Ask whether a peptide "gets into cells" and you have asked three questions at once. Is it associated with the cell? Is it inside a membrane-bounded compartment? Is it in the cytosol, where most intracellular targets are? Those are different measurements with different failure modes, and the history of this field is largely the history of a community discovering that its favorite assay answered the first question while everyone read it as an answer to the third.

The result that reset the literature

Through the 1990s, cationic peptides were widely reported to cross the plasma membrane by a route that was not endocytosis: uptake appeared to continue at low temperature and in the presence of drugs that block active transport, and it was reported to be much faster than known endocytic processes [2]. A 2003 reevaluation of Tat 48–60 and nonaarginine reported that cell fixation, even under mild conditions, leads to artifactual uptake of these peptides. The same work reported that flow cytometry could not be used validly without a trypsin digestion step to remove membrane-adsorbed peptide, that fluorescence microscopy on live unfixed cells gave a characteristic endosomal distribution, and that uptake was inhibited by low temperature and by depletion of the cellular ATP pool [1]. So the pattern people had been reading as cytosolic and nuclear filling was, in live cells of the same kind, punctate. A review from the same group added the second confound: strong ionic interaction with the cell surface led to overestimation of the recorded cell-associated fluorescent signal [2].

The correction was not total, and it is worth being precise about what survived. Time-lapse confocal imaging of nonfixated fibroblasts reported that a fluorescent derivative of the HIV-1 Tat protein transduction domain entered cytoplasm and nucleus within seconds, which those authors read as arguing against the fixation artifact as the whole explanation; enzymatic removal of extracellular heparan sulfate reduced the rate of both uptake and the surface aggregate formation they observed [3]. A separate re-evaluation of the S4(13)-PV peptide, using cells lacking heparan sulfate proteoglycans as a comparator, concluded that endocytosis was evident only at low peptide concentrations and that efficient uptake occurred mainly by an alternative, non-endocytic mechanism, which those authors described as most likely direct penetration [4]. Keep that hedge. The consensus that emerged is not "endocytosis, always" but that the route is a function of the peptide, the cell type, the cargo and the concentration [5].

What the three pathways look like as an experiment

Clathrin-mediated endocytosis, caveolar and other raft-associated routes, and macropinocytosis are distinguished operationally, by what perturbs them and what marker they carry, not by anything visible in a peptide channel. Conventionally — and this is convention, not a finding — labeled transferrin marks the clathrin route, cholera toxin B subunit marks raft and caveolar uptake, and a high-molecular-weight fluorescent dextran marks bulk fluid-phase uptake including macropinocytosis. Colocalization of peptide with a marker is evidence about a shared compartment; it is not evidence that the peptide entered by that pathway, because compartments merge downstream of entry.

Macropinocytosis has a second signature: it is actin-driven and accompanied by membrane ruffling. Uptake of octaarginine by HeLa cells was reported to be significantly suppressed by the amiloride analog EIPA and by the actin polymerization inhibitor cytochalasin D, and the peptide itself induced significant rearrangement of the actin cytoskeleton; penetratin, in the same study, was less sensitive to EIPA and did not produce those distinct actin effects [6]. That contrast is the useful part. Two peptides both called cell-penetrating gave different answers to the same panel in the same cells, which is the reason a pathway assignment cannot be inherited from a related sequence.

Low temperature and energy depletion: necessary, not sufficient

Incubation at 4 °C and ATP depletion are the standard first cut, because vesicular budding and trafficking are energy-dependent and largely stop, while passive partitioning into and across a bilayer does not. Both were reported to inhibit Tat and nonaarginine uptake, consistent with endocytosis [1].

The control has a well-known weakness: 4 °C changes far more than endocytosis. Membrane fluidity, lipid phase behavior, membrane potential and the entire cytoskeleton change with it, so a reduction at low temperature does not identify which of those mattered. And a residual signal at 4 °C is not automatically a direct-translocation result either. Reporting on a hexanoylated octaarginine, one group observed a significant cytosolic distribution at 4 °C, less prominent than at 37 °C, and described this as suggesting a potential ability for direct penetration [7]. That is a suggestion drawn by those authors for that acylated analogue, and it should not be carried over to unmodified oligoarginine.

Chemical inhibitors and what a percentage licenses

Inhibitor panels are the most-used and least-trustworthy element of this literature. A characterization of four commonly used agents — chlorpromazine, genistein, methyl-beta-cyclodextrin and potassium depletion — across five cell lines found that chlorpromazine and, to a lesser extent, methyl-beta-cyclodextrin significantly decreased cell viability in some lines even after short incubations and at concentrations routinely used to inhibit endocytosis. Tracking two probes reported to enter by clathrin-dependent and clathrin-independent mechanisms, the authors showed poor specificity for inhibiting distinct pathways, found the effects highly cell-line dependent, and questioned the significance of studies using these agents without very stringent controls [8].

The dynamin inhibitors fail in an instructive way. In fibroblasts engineered to lack dynamin 1, 2 and 3, fluid-phase endocytosis and peripheral membrane ruffling were not impaired by the absence of all three genes — yet dynasore and Dyngo-4a robustly inhibited both processes in the triple-knockout cells as well as in wild-type cells [9]. An inhibitor that blocks a process in cells lacking its nominal target is reporting something other than that target.

This is where the arithmetic goes wrong most often, so state the denominator before writing any number. If an inhibitor reduces cell-associated peptide signal by 60%, the denominator is the signal in matched vehicle-treated cells run in the same plate on the same day — not the peptide added to the well, and not an untreated culture measured last week. And that 60% is not a bound in either direction. Incomplete blockade of the pathway pushes the true contribution higher than the measured drop; off-target inhibition of other routes, or of the cell's general capacity to do anything, pushes it lower. A partial inhibition gives a point estimate resting on two assumptions that the inhibitor literature repeatedly reports as unmet [8].

Separating what is stuck on from what is in

A cationic peptide binds glycosaminoglycans on the cell surface avidly, and that bound population is in the same optical section and the same flow cytometry gate as anything internalized. Two removal steps are standard practice: brief trypsin treatment, which cleaves the adsorbed peptide, and a heparin or high-salt wash, which competes it off. Trypsinization before flow cytometry was the step whose omission the 2003 reevaluation identified as invalidating the readout [1]. Neither removal is complete, and trypsin also cleaves the receptors, so the step should be reported with its concentration and duration rather than named.

Fixation deserves its own rule. If you fix, you cannot make a claim about subcellular distribution unless you have shown the identical pattern in live cells under your own conditions. Live-cell imaging with a membrane-impermeant viability dye in a separate channel, and gating on it, is the cheaper answer.

Whole-cell signal cannot see the compartment

Here is why total internalized peptide and cytosolic peptide are different quantities, in arithmetic. Let a cell internalize N molecules and have volume V. The whole-cell average concentration is N/V. If all N sit in an endosomal lumen occupying a fraction f of the cell volume, the luminal concentration is N/(fV), which is the whole-cell average divided by f — larger than the average by exactly 1/f — while the cytosolic concentration is essentially zero. Take f = 0.01 as an illustrative round number rather than a measurement: a whole-cell average of 1 µM then corresponds to about 100 µM in the lumen and roughly nothing in the cytosol. The direction matters and is easy to invert. Confinement raises the local concentration above the whole-cell average and lowers the accessible concentration below it. A single fluorescence number is compatible with both extremes.

Measured directly, the two do come apart. Across a panel of hydrocarbon-stapled peptide variants carrying different arginine arrays, one group found no relationship between overall uptake judged by flow cytometry and cytosolic access determined by fluorescence correlation spectroscopy, and concluded that overall uptake and endosomal release are fundamentally different processes [10]. Using the same technique to compare seven previously reported cell-penetrating peptides carrying a SNAP-tag cargo, another study reported that when delivered by the miniature protein ZF5.3 the SNAP-tag cargo reached a cytosolic concentration as high as 250 nM, generally at least 2-fold and as much as 6-fold higher than with any other vehicle evaluated [11]. Note the asymmetry in that pair of figures: the 2-fold is the authors' general lower comparison and the 6-fold the largest they saw, so quoting either alone misstates the result.

Readouts that do not depend on where the dye is

The strongest designs replace the fluorophore with a function that can only occur in the cytosol or nucleus. A transducible Cre recombinase reporter read on live cells was used to conclude that Tat-fusion proteins are internalized, after an initial ionic cell-surface interaction, by lipid raft-dependent macropinocytosis, with transduction independent of the interleukin-2 receptor/raft, caveolar and clathrin routes and of phagocytosis; the same work built a pH-sensitive fusogenic dTAT-HA2 peptide that markedly enhanced escape of the Tat-Cre from macropinosomes [12]. An enzymatic or recombination event is a threshold, not a concentration, so it answers "did any reach the cytosol" rather than "how much."

For amount, label-independent quantification helps. A fluorometric protocol on cell lysates designed to distinguish endocytosis from direct membrane translocation, applied to penetratin, Tat and R9, converged with mass spectrometry quantification of the same peptides, while flow cytometry was limited by fluorescence quenching that depended on the cell line and set in above a peptide-to-cell ratio the authors specify for a carboxyfluorescein-labeled penetratin. Both techniques indicated two alternative routes, translocation and endocytosis, whose relative contribution depended on cell-surface sugars and on peptide concentration; for penetratin, those authors reported translocation at low concentration and endocytosis at high micromolar concentrations [13].

Compare that with the S4(13)-PV finding above, where endocytosis was the low-concentration behavior and the non-endocytic route dominated at higher concentration [4]. The two point in opposite directions along the concentration axis, for different peptides in different cells. Both cannot be generalized into a rule, and neither claimed to be one.

A minimum control set

None of the following is a finding; it is the set of comparisons without which the measurement is not interpretable.

  • Live cells for any subcellular-distribution claim, with fixed-cell images shown only alongside the live equivalent.
  • A surface-removal step — trypsin, heparin or high salt — with concentration and duration reported, plus the no-removal condition for comparison.
  • Viability measured under the exact inhibitor concentration, time and cell line used, not from the supplier's datasheet.
  • A positive-control marker for each pathway invoked, run in the same wells, to show the inhibitor did what it was chosen to do.
  • At least one genetic perturbation where a pharmacological claim is central.
  • A concentration series, since route assignments have repeatedly changed with concentration [5].
  • One readout that does not depend on a fluorophore's location: a recombination, enzymatic or mass spectrometric endpoint.

What is not known

There is no agreed quantitative partition of any peptide's uptake between direct translocation and endocytic routes that holds across cell lines. The endosomal escape step — the fraction of internalized material that reaches the cytosol — is measured directly by only a few techniques and is not routinely reported at all. Whether the surface aggregation and heparan sulfate engagement described for cationic peptides [3] is a required intermediate or an accompaniment is unresolved. And there is no consensus threshold above which a peptide "penetrates": the number depends entirely on which of the three opening questions the assay answered.