donors (n = 243) | 32 (2–120) |
IM, informative marker.
Linearity Study
The quantification range and linearity of the NGS chimerism assay was determined. In brief, DNA samples of two unrelated individuals (one as donor, another one as recipient) and DNA samples of three unrelated individuals (two of them as double donors equally mixed, third one as recipient) were serially diluted into percentage donor at 0%, 0.1%, 0.2%, 0.5%, 1%, 2%, 5%, 10%, 25%, 50%, and 100%. Three independent serial dilution preparations were then assayed in three independent runs. The results were compared between the experimental percentage of chimerism (observed percentage single donor or observed percentage double donors) and the expected values (expected percentage single donor or expected percentage double donors). The correlation coefficient was R2 = 0.9987 (single donor) and R2 = 0.9980 (double donor) for the entire range (0% to 100% donor chimerism) (Figure 5, A and C) and R2 = 0.9995 (single donor) and R2 = 0.9569 (double donor) for low values (0% to 5% donor) (Figure 5, B and D), indicating good linearity and accurate performance of the assay across the complete range (0% to 100%).
Figure 5.
Linearity study of next-generation sequencing–based chimerism assay, based on three independent runs. For single donor-recipient pair (two genomes), representative data shown in the range of 0% to 100% donor (A) and zoomed in on 0% to 5% donor (B). For double donor-recipient (three genomes), representative data shown in the range of 0% to 100% total donor (C) and zoomed in on 0% to 5% total donor (D).
Analytical Sensitivity
The analytical sensitivity of the assay and minimum DNA input was evaluated. Single-donor and multidonor (three genomes) chimeric samples, with chimerism ranging from 0% to 5% donor, were analyzed and calculated as described in Materials and Methods. The analytical sensitivity of the chimerism assay is 0.2% donor for single donor and 0.5% donors for double donors (<20% CV) (Supplemental Tables S2 and S3).
The effect of input DNA concentration was also evaluated. Artificial chimeric samples from either two unrelated individuals at 5% donor or three unrelated individuals at 10% of total donors were serially diluted into 0.5, 1, 2, 5, and 10 ng/μL. Each level of the DNA dilutions was tested in three independent runs. The NGS assay accurately and consistently detected 5% single donor or 10% total donor in chimeric samples down to 1 ng/μL of DNA (7.2% and 10.9% CV, respectively) (Supplemental Tables S4 and S5).
Furthermore, artificial DNA mixture at 0.2% donor was diluted into 1 and 5 ng/μL, respectively. Both DNA dilutions were tested in 19 independent NGS runs. The results demonstrated a consistent sensitivity level (0.2% donor) at 5 ng/μL (Figure 6A), but not at 1 ng/μL DNA (Figure 6B).
Figure 6.
Sensitivity study of chimerism testing. Artificial DNA mixture at 0.2% donor was repeated for 19 individual next-generation sequencing (NGS) runs by 1 ng/μL (A) and 5 ng/μL (B) of DNA concentration. LOD, limit of detection.
Reproducibility
Precision (within run) was assessed by including 13 clinical samples (including single- and double-donor transplants) over a range of mixed chimerism percentages in triplicate, indexed using different sample barcodes in a single run. The DNA concentrations across the 13 samples ranged from 1.29 to 88.8 ng/μL. Chimerism results were highly reproducible with either single or double donors (Figure 7, A and B), and CVs ranged from 0.4% to 12.7%. The interrun reproducibility (between runs) was also assessed by including the above 13 clinical samples in three independent assay runs (from original DNA through sequencing and data analysis). The consistent detection of donor percentage across three different assay runs showed a high degree of interrun reproducibility on the 13 clinical samples (Figure 7, C and D), and CVs ranged from 0.5% to 18.4%. Highest CV was seen at lower percentage chimerism (<1% donor chimerism). The results indicated the assay had an excellent intrarun and interrun reproducibility over a broad range of chimerism results (percentage donor).
Figure 7.
Reproducibility study of chimerism testing in three independent runs. Chimerism results for intrarun (within-run) reproducibility study with single donor (A) or double donors (B) and for interrun (between-run) reproducibility study with single donor (C) or double donors (D).
Accuracy Study
To assess the accuracy of the NGS chimerism assay, artificial chimeric mixtures (n = 148) of donors (including single, double, and triple donors) and recipient DNA were prepared, ranging from 0% to 100% donor. The percentage donor, as measured by the NGS method, had excellent correlation with expected values (Figure 8). The slope of the linear regression line fit was 1.002; R2 value was 0.9966 across the entire range of percentage donor (Figure 8A), and in the range of 0% to 5% donor, R2 value was 0.927 (Figure 8B). The percentage donor by NGS was close to the expected values, and maximum absolute difference was 7.7 (at 51% donor) (Figure 8C). Overall, the quantification of the NGS chimerism assay was reliable and accurate over a complete range of measurements (0% to 100% donor).
Figure 8.
Accuracy study (trueness) of next-generation sequencing (NGS)–based chimerism assay. Artificial chimeric mixtures of donor and recipient DNA, ranging from 0% to 100% donor, were tested using NGS-based chimerism assay (A); zoomed in on 0% to 5% donor (B); and zoomed in on 40% to 60% donor (C). n = 148 (A).
The NGS method was compared to STR-PCR by testing 84 samples, including single-donor (n = 40) or double-donor (n = 30) transplants and proficiency samples (n = 14), ranging from 0% to 100% total donor. The NGS method demonstrated high concordance to STR-PCR method in entire range of donor chimerism (R2 = 0.9991) (Figure 9A). Below 5% donor, correlation of NGS results with those of STR-PCR was lower (R2 = 0.4474), noting that this is below the sensitivity level of STR-PCR (Figure 9B). Furthermore, Bland-Altman analysis (Figure 9C) showed good agreement between NGS and STR-PCR over the entire range of percentage donor. Additionally, minor recipient or minor donor chimerism (<5%) was consistently detected by NGS method with higher sensitivity, but not by STR-PCR.
Figure 9.
A and B: Correlation of percentage donor chimerism results between next-generation sequencing (NGS) and short-tandem-repeat (STR) PCR based on testing of patient samples in the range of 0% to 100% donor (A) and zoomed in on 0% to 5% donor (B). C: Bland-Altman plot of difference between NGS and STR-PCR results versus average of results; dotted lines indicate 95% limits of agreement.
The NGS method was compared with qPCR by testing 85 samples, ranging from 0% to 100% donor (two genomes). The methods demonstrated high concordance. The slope of the linear regression line fit was 1.005; R2 value was 0.9908 across the entire range of donor chimerism (Figure 10A). Correlation of NGS results with those of qPCR was lower (R2 = 0.4946) in the range of 40% to 60% donor chimerism (Figure 10C), but higher below 5% donor chimerism (R2 = 0.8507) (Figure 10B). The greatest differences between the two methods were seen in the midrange (40% to 60% donor) because of less accuracy by qPCR method; based on Bland-Altman analysis, the average difference was 0.648, and limits of agreement were −5.9 to 7.2 (Figure 10D).
Figure 10.
A–C: Correlation of percentage donor chimerism results between next-generation sequencing (NGS) and real-time quantitative PCR (qPCR) based on testing of patient samples in the range of 0% to 100% donor (A); zoomed in on 0% to 5% donor (B); and zoomed in on 40% to 60% donor (C). D: Bland-Altman plot of difference between qPCR and NGS results versus average of results; dotted lines indicate 95% limits of agreement. Largest differences between NGS and qPCR occurred between 20% to 75% donor chimerism.
Discussion
The aim of this study is to validate the performance of an NGS-based chimerism assay and develop a technical framework for validating high-sensitivity chimerism assays used in a clinical setting to monitor patients after alloHCT with single or multiple donors and to detect microchimerism. We include considerations for assessing quality metrics.
The multidonor library preparation commercial kit (ScisGo Chimerism Multi-Donor Assay) and resulting NGS sequencing runs performed consistently, allowing us to establish minimum quality metric parameters and acceptable ranges for performance at the sequencing run level, including NGS library concentration, cluster density, cluster passing filter rate, percentage >Q30 (Figures 1 and 2 and Supplemental Table S1), and at the sample level, including average coverage, 95% CI, and LOD (Figures 3 and 4). Parameters that were specifically associated with quality included average coverage per amplicon pool per sample, LOD%, and 95% CI (Figures 3 and 4). Establishing acceptability criteria to ensure quality performance over time can be challenging for NGS-based assays. When a validation involves such a high-complexity assay with high precision, typically the data set is based on a limited number of initial runs performed by a limited number of technical staff, leading to less variation than seen in routine operations. This should be considered when adopting quality metrics, with continued monitoring once put into routine use.
As with all quantitative, highly sensitive NGS-based assays, depth of coverage is a critical determinant of sensitivity. As per manufacturer's recommendations, the default threshold for minimum coverage of an IM was set at 1000×. A threshold of 100× was validated with accurate results, albeit reduced sensitivity and increased background noise. Low coverage can be associated with low DNA concentration, often seen in samples enriched for a low-frequency cell lineage (eg, B-cell subset enrichment), but also occasionally seen when using buccal swab samples as source DNA for identity testing with poor or inadequate collection. Validating a lower threshold that achieves accurate results is important if these sample types will be used for identity testing in chimerism analysis.
The commercial assay used in this study (ScisGo Chimerism Multi-Donor Assay) demonstrated high sensitivity and accuracy compared with STR and qPCR across a range of chimerism in the setting of single donor (ie, two genomes) and double donor (ie, three genomes, attributable to double umbilical cord transplants or secondary transplant). The analytical sensitivity was 0.2% chimerism for single donor and 0.5% chimerism for double donors (CV <20%), with reduced sensitivity resulting from fewer informative markers among three genomes. This is comparable to analytical sensitivities achieved by qPCR- and droplet digital PCR–based methods,4,12, 13, 14, 15 and greater than sensitivities achieved by STR-PCR. Assays that can accurately detect <1% chimerism (termed microchimerism) are generally classified as high-sensitivity assays. Although there are published reports of association of microchimerism with disease relapse,15,16 there is no overall consensus of its clinical utility in post-transplant monitoring and/or risk mitigation of relapse (see the study by Haugaard et al17).
Pre-analytical factors that affect sensitivity include not only DNA concentration, but also percentage chimerism (ie, accuracy of microchimerism detection/quantitation is limited in samples with low DNA yield, ie, patients with low white blood cell counts).9 High accuracy and precision at 0.2% chimerism were achievable when DNA input concentration was at least 5 ng/μL but was less reliable when DNA input concentration was <1 ng/μL. As clinicians frequently send for chimerism testing in patients after HCT with decreasing cell counts, validation should include samples of lower DNA concentration and lower percentage chimerism, as this will affect sensitivity. DNA concentrating methods (such as silica spin columns) can help to address these clinically challenging samples.
When validating a highly sensitive assay and verifying the accurate, precise detection of microchimerism, it is important to consider whether recipient microchimerism or donor microchimerism is the purpose for testing. In contrast to alloHCT, where increasing recipient chimerism may be an early marker of relapse, detection of increasing donor-derived cell-free DNA may indicate allograft rejection in solid organ transplants.18,19 The distinction is that informativity algorithms depend on the percentage chimerism of the minor genome. Other software may assign informativity based on recipient genome.20,21 In either case, each end of the percentage chimerism spectrum can be based on different markers. Understanding the algorithmic conditions/decisions of the pipeline/software is critical.
Likewise, a thorough performance assessment should include samples that test all clinically relevant ranges of mixed chimerism. Samples should approximate clinically important scenarios, wherein the recipient is the minor genome, such as rising/increasing recipient chimerism. In some cases, re-emergence of recipient chimerism merely reflects autologous hematopoiesis, not malignant relapse. Increasing recipient chimerism, in comparison to a single time point of mixed chimerism, may be more predictive of relapse.16,22,23 Accurate and precise quantitation is critical for an assay used to monitor chimerism dynamics. The validation should address this.
In this patient population, chimerism analysis involving three or more genomes is frequently required, due to patients undergoing double cord blood transplants, or retransplantation. Depending on the transplantation strategy (conditioning regimen and graft manipulation/graft-versus-host disease prophylaxis), there is a period wherein more than two genomes are detectable. The capability of the assay and analysis software to distinguish genomes and accurately calculate percentage chimerism in these cases should be established. In this study, in the setting of three genomic contributors, analytical sensitivity was 0.5% compared with 0.2% in the single-donor setting.
Chimerism analysis of NGS data requires a bioinformatics pipeline and is seemingly more complex than that of STR or qPCR. However, the steps of analysis are the same: marker classification (homozygous/heterozygous), informative marker determination for the donor/recipient pair, and then percentage chimerism calculation. Marker informativity is influenced by the method used and genetic marker type (for instance, stutter peaks in STR, allele bias in SNPs). These must be considered in the algorithm used for marker classification and determination of informativity. For NGS-based methods involving biallelic SNPs and InDels, allele amplification bias is an important factor. Thresholds for homozygosity classification need to take this into account. Allowing for adjustment of thresholds in the software on a per-sample basis can allow for scenarios needing less stringent settings, such as post-transplant buccal swabs as recipient-identity samples, used when pre-HCT recipient DNA is not available. It is inevitable that replacement of STR by NGS-based chimerism methods will necessitate repeating identity testing by NGS for all patients, including those whose transplants occurred several years ago and pre-HCT DNA may be exhausted (A.G.B., F.Y., D.G., W.N., unpublished data).
Regarding informative markers, more than three IMs were identified in all cases of single-donor and double-donor transplants, as required by some accrediting agencies, such as the College of American Pathologists. However, in cases with more than three genomes, such as triple-donor transplants, only two cases that had less than three IMs were identified, both of which were haploidentical double cord blood transplants. Current standards do not have provisions for testing chimerism in the setting of multiple genetic contributors because most commonly used methods at present, such as STR-PCR and qPCR, are not reliable in accurately assessing chimerism in this setting. Revisions of accreditation standards and working group recommendations24,25 pertinent to NGS-based chimerism analysis, including provisions for testing in the setting of multiple donors as well as other quality metrics investigated in this study, are warranted.
Another important consideration for NGS data analysis is the user interface. It should provide a clear, easily accessible organized view of quality metrics in addition to chimerism results. In the software used in this study, a heat map of marker divergent coverage per sample and bar graph of average coverage per amplicon pool per sample provide easy-to-see overview of sequencing run quality metrics (Supplemental Figure S1). The user interface should have a clearly organized layout for review of metrics (depth of coverage, read bias) for informative markers, allowing for manual deselection of outliers (Supplemental Figure S2). Additionally, software flags should be set to warn in samples with high noise (based on LOD, calculated from homozygous, homologous markers) and in markers with divergent coverage. The user interface should also include a view of a patient's previous results, so that trends/changes can be observed.
Conclusion
The NGS-based multidonor chimerism assay had high accuracy and precision across a full range of chimerism with improved analytical sensitivity compared with STR and comparable sensitivity to qPCR (LOD, 0.2% for single-donor transplants, 0.5% for multiple-donor transplants). Beyond a rigorous performance assessment of accuracy, sensitivity, linearity, and precision, a validation for routine clinical use of NGS-based chimerism assays should evaluate the impact of various factors, such as DNA input, marker informativity, number of genomes expected, and level of chimerism, in the sample. Finally, the clinical purpose for testing, such as engraftment monitoring (including early mixed chimerism) and/or accurate detection of microchimerism, should drive the design of the validation study and acceptance criteria.
Author Contributions
A.G.B. and F.Y. are the guarantors of this work and, as such, had full access to all data in the study and take responsibility for the integrity of the data and accuracy of the data analysis. A.G.B. conceptualized the study, performed validation, developed methods, curated, analyzed, and visualized data, performed investigations, wrote, reviewed, and edited the manuscript, administered the project, acquired funding, obtained resources, and supervised the study; W.N. performed investigations, reviewed and edited the manuscript, and obtained resources; D.G. performed investigations, reviewed and edited the manuscript, and obtained resources; M.A. conceptualized the study, developed methods, performed investigations, curated data, reviewed and edited the manuscript, and obtained resources; and F.Y. conceptualized the study, developed methods, performed validation, curated, analyzed, and visualized data, performed investigations, wrote, reviewed, and edited the manuscript, administered the project, obtained resources, and supervised the study. All authors revised the manuscript and approved the final version of the manuscript.
Disclosure Statement
W.N. and D.G. own stock in and are employees of Scisco Genetics Inc. M.A. has received compensation as a member of the scientific advisory board for CareDx, Inc.
Footnotes
Supported in part through the NIH/National Cancer Institute Cancer Center Support grant P30 CA008748 (A.G.B. and F.Y.).
Supplemental material for this article can be found at http://doi.org/10.1016/j.jmoldx.2024.07.003.
Contributor Information
Amanda G. Blouin, Email: blouina@mskcc.org.
Fei Ye, Email: yef@mskcc.org.
Supplemental Data
Software user interface quality metric graphs. A: Heat map, example: read bias (R1 bias <0.3, green bars; R1 bias >0.7, blue bars) and divergent coverage (red bars) for each marker per sample in a next-generation sequencing (NGS) run. In this example, a negative control sample with contamination shows read bias and divergent coverage for several markers (asterisk). B: Coverage plot, example: total reads per amplicon (AMP) pool per sample in an NGS run. In this example, a negative control sample with contamination shows amplification in multiple amplicon pools (asterisk).
Representative chimerism results in ScisCloud Software user interface. A: Sample identifier (ID), FASTQ file date and time, analysis settings including threshold 1000×, Q: 20 (the minimum Illumina basecall quality score 20), read bias 40%; and several tools for filtering, editing parameters, recomputing, and viewing case report. B: Case result displays the calculated percentage chimerism for the patient (in this example, 79.61%) and donor (in this example, 20.39%). In addition, data analysis quality metrics show informative markers (N), depth with SD, 95% CI, and limit of detection (LOD). C: With the informative marker list open, each marker list item shows the marker ID, (eg, sgi00019) and the genotypes of the donor (D) and the recipient (P) in the sample. For single-nucleotide polymorphism markers, the genotypes are written in terms of the four nucleotides, A, C, G, and T. For insertion/deletion markers, the genotypes are written as S for single-nucleotide insertion, M for multiple-nucleotide insertion, or N for none, which represents a deletion of either single base (when the other type is S) or multiple bases (when the other type is M). In addition, each informative marker shows chromosomal location, read bias, depth, and percentage of chimerism results.
References
- 1.Kanate A.S., Majhail N.S., Savani B.N., Bredeson C., Champlin R.E., Crawford S., Giralt S.A., LeMaistre C.F., Marks D.I., Omel J.L., Orchard P.J., Palmer J., Saber W., Veys P.A., Carpenter P.A., Hamadani M. Indications for hematopoietic cell transplantation and immune effector cell therapy: guidelines from the American Society for Transplantation and Cellular Therapy. Biol Blood Marrow Transplant. 2020;26:1247–1256. doi: 10.1016/j.bbmt.2020.03.002. [DOI] [PubMed] [Google Scholar]
- 2.Blouin A.G., Askar M. Chimerism analysis for clinicians: a review of the literature and worldwide practices. Bone Marrow Transplant. 2022;57:347–359. doi: 10.1038/s41409-022-01579-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Blouin A.G., Ye F., Williams J., Askar M. A practical guide to chimerism analysis: review of the literature and testing practices worldwide. Hum Immunol. 2021;82:838–849. doi: 10.1016/j.humimm.2021.07.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Alizadeh M., Bernard M., Danic B., Dauriac C., Birebent B., Lapart C., Lamy T., Le Prise P.Y., Beauplet A., Bories D., Semana G., Quelvennec E. Quantitative assessment of hematopoietic chimerism after bone marrow transplantation by real-time quantitative polymerase chain reaction. Blood. 2002;99:4618–4625. doi: 10.1182/blood.v99.12.4618. [DOI] [PubMed] [Google Scholar]
- 5.Tyler J., Kumer L., Fisher C., Casey H., Shike H. Personalized chimerism test that uses selection of short tandem repeat or quantitative PCR depending on patient's chimerism status. J Mol Diagn. 2019;21:483–490. doi: 10.1016/j.jmoldx.2019.01.007. [DOI] [PubMed] [Google Scholar]
- 6.Hindson B.J., Ness K.D., Masquelier D.A., Belgrader P., Heredia N.J., Makarewicz A.J., et al. High-throughput droplet digital PCR system for absolute quantitation of DNA copy number. Anal Chem. 2011;83:8604–8610. doi: 10.1021/ac202028g. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Mika T., Baraniskin A., Ladigan S., Wulf G., Dierks S., Haase D., Schork K., Turewicz M., Eisenacher M., Schmiegel W., Schroers R., Klein-Scory S. Digital droplet PCR-based chimerism analysis for monitoring of hematopoietic engraftment after allogeneic stem cell transplantation. Int J Lab Hematol. 2019;41:615–621. doi: 10.1111/ijlh.13073. [DOI] [PubMed] [Google Scholar]
- 8.Pedini P., Kouba N., Riquier M., Simon S., Basire A., Fina F., Galambrun C., Chiaroni J., Picard C. Droplet digital PCR: a new technology for detection and quantification of chimerism after allogenic hematopoietic stem cell transplantation. Biomed J Sci Tech Res. 2019;13:10065–10068. [Google Scholar]
- 9.Pettersson L., Vezzi F., Vonlanthen S., Alwegren K., Hedrum A., Hauzenberger D. Development and performance of a next generation sequencing (NGS) assay for monitoring of mixed chimerism. Clin Chim Acta. 2021;512:40–48. doi: 10.1016/j.cca.2020.10.034. [DOI] [PubMed] [Google Scholar]
- 10.Vynck M., Nollet F., Sibbens L., Lievens B., Denys A., Cauwelier B., Devos H. Performance assessment of the devyser high-throughput sequencing-based assay for chimerism monitoring in patients after allogeneic hematopoietic stem cell transplantation. J Mol Diagn. 2021;23:1116–1126. doi: 10.1016/j.jmoldx.2021.05.017. [DOI] [PubMed] [Google Scholar]
- 11.Kim J., Hwang I.S., Shin S., Choi J.R., Lee S.T. SNP-based next-generation sequencing reveals low-level mixed chimerism after allogeneic hematopoietic stem cell transplantation. Ann Hematol. 2018;97:1731–1734. doi: 10.1007/s00277-018-3325-6. [DOI] [PubMed] [Google Scholar]
- 12.Pedini P., Cherouat N., Basire A., Simon S., Budon L., Pourtein M., Grondin S., Moskovtchenko P., Chiaroni J., Michel G., Frassati C., Picard C. Evaluation of next-generation sequencing and crystal digital PCR for chimerism monitoring of post-allogeneic hematopoietic stem cell transplantation. Transplant Cell Ther. 2021;27:89.e1–89.e10. doi: 10.1016/j.bbmt.2020.09.023. [DOI] [PubMed] [Google Scholar]
- 13.Fortschegger M., Preuner S., Printz D., Poetsch A.R., Geyeregger R., Pichler H., Lawitschka A., Lion T. Detection and monitoring of lineage-specific chimerism by digital droplet PCR-based testing of deletion/insertion polymorphisms. Biol Blood Marrow Transplant. 2020;26:1218–1224. doi: 10.1016/j.bbmt.2020.02.016. [DOI] [PubMed] [Google Scholar]
- 14.Kim S.Y., Jeong M.H., Park N., Ra E., Park H., Seo S.H., Kim J.Y., Seong M.W., Park S.S. Chimerism monitoring after allogeneic hematopoietic stem cell transplantation using quantitative real-time PCR of biallelic insertion/deletion polymorphisms. J Mol Diagn. 2014;16:679–688. doi: 10.1016/j.jmoldx.2014.06.005. [DOI] [PubMed] [Google Scholar]
- 15.Valero-Garcia J., González-Espinosa M.D.C., Barrios M., Carmona-Antoñanzas G., García-Planells J., Ruiz-Lafora C., Fuentes-Gálvez A., Jiménez-Velasco A. Earlier relapse detection after allogeneic haematopoietic stem cell transplantation by chimerism assays: digital PCR versus quantitative real-time PCR of insertion/deletion polymorphisms. PLoS One. 2019;14 doi: 10.1371/journal.pone.0212708. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Sellmann L., Rabe K., Bünting I., Dammann E., Göhring G., Ganser A., Stadler M., Weissinger E.M., Hambach L. Diagnostic value of highly-sensitive chimerism analysis after allogeneic stem cell transplantation. Bone Marrow Transplant. 2018;53:1457–1465. doi: 10.1038/s41409-018-0176-7. [DOI] [PubMed] [Google Scholar]
- 17.Haugaard A.K., Kofoed J., Masmas T.N., Madsen H.O., Marquart H.V., Heilmann C., Müller K.G., Ifversen M. Is microchimerism a sign of imminent disease recurrence after allogeneic hematopoietic stem cell transplantation? a systematic review of the literature. Blood Rev. 2020;44 doi: 10.1016/j.blre.2020.100673. [DOI] [PubMed] [Google Scholar]
- 18.Agbor-Enoh S., Shah P., Tunc I., Hsu S., Russell S., Feller E., Shah K., Rodrigo M.E., Najjar S.S., Kong H., Pirooznia M., Fideli U., Bikineyeva A., Marishta A., Bhatti K., Yang Y., Mutebi C., Yu K., Kyoo Jang M., Marboe C., Berry G.J., Valantine H.A. Cell-free DNA to detect heart allograft acute rejection. Circulation. 2021;143:1184–1197. doi: 10.1161/CIRCULATIONAHA.120.049098. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Huang E., Sethi S., Peng A., Najjar R., Mirocha J., Haas M., Vo A., Jordan S.C. Early clinical experience using donor-derived cell-free DNA to detect rejection in kidney transplant recipients. Am J Transplant. 2019;19:1663–1670. doi: 10.1111/ajt.15289. [DOI] [PubMed] [Google Scholar]
- 20.Liacini A., Tripathi G., McCollick A., Gravante C., Abdelmessieh P., Shestovska Y., Mathew L., Geier S. Chimerism testing by next generation sequencing for detection of engraftment and early disease relapse in allogeneic hematopoietic cell transplantation and an overview of NGS chimerism studies. Int J Mol Sci. 2023;24 doi: 10.3390/ijms241411814. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Kakodkar P., Zhao Y., Pan H., Wu F., Pearce T., Webster D., Elemary M., Sabry W., Kwan L., Pelzer L., Bosch M., Sherwood K.R., Lan J., Tran J., Liwski R., Keown P., Mostafa A. Validation of next-generation sequencing-based chimerism testing for accurate detection and monitoring of engraftment in hematopoietic stem cell transplantation. Front Genet. 2023;14 doi: 10.3389/fgene.2023.1282947. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Bader P., Kreyenberg H., Hoelle W., Dueckers G., Handgretinger R., Lang P., Kremens B., Dilloo D., Sykora K.W., Schrappe M., Niemeyer C., Von Stackelberg A., Gruhn B., Henze G., Greil J., Niethammer D., Dietz K., Beck J.F., Klingebiel T. Increasing mixed chimerism is an important prognostic factor for unfavorable outcome in children with acute lymphoblastic leukemia after allogeneic stem-cell transplantation: possible role for pre-emptive immunotherapy? J Clin Oncol. 2004;22:1696–1705. doi: 10.1200/JCO.2004.05.198. [DOI] [PubMed] [Google Scholar]
- 23.Jacque N., Nguyen S., Golmard J.L., Uzunov M., Garnier A., Leblond V., Vernant J.P., Bories D., Dhédin N. Chimerism analysis in peripheral blood using indel quantitative real-time PCR is a useful tool to predict post-transplant relapse in acute leukemia. Bone Marrow Transplant. 2015;50:259–265. doi: 10.1038/bmt.2014.254. [DOI] [PubMed] [Google Scholar]
- 24.Lion T., Watzinger F., Preuner S., Kreyenberg H., Tilanus M., de Weger R., van Loon J., de Vries L., Cave H., Acquaviva C., Lawler M., Crampe M., Serra A., Saglio B., Colnaghi F., Biondi A., van Dongen J.J., van der Burg M., Gonzalez M., Alcoceba M., Barbany G., Hermanson M., Roosnek E., Steward C., Harvey J., Frommlet F., Bader P. The EuroChimerism concept for a standardized approach to chimerism analysis after allogeneic stem cell transplantation. Leukemia. 2012;26:1821–1828. doi: 10.1038/leu.2012.66. [DOI] [PubMed] [Google Scholar]
- 25.Clark J.R., Scott S.D., Jack A.L., Lee H., Mason J., Carter G.I., Pearce L., Jackson T., Clouston H., Sproul A., Keen L., Molloy K., Folarin N., Whitby L., Snowden J.A., Reilly J.T., Barnett D., United Kingdom National External Quality Assessment Service for Leucocyte Immunophenotyping Chimerism Working Group Monitoring of chimerism following allogeneic haematopoietic stem cell transplantation (HSCT): technical recommendations for the use of short tandem repeat (STR) based techniques, on behalf of the United Kingdom National External Quality Assessment Service for Leucocyte Immunophenotyping Chimerism Working Group. Br J Haematol. 2015;168:26–37. doi: 10.1111/bjh.13073. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Software user interface quality metric graphs. A: Heat map, example: read bias (R1 bias <0.3, green bars; R1 bias >0.7, blue bars) and divergent coverage (red bars) for each marker per sample in a next-generation sequencing (NGS) run. In this example, a negative control sample with contamination shows read bias and divergent coverage for several markers (asterisk). B: Coverage plot, example: total reads per amplicon (AMP) pool per sample in an NGS run. In this example, a negative control sample with contamination shows amplification in multiple amplicon pools (asterisk).
Representative chimerism results in ScisCloud Software user interface. A: Sample identifier (ID), FASTQ file date and time, analysis settings including threshold 1000×, Q: 20 (the minimum Illumina basecall quality score 20), read bias 40%; and several tools for filtering, editing parameters, recomputing, and viewing case report. B: Case result displays the calculated percentage chimerism for the patient (in this example, 79.61%) and donor (in this example, 20.39%). In addition, data analysis quality metrics show informative markers (N), depth with SD, 95% CI, and limit of detection (LOD). C: With the informative marker list open, each marker list item shows the marker ID, (eg, sgi00019) and the genotypes of the donor (D) and the recipient (P) in the sample. For single-nucleotide polymorphism markers, the genotypes are written in terms of the four nucleotides, A, C, G, and T. For insertion/deletion markers, the genotypes are written as S for single-nucleotide insertion, M for multiple-nucleotide insertion, or N for none, which represents a deletion of either single base (when the other type is S) or multiple bases (when the other type is M). In addition, each informative marker shows chromosomal location, read bias, depth, and percentage of chimerism results.