Droplet digital PCR: introduction and application

Figure 5-29 Schematic representation of the probe-based fdC detection workflow. Two strands of the duplex are differentiate by red and black lines; black segments represent fdC probe; gray and purple segment represent reporter strand; arrows represent primers; segments with end points represent TaqMan probes.

5.4.2. Droplet digital PCR: introduction and application

123

Droplet digital PCR (ddPCR) came up in 2011.³⁴⁸ Instead of analysing the PCR process, ddPCR measures the absolute quantity by counting nucleic acid molecules encapsulated in discrete water-in-oil droplet partitions. Figure 5-30 shows the ddPCR workflows. First, the reaction mixture is prepared as in rtPCR, either with EvaGreen or using TaqMan probe contains FAM, HEX or VIC. Second, uniform droplets are generated from the reaction mixture. Each microliter (μL) generates 1000 droplets.

The target and background DNA samples are distributed randomly into the droplets together with the primers, the TaqMan probe, dNTP, and the polymerase. The process obeys the Poisson distribution. Third, the droplet is transferred to a PCR plate, and PCR cycle is conducted in the same manner as conventional PCR. Last, the PCR plate is placed in a droplet reader machine, which analyzes each droplet using a double wavelength detector.

Figure 5-30 Schematic representation of the ddPCR workflow. The figure is adapted from reference.³⁴⁸

More and more research taking advantage of the ddPCR principle came out in the past three years. A recent report applied the drop-based microfluides for the isolation and sequencing of the viral genome.³⁴⁹ Compared to the previous viral detection and sequencing method, droplet-based microfluidics followed by whole-genome amplification is a simple and applicable method to unknown viruses. This platform enables efficient isolation of single viral species from a mixture of other viruses and DNA contaminants, which was not feasible before.

Compared to rtPCR, ddPCR achieves absolute quantification of the amplicons with high accuracy. In this project, ddPCR is applied to develop a probe-based fdC

detection method. The ligation condition will be further optimized for the ddPCR method.

5.4.2.2. ddPCR: optimization and quantification modeling

The encapsulation maximum of one target amplicon in one droplet to generate a positive or negative signal is the ideal scenario for our situation. If a droplet contains more than one detection amplicon, for example, one contains one fdC and one cytosine, it will show a positive signal, and the negative cytosine signal vanishes.

The probability that two or more detection amplicons getting into one droplet can be calculated according to the Poisson distribution, i.e. a discrete random variable X complies the Poisson distribution with parameter λ > 0, if, for k = 0, 1, 2, …, the probability mass function of X is given by:

; λ Pr X λ

! where e is Euler's number and k! is the factorial of k.

Table 5-10 Poisson distribution probabilities of genome copies in the droplet.

input

λ k

ng 1 2 3 4 5

3 0.05 4.8% 0.1% 0.0% 0.0% 0.0%

6 0.10 9.0% 0.5% 0.0% 0.0% 0.0%

9 0.15 12.9% 1.0% 0.0% 0.0% 0.0%

10 0.17 14.1% 1.2% 0.1% 0.0% 0.0%

15 0.25 19.5% 2.4% 0.2% 0.0% 0.0%

20 0.33 23.9% 4.0% 0.4% 0.0% 0.0%

30 0.50 30.3% 7.6% 1.3% 0.2% 0.0%

40 0.67 34.2% 11.4% 2.5% 0.4% 0.1%

For example, the mass of a mouse genome is approximately 3.0 pg (3.0×10^-12 g). If 30 ng for a 20 μL reaction is used, 10,000 genomes will be distributed into 20,000 droplets. So, λ equals to 10,000 / 20,000 = 0.50. Let X = 1, then f (1; 0.50) = 0.303; let X = 2, then f (2; 0.50) = 0.076. This means 30.3% of the droplets, instead 50%, of the

125

droplet, contain a single copy while 7.6% of the droplets contain two copies.

Extensive distribution probabilities are listed in Table 5-10. If less than 9 ng is settled in a 20 μL reaction, the probability to have two copies inside one droplet will be lower than 1%. For the ease of calculation, 6 ng gDNA is input in each reaction of 20 μL, corresponding to ca. 90 copy/μL.

Figure 5-31 ddPCR output and modelling: a) 2-D plot of droplet fluorescence for illustration; b) clusters separation in four quadrants; c) algebraic simplification of counting numbers of the clusters.

As shown in Figure 5-31, for ease of nomination, Ch1+Ch2+ (yellow) refers to droplets with both positive signals; Ch1+Ch2- (blue) refers to droplets with only detection (reporter strand) signal; Ch1-Ch2+ (green) refers to droplets with only reference (gDNA) signal; Ch1-Ch2- (black) refers to droplets without target locus;

AD refers to all the droplets accepted; resolution refers to the separation of the clusters.

In principle, Ch1+Ch2+ shows the droplets that containing fdC site in the target locus;

Ch1+Ch2- indicates false-positive due to unspecific amplification; Ch1-Ch2+ shows the droplets contain non-fdC sites. Ideally, the ratio Ch1+Ch2+ to Ch2+ should be dependent on the gDNA sample. In reality, however, how the gDNA is treated and how the PCR is conducted impacts the resolution and signal to noise ratio. Therefore, it is crucial to optimize crosslinking, ligation and PCR conditions to separate the four clusters well.

A series of fdC probe concentrations were tested with the same reporter strand concentration using Tdg^-/- cells as a positive control and Dnmt TKO cell where there should be no fdC as a negative control. The results are judged by the ratio Ch1+Ch2+

to Ch2+. The background noise arose with increasing amount of fdC probe. To get a lower noise and good differentiation between the positive and negative control; also consider the amount of probe available, a 100 nM concentration was chosen for the further experiment.

From these experiments, it is concluded that ligation will not break the crosslinker (Figure 5-28c); Ampligase is superior to Taq Ligase in both specificity, i.e. Ch1+Ch2+

/ Ch1+Ch2-, and efficiency, i.e. Ch1+Ch2+ / Ch2+; ligation temperature should approach melting temperature of the reporter strand hybridization part; dilution of the ligation reaction mixture makes no difference in ligation specificity and efficiency.

Figure 5-32 1-D plot of droplet fluorescence of detection primer pair (Ch1) and reference primer pair (Ch2) at 68-60°C. Well A 68.0°C, B 67.4°C, C 66.4°C, D 64.9°C, E 63.1°C, F 61.6°C, G 60.6°C, H 60.0°C. PCR was conducted for 35 cycles.

The lower the PCR annealing temperature was set, the more background noise in Ch1 but the better resolution in Ch2 (Figure 5-32) was obtained. 64°C was finally chosen as annealing temperature. When increasing the numbers of PCR cycles, the higher the Ch2 resolution; the faster the PCR ramps, the higher the resolution in Ch1 and Ch2 (Figure 5-33). 2°C/s and 1 nM reporter strand were chosen for an ideal outcome. Ch1

127

primer amplification promotes Ch2 primer replication in droplet numbers slightly and lowers Ch2 resolution (Figure 5-34). For the same sample resulted from the two-step reaction from replicates the Ch1+Ch2+ / Ch2+ ratios are stable and reproducible.

Figure 5-33 1-D plot of droplet fluorescence of different temperature ramping with varied reporter strand concentration: a) 0.4°C/s; b) 2°C/s. Reporter strand concentration: well A, 2 nM, well C, 1 nM;

well E, 0.5 nM. PCR was conducted at 64°C for annealing and 35 cycles.

Figure 5-34 Double and single primer pair test: a) raw numbers of ddPCR results; b) 1-D plot of droplet fluorescence showing double primer pairs (well A), only detection primer pair ODN 35/36 (well G), only reference primer pair ODN 38/39 (well H).

Positive and negative controls were preformed on ddPCR. For negative control, a probe of position 2 (ODN 41, Chr 15, 8,846,677^th, MM9, see Chapter 5.4.3) instead of probe ODN 32 was used (Figure 5-35a); no reporter strand ODN 33 (Figure 5-35b) or no ligase (Figure 5-35c) was added. None of the three negative controls showed a Ch1+Ch2+ cluster, confirming that the correct combination of probe for the right position, the reporter strand and the ligase are the prerequisites for the detection.

When a ligated reporter strand ODN 34 was added to the reaction mixture, which served as a positive control, Ch1+Ch2+ and Ch1+Ch2- clusters appeared.

129 Figure 5-35 2-D plot of droplet fluorescence for negative and positive control of position 1: a) probe ODN 41 was used; b) no reporter strand ODN 33; c) no ligase; d) positive control. Condition: 80 U Taq Ligase, PCR extended at 64°C, 40 cycles.

Figure 5-36 Agarose gel shows gDNA degradation: Line 1, log 2 marker; line 2,3, gDNA (150 ng) after crosslinking; line 4,5, gDNA (150 ng) after ligation cycle, 95°C for 3min, then 10 cycles of 94°C for 1min and 60°C for 1h; liner 6,7, gDNA (150 ng).

Without considering the dissociation of the ligated products, the unreacted probe, which remained in the system, will cause unspecific amplification, i.e. Ch1+Ch2 -signals. Catalyst, acid buffer, and ligation cycles will cause gDNA degradation (Figure 5-36), giving more Ch1+Ch2- false-negative signals. However, Ch1+Ch2- and Ch1+Ch2- / Ch1-Ch2- resolution do not play a role in the mathematical modelling.

Assuming that all fdC at the target site is converted to the reporter strand via crosslinking and ligation, the yield is 100%. Assume that there are less than 150

copies in 1 μL so that the Poisson distribution is exclusive in our model.

Let a = Ch1+Ch2-, b = Ch1+Ch2+, c = Ch1-Ch2+, (Figure 5-31c) A = Accepted droplet for the experiment entry, respectively, a’, b’, c’, and A’ for the control, i.e. TET knockout cell line.

Let η = fdC content of the target site.

Then,

η a b ′

′ /η

where a refers to the degraded gDNA copy containing fdC at the target site that does not show in Ch2, a /η refers to all the degraded gDNA copies.

So,

η b ′ b′

Herein, in this ideal model, without considering the dissociation of the ligated products, η is independent of a, a’, and c’, i.e. genome degradation and unspecific ligation do not affect fdC percentage. Ch1+Ch2- only be resulted from the dissociation of the ligated products. Also, As shown in Figure 5-35 a-c, b’ can be omitted. Simplified η'

η′ a b

is calculated to indicate relative abundance of fdC at the target site.

To this point, the ddPCR-based profiling method was verified and optimized; a mathematical model was established.