The present invention provides methods relating to detecting donor cell-free DNA in the circulation of an organ transplant recipient for the early identification of transplant rejection.

Background of the invention

Every year 100.000 patients undergo organ transplantation all over the world as the only definitive treatment for end-stage organ failure.

Independently on the organ transplanted, graft rejection is a major open problem for these patients, occurring when the graft is recognized as foreign, attacked and rejected by the recipient's immune system, causing cellular damage and graft failure. Postoperative care consists of a painful surveillance that lasts their lifetime, with frequent organ biopsies, necessary to guarantee a survival that is currently expected at 15 years.

Currently, the main monitoring methods for organ rejection upon transplantation are two:

a) tissue biopsy, that is an invasive procedure by which tissue cells are taken directly from the transplanted organ. This procedure carries several critical aspects: risk for patients, uncomfortable to be performed by doctors, expensive, feasible with sedation or anesthesia in an hospital setting, samples could be taken on unaffected graft tissue because rejection process is often focal and the number of samples taken per procedure is limited (from 5 to 8), with at least one non informative sample because it is possible to take a sample that is not included in the rejected area (useless).

b) biomarkers, whose efficiency varies for different organs and are dependent from patient features. For example approximately 50% of the function of the transplanted kidney can be lost before measurable increase of serum creatinine (kidney biomarker), depending on sex, muscle mass, or ethnicity. Conventional tests for liver function do not specifically assess acute cellular rejection (ACR).

A particular case is the example of the heart wherein approximately 7,000 of transplant/year are heart transplants (WHO Global Observatory on Donation and Transplantation).

The endomyocardial biopsy (EMB) remains the gold standard for acute rejection surveillance after Htx, but this technique entails many negative and critical features. Therefore, in solid organ transplantation, therapeutic interventions may be delayed by late diagnosis. Allogenic rejection is diagnosed and graded by histology study of the allograft biopsy (1 ). Acute cardiac allograft rejection remains common during the first year post-transplantation, with an incidence of over 40%, and represents a leading cause of mortality during this period, responsible for approximately 12% of fatalities (1 ). Moreover, an episode of rejection occurring during the first year, even when apparently successfully treated, confers higher two-year and four-year mortalities in those surviving beyond the first year, and generates independent risk factors for allograft vasculopathy (1 ).

The mechanisms of acute rejection in pediatric Htx recipients are similar to those occurring in adults. While infants and young children have lower acute rejection rates, adolescents have the highest acute rejection rates, with many episodes > Grade 2R (classification ISHTL 2010, moderate acute cellular rejection) asymptomatic and detected only by surveillance EMB (2).

The EMB is an invasive technique and suffers from inter-observer variability, high cost, potential complications, and significant patient discomfort (3). Histological analysis of right ventricular myocardial tissue obtained with EMB remains the‘gold standard’ technique for acute rejection surveillance; during the first postoperative year patients undergo frequent biopsies. There is no consensus on the optimal frequency of surveillance with EMB, and EMB schedules vary between Htx centers. The frequency of EMB is highest in the first 3 postoperative months with a decreasing frequency thereafter. The first year after Htx the patient undergoes about 20 EMBs. This schedule is based on the observation that the risk of allograft rejection is highest in the first 6 months and decreases sharply after 12 months. The usefulness of surveillance EMB in all patients later than 1 year after transplant is subject of debate. If the patient manifests a clinical picture consistent with allograft rejection, then it is appropriate to perform EMB, as the results may dictate changes in therapy (4). Conflicting data exist on the diagnostic yield and need for surveillance EMB in pediatric recipients. In single centers the rates of acute rejection on surveillance EMB ranges from 0.3% to 1.4% in the first year post-transplant and from 0% to 10% thereafter. Due to very low rates of rejection on surveillance EMB beyond 5 years, there is increasing consensus that EMB beyond 5 years have little usefulness in asymptomatic patients. Given the increased risk of complications in pediatric recipients, many center minimize the number of surveillance EMB in very small children and avoid them altogether in infants, while at a few pediatric center no routine surveillance EMB are performed in pre-adolescents. Furthermore, due to sampling error related to the patchy nature of acute rejection, variability in the interpretation of histological findings and non-routine screening for antibody-mediated rejection, ‘biopsy negative’ acute rejection (hemodynamic features suggestive of significant acute reject but apparently normal EMB) is reported to occur in up to 20% of patients. Substantial inter-observer variability exists in the grading of heart biopsies, and acute rejection may be missed when taking small samples of myocardial tissue, owing to the inhomogeneous distribution of inflammatory infiltrates and graft damage.

Furthermore, EMB is invasive, with an associated complication rate of approximately 0.5% (including myocardial perforation, pericardial tamponade, arrhythmia, access- site complications and significant tricuspid regurgitation) (5), it is expensive and it is disliked by patients, factors that prevent more frequent procedures and, thus, limit optimal titration of immunosuppressive therapy.

In children, deep sedation or general anesthesia is generally required to achieve safe vascular access and to perform the EMB. The procedure is particularly challenging in very small infants. Overall risk of serious complications with EMB in children is 0.6%

(6).

In synthesis, the gold standard procedure has the advantage that it is possible to make the diagnosis of both acute rejection and infections.

However, said procedure is very invasive with an associated risk for the patient, has an inter-observer variability, needs hospitalization and a high specialized staff and depends on patients health status.

Due to the above features, the gold standard (EMB) does not reach efficient screening and transplantation surveillance: not suitable for kids and for patients with critical physical condition, increase of the hospital saturation problem, procedure complex and time-consuming, and it involves several risks for the patient. The ability to rapidly and reliably measure graft integrity is required to effectively adjust therapy in individual patients (i.e. to provide personalized immunosuppression) and thereby improve long-term graft survival. The fact that organ transplants are also genome transplants opens up the possibility of monitoring for allograft injury and integrity by quantification of graft-derived circulating cell-free DNA (cfDNA). In 2014 De Vlaminick et al. (7) conducted a prospective cohort study (65 patients, 565 samples) aimed at stating the potential effectiveness of cfDNA as biomarker in Htx. They used Next Generation Sequencing (NGS) technique, sequencing the whole genome, with a considerable slowdown of process time and subsequent bioinformatic analysis.

In said method the genome of the transplant donors and recipients were characterized using SNPs genotyping, on average 53,423 SNP markers, and the SNP positions with single-base alleles that were distinct between the donor and recipient and homozygous within each individual, allowed discrimination of donor- and recipient-derived sequences.

They concluded that the method was very specific (93%) but costs due to the high number of the SNPs markers used and implementation difficulties were concrete obstacles to translate these findings in clinical practice.

Snyder et al in 2011 (8) discloses a report wherein the cfDNA circulating in the blood of heart transplant recipients has been analyzed and wherein the increased levels of cfDNA from the donor genome has been observed in patients after solid organ transplant rejection, using a shot gun sequencing to measure SNP differences between individuals to quantify the donor DNA signal. The authors reported the presence of donor cfDNA at baseline (heart transplant day 0) between 0 and 1 % and that the quantity of the donor cfDNA increases with rejection to at least 3-6% in heart transplantation.

The method disclosed in the document is based on the sequencing of the donor and recipient DNA and on the analysis of the informative SNP, without a selection of specific list of SNPs that can be used in a method of monitoring the status of a transplanted organ in a subject and applicable to all the samples tested.

Ollerich et al. (9) found no elevation of cfDNA in 5 patients with either intrahepatic or drug-induced cholestasis, or after liver transplant cholangiopathy who had no clinical signs of rejection. This was in contrast to conventional markers like AST and [gamma]-GT, which were highly elevated in these non-rejection conditions.

Sigdel et al in 2013 (10) used a ChrY-based sequencing method using urine from kidney transplant recipients. The method was not suitable for routine use because of the need to have specific gender (male to female) donor/recipient pairs.

Beck et al. in 2013 (1 1 ) and in WO2014/1941 13 used a droplet digital PCR (ddPCR) method as candidate technique for the determination of cfDNA percentages and donor/recipient discrimination. In particular, the digital droplet PCR allows for a rapid quantification of donor cfDNA in the circulation of transplant recipients as a potential universal biomarker of graft injury, by selecting no fewer than 30-35 known SNPs for high minor allelic frequencies. The accuracy of ddPCR was very promising, but the costs and the preparation time are very limiting.

Furthermore, it has been observed that said method it has the limit to be dependent on the technology of the digital droplet PCR, and it is difficult to be reproduced if the data of different laboratories are compared. Finally, the method is applicable in case of presence of a baseline with which to compare subsequent detections and to calculate a trend of variation in the ratio of donor /recipient cfDNA.

Groskovic et al in 2016 (12) and WO2015/138997 developed an assay that indirectly quantifies the fraction of cfDNA in both unrelated and related donor-recipient pairs, with a test time of 3 days, using NGS and a panel of a maximum of 266 different SNPs.

Said documents disclose a method of monitoring the status of an allograft in a transplant recipient and of adjusting immunosuppressive therapies being administered to the transplant recipients, based on the analysis of different SNPs panels for each patient tested, comprising a minimum of about 195 to a maximum of about 266 different SNPs.

In other words, it is assumed that the majority signal from the cell-free DNA sample is recipient-derived DNA and that the minority signal is donor-derived DNA, and this information can be used to calculate the levels of donor-derived DNA in the cell-free DNA sample.

Furthermore, the calculation of the percentage of the donor-derived cell-free DNA with respect to the recipient cell-free DNA in WO2015/138997, is determined on the basis of the variance in the SNP allelic distribution, by comparing said allele distribution to the expected homozygous or heterozygous distribution patterns which may be affected by sequencing errors or PCR artifacts, resulting in low precision of the final result.

Gordon et al in 2016 (13) developed an assay of 124 SNP for NGS application.

Both these last two methods applies an analysis algorithm based on the comparison between a priori probability of presence of variant allele in the SNP panel, and the a posteriori probability calculated after NGS, to inference the presence of donor DNA. However, said methods are costly, require a long preparation, sequencing and analysis time and they need to know the genotype of the donor, which is not often accessible, or are subjected to the calculation of thresholds of significance for the inference of the presence of rejection, due to starting hypothesis on the presence or absence of donor cfDNA.

Finally, the methods are applicable only in case of presence of a baseline with which to compare subsequent detections.

There is therefore the need of a less invasive, less risky, more precise and always applicable procedure for the monitoring of patient undergoing transplantation, and of a more reliable, standardized and economic method.

Brief description of the figures

Figure 1 shows the entire test procedure, from blood sampling to donor fraction calculation and algorithm.

Figure 2. Computer readable instructions: from raw data to DF%.

Figure 3 shows the results of the simulation of a mixture of cfDNA different fractions in 2 different individuals (simulated donor and simulated receiver).

X axis: % of donor cfDNA inserted in the simulated mixture (FD mixture simulated).

Y axis: % of cfDNA of the simulated donor measured with our method (FD measured).

Figure 4 shows the results of the simulation in vitro of 6 mixture of cfDNA from 2 subjects (mimicking recipient and donor), in known percentage of cfDNA of donor in recipient.

X axis: % of donor cfDNA inserted in the in vitro mixture (% expected).

Y axis: % of cfDNA of the donor measured with our method. Figure 5 shows ROC curve derived from ROC analysts (Receiver Operating Characteristic) representing the accuracy evaluation of the machine learning model after training with a simulated dataset, and tested with a random cross validation test set (20-fold validation). X axis: False Positive Rate (1 -Sensitivity). Y axis: True positive rate (Sensitivity).

Figure 6 shows test results on 6 real samples from 3 patients who underwent Htx. T 1 was 2 weeks after Htx, T2 after 3 weeks. Labels represent EMB classification according to ISHLT guidelines. % represent Donor Fraction calculated using our test. DF%= % of donor fraction cfDNA calculated using our test.

Definitions

Unless otherwise defined, all terms of art, notations and other scientific terminology used herein are intended to have the meanings commonly understood by those persons skilled in the art to which this disclosure pertains. In some cases, terms with commonly understood meanings are defined herein for clarity and/or for ready reference; thus, the inclusion of such definitions herein should not be construed to represent a substantial difference over what is generally understood in the art.

The term“circulating cell-free DNA” herein refers to the fragments of DNA/free DNA of 25 nucleotides of longer that are not associated with intact cells, released in the bloodstream after the apoptosis of cells.

A“single nucleotide polymorphism (SNP)" in the context of this invention refers to the presence of an alternative base in specific positions of DNA sequence, chosen to characterize DNA genotype of an individual with respect with another. Such a SNP is a geno-marker for donor material. The term “variant” herein refers to the presence of an alternative DNA base in a particular position of DNA sequence, referring to universally recognized reference sequence (GRCH37).

The term“allele frequency” herein refers to the relative frequency (percentage) of an alternative base in a particular position of the DNA, referring to universally recognized reference sequence (GRCH37), calculated on the total amount of sequences in that position.

The terms“approximately” and“about” herein refers to the range of the experimental error, which may occur in a measurement.

The terms“comprising”, “having”, “including" and “containing” are to be construed open-ended terms (i.e. meaning “including, but not limited to”) and are to be considered as providing support also for terms as“consist essentially of,“consisting essentially of,“consist of or“consisting of.

The terms“consist essentially of,“consisting essentially of are to be construed as semi-closed terms, meaning that no other ingredients which materially affects the basic and novel characteristics of the invention are included (optional excipients may thus be included).

The terms“consists of,“consisting of are to be construed as closed terms.

Description of the invention

It has been surprisingly found that by sequencing a panel of 94 single nucleotide polymorphisms (SNPs) from cell-free DNA from a sample obtained by a subject who is the recipient of an organ transplant, it is possible to differentiate between donor- derived cell-free DNA and recipient-derived cell-free DNA, and therefore to correlate the graft damage to the quantity of graft cell-free DNA released from damaged cells in patients. In particular, it has been demonstrated that the selection of a pre-determined number of SNPs, in particular of a panel of 94 SNPs reduce the possibility to obtain the false positive during the sequencing phase, that could be very high analyzing cfDNA, that is usually very fragmented.

This is based on the fact that, during acute rejection the recipient's immune system recognizes the transplanted organ as alien, it attacks the organ and causes cellular damage, thus resulting in dispersion of nuclear DNA from the donor organ cells in the recipient’s bloodstream.

This DNA, called donor“circulating cell-free DNA (cfDNA)", is distinctive for the donor and it can be recognized from the one of the recipient, physiologically present in a small quantity in recipient blood stream, because of normal necrosis of cells like endothelial ones.

The greater is the amount of donor cfDNA in the recipient blood, the greater is the graft damage caused by the recipient immune system, therefore calculating the amount of said cfDNA it is possible to correlate the graft damage to the level of graft DNA released from damaged cells in patient.

in particular, our method is based on a simple biological observation related to acute and chronic rejection: the immune response to allograft that damages graft cells, which release cellular residues in the bloodstream. These cellular residues include highly fragmented but detectable cell-free DNA in plasma. This damage induces identifiable and quantifiable chimerism in the blood of the recipient. The assumption is that, during acute rejection, the increased amount of molecules derived from the donor (and therefore derived from the implanted organ) are present in the blood stream of the recipient and increases with the increase of the extent of organ damage. Our goal is to directly interrogate the graft status, individuating and measuring the DNA fingerprint of dying allogenic cells in the cell-free DNA circulating in the recipient's plasma. If the fingerprint of DNA from the donated organ (compared with the recipient's genome) is characterized, then the presence and level of “donor cfDNA" can be monitored over time. Changes in organ status can be detected as changes in the donor cfDNA levels. The rationale for this approach arises from the observation that both acute and chronic rejection processes are associated with apoptosis of organ-specific cell types within the allograft. An increase of graft DNA from 1 % to 5% during rejection episodes has been reported in the circulation of stable heart transplant recipients, while an increase of graft DNA > 5% is indicative of a beginning of rejection.

From the experiments made in our laboratory we observed that sequencing an high number of SNP, such as different panels from 138 to 155 SNPs (see Table n. 1 ), we had an high number of false positive and we obtained a method with a lower precision and very laborious.

Based on said results we identified a method with a panel of 94 SNPs previously identified that was able to differentiate between two unrelated individuals.

In particular, the system of our method is composed by a panel of specific genetic markers, named SNPs (single nucleotide polymorphisms), and by an analysis algorithm specifically designed to discriminate two individuals, the donor and the recipient, using a specific identified panel of 94 SNPs and quantifying the cfDNA fraction from the donor.

For each experiment, we process a pair of genomic DNA, from the donor and from the recipient, and the total cfDNA from the recipient (individual 1 ) blood samples. The aim is to identify the presence of the donor (individual 2) cfDNA, using the genomic the DNA from recipient and from donor as reference.

In case of impossibility of process genomic DNA sample from donor, we process genomic DNA and total cfDNA, using the genomic DNA of the recipient as unique reference.

The advantages of the new method developed is that with at least one SNP of difference between the donor and the recipient it is possible to identify the presence of the cfDNA of the donor in a blood sample and to quantify it, without the need of donor genomic DNA as reference.

Furthermore, our method has a low limit of detection (1 %) and therefore it is possible to identify up to 1% minimum donor cfDNA fraction in total cfDNA (recipient + donor), and a very high sensitivity (>98%) and a low incidence of false negative (£1 %). Therefore the algorithm used in the method is able to correctly identify SNPs positions in the panel and to identify variants in that positions and to calculate the percentage of the donor-derived cell free DNA with respect to the recipient cell-free DNA, using the recipient genomic DNA as reference.

One aim of the present invention is therefore a method of monitoring the status of a transplanted organ in a subject, comprising:

a) providing cell-free DNA from a biological sample obtained from a subject who is the recipient of an organ transplant from a donor;

b) sequencing a panel of 94 single nucleotide polymorphisms (SNPs) from the cell- free DNA; and

c) calculating the quantity the donor-derived cell-free DNA in said biological sample. According to a preferred embodiment, the cell-free DNA isolated in step a) from the transplant recipient can be extracted from methods and protocols for DNA extraction well known in the art.

Said method are preferably selected from Qiagen QIAamp MinElute ccfDNA Kits, Maxwell RSC cfDNA Plasma Kit, Invitrogen MagMAX Cell-Free DNA Isolation Kit. In a preferred embodiment of the present invention the panel of 94 single nucleotide polymorphisms sequenced at point b) include the SNP listed in Table 3.

The selection of said panel of 94 SNP was made identifying the SNPs that are useful to identify uniquely one individual from another (45 SNP identified as iiSNP in forensic genetic (14)) and that have an high variability, easily to be mapped and with a fixation index of 0.06 (49 SNP, selected from https://alfred.med.vale.edu/, last accessed April 2017 ).

In a further preferred embodiment, the SNP sequencing is made with the primers listed in Table 4.

According to the present invention, the methods of sequencing DNA disclosed in step b) are selected from the next generation sequences method and instruments known in the art (i.e lllumina platforms, Ion-Torrent, MinlON, GeneRead, PacBio).

Preferably, said methods of sequencing the DNA are selected from amplicon method, capture method, enrichment method, pyrosequencing, incorporation of nucleotides, semiconductor technologies, nanopore real time reading and sequencing methods without PCR.

The bioinformatics analysis used in the method of the present invention is similar to the one used to identify somatic mutations in tissue. In particular, said bioinformatics analysis compares recipient genomic DNA with cfDNA, searching for variants in cfDNA. The presence of those variants can be associated only to sequencing errors or donor DNA coming from the graft cells destruction due to rejection. The analysis excludes errors identifying real variants in specific positions (SNPs).

According to a preferred embodiment, the method of the present invention is characterized in that step c) calculate the differences in SNP positions by identifying and comparing variations in specific nucleotide positions between the donor-derived cell-free DNA and the recipient-derived cell-free DNA using the recipient genomic DNA previously characterized, as reference.

In a further preferred embodiment the step c) of the method according to the present invention is based on the calculation of cfDNA donor fraction with the formula:

DFi%= (AF (SNPi) - AF(SNPi _rec )) / (AF(SNPi _don ) - AF(SNPi _rec )) *100 for each SNPi

and total DF% is calculated as: DF% = mean DFi% ± SD DF

where:

AF(SNPi) is the allele frequency of the SNP considered, calculated by the algorithm for the total cfDNA mixture (donor+recipient), and

AF(SNP _d on) and AF(SNP _rec ) is the allele frequency of the SNP considered in the donor and recipient cfDNA respectively, according to their genotype,

SNPs where the transplant, material and recipient are homozygous, but with different alleles, can then be used for future determination of graft cfDNA percentage.

Preferably, said algorithm is composed by 2 parts: the first part is aimed at identifying the 94 variants (SNPs panel), the second part is aimed at identifying genotype of donor and recipient, distinguishing cfDNA reads from 2 individuals, and calculate donor fraction.

The steps used in the first part are: • Alignment of the sequence to the GRCH37 human genome using a well- known algorithm selected from BWA, BWA-MEM or Novoalign.

• Pre-processing of the aligned sequence to reduce sequencing or alignment errors using dedicated open source algorithms selected from GATKv3.7 or Picard v2.7.

• Variant Calling using a known open source algorithm for variant identification in a case-control analysis selected from Mutect2, VarDictJava, GATK v3.7, Varscan v2.3.9 or Freebayes v0.9.10.

• Post-processing and feature extraction: during post-processing, variants obtained from each variant caller are merged into a single dataset, reporting all features in a tsv file format.

The second part of analysis is aimed at identifying genotype of donor and recipient, distinguishing cfDNA reads from 2 individuals, and calculate donor fraction.

The steps used in second part are:

• Variant filtering: wherein variants are classified as PASS or FILTER to identify TP and FP on the basis of quality of sequencing and coverage. In this step the 94 SNPs are identified (as variant annotation) and considered for further analysis.

• Genotype identification: in case of absence of donor DNA, the algorithm applies a supervised machine learning inference model, preferably based on the posterior Bayes probability (Naive Bayes), to classify each new SNP independently. The machine learning can be applied using a Python based package such as Scikit-Leam, Teano, Tensor Flow or Weka, STATA, Orange or R software for data analysis. • Donor fraction calculation: the donor fraction is calculated using the formula described above for each SNP. The mean of donor fraction of all informative SNPs is calculated. In case of absence of donor genotype, a weighted mean value for each genotype is calculated.

• Reporting of results: for each sample a report is create containing the SNP identified, the genotype, the presence of variants referable to donor and the amount of donor cfDNA (DF%).

A scheme of the entire algorithm is represented in Figure 2.

AF(SNP ) and AF(SNP _ire c) for each SNP found in the tested sample is calculated by the algorithm on the basis of the sequence obtained by the instrument.

In case of the availability of the genomic donor DNA, also the AF(SNP _id0 n) is calculated by the algorithm and the unique unknown variable in the formula is DF, otherwise AF(SNP _jdon ) is deduced using a supervised machine learning model using as attributes quality of the sequence in that position, AF(SNPi _rec ) and AF(SNP ) and relative genotypes.

Preferably, said supervised machine learning model is based on the posterior Bayes probability (Naive Bayes), to classify each new SNP independently.

In a preferred embodiment, in the method according to the present invention the transplanded organ is selected from heart, lung, kidney, liver or pancreas.

In particular, the method of the present invention is applicable to all the transplanted organs, because the presence of cfDNA is organ independent.

According to the method of the present invention, the number of SNPs sequenced will be sufficient to discriminate between recipient and donor alleles even in related individuals (excepting twins). In a preferred embodiment the biological sample used on the claimed method is a blood sample from the recipient, and a blood or tissue sample from donor, if present. According to the present invention, the diagnosis of the status of the transplanted organ in the subject is made evaluating the percentage and/or the quantity of the donor-derived cell-free DNA with respect to the recipient cell-free DNA, using the recipient genomic DNA as reference.

Preferably, a change in the percentage of the donor-derived cell-free DNA during a period of time is indicative of the status of the transplanted organ.

According to the method of the present invention, an increase in the levels of the donor-derived cell-free DNA over the time interval is indicative of transplant rejection, a need for adjusting immunosuppressive therapy, and/or a need for further investigation of the transplanted organ status.

Preferably, an increase in level of cfDNA comprised between 5 and 15% is indicative of probably presence of rejection, more preferably an increase of 10 to 25% is indicative of clinically significant rejection.

In a further embodiment, a decrease in the levels of the donor-derived cell-free DNA over the time interval is indicative of transplant tolerance, a need for adjusting immunosuppressive therapy, and/or a need for further investigation of the transplanted organ status; while a value under 5% or no change in the levels of the donor-derived cell-free DNA over the time interval is indicative of stable transplant rejection status and/or opportunity for adjusting immunosuppressive therapy.

Preferably, a decrease in level of cfDNA comprised between 10 and 25 % from the positive evaluation is indicative of efficacy of immunosuppressive therapy. More preferably, said decrease is comprised between 15 and 20%. According to the present invention, the biological samples may be taken from a transplant recipient over a period of time (i.e. over a time interval). The time at which samples are taken from the transplant recipient following the transplant event may vary. Samples may be taken from a transplant recipient at various times and over various periods of time for use in determining the status of the allograft according to the methods of the present disclosure. For example, samples may be taken from the transplant recipient within days and weeks after, about three months after, about six months after, about nine months after, or less than one year after the transplant event. Samples may be taken from the transplant recipient at various times before the one year anniversary of the transplant event, at the one year anniversary of the transplant event, or at various times after the one year anniversary of the transplant event. For example, at the one year anniversary after a transplant, samples may begin to be taken from the transplant recipient at month 12 (i.e. the one year anniversary of the transplant event) and continue to be taken for periods of time after this.

In some embodiments, a transplant recipient has biological samples taken for one to three consecutive months, starting at the one year anniversary of the transplant event (i.e. 12 months after the transplant event), providing a total of 4-6 samples for analysis taken over a three month time period, with samples being collected about every two weeks. In some embodiments, a transplant recipient has samples of bodily fluid taken once a week for one to three consecutive months, starting at the one year anniversary of the transplant event (i.e. 12 months after the transplant event), providing a total of twelve samples for analysis taken over a three month time period. The total duration of obtaining samples from a transplant recipient, as well as the frequency of obtaining such samples, may vary and will depend on a variety of factors, such as clinical progress. For example, a transplant recipient may have samples obtained for analysis of cell-free DNA for the duration of their lifetime.

Appropriate timing and frequency of sampling will be able to be determined by one of skill in the art for a given transplant recipient.

The methods of the present disclosure may be used to predict the risk of future transplant rejection such as, for example, the risk of rejection within the following 3-6 months after analysis of samples from the transplant recipient.

The methods of the present disclosure may also be used provide an assessment of the immune status of the transplant recipient, which may be used to guide decisions regarding immunosuppressive therapy in the transplant recipient. The methods of the present disclosure may also be used to guide decisions related to adjustment of immunosuppressive therapies being administered to the transplant recipient, with regard to the presence or level of rejection identified.

Additional benefits and/or uses of the methods of the present disclosure will be readily apparent to one of skill in the art.

In a further embodiment, the invention provides a method of monitoring the status of a transplant in a transplant recipient to evaluate immunosuppressive therapy where the method comprises quantifying the amount of the donor derived cell-free DNA in a sample from the transplant recipient at desired time points and adjusting the immunosuppressive therapy, e.g., adjusting the amount of immunosuppressive drug. Thus, the lowest dose of an immunosuppressive drug can be identified for that individual patient.

In a further aspect, the invention provides a method of monitoring the status of a transplant in a transplant recipient to evaluate reperfusion injury to the transplant, in such embodiments, the amounts of graft cfDNA are determined over a time course, for example, a time course of days or weeks up to a month following transplant.

Preferably, cfDNA is monitored over the first 7 days after engraftment. In a further embodiment, the method comprises determining the level of graft cfDNA over a course of seven days, or up to 30 days following transplant.

In a further aspect, the invention provides a method of monitoring the status of the transplanted graft with respect to Antibody-Mediated Rejection (AMR) in a transplant recipient to evaluate status of the graft. In such embodiments, the amounts of graft cfDNA are determined after 2 years from the transplanted. The presence of AMR will determine changes in the immunosuppressive therapy strategy.

A further embodiment of the present invention relates to a computer program product comprising a computer- usable medium having computer-readable program codes or instructions embodied thereon for enabling a processor to carry out the analysis and correlating functions as described above.

A further embodiment of the present invention is a computer medium comprising instructions which, when executed by a computer, cause the computer to carry the following steps:

(i) receiving raw data from SNP sequences obtained from the method of claim 1 ;

(ii) aligning said sequencing data with the referring genome;

(iii) preprocessing and preparing raw data for variant calling;

(iv) identifying SNP variants and calculating the genotype of the recipient for these SNP, identifiable as not mutated (0/0), mutated heterozygous (0/1 ) or mutated homozygous (1/1 ), through a germline variant calling process; (v) calculating the genotype of the cfDNA mixture and comparing it with the calculated recipient one through a somatic variant calling process;

(vi) calculating the donor fraction according to the following formula:

DFi%= (AF (SNP _| ) - AF(SNP _jrec )) / (AF(SNP _id0 n) - AF(SNP _irec )) x100

DF% = mean DF _{i %} ± SD DF

Preferably, said method is a computer readable list of instructions according to Figure 2 recorded to cause a computer to perform the steps reported above.

Examples

Example 1.

SNP panel design

Using our bioinformatics background and new knowledge developed in forensic genomics on individual discrimination (15), we designed a Single Nucleotide Polymorphisms (SNPs) panel, containing so-called individual identification SNPs. The goal of the SNP panel design solution was to be able to discriminate the presence of fragments of DNA of a different person from the one in which the blood sampling was done.

Before identifying the more appropriate number of SNP that lead to the reliable, precise and economic method according to the present invention, we started using a different number of SNP, variable from 138 to 155 SNPs, having as principal characteristic an high variability and a difference homozygous/homozygous between donor and recipient.

As reported in Table n. 1 this lead to a method with very low precision and accuracy (from 40-50%) and an high number of false positive.

Table 1 . Results of simulation of different concentrations of cfDNA of individual 1 and 2 and a panel of SNP from 138-155: statistics. cfDNA Concentration: concentration of an individual's cfDNA, in the total cfDNA of the two individuals.

NUM (SNPs): number of SNP identified;

PRECISION: or positive predictive value, proportion of positive call, really positive ACCURACY: proportion of true results among the total number of cases examined.

As it can be observed form Table n. 1 , the method based on a panel of 138-155 SNPs was not sufficient to identify clearly between the 2 individuals, due to the high number of false positive detected. Therefore, a lower number of more precise SNP should be used to increase performance of the method.

Based on the non-optimal results obtained with a panel of an higher number of SNPs, we made the same experiments with an identified panel of 94 SNPs previously identified as suitable for differentiating between any two unrelated individuals. We choose SNPs with a high frequency of mutation in general population (independently of gender or ethnicity), to be sure to find a minimal percentage of SNPs of the panel, different in every combination of two individuals. The rationale was to discriminate cfDNA of the donor through identification of SNPs different from the recipient. The obtained panel is composed by 94 SNPs. This combination of SNPs has an extremely low probability for two unrelated individuals having identical genotypes. All SNPs have an average heterozygosity >0.4 and /or a Fst>0.06. Fixation index (Fst) is a measure of population differentiation due to genetic structure, the values range from 0 to 1. A zero value implies complete panmixia (random mating); that is, that the two populations are interbreeding freely; a value of one implies that all genetic variation is explained by the population structure, and that the two populations do not share any genetic diversity. The chosen value makes this a universally applicable panel irrespective of ethnicity or ancestry (Table 1 ). The frequencies of SNPs were verified in 1000Genomes (http://www.internationalqenome.orq/1000-genomes- browsers/, last accessed 19/1 st/2019).

Table 2. Results of simulation of different concentrations of cfDNA form individual 1 and 2 and a panel of 94 SNPs: statistics.

NUM SNPs: number of SNP identified;

PRECISION: or positive predictive value, proportion of positive call, really positive ACCURACY: proportion of true results among the total number of cases examined.

The results obtained with this new panel were very promising in terms of accuracy and precision (95-100%), with an high accuracy and precision with respect to the panel of 138-155 SNPs reported in Table 1 .

Table 3. List of 94 SNPs panel.

(https://www.ncbi.nlm.nih.aov/Droiects/SNP/. last accessed 19/1 st/2019), Fst, heterozigosity and reference populations for calculation from ALFRED (https://alfred.med.vale.edu/ , last accessed 19/1 st/2019).

Each of the SNPs in the panel reported in Table 3 can be present in the donor and recipient as not mutated (0/0 or reference homozygous), mutated homozygous (1/1 or alternative homozygous), or heterozygous 0/1 or 1/0.

In order to discriminate uniquely the 2 unrelated individuals, only combinations specifically different from recipient and donor such as 0/0 - 0/1 or 0/0 - 1/1 are useful for the analysis.

Analysis algorithm

An ad hoc analysis pipeline of NGS data was developed using software packages developed in Unix Bash, including several bioinformatic tools and Python scripts. The pipeline is based on the contemporary calculation of genotype (GT) of samples analyzed, using a germline and a somatic pipeline for NGS identification of variants. The germline pipeline is focused on identification of genotype in genomic DNA of individual 1.

Pipeline for parallelized germline genotype identification and somatic identification of differences is reported below.

• Alignment: FASTQ files coming out from l!lumina sequencer are aligned using BWA-MEM v 0.7.15 tool.

• Pre-processing: SAM file obtained are then converted in BAM files and sorted using picard v2.7. Read Group are added. Target Amplicon analysis includes a data-cleaning step to reduce False Positive and False Negative using Base Quality Score Recalibration using GATK v3.7 and BAM file are indexed using picard v2.7.1

• Variant Calling: this step involves the usage of three variant callers to increase sensitivity. Somatic pipeline performs a case-control analysis using Mutect2, VarDictJava e VarScan v2.3.9, where case is cfDNA and control is the genomic DNA. All variant callers are used without filters. The genomics DNA genotype is also called by three variant callers (GATK v3.7, Varscan v2.3.9 and Freebayes v0.9.10) for germline analysis. The genomic DNA gives the genotype information about individual 1 , to be used in reporting phase. All variant callers are used with default parameters.

• Post-processing and feature extraction: during post-processing, variants obtained from each variant caller are merged into a single dataset, reporting all features in a tsv file format.

• Variant filtering: variants are classified as PASS or FILTER to identify TP and FP on the basis of quality of sequencing and coverage. The 94 SNPs are identified (variant annotation) and considered for further analysis.

• Genotype identification: in case of absence of donor DNA, the algorithm applies a supervised machine learning inference model based on the posterior Bayes probability (Naive Bayes), to classify each new SNP independently. Weka 3.8.1 was used for calculation.

• Donor fraction calculation: donor fraction is calculated using the formula described above for each SNP. The mean of donor fraction of all informative SNPs is calculated. In case of absence of donor genotype, a weighted mean value for each genotype is calculated.

• Reporting of results: For each sample a report is create containing SNP identified, genotype, presence of variants referable to donor and amount of donor cfDNA (DF%).

Materials and Methods cfDNA extraction

Whole blood (6 ml) was collected in blood collection tubes (BD tubes with EDTA). Blood was centrifuged at either 4,000 rpm (larger volumes) or 13,000 rpm (smaller volumes) at 4°C for 15 min, the plasma removed, and centrifuged again at 13,000 rpm at 4°C for 15 min and the supernatant frozen at -80°C until used.

In 10 blood samples (3ml) from 10 unrelated voluntary donors, plasma was separated and cfDNA was purified using both a semi-automated or completely automated method: Qiagen (QIAamp Circulating Nucleic Acid Kit) and Promega (Maxwell® RSC ccfDNA Plasma Kit), according to both protocols, respectively. cfDNA concentration was measured using the Qubit dsDNA HS Assay kit (Invitrogen, Life Technologies, CA, USA) on a Qubit 2.0 Fluorometer (invitrogen, Life Technologies, CA, USA) according to manufacturer’s instructions.

Libraries preparation and sequencing

The cfDNA assay is based on targeted amplification of DNA regions harboring 94 SNPs. cfDNA extracted from 1 mL plasma or reference materials was pre-amplified in a single multiplex reaction with 270 primer (see Table 2) pairs for 15 cycles. Preparation protocol used were lliumina Truseq Custom Amplicon for target sequencing. Index sequences and lliumina sequencing adapters were added to each sample DNA by PCR, and the sample was qualified and quantified by Qubit 2.0 Fluorometer (Invitrogen, Life Technologies, CA, USA) and an Agilent Technologies 2100 Bioanalyzer (Agilent, Santa Clara, CA), according to preparation protocols. Up to 24 amplified samples were pooled in equimolar amounts, purified using Agencourt AM Pure XP beads (Beckman Coulter, Brea, CA), and sequenced on an lliumina MiSeq instrument. Table 4 reports SNP primers of target amplicons used for experiments. Some SNPs are covered by more than one target to ensure the complete coverage of the panel.

Table 4. Target (primers) used in experimental setting.

Synthetic study design

SNP panel validation

Using Minor Aliele Frequency (MAF) in general population reported in 1000Genome (1000G) project to calculate the probability of mutation of each allele of each of the 94 SNPs in our panel (Table 3), we created 10000 simulated genotypes (0/0, 0/1 , 1/1 ), under the hypothesis of independence of mutation between SNPs:

p{$np 1 =0/1 ,snp2=0/1 )=p (snp 1 =0/1 )p (snp2=0/1)

that is the joint probability of the presence of two SNPs is equal to the product of the single probability.

Calculations were made for each SNP in each simulated sample as:

P{SNPF0/0)= p(AF(SNP,)) x p(MAF1000G (SNP/)) where AF(SNP, =0/0) =0

p(SNPj=0/1)= p(AF(SNP,)) x p(MAF1000G (SNP _/ )) where AF(SNP, =0/1) =0.5 p(SNPj=1/1)~ p(AF( SNP,)) x p(MAF1000G(SNP _/ )) ² where AF (SNP, =1/1) =1 Our aim was to estimate the probability of each SNP to be mutated in the simulated population, in a genotype useful for our evaluation.

Table 5. SNPs list. 1000G MAF and calculated frequencies according to the formula above.

MAF is the allele frequency of the minor represented base variant in the position of each SNP in a sequenced population.

P(SNP) in the population was calculated according the Hardy Weinberg Theory as: p(0/0)={ 1 -MAF 10OOG) ²

p(0/1 )=2x(1 -MAF 1000G)xM AF 1000G

p(1 /1 )=MAF1000G ²

Limit of detection; in silico evaluation

We created 10 simulation FASTQ files mimicking the characteristics (coverage, quality, error rate) of the raw files coming from the Miseq IIEumina instrument, using Simulvar v1.0 tool. The choice of the features was to cover the spectrum of possible output of the instrument (low coverage, high error rate to high coverage, low error rate). From the 10 initial files we created 5 pairs of BAM files with the 94 SNPs genotype obtained from 5 (randomized) of the 10000 genotypes simulated from general population allele frequency.

The aim was to mimic the different concentrations (in a known percentage between 0.1 % and 5%) of donor in recipient cfDNA. Obtained files were analyzed using the analysis algorithm disclosed above.

Limit of detection: in vitro evaluation

In order to confirm the previous result, we created 6 control samples, with a predefined quantity of individual 1 cfDNA in an amount individual 2 (1 %, 5%, 10%, 25%, 50%, 100%), to test the limits of detection of our test methodology in a calculated, progressively diluted mixture of real couples of cfDNA samples.

Pilot study design

An early stage pilot study was performed on 23 sample series from heart transplanted patients undergoing EMB monitoring, to confirm the concept and check results. We tested the efficacy to precisely detect cfDNA of the donor (individual 1 ) in recipient (individual 2) blood.

Currently we developed an operating fab prototype, based on the specific designed genetic markers panel and the innovative algorithm.

Panel design validation

The process works correctly individuating the geno-marker of the individual 1 /individual pair of cfDNA. In particular the probability to individuate at least one SNP with an informative genotype is reported in Table 6 and it is a probability near 1 both in case of homozygosis (0/0 and 1/1 ) and in case of heterozygosis (0/1 and 0/0).

As for design of the panel, at least the presence of 1 SNP genotype informative combination is sufficient to discriminate the DNA of different individuals.

Table 6. Probability of incidence of paired genotypes in a simulated 10000 individual population. Genotype individual 1 Genotype individual 2 Probability of incidence

o/o | 1/1 0.9999

0/0 0/1 0.9945 o/i | o/o 0.9945

1/1 0/0 0.9999

0/0 reference homozygous, 1/1 alternative homozygous, 0/1 or 1/0 heterozygous

Limit of detection: in silico evaluation

We identified a limit of detection of 1 % of the developed method {minimum quantity of individual 1 cfDNA we can identify in the total of cfDNA, containing individual 1 + individual 2 cfDNA). In real samples, we expect an increase of the donor DNA fraction within the total cfDNA concurrent with the presence of rejection and proportional to the degree of rejection. Table 7 and Figure 3 represent the results of this evaluation. Statistics reported refers to the capacity of the algorithm to correctly identify SNPs in the reported fraction of donor DNA.

Table 7. Results of simulation of different concentrations of individual 1 and 2: statistics and 94 SNP.

DETECTION RATE: % of variant identified on the total 94 SNP panel;

SENSITIVITY: rate of true positive identified on the total of called positive;

MISSING RATE: number of false negative in the total of called negative by the algorithm.

Limit of detection: in vitro evaluation

We identified a limit of detection of 1 % (minimum quantity of individual 2 cfDNA we can identify in the total of individual 1 cfDNA), coherent with simulated results. Limit of the evaluation was the technical difficulty of preparing the samples with a concentration <1 %. Results are reported in Figure 4. The algorithm was able to identify the correct individual 2 fraction in in vitro mixture, with an error of 0.01 %. Machine learning process results for prediction of donor genotype GT

To test the method accuracy and performance we used a machine learning supervised approach (Naive Bayes) on simulated data as training set and real data (with known donor GT for each SNP in the panel) as test set. In Table 8 are reported results of the performance of the trained model. Figure 5 shows ROC (Receiver Operating Characteristic) curve, area under the curve was 0.991 .

Table 8. Real GT VS GT predicted by the trained machine learning model.

Predicted GT

Reai GT 0/0 0/1 1/1

0/0 95,7 % 2.5 % 0 %

0/1 3.0 % 97 % 3.7 %

1/1 1.2 % 0.5 % 96.3 %

Finally, we applied the machine learning trained model on synthetic dataset (unknown donor GT, but known donor cfDNA fraction).

In Table 9 are reported results of the machine learning model.

Table 9. Donor fraction percentage calculated using GT predicted by the trained machine learning model on the synthetic dataset VS known percentage for same samples.

Earlv stage pilot study

To test our pipeline in a real setting, gold standard results (derived from endomyocardial biopsy EMB) classified according to ISLHT standards (Stewart S, et al. Revision of the 1990 working formulation for the standardization of nomenclature in the diagnosis of heart rejection. J Heart Lung Transplant. 2005;24:1710-20) were compared with our test in a cohort of 4 sample, in which donor blood were genotyped and donor GT used to calculate expected donor fraction. Calculated donor fraction was estimated using donor genotype inference using machine learning.

In Table 10 are reported Donor Fraction percentage calculated from the algorithm, preliminary classification and comparison with gold standard.

Table 10. Donor fraction calculated with known GT and calculated GT on a real dataset.

Finally, we applied our test on a cohort of 4 samples. Donor genomic DNA was not available, and we used the machine learning approach to calculate donor genotype. Donor fraction was then calculated with inferenced donor genotype. Results of the test and comparison with gold standard is reported in Figure 6. For each patient 2 samples were collected, TO after 15 days and T1 after 21 days from heart transplant. Conclusions.

Our system is useful to identify and measure correctly cfDNA of an individual in the total cfDNA extracted from a blood sample of another individual, starting from a percentage of 1 % of concentration, with a very high sensitivity (98%).

In the hypothesis of transplant, our system can recognize the presence of unrelated donor cfDNA in the total amount sampled from the receiver and quantify it in order to identify the presence of rejection.

It can be used also to monitor during time after transplant the rejection, and the immunosuppressive therapy efficacy. References

1. Miller CA, Fildes JE, Ray SG, Doran H, Yonan N, Williams SG, Schmitt M.

Non-invasive approaches for the diagnosis of acute cardiac allograft rejection. Heart. 2013 Apr;99(7):445-53.

2. Costanzo MR, Dipchand A, Starling R, et al. The International Society of Heart and Lung Transplantation Guidelines for the care of heart transplant recipients. J Heart Lung Transplant. 2010; 29: 914-956.

3. Saraiva F, Matos V, Gongalves L, Antunes M, Providencia L. Complications of endomyocardial biopsy in heart transplant patients: A retrospective study of 21 17 consecutive procedures. Transplant Proc.201 1 ;43:1908-1912.

4. Stehlik J, Starling RC, Movsesian MA et al. Utility of longterm surveillance endomyocardial biopsy: a multiinstitutional analysis. J Heart Lung Transplant 2006; 25(12):1402-1409.)

5. Deckers JW, Hare JM, Baughman KL. Complications of transvenous right ventricular endomyocardial biopsy in adult patients with cardiomyopathy: a seven-year survey of 546 consecutive diagnostic procedures in a tertiary referral center. J Am Coll Cardiol 1992; 19(1 ):43-47.

6. Pophal SG, Sigfusson G, Booth KL et al. Complications of endomyocardial biopsy in children. J Am Coll Cardiol 1999; 34(7):2105-2110

7. De Vlaminck I, Valantine HA, Snyder TM, et al. Circulating cell-free DNA enables noninvasive diagnosis of heart transplant rejection. Sci Transl Med. 2014;6:241 ra77.

8. Snyder TM, Khush KK, Valantine HA, Quake SR. Universal noninvasive detection of solid organ transplant rejection. Proc Natl Acad Sci U S A.201 1 ;108:6229-34. Oellerich M, Kanzow P, Beck J, et al.. Graft-derived cell-free DNA (GcfDNA) as a sensitive measure of individual graft integrity after liver transplantation. Am J Transplant. 2014; 14(suppl !):874. - Sigdel TK, Vitalone MJ, Tran TQ, et al.. A rapid noninvasive assay for the detection of renal transplant injury. T ransplantation. 2013;96:97-101. . Beck J, Bierau S, Balzer S, et al.. Digital droplet PCR for rapid quantification of donor DNA in the circulation of transplant recipients as a potential universal biomarker of graft injury. Clin Chem. 2013;59:1732-1741. . Grsko ic M et al. Validation of a Clinical-Grade Assay to Measure Donor- Derived Cell-Free DNA in Solid Organ Transplant Recipients. J Mol Diagn. 2016;18:890-902.

. Gordon PM, et al. An Algorithm Measuring Donor Cell-Free DNA in Plasma of Cellular and Solid Organ Transplant Recipients That Does Not Require Donor or Recipient Genotyping. Front Cardiovasc Med. 2016;3:33. eCollection 2016. . Pakstis AJ, et al. SNPs for a universal individual identification panel. Hum Genet. 2010;127:315-24. . Kidd KK, et al. Expanding data and resources for forensic use of SNPs in individual identification. Forensic Sci Int Genet. 2012;6(5):646-52.