Coupling Exponential to Linear Amplification for Endpoint Quantitative Analysis

Abstract Exponential DNA amplification techniques are fundamental in ultrasensitive molecular diagnostics. These systems offer a wide dynamic range, but the quantification requires real‐time monitoring of the amplification reaction. Linear amplification schemes, despite their limited sensitivity, can achieve quantitative measurement from a single end‐point readout, suitable for low‐cost, point‐of‐care, or massive testing. Reconciling the sensitivity of exponential amplification with the simplicity of end‐point readout would thus break through a major design dilemma and open a route to a new generation of massively scalable quantitative bioassays. Here a hybrid nucleic acid‐based circuit design is introduced to compute a logarithmic function, therefore providing a wide dynamic range based on a single end‐point measurement. CELIA (Coupling Exponential amplification reaction to LInear Amplification) exploits a versatile biochemical circuit architecture to couple a tunable linear amplification stage – optionally embedding an inverter function – downstream of an exponential module in a one‐pot format. Applied to the detection of microRNAs, CELIA provides a limit of detection in the femtomolar range and a dynamic range of six decades. This isothermal approach bypasses thermocyclers without compromising sensitivity, thereby opening the way to applications in various diagnostic assays, and providing a simplified, cost‐efficient, and high throughput solution for quantitative nucleic acid analysis.


Supporting Information
Coupling Exponential to Linear Amplification for Endpoint Quantitative Analysis Coline Kieffer, Yannick Rondelez, Guillaume Gines*         Unnormalized endpoint linear signal (C) show a substantial deviation from the expected proportionality with At (R² = 0.89).The proportionality is almost perfect when dividing rTω signal by its signal at saturation for the same well (E).

Table of Contents
However, such normalization is not possible in a standard endpoint experiment, as the saturation is reached after the experiment is stopped (here the endpoint time is 500 minutes while the saturation is reached in about 1000 minutes).
Alternatively, the rTω signal can be normalized by the rTα signal for the same endpoint time (D), which significantly improves the correlation with At.This indicates that most of the observed dispersion results from instrumental noise, more specifically inter-well variations of the optical signal, which applies to both rTα and rTω (F).In the absence of aTα, we noticed that the linear amplification rate decreases over time and this effect is all the more important as αkβ concentration is increased.This suggests that αkβ can slowly self-activate and that this leaky reaction gradually consumes the pool of β, eventually reducing the ω production rate.When loaded by aTα, the inhibition is substantially more efficient than the self-activation process.As expected, the higher [αkβ], the lower the inhibition delay following α exponential amplification.Table S1.Nucleic acid sequences used throughout this study."*" and "p" denote phosphorothioate backbone modification and 3' phosphate moiety, respectively.Upper and lower cases correspond to deoxyribonucleoside and ribonucleoside, respectively.2'OMeU corresponds to 2'-O-methyluridine.

Seq. ID Sequence Function
Sequences related to PEN-DNA reactions

Figure S3 .
Figure S3.Detailed chemical reactions of the CELIA network for microRNA detection (Figure 5).The dotted arrow indicates that reaction intermediates are not represented for the sake of clarity.

Figure S4 .
Figure S4.Detailed chemical reactions of the CELIA network embedding a signal inverter function (Figure 4). 4

Figure S5 .
Figure S5.Detailed chemical reactions of the linear amplification circuit for let-7a detection (cf. Figure S13).

Figure S6 .
Figure S6.Extended data from Figure 3. (A) Amplification curves for individual samples spiked with various concentrations of pTα and α.The linear amplification time trace (red curve) is fitted with a piecewise function (black dashed curve) () = {,   ≤ ; . +  − .,   >  , where  is the baseline constant,  is the amplification time and  is the linear amplification gain.The corresponding exponential α amplification curve is represented as the gray curve.On the top on each panel are indicated the concentration of pTα (from 0 to 7 nM) and α (0 or 50 pM) (B) Zoom on early emergence of the rTω fluorescence demonstrates the sharp transition from a nearnull to a constant ω production, concomitantly to the detection of the exponential amplification.(C) Amplification time as a function of [pTα].(D) Endpoint (1000 minutes) fluorescence of rTω as a function of [pTα].(E) The linear amplification gain extracted from the fit is plotted as a function of the amplification time.The gain is constant (CV ~5%) and shows no correlation with the amplification time, demonstrating the persistence of the DNA-enzyme reaction network toward long incubation time at 50 °C.

Figure S7 .
Figure S7.Linear signal normalization.The let-7a detection circuit is spiked with various concentration of let-7a, resulting in the modulation of the α amplification time.(B) Time traces of the exponential amplification (rTα fluorescence).The blue line depicts the signal threshold for extracting the amplification time (At).(C) top: Nonnormalized time traces of the linear amplification (rTω).bottom: endpoint signal (t = 500 min) as a function of At. (D) top: Time traces of the linear amplification (rTω) normalized with respect to the endpoint fluorescence of rTα (t=500 min).bottom: normalized endpoint ω signal (t = 500 min) as a function of At. (E) top: time traces of the linear amplification (rTω) normalized with respect to the ω signal at saturation (~1000 minutes).bottom: normalized endpoint ω signal (t = 500 min) as a function of At. (F) Maximum fluorescence at saturation of rTω versus rTα.

Figure S8 .
Figure S8.Tunability of the linear amplification module in the inverter function.(A) Architecture of the linear amplification module.(B) Real-time monitoring of the linear amplification module for various concentrations of βtoω and β. (C) Array plot of the production rate of ω.The more βtoω converter template or β input, the faster the production of ω.

Figure S9 .
Figure S9.Inhibition of ω production by the killer template αkβ.(A) The inverter circuit, in absence or presence of αkβ (5 nM) is run for various concentrations of βtoω and β. (B) Time traces of the exponential (green) and linear (red) amplification reactions in absence (light color) and presence (dark color) of αkβ.(C) Array plot of the inhibition delay, which correspond to difference between the amplification time and the time the linear amplification rate goes below 10 pM of consumed rTω per minute.The more βtoω or β input, the longer it takes for the killer template to reach the quasi-complete inhibition of ω production.For high concentrations of βtoω and β (gray boxes), rTω is entirely consumed before the inhibition reaction crosses the 10 pM/min threshold, which prevents from determining the inhibition delay.

Figure S10 .
Figure S10.Effect of the killer template concentration on the inhibition of ω production.(A) The inverter circuit in the absence (to test the kT leakage) and presence (to test the kT inhibition efficiency) of aTα is run in the presence of a varying concentration of αkβ.(B) Time traces of the exponential (top) and linear (bottom) amplification reaction in the absence (left) or presence (right) of aTα.In the absence of aTα, we noticed that the linear amplification rate

Figure S11 .
Figure S11.Extended data from Figure 5. (A) Amplification curves for individual samples spiked with various concentrations of pTα and α.The linear amplification time trace (red curve) is fitted with a piecewise function (black dashed curve) () = {a.x,   ≤  + ; .( + ) +  ! .( +  + ),   >  +  , where  is the amplification time extracted from the exponential amplification curve (represented as a blue triangle),  is the inhibition delay,  and ′ are initial linear amplification gain and the residual linear rate, respectively.The corresponding exponential α amplification curve is superimposed as the green curve.On the top on each panel are indicated the concentration of pTα (from 0 to 10 nM) and α (0, 50 or 200 pM).(B) Linear amplification gain (extracted from the fit) as a function of At.We observe that for short At, this value is correlated to the gain.This is explained by acceleration of the ω production in the early phase of the linear amplification (cf.also FigureS10), which introduces a

Figure S13 .
Figure S13.Strand displacement linear amplification for miRNA detection.(A) The linear amplification is achieved here by omitting the exponential amplification module and by directly connecting the output of the miRtoα converter template to the reporter template rTα.(B) Time traces of the linear amplification of a triplicate experiment for various

Figure
Figure S14.RT-qPCR miRNA calibration curves.(A) for let-7a and (B) for miR-203.The left panels represent qPCR amplification curves for a technical triplicate experiment.The right panels show the extracted calibration curves.All data points (except the no template control (NTC) were fitted with a linear regression (black line).The colored shaded area represents the 95 % confidence interval on the fit parameters.

Figure S15 .
Figure S15.Extended data from the Figure 4D.(A) Expected concentration versus concentration measured from the endpoint rT fluorescence for let-7a (left) and miR-203a (right).(B) 2D pattern of expected concentration (disks) versus concentration measured from the real-time amplification time traces (crosses).(C) Fold differences distribution computed for the endpoint (ep) or real-time (rt) readout.

Figure S17 .
Figure S17.Example of template design."*" are phosphorothioate bonds for 5'end protection."P" indicates 3' phosphate modification.Small protrusion indicates the nick site on the opposite strand.

Figure S18 .
Figure S18.Experimental conditions used in this study.