A handheld luminometer with sub-attomole limit of detection for distributed applications in global health

Abstract

Luminescence is ubiquitous in biology research and medicine. Conceptually simple, the detection of luminescence nonetheless faces technical challenges because relevant signals can exhibit exceptionally low radiant power densities. Although low light detection is well-established in centralized laboratory settings, the cost, size, and environmental requirements of high-performance benchtop luminometers are not compatible with geographically-distributed global health studies or resource-constrained settings. Here we present the design and application of a ~$700 US handheld, battery-powered luminometer with performance on par with high-end benchtop instruments. By pairing robust and inexpensive Silicon Photomultiplier (SiPM) sensors with a low-profile shutter system, our design compensates for sensor non-idealities and thermal drift, achieving a limit of detection of 1.6E-19 moles of firefly luciferase. Using these devices, we performed two pilot cross-sectional serology studies to assess sars-cov-2 antibody levels: a cohort in the United States, as well as a field study in Bangladesh. Results from both studies were consistent with previous work and demonstrate the device's suitability for distributed applications in global health.

Fig 1. Design overview and demonstration of operating principle.

A) Overview of the handheld luminometer: A removable rear access lid (left) allows access to the sample holder, accepting two clear, thin-walled PCR tubes. The user interface consists of a display panel, buttons, and a beeper. B) Rear view of the assembly with the access lid removed, showing the tube holder with its hinged lid. PCR tubes containing samples slide downward into the built-in recesses leading to the measurement cavities. C) Rotationally-exploded view of the internal sample cavity and shutter system. The tube holder (left, light and dark blue) houses the sample tubes in their own respective optical cavities coated in reflective mylar film (shown in yellow), increasing optical collection efficiency. The swinging shutter flags are low-profile, laser-cut stainless steel, fitting inside the 2 mm gap between the sensors and sample cavity, in order to preserve the high collection efficiency resulting from positioning the sample as close as possible to the sensor. D) Raw, shutter-modulated voltage traces (blue points, left y-axis) showing signal levels with varying amounts of incident optical power (black text with leaders) produced by filtered light from a feedback-stabilized arc lamp (Excelitas Exacte) imaged onto the luminometer sensor. Individual gated data points (red circles, right y-axis) are computed on the luminometer as described in the main text. Inset: Summary of gated voltage measurements vs. incident optical power, demonstrating the linear relationship between gated signal and incident optical power, which is independent from drift in sensor dark current. Raw data for D) is provided in S1 Data.

Fig 2. Basic performance characteristics.

The luminometer’s limit of detection is plotted for three contexts. A) Titration of firefly luciferase using luciferase assay system (red dashed lines), and the Bright-Glo assay (solid blue lines). ‘X’ markers represent channel A and ‘O’ markers represent channel B; the two channels on the same device. B) To simulate the spLUC assay, the LOD for eight luminometers was validated and normalized by titration of nanoluciferase into whole blood. Individual devices are shown in light gray with internal standard error of the mean (sem) error bars, while the fleet average is shown in black with the standard deviation across devices shown as error bars. Data with random noise fluctuations below zero are not shown due to the log scale. C) Radiometric testing results, shown as luminometer signal vs. incident optical power. In order to visualize the noise floor on a log-scale in the presence of stochastic noise that causes some data to become negative, a small additive offset of 3 RLU was applied to the data in all panels.

Fig 3. Performance under temperature stress.

A) Raw output time traces from the transimpedance amplifier are shown as solid traces at three temperatures (top plot), each after a 1 hr pre-incubation of the luminometer at the given temperature. The corresponding gated traces, which include dark current subtraction and ECC compensation, are shown as dotted lines (bottom plot). Gated traces are shown on separate axes due to the difference in amplitude. Inset: Zoom of the raw trace showing the ECC effect in the raw, uncompensated data. B) Allan deviation is used to estimate the lower bound of the instrument noise floor vs. integration time, for all three temperatures. Raw data is shown in solid lines (upper traces) and gated data is shown in dotted lines (lower traces). 1 RLU corresponds to approximately 1 fW of optical power, or ~3E-20 moles of luciferase. Colors in all panels correspond to the ambient temperature.

Fig 4. Detecting antibody levels to COVID-19 with a luciferase-based assay on the handheld luminometer.

A) Schematic of the point-of-care spLUC assay to detect antibodies. Lyophilized protein-based biosensors are reconstituted before the addition of capillary blood, after which luminescent substrate is added, red blood cells are pelleted and the signal is captured on the luminometer. B) Titration of the international serology standard on both the Spike (S) and Nucleocapsid (N) based biosensors. Dotted lines represent the cutoff value for determining a positive signal (purple, S: 22 RLU; blue, N: 32 RLU) as determined by negative control serum. C) Comparison of serology assays on a cohort of 26 vaccinated individuals. The signal from the spLUC assay, which is read out on the handheld luminometer, is compared to the Nirmidas lateral flow assay (LFA) and dried-blood spot based ELISA assay. Both assays show strong correlation to the spLUC assay (LFA: R = 0.84, ELISA: R = 0.91).

Fig 5. Bangladesh pilot serology data.

a) Raw S- and N- spLUC signals are individually plotted, sorted in ascending order along with positivity thresholds. Negative values were clipped at a value of ‘1’ for display on the logarithmic axis. The horizontal lines represent the negative cutoff values for each sensor (red dotted line: S-threshold = 20; blue dashed line: N-threshold = 25) that were determined by using mock negative control blood. b) Distributions of spLUC signals for S-sensors are shown as a function of the number of vaccine doses received by each individual. Horizontal lines are used to display quantiles of the distributions, with 10 and 90 percentiles denoted by thin lines, and 50 percentile (median) denoted by bold lines. c) Same as b), but for N-sensors. For clarity, two outlier data points at signal levels of 6,313 and 14,445 RLU were artificially lowered into the displayed range (dashed box).