Label-Free Cellular Assays: Impedance and Dynamic Mass Redistribution

A cAMP accumulation plate and a calcium flux plate each ask the cell one question and record one answer. A label-free whole-cell assay asks nothing in particular. It seats the monolayer on a transducer and records how the cells rearrange their mass and their contact with the substrate after you add a ligand. Two sensor families dominate: electrical impedance measured across thin-film electrodes beneath the cells, and optical resonant waveguide grating (RWG) reading dynamic mass redistribution (DMR). Neither has any pathway selectivity built into it. That is what makes the readout broad, and it is also why a trace on its own does not tell you which transducer moved.

What follows is what each sensor physically integrates, what the time course carries that a single-timepoint read discards, how to recover pathway identity by perturbation rather than by inspection, and where cell background and adhesion state quietly set the size of the response you are about to call pharmacology.

What the impedance sensor is actually integrating

The measurement is old and its physical basis is unusually well characterised. Giaever and Keese cultured mammalian fibroblasts on evaporated gold electrodes in an alternating field at 4000 Hz, with a large counter electrode of roughly 2 cm² and a small sensing electrode of roughly 3 × 10⁻⁴ cm²; the small population of cells attaching and spreading on the small electrode had a marked effect on the measured impedance and also caused it to fluctuate with time, and the amplitude of those fluctuations was greatly reduced by cytochalasin B at 10 µM, which suggested they are a consequence of cell movement ^[1]. That single control is the licence for reading an impedance trace as a cytoskeletal quantity rather than an electrical curiosity.

What the current is doing under the cells was formalised later. Lo, Keese and Giaever measured transepithelial impedance of MDCK layers over 20–50,000 Hz and built a model considering three pathways for current flowing from the electrode through the cell layer: in through the basal and out through the apical membrane, in through the lateral and out through the apical membrane, and between the cells through the paracellular space ^[2]. Fitting that model to the spectrum returns the resistivity of the cell layer, the average distance between the basal cell surface and the substratum, and the capacitance of apical, basal and lateral membranes; applied to removal of Ca²⁺ from confluent MDCK layers, it showed junction resistance between cells dropping and the basal-to-substratum distance increasing ^[2]. Those are MDCK epithelial parameters under Ca²⁺ withdrawal, not a general calibration you can transfer to a HEK monolayer, but they identify the three physical quantities any impedance change has to be some combination of: junctional tightness, the height of the fluid gap under the cell, and membrane area.

Frequency chooses which of those you are most sensitive to. Wegener, Keese and Giaever showed that high-frequency capacitance measurements at 40 kHz are best suited to follow increasing surface coverage of the electrode as cells spread, and used the resulting time resolution to analyse spreading kinetics of MDCK strain II cells on different protein coatings ^[3]. If your instrument reports a single composite index at a fixed frequency, you have accepted a fixed weighting of those three contributions and cannot recover them afterwards.

The sensitivity floor is set by geometry rather than by optics. Tracking impedance over time, Giaever and Keese found fluctuations that are a direct measure of cell motion, and by comparing the measured impedance with a theoretical model concluded that under those circumstances average motions of the cell layer of 1 nm can be inferred; those fluctuations continued even when the cell layer became confluent ^[4]. Two consequences for assay design follow. First, the baseline of a healthy confluent well is not flat, and the noise band you subtract is itself biology. Second, resolution at the nanometre scale means the sensor will report anything that moves cells, including additions that have nothing to do with your receptor.

What the optical sensor is actually integrating

RWG biosensors are sensitive to redistribution of cellular contents in both directions perpendicular and parallel to the sensor surface. Fang and colleagues showed that changes in the shape of the resonant peak are mainly contributed by stimulation-modulated inhomogeneous redistribution of cellular contents parallel to the sensor surface, while the shift in incident angle primarily reflects the stimulation-triggered DMR perpendicular to the sensor surface ^[5]. The standard DMR trace is the resonance shift — reported as a change in incident angle or in resonant wavelength depending on the reader — so it is dominated by the perpendicular component; the parallel component is present in the peak shape and is discarded unless the instrument exposes it.

The quantity behind the angle shift is refractive index: DMR relies on detection of refractive index alterations on biosensor-coated microplates that originate from stimulus-induced changes in the total biomass proximal to the sensor surface ^[6]. The word doing the work is proximal. Lieb and colleagues state the constraint explicitly — the sensitivity of the optical readouts is confined to a distance of 100–200 nm from the surface, depending on the wavelength of the incident light ^[7]. Mass that moves within that shell is counted; mass that moves at the apical surface is not. A DMR trace is therefore a weighted report on the basal few hundred nanometres of the cell, weighted again by how tightly that face sits against the sensor.

What the kinetic trace carries that an endpoint read throws away

The distinguishing feature is not sensitivity but shape. Stimulating the β₂-adrenergic receptor in living cells with isoproterenol generated a complex, multi-featured impedance response over time, and Stallaert and colleagues treated the features rather than the amplitude as the data ^[8]. An endpoint read collapses that to one number chosen by a clock.

Shape can separate ligands. Multi-parameter analysis of DMR in human epidermoid A431 cells responding to diverse β₂-adrenoceptor ligands revealed distinct patterns in activation and signalling of the receptor induced by different agonists, and sequential and co-stimulation assays categorised various ligands for their ability to modulate signalling induced by catechol ^[9]. Working from impedance in a different system, unsupervised clustering of the responses to a collection of β-adrenergic ligands revealed the existence of five distinct compound classes ^[8]. Both results are classifications within one cell background; neither licenses carrying a signature shape across cell lines.

Because the readout is non-invasive, the well survives the experiment. Yu and colleagues note that the non-invasive nature of the impedance readout offers the added advantage of performing multiple additions in the same well, allowing events such as desensitisation and receptor cross-talk to be studied ^[10]. That is the practical inverse of a lysis endpoint: pre-stimulate, wash or not, restimulate, and read the second response against the first in the same population.

The cost is timing discipline. Real-time DMR recordings run in 384-well plates and can be completed in 2 h or extended to several hours depending on the biological question, with the entire procedure including cell harvesting taking 1–2 d ^[6]. Decide before unblinding which feature you will quantify — amplitude at a fixed time, the extremum, the time at which it occurs — because a trace with several phases offers several defensible answers, and choosing after seeing the curves is how a concentration–response fit becomes a story.

Deconvolution: the readout is pathway-agnostic by construction

Since neither sensor contains a pathway-specific element, pathway identity has to be imported by perturbation. Schröder and colleagues set out a strategy to map the cellular mechanisms that define label-free responses; the holistic nature of DMR measurements enabled them to probe GPCR functionality along all four G-protein signalling pathways, which they described at the time as beyond the reach of most other assay platforms, and to define heterotrimeric G proteins as triggers for the complex optical fingerprints ^[11]. The general form of that experiment is: block one arm, rerun the ligand, and attribute the part of the trace that disappears.

Stallaert and colleagues ran it on impedance. Selective pharmacological inhibition of specific arms of the β₂AR signalling network revealed the differential contribution of G_s-, G_i- and Gβγ-dependent signalling events, including activation of the canonical cAMP and ERK1/2 pathways, to specific components of the impedance response; further dissection revealed an essential role of intracellular Ca²⁺ in the impedance response and led to the identification of a β₂AR-promoted Ca²⁺ mobilisation event that had not previously been described for that receptor ^[8]. Note what that last point implies about interpretation: a component of the trace can be real, reproducible, and attributable to a transducer nobody expected in that system.

The tool set is small and each member has a documented specificity. Pertussis toxin is used extensively to probe and inhibit the signalling of G_i/o proteins ^[13]. On the G_q side, YM-254890 blocked Ca²⁺ mobilisation mediated by several Gα_q/11-coupled receptors but not by a Gα_i- or Gα₁₅-coupled receptor, and the data indicated that it blocks the exchange of GDP for GTP in Gα_q/11 activation ^[12]. The depsipeptide FR900359 was systematically characterised as a selective inhibitor of G_q/11/14 over all other mammalian Gα isoforms ^[13]. Both G_q tools act upstream of phospholipase C, so a loss of signal after either one places the requirement at the G protein, not at the effector.

Receptor dependence is a separate control from pathway dependence and needs its own reagent. In A431 cells, the DMR response to the peptide agonist bradykinin showed saturability, and specific concentration-dependent inhibition by the B₂-selective antagonist HOE140 suggested that the optical signature is a direct result of B₂ receptor activation; modulation of several intracellular targets with specific inhibitors then suggested signalling via at least dual G_s- and G_q-mediated pathways which counter-regulate each other, and the signature was sensitive to cholesterol depletion by methyl-β-cyclodextrin ^[14]. Two lessons travel from that study. Counter-regulating arms mean the measured amplitude is a difference, so losing one arm can make a response larger. And a manipulation as generic as cholesterol depletion changes the trace, which is why any inhibitor must also be run alone, without agonist, and ideally in a receptor-negative parental line: pharmacological modulators of signalling pathways are how the specificity of an impedance measurement is demonstrated in the first place ^[10].

Why the trace does not name the G protein

Impedance profiles do carry coupling information under favourable conditions. Peters and Scott describe cellular impedance as able to qualitatively distinguish G_s, G_i/o and G_q signalling in a single assay, and used it to confirm cell-specific G_s and G_q coupling for the melanocortin-4 receptor and dual G_i and G_s signalling with the cannabinoid-1 receptor ^[15]. In a three-platform comparison using G_i-coupled M2, G_q-coupled M1 and G_s-coupled CRF1 receptors, the impedance instrument was uniquely able to qualitatively distinguish G-protein coupling and reveal dual signalling by CRF1 among the three instruments tested ^[16].

The limit sits in the same papers. In CB1-expressing HEK cells, G_i- and G_s-like responses combined to yield a novel impedance profile, and the balance of G_i versus G_s signalling depended on the cell line ^[15]. A receptor engaging more than one transducer therefore produces a shape that is not the shape of either component and is not stable across backgrounds. Reading coupling off a curve is a hypothesis; the assignment comes from the inhibitor series. The field-level version of the same caveat is that issues of data comparability and interpretability have limited wider implementation of cell-based electrical impedance measurements ^[17].

Cell background, receptor expression and adhesion set the response

The most useful single finding for experimental design is the dissociation between trace shape and ligand rank order. Cell-specific signalling was observed with the endogenous D1 receptor in U-2 and SK-N-MC cells, yet the pharmacological profile of partial and full agonists was similar in both cell lines ^[15]. The shape belongs to the cell; the relative pharmacology of a ligand series can survive a change of background. Do not assume the converse holds for a receptor whose coupling balance is itself background-dependent.

Adhesion is the confound specific to these platforms, and it is larger for the optical readout. Comparing ECIS, RWG/DMR and SPR on the human histamine H₁ receptor in U-373 MG and HEK 293T cells, Lieb and colleagues found that reduced cell adhesion to the sensing surface affected both the optical and the impedance-based readout but became much more obvious for the RWG- and SPR-based assays; co-expression of the receptor with the adhesion protein hMSR1 in HEK 293T cells strongly enhanced the signal compared with receptor expression alone, which they interpret as tighter adhesion shortening the distance between membrane and substrate given the 100–200 nm optical sensing depth ^[7]. An engineered clone that adheres more tightly can give a bigger DMR response for reasons that have nothing to do with its receptor.

Confluence enters twice. The impedance baseline of a confluent layer still fluctuates ^[4], and capacitance at high frequency tracks electrode surface coverage as cells spread ^[3], so seeding density and the interval to confluence set the pre-stimulus operating point of every well. Plate a density series once per cell line and per plate format and pick the point where the baseline is stable rather than the point where it is highest.

Receptor expression is the confound these sources do not quantify for you. What they establish is that endogenous-level responses are detectable: an endogenous D1/D5 receptor response was picked up with similar high sensitivity by both impedance and optical biosensors ^[16], and impedance assays for GPCRs work with both engineered and non-engineered cell lines ^[10]. That is a statement about detectability, not about how amplitude scales with receptor number. If you compare a recombinant clone against a native line, expression is an unmeasured variable in the comparison, and a difference in maximal response is not by itself evidence of anything about the ligand.

Set against cAMP and calcium

Second-messenger assays are extremely efficient at capturing one clearly defined cellular readout but, as Grundmann and Kostenis put it, do not authentically portray the events taking place in living cells as a whole; DMR is presented as a good tool for GPCR signalling if combined with diligent assay adjustments ^[18]. The receptors addressed by DMR are exactly those traditionally examined with endpoint assays based on detection of individual second messengers such as cAMP, Ca²⁺ or inositol phosphates ^[6], and the impedance readout includes information on multiple signalling events instead of the activity of a single reporter, which is why it is attractive for receptors whose cascades are unknown ^[17].

Three practical consequences. A label-free response that survives an antagonist and a receptor-negative control is evidence that something receptor-dependent happened, which is worth more than a negative cAMP plate when you do not know the coupling. It is not a substitute for a proximal readout. In a comparison of three label-free instruments across G_i-, G_q- and G_s-coupled receptors, all three were robust in agonist and antagonist modes and yielded comparable potencies and assay variance ^[16]; that agreement was between label-free platforms rather than between a label-free and a second-messenger readout, so a label-free potency matching a cAMP potency is not by itself confirmation of a mechanism. And a divergence between a label-free EC₅₀ and a cAMP EC₅₀ has at least four ordinary explanations before bias — different amplification between the two readouts, different read times on a non-equilibrium response, a second transducer contributing to the integrated trace, and adhesion or confluence differences between the two plate formats. Rule those out with the inhibitor series and matched plating before the word bias enters the figure legend.

The measurement licenses one conclusion cleanly: under these conditions, in this cell background, this ligand produced a receptor-dependent, time-resolved rearrangement of cell mass or cell–substrate contact of this size and this shape. Everything mechanistic beyond that comes from what you knocked out to make part of the trace disappear.