Short Keynote Paper: Single Molecule Detection of Protein Biomarkers to Define the Continuum From Health to Disease

Improving the Accuracy, Robustness, and Dynamic Range of Digital Bead Assays

We report methods that improve the quantification of digital bead assays (DBA)─such as the digital enzyme-linked immunosorbent assay (ELISA)─that have found widespread use for high sensitivity measurement of proteins in clinical research and diagnostics. In digital ELISA, proteins are captured on beads, labeled with enzymes, individual beads are interrogated for activity from one or more enzymes, and the average number of enzymes per bead (AEB) is determined based on Poisson statistics. The widespread use of digital ELISA has revealed limitations to the original approaches to quantification that can lead to inaccurate AEB. Here, we have addressed the inaccuracy in AEB due to deviations from Poisson distribution in a digital ELISA for Aβ-40 by changing the AEB calculation from a fixed threshold between digital counting and average normalized intensity to a smooth, continuous combination of digital counting and intensity. We addressed issues with determining the average product fluorescence intensity from single enzymes on beads by allowing outlier, high intensity arrays to be removed from average intensities, and by permitting the use of a wider range of arrays. These approaches improved the accuracy of a digital ELISA for tau protein that was affected by aggregated detection antibodies. We increased the dynamic range of a digital ELISA for IL-17A from AEB ∼25 to ∼130 by combining long and short exposure images at the product emission wavelength to create virtual images. The methods reported will significantly improve the accuracy and robustness of DBA based on imaging─such as single molecule arrays (Simoa)─and flow detection.

Digital detection of proteins

This paper reviews methods for detecting proteins based on molecular digitization, i.e., the isolation and detection of single protein molecules or singulated ensembles of protein molecules. The single molecule resolution of these methods has resulted in significant improvements in the sensitivity of immunoassays beyond what was possible using traditional "analog" methods: the sensitivity of some digital immunoassays approach those of methods for measuring nucleic acids, such as the polymerase chain reaction (PCR). The greater sensitivity of digital protein detection has resulted in immuno-diagnostics with high potential societal impact, e.g., the early diagnosis and therapeutic intervention of Alzheimer's Disease. In this review, we will first provide the motivation for developing digital protein detection methods given the limitations in the sensitivity of analog methods. We will describe the paradigm shift catalyzed by single molecule detection, and will describe in detail one digital approach - which we call digital bead assays (DBA) - based on the capture and labeling of proteins on beads, identifying "on" and "off" beads, and quantification using Poisson statistics. DBA based on the single molecule array (Simoa) technology have sensitivities down to attomolar concentrations, equating to ∼10 proteins in a 200 μL sample. We will describe the concept behind DBA, the different single molecule labels used, the ways of analyzing beads (imaging of arrays and flow), the binding reagents and substrates used, and integration of these technologies into fully automated and miniaturized systems. We provide an overview of emerging approaches to digital protein detection, including those based on digital detection of nucleic acids labels, single nanoparticle detection, measurements using nanopores, and methods that exploit the kinetics of single molecule binding. We outline the initial impact of digital protein detection on clinical measurements, highlighting the importance of customized assay development and translational clinical research. We highlight the use of DBA in the measurement of neurological protein biomarkers in blood, and how these higher sensitivity methods are changing the diagnosis and treatment of neurological diseases. We conclude by summarizing the status of digital protein detection and suggest how the lab-on-a-chip community might drive future innovations in this field.

Ultrasensitive multiplexed chemiluminescent enzyme-linked immunosorbent assays in 384-well plates

We have developed an ultrasensitive multiplexed immunoassay using 384-well microtiter plates capable of detecting proteins at subfemtomolar concentrations that requires as little as 2.5 μL of sample. Arrays of up to 4 capture antibodies were patterned on the bottom of the wells of a 384-well plate either by directly printing the capture antibodies or by printing anti-peptide tag anchor antibodies and incubating these arrays with capture antibodies conjugated to the corresponding peptide tags ("customized" assays). Samples were incubated with the antibody arrays and shaken orbitally at 2000 rpm to achieve the greatest sensitivity. Chemiluminescence (CL) from immunocomplexes labeled with horseradish peroxidase was imaged across the entire plate to quantify the amount of protein bound to each antibody spot of the arrays. The 384-well assay had a throughput 5-fold >96-well plates that was achieved from simultaneous imaging of CL in all 384-wells and the use of automated pipettors to allow parallel processing of 384 assays. We developed 4 assays based on the 384-well CL ELISA: a direct print assay for IL-10 (limit of detection (LOD) = 0.075 fM); a customized assay for IL-6 (0.22 fM); a customized pharmacokinetic (PK) assay for measuring adalimumab (7.3 pg/mL); and a customized 4-plex assay for IL-5 (0.1 fM), IL-6 (0.52 fM), IL-10 (0.2 fM), and TNF-α (3.2 fM). The sensitivity and precision of the cytokine assays were comparable to current ultrasensitive protein detection methods in 96-well formats. The PK assay for adalimumab was 650 times more sensitive than a commercially available 96-well plate ELISA. We used the 384-well CL ELISAs to measure endogenous levels of the cytokines in the serum and plasma of healthy humans: the mean concentrations and precision were comparable to those from 96-well immunoassays. This 384-well format with subfemtomolar sensitivity will enable ultrasensitive multiplexed immunoassays to be performed with higher throughput and lower sample volumes than currently possible, a particularly important capability for clinical studies in drug development.

Fluid Biomarkers of Frontotemporal Lobar Degeneration

A timely diagnosis of frontotemporal degeneration (FTD) is frequently challenging due to the heterogeneous symptomatology and poor phenotype-pathological correlation. Fluid biomarkers that reflect FTD pathophysiology could be instrumental in both clinical practice and pharmaceutical trials. In recent years, significant progress has been made in developing biomarkers of neurodegenerative diseases: amyloid-β and tau in cerebrospinal fluid (CSF) can be used to exclude Alzheimer's disease, while neurofilament light chain (NfL) is emerging as a promising, albeit nonspecific, marker of neurodegeneration in both CSF and blood. Gene-specific biomarkers such as PGRN in GRN mutation carriers and dipeptide repeat proteins in C9orf72 mutation carriers are potential target engagement markers in genetic FTD trials. Novel techniques capable of measuring very low concentrations of brain-derived proteins in peripheral fluids are facilitating studies of blood biomarkers as a minimally invasive alternative to CSF. A major remaining challenge is the identification of a biomarker that can be used to predict the neuropathological substrate in sporadic FTD patients.

Printed Electrodes in Microfluidic Arrays for Cancer Biomarker Protein Detection

Medical diagnostics is trending towards a more personalized future approach in which multiple tests can be digitized into patient records. In cancer diagnostics, patients can be tested for individual protein and genomic biomarkers that detect cancers at very early stages and also be used to monitor cancer progression or remission during therapy. These data can then be incorporated into patient records that could be easily accessed on a cell phone by a health care professional or the patients themselves on demand. Data on protein biomarkers have a large potential to be measured in point-of-care devices, particularly diagnostic panels that could provide a continually updated, personalized record of a disease like cancer. Electrochemical immunoassays have been popular among protein detection methods due to their inherent high sensitivity and ease of coupling with screen-printed and inkjet-printed electrodes. Integrated chips featuring these kinds of electrodes can be built at low cost and designed for ease of automation. Enzyme-linked immunosorbent assay (ELISA) features are adopted in most of these ultrasensitive detection systems, with microfluidics allowing easy manipulation and good fluid dynamics to deliver reagents and detect the desired proteins. Several of these ultrasensitive systems have detected biomarker panels ranging from four to eight proteins, which in many cases when a specific cancer is suspected may be sufficient. However, a grand challenge lies in engineering microfluidic-printed electrode devices for the simultaneous detection of larger protein panels (e.g., 50-100) that could be used to test for many types of cancers, as well as other diseases for truly personalized care.

Digital enzyme-linked immunosorbent assays with sub-attomolar detection limits based on low numbers of capture beads combined with high efficiency bead analysis

We report the development of digital enzyme-linked immunosorbent assays (ELISAs) based on single molecule arrays (Simoa) with improved sensitivities over conventional digital ELISA, enabling detection of proteins at sub-attomolar concentrations. The improvements in sensitivity were based on using fewer beads to capture the target proteins (≤5000 vs.∼500 000 beads) that increased the ratio of molecules to beads, and increasing the fraction of beads that were analyzed (bead read efficiency) from ∼5% to ∼50%. Bead read efficiency was increased by: a) improving the loading of beads into arrays of microwells by combining capillary and magnetic forces in a method called magnetic-meniscus sweeping (MMS); b) using a centrifugal washer to minimize bead loss during the assay; and, c) improved optics and image analysis to enable the analysis of more microwells. Using this approach, we developed an assay for IL-17A with a limit of detection (LOD) of 0.7 aM, 437-fold more sensitive than standard digital ELISA. A digital ELISA with improved sensitivity was used to measure IL-17A in 100 serum and plasma samples with 100% detectability, compared to 51% for standard digital ELISA. Low numbers of capture beads yielded improved LODs for IL-12p70 (0.092 aM), p24 (9.1 aM), and interferon alpha (45.9 aM). IL-4 and PSA showed no improvements in sensitivity using fewer beads, primarily due to low antibody loading on beads and increased non-specific binding, respectively. The results were consistent with a kinetic model of binding that showed that combining capture antibodies with high on-rates with high antibodies per bead yields the greatest improvement in sensitivity.