An internalization assay never watches a receptor go inside. It measures something that changes when the receptor goes inside — a biotin tag that stops being reachable, an antibody complex that stops being strippable, a fluorophore that gets quenched or stops being quenchable — and reports that change as a number. Which change you pick determines what the number can mean. This guide covers the main surface-labeling strategies, the pH-sensitive reporters that dominate live-cell work, how recycling is separated from degradation, and the specific dependencies buried inside any internalization percentage.

What is actually being measured

A receptor is the same molecule on both sides of the membrane. What changes when it crosses is accessibility: to reagents in the medium, to the extracellular buffer, and to the acidifying lumen it now sits in. Methods sort themselves by which of those changes they exploit. Rosendale and Perrais classified imaging approaches to endocytosis into four families: distinguishing extracellular from internalized cargo in fixed samples, monitoring endosomal acidification, measuring the turnover of endocytic zones, and detecting vesicle scission. The last three are fluorescence-based and used in living cells [1].

That taxonomy is worth holding onto, because two of those families are often treated as interchangeable when they are not. Acidification and scission are different events separated in time. A method reporting acidification reports scission only if the interval between them is short and constant, an assumption rarely tested in the system at hand.

Surface labeling: making outside distinguishable from inside

One long-standing strategy is to tag the surface pool covalently and then ask how much of the tag has become inaccessible. Cho and Roche describe the biochemical version using sulfo-NHS-SS-biotin, a water-soluble NHS-ester reagent carrying a cleavable disulfide, which makes the labeling reversible; they note that biotinylation is rapid and stable, does not transfer from cell to cell, and that the small size of the biotin probe does not affect cell function. Their framing of the underlying biology matters as much as the chemistry: most, if not all, plasma membrane proteins continuously undergo endocytosis, and many rapidly recycle from endosomes back to the cell surface, which is how apparently stable surface expression is maintained [2].

The reversibility is the whole trick. Label the surface, allow uptake, then apply a membrane-impermeant reducing agent to strip the biotin that is still exposed. What resists stripping was internal at the moment the reducing agent was applied. The choice and concentration of reducing agent, and the number of stripping rounds, are laboratory conventions rather than findings, and stripping efficiency should be measured on a control held at low temperature rather than assumed.

The antibody equivalent — label the surface with an antibody against an extracellular epitope, then discriminate internal from external — carries a trap that Rigo and Vinante state plainly. A fluorochrome-bearing antibody bound to a receptor that translocates from the membrane to the cytoplasm continues to emit light, so the cell remains equally positive for that marker even if the receptor is strongly downregulated or no longer detectable on the membrane. Fluorescence is lost only when the receptor is shed. They built a flow cytometry procedure around that distinction, combining standardized hyperosmolar sucrose, acidic treatment, and staining at different times, and validated it across three receptors with different trafficking patterns [3]. The consequence is blunt: whole-cell fluorescence is not an internalization readout, and a protocol without a surface-stripping or surface-only staining step is measuring binding.

Resonance energy transfer avoids stripping altogether by putting the discriminating reagent in the medium. Levoye and colleagues labeled cell-surface receptors carrying a SNAP-tag with a luminescent terbium cryptate donor and supplied a fluorescein acceptor in the culture medium; internalization removes the donor from the acceptor pool and produces a quantifiable reduction in energy transfer. They reported that both constitutive and ligand-induced internalization could be followed as a function of time and ligand concentration across G protein-coupled receptors belonging to different classes, and reported a Z'-factor of 0.45 for the assay, which they took as indicating suitability for high-throughput screening [4]. A Z'-factor describes separation between plate controls, not trafficking: it says the assay window is workable, not that the biology is well modeled.

pH-sensitive fluorophores

The dominant live-cell approach exploits the fact that the endocytic lumen is acidic relative to the medium. Miesenböck, De Angelis and Rothman generated pH-sensitive mutants of green fluorescent protein — pHluorins — by structure-directed combinatorial mutagenesis, with the explicit aim of exploiting the acidic pH inside secretory vesicles to monitor vesicle exocytosis and recycling; linked to a vesicle membrane protein, they were sorted to secretory and synaptic vesicles and reported transmission at individual boutons [5].

Tagged onto a receptor's extracellular face, superecliptic pHluorin (SEP) is bright at the cell surface and dims as the compartment carrying it acidifies. That is a signal-loss readout, and signal loss has many causes. The pulsed-pH protocol inverts the logic to remove the ambiguity. Sposini and colleagues describe SEP as brightly fluorescent at pH 7.4 and not fluorescent below pH 6.0; a surface-resident SEP therefore dims when the cells are bathed in a pH 5.5 buffer, whereas an internalized SEP remains fluorescent even during application of that buffer. Fast perfusion exchanges the two extracellular solutions every 2 seconds, which sets the temporal resolution, and they note the protocol detects endocytic vesicles without a priori hypotheses about the dynamics of vesicle formation [6].

The distinction is the point. In the steady-state design the discriminating agent is endosomal acidification, so the measurement inherits whatever lag the proton pump imposes; in the pulsed design it is a buffer you control on a clock you set. Two experiments with the same reporter can be measuring different things.

Color has consequences too. Shen and colleagues developed pHuji, a red fluorescent protein with pH sensitivity approaching that of SEP and sufficient for detecting single exocytosis and endocytosis events. Using pHuji together with SEP, they imaged clathrin-mediated internalization of the transferrin receptor and the beta-2 adrenergic receptor simultaneously in two colors, and found the two receptors to be differentially sorted at the time of endocytic vesicle formation [7]. That result is a warning about generalization: an internalization rate measured for one cargo is a statement about that cargo, not about the cell's endocytic capacity.

Reading fate from the pH of the compartment

If lumenal pH varies across the endocytic pathway, it can in principle be used to assign a compartment. Barriere and Lukacs built fluorescence ratiometric image analysis on exactly that premise, determining the post-endocytic fate and transport kinetics of transmembrane proteins from single-cell pH measurement of the compartments holding the internalized cargo. Their stated reasoning is conditional and the hedge should survive: the pH of a cargo-carrying transport vesicle could be taken as an indicator of that vesicle's identity, given that endocytic organelles have characteristic vesicular pH values. Cargo is labeled through a primary antibody against its extracellular domain plus a pH-sensitive FITC-conjugated secondary, and they validate the assignment against colocalization with organellar markers [8].

That validation is not decoration. A pH value assigns a compartment only if the pH distributions of the candidate compartments are separable in that cell type, and separability is a property of the preparation, not a constant.

Recycling versus degradation

Once internalized, cargo can return to the surface, be routed onward for degradation, or move retrogradely. Measuring the split is harder than measuring uptake, and the errors run in a predictable direction. Blagojević Zagorac and colleagues compared several monoclonal-antibody-based recycling protocols on a clathrin-dependent cargo (the transferrin receptor) and a clathrin-independent one (major histocompatibility class I molecules). Both direct and indirect detection of recycled antibody-protein complexes at the cell surface underestimated the recycling pool, they report, especially for clathrin-dependent proteins, which they describe as rapidly reinternalized after recycling. Capture-based protocols, on their account, require Alexa Fluor 488-conjugated or FITC-conjugated secondary antibodies together with inhibitors of endosomal acidification and degradation, and a protocol using Alexa Fluor 488-labeled primary antibody with anti-Alexa Fluor 488 quenching gave the same quantitative assessment as the antibody-capture protocols [9].

Read the direction carefully. If surface-appearance protocols underestimate, then a recycling number obtained that way is a lower bound rather than an unbiased estimate. If a manipulation appears to reduce recycling, one candidate explanation is that it accelerated reinternalization.

The framing of recycling has also shifted. Cullen and Steinberg describe how endocytic recycling was historically regarded as relatively passive, with most internalized integral proteins returning by a nonspecific, sequence-independent bulk membrane flow pathway, and how the discovery of sequence-specific sorting motifs together with the identification of cargo adaptors and associated coat complexes has increasingly challenged that view [10]. A recycling fraction is therefore a measurement of a sorting decision, and a mutation in a sorting motif can change it without changing anything about uptake.

Degradation readouts have their own trap: they report on whichever part of the molecule your label is attached to. Henry and colleagues examined lysosomal trafficking of delta-opioid receptors and found that these receptors can downregulate effectively when their ubiquitination is prevented, in contrast to a number of other signaling receptors. Internalized receptors traversed the canonical multivesicular body pathway to intralumenal vesicles; ubiquitination stimulated but was not essential for transfer to those vesicles and for proteolysis of the receptor endodomain; and ubiquitination played no detectable role in early sorting out of the recycling pathway. Notably, the receptors underwent extensive proteolytic fragmentation in the ectodomain even when ubiquitination was prevented or intralumenal vesicle formation itself was blocked [11]. The authors state that these findings support the hypothesis that mammalian cells can control the cytoplasmic accessibility of internalized signaling receptors independently from their ultimate trafficking fate. The assay-design implication is separate and is ours: an ectodomain-tag readout and an endodomain-tag readout of the same receptor need not agree, and a disagreement between them is information rather than noise.

What an internalization number depends on

Before reporting a percentage, name the denominator. "Forty percent internalized" can mean forty percent of the surface pool labeled at time zero, forty percent of total cellular receptor, or forty percent of the labeling reagent added — three different quantities that diverge as soon as recycling, degradation or new delivery is running. None of them is wrong; only the unstated one is.

Beyond the denominator, an internalization figure carries at least these dependencies:

  • The time window, and whether the value is an initial rate or an endpoint. An endpoint reached after recycling has begun is a net number in which uptake and return have already been summed.
  • Degradation inside the window, which removes label from the numerator and makes uptake look slower than it was.
  • The valency of the labeling reagent. A bivalent antibody can engage two receptors at once; a monovalent ligand cannot. That is a difference in the physical starting condition, whatever it does downstream.
  • Expression level, and whether the tagged construct is overexpressed or replaces the endogenous protein.
  • Temperature, both during labeling and during chase, since the low-temperature condition used as the zero-uptake control is a control only if uptake is genuinely arrested there.
  • The identity of the cargo, per the two-color result above [7].

The most transferable form of the answer is a rate constant rather than a percentage. Waters and colleagues measured surface-bound, internalized and free labeled epidermal growth factor as functions of time on fetal rat lung cells under two protocols and fitted a mathematical model of endocytosis, using sensitivity analysis to ask which parameters were well defined. Three were, in the combination of the two approaches: the endocytic rate constant, the association rate constant and the dissociation rate constant. The endocytic rate constant was found to be independent of initial surface receptor occupancy in the temperature-shift protocol, across occupancies from approximately 200 to approximately 18,000 receptors per cell — roughly a 90-fold span. In the constant-temperature protocol, with initial free ligand concentration varied from 0.2 to 50 ng/mL, a 250-fold span, there was some indication that it increased with ligand concentration [12]. Both hedges belong to the record. Occupancy independence was demonstrated for one protocol on one cell type, and the indication of a concentration dependence comes from the other protocol alone — the temperature-shift experiments varied occupancy in medium initially devoid of free ligand, so they constrain a different variable rather than confirming or contradicting that indication.

Rate constants, and one worked comparison

Rate constants also make perturbations comparable. Presley and colleagues used quantitative fluorescence microscopy to locate which step of recycling is affected by the vacuolar proton pump inhibitor bafilomycin A1 in Chinese hamster ovary cells expressing transfected human transferrin receptors. Removal of transferrin from sorting endosomes and its accumulation in the peri-centriolar endocytic recycling compartment proceeded normally. The rate constant for exit of transferrin receptors from recycling endosomes fell from 0.063 per minute in untreated cells to 0.034 per minute with bafilomycin A1, while bulk membrane recycling, followed as efflux of an internalized fluorescent lipid analog, fell from 0.060 to 0.046 per minute. They concluded that bafilomycin A1 slows bulk membrane flow but causes additional inhibition of receptor recycling dependent on a cytoplasmic-domain peptide motif [13].

The arithmetic, with the denominator named: taking the untreated value as the denominator in each case, the receptor rate constant fell by 0.029 per minute, about 46 percent of 0.063, while the bulk rate constant fell by 0.014 per minute, about 23 percent of 0.060. The receptor-specific fractional slowing is therefore about twice the bulk one, and this is a comparison of fractions, not of absolute rate constants — though here the absolute decreases happen to stand in a similar ratio. If, and only if, exit behaves as a single first-order process from one pool, the two receptor rate constants correspond to half-times of about 11.0 minutes untreated and about 20.4 minutes with bafilomycin A1; that conversion is ours, not the authors'.

What remains convention

Several load-bearing pieces of these protocols are habit. Stripping and quenching efficiencies are usually assumed rather than titrated. The low-temperature block is used as a zero-uptake baseline on the strength of tradition more often than of a measurement in the cell line at hand. The interval between vesicle scission and lumenal acidification is treated as short and uniform even though it is a property of the cell and the compartment. And a compartment identity read from pH is only as good as the separability of the pH distributions being distinguished, which is why colocalization validation accompanies the ratiometric method rather than replacing it [8].

The honest summary is that internalization assays measure accessibility, and accessibility is a proxy. Reporting the proxy explicitly — what was labeled, what was stripped, what the denominator was, and over what window — is what makes one laboratory's number comparable with another's.