Islet β cell death precedes development of type 1 diabetes, and detecting this process may allow for early therapeutic intervention. Here, we provide a detailed description of how to measure differentially methylated INS DNA species in human serum as a biomarker of β cell death.
The death of islet β cells is thought to underlie the pathogenesis of virtually all forms of diabetes and to precede the development of frank hyperglycemia, especially in type 1 diabetes. The development of sensitive and reliable biomarkers of β cell death may allow for early therapeutic intervention to prevent or delay the development of diabetes. Recently, several groups including our own have reported that cell-free, differentially methylated DNA encoding preproinsulin (INS) in the circulation is correlated to β cell death in pre-type 1 diabetes and new-onset type 1 diabetes. Here, we present a step-by-step protocol using digital PCR for the measurement of cell-free INS DNA that is differentially methylated at cytosine at position -69 bp (relative to the transcriptional start site). We demonstrate that the assay can distinguish between methylated and unmethylated cytosine at position -69 bp, is linear across several orders of magnitude, provides absolute quantitation of DNA copy numbers, and can be applied to samples of human serum from individuals with new-onset type 1 diabetes and disease-free controls. The protocol described here can be adapted to any DNA species for which detection of differentially methylated cytosines is desired, whether from circulation or from isolated cells and tissues, and can provide absolute quantitation of DNA fragments.
Type 1 diabetes (T1D) is an autoimmune disease that is characterized by the destruction of insulin-producing islet β cells by autoreactive T cells1. The diagnosis of T1D is typically made upon measurement of hyperglycemia (blood glucose > 200 mg/dl) in a lean, young individual, who might present with ketoacidosis as evidence of insulin deficiency. At the time of diagnosis of T1D, there is evidence for substantial loss of β cell function and mass (from 50 – 90%)2. In clinical studies, several immune modulatory drugs that were instituted at the time of diagnosis resulted in the stabilization of β cell function (and presumably mass), but none have resulted in clinical remission of disease, a finding that has raised the call for the development of biomarkers for earlier detection of the disease and for the longitudinal tracking of effectiveness of combination therapies3,4. Efforts by international consortia, such as the Human Islet Research Network Consortium at the National Institutes of Heath5, have emphasized the need to develop biomarkers that focus on β cell stress and death in T1D.
In line with these efforts, our group and others have recently developed biomarker assays that measure circulating, epigenetically modified DNA fragments that arise primarily from dying β cells6–9. In all of the published assays to date, the focus has been on the quantitation of the human gene encoding preproinsulin (INS), which demonstrates greater degrees of unmethylated CpG sites in the coding and promoter regions compared to other cell types. The liberation of unmethylated INS DNA fragments was hypothesized as arising primarily from dying (necrotic, apoptotic) β cells. Our recent studies showed that in youth, elevations in both unmethylated and methylated INS DNA at position -69 bp (relative to the transcriptional start site) were observed in new-onset T1D, and together served as specific biomarkers for this population6. These biomarker assays involve the isolation of cell-free DNA from serum or plasma using commercial spin kits, followed by a bisulfite conversion of the isolated DNA (to convert non-methylated cytosines to uracils, leaving methylated cytosines intact).
In this report, we describe the technical aspects of serum sample collection, isolation of cell-free DNA from serum, bisulfite conversion, and performance of droplet digital PCR (henceforth, digital PCR) for differentially methylated INS DNA.
Ethics Statement: Protocols were approved by the Indiana University Institutional Review Board. Parents of subjects provided written informed consent, and children older than 7 years provided assent for their participation.
1. Serum Processing
NOTE: The assay as described has been rigorously tested using human serum isolated as follows.
2. Serum DNA Extraction
NOTE: DNA is extracted with a DNA extraction kit using 50 µl of serum (recommended), following manufacturer's protocol with some modifications.
3. Bisulfite Conversion
NOTE: Bisulfite conversion is performed using a bisulfite conversion kit, following the manufacturer's protocol with some modifications.
4. Multiplex Digital PCR
5. Data Analysis
To interpret data appropriately, we use plasmid controls for both the unmethylated and methylated target INS DNA in each digital PCR run. These controls ensure that signals corresponding to methylated and unmethylated DNA are clearly distinguishable. Figure 1 shows the 2-D scatter plots corresponding to droplets for plasmid controls containing bisulfite-converted unmethylated INS DNA (Figure 1A), methylated INS DNA (Figure 1B), and a 1:1 mixture of the two plasmids (Figure 1C). Plasmid containing unmethylated INS is indicated by the droplets positive for the 6-Carboxyfluorescein (FAM) signal, whereas plasmid containing methylated INS is indicated by the droplets positive for the VIC signal. In the 1:1 plasmid mixture, a population of double-positive droplets is seen, corresponding to droplets containing both species of DNA. To distinguish positive droplets from negative droplets, data are gated using these 2-D plots. Note that in Figure 1A, there is a slight shift of the FAM-positive droplets into the VIC channel, indicating that the probe for the methylated INS DNA exhibits some cross-reactivity for the unmethylated INS DNA. Similarly, in Figure 1B, there is a very slight shift of the VIC-positive droplets into the FAM channel, indicating cross-reactivity of the probe for the unmethylated DNA with methylated DNA. A major advantage of digital PCR is it can still discriminate specific signals even when probes are not 100% specific. For appropriate Poisson statistical calculations by the software, each PCR reaction should partition into at least 10,000 total droplets, display a cluster of negative droplets, and show a clear amplitude difference between negative and positive droplets (for reliable gating). To demonstrate linearity of our primers, we performed mixtures of the two plasmids across several orders of magnitude. As shown in Figure 2, varying concentrations of one plasmid can be linearly detected in the presence of a constant amount of the second plasmid. For new labs performing this procedure, we recommend constructing a similar mixture experiment to ensure that the assay is operating in a linear detection fashion.
Next, we obtained serum of three subjects with new-onset T1D (within 2 days of diagnosis) and three control individuals without T1D (see Table 1 for clinical characteristics). Protocols were approved by the Indiana University Institutional Review Board. Parents of subjects provided written informed consent. Figure 3 shows the quantitation of unmethylated and methylated INS DNA copy numbers/µl serum, converted to log10 from these controls and T1D subjects. The data show that individuals with T1D exhibit elevated levels of both methylated and unmethylated INS DNA compared to controls, similar to data reported previously6.
Figure 1: Representative 2-D Plots of Plasmid Controls. Plasmid standards were generated by cloning fragments of bisulfite-converted INS DNA harboring unmethylated or methylated cytosine at position -69 bp relative to the INS transcriptional start site. Shown are 2-D plots using plasmid containing unmethylated (A) and methylated (B) INS DNA and for a 1:1 mixture of the two plasmids (C). Arrows identify the unmethylated, methylated, and unmethylated + methylated (double- positive) INS DNA-containing droplets. FAM = Target 1; VIC = Target 2. Please click here to view a larger version of this figure.
Figure 2: Quantitation of DNA Targets using Control Plasmid Mixtures. Plasmid standards containing cloned fragments of bisulfite-converted INS DNA harboring unmethylated or methylated cytosine at position -69 bp relative to the INS transcriptional start site were mixed at the various combinations shown, then subjected to multiplex digital PCR. The figure shows the quantitation of target DNA fragments, presented as copies/µl; r2 = 0.989 for unmethylated INS DNA and r2 = 0.998 for methylated INS DNA. Please click here to view a larger version of this figure.
Figure 3: Circulating Unmethylated and Methylated INS DNA Levels in Controls and Subjects with T1D. Serum samples were collected from 4 youth with new-onset T1D and from 4 disease-free, unrelated controls, then processed for measurement of differentially methylated INS DNA by digital PCR. Shown are results of digital PCR assays for unmethylated (A) and methylated (B) INS DNA. Data is displayed as individual points and mean ± SEM. Statistics were analyzed by a two-tailed parametric Students t test. Please click here to view a larger version of this figure.
Controls | T1D at onset | |
Age (years) | 10.3 ± 1.3 | 9.2 ± 1.0 |
Female/Male | 1:3 | 1:3 |
BMI Z-Score | 0.30 ± 0.9 | 0.74 ± 0.9 |
HbA1C (%) | 11.1 ± 0.6 | |
C-Peptide (pmol/L) | 285.5 ± 37.0 |
Table 1: Demographics of Controls and Subjects with T1D.
Methylation of cytosines by DNA methyltransferases allows for the epigenetic control of transcription at many genes. The INS gene in humans is almost exclusively expressed in islet β cells, and there appears to be a correlation between the frequency of methylation of cytosines in the INS gene to silencing of its transcription11. As such, most cell types show substantially higher frequencies of methylation of the INS gene at various cytosines compared to β cells11–13. It has been proposed that the prevalence of β cell death could be monitored by analysis of the serum levels of cell-free unmethylated INS DNA, which would arise largely from dying β cells that liberate their DNA12.
The bisulfite reaction converts unmethylated cytosines into uracils, leaving methylated cytosines unchanged14. Sequencing or PCR analysis of bisulfite-converted DNA therefore allows determination of whether the original cytosine was methylated or not. Methylation-specific PCR using dye-based technology exploits differences in PCR efficiency owing to 3' base-pair mismatches between primers and template, and thereby allows distinction of unmethylated vs. methylated cytosines following bisulfite conversion of DNA15. Accordingly, dye-based PCR was used to quantitate the relative ratio of unmethylated to methylated INS in early studies of T1D subjects12,16. However, dye-based PCR technology has several drawbacks, including limited specificity, lack of absolute quantitation of DNA fragment concentrations, and the inability to multiplex (i.e., detect both methylated and unmethylated DNA species in the same reaction). Other groups have performed direct sequencing of cell-free DNA fragments, a technique that allows monitoring of several differentially methylated CpG sites with much greater specificity9. However, the sequencing technique requires large sample volumes, is not easily amenable to use of banked samples, and is expensive.
As a result of these limitations, our group and others have employed digital PCR, a technique that provides absolute quantitation of DNA fragments, requires only microliter quantities of serum from fresh or banked samples, is amenable to multiplexing, shows greater sensitivity than traditional quantitative PCR, and is less expensive than direct sequencing. Digital PCR technology involves the use of conventional primer/probe assays and a microfluidics-based partitioning of a water-in-oil PCR reaction into ~ 20,000 droplets. Following thermal cycling of the partitioned reaction, the droplets are subsequently analyzed by a flow cytometer to identify droplets with positive and negative signals. Poisson statistics is used to calculate absolute quantities (copy numbers) of each DNA species. The conceptual details of digital PCR have been described elsewhere17.
This protocol describes the measurement of β cell death in humans by measuring differentially methylated INS DNA by digital PCR. Our data presented here and elsewhere6 show that primer/probe combinations that detect differential methylation of cytosine at position -69 (relative to the transcriptional start site of INS) exhibit sufficient specificity for either the unmethylated cytosine (FAM probe) or the methylated cytosine (VIC probe) (see Figure 1A and B). Mixtures of plasmids containing both methylated and unmethylated INS fragments at position -69 bp (i.e., bisulfite-converted DNA having either uracil or cytosine at position -69 bp) exhibit droplets containing one or the other plasmid, or both plasmids (double-positive signals in Figure 1C). The Poisson statistical calculation utilizes the droplet numbers that are positive for a given probe (whether single-positive or double-positive) to derive the copy numbers of each methylated or unmethylated DNA fragment.
When utilizing human serum for the measurement of differentially methylated cytosines at position -69 bp in cell-free DNA (Figure 3), our data demonstrate two key features: (1) levels of methylated INS are 5-10-fold higher in these subjects than unmethylated INS (regardless of diagnosis), and (2) absolute levels of both unmethylated and methylated INS are higher in subjects with new-onset T1D compared to controls. Whereas the higher levels of methylated INS compared to unmethylated INS likely reflect the greater burden of cell-free DNA emanating from non-β cells, the higher levels of both species of INS in new-onset T1D subjects likely reflects both an increase in prevalent β cell death (unmethylated INS), and possibly death/turnover of other cell types that are associated with the disease state (e.g., innate and adaptive immune cells). Further speculation on the elevations of both species of INS in new-onset T1D subjects was discussed previously6. Regardless of the sources of both species of INS in serum from these individuals, it is notable that measurement of ratios of unmethylated to methylated INS (as reported by other investigators using dye-based PCR) would not have picked up the significant differences in absolute levels described here and in our recent publication6, thus emphasizing the power of digital PCR for this type of biomarker analysis.
Some precautions and limitations should be noted. We feel it is important that blood is processed within 4 hr of collection and either used immediately for DNA isolation or stored at -80 ºC for future use. In unpublished studies, we have observed some loss of DNA recovery after 4 hr of collection, perhaps owing to DNA degradation, as seen in studies of long-term storage of samples18. Because of the increased sensitivity of digital PCR, special care should be taken to avoid contamination of samples that would otherwise not be noticeable in traditional qPCR. Finally, a key limitation of the technique is the analysis of samples in which droplet generation does not exceed ~10,000 droplets. Such samples do not provide reliable Poisson statistics, and as such may be considered uninterpretable. Multiple causes of such low droplet generation might include technical errors in sample handling, technical issues with the droplet generator, or issues related to quality of reagents used.
In summary, this report describes the detailed protocol for measurement of differentially methylated, cell-free INS DNA at position -69 bp by digital PCR. The protocol described here can be adapted to any DNA species for which detection of differentially methylated cytosines is desired, whether from circulation or from isolated cells and tissues, and can provide absolute quantitation of DNA fragments.
The authors have nothing to disclose.
This work was supported by National Institutes of Health grant UC4 DK104166 (to RGM). We wish to acknowledge the assistance of the Indiana Diabetes Research Center Translation Core supported by National Institutes of Health grant P30 DK097512.
Red Top Vacutainer | Beckon Dickinson | 366441 | no additive, uncoated interior, 10 ml |
Cryovial Tube | Simport | T310-3A | polypropylene, screw cap tube, any size |
QIAamp DNA Blood Mini Kit | Qiagen | 51106 | |
Poly(A) | Sigma | P9403 | disloved in TE buffer (10 mM Tris-Cl pH 8.0 + 1 mM EDTA) to 5 µg/µl |
Absolute Ethanol (200 Proof) | Fisher Scientific | BP2818-500 | |
DPBS (with CaCl and MgCl) | Sigma | D8662 | |
0.2 mL PCR 8-strip Tubes | MidSci | AVST | |
8-strip Caps, Dome | MidSci | AVSTC-N | |
EZ DNA Methylation-Lightning Kit | Zymo | D5031 | |
ddPCR Supermix for Probes (No dUTP) | Biorad | 1863024 | |
Buffer Control for Probes | Biorad | 1863052 | |
Human Unmethylated/Methylated Primer/Probe mix | Life Technologies | AH21BH1 | |
EcoR1 | New England Biolabs | R0101L | |
twin.tec PCR Plate 96, semi-skirted | Eppendorf | 951020346 | |
Pierceable Foil Heat Seal | Biorad | 1814040 | |
PX1 PCR Plate Sealer | Biorad | 1814000 | |
QX200 AutoDG Droplet Digital PCR System | Biorad | 1864101 | |
Automated Droplet Generation Oil for Probes | Biorad | 186-4110 | |
DG32 Cartridge for Automated Droplet Generator | Biorad | 186-4108 | |
Pipet Tips for Automated Droplet Generator | Biorad | 186-4120 | |
Pipet Tip Bins for Automated Droplet Generator | Biorad | 186-4125 | |
C1000 Touch Thermal Cycler | Biorad | 1851197 | |
QX200 Droplet Reader | Biorad | 186-4003 | |
ddPCR Droplet Reader Oil | Biorad | 186-3004 |