Patients were recruited from Sahlgrenska University Hospital, Gothenburg, Sweden as a part of the prospective BIODRAFT-trial (NCT03477383). All patients or caregivers provided informed consent and the study was approved by the institutional review board (no 014-16). Patients were eligible if they underwent HTx between 2016 and 2018. Blood samples were drawn coincidental to endomyocardial biopsies (EMB) during the first year after HTx. Access to donor blood samples was granted. Clinical data were extracted from medical records.

Blood was collected immediately before catheterization in 10 mL Cell-Free DNA^® BCT (Streck, La Vista NE, USA). The samples were agitated for 10 s, shipped, and stored at room temperature for no longer than 7 days before plasma isolation. Plasma was separated from cells by centrifugation at 2,000 g for 15 min. The plasma fraction was transferred to a new tube followed by a second centrifugation at 16,000 g for 10 min. Centrifugations were performed at 20°C. The plasma was transferred to a collection tube and frozen at −80°C. Genomic DNA was prepared from the leukocyte fraction of the blood samples using the DNeasy Blood & Tissue Kit (Qiagen GmbH, Hilden, Germany). cfDNA was extracted using the QIAamp^® Circulating Nucleic Acid Kit (Qiagen) on the QIAvac 24 Plus vacuum manifold (Qiagen). cfDNA was eluted in 20 μL AVE buffer per ml plasma. Samples were stored at −20°C until analysis. Concentrations of cfDNA were quantified with the Qubit^® 3.0 Fluorometer (Thermo Fisher Scientific, Waltham MA, USA), fragment sizes were analyzed with the 4200 TapeStation (Agilent technologies, Santa Clara CA, USA).

35 previously published SNP assays [29] were selected for this study. For detection of the Y-chromosome the Human Y-Chromosome Specific Assay (TATAA Biocenter, Gothenburg, Sweden) targeting the TSPY1 gene was used. Probe and primer sets (Integrated DNA Technologies Inc., Carolville, IA, USA) were designed using HEX (Hexachlorofluorescin) and FAM (Carboxyfluorescin).

Genomic DNA extracted from white blood cells was used. The donor was investigated with respect to the homozygous alleles found in the recipient. In sex mismatched HTx (female recipient, male donor), the Y-chromosome was used.

Preamplification using pooled primers for all 35 SNP and the Y-chromosome was conducted on cfDNA corresponding to 2 mL of patient plasma. 40 μL cfDNA, 45 μL Q5 Hot Start High-Fidelity 2x Master Mix (New England BioLabs, Ipswich, MA, USA), 3.6 μL primerpool (0.04 μmol) and 1.4 μL water were used in a total volume of 90 μL. Amplification was applied on a T100 Thermal Cycler (Bio-Rad, Hercules, CA, USA): 98°C for 3 min, followed by 10 cycles (98°C for 20 s, 63°C for 3 min and 72°C for 30 s). After the final extended (10 min) elongation step, the samples were stored at −20°C until analysis.

dPCR on non-amplified cfDNA was conducted on all patient samples using a targeted SNP assay. 10 μL of eluted patient cfDNA, corresponding to 0.5 mL of blood plasma, were used with 11 μL ddPCR Supermix for Probes (No dUTP) (Bio-Rad), 0.5 μL primer/probe-mix (900 nmol of each primer and 250 nmol of each probe) and 0.5 μL water in a total volume of 22 µL. Negative control with a no template control (NTC) with water was included as well as a positive control with a sample of genomic DNA (gDNA). Amplification was applied on a T100 Thermal Cycler (Bio-Rad): 95°C for 10 min, followed by 40 cycles (95°C for 30 s, 59°C/61°C for 1 min), 98°C for 10 min. Analysis was performed using the QX200 AutoDG Droplet Digital PCR System (Bio-Rad). Using negative and positive controls for cluster detection, manual fluorescent thresholds were placed. Analysis was conducted using QuantaSoft Analysis Pro v1.0 (Bio-Rad) to calculate absolute droplet counts as well as target DNA-concentration for rd-cfDNA. Target concentrations were expressed as copies per ml plasma.

dPCR on the preamplified cfDNA (PA-dPCR) was conducted using dilutions, based on the target concentrations of non-amplified cfDNA. All identified SNP were used. 10µL amplified cfDNA was diluted in purified water to acquire the desired concentration. All experiments were conducted as triplets, according to the protocol in 2.6. Target concentrations were expressed as copies per µl PCR-reaction, DF was calculated as dd-cfDNA/(rd-cfDNA + dd-cfDNA).

The initial dPCR is conducted on known concentrations directly corresponding to the amounts of isolated plasma (3 mL plasma is eluted in 60 μL, 4 mL plasma in 80 μL), see Supplementary Figure S1). This allows for the determination of absolute copy numbers for the recipient (cp/mL plasma). Using the donor fraction from the PA-PCR, by multiplying with absolute copy numbers for the recipient, total copy number for the donor is calculated (cp/mL plasma).

The efficiency of target-specific preamplification was determined using a cfDNA control from normal donor plasma, in the range of 0.5–32 ng. Preamplification was performed in 30 μL reactions, using Q5 Hot Start High-Fidelity 1x Master Mix, 40 nM of each primer and template cfDNA at seven different concentrations (32, 16, 8, 4, 2, 1, and 0.5 ng/μL) in triplicate. The same amplification protocol was used as above. After the final extended (10 min) elongation step, the samples were immediately frozen on dry ice, slowly thawed on ice, diluted 1:20 in 1x TE buffer (ThermoFisher Scientific, Waltham, MA, USA) and stored at −20°C until analysis. qPCR (real-time quantitative PCR) was performed in 10 μL reactions utilizing 1x TATAA SYBR GrandMaster Mix (TATAA Biocenter) with 400 nM of each primer (Integrated DNA Technologies) and 2 μL diluted preamplification product as template. qPCR was performed in triplicates using the CFX384 Touch Real-Time PCR Detection System (Bio-Rad): 95°C for 10 min, followed by 50 cycles of amplification (95°C for 15 s and 60°C for 1 min). Melting curve analysis was performed in the range of 65°C to 95°C, 0.5°C per 5 s increments. Cycles of quantification (Cq) values were determined by the second derivative maximum method. Limit of blank (LOB), limit of detection (LOD) and limit of quantification (LOQ) were defined according to Armbruster et al [33] and determined as published by our group [34]: LOB 0.016% DF, LOD 0.055% DF, with LOQ = LOD.

Quantification of cfDNA in the dPCR-experiments is embedded in the software Quantasoft Pro (Bio-Rad) [35]. Continuous data are reported as means with standard deviation (SD) or as medians with interquartile range (IQR, q1-q3). Range is also reported when appropriate. In the boxplots, the horizontal line represents the median value, the top and bottom of each box show the upper and lower limit of the IQR, and the whiskers represent the range. Receiver-operating characteristic (ROC) curves were used to assess the sensitivity and specificity of DF and dd-cfDNA to predict treated rejection. Correlations (Pearson and Spearman) report the correlation coefficient r, CI₉₅ of r and R². Early (7–14 days) and late (>14 days) samples were compared using a linear mixed effects (LME) model. Other groups were also compared using a LME model, with adjustment for days from transplantation. A two-tailed p-value of <0.05 was considered statistically significant. The LME models were computed using R Statistical Software (v4.3.1) [36]. All other statistical analysis was performed using the GraphPad Prism Software (GraphPad Software Inc., version 10.0.2, GraphPad Software, Boston, Massachusetts USA).

The 52 patients generated 580 venous samples during their 1 year follow-up. 23 samples were excluded (hemolytic sample, too little plasma yield, high technical error rate, too few droplets generated). The population consisted of 41 adults and 11 children who were aged 1–68 years, (median 52.5), 69% of patients were male, and the median BMI was 24.9. The indication for transplant was dilated cardiomyopathy in 58%, and ventricular assist device was used in 29%. Median donor age was 49.5 years, median donor BMI was 23.8, median donor heart ischemic time was 183.5 min. More detailed patient and donor specifications can be seen in the Supplementary Tables S1, S2. Median time to first biopsy was 10.5 days (IQR 9–12). Blood samples taken during the first 14 days after HTx (n = 48) were analyzed by dPCR but excluded from general statistical analysis except for when otherwise stated.

The detailed workflow and calculation of cfDNA-levels from recipient, donor and DF can be seen in Supplementary Figure S1. After sample collection, cfDNA was extracted within a time frame of 7 days [37]. cfDNA-analysis was conducted on bundled samples after the patients had left the study. Results were available with 48 h. A mean of 4 mL plasma was obtained from each sample (SD 0.57, range 1.25–5.70). The median for the fluorometrically determined cfDNA-concentration was 34.20 ng/mL plasma (range 5.16–2,856, IQR 20.45–57.60). The 509 samples showed a median of 9,905 copies/mL plasma rd-cfDNA concentration (range 1,245-219,754, IQR 5,137-17,596), the median for dd-cfDNA was 9.31 copies/mL plasma (range 0.43–348.5, IQR 5.06–21.71, mean 18.9). Donor fraction showed a median of 0.09% (range 0.003–3.34, IQR 0.05–0.21). See Supplementary Figure S2.

Of the 557 samples, 48 samples were excluded as early samples (<15 days) and 18 due to reasons that impaired interpretation (malignancy, severe infections). This resulted in 491 samples that were suitable for evaluation. Acute reaction was seen in 13 biopsy-matched samples from 7 patients, see Figure 1. One sample showed rd-cfDNA levels of 220,000 copies/mL, >20 times the median (see Study Protocol and General Results), falsely lowering the DF, and was thus excluded from analysis. Median levels were significantly higher during rejection episodes: absolute levels of dd-cfDNA showed a median of 23 copies/mL (IQR 10.6–49.8) during rejection compared to 8.8 (IQR 4.7–19.8) during quiescence (p = 0.0001). Using a cut-off of 7.5 copies/mL, sensitivity was 92% and specificity was 43% (AUC 0.75; 95% CI 0.63–0.87, PPV = 0.04, NPV = 0.99). DF was also elevated in rejection, with a median level was 0.19 (IQR 0.13–0.56) compared to 0.08 (IQR 0.05–0.19) during quiescence (p < 0.0001). Using a cut-off of 0.1%, sensitivity was 92% and specificity was 56% (AUC 0.78, 95% CI 0.68–0.88, PPV = 0.05, NPV = 0.99). See Figure 2 and Supplementary Figure S7.

Early samples (day 7–14, n = 48) were compared to later samples (day 15–400, n = 509). The early samples showed significantly higher levels (p < 0.0001) for both rd-cfDNA (median 18,135 versus 9,905 copies/mL), dd-cfDNA (median 48.7 versus 9.3 copies/mL) and DF (median 0.31% versus 0.09%), see Supplementary Figure S3.

Samples from adults (n = 475) were compared to samples from children (n = 82), including early samples. Levels of rd-cfDNA did not differ significantly (adults median 11,091 copies/mL versus 7,177, p = 0.053). Levels of dd-cfDNA did not differ significantly (adults median 11.0 copies/mL versus children 9.8, p = 0.99). DF was significantly lower in adults (median 0.09 versus 0.13, p = 0.03), see Supplementary Figure S4.

Female patient samples (n = 171) were compared to samples from male patients (n = 386). As compared with females, males had significantly lower levels of rd-cfDNA (median 8,888 copies/mL versus 15,943, p = 0.0004). There were no significant differences for dd-cfDNA (males median 9.8 copies/mL versus 13.2, p = 0.13) and DF (males median 0.12% versus 0.09%, p = 0.08), see Supplementary Figure S5.

The rd-cfDNA results of the initial dPCR (copies/mL) and the PA-dPCR (copies/µL) were compared including results from samples taken during the first 14 days after HTx (n = 557). The results showed a very high correlation (Pearson r = 0.97, CI₉₅ (0.97; 0.98), R² = 0.95, p < 0.0001; Spearman r = 0.95, CI₉₅ (0.94; 0.96), p < 0.0001). The levels of rd-cfDNA from PA-dPCR were also correlated with the fluoroscopic measurements of DNA-concentration in the initial plasma samples using Qubit. The results showed a very high correlation (Pearson r = 0.95, CI₉₅ (0.94; 0.95), R² = 0.90, p < 0.0001; Spearman r = 0.93, CI₉₅ (0.92; 0.94), p < 0.0001).

The efficiency of preamplification was determined using a cfDNA standard and qPCR to monitor individual SNP assays as previously published by our group (XX). The qPCR profiles for all SNP assays including the Y-chromosome are seen in Supplementary Figure S6. No changes in allelic distribution for the SNP assays could be detected within the range of cfDNA concentrations.

Patient 1: A 27 year-old patient underwent uneventful HTx. The clinical course was unremarkable except for suspected bacterial infection on day 3 and day 20 as well as Influenza A infection on day 270. A total of 11 scheduled endomyocardial biopsies were obtained, none of which showed signs of rejection warranting treatment. The results are shown in Figure 3.

Patient 2: A 28 year-old patient underwent uneventful HTx. Scheduled routine biopsy on day 137 showed a 3R rejection and treatment was initiated. Further biopsies revealed resolution of the rejection. Infection episodes occurred on day 4 and day 94 (respiratory infections), day 125 (infection with enterovirus) and day 164 (urinary tract infection). The results are shown in Figure 4.