Phage Display: Discovering Tumor-Homing Peptides

Tumor-Targeting Peptides

10^9+ peptides screened

A single phage display library can contain over a billion unique peptide sequences. Injected into tumor-bearing animals, only the sequences that bind tumor tissue survive selection, producing peptides with natural homing specificity.

Pasqualini & Ruoslahti, Nature, 1996

Pasqualini & Ruoslahti, Nature, 1996

View as image

Most cancer drugs fail not because they cannot kill tumor cells but because they cannot reach them. The delivery problem, getting enough drug into tumors without poisoning healthy tissue, remains the central obstacle in oncology. Tumor-homing peptides offer a solution: short amino acid sequences that bind selectively to markers on tumor vasculature or cancer cells, guiding attached drugs or imaging agents to the right location. The primary method for discovering these peptides is phage display, a technique that screens libraries of over a billion random sequences against actual tumor tissue in living animals. For a broader view of tumor-targeting peptide classes, see The Expanding Library of Tumor-Targeting Peptides.

Pasqualini and Ruoslahti published the foundational demonstration of in vivo phage display for organ targeting in Nature in 1996, showing that peptides capable of selective homing to specific organs could be isolated from random libraries injected into living mice.^[1] This approach, extended to tumor models within two years, created a discovery platform that has produced hundreds of validated tumor-homing sequences. The method works because it requires no prior knowledge of the target molecule. The peptides find their targets through competitive binding in a real biological environment, capturing interactions that rational drug design cannot predict.

What phage display is and how it works

Phage display uses bacteriophages, viruses that infect bacteria, as molecular scaffolds. A library is created by inserting random DNA sequences into a phage coat protein gene, so each phage particle displays a unique peptide sequence on its surface. The M13 filamentous phage is the most commonly used scaffold, with peptides typically displayed on the pIII minor coat protein (5 copies per phage, higher affinity) or the pVIII major coat protein (up to 2,700 copies, higher avidity).^[2]

A standard library contains 10^9 or more unique peptide sequences, each 7 to 12 amino acids long. This diversity means the library samples an enormous fraction of possible peptide sequences at that length. The library is a physical collection of phage particles, each carrying its own identifying DNA. After selection, the "winning" peptides are identified simply by sequencing the DNA of the recovered phage. George Smith, who first demonstrated peptide display on phage surfaces in 1985, shared the 2018 Nobel Prize in Chemistry for this work.^[3]

The selection process, called biopanning, operates by Darwinian logic. Phages displaying peptides that bind a target are retained; those that do not are washed away. After amplification in bacteria, the enriched pool is reapplied to the target. Three to five rounds of this cycle progressively concentrate the best-binding sequences from billions of candidates down to a handful.

In vivo biopanning: selection inside living animals

The critical advance that made phage display relevant to tumor targeting was moving the selection from a test tube to a living animal. In vitro biopanning (selection against purified proteins or cultured cells) identifies peptides that bind specific molecules. In vivo biopanning identifies peptides that actually reach and accumulate in target tissues when injected into the bloodstream.^[1]

Pasqualini and Ruoslahti (1996) injected a phage display library intravenously into mice and allowed it to circulate for a defined period (typically 5-15 minutes). They then perfused the animals to wash out unbound phage, harvested target organs, and recovered bound phage by infection of bacteria. The recovered phage were amplified and reinjected for additional rounds. After three rounds, organ-specific peptides emerged: sequences that homed preferentially to brain or kidney vasculature with up to 13-fold selectivity over control organs.^[1]

This in vivo approach captures information that in vitro methods miss. A peptide must survive serum proteases, avoid liver and spleen clearance, reach the tumor vasculature, and bind in the presence of competing blood proteins. The selection pressure is biological reality, not an artificial surface. Peptides that emerge from in vivo biopanning have already demonstrated pharmacokinetic viability.

Extending this to tumor models followed quickly. By injecting libraries into mice bearing human tumor xenografts, researchers identified peptides that home specifically to tumor vasculature. The tumor vasculature is an attractive target because it is directly accessible from the bloodstream, it differs molecularly from normal vasculature (overexpressing integrins, aminopeptidase N, and other markers), and it is shared across many tumor types regardless of the specific cancer.

Key peptides discovered through phage display

RGD and the integrin family

The tripeptide motif arginine-glycine-aspartate (RGD) was identified through phage display as a binding sequence for integrin receptors, particularly alphavbeta3 and alphavbeta5, which are overexpressed on angiogenic tumor vasculature. The cyclic version, RGD-4C (CDCRGDCFC), showed improved binding affinity and selectivity for tumor blood vessels over the linear sequence. Integrins are surface receptors that mediate cell-matrix adhesion, and their overexpression on tumor endothelial cells makes them one of the most validated targets for vascular homing.^[4]

Liu et al. (2017) reviewed the full landscape of tumor-targeting peptides from combinatorial libraries, documenting how RGD-based peptides have been conjugated to chemotherapy drugs, nanoparticles, and imaging agents. Radiolabeled RGD peptides have entered clinical trials for PET imaging of tumor angiogenesis, providing a non-invasive readout of integrin expression that can guide treatment decisions.^[4]

For a deep dive on RGD peptide biology and clinical applications, see RGD Peptides: Homing In on Tumor Blood Vessels via Integrins.

NGR: targeting aminopeptidase N

The asparagine-glycine-arginine (NGR) motif was discovered through in vivo phage display as a homing peptide for tumor vasculature. Its receptor was identified as aminopeptidase N (CD13), an enzyme overexpressed on endothelial cells during angiogenesis. The cyclic form, CNGRC, demonstrated selective binding to tumor blood vessels in breast carcinoma, melanoma, and Kaposi's sarcoma models. NGR peptides have been used to deliver tumor necrosis factor-alpha (TNF-alpha) selectively to tumor vasculature, achieving anti-tumor effects at doses too low to cause systemic toxicity.^[4]

iRGD: the tumor-penetrating breakthrough

The most significant recent advance in phage display-derived tumor targeting is iRGD (CRGDKGPDC), a peptide that combines two functions in one sequence. The RGD motif binds alphavbeta3/alphavbeta5 integrins on tumor vasculature (homing function), and after proteolytic cleavage at the tumor site, the exposed C-terminal RGDK sequence binds neuropilin-1 (NRP-1), triggering the C-end Rule (CendR) pathway that drives tissue penetration. This dual mechanism allows iRGD to not only find tumors but actively transport attached cargo through the vascular wall and into the tumor parenchyma.^[5]

Singh et al. (2024) demonstrated the practical value of iRGD by conjugating it to camptothecin, a topoisomerase inhibitor with poor tumor penetration. The iRGD-camptothecin conjugate achieved approximately 3-fold higher tumor accumulation compared to free drug and showed enhanced antitumor activity in colon cancer xenograft models. The conjugate exploited the sequential binding mechanism: integrin-mediated vascular homing followed by neuropilin-mediated tissue penetration.^[5]

Wang et al. (2018) used the same iRGD strategy to enhance thymosin alpha-1, an immunomodulatory peptide, by creating a Talpha1-iRGD fusion protein. The fusion retained thymosin's immune-stimulating activity while gaining tumor-specific delivery, producing enhanced anti-melanoma effects in mouse models compared to unconjugated thymosin alpha-1.^[6]

Cyclic libraries: constraining peptides for higher affinity

Linear peptides are flexible, which means they pay an entropy penalty when binding a target because they must adopt a defined conformation from many possible states. Cyclic peptide libraries, where the displayed peptide is constrained by a disulfide bond or chemical crosslink, reduce this penalty and produce hits with substantially higher binding affinity.

Deyle et al. (2017) reviewed phage selection of cyclic peptides and found that cyclic hits typically show 10-100 fold higher affinities than linear counterparts against the same target. Beyond affinity, cyclization also improves metabolic stability: cyclic peptides resist serum proteases far better than linear sequences, a critical advantage for any peptide that must survive in the bloodstream long enough to reach a tumor.^[7]

Bicyclic peptide libraries, where two loops are constrained by a chemical scaffold, push this further. These structures can achieve antibody-like affinities (low nanomolar) while maintaining the small size (1-2 kDa) that enables tissue penetration inaccessible to antibodies (150 kDa). Several bicyclic peptides from phage display have entered clinical development as tumor-targeting agents.^[7]

Beyond the vasculature: tissue-penetrating peptides

Classical in vivo phage display recovers peptides that bind the vascular surface because unbound phage are washed out before they can penetrate tissue. This means the technique is biased toward finding vascular homing peptides, missing sequences that could penetrate into the tumor parenchyma, where most cancer cells reside.

Pemmari et al. (2025) addressed this limitation by combining in vivo phage display with microdialysis-based recovery. Instead of homogenizing the entire tumor to recover phage, they implanted microdialysis probes in the tumor parenchyma and collected only phage that had extravasated from the vasculature and reached the interstitial space. This approach selected specifically for tissue-penetrating peptides rather than just vascular binders. The recovered sequences were different from those found by conventional biopanning, confirming that the two methods access different molecular targets.^[8]

Kang et al. (2020) demonstrated the value of brain tumor-homing peptides discovered through phage display. A four-amino-acid sequence identified by screening against glioblastoma delivered therapeutic nanoparticles across the blood-brain barrier and into brain tumors in mice, a delivery challenge that most antibody-based strategies cannot solve because of size constraints. The tetra-peptide improved median survival in glioblastoma-bearing mice compared to untargeted nanoparticles.^[9]

Modern enhancements to the platform

High-throughput sequencing replaces clone picking

Traditional phage display required picking individual phage clones after biopanning and sequencing them one by one, limiting analysis to dozens or hundreds of sequences. Next-generation sequencing (NGS) now enables sequencing of millions of recovered phage in a single experiment. This transforms biopanning from a search for individual "winner" sequences into a population-level analysis of enrichment patterns. Sequences that are statistically enriched above background across multiple rounds can be identified with high confidence, even if no single clone dominates the output.^[8]

Nanotechnology integration

Goracci et al. (2020) reviewed how phage display-derived peptides are being integrated into cancer immunotherapy platforms. Tumor-homing peptides identified by biopanning are conjugated to nanoparticles carrying immune checkpoint inhibitors, cytokines, or tumor antigens, creating targeted immunotherapy delivery vehicles. The peptide provides tumor selectivity; the nanoparticle provides drug payload capacity. Phage particles themselves have also been explored as vaccine platforms, displaying tumor antigens on their surfaces to stimulate anti-tumor immune responses.^[10]

New targets: oncofetal proteins

Lingasamy et al. (2025) demonstrated targeting of fibronectin Extra Domain-B (Fn-EDB), an oncofetal protein expressed in tumor stroma but absent from adult healthy tissues. The PL2 peptide, identified through affinity selection, homes to Fn-EDB-positive tumors and simultaneously binds neuropilin-1, achieving dual-target selectivity. This study illustrates how the tumor-homing peptide field continues to expand into new target biology beyond the original integrin and aminopeptidase targets.^[11]

Limitations of phage display for tumor targeting

Phage display is powerful but not without constraints. The M13 phage scaffold limits displayed peptides to sequences compatible with phage assembly; some peptide sequences disrupt coat protein folding and are underrepresented in libraries. The in vivo selection depends on the specific tumor model used, and peptides selected against one xenograft may not transfer to other tumor types or to human cancers. The transition from mouse models to human patients remains a major translational gap.

The short circulation time of phage in vivo (clearance by liver and spleen) biases selection toward peptides that bind rapidly, which may not be the tightest binders. Extended circulation versions using PEGylated phage or modified injection protocols are being explored but add complexity to an already demanding experimental workflow.

Finally, phage display identifies peptides that bind to tumor tissue but does not itself reveal the molecular identity of the target receptor. Target deconvolution, determining what the peptide actually binds to, requires separate experiments (affinity chromatography, cross-linking mass spectrometry, or genetic approaches). Some clinically advanced tumor-homing peptides still have incompletely characterized receptors.

For related topics, see How Cell-Penetrating Peptides Escape Endosomes, How Amphipathic Peptides Target Cancer Cell Membranes, and Bombesin Peptides: Targeting Receptors Overexpressed in Cancer.

The Bottom Line

Phage display is the dominant discovery platform for tumor-homing peptides, enabling unbiased selection of binding sequences from libraries of over a billion candidates. The technique has produced validated peptides targeting integrins (RGD), aminopeptidase N (NGR), and neuropilin-1 (iRGD), with clinical applications in drug delivery, imaging, and immunotherapy. Recent advances in cyclic libraries, microdialysis-coupled biopanning, and next-generation sequencing are expanding both the quality and scope of peptides that can be discovered.

Frequently Asked Questions

What is phage display for peptide discovery?

Phage display is a laboratory technique that uses bacteriophages (bacterial viruses) to screen billions of random peptide sequences for binding to a specific target. Each phage particle displays a unique peptide on its surface and carries the DNA encoding that peptide. Through iterative rounds of selection called biopanning, peptides that bind the target are enriched while non-binders are eliminated. George Smith developed the technique in 1985 and received the 2018 Nobel Prize for it.

How does in vivo biopanning work?

In vivo biopanning injects a phage display library directly into the bloodstream of a living animal, typically a mouse bearing a human tumor. After 5-15 minutes of circulation, unbound phage are washed out by cardiac perfusion. Phage bound to tumor vasculature are recovered, amplified in bacteria, and reinjected for additional rounds. After 3-5 rounds, peptides that specifically home to tumor tissue are enriched. Pasqualini and Ruoslahti (1996) first demonstrated this approach in Nature, achieving 13-fold organ selectivity.

What is the RGD peptide and why does it target tumors?

RGD (arginine-glycine-aspartate) is a tripeptide motif discovered through phage display that binds alphavbeta3 and alphavbeta5 integrins overexpressed on tumor blood vessel endothelium during angiogenesis. The cyclic form RGD-4C shows improved binding affinity and tumor selectivity. RGD peptides have been conjugated to chemotherapy drugs and radiolabeled for PET imaging of tumor angiogenesis, with several entering clinical trials (Liu et al., Advanced Drug Delivery Reviews, 2017).

What is iRGD and how does it penetrate tumors?

iRGD (CRGDKGPDC) is a cyclic peptide that combines tumor homing with tissue penetration. Its RGD motif binds integrins on tumor vasculature, then proteolytic cleavage exposes a C-terminal sequence that activates the CendR pathway through neuropilin-1, driving transport across the vascular wall into tumor tissue. Singh et al. (2024) showed that iRGD-camptothecin conjugates achieved 3-fold higher tumor accumulation compared to free drug in colon cancer models.

Why are cyclic peptide libraries better than linear ones?

Cyclic peptide libraries produce hits with 10-100 fold higher binding affinities than linear libraries because the constrained backbone reduces the entropy penalty of target binding. Cyclization also improves resistance to serum proteases, extending the peptide's half-life in the bloodstream. Bicyclic peptides, with two constrained loops, can achieve antibody-like affinities while remaining small enough to penetrate tumor tissue (Deyle et al., Accounts of Chemical Research, 2017).

Can phage display find peptides that penetrate tumors, not just bind the surface?

Yes. Pemmari et al. (2025) combined in vivo phage display with microdialysis probes implanted in the tumor parenchyma, recovering only phage that had extravasated and penetrated into the tissue interstitium. This method selects specifically for tissue-penetrating peptides rather than vascular surface binders. The recovered sequences differed from conventional biopanning results, confirming that the two approaches access different molecular targets.