Single-molecule ELISA.

The need for assay methods of sufficient sensitivity to determine the low concentrations of hormones present in body fluids led to the original development of immunoassays and analogous “binding” (or “ligand”) assays in the late 1950s and early 1960s. These methods depend on the use of a binding agent (also commonly referred to by other terms such as “receptor,” “binding reagent,” and “analyte-specific reagent”), a substance used to recognize and bind the target analyte. Typical binding agents include antibodies, antigens, cell receptors, and serum binding proteins. Immunoassays still constitute the most widely used class of binding assays, although microarray-based nucleic acid assays, employing oligonucleotides as binding agents, are rapidly increasing in popularity. A principal objective in this field since the emergence of these assays has been to increase their sensitivities, especially in their application to certain analytes. Typifying such attempts, Rissin et al. (1 ) have recently reported a new approach to the further improvement of immunoassay sensitivities, claiming that with the use of an ELISA-type system, they were able to “detect serum proteins at subfemtomolar concentrations” and to increase the sensitivity of measurements “using a typical ELISA plate reader by a factor of about 68 000.” But before discussing the novel features of Rissin et al.’s approach, we should briefly examine the concept of sensitivity and the meaning of the term “sensitive” to describe the performance of a binding assay— or indeed that of any measurement system. Many workers in this area, including Rissin et al., identify sensitivity with the lower limit of detection (LoD) of an assay. However, certain bodies, including the American Chemical Society and the International Union of Pure and Applied Chemistry, have formally defined sensitivity as the slope of the dose–response curve [or the response/dose (R/D) ratio—an intrinsically meaningless concept with which we strongly disagree (see Fig. 1)]. The reason for our opposition to this concept is, in short, that the more sensitive of 2 or more systems has been regarded by scientists since the 1850s as the system that detects and measures the smaller amount of that which the systems are intended to measure, i.e., the system that exhibits the lower LoD. Though some have argued (2 ) that increasing the response-curve slope reduces the LoD, this supposition is not generally true. The fundamental difference between these 2 concepts of sensitivity has not only led to past controversy and debate (3 ), but the slope concept has also profoundly influenced the design of immunoassays and analogous methods. Equating sensitivity with the response-curve slope or the R/D ratio has, in practice, led to the use of relatively high antibody concentrations, typically approximating 1/K (where K is the affinity constant governing the binding reaction under the conditions used in the assay) in competitive assays (e.g., RIAs) and 20/K in noncompetitive immunometric methods (e.g., sandwich assays), this generally implying the capture of approximately 40%–50% and 90%, respectively, of the analyte in a sample. (Note that immunoassays that rely on the use of radioisotopically labeled antibodies are generally termed IRMAs but are sometimes of competitive design.) But an important factor that affects an assay’s LoD is the presence of noise in the system, i.e., the variation in the signal generated by a blank sample containing 0 analyte, generally represented by the SD of the blank measurement. Thus a key determinant of an assay’s LoD (as of any measuring system) is the signal/noise ratio, where “signal” here refers to the signal deriving from the target analyte. Clearly the lower the analyte concentration, the lower the signal/noise ratio, it being commonly accepted that a ratio of 3 defines the LoD. Rissin et al. clearly recognize the importance of maximizing the signal/noise ratio to maximize immunoassay sensitivity. Their approach to achieving this objective (termed digital ELISA) is to count individual target molecules that are captured by antibody on a solid support (this support comprising thousands of microscopic beads, each 2.7 m in diameter), the captured molecules being subsequently bound by a second, enzyme-labeled antibody to form an antigen/antibody sandwich. By using large numbers of beads, the authors ensure both that a high proportion (approximately 70%) of analyte molecules in a sample is cap1 University College, London, UK; 2 Northwestern University, Evanston, IL. * Address correspondence to this author at: Windeyer Institute of Clinical Science, University College London, Mortimer Street, London W1T 4JF, UK. E-mail [email protected]. Received September 29, 2010; accepted October 4, 2010. Previously published online at DOI: 10.1373/clinchem.2010.152850 3 Nonstandard abbreviations: LoD, limit of detection; R/D, response/dose. Clinical Chemistry 57:3 372–375 (2011) Perspective