Research Article: Statistical models discriminating between complex samples measured with microfluidic receptor-cell arrays

Date Published: April 8, 2019

Publisher: Public Library of Science

Author(s): Ron Wehrens, Margriet Roelse, Maurice Henquet, Marco van Lenthe, Paul W. Goedhart, Maarten A. Jongsma, David T. Eddington.


Data analysis for flow-based in-vitro receptomics array, like a tongue-on-a-chip, is complicated by the relatively large variability within and between arrays, transfected DNA types, spots, and cells within spots. Simply averaging responses of spots of the same type would lead to high variances and low statistical power. This paper presents an approach based on linear mixed models, allowing a quantitative and robust comparison of complex samples and indicating which receptors are responsible for any differences. These models are easily extended to take into account additional effects such as the build-up of cell stress and to combine data from replicated experiments. The increased analytical power this brings to receptomics research is discussed.

Partial Text

Receptomics research with microfluidic receptor cell arrays aims to measure purely biological responses without a complicated biological system surrounding it [1–3]. For example, the human tongue can be emulated on a chip by an array containing G-protein coupled receptors (GPCRs), e.g., in the form of reconstituted receptor proteins [4] or vesicles [5]. Another possibility is formed by living cells expressing the genes coding for particular GPCRs, produced by either reverse-transfecting a generic cell line on the chip [6] or spotting pre-transfected cells [7]. Such a tongue-on-a-chip allows direct access to the original taste signal, before the signal is further transmitted via neurons, and processed and interpreted by the human brain. Thus, while a taste panellist would define a sample as bitter or sweet, a tongue-on-a-chip provides direct quantitative information on which taste receptors are triggered, and by how much. On a more general level, receptomics enables identifying compounds or extracts activating or blocking specific receptors active in taste sensation as well as in many other processes. Humans have a wide palette of receptor proteins. Even considering only GPCRs there are more than 800 receptors for the detection of hormones, neurotransmitters, tastants, odorants, and others. Since all receptors play an important role in human physiology, there is an advantage to a receptomics approach aiming at combining different receptors on a single chip, allowing the researcher to study the role of a compound or extract in a wider perspective by including all or at least the most relevant receptors.

Three experiments, each executed three times, serve to illustrate our approach. In the first type of experiment (A), a quality control (QC) mixture of four compounds, chosen because they are known to hit specific bitter receptors, was injected at nine different dilutions (see Table 1). The sample with the lowest concentration was injected first; each subsequent injection had a double concentration of the QC mixture. In addition, a blank injection, and a 2 μM ATP injection were performed. The ATP injection elicits a host-cell response, different for each receptor type. The second type of experiment (B) was set up to compare 2 μM ATP injections with injections where the ATP sample was spiked with the same QC mixture as in (A). Also here a blank injection (containing only assay buffer) was included. This experiment is a simple example of a case where a specific response should be estimated in the presence of a constant background (the host-cell response). Table 2 gives an overview of the injections in both experiment types.

Data from sensors based on live cells often lead to highly variable results. At the cell level, transfection efficiency and cell-cycle differences cause variation in protein expression, leading to variation at the spot level. This can be prominent if relatively few cells make up a signal on a receptor-cell microarray. In addition, in the analysis of complex samples both receptor-specific and generic host-cell responses are often encountered.

In the field of receptomics, microfluidic receptor-cell arrays are a valuable tool in investigating human responses to food and in trying to understand the relation between chemical composition and taste in food stuffs [6, 21]. We have shown that careful experimental design, data processing and statistical analysis can lead to highly informative results, opening the way to many diverse applications. One particularly interesting possibility is to use these cell arrays as extensions of taste panels for prescreening: in principle, many quantitative comparisons can be made in a rapid and cost-effective way. This receptomics tool can be further extended to other receptors of the GPCR gene family and ion channels, both of humans and other organisms.




Leave a Reply

Your email address will not be published.