Cell-Penetrating Peptides: How Peptides Get Inside Cells

A cell membrane is designed to keep things out. It is an oily double layer that a charged, water-loving molecule the size of a peptide should not be able to cross. So the discovery, in 1988, that a viral protein simply walked into cultured cells was startling — and the short peptide sequences responsible became one of molecular biology’s most useful tools, and one of its most over-promised ones.

Cell-penetrating peptides are the reason some research compounds can reach a target inside the cell at all. They are also the subject of a cautionary tale about microscopy artefacts that reset the field in 2003, and of a delivery bottleneck nobody has solved. This guide covers what they are, how they work, what went wrong with the early evidence, and where they stand today.

What a cell-penetrating peptide is

A cell-penetrating peptide is a short peptide — usually somewhere between five and thirty residues — that crosses the plasma membrane, and that can drag something else across with it. That second property is the useful one. Plenty of interesting molecular targets sit inside the cell: a transcription factor, a protein–protein interface, a stretch of mRNA. A peptide designed to block one of them is useless if it cannot get past the membrane, and most peptides cannot.

The literature also calls these sequences protein transduction domains, or PTDs. The two terms are used interchangeably in much of the field, though “PTD” is sometimes reserved for the specific internalising domain of a named protein — which is why a research compound built on one may be called a PTD fusion rather than a CPP conjugate.

What CPPs are not is selective: a cationic CPP will enter more or less any cell type it meets. That is a feature in a culture dish and a problem in a body, and it is one reason the technology has been far more successful as a laboratory reagent than as a drug-delivery platform.

Two accidental discoveries

In 1988, two groups published back-to-back in the same issue of Cell — Frankel and Pabo, and Green and Loewenstein — and both reported the same unexpected thing: the HIV-1 Tat trans-activator protein could enter cultured cells on its own. Nobody had been looking for a membrane-crossing peptide; they were studying how a virus switches on its own genes.

Three years later, Alain Prochiantz’s group reported that the Antennapedia homeobox peptide was taken up by cultured neurons and altered their morphogenesis. In 1994, Derossi and colleagues narrowed the responsible region to the third helix of the homeodomain, under a title that stated the finding plainly: the third helix “translocates through biological membranes”. That 16-residue peptide, RQIKIWFQNRRMKWKK, became penetratin.

In 1997, Vivès, Brodin and Lebleu narrowed Tat down the same way. Synthesising peptides across residues 37 to 60, they found that “all peptides containing the basic domain are taken up by cells within less than 5 min”, while “a peptide with a full alpha-helix but with a truncated basic amino acid cluster is not taken up by cells”. The basic, arginine- and lysine-rich stretch was doing the work, not the helix. Laboratories still cite that stretch with different boundaries — residues 47–57, 48–60 or 49–57 — so a “Tat peptide” in one paper is not always the same molecule as in another.

Why arginine matters

In 2001, Futaki and colleagues pushed the logic to its conclusion. Testing a set of arginine-rich peptides, including sequences from Tat and from the unrelated HIV Rev protein, they found they all behaved like the Tat peptide despite sharing no common sequence or structure beyond their arginines. Around eight arginine residues gave the most efficient translocation. Arginine is more favourable than lysine for this — the guanidinium group, with its distinctive hydrogen-bonding geometry, rather than positive charge alone.

That is why polyarginine peptides such as R8 and R9 became standard reagents, and why so many engineered CPPs are simply arginine-rich. It is also where the story gets complicated: the same paper reported uptake at 4 °C and read that as energy-independent direct entry — an interpretation that had to be revisited two years later.

The fixation artefact that reset the field

For about fifteen years the standard evidence that CPPs entered cells was a fluorescence micrograph: label the peptide, add it to cells, fix the cells, look at where the signal is. It was usually in the cytosol and the nucleus, and the inference seemed unavoidable.

In 2003, Richard and colleagues published a reassessment in the Journal of Biological Chemistry under a title that said what it was: “Cell-penetrating peptides: a reevaluation of the mechanism of cellular uptake”. Working with the Tat 48–60 peptide and nona-arginine, they reported that “cell fixation, even in mild conditions, leads to the artifactual uptake of these peptides”. Fixing the cells redistributed membrane-bound peptide into the nucleus, manufacturing exactly the picture everyone had been interpreting. Flow cytometry had a parallel problem: the numbers were misleading unless “a step of trypsin digestion of the cell membrane-adsorbed peptide is included in the protocol”, because surface-stuck peptide was being counted as internalised.

In live, unfixed cells with surface peptide removed, the picture changed: the peptides appeared in punctate, endosome-like compartments rather than diffusely in the cytosol, and uptake fell at low temperature and under ATP depletion. The same held for Tat-conjugated peptide nucleic acids, and the authors concluded that endocytosis has a major role in internalisation. Related doubts had surfaced earlier — a 2001 correspondence in Nature Biotechnology asked whether the apparent nuclear homing of the related protein VP22 was an artefact, and a 2002 paper reported that positively charged DNA-binding proteins cause only an apparent membrane translocation.

How they actually get in

The current picture is that there is no single route. Both direct translocation and endocytosis happen, and which dominates depends on the peptide, its concentration, the cargo and the cell type — which is also why published results so often conflict.

Three variables keep turning the answer over. Concentration is “a supreme factor, since it can trigger different uptake pathways” — and it does not even move in a consistent direction, since penetratin shows the opposite pattern to the general rule. Cargo matters: “the size of cargo fused to TAT had a great influence on the uptake mechanism”, with larger Tat–protein conjugates trending toward caveolar uptake. And cell type matters; even confluence changed uptake in CHO cells. A 2020 review of internalisation mechanisms concludes the question “still remains an unsolved piece of the puzzle”.

The endosome problem

If uptake is mostly endocytic, then getting into the cell is not the same as getting into the cell’s interior. Cargo inside an endosome is still topologically outside the cytosol, and endosomes mature into lysosomes, where their contents are degraded. “The release from endosomes seems to be a limiting step in the endocytic uptake of CPPs”, as the same 2020 review puts it.

How limiting was not well quantified until a 2021 study in Nature Communications. Teo and colleagues built a split-luciferase assay able to measure how much protein cargo actually reaches the cytosol, and the numbers are sobering: only about 2% of GFP reached the cytosol in HEK293 cells and about 7% in HeLa cells, with “the majority of material delivered to cells using EEPs” — endosomal escape peptides — “endocytosed and becomes trapped inside endosomes”.

The study’s central finding is the one most often garbled. The cationic CPPs tested did not improve escape efficiency. They increased total cell association by 29- to 49-fold, so more cargo got in, but the proportion that escaped the endosome stayed the same or fell. In the authors’ words, positively charged CPPs “enhance cytosolic delivery as a result of increased non-specific cell membrane association” — “rather than increased endosomal escape efficiency”. More in the door, not more through the second door.

The strategies developed to force escape show how hard it is: fusogenic lipids such as DOPE that change phase at endosomal pH, pH-sensitive peptides such as HA2, “proton sponge” chemistry using histidine or PEI, endosomolytic agents such as chloroquine, and photochemical internalisation. All are interventions bolted on to the delivery system rather than properties of it.

The three classes

CPPs are conventionally sorted by physical chemistry rather than by origin.

• Cationic. Tat, R8, penetratin, DPV3 and DPV6. Uptake driven by positive charge, in practice by arginine content; polyarginine 8- to 10-mers are the most widely studied “because of their high efficiency in cellular uptake”.
• Amphipathic. Distinct polar and non-polar regions: transportan, MAP, pVEC, MPG and the chimeric Pep-1, whose lysine-rich domain comes from the SV40 large T antigen nuclear localisation sequence. These can penetrate directly at lower concentrations than cationic ones.
• Hydrophobic. The smallest group — Bip4, C105Y, melittin, gH625. “There are relatively few numbers of hydrophobic CPPs.”

Many research-compound CPPs are also built with D-amino acids in a reversed sequence — a “retro-inverso” design. Because proteases are stereospecific, D-peptides resist breakdown, and in an extended conformation a retro-inverso analogue’s side chains can align with the parent’s. The caveat matters: a 2021 review warns that “given their general incapability to adopt the 3D structure of the parent peptides their application should be careful evaluated”. It resists proteolysis and sometimes preserves activity; it is not a general solution. Our guide to D-amino acids covers the stereochemistry, and peptide half-life engineering the wider family of stability tricks.

What CPPs are used for in the laboratory

Stripped of therapeutic ambition, CPPs are a good reagent. In cell culture they have been used to deliver peptides, proteins, nucleic acids and nanoparticles; functional antibodies and antibody fragments; phosphorodiamidate morpholino oligomers, using arginine-rich bicyclic peptides; siRNA, using a tricyclic CPP; and gene-editing components — a 2014 Genome Research paper disrupted genes by treating cells with CPP-conjugated recombinant Cas9 protein, with the guide RNA complexed to CPP as condensed, positively charged nanoparticles.

Two research compounds in our own catalogue exist because of this chemistry.

FOXO4-DRI is the senolytic peptide from Baar and colleagues’ 2017 paper in Cell. The design problem was that the target is a protein–protein interaction inside the nucleus: FOXO4 holding p53 away from its apoptotic role in senescent cells. The solution was a peptide covering the p53-interaction region of FOXO4, built in a D-retro-inverso conformation and fused to HIV-Tat for cellular uptake — so the penetrating moiety is Tat-derived, not FOXO4-derived. In senescent cells the paper reports it “selectively causes p53 nuclear exclusion and cell-intrinsic apoptosis”, and in mice it neutralised doxorubicin-induced chemotoxicity and restored fitness, fur density and renal function in fast-ageing and naturally aged animals. That evidence is cells and mice; it has not been tested as a medicine in people. Our FOXO4-DRI guide covers it in detail.

PTD-DBM names its own architecture: a protein transduction domain joined to the Dishevelled-binding motif of CXXC5. CXXC5 dampens Wnt/β-catenin signalling by binding Dishevelled; the peptide competes for that interaction, and the PTD is what gets it inside. Kang-Yell Choi’s group at Yonsei University described the construct as including “a PTD for enhanced protein delivery, a linker for flexibility, DBM, and lysine conjugated with FITC for visualization”. Its evidence is also preclinical — cultured cells and mice — and PTD-DBM is not an approved treatment for anything.

What happened in the clinic

A 2020 review in Frontiers in Pharmacology stated it plainly: “no CPPs or CPP/cargo complexes have been approved by the US Food and Drug Administration”. We are not aware of one since.

That is not for want of trying. XG-102, also called D-JNKI-1 or brimapitide, is a Tat-fused JNK-inhibiting peptide in a D-amino-acid form; it reached phase 3 for postoperative ocular inflammation, where its effect was reported as lower than dexamethasone eye drops. AM-111, a Tat-conjugated D-JNKI-1 for acute inner-ear hearing loss, also reached phase 3; its HEALOS trial “did not meet the primary efficacy endpoint” at either dose. AZX100, an anti-fibrotic peptide for dermal scar reduction, was abandoned and its licence terminated in 2014.

What’s established and what isn’t

Sourcing CPP-based reagents in Canada

CPP-fused peptides are longer and more complex than the compounds most research buyers handle, which changes what a useful certificate looks like.

• Ask for the full sequence and expected mass for your lot. A Tat-fused, D-retro-inverso peptide can run past forty residues; without the sequence you cannot check a reported mass against anything.
• Expect a resolved chromatogram, not just a percentage. Long synthetic peptides accumulate deletion and truncation products, so peak shape and resolution matter more than a bare figure. See how to read an HPLC chromatogram.
• Remember what mass alone cannot tell you. Mass spectrometry cannot distinguish stereoisomers, so it cannot confirm that a D-retro-inverso peptide is in fact D-form. That needs chiral or derivatisation methods. See how to read a peptide mass spectrum.
• Match the batch number on the certificate to your vial, and keep it with your lot records. Our COA guide covers the rest.
• Store cold, dry and dark, and keep material lyophilized until use. See the peptide storage guide and what lyophilized means.

PTD-DBMA protein transduction domain fused to the CXXC5 Dishevelled-binding motifView product FOXO4-DRIA Tat-fused, D-retro-inverso peptide aimed at the FOXO4–p53 interactionView product

Frequently asked questions