Between January 1st 2018 and September 31st 2021, 195 consecutive adult patients received an Allo-HSCT at the Humanitas Cancer Center (Rozzano, Italy) for hematologic malignancies. All adult patients with hematologic malignancies undergoing their first allo-HSCT from any graft source (blood marrow or peripheral blood stem cells) and after any type of conditioning regimen were included. The hematopoietic cell transplantation comorbidity index (HSCT-CI) was calculated according to Sorror et al. [24]. Patients were routinely monitored for complete blood counts, LDH, and urine protein levels at least once a week. Blood pressure was monitored daily during hospital admission and on every outpatient visit. Screening features for TA-TMA comprised the seven EAABMT criteria: anemia, thrombocytopenia, elevated LDH, hypertension, schistocytes, proteinuria and sC5b-9 plasma levels (marker of complement activation). When at least three of these characteristics occurred simultaneously or remained unexplained over time, further investigations were carried out, and TA-TMA was diagnosed when at least four of the seven EAABMT features occurred twice within 14 days, as recommended by Schoettler et al. [2]. A total of sixteen patients were suspected of TA-TMA. Their demographic characteristics are summarized in Table 1. ADAMTS13 (a disintegrin and metalloproteinase with a thrombospondin type 1 motif, member 13), soluble C5b-9 (sC5b-9) complex and complement genetic studies were evaluated for blood samples obtained at the time of TA-TMA suspicion.

The study was approved by the Lombardia 5 Ethics Committee (no. TAM ONC/OSS-02/2025). All patients gave their written informed consent and all procedures were performed in line with the Helsinki Declaration and the code of good clinical practice.

Blood samples were collected from an antecubital vein by clean venipuncture with minimal stasis using a 21 G butterfly needle in tubes containing ethylenediaminetetraacetic acid (EDTA: final concentration 10 mM). The samples were centrifuged within one hour for ten minutes at 2000 g at room temperature. The plasma was removed, divided into aliquots and stored at −80 °C until tested. Plasma levels of soluble C5b-9 (sC5b-9) were measured using a solid-phase assay (MicroVue Complement SC5b-9 Plus EIA kit, Quidel Corporation, San Diego, CA, USA) whose intra- and inter-assay coefficients of variation were respectively 6.8% and 13.1%; the lowest detection limit was 3.7 ng/mL. For sC5b-9 measurements, twenty healthy subjects (ten women and ten men; median age: 48 years; range: 20–70 years) served as normal controls.

Anti-factor H autoantibodies were determined by ELISA, using purified factor H for capture and anti-human IgM, IgG, and IgA antibodies for detection [18]. Purified factor H (Calbiochem, EMD Chemicals, San Diego, CA, USA; 10 mg/mL in phosphate-buffered saline at pH 7.4) was coated overnight onto microtitration plates, and, after washing to avoid non-specific binding, the wells were coated with bovine serum albumin. After additional washes, 1:20 diluted serum samples were added and incubated at room temperature for 45 min. After further washing, the factor H-bound immunoglobulins were detected by class-specific mouse monoclonal anti-IgM, -IgG, or -IgA antibodies (Sigma Aldrich, St Louis, MO, USA). Peroxidase-conjugated anti-mouse antibodies (Sigma-Aldrich), developed with ortho-phenylenediamine, were used for detection. Optical density (OD) was measured at 490 nm. Results were expressed in units per mL (U/mL) with reference to an internal standard (serum from a patient with a high anti-factor H antibody titer) arbitrarily fixed at 100 U/mL. In order to avoid the confounding effect of natural antibodies, it was decided to use the maximum level observed in normal subjects as the cut-off between normal and abnormal levels.

ADAMTS13 activity plasma levels were measured with a fluorescence resonance energy transfer assay that uses a synthetic 73-amino acid peptide (FRETS-VWF73), as previously described [25]. The intra- and inter-assay coefficients of variation were 6% and 9.5%, respectively.

Genomic DNA was isolated from blood samples using the QIAsymphonySP automated platform (Qiagen GmbH, Hilden, Germany). DNA quality was assessed using the NanoDrop™ 2000c spectrophotometer (ThermoFisher, Waltham, MA, USA) while quantification was performed using the Qubit dsDNA HS assay kit (Thermo Fisher) and the Qubit Fluorimeter (Thermo Fisher). For nanopore sequencing, DNA integrity was also evaluated using the ScreenTape Genomic DNA kit (Agilent Technologies, Santa Clara, CA, USA) and the 4150 TapeStation System (Agilent Technologies). All measurements were carried out following manufacturer instructions.

Nucleotide variations were detected by NGS on the MiSeq platform (Illumina) using the ‘targeted sequencing’ technique (SureSelect XT HS RevB ILM – Tier1 Target Enrichment System, Agilent Technologies) employing a multiple gene custom panel comprising CFH (NM_000186.4), MCP/CD46 (NM_002389.4), CFI (NM_000204.5), C3 (NM_000064.4), CFB (NM_001710.6), THBD (NM_000361.3), DGKE (NM_003647.3), CFHR1 (NM_002113.3), CFHR3 (NM_021023.6), CFHR5 (NM_030787.4), with average depth in target regions greater than 100×. To identify putative causative variants, NGS data were processed and filtered using the Agilent Alissa suite, specifically the Align & Call and Interpret modules (Agilent Technologies). Variants with a minor allele frequency (MAF) <3% (1000 Genome Phase 3 and gnomAD databases) were selected. All variants identified by NGS analysis were then confirmed by the standard Sanger sequencing method. The potential impact of amino acid changes was assessed by in silico analysis (VarSome [https://varsome.com], Franklin by genoox [https://franklin.genoox.com/]) to predict the functional significance of unpublished or uncommon variants. Variant frequencies were also cross-referenced against the ExAC database (Exome Aggregation Consortium, [http://exac.broadinstitute.org]) to evaluate their prevalence in the general population. The genetic variant classification of the database of complement genetic variants (http://www.complement-db.org/home.php) was also considered [19,20].

Multiplex ligation-dependent probe amplification assay (MLPA kit P236, MRCHolland, Amsterdam, The Netherlands) was used to identify CFHR3/CFHR1 copy number and macro rearrangements such as CFH/CFH-Related hybrid genes. Raw data were analyzed by Coffalyser.net (https://www.mlpa.com) and the relative dosage ratio was calculated [19,20].

Prior to library preparation, 3 µg of genomic DNA was sheared using g-Tubes (Covaris, Woburn, MA, USA) by centrifugation at 3500 x g for 1 min at room temperature, yielding 10-kb fragments. Library preparation was carried out using the Ligation Sequencing Kit SQK-LSK114 (Oxford Nanopore Technologies, Oxford, UK). The manufacturer protocol was modified slightly by increasing the incubation times of the DNA repair, end prepare, and ligation steps.

PromethION flow cells (FLO-PRO114; Oxford Nanopore Technologies) were loaded on a PromethION 24 instrument. Following flow cell priming, 50 fmol of library was loaded per flow cell according to the manufacturer’s instructions. DNA sequencing, data acquisition, and real-time basecalling were performed for 72h using MinKNOW software with default settings.

Data analysis was initially performed using the EPI2ME software (Oxford Nanopore Tecnologies) installed on the PromethION 24 instrument, using default settings. The wf-human variation pipeline was used to simultaneously call Single Nucleotide Polymorphisms (SNP), structural variants, Copy Number Variants (CNV) and Short Tandem Repeats (STR) using the CRCh38 as reference genome.

Then, data were blindly submitted to another center for independent analysis using a different pipeline. In this second pipeline, the sequencing data were aligned against the GRCh38 reference with Minimap2 [26] and the alignments were analyzed with CNVKit [27], a software designed to identify CNV. The CNVKit is primarily focused on hybrid capture sequencing data. Although a default analysis pipeline is provided by the developers, the tool execution was customized to better cope with the analyzed data. The workflow included:

• i.

  Preparation of the target region with the target –split command (default value of 267 was used for –average-size parameter).

• ii.

  Use of a blank anti-target for both reference definition and sample analysis.

• iii.

  Definition of the reference copy number from the alignment profiles of the wt samples selected by a visual inspection carried out with Integrative Genomics Viewer (IGV) of the CHFR1–5 locus [28]

• iv.

  Sample analysis with the serial execution of coverage, fix, segment (using the Hmm-germline method) and call (using custom thresholds for copy number calling: −1.1, −0.6, 0.4, 0.7) commands.

Due to a non-normal distribution of data according to the Kolmogorov-Smirnov test, quantitative variables are expressed as median, ranges (minimum–maximum) and/or interquartile range (IQR). Categorical variables are expressed as an absolute number. The nonparametric Mann-Whitney U test was used to compare different groups. Statistical significance was defined as a p-value <0.05. Outcomes were defined according to the European Society for Blood and Marrow Transplantation (EBMT) statistical guidelines. Cumulative incidence estimates were used to account for competing risks in the study population. Differences between cumulative incidence curves in the presence of relapse as competing risk, were tested using the Gray method [29]. Results are presented as 1-year non-relapse mortality. Statistical analyses were performed using SAS software version 9.4 (SAS Institute Inc, Cary, NC, USA) and IBM Statistical Package for Social Sciences (SPSS) version 29.0 (IBM Corp, Armonk, NY, USA). The impact of TA-TMA on overall survival was analyzed employing the Cox proportional hazard model where TA-TMA was evaluated as the time dependent variable (STATA version 14.2, StataCorp, TX, USA).

Of the 16 patients with suspected TA-TMA (Table 1), one (Patient #1) presented ADAMTS13 levels <3%. Consequently, a diagnosis of thrombotic thrombocytopenic purpura (TTP) was made, and the patient was excluded from the study. The remaining 15 patients (Table 2) comprised the cohort with suspected TA-TMA. The median time to suspected diagnosis was 80 days post-transplant (IQR: 57–95 days). For 13 of these patients, TA-TMA was suspected within the first 100 days post-transplant, while the remaining two were identified at a later stage (170 and 220 days) with both the latter having concurrent chronic graft-versus-host disease (GvHD). Concomitant post-transplant complications occurred in all patients; either acute or chronic GvHD in 11, viral infections in seven and probable invasive fungal disease in one (Table 2). More details are reported in Table 3. Of the 15 patients, five had <4 EAABMT criteria, ruling out TA-TMA; one of them died because of disease relapse and one for acute GvHD. Ten patients had ≥4 EAABMT criteria and were diagnosed with TA-TMA: four criteria in five patients and >4 in the other five (Table 2). Table 4 shows the values of hemoglobin, platelets, lactate dehydrogenase, Coombs test, blood pressure, prothrombin time and activated thromboplastin time. Six of these ten patients were switched from cyclosporine to sirolimus with four showing clinical improvement, one of whom also received eculizumab treatment (Table 3). At 1-year post-transplant, seven of the ten patients had died, one because of disease relapse and six due to post-transplant side effects (Table 3). The cumulative incidence of 1-year non-relapse mortality was 20% in patients with <4 EAABMT criteria and 60% in patients with ≥4 EAABMT criteria (Figure 1A) (p-value = 0.07). Overall survival was further compared between the cohort of ten patients with confirmed diagnosis of TA-TMA and 185 subjects without TA-TMA using the Cox proportional hazard model where TA-TMA was analyzed as the time dependent variable: patients developing TA-TMA had five-fold increased risk of death compared with those without TA-TMA (hazard ratio [HR]: 5.19; 95% confidence incidence: 2.45–10.98; p-value <0.001) Table 5

Table 2.

Clinical and laboratory characteristics of patients with suspected transplant-associated thrombotic microangiopathy (TA-TMA) after exclusion of the patient with thrombotic thrombocytopenic purpura.

aGvHD: acute graft versus host disease; Allo-HSCT: allogeneic hematopoietic stem cell transplantation; BKV: polyomavirus BK; CMV: cytomegalovirus; cGvHD: chronic graft versus host disease; EAABMT criteria: criteria from the consensus of the world’s leading societies for Blood and Marrow Transplant Research; G2/G3: grade 2 or 3 aGvHD; HHV6: herpes virus 6; IFD: invasive fungal disease.

Patients with confirmed diagnosis of transplant-associated thrombotic microangiopathy (TA-TMA) are highlighted in grey.

Table 3.

Timing of graft-versus-host disease, veno-occlusive disease (VOD), infections and initiation of immunosuppressive therapy from the diagnosis of transplant-associated thrombotic microangiopathy (TA-TMA).

Patients with confirmed diagnosis of transplant-associated thrombotic microangiopathy (TA-TMA) are highlighted in grey. The minus sign refers to the days before the diagnosis of TA-TMA and the plus sign refers to the days after the diagnosis of TA-TMA.

aGvHD: acute GvHD; BP: blood pressure; BSI: blood stream infection; cGvHD: chronic GVHD; Hb: hemoglobin; LDH: lactate dehydrogenase; NA: not available; PLT: platelet; SARS-CoV2: severe acute respiratory syndrome coronavirus 2.

Table 4.

Laboratory and clinical characteristics of patients with transplant-associated thrombotic microangiopathy (TA-TMA).

Hb: hemoglobin; PLT: platelet; LDH: lactate dehydrogenase; DAT: direct antiglobulin test (direct Coombs test); NV: normal values; PT/aPTT: prothrombin time/activated thromboplastin time.

Patients with confirmed diagnosis of transplant-associated thrombotic microangiopathy (TA-TMA) are highlighted in grey.

Figure 1.

A. Incidence of non-relapse mortality in transplant-associated thrombotic microangiopathy (TA-TMA). The figure shows the cumulative incidence of 1-year non-relapse mortality of 15 patients with suspected diagnosis of TA-TMA that was confirmed in ten (TA-TMA) and not confirmed in five (No TA-TMA). B. Plasma levels of soluble C5b-9 (sC5b-9) in TA-TMA. The figure shows plasma levels of sC5b-9 in ten patients with TA-TMA, five patients without TA-TMA and 20 healthy subjects that served as controls. Boxes show median value and interquartile range while whiskers show 95% confidence interval.

Table 5.

Chimerism and supplementary donor characteristics.

CB: Cord blood; Haplo: Haploidentical transplant; MRD: Matched related sibling; MUD: Matched unrelated donor; NA: not available; UD: Unrelated donor (≥7/8 disparity).

Patients with confirmed diagnosis of transplant-associated thrombotic microangiopathy (TA-TMA) (EAABMT criteria ≥4) are highlighted in grey.

Consistent with TA-TMA diagnosis (≥4 EAABMT criteria), the ten patients with TA-TMA had sC5b-9 plasma levels (median 610 ng/mL; range 210–3313 ng/mL) higher than the five patients with <4 EAABMT criteria (257 ng/mL; 152–368 ng/mL) (Figure 1B; p-value = 0.040). In patients with TA-TMA, sC5b-9 plasma levels were also higher than those of 20 healthy controls (131 ng/mL; 75–242 ng/mL) (p-value <0.001). None of the patients in this cohort tested positive for anti-factor H autoantibodies (IgG, IgM, or IgA).

At the time of genetic testing, full chimerism was evident in all 15 patients (Table 5).

Table 6 shows complement genetic findings of the ten patients with TA-TMA and the five patients with <4 EAABMT criteria. All complement-related genetic variants identified in the 15 patients via targeted NGS were rare (MAF <1%) and classified as variants of uncertain significance (VUS). Seven variants were found in the group of ten TA-TMA patients (two patients had two variants and three patients had one variant) whereas only one variant was found in the group of five patients with <4 EAABMT criteria (Table 6 & Figure 2A). All variants were in the heterozygous state. The CFHR3-CFHR1 deletion, identified by MLPA (Table 6), was present in four of ten patients with TA-TMA (one as homozygous and 3three as heterozygous) and in two of the five patients with <4 EAABMT criteria (all heterozygous). The slight difference in allele frequency (25% in patients with TA-TMA versus 20% in patients with <4 EAABMT criteria) did not reach statistical significance.

Table 6.

Complement genetic data of 15 patients with suspected transplant-associated thrombotic microangiopathy (TA-TMA). Nanopore sequencing confirmed targeted next generation sequencing using the EPI2ME pipeline, whereas, to confirm MLPA deletions it was necessary to use the newly developed customized pipeline based on CNVKit.

Patients with confirmed diagnosis of transplant-associated thrombotic microangiopathy (TA-TMA) (EAABMT criteria ≥4) are highlighted in grey.

Figure 2.

A. Variants in complement-related genes identified with next generation sequencing and confirmed with nanopore sequencing. The figure shows the single nucleotide variants (circles, squares, triangles and diamond) within the exons (blue rectangles) of the single complement- related genes (C3: NM_000064.4; DGKE: NM_003647.3, THBD: NM_000361.3; CFH: NM_000186.4; CFHR3: NM_021023.6. 3′UTR: 3′ untranslated region. B. Rapid genomic analysis in three patients with transplant-associated thrombotic microangiopathy (TA-TMA). The figure shows read coverage data across the genomic region encompassing the CFHR3 and CFHR1 genes for three different samples from patient no 3 (wild type), patient no 14 (heterozygous deletion) and patient no 13 (homozygous deletion). The dashed vertical lines indicate the breakpoints of the structural variants spanning the two CFHR genes, as automatically calculated by the CNVkit tool.

The 15 samples were sequenced with nanopore technology (Figure 2B) and analyzed by combining the EPI2ME and CNVKit pipelines. A custom in-house quality control tool was applied to ensure accurate variant calling within the complex CFHR1–CFHR5 region. This analysis yielded a mean N50 of approximately 8400, a mean depth of 37×, and a mean alignment quality of 59.6. A drop in alignment quality was observed in the CFHR1–CFHR3 region (39.4). The EPI2ME wf-human-variation workflow confirmed the results obtained with targeted NGS (Table 6) but failed to identify deletions detected by MLPA. In contrast, the CNVKit analysis focused on the CFHR3–CFHR1 region, identified aberrations matching the MLPA location and zygosity for all six carrier samples. In the remaining samples, the tool produced no false positive calls, achieving 100% sensitivity and specificity for the analysis. Reproducibility tests performed on replicas of the same samples gave fully concordant results.