This is a single-center prospective study performed on 100 bilateral lung transplant recipients who consecutively underwent scheduled monitoring after surgery from October 2022 to October 2024 at the University Hospital of Padova. All patients underwent routine post-transplant monitoring with regular clinical visits, spirometry, imaging, surveillance bronchoscopy with BAL and TBB, and DSA measurements. All BAL and plasma samples analyzed in this study were collected exclusively during these scheduled surveillance visits. No samples were obtained during clinically indicated procedures prompted by acute clinical symptoms, spirometric decline, or radiologic abnormalities.

At our center, surveillance protocol includes scheduled medical appointments as previously reported (12) and cause evaluations in case of respiratory exacerbations. Precise inclusion/exclusion criteria are reported in Figure 1 . The “Strengthening the Reporting of Observational Studies in Epidemiology” (STROBE) statement guidelines for reporting observational studies were followed. All patients provided informed consent, and the study was approved by our Institutional Review Board.

Plasma and BAL samples underwent standard culture evaluations for bacteria, mycobacteria, and fungi, including molecular assessments for viruses and non-cultivable bacteria. Infection was defined according to the International Society for Heart Transplant (ISHLT) guidelines as the detection of a clinically significant organism in microbial cultures, accompanied by mucopurulent discharge observed during bronchoscopy or focal radiographic abnormalities on thoracic imaging. Additionally, an inflammatory response, demonstrated by lymphocytosis and/or neutrophilia on cytological evaluation, was required to support the diagnosis (13). In contrast, colonization was defined as the presence of microorganisms detected on microbial cultures (10⁴ colony-forming unit - CFU) without evidence of mucopurulent discharge, focal radiographic abnormalities, or an associated inflammatory response, indicating a non-pathogenic or commensal state without active disease.

Anti-HLA antibodies were evaluated at the time of the reference biopsy. Anti-HLA IgG reactivity was analyzed with bead-based assays using the Werfen, Single Antigen Bead kits (Waukesha Wi, USA). Single antigen results with a mean fluorescence intensity (MFI) greater than 500 were considered as positive.

The histopathologic evaluation of TBB specimens was performed according to the revised ISHLT working group recommendations and the LASHA template (12). Morphological features suggestive of antibody mediated rejection (AMR) were systematically reported in accordance with LASHA and further discussed in MDT meetings. The certainty of AMR was determined based on the ISHLT consensus statement (14).

At our center a MDT meets weekly to review and discuss patient data, ultimately reaching a consensus on the diagnosis of the observed graft lesion to guide appropriate clinical management. The MDT report includes a morphological description of the biopsy (according to LASHA), humoral and infectious data, imaging findings and cytological information of the BAL. All MDT reports for enrolled patients were independently adjudicated by a predefined expert panel. This panel included two pulmonologists (FM, FB), one immunologist (EC), one radiologist (CG), and two pathologists (FC, FP), and was further supported by an external expert in lung transplantation and antibody-mediated rejection (DL, Stanford University), who provided additional oversight and expertise in complex cases. Four categories were identified: 1) No injury (group 1), 2) Immunological injury: including acute cellular rejection (ACR), AMR, and Chronic Lung Allograft Dysfunction (CLAD) (group 2), 3) non-immunological injury: including infections or gastroesophageal reflux disease (GERD)-lung aspiration (group 3) and 4) Mixed immunological and non-immunological injury (group 4).

Cases with ACR<1 without allograft dysfunction and possible subclinical AMR were included in group 1. The MDT was blinded to dd-cfDNA% data.

Plasma and BAL samples were collected using Streck Cell-Free DNA tubes and processed within two hours of collection. After centrifugation, aliquots were stored at −80°C. Extraction of cfDNA was performed using the QIAamp MinElute ccfDNA kit (QIAGEN, Venlo, Netherlands), and downstream analysis was conducted using the AlloSeq^® cfDNA kit (CareDx, CA, USA). Next-generation sequencing (NGS) was performed on the Illumina MiniSeq 3000 platform (2 × 150 bp). Library preparation, single nucleotide polymorphism (SNP) genotyping, and paired-end read mapping were used to quantify the proportion of dd-cfDNA, expressed as the percentage of donor-specific SNPs among total SNPs. The software and quality control pipelines provided by CareDx were used to determine dd-cfDNA percentages. For plasma, the 1% threshold was predefined by the manufacturer and supported by previous literature as the clinically relevant cut-off for detecting allograft injury in solid organ transplantation. For BAL fluid, where no established literature standard exists, a cut-off of 10% was defined based on cross-validated ROC analysis and supported by a subset of pre-/post-transplant paired controls. Additionally, we performed post-hoc analyses using alternative thresholds to assess how test characteristics varied across cutoffs; results consistently supported the selection of the 10% value (see Supplementary Methods for detailed rationale and performance metrics). This threshold was derived from the average of fold-specific Youden index optima in a 10-fold cross-validation framework (mean ≈ 9.7%) and rounded to 10% for clinical interpretability.

For further technical details, including pre-analytic handling, sequencing metrics, SNP filtering, and validation of BAL-specific dd-cfDNA quantification, please refer to the Supplementary Material .

Plasma samples were collected concurrently with TBBs in all patients. Additionally, a subset of patients underwent repeated plasma sampling for dd-cfDNA during their scheduled visits. In a subset of patients, aliquots of BAL were collected for this analysis.

All assessment data were meticulously recorded within the dedicated REDCap platform.

Continuous variables were expressed as median (interquartile range), and categorical variables as counts (percentages). Mann-Whitney, Pearson’s Chi-squared test, Wilcoxon rank sum test, and Fisher’s exact test were employed as appropriate. Correction for multiple comparisons was made by adjusting the false discovery rate (FDR) to control the type I error rate.

Among 300 Caucasian recipients regularly visited at our center between October 2022 to October 2024, 100 patients were consecutively enrolled following the exclusion/inclusion criteria ( Figure 1 ); 61% were males, with a median age of 50 years (IQR 40–63). The median time from transplantation to the collection of plasma samples was 10 months (IQR 3–45 months). Major demographic characteristics are reported in Table 1 . As for graft injury, 52% of the cohort showed some form of graft damage: 15% categorized as Group 2 (immunological injury), 28% as Group 3 (non-immunological injury) and 9% as Group 4 (both immunological and non-immunological injury). Among the patients in the immunological group, one was graded as ACR (grade A3), four as AMR (3 possible, 1 definite), and 10 CLAD (1 restrictive, 7 obstructive, 2 mixed; one with concurrent AMR). In AMR cases, the most frequently observed histological lesions included severe septal widening, moderate/severe granulocyte septal infiltration, presence of hyaline membranes, septal or intra-alveolar granulation tissue plugs (organizing pneumonia)/pneumocyte hypertrophy, and high scores of hemosiderophages. The remaining 48% of patients showed no signs of graft injury (Group 1) Table 2 .

Anti-HLA antibody testing was available for the entire cohort. Prior to transplantation DSA were not observed in 79% of cases whilst 21% had non-donor-specific anti-HLA antibodies. Following transplantation de novo DSA were detected in only 7% of cases. A complete dataset with typing for 4 HLA loci (HLA A, B, DRB1 and DQB1) was available for 62 cases. Among the 62 typed donor-recipient pairs, the distribution of mismatches (MM) ranged from 3/8 to 8/8: In particular, there 3 cases (5%) with 3/8 MM; 5 cases (8%) with 4/8 MM; 10 cases (16%) with 5/8 MM; 12 cases (20%) with 6/8 MM; 20 cases (32%) with 7/8 MM; 12 cases (20%) with 8/8 MM. These results show that a high degree of HLA MM (7/8 or 8/8) was present in more than 50% of the overall cohort of transplanted patients. In addition, when comparing the populations of transplanted patients with or without DSA, in both populations we observed a high degree of HLA MM (7/8 and 8/8) in 60% and 51% of cases, respectively. The small sample size, however, did not enable statistical evaluations.

A total of 127 dd-cfDNA levels were measured in the 100 patients with median plasma dd-cfDNA level of 0.58% (IQR 0.29%–1.65%). Forty-six percent of samples exceeded the clinically relevant threshold of 1%. The dd-cfDNA levels in plasma, assessed across different patient groups, showed a clear association with graft injury. Plasma dd- cfDNA levels were significantly higher in patients with graft injury, with a median of 1.27% (IQR 0.49%–3.65%) compared to 0.39% (IQR 0.22%–0.61%) in those without graft injury (p < 0.001). Stratifying by injury type, patients with immunological injury (Group 2) exhibited the highest plasma dd-cfDNA levels, with a median of 2.67% (IQR 1.38%–4.06%), while non-immunological cases showed lower levels (median 0.68%, IQR 0.34%–1.15%). In Group 2, the highest median values of dd-cfDNA were observed in AMR cases (median: 2.77%, IQR 1.25%–4.9%). Patients with both immunologic and non-immunologic injuries exhibited the highest median plasma dd-cfDNA levels, with a median of 3.39% (IQR 1.39%–3.93%) ( Table 2 ).

The ROC curve of Group 2 showed an area under the curve (AUC) AUC of 0.93 (95% CI 0.88-0.99) and a sensitivity of 100% (positive predictive value - PPV= 71%, negative predictive value - NPV=100%) and overall accuracy of 90% (p = 0.003). In contrast, the ROC curve of Group 3 showed an AUC of 0.63 (95%CI 0.50-0.76) and a much lower sensitivity of 28%, (PPV=57%, NPV=68%) with an overall accuracy of 66%. No difference was observed in terms of percentage level comparing different forms of infections (bacterial, vs viral vs fungal) and infections with other injuries such as lung aspiration. Even insignificant multiple agents (viruses and/or bacterial) showed a higher dd-cfDNA level compared to single agents (median: 2.98, IQR: 0.28-3.35 vs median: 1.12, IQR: 0,29-1.39; p = ns).

The ROC curve for patients of Group 4 showed an AUC of 0.90 (95% CI 0.79-1.0), a sensitivity of 87% (PPV=54%, NPV=98%) and accuracy of 88% Figure 2 , Table 3 .

Twenty-seven serial plasma samples were obtained from patients with more than two assessments. These follow-up samples were collected during scheduled surveillance visits as part of routine post-transplant monitoring and were not specifically driven by rejection episodes, non-response to therapy, or persistent dysfunction. In general, in our center, if an episode of ACR is diagnosed, patients are treated and re-evaluated after 40 days with repeat transbronchial biopsies and pulmonary function tests. If histological signs of rejection persist, a second course of high-dose steroids is administered. Should the rejection fail to fully regress after an additional 40 days, and after ruling out other potential causes of non-response such as Cytomegalovirus infection or subtherapeutic immunosuppressant levels, a change in maintenance immunosuppressive regimen is considered. Of these recipients only one showed a severe graft injury (concomitant A3Bx + multiple viral infections) at the last surveillance follow-up: at this time a high dd-cfDNA level (2.98) was detected Figure 3 . This patient experienced a bacterial/fungal infection with an associated mild sequalae of ischemia/reperfusion injury scored by LASHA ≤2 (dd-cfDNA: 0.76), which resolved by the time of the second scheduled visit (dd-cfDNA: 0.45).

Univariate analysis showed that time from lung transplantation, native diseases other than obstructive and restrictive forms, mainly including idiopathic pulmonary hypertension, and donor age were associated with higher dd-cfDNA values (p < 0.0001, p = 0.024, p = 0.03, respectively) ( Supplementary Table 2 ). However, after the multivariate regression analysis the only parameter associated with increased dd- cfDNA levels (Beta = 0.02, 95% CI [0.01, 0.03], p < 0.001), without any difference among groups was the time from lung transplantation ( Supplementary Table 3 ). All other donor or recipient characteristics showed no significant influence on dd- cfDNA levels in all groups.

A total of 60 BAL dd-cfDNA levels were measured in 60 patients with concomitant plasma evaluation. Four samples were not processed due to inadequate DNA. The cut-off of BAL was 10%. The dd-cfDNA levels in BAL fluid were substantially higher than those in plasma, with a median of 12% (IQR 5%–23%). Unlike plasma dd-cfDNA, BAL dd-cfDNA levels did not significantly differentiate between patients with and without graft injury (p = 0.060).

After stratification by injury type, Group 3 patients showed a better sensitivity than plasma dd-cfDNA (43% vs 28%), even if low specificity ( Supplementary Table 4 ). The cutoff for multiple infections (bacterial and viral) was 16.17% (IQR 12–20), significantly higher than that for single infections, which was 1.88% (IQR 0.7–1.9) (P = 0.037). Aspiration did not influence the significance level.

All performance metrics presented in this section refer exclusively to the 60 patients who had paired plasma and BAL sampling. Within this subcohort, the performance of plasma dd-cfDNA alone for detecting non-immunological injury mirrored the results from the larger 100-patient cohort, with a sensitivity of 28%. The addition of BAL dd-cfDNA (using a 10% cut-off) improved sensitivity for detecting non-immunologic injury to 71%, recovering 4 out of 8 cases that plasma alone had missed. Importantly, this diagnostic gain was limited to Group 3 (infections or aspiration-related injuries); there was no added benefit observed in Group 2 (immunological) or Group 4 (mixed) injuries. Thus, whilst adding BAL had no value in the case of immune-mediated damage, our findings demonstrate a clear complementary role of BAL dd-cfDNA in non-immunological cases.

Additionally, we evaluated the performance of the combined molecular rule (plasma >1% or BAL ≥10%) against histological outcome. The combined dd-cfDNA approach showed a sensitivity of 65% (95% CI: 42–82%), specificity of 85% (63–96%), PPV of 83% (60–95%), NPV of 68% (45–86%), and overall accuracy of 74% (59–86%). The molecular assay missed 35% of biopsy-positive cases.

The Cox proportional hazards regression identified two significant predictors of adverse outcomes (development of CLAD or death), in LT recipients: older recipient age and chronic rejection-associated lesions described by LASHA. Recipient age was associated with a 5% increased risk of adverse events for each additional year (Hazard ratio - HR = 1.0467, 95% CI: 1.0075–1.087, p = 0.019). Suggestive histological lesions of chronic rejection increased the risk of CLAD development (HR = 6.50, 95% CI: 2.69–156.9, p = 0.0102) and in general of adverse outcome (HR = 3.48, 95% CI: 1.03–11.7, p = 0.044). Random Forest survival analysis showed that plasma dd-cfDNA higher than 1% was among the most important variables in influencing CLAD and death. In addition, we assessed the direct association between plasma dd-cfDNA levels and CLAD using a univariable logistic regression model. Applying the pre-specified 1% plasma dd-cfDNA threshold, recipients with values above this cut-off had approximately 16 times higher odds of being diagnosed with CLAD compared to those below it (Odds Ratio = 16.0, 95% CI = 2.6–305, p = 0.012). Although the confidence interval is wide due to the limited number of CLAD cases (n = 10), the lower bound of the interval remains well above 1, indicating a statistically meaningful association between elevated plasma dd-cfDNA and CLAD status. ( Figures 4A, B ).