EP4211265A1

EP4211265A1 - Methylation detection assay

Info

Publication number: EP4211265A1
Application number: EP21773407.8A
Authority: EP
Inventors: Guy Van Camp; Marc Peeters; Ken OP DE BEECK; Arvid SULS; Isabelle Neefs; Janah Vandenhoeck
Original assignee: Universitair Ziekenhuis Antwerpen; Universiteit Antwerpen
Current assignee: Universitair Ziekenhuis Antwerpen; Universiteit Antwerpen
Priority date: 2020-09-10
Filing date: 2021-09-10
Publication date: 2023-07-19
Also published as: CA3193631A1; AU2021339002A1; AU2021339002A9; WO2022053637A1

Abstract

The present invention in general relates to the field of DNA methylation detection. More in particular the current invention provides a method for DNA methylation detection using a combination of methylation-specific restriction enzymes (MSRE) and single molecule Molecular Inversion Probes (smMIPs). The present invention also provides uses of said DNA methylation detection method such as but not limited to the field of cancer diagnosis and/or monitoring.

Description

METHYLATION DETECTION ASSAY

FIELD OF THE INVENTION

BACKGROUND TO THE INVENTION

Each year, an estimated 8.2 million people die of cancer worldwide. With appropriate detection methods and treatment, many of these deaths would be avoidable. Due to the high incidence and mortality rates, early and accurate diagnosis is paramount for a quick and adequate treatment of patients. Until recently, no truly non-invasive diagnostic methods for the detection of cancer existed. An attractive novel method is the detection of abnormally expressed biological markers manifested during carcinogenesis in so called “liquid biopsies”. Liquid biopsy is a technique in which non-solid biological tissues such as urine, stool or peripheral blood, are sampled and analyzed for disease diagnosis.

The analysis of Circulating tumor DNA (CtDNA) in liquid biopsy samples of cancer patients is not new and has been performed in the past. However, until now, a strong focus existed on the detection of tumor specific mutations, which has several limitations. One of the problems with liquid biopsy nucleic acid biomarkers is the limited sensitivity for early detection. Indeed, in early stages of carcinogenesis, many tumor types have low concentrations of CtDNA. Sensitivity can be increased by measuring a multitude of markers simultaneously. The use of methylation markers instead of mutation markers has many advantages and is understudied. However, to date, no efficient techniques exist allowing multi-region methylation analysis in plasma or other types of liquid biopsies.

Most methylation assays currently used are based on bisulfite conversion of DNA, which allows for identification of methylated C’s, as they are protected from conversion to T. A major challenge in bisulfite conversion methods is the degradation of DNA caused by bisulfite. Even carefully controlled conditions for complete conversion lead to the degradation of about 90% of the incubated DNA. While this is not a problem when large amounts of DNA are available, it is unacceptable for liquid biopsies, as they yield very limited amounts of DNA. In addition, careful control of the reaction parameters is often problematic, which has led to a poor reputation of bisulfite conversion in terms of robustness and reliability. Therefore, liquid biopsy analysis requires a bisulfite-free analysis. The alternatives for bisulfite conversion methods are immunoprecipitation methods and methods based on methylation-specific restriction enzymes (MSRE). Immunoprecipitation methods such as MeDIP or Methyl-cap are based on me-CpG recognizing antibodies or methyl- binding proteins. Unfortunately, the antibody or methyl binding proteins are imperfect, and introduce false positive results and unwanted bias towards certain regions of the genome.

Therefore, there is a continuing need in the art for novel methods for DNA methylation detection, which are particularly suitable for liquid biopsies containing low amounts of DNA. In the present invention, we have surprisingly found that by combining methylation-specific restriction enzyme-assisted digestion of DNA for obtaining a set of DNA fragments, and single molecule Molecular Inversion Probe-assisted detection of said DNA fragments, a very powerful method for DNA methylation detection is obtained. This method cannot only suitably be used in the field of cancer diagnostics and/or monitoring, but is applicable to all fields in which DNA methylation detection is relevant. In particular, this novel approach has the potential to increase sensitivity up to 1000-fold compared to current methylation detection methods, while reducing the cost more than a 100-fold.

SUMMARY OF THE INVENTION

In a first aspect, the present invention provides a method for the detection of DNA methylation comprising the steps of: a) providing a sample comprising one or more DNA molecules; b) digesting said one or more DNA molecules by using one or more methylation-specific restriction enzymes (MSRE), thereby obtaining a set of DNA fragments; c) capturing and amplifying said set of DNA fragments using single molecule Molecular Inversion Probe technology (smMIP); thereby obtaining a set of amplified DNA fragments; d) detecting said amplified DNA fragments; thereby detecting DNA methylation.

In a further aspect, the present invention provides a method for the detection of DNA methylation comprising the steps of: a) providing a sample comprising one or more DNA molecules; b) digesting said one or more DNA molecules by using one or more methylation-specific restriction enzymes (MSRE), thereby obtaining a set of DNA fragments; c) capturing and amplifying said set of DNA fragments using one or more single molecule Molecular Inversion Probes (smMIPs); each spanning at least one methylation site in the DNA; thereby obtaining a set of amplified DNA fragments; wherein each smMIP comprises one or more specific alignment tag(s); d) detecting said amplified DNA fragments; thereby detecting DNA methylation. In a specific embodiment, said smMIP technology comprises the use of one or more smMIPs each spanning at least one methylation site in the DNA.

In another particular embodiment, each of said smMIPs (further) comprises:

- a specific extension probe, which specifically hybridizes to a first site next to a methylation site;

- a common backbone, which is not capable of hybridizing to said DNA molecules, and

- a specific ligation probe, which specifically hybridizes to a second site next to said methylation site, and

- a specific molecular tag, in particular an alignment tag.

In a specific embodiment, said alignment tag(s) comprise at least four random nucleotides.

In another embodiment, said alignment tag(s) is/are located between the common backbone and the extension probe or between the common backbone and the ligation probe, or both.

In a particular embodiment, said smMIP comprises two alignment tags each comprising at least 4 nucleotides located between the common backbone and each of said probes, or one alignment tag comprising at least 8 random nucleotides between the common backbone and said ligation or extension probe.

In yet a further embodiment, step c) of the method of the present invention comprises the steps of: c1) hybridizing said one or more smMIPs to said DNA fragments; c2) performing an extension reaction from the extension probe(s) across the methylation site(s) in the direction of the ligation probe(s); c3) performing a ligation reaction using the ligation probe(s), thereby obtaining a circular DNA fragment; c4) performing an exonuclease treatment to digest non-circular DNA fragments; c5) performing an amplification reaction using common primers capable of hybridizing to said common backbone; thereby obtaining a set of amplified DNA fragments.

In yet a further embodiment of the present invention, each of said amplified DNA fragments comprises a specific extension probe sequence, a methylation site sequence, a specific ligation probe sequence, sequencing adaptors, primer binding sites, and a specific molecular (in particular alignment) tag sequence; or the complement thereof.

In a further embodiment, step d) of the method of the present invention comprises the step of detecting said amplified DNA fragments using next generation sequencing. In another particular embodiment, said method comprises the use of at least 2 methylationspecific restriction enzymes.

In a further embodiment, said methylation-specific restriction enzymes are only capable of digesting unmethylated DNA regions.

In another embodiment, said one or more MSRE are selected from the list comprising: Hpall, HinP11, Acil, HpyCH4IV and combinations thereof.

In a further embodiment, the method of the present invention further comprises one or more control smMIPs which do not span a CpG region and/or one or more control smMIPs which span a CpG region that does not include a restriction site for said MSRE.

In yet a further embodiment, said sample is a solid or liquid biopsy sample from a subject, in particular a liquid biopsy.

In a further aspect, the present invention provides the use of a combination of methylationspecific restriction enzymes (MSRE) and single molecule Molecular Inversion Probe technology (smMIP) in the detection of DNA methylation in a biological sample.

In a further aspect, the present invention provides the use of a combination of methylationspecific restriction enzymes (MSRE) to obtain a set of DNA fragments; and single molecule Molecular Inversion Probes (smMIPs) to capture and amplify said set of DNA fragments; in the detection of DNA methylation in a biological sample.

BRIEF DESCRIPTION OF THE DRAWINGS

Fig. 1 : shows the principal of methylation-specific restriction digestion (MSRE)

Fig. 2 : shows a typical smMIP design according to the present invention.

Fig. 3 : shows a schematic overview of the smMIP technology according to the present invention.

Fig. 4: Efficiency of the subset of smMIPs. Top 52 smMIPs are 100x more efficient compared to all smMIPs, improving the total efficiency of the assay

Fig. 5: Repeatability of the assay. Panel A and B show the samples that were not digested by the MSREs (‘uncut’). Panel C and D show the samples that were digested by the MSREs (cut). Panel A and C are non-CpG control smMIPs and Panels B and D are CpG smMIPs. Samples 1-10 are DNA samples originating from human blood. Sample 11 is DNA originating from an artificially methylated blood sample and sample 12 is DNA isolated from a CRC cell line. The relative coverage of the smMIPS per sample is calculated as following: A) the sum of the counts in an uncut sample for non-CpG smMIPs, divided by the total counts for all non-CpG smMIPs in the uncut samples B) the sum of the counts in an uncut sample for CpG smMIPs, divided by the total counts for all CpG smMIPs in the uncut samples C) idem as A, but for cut samples and D) idem as B but for cut samples.

Fig. 6: Correlation between the relative counts per smMIP for uncut samples of two independent runs. The average relative coverage per smMIP is shown for run A and run B.

Fig. 7: Correlation between the relative counts per smMIP for cut samples of two independent runs. Panel A shows the relative counts per smMIP in a low methylated, cut sample and Panel B shows the relative counts per smMIP in a highly methylated, cut sample.

Fig. 8: Counts per smMIP in different digested control samples. X-axis: 32 CpG smMIPs;

Y-axis: absolute counts per smMIP per sample. Example of a nano v2 run on the MiSeq.

Fig. 9: Results of the qPCR. Different sample conditions for a complete MSRE digestion were tested by qPCR. Distinct input amounts of DNA (ng) were digested in distinct end volumes (pL). Thereafter, fragmentation was examined by qPCR with primer pairs around the MSRE recognition sites. Uncut samples were included as reference. Primer3 was left out of the analysis as the melt-curves showed this primer is poorly performing.

Fig. 10: Comparison of normal DNA (healthy blood) and colorectal cancer tissue DNA. smMIPs designed for capture of hypermethylated sites are shown in grey. smMIPs that capture non CpG sites in black. Blood samples (B) are shown on the left part of the figure, colorectal cancer samples (CRC) on the right. Comparing the bars, a clear difference between normal and cancer can be made.

Fig. 11 : Principal component analysis (PCA) plot. Blood samples (circles) are clustered on the left side and colorectal cancer samples (triangles) on the right side. There is a very clear distinction between the two groups.

Fig. 12: Absolute read counts for non CpG and CpG smMIPs in MSRE-digested samples.

In this experiment, all samples were MSRE digested and enriched using a selection of +/- 800 efficient smMIPS. The absolute read counts are given for gDNA extracted from normal blood samples (left) and compared to gDNA extracted from cancer cell lines (right). It is clear that a perfect distinction can be made based on the CpG smMIPs (light bars). nonCpG smMIPs are also shown (dark bars) and theoretically should not differ between cancer and normal samples

Blood = gDNA extracted from whole blood from healthy volunteers. Methylated blood = one of these blood samples, in vitro methylated with the CpG methyltransferase Sss.l enzyme. CL = gDNA extracted from cell lines. CRC= colorectal cancer, BRCA= breast cancer, LUNG= lung cancer, HNSC= head and neck squamous cell cancer, PANC = pancreatic cancer and PRCA= prostate cancer. Fig. 13: Principal component plot of the first two PC’s.

Based on the experiment displayed in figure 12, a subselection of the 500 most discriminating smMIPS was performed and the first two principal components (PCs) were plotted. Blood samples are shown as circles, cancer samples as triangles and an artificially methylated blood sample as a square. The figure shows that a perfect separation exists between blood samples and cancer cell lines.

Fig. 14: Proof of principle showing benefits of removing PCR duplicates in data analysis. Here, the results of analyses with and without duplicate removal are compared using the same raw input data (see equation 1 and 2). Proportion ratios of reads per sample (Panel A) or per smMIP (Panel B) are displayed. Proportion ratios between 0.95 and 1.05 are in white, one interval further in light grey and others dark grey. Each bar indicates the number of samples (A) or smMIPs (B) within the specified interval of proportion ratios. Intervals are half open, upper limit included..

Fig. 15: Refinement of the proportion analysis; this time comparing ratio per smMIP per sample, divided by the reads per smMIP (Panel A) or divided by the reads per sample (Panel B). This analysis is similar to the one shown in fig. 14, albeit more refined (see also equations 3 and 4). The conclusion from this analysis is that a molecular tag reduces technical noise within an experiment. Proportion ratios between 0.95 and 1.05 are in white, one interval further in light grey and others in dark grey. Each bar indicates the number of smMIPs per sample within the specified interval of proportion ratios. Intervals are half open, upper limit included.

DETAILED DESCRIPTION OF THE INVENTION

As already detailed herein above, the present invention relates to a method for the detection of DNA methylation comprising a combination of methylation-specific restriction enzyme-assisted (MSRE) digestion of DNA thereby obtaining a set of DNA fragments, followed by single molecule Molecular Inversion Probe-assisted (smMIP) amplification and detection of said DNA fragments.

Accordingly, in a first aspect, the present invention provides a method for the detection of DNA methylation comprising the steps of: a) providing a sample comprising one or more DNA molecules; b) digesting said one or more DNA molecules by using one or more methylation-specific restriction enzymes (MSRE), thereby obtaining a set of DNA fragments; c) capturing and amplifying said set of DNA fragments using single molecule Molecular Inversion Probe technology (smMIP); thereby obtaining a set of amplified DNA fragments; d) detecting said amplified DNA fragments; thereby detecting DNA methylation.

More in particular, the present invention provides a method for the detection of DNA methylation comprising the steps of: a) providing a sample comprising one or more DNA molecules; b) digesting said one or more DNA molecules by using one or more methylation-specific restriction enzymes (MSRE), thereby obtaining a set of DNA fragments; c) capturing and amplifying said set of DNA fragments using one or more single molecule Molecular Inversion Probes (smMIPs) each spanning at least one methylation site in the DNA; thereby obtaining a set of amplified DNA fragments; wherein each smMIP comprises one or more specific alignment tag(s); d) detecting said amplified DNA fragments; thereby detecting DNA methylation.

As used herein and unless otherwise specified, the term “alignment tag” may also refer to “alignment sequence”, “single molecule tag”, “molecular tag”, or “tag” and these terms can be used interchangeably throughout the application. At any instance, the ‘alignment’ tag is meant to be a nucleotide sequence which allows the alignment of amplified DNA fragments into particular consensus read sequences. Accordingly, using sequencing tools such as next generation sequencing, the alignment tags allow sequence reads containing the same alignment tag (and thus originating from the same capture) to be merged into one consensus read sequence. Accordingly, duplicate reads can be identified and filtered out thereby removing PCR and sequencing artifacts and enabling detection of low-frequency and sub clonal genetic variation. This results in an ultra-sensitive targeted sequencing method exhibiting the specificity and multiplexing advantage of the MIPs and the quantitation ability of the “single molecule tagging” approach. In a particular embodiment, the alignment tag described in the present invention does not function as a detection probe. More specifically, tag sequences as described in for example US2006292585 or WO2012112970 are recognized by an array of tag probes that are complementary to the tag sequences in the MIPs and facilitate detection of PCR amplified sequences. This is in clear contrast to the alignment tags described herein, which serve the purpose of aligning amplified DNA fragments into particular consensus read sequences avoiding duplicate reads.

The inventors have thus developed a method to enrich and assess specific DNA loci for methylation content using a protocol that combines MSREs and smMIPs. The technology has the advantage to increase sensitivity compared to current existing technologies, while reducing costs significantly.

Selection of relevant detection biomarkers allows the assay to be either pan-cancer or cancer specific. This approach has the potential to increase sensitivity 100 to 1000-fold compared to current technologies while reducing the cost more than a 100-fold. The current invention is a technique for multiplex analysis of a selected number of methylation sites in the genome, allowing sensitive analysis of even small quantities of DNA at an affordable cost. The field of liquid biopsies for the early detection of cancer is booming, with large investments worldwide in the biotech and pharma sector. The analysis and importance of the methylome is also an expanding field, not only in oncology but also well beyond. Methylation of DNA in particular and epigenetics in general are linked to a wide range of diseases and health conditions. As such, the potential application of this technology is broad, with a first focus on oncology diagnostics.

This technology has many applications, such as the detection of methylated CpG signatures in cancer research. As this technology will be much more sensitive compared to current protocols, it has also applications in the detection of methylated DNA fragments in liquid biopsies. Accordingly, in a further embodiment, said sample is a solid or liquid biopsy sample from a subject, in particular a liquid biopsy. Alternatively, said sample may also be derived from a cell line. As stated already, this novel detection assay is specifically suitable as a cancer detection assay, which allows for high resolution methylation detection in tissues, plasma or other biological matrices (e.g. blood, urine, saliva ...) of cancer patients.

MSREs have been used for a very long time for analysis of methylation in specific regions of the genome, and more recently, also for genome wide analysis. In contrast to antibodies, restriction enzymes are ultimately specific and predictable in their action. In the presented protocol, we use MSREs to digest genomic DNA. These MSREs are very specific and cut the DNA only when it is unmethylated and not when it is methylated, as illustrated in figure 1 . As used herein, MSRE may refer to “Methylation-Specific Restriction Enzymes” or “Methylation-Sensitive Restriction Enzymes” and the terms can be used interchangeably. Using MSREs instead of bisulfite, which is a widely used compound used in methylation analysis, has strong advantages. The main advantage is that MSREs do not degrade DNA, in contrast to bisulfite.

As used herein, the terms "restriction endonucleases" and "restriction enzymes" refer to bacterial enzymes each of which cut double-stranded DNA at or near a specific nucleotide sequence.

After digestion, specific “a priori” selected DNA loci are enriched and sequencing libraries are generated using smMIPs. smMIPs technology is used to capture and enrich specific DNA fragments, resulting in DNA libraries that can be sequenced. If used with a sufficiently high annealing temperature, smMIPs are highly specific. smMIPs are typically designed to anneal at a temperature of about 60°C. However, this may also be altered for example when a lot of unwanted ‘side products’ are generated, wherein the temperature can be slightly increased. Accordingly, typical smMIPs have an annealing temperature of about 50 °C - 70°C, more in particular about 55 °C - 65 °C, such as about 55 °C, about 56 °C, about 57 °C, about 58 °C, about 59 °C, about 60 °C, about 61 °C, about 62 °C, about 63 °C, about 64 °C, about 65 °C.

In addition, typically smMIPs are allowed to capture the DNA in a reaction at 60°C, for over 20 hours. We have now surprisingly found though that a cycled capture reaction of 5 times 4 hours at 60 °C with an intermediate denaturation step results in an increase of the percentage of unique reads. Accordingly, in a particular embodiment, the smMIP reaction comprises a cycled capture reaction of x times y hours at a predefined temperature, which is as discussed above; wherein x is selected from 3-6, in particular 5; and y is selected from 2-6, in particular 4.

Moreover, smMIPs are highly multiplexable, so that the combination of both steps results in the potential detection of several thousands of CpG methylation sites in a single sequencing run. smMIPs based approaches have been used for several applications, in combination with various techniques including microarray or next generation sequencing. These applications include SNP genotyping, Copy Number Variation quantification, and resequencing of genomic regions. smMIPs are extremely suitable for multiplex analyses, with routine multiplexing up to 50.000 being reported. For smMIPs design, we are routinely using a bioinformatics pipeline that was originally developed at the Radboud University in the Netherlands. In this pipeline, parameters for smMIPs design can be easily adapted.

Specific smMIPs at carefully chosen informative CpG sites in the genome will be generated, and hybridized to the digested DNA that was cut with the methylation-specific restriction enzymes. Subsequent extension and ligation of the smMIPs will only take place if the CpG site is methylated, because if it is unmethylated, it will be cut by the action of the restriction enzymes and no hybridization can take place. Accordingly, in a specific embodiment, said smMIP technology comprises the use of one or more smMIPs each spanning at least one methylation site in the DNA.

In a particular embodiment, each of said smMIPs comprises:

- a common backbone, which is not capable of hybridizing to said DNA molecules;

- a specific molecular tag, in particular an alignment tag.

In a particular embodiment, a set of amplified DNA fragments is obtained using one or more smMIPs each spanning at least one methylation site in the DNA; wherein each smMIP comprises one or more specific alignment tag(s);

In another particular embodiment, each of said smMIPs further comprises:

- a common backbone, which is not capable of hybridizing to said DNA molecules, and;

- a specific ligation probe, which specifically hybridizes to a second site next to said methylation site.

Such smMIP design in accordance with the present invention is for example illustrated in figure 2. In the present case, the extension probe hybridizes 5’ of the targeted region (i.e. including the methylation site), and the ligation probe hybridizes 3’ of said targeted region. Both probes are attached to each other by means of a non-complementary backbone. Said backbone is designed such as to allow hybridization of library amplification primers, for amplification of the targeted region. These amplification primers contain P5 and P7 illumina sequences so that the resulting PCR product can anneal to the flow cell. Finally, the smMIP comprises a specific molecular tag as part of the backbone. While in the present figure the molecular tag is indicated to be present between the ligation probe and one of the library amplification primer sites, it may alternatively be present anywhere in the smMIP, specifically in the backbone region. In a particular embodiment, said molecular tag(s) is/are located between the common backbone and the extension probe or between the common backbone and the ligation probe, or both.

The smMIPs of the present invention are in particular designed such as to have a relatively small targeted region (about 50 bp), compared to standard smMIPs (typically at least 100 bp). In particular the smMIPs of the present invention are designed to have a targeted region of about 40 - 60 bp, such as about 40 bp, about 45 bp, about 50 bp, about 55 bp, about 60 bp. This is in particular advantageous in the context of liquid biopsies, containing fragmented DNA. As used herein, the term "primer" refers to an oligonucleotide, whether occurring naturally as in a purified restriction digest or produced synthetically, which is capable of acting as a point of initiation of nucleic acid sequence synthesis when placed under conditions in which synthesis of a primer extension product which is complementary to a nucleic acid strand is induced, i.e. in the presence of different nucleotide triphosphates and a polymerase in an appropriate buffer ("buffer" includes pH, ionic strength, cofactors etc.) and at a suitable temperature. One or more of the nucleotides of the primer can be modified for instance by addition of a methyl group, a biotin or digoxigenin moiety, a fluorescent tag or by using radioactive nucleotides.

As used herein, the term "target sequence" refers to a specific nucleic acid sequence to be detected and/or quantified in the sample to be analysed.

The smMIP reaction typically comprises 3 separate steps, as shown in figure 3. During a first capture reaction (Fig. 3A), the smMIPs are allowed to hybridize to the fragmented DNA. The ‘gap’ defined by the targeted region in the smMIP is subsequently filled-in using a DNA polymerase, followed by a ligase thereby obtaining a circular smMIP. This step typically takes about 16 to 22h.

In a second stage, and in order to ‘clean-up’ all linear DNA fragments, an exo-nuclease reaction is performed, thereby, only circular smMIPs are retained (Fig. 3B). This step typically takes about 1 h.

Finally, these circular smMIPs are amplified by means of a PCR reaction using the amplification primers recognizing sequences in the backbone of the smMIPs (Fig. 3C). The single molecular tag can be used during the subsequent detection reaction for identification of the fragments, such as for example during multiplex next-generation sequence analysis.

Accordingly, in a further embodiment of the present invention, step c) comprises the steps of: c1) hybridizing said one or more smMIPs to said DNA fragments; c2) performing an extension reaction from the extension probe(s) across the methylation site(s) in the direction of the ligation probe(s); c3) performing a ligation reaction using the ligation probe(s), thereby obtaining a circular DNA fragment; c4) performing an exonuclease treatment to digest non-circular DNA fragments; c5) performing an amplification reaction using amplification primers capable of hybridizing to said common backbone; thereby obtaining a set of amplified DNA fragments.

As used herein the term "PCR" refers to the polymerase chain reaction. The PCR amplification process results in the exponential increase of discrete DNA fragments whose length is defined by the 5' ends of the oligonucleotide primers. As used herein, the terms "hybridisation" and "annealing" are used in reference to the pairing of complementary nucleic acids.

Following the method of the present invention, in a particular embodiment each of said amplified DNA fragments thus comprises a specific extension probe sequence, a methylation site sequence, a specific ligation probe sequence, sequencing adaptors, primer binding sites, and a specific molecular tag sequence; or the complement thereof.

These amplified DNA fragments can then be detected using any suitable methodology, such as by means of next generation sequencing. Accordingly, in a further embodiment, step d) of the method of the present invention comprises the step of detecting said amplified DNA fragments using next generation sequencing. The molecular tag sequence can be used in this stage to align the sequences into particular consensus read sequences. More specifically, sequence reads containing the same tag (and thus originating from the same capture) are merged into one consensus read sequence which can be detected using next-generation sequencing. In the context of the present invention, the tag sequence is a unique barcode that is incorporated in each smMIP allowing true molecule counting. Due to the obligatory PCR amplification step, which is necessary to obtain enough DNA templates for subsequent (next-generation) sequencing, PCR-duplicates will arise. Using this single molecule tag, it is possible to account for duplicates and as such, precise quantification is achieved. This accurate counting would not be possible when only MIPs are used. Accordingly, in another particular embodiment, a smMIPs comprises one or more specific alignment tag(s). The specific tag sequence can have a sequence variation as provided herein and may comprise at least 4, 5, 6, 7, 8, 9, 10, ... random nucleotides.

In a specific embodiment, the alignment tag(s) comprises at least 4 random nucleotides.

In a particular embodiment, the smMIP comprises two alignment tags, each comprising at least 4 nucleotides (i.e. 2 separate tags), located near the ligation and extension probe. Alternatively, the smMIP comprises one alignment tag comprising at least 8 random nucleotides located near the ligation or extension probe. In a preferred embodiment, said alignment sequence comprises two tags of at least 4 nucleotides, wherein one of the tags is located between the common backbone and the ligation probe, and the other tag is located between the common backbone and the extension probe. In a preferred embodiment, the alignment sequence comprises 8 random nucleotides located between the common backbone and either the ligation or the extension probe. Accordingly potential sequence variations erroneously introduced during the PCR reaction as well as sequencing artefacts can be easily detected and ignored. Therefore, the eventual sequence reads are much more reliable (see last section of the examples).

It is advantageous and moreover technically feasible in the method of the present invention to use multiple MSRE to fragment the DNA. Accordingly, it allows multiple potential methylation sites to be detected simultaneously, even if these are cut by different MSRE. Therefore, in another particular embodiment, the method of the present invention comprises the use of at least 2 methylation-specific restriction enzymes.

The MSRE according to the present invention are in particular characterized in digesting unmethylated DNA regions. Accordingly, unmethylated DNA regions are digested and can no longer bind the smMIPs, thereby leaving only methylated regions to be detectable, providing an excellent tool for the identification of such methylated regions. Particularly suitable MSRE in the context of the present invention may be selected from the non-limiting list of Hpall, HinP11, Acil, HpyCH4IV and combinations thereof.

Since methylation typically occurs in CpG regions, the present invention may further comprise the use of control smMIPs which do not span a CpG region. These regions are thus typically not digested by the methylation-specific enzymes and the detection thereof can be used to assess and/or monitor the reaction, or be used as a reference quantification marker. In another embodiment, the present invention may further comprise the use of control smMIPs that span a CpG region that do not include a restriction site for the MSREs (i.e. noRS smMIPs).

In some circumstances, a sample may contain substances that interfere with a subsequent restriction, amplification and/or detection step. In a method according to the invention, such interference may be avoided by extracting the DNA from a sample prior to digestion. Hence, the invention relates to a method as described above, wherein the DNA is extracted from the sample before allowing the DNA to be cut by the MSRE.

In summary, the present invention is particularly directed to the use of a combination of methylation-specific restriction enzymes (MSRE) and single molecule Molecular Inversion Probe technology (smMIP) in the detection of DNA methylation in a biological sample.

EXAMPLES

Selection of differentially methylated regions (DMRs) for incorporation in a MSRE- smMIPS pan-cancer detection panel

We selected a most optimal set of DMRs and developed this set into a MSRE-smMIPs-seq pancancer assay. Using the pipelines that are used to generate the data, an analysis for a total of 14 cancer types is performed on the basis of large datasets of 14 common cancer types, available from TOGA as well as the Gene Expression Omnibus database, and common DMRs are selected. Subsequently, our smMIPs design bioinformatics pipeline is specifically adapted towards the selection of smMIPs with optimal annealing temperatures covering both DMRs and adapter sequences, and the higher annealing temperature. We design specific smMIPs for 1000 DMRs that rank highest in the design pipeline, and these are combined in a single assay. In a balancing run for this assay, the performance of each individual smMIP is evaluated. Subsequently, a new assay is developed containing 500 well performing smMIPs, and individual smMIP concentrations are adapted for uniform performance. A bioinformatics pipeline for the analysis of the sequencing data is developed. This pipeline is specifically tailored towards the accurate quantification of the methylation status of each DMR/CpG region.

Validation of the MSRE-smMIPs-seq pan-cancer assay in breast, colon and lung cancer

After generation and optimization of the new MSRE-SMIPS-seq pan-cancer assay, and final selection of DMRs, the assay is validated using tissue and finally liquid biopsy samples from breast, colon and lung cancer patients as well as in blood samples from healthy patients in a first phase. First, the assay is validated in 50 untreated tumor vs normal tissue samples for the 3 most common cancer types (breast, colon and lung cancer). All samples are readily available from the Antwerp tumor bank. Next, the assay is tested on liquid biopsies from respective cancer patients (n=50 for each tumor type; also available in Antwerp tumor bank) and compared with liquid biopsies from healthy persons (n=50). CtDNA is extracted from all collected plasma samples using the QIAamp Circulating Nucleic Acid Kit (QiaSymphony) and analyzed. Data analysis is performed in-house.

Proof of concept, technical development and optimization of MSRE-smMIPs-seq technique:

A working principle of the technology has been obtained, and experiments using the bestperforming smMIPs are performed. Meanwhile, the construction of the bioinformatics pipeline for the MSRE-smMIP-seq data analysis is initiated. Clinical samples from cancer patients (readily available from the biobank of the UZA) are tested to obtain the proof-of-principle for the technique.

The combination of different MSREs can efficiently be used for application in the methylation sequencing assay

We analyzed the efficiency of combining four different MSREs (Hpall, HinPI I, Acil and HpyCH4IV) in a methylation detection assay. The efficiency of the cutting reaction was proved by performing a qPCR on a LightCycler 480 machine (Roche) with two control samples. The negative control consisted of A phage DNA, which is unmethylated. The positive control was artificially methylated A phage DNA. Artificial methylation was performed using Methyl Sss.l transferase (New England Biolabs). Primer pairs were designed for the specific restriction sites of each MSRE. Both controls were cut and uncut. Uncut samples underwent the same experimental conditions (i.e. incubation temperature and timing, buffers) but without the MSREs. Results are given in table 1 .

Table 1: Results obtained from the qPCR

For the positive controls, the mean obtained ct-values are expected to lay closely together, as the selected MSREs do not cut methylated DNA fragments. This is clearly the case, which proves that methylation effectively blocks the combined MSRE digestion. For the negative control, the cut samples should behave like a bianco sample, as it is expected all DNA is digested. We obtained a lower ct-value, indicating the DNA was not 100% digested. However, a difference in ct-values from cut and uncut samples of 17 was obtained. This means 1 in 2¹⁷= ~131 ,000 DNA molecules were not digested, suggesting that the combined MSRE-cut reaction is very efficient.

Demonstrating the first working principle and the feasibility of the protocol using several control samples

We demonstrated a first working principle and feasibility of the protocol. In these experiments, several control samples were used, including DNA extracted from 1) pancreatic cell lines, 2) primary lung tumor tissue, 3) human blood samples and 4) non-invasive prenatal test (NIPT) samples. The latter samples contained cfDNA and were included to test the feasibility of the approach in liquid biopsies. In total, 192 double-tiled designed smMIPs were used to detect 66 genome-wide differentially methylated regions (DMRs) and 30 non-CpG control regions. The control regions do not contain CpG sites and will therefore always be captured by smMIPs, irrespective of the MSRE digest. As such, these smMIPs can be used to estimate total DNA concentration in the sample.

After library enrichment, next-generation sequencing was performed using the Miseq reagent v2 nano kit. The data was analyzed with an in-house adapted bioinformatics pipeline. For each sample, the number of reads per smMIP is counted. The results that were obtained in these first runs showed a high number of reads on tumor, cell line and blood DNA. The NIPT samples obtained a lower amount of sequencing reads as expected. NGS run qualities ranged from 71 ,6 to 94,6 %1Q3O. Interestingly, some smMIPs gave more reads than others across samples and across experiments, showing that some smMIPs are more efficient than others. The efficiency of a smMIP is considered an inherent property, as it was shown to be independent of the composition of the total smMIP pool. A selection of smMIPs was tested both independently (capture reaction performed separately) and together in one pool (capture reaction performed in the pool). We observed that the effect of one smMIP being more efficient than another remained when analyzed separately. As such, only efficient smMIPs can be selected to be used for analysis of liquid biopsies.

Optimization of the protocol altering several wet-lab parameters

Currently, a selection of the most efficient smMIPs (32 CpG smMIPs and 20 control smMIPs) is used to further optimize the technique. Several runs have been performed using DNA extracted from human blood samples and a colorectal cell line (HT29, passage 13). One of the human blood samples was artificially methylated using the Methyl Sss.l transferase (New England Biolabs). i) Optimizing the enzymatic DNA cutting reaction

In liquid biopsies, the low concentration of DNA is one of the significant hurdles to overcome when designing new technologies. To show that our technology can be used for low input of DNA, the enzyme digest reaction was performed using different input concentrations in different end-volumes. An overview is given in table 2. For this preliminary experiment, as in previous experiments, A phage DNA was used. Uncut phage DNA was used as a positive control.

Table 2: overview of different sample conditions that were used in the optimization experiment

A qPCR was performed using primers designed for the specific restriction sites of each specific MSRE. Results are given in figure 9. Results of a similar qPCR results are already described above. Of interest here is that the enzyme digest performs very similar for all sample conditions. It can be concluded that this reaction is very robust, and applicable in samples with low DNA input. A DNA amount input of 5 ng was tested successfully and as can be seen on the figure, this low input amount performs equally well as a high input of 1 pig DNA. Together, these data suggests that the enzyme digestion is suitable for liquid biopsies. ii) Optimizing the efficiency of the assay by selection of a subset of smMIPs

As previously defined, a selection of 52 smMIPs was made to further optimize the assay. Figure 4 shows the balancing curve for all smMIPs (blue) compared to the top 52 (red). The x-axis represents all 192 smMIPs and the y-axis the relative coverage per smMIP, as a fraction of counts. The fraction of counts is determined as the absolute counts per smMIP in a sample divided by the total number of counts in that sample. As such, samples and smMIPs with different numbers of counts can be compared. When all smMIPs are taken into account, the most efficient smMIP (=smMIP with highest relative coverage) is 10,000 x more efficient than the least efficient smMIP. For the subset of 52 smMIPs, this difference is reduced to a 100-fold. In this way, the efficiency of the whole assay was improved.

Hi) Optimizing the efficiency of the assay using cycled capture reactions

Another parameter that allows the optimization of the assay, is the capture reaction. In several papers, it has been described that smMIPs capture the DNA in a reaction at 60°C, for over 20 hours. We used a cycled capture reaction, meaning we split the 20 hours into 5 times 4 hours. A denaturation step was implemented in between each cycle of 4 hours. Compared to a run with the exact same experimental conditions (except the capture reaction) and the same samples, we observed an increase of the percentage of unique reads (5% vs 12%). We are currently optimizing the number of cycles and total hours for the capture reaction in order to obtain as much unique reads as possible and as such to increase the efficiency of the assay.

Repeatability of the assay

Preliminary results for the replicability and reproducibility of the assay have been obtained. Considering the repeatability, the exact same experiment was conducted by two independent researchers. Results are shown in figure 5. The results are given for both cut and uncut samples, for the control (=non-CpG smMIPs) and CpG smMIPs.

The experiments were executed independently from each other and in parallel. Every step, beginning from the MSRE digest until the analysis was performed separately. In almost all samples, the results between two independent researchers (blue vs orange) are very similar. These preliminary results suggest the assay is robust, reliable and replicable. It is also clear from figure 5D that the relative coverage for CpG smMIPs is high in methylated samples and very low in unmethylated samples, as expected.

Reproducibility of the assay

The MSRE-smMIP-seq workflow has been performed several times by independent researchers. Preliminary results for the reproducibility demonstrate that the assay is reproducible. Results are given in figures 6 and 7.

Two independent runs (A and B) were executed by the same researcher. The relative coverage of the smMIPs is calculated as stated before. The relative counts of run A are plotted against the relative counts of run B. In figure 6, the average relative coverage per smMIP over all uncut samples was calculated and plotted. The average can be taken since for uncut (=not digested) samples, smMIPs behave similarly. There is a very strong correlation of the average relative counts in run A and B, with a correlation coefficient r^= 0,9894. The strong correlation between the results of the independent runs shows that the MSRE-smMIP method is reproducible.

Figure 7 shows the relative counts per smMIP in a low (A) versus high (B) methylated sample, in two independent runs executed by one researcher. For both samples, there is a strong correlation between the results of run A and B (r²= 0,9922 in the low methylated sample, r^= 0,9718 in the high methylated sample). The strong correlation between the relative coverages of run A and B again suggests that the method is very likely to be reproducible.

Methylation detection using CpG smMIPs is working effectively in digested samples

As was already highlighted in the part on repeatability of the assay, the CpG smMIPs show a higher number of counts and a higher relative coverage for methylated samples (samples 12 and 13, figure 5D) compared to human blood DNA samples. This is expected, since DNA from human blood samples is only low methylated. The CpG sites in these samples are digested and unavailable for capturing by the smMIPs. In the artificially methylated DNA blood sample and the DNA extracted from the cell line, the CpG sites are methylated, preventing digestion by the MSREs and enabling subsequent smMIP capturing. In figure 8, an example is given. For all 32 CpG smMIPs, the absolute number of counts per smMIP is given in 3 different samples (DNA from CRC cell line, artificially methylated human blood and human blood ). The high methylated samples show a higher number of absolute counts compared to the low methylated sample.

Performance of the technology in clinical colorectal cancer tissue DNA compared to normal blood DNA

Previous experiments showed a clear difference in highly methylated samples compared to low methylated samples. In a follow-up experiment, colorectal cancer tissue and blood samples were used to illustrate the ability of the technology to discriminate between cancer and normal samples.

For this experiment, DNA was extracted from 14 blood samples and 13 colorectal cancer tissue (FFPE-material). The earlier described combination of 52 smMIPs was used again. 32 of these smMIPs are designed to capture hypermethylated sites in tumor cells that were previously defined using TOGA data. This database provides epigenomic data through 450K micro arrays. Differential methylation is defined based on the B-values (output of a 450K micro array). In the original smMIP design, B-values with a maximum of 0.25 were selected for normal tissue. For cancer, B-values of at least 0.5 were selected. To discriminate high (hyper) and low (hypo) methylation, the largest difference in methylation B-values of normal adjacent tissue and cancer tissue was used, with a minimal difference of 0.25. By using the smMIPs designed for these CpG sites, we expect to find a high number of counts for CpG smMIPs in cancer tissue (as they capture the hypermethylated sites), while a low number of counts for CpG smMIPs is expected for normal tissue. For this experiment, normal blood samples were used instead of normal adjacent tissue. In figure 10, the result of the experiment is shown. All samples have been treated with MSRE. Blood samples (B) are shown on the left side and colorectal cancer tissue (CRC) on the right side. As expected, the CRC tissue samples show high read counts for the CpG specific smMIPs while the blood samples have lower counts for these smMIPs. The data prove that a very clear distinction can be made between normal and cancer using the technology. This difference is also demonstrated in figure 11 .

The counts for all CpG smMIPs were used in a principal component analysis (PCA) (figure 11). In this plot, the preliminary data are plotted on the first two principal components, that account for 87% of the variance in the data. Blood samples (circles) and CRC samples (triangles) can clearly be distinguished from one another. Blood samples are clustered more closely together than CRC samples, which can be explained due to the inherent variability in cancers. The sample type was not considered while performing the principal components analysis, indicating that the visualized difference between the blood and CRC samples is inherent.

Performance of the technology in several cell line DNA samples compared to normal blood DNA samples

In figure 12, the absolute read counts are given for gDNA extracted from normal blood samples (left) and compared to gDNA extracted from cancer cell lines (right). From this graph, it is clear that a perfect distinction can be made based on the CpG smMIPs (light bars). nonCpG smMIPs are also shown (dark bars) and theoretically should not differ between cancer and normal samples, which is the case here. In this experiment, a selection (based on efficiency) of +/- 800 CpG smMIPs were used.

A principal component analysis was then performed on the 500 best discriminating smMIPs. In figure 13, the first two principal components are shown. It is again clear that a perfect separation is found between blood samples (circles) and cancer samples, including the methylated blood sample (triangles and square). Interestingly, the blood samples are all grouped very closely, while the cancer samples are much more spread out.

Taken together, these results are a clear proof-of-concept that the technology is working as theoretically predicted. Taken all data into account, we are confident our technology will be applicable in liquid biopsies as well.

Proof of principle for single molecule Molecular Inversion Probe (smMIP) in removing PCR artefacts

To demonstrate the benefit of using single molecule tags to reduce technical noise caused by PCR amplification, we compared the results of analyses with and without duplicate removal using the same raw input data. The analysis without duplicate removal simulates an analysis without single molecule tag since the tag is not considered when duplicates are not removed.

First, the proportion of the reads per sample compared to the total amount of reads was compared between the analyses. Equation 1 below gives the example for sample X:

Equation 1 (left): Example of proportion ratio per sample. All reads associated with sample X are counted and divided by the total number of reads with and without duplicates. Then, the proportion without duplicates is divided by the proportion with duplicates to get the proportion ratio. Equation 2 (right): Example of proportion ratio per smMIP. All reads associated with smMIP N are counted and divided by the total number of reads with and without duplicates. Then, the proportion without duplicates is divided by the proportion with duplicates to get the proportion ratio.

If the single molecule tag would have no effect, it would be expected that for each sample, the proportion ratio approximately equals one. In figure 14; Panel A, we can see the proportion per sample grouped in bins of 0.1 around 1. The largest number of samples is indeed in the group from 0.95 to 1 .05; however, there are some samples that are incorrectly represented when not accounting for duplicates.

Next, this proportion analysis was repeated, this time comparing the proportion of reads per smMIP in both analyses (equation 2). Again, for many smMIPs the proportion is close to one, but now there are significantly more smMIPs that have a different count proportion when taking duplicates into account (figure 14; Panel B). Additionally, this distribution is skewed, so smMIPs with a proportion over one deviate more than smMIPs with a proportion under one.

Lastly, the proportion analysis was performed more fine-grained, this time looking into the proportion of reads per smMIP per sample, compared to the total reads per smMIP and per sample (equations 3 and 4 respectively):

Equation 3 (upper): Example of proportion ratio per smMIP per sample, divided by the reads per smMIP. All reads associated with smMIP N, sample X are counted and divided by the total number of reads of smMIP N with and without duplicates. Then, the proportion without duplicates is divided by the proportion with duplicates to get the proportion ratio. Equation 4 (bottom): Example of proportion ratio per smMIP per sample, divided by the reads per sample. All reads associated with smMIP N, sample X are counted and divided by the total number of reads of sample X with and without duplicates. Then, the proportion without duplicates is divided by the proportion with duplicates to get the proportion ratio.

Here, the differences are even more prominent. In figure 14 Panel A and B, it is clear that there are a lot of smMIPs in samples that diverge when correcting for duplicates. Interestingly, when comparing per sample, this effect is more prominent. Both distributions are also slightly skewed.

From this comparison, we can conclude that using a single molecule tag clearly reduces the technical noise in an experiment. This technical noise, caused by PCR amplification before sequencing to acquire the needed DNA concentration for sequencing, becomes more apparent when analysing the data in more detail. Additionally, overrepresentation of certain smMIPs because of a PCR bias is also fixed.

Claims

1 . A method for the detection of DNA methylation comprising the steps of: a) providing a sample comprising one or more DNA molecules; b) digesting said one or more DNA molecules by using one or more methylation-specific restriction enzymes (MSRE), thereby obtaining a set of DNA fragments; c) capturing and amplifying said set of DNA fragments using one or more single molecule Molecular Inversion Probes (smMIPs) each spanning at least one methylation site in the DNA; thereby obtaining a set of amplified DNA fragments; wherein each smMIP comprises one or more specific alignment tag(s); d) detecting said amplified DNA fragments; thereby detecting DNA methylation.

2. The method as defined in claim 1 , wherein each of said smMIPs further comprises:

3. The method as defined in claim 1 ; wherein said alignment tag comprises at least four random nucleotides.

4. The method as defined in anyone of claims 2 to 3; wherein said alignment tag(s) is/are located between the common backbone and the extension probe or between the common backbone and the ligation probe, or both.

5. The method as defined in anyone of claims 2 to 4; wherein said smMIP comprises two alignment tags each comprising at least 4 nucleotides located between the common backbone and each of said probes, or one alignment tag comprising at least 8 random nucleotides between the common backbone and said ligation or extension probe.

6. The method as defined in anyone of claims 2 to 5 wherein step c) comprises the steps of: c1) hybridizing said one or more smMIPs to said DNA fragments; c2) performing an extension reaction from the extension probe(s) across the methylation site(s) in the direction of the ligation probe(s); c3) performing a ligation reaction using the ligation probe(s), thereby obtaining a circular DNA fragment; c4) performing an exonuclease treatment to digest non-circular DNA fragments; c5) performing an amplification reaction using amplification primers capable of hybridizing to said common backbone; thereby obtaining a set of amplified DNA fragments.

7. The method as defined in claim 6; wherein each of said amplified DNA fragments comprises a specific extension probe sequence, a methylation site sequence, a specific ligation probe sequence, sequencing adaptors, primer binding sites, and one or more specific alignment tag sequence; or the complement thereof.

8. The method as defined in claim 1 ; wherein step d) comprises the step of detecting said amplified DNA fragments using next generation sequencing.

9. The method as defined in anyone of claims 1 to 8, wherein said method comprises the use of at least 2 methylation-specific restriction enzymes.

10. The method as defined in anyone of claims 1 to 9; wherein said methylation-specific restriction enzymes are only capable of digesting unmethylated DNA regions.

11. The method as defined in anyone of claims 1 to 10, wherein said one or more MSRE are selected from the list comprising: Hpall, HinP11, Acil, HpyCH4IV and combinations thereof.

12. The method as defined in anyone of claims 1 to 11 , further comprising one or more control smMIPs which do not span a CpG region and/or one or more control smMIPs which span a CpG region that does not include a restriction site for said MSRE.

13. The method as defined in anyone of claims 1 to 12, wherein said sample is a solid or liquid biopsy from a subject, in particular a liquid biopsy.

14. The method as defined in claim 13, wherein the subject is a mammal, in particular a human being.

15. Use of a combination of methylation-specific restriction enzymes (MSRE) to obtain a set of DNA fragments; and single molecule Molecular Inversion Probes (smMIPs) to capture and amplify said set of DNA fragments in the detection of DNA methylation in a biological sample.