In Vivo Translation of Peptide-Targeted Drug Delivery Systems Discovered by Phage Display

Abstract

Therapeutic compounds with narrow therapeutic windows and significant systemic side effects benefit from targeted drug delivery strategies. Peptide-protein interactions are often exploited for targeting, with phage display a primary method to identify high-affinity peptide ligands that bind cell surface and matrix bound receptors preferentially expressed in target tissues. After isolating and sequencing high-binding phages, peptides are easily synthesized and chemically modified for incorporation into drug delivery systems, including peptide-drug conjugates, polymers, and nanoparticles. This review describes the phage display methodology to identify targeting peptide sequences, strategies to functionalize drug carriers with phage-derived peptides, specific examples of drug carriers with in vivo translation, and limitations and future applications of phage display to drug delivery.

Graphical Abstract:

Phage display is used to identify peptides that bind target molecules, such as cell receptors expressed in diseased tissues, for generating targeted drug delivery systems.

Motivation

Drug delivery strategies are designed to address challenges associated with drug bioavailability, stability, solubility, and toxicity, as well as physiological barriers to cellular entry.^1–3 Compounds with narrow therapeutic indices are particularly well-suited for administration by targeted drug delivery systems to avoid off-target effects in healthy tissues.⁴ Early drug targeting systems exploited antibody-functionalized liposomes to enhance cell uptake and reduce systemic off-target accumulation.^{1, 5} Recent efforts focus on peptides as targeting ligands, which can exhibit homology to antibodies and other native proteins, but are more economical and reproducible and are less likely to induce an immune response. ^{6, 7}

This review focuses on targeting peptides specifically identified through phage display, as the process can be used to screen against a range of targets such as whole cells or proteins to produce targeting ligands with affinities in the nanomolar range.^{8, 9} Phage display is a multi-step selection and amplification process in which a random pool of 10⁹-10¹⁰ random peptide sequences, each 7–12 amino acids long, is narrowed in as few as 3 rounds over a span of weeks.¹⁰ This review describes the phage display process, highlights examples of phage display in targeted drug delivery, acknowledges limitations to phage display, and provides perspective on future applications of phage display in drug delivery.

Phage Display

Targeting peptides are often isolated via phage display. Bacteriophages, or more colloquially “phages,” are viruses that infect and replicate within a host bacterium. Phages’ proteins are surrounded by their encoding DNA, directly linking phenotype to genotype.⁹ The first, and now most commonly used, phage for screening peptides libraries is the filamentous M13 bacteriophage; others include T4, T7, and λ phages.^11–13 DNA sequences can be inserted into a virus without disrupting normal function, leading to synthesis of the encoded peptide on the surface of the virus.⁹ In Smith’s seminal report, foreign DNA fragments were inserted into Gene III of M13 phage, resulting in expression of a foreign peptide on the pIII minor coat protein.¹¹ Typically, five peptide clones display the N-terminus of the foreign peptide, and the C-terminus is connected to the rest of pIII buried within the virion.⁹ The display of only five peptides limits confounding multivalent effects, unlike modification of Gene VIII, which can result in up to 2700 copies of peptides displayed on pVIII major coat proteins.¹⁴ The significant avidity of multicopy pVIII phage display libraries may overshadow relatively weak affinity of a single peptide clone.^{9, 13} In other words, target binding of a low-affinity peptide causes neighboring low-affinity peptides to bind to the target, yielding a higher functional affinity (avidity) than the sum of individual peptides.^{15, 16} The contribution of avidity can be measured using techniques such as isothermal calorimetry or surface plasmon resonance (SPR) spectroscopy. Avidity effects are minimized by reducing the number of displayed peptide clones via phages (pVIII or pIII) with both wild-type and recombinant genes.¹⁷

The process of phage display begins with generating a random peptide library (Figure 1A). Oligonucleotides (NNK)_x(GGC)_y are inserted upstream of wild-type Gene III. Each “NNK” is a codon, where “N” is any nucleotide (A, C, G, or T) and “K” is usually either G or T to limit the number of possible codons.^{13, 18} This yields 32 different codons that translate into “x” amino acids in the displayed peptide, and “(GGC)_y” translates into “y” glycine spacers between foreign peptide and pIII. It is noted that this method may not generate phages displaying every possible amino acid combination, as there is bias against amino acids with only one codon, and some sequences may interfere with normal coat protein folding.^{9, 18, 19} After generating a library, phages are exposed to immobilized molecular probes ranging from small molecules to proteins to whole cells in a process known as panning.¹² The method of panning is a multi-step process described in detail by Kay et al. and Freund et al.^{13, 20} Briefly, the target is immobilized and phages are introduced in solution (Figure 1B). Unbound phages are washed away, and bound phages are stripped from the target using acidic, basic, or denaturing conditions. Recovered M13 phages are infected into E. coil to amplify (Figure 1D), and the series repeats for a chosen number of rounds.¹⁴ Sequencing the DNA of phages after the final selection round identifies high-binding peptide sequences (Figure 1E).⁹ For example, a random hexapeptide library of 3×10⁸ phages was panned against 3-E7 monoclonal antibodies, resulting in 51 high-binding peptide sequences after three rounds of selection and amplification.¹⁸ The antibody 3-E7 is specific to β-endorphin, which has an N-terminus beginning with Tyr-Gly-Gly-Phe, and 94% of the recovered phages displayed peptides beginning with Tyr-Gly. This highlights the ability of phage display to identify specific epitopes out of a highly diverse library.²¹

Figure 1.

Phage display methodology. A random library of clones (A) is panned in vitro (B) or in vivo (C) to identify target-bound phages. Phages are eluted from the target, amplified (D), and re-panned in multiple rounds of selection. Phages present in the final round of selection are sequenced to identify high-binding peptide clones (E).

More robust phage selection has been achieved through parallel in vitro and in vivo screens. A thorough in vitro and in vivo phage display selection process by Whitney et al. identified tumor-targeting peptides.²² In this study, hexahistidine-tagged phages (Figure 1A) were mixed with liver, kidney, and tumor extracts to identify sequences that were cleaved by only tumor enzymes (Figure 1B). The positively selected phages, which were those cleaved by tumor extracts but not liver and kidney extracts, were amplified and screened a total of six rounds (Figure 1D). In parallel, the library was injected into a transgenic breast cancer murine model, and cleaved phages were isolated from tumor extracts (Figure 1C). The recovered phages were amplified and injected into additional tumor-bearing mice for a total of 15 rounds (Figure 1D). One amino acid sequence, RLQLKL, was identified in both in vitro and in vivo screens. Further investigation demonstrated the utility of the peptide as an imaging agent for primary and metastatic tumors.²²

Applications of Phage Display to Targeted Drug Delivery Systems

A wide variety of targeting peptide sequences have been identified through phage display. From the perspective of targeted drug delivery, cell receptors and extracellular matrix proteins that are specific to or enhanced by diseased tissues are used as bait.²³ Delivery systems functionalized with these targeting peptides are designed to preferentially distribute to target tissues upon systemic administration, limiting off-target carrier accumulation and adverse drug effects. Classes of drug delivery systems discussed in this review are highlighted in Figure 2 and include monovalent/multivalent peptide-drug conjugates, peptide-polymer-drug conjugates, monovalent/multivalent peptide-liposome conjugates, and peptide-micelle conjugates. Specific examples of chemical strategies to incorporate targeting peptides into delivery systems and applications that highlight the benefits of each system are summarized in Tables 1–2.

Figure 2.

Configurations of drug delivery systems with targeting ligands (green chevrons) and drug payloads (orange crosses). Highlighted are simple monovalent peptide-drug conjugates (A), multivalent peptide-drug conjugates (B), multivalent peptide-polymer-drug conjugates (C), monovalent (D1) and multivalent (D2) peptide-liposome conjugates, and peptide-micelle conjugates (E).

Table 1:

Summary of studies investigating phage display to design targeted peptide conjugates.

Indication	Target	Targeting Peptide	Result	Ref
Bioactive Targeting Peptides
Kaposi’s sarcoma xenograft in nude mice	Gelatinase B	CTTHWGFTLC	Inhibited endothelial and tumor cell migration in vitro. Prevented tumor growth in vivo, whereas tumor size increased five-fold after control peptide treatment.	24
Mammary carcinoma murine model	Aminopeptidase A (APA)	CPRECESIC	Inhibited APA enzymatic activity, suppressed endothelial cell migration and proliferation, and inhibited capillary tube formation in vitro. Suppressed neovascularization and inhibited tumor growth in vivo.	25
Breast cancer and melanoma xenografts in nude mice	Tumor and endothelial cell lymphatics	CGNKRTRGC	Induced apoptosis in vitro. Homed to tumors and metastatic lesions, inhibited tumor growth, and reduced tumor lymphatics in vivo.	26, 27
Small cell lung cancer and prostate cancer xenografts in nude mice	E2F-1 transcription factor	HHHRLSH	Conjugated to penetratin to facilitate cell uptake, peptide was cytotoxic to a range of tumor cells in vitro. Encapsulated in liposomes to increase serum stability, peptide inhibited tumor volume growth.	28, 29
Monovalent Peptide-Drug Conjugates
Breast carcinoma xenograft in nude mice	Aminopeptidase N (APN), or CD13, in tumor vasculature	CDCRGDCFC-Dox and CNGRC-Dox (doxorubicin)	Tumor reduction by one-fourth to one-fifth the size of control treated tumors, fewer metastases, and longer survival relative to untargeted Dox	30
Lymphoma and melanoma in murine model	APN30	CNGRCG-TNF1–157	Efficacy:toxicity ratio 14 times greater than untargeted TNF. Similar pharmacokinetics as untargeted TNF.	31
Prostate cancer xenograft in nude mice	LNCaP cells	GTIQPYPFSWGY-D(KLAKLAK)2	Induced LNCaP cell death in vitro. Exhibited high biodistribution to tumors in vivo.	32
Colon carcinoma in murine model	Tumor-associated macrophages (TAMs)	YEQDPWGVKWWY GGGS-D(KLAKLAK)2	M2 macrophage specific binding and uptake in vitro. Delayed mortality and selective elimination of TAMs in vivo.	33
Multivalent Peptide Conjugates
Non-small cell lung cancer xenografts in nude mice	αvβ6 integrin on H2009 cells34	RGDLATLRQLAQE DGVVGVR35	Greater affinity and serum stability of tetrameric peptides (relative to mono-, di-, and trimeric) to cells in vitro. Five-fold greater tumor-to-blood ratio for tetrameric peptides relative to scrambled control in vivo.	36
Non-small cell lung cancer xenografts in nude mice	Human NCI-H460 lung cancer cells	CRCPLSHSLICYC	Time- and dose-dependent uptake by lung cancer cells in vitro. Higher tissue distribution to xenografts relative to untargeted dendrimers in vivo.	37
Bladder carcinoma xenografts in nude mice	Carbohydrate-associated antigens on tumor cells38	SGPG[VG-SNRDARR-G-(VGVPG12)]5	1.3-fold greater binding than monomeric SNRDARR and 17-fold greater binding than controls to bladder cancer cells in vitro. Time-dependent, 2.3-fold greater tumor accumulation relative to controls in vivo.	39
Breast carcinoma xenografts in nude mice	Interleukin-4 receptor40	[VG-RKRLDRN-G-(VGVPG)12]6	Binding to and uptake by breast carcinoma cells in vitro. Exerted cytotoxicity by blocking IL-4 receptors. Tumor tissue retention of at least 24 h in vivo.	41
Prostate cancer and breast cancer xenografts in nude mice	Clotted plasma proteins and fibrinogen42 and Tumor lymphatics27	CREKA and CGNKRTRGC	Extravasation into tumor tissue (both prostate- and breast-derived tumors) in vivo.	43
Multivalent Peptide-Polymer Conjugates
Melanoma metastasis to lung in murine model	E-selectin44	DITWAQLWDLMK	Binding to brain capillary endothelium cells in vitro. pH-sensitive hydrazide-mediated release of doxorubicin or D(KLAKLAK)2 from 34 kDa copolymers prolonged survival.	45
Primary lung carcinoma and melanoma pulmonary metastases	Vascular Endothelial growth factor receptor-1 (VEGFR-1)46	WHSDMEWWYLLG	Binding to VEGFR-1 expressing endothelial cells in vitro. Tumor-specific doxorubicin delivery inhibited tumor growth and extended survival.	47
Fracture in murine model	Tartrate-resistant acid Phosphatase (TRAP)6	TPLSYLKGLVTVG	Enhanced TRAP binding by random polymers relative to gradient polymers in vitro. Greater fracture persistence (at least one week) by 65 kDa random polymers.	48

Table 2:

Summary of multivalent nanoparticles (NPs) with phage-derived targeting peptides.

Indication	Target	Targeting Peptide	Nanostructure and size	Result	Ref
Post-assembly peptide conjugation
Femur fracture in murine model	Tartrate-resistant acid phosphatase6	TPLSYLKGLV TVG	Micelle formation of PSMA-PS diblock copolymers by solvent exchange; carbodiimide peptide conjugation to form 50 nm micelles	Two-fold greater fracture accumulation relative to scrambled control in vivo. Fracture-specific drug delivery expedited healing and improved bone mechanical properties.	62
Liver-colonizing Colon 26 NL-17 cells in murine model	Human endothelial progenitor cells (hEPCs)	ASSHN	DPPC, CHOL, and DSPE-PEG2000-MAL liposomes modified with ASSHNC peptide to form 101 nm particles	Intratumoral distribution in vivo. Tumor-specific doxorubicin delivery reduced tumor volume more than untargeted control.	63
Neuroblastoma xenografts in nude mice	APA25 and APN30	CPRECES and CNGRC	Dual conjugation of peptides to HSPC, CHOL, DSPE-PEG2000-NH2 liposomes via BS3 cross-linker to form 90–115 nm liposomes	Superior tumor uptake of dual-targeted liposomes in vivo. Tumor-specific doxorubicin delivery led to tumor vasculature-associated endothelial cell and pericyte death and significantly greater life spans of mice.	64
Lung cancer xenografts in nude mice	Tumors responding to VEGF inhibitor treatment	HVGGSSV65	100 nm liposomes composed of DSPC, CHOL, and MAL-PEG2000-DSPE modified with cysteine-functional peptide	Greater accumulation in irradiated tumors and time-dependent increase in doxorubicin delivery to tumors in vivo. Tumor-specific doxorubicin delivery induced apoptosis and delayed tumor growth.	66
Drug delivery through BBB	Brain	TGNYKALHP HNG	Cysteine-modified peptide conjugated to MAL-PEG-PLA micelles to form 110–120 nm particles	Uptake by brain-derived endothelial cells in vitro. High brain-to-liver and brain-to-spleen for TGN-NPs in vivo. Dye delivery to brain ~2- to 4-fold greater than controls.	67
Breast carcinoma xenografts in nude mice, glioma murine model	Glucose-regulated protein GRP78	GIRLRG	Cysteine-modified peptide conjugated to allyl-functional poly(ester) NPs	Accumulation in irradiated tumors in vivo. Tumor-specific paclitaxel delivery resulted in 50% greater apoptosis and delayed tumor tripling time by 55 days.	68
Left carotid injury in rats	Human collagen IV	KLWVLPK	Cysteine-modified peptide conjugated to shell MAL-PEG-DSPE / core PLA-paclitaxel liposomes to form 60 nm particles.	Exhibited 10–12 day drug release and binding to injured aortas (percutaneous angioplasty) in vitro. Spatiotemporally control distribution to injured vessels in vivo.	69
Pre-assembly peptide conjugation
Neuroblastoma, lung, ovarian, and prostate tumor xenografts in nude mice	APN30	CGNGRGGVR SSSRTPSDKY C	Peptide conjugated to DSPE-PEG2000-MAL and combined with HSPC, CHOL, and DSPE-PEG2000 to form 90–115 nm liposomes	Dox-loaded NGR-liposomes exhibited high tumor uptake in vivo, tumor volumes were reduced, and mice achieved longer survival.	70–72
Glioblastoma xenograft in nude mice	VAV3 receptor73	SSQPFWS	MAL-PEG-PLA modified with SSQPFWSC peptides to form 23 nm micelles	High cellular uptake efficiency and tumor spheroid penetration in vitro. Tumor vasculature accumulation in vivo. Tumor-specific paclitaxel delivery inhibited tumor growth more than untargeted control.	74
Carotid artery ligation atherosclerosis murine model	Filamin-A in atheroprone regions of disturbed blood flow	GSPREYTSY MPH	Peptide conjugated to DPHE and combined with DPPC to form 64 nm liposomes	Five-fold greater accumulation in disturbed flow regions in vivo. Artery-specific BH4 delivery decreased superoxide concentration and reduced plaque burden.	75
Liver-colonizing Colon 26 NL-17 cells and Meth A sarcoma in murine model	Angiogenic vessels in dorsal air sac murine model	APRPG	Stearoyl-peptide combined with DSPC and CHOL to form 100 nm Liposomes	Binding to human islet cell and glioblastoma tumors in vitro. Time-dependent tumor accumulation in vivo. Tumor vessel-specific doxorubicin slowed tumor growth and increased life span.	76
Carotid artery Ligation atherosclerosis murine model	Non-muscle myosin heavy chain IIA on endothelial cells in atheroprone vasculature	CLIRRTSIC	Peptide conjugated to NHS-PEG-MAL, then MAL reacted with disulfide-Crosslinked branched PEI. Complexed with siRNA to form 200 nm polyplexes	Binding to endothelial cells exposed to oscillatory shear stress in vitro. Selective binding to and gene knockdown in regions of disturbed flow in vivo.	77
RIP-Tag2 transgenic pancreatic cancer murine model	Pancreatic islet tumor78, Potentially Frizzled-5 Receptor	CKAAKN	Peptide conjugated to 6-(maleimidyl)-Hexanoic acid(trisnor-squalenylidene)-hydrazide. SQ-CKAAKN mixed with SQ-gemcitabine to form 130–170 nm NPs	Binding to secreted frizzled related protein-4 and low complement activation in vitro. Tumor-specific gemcitabine delivery reduced tumor burden by 30% and vessel area by 60% compared to untargeted.	79
Gene delivery to the brain	Brain67	TGNYKALHP HNG	Peptide conjugated to MAL-PEG-pDMAEMA. Complexed with DNA to form 80 nm micelles	Uptake by brain capillary endothelial cells in vitro. Brain-specific DNA delivery qualitatively enhanced gene expression in the brain relative to untargeted.	80

Table 2 Abbreviations: APA, Aminopeptidase A. APN, Aminopeptidase N. BBB, Blood-brain barrier. BH4, Tetrahydrobiopterin. BS3, bis(sulfosuccinimidyl)suberate. CD, Cluster of differentiation. CHOL, Cholesterol. DOTAP, 1,2-dioleoyl-3-trimethylammonium-propane. DPHE, 1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine-N-hexanoylamine. DPPC, Dipalmitoylphosphatidylcholine. DPPG, 1,2-dipalmitoyl-sn-glycero-3-[phospho-rac-(1-glycerol)]. DSPC, 1,2-distearoyl-sn-glycero-3-phosphocholine. DSPE, 1,2-distearoylphosphatidylethanolamine. ePC, L-α-phosphatidylcholine. HSPC, Hydrogenated soybean phosphatidylcholine. MAL, Maleimide. MCF, Michigan Cancer Foundation. NHS, N-hydroxysuccinimide. pDMAEMA, poly(2-(dimethylamino) ethyl methacrylate). PE, 1,2-distearoyl-sn-glycero-3-phosphoethanolamine. PEG, poly(ethylene glycol). PEI, poly(ethyleneimine). PLA, poly(lactic acid). PSMA, poly(styrene-alt-maleic anhydride). PS, poly(styrene). SPDP, N-succinimidyl 3-(2-pyridyldithio)proprionate. SQ, squalene. VEGF, endothelial growth factor.

Peptide conjugates

Phage display has identified bioactive peptides with intrinsic targeting ability, but these peptides are uncommon. Frequently, various small molecules exhibit therapeutic potential, but many have poor pharmacokinetic profiles. These molecules can be combined with phage-derived peptides to form peptide-drug conjugates (Figure 2A) to realize their potential. Consideration must be given to the linkage between peptides and drugs to prevent binding interference or drug inactivation. Often, the affinities of synthetic peptides are orders of magnitude lower than phages, motivating the synthesis of multivalent peptide conjugates (Figure 2B) and multivalent peptide-polymer conjugates (Figure 2C). A summary of studies investigating bioactive targeting peptides and peptide-drug conjugates is given in Table 1.

Bioactive targeting peptides

Peptides alone can exhibit inherent therapeutic effects. A common characteristic of the bioactive peptides summarized in Table 1 is that the sequences contain two cysteine amino acids that form disulfide bonds and induce cyclicization. The cyclic structure is vital to function; for example, cyclic CTTHWGFTLC peptides specifically inhibited the enzymatic activity of both gelatinases, but linear peptides did not.²⁴ This is likely due to the structural constraints cyclicization places on a peptide, which facilitates proper amino acid orientation to bind to target epitopes.⁴⁹ Cyclic peptides also exhibit greater serum stability than their linear counterparts.⁵⁰ Of note, not all cyclic peptides are bioactive, as the control cyclic peptide GACVFSIAHECGA used by Arap et al. showed no tumor homing or activity.³⁰ An advantage of bioactive targeting peptides is the simplicity of synthesis, as solid-phase peptide synthesis and recombinant DNA technology are easily accessible methods. However, peptides are less than 10 kDa and freely filtered by the kidneys, leading to low plasma half-lives and limited tissue residence time.⁵¹

Monovalent peptide conjugates

Peptide conjugation to small molecules with poor pharmacokinetics enhances target tissue accumulation. A common peptide-drug conjugation route employs carbodiimide chemistry to functionalize the N- or C-termini of peptides with small molecules that contain amine or carboxylic acid functionalities, such as the chemotherapeutic doxorubicin (Dox)³⁰, or protein therapeutics with poor intrinsic target tissue distribution. In the case of protein and peptide therapeutics, it is also possible to synthesize the entire peptide-drug conjugate in one step.³¹ However, a limitation to phage display is that synthesized peptides often exhibit affinities orders of magnitude lower (i.e. dissociation constants (K_D) orders of magnitudes higher) when avidity effects are removed, and monovalent peptide conjugates may not provide efficient target tissue distribution.

Multivalent peptide conjugates

Multivalent peptides are more akin to phage displays and antibodies, both of which exhibit avidity effects that yield affinity in the nanomolar or picomolar range.^{16, 52} Avidity effects seem to be significant for multivalent peptide conjugates with four or more peptides copies.^{36, 53} The benefit of multivalent peptide conjugates relative to monovalent peptide conjugates is highlighted by the peptide HVWMQAPGGG, which was identified to have high affinity for collagen using phage display.⁵⁴ Analysis of binding kinetics by SPR spectroscopy identified rapid association and dissociation of the peptide, resulting in K_D of only 61 µM. When the peptide was incorporated into a fifth-generation dendrimer, affinity was dramatically increased 100-fold to 550 nM.

Although peptide-targeted dendrimers provide versatility in drug delivery by enabling hydrophobic drug encapsulation, covalent drug conjugation, and siRNA complexation,^{55, 56} a major limitation is the cationic charge of dendrimers that can cause cytotoxicity due to membrane disruption.⁵⁵ Surface engineering strategies have improved the cytocompatibility of dendrimers by masking cationic charges,⁵⁷ but upon injection, dendrimers tend to aggregate and accumulate in the lungs.⁵⁸ While this characteristic may be useful for drug delivery to the lungs, it is inefficient for delivery to other tissues and may cause toxicity. As a result, there are limited clinical applications of dendrimers as drug delivery vehicles.⁵⁹

Multivalent polymer conjugates

Underscored by multivalent peptide-drug conjugates, multi-peptide incorporation into a carrier enables much greater binding avidity versus singular peptide binding affinity. However, when peptides are hydrophobic, aggregation may obscure binding and poor solubility can hinder applicability to drug delivery.⁶⁰ Copolymer conjugates enable functionalization of a core hydrophilic polymer, such as commonly used poly(ethylene glycol) (PEG) or N-(2-hydroxypropyl)methacrylamide (HPMA), with targeting peptides. As an added advantage, copolymers can also incorporate multiple drug molecules to enhance drug payload (Figure 2C). Two methods are employed for peptide functionalization: peptide incorporation during polymerization and post-polymerization peptide incorporation. Based on monomer functionalities and reactivity ratios, gradient and uniform peptide incorporation can be achieved. A study comparing these two polymer architectures identified greater, but less stable, target binding of gradient copolymers with increasing peptide content, but greater binding stability with random copolymers above 15% percent peptide content.⁴⁸ In vivo, the fracture-targeted random copolymers persisted in bone for at least one week, whereas gradient copolymers were cleared, demonstrating that the two-step synthesis scheme is more promising for fracture targeting.

One benefit to post-polymerization functionalization is the ability to use polymer functionalities to incorporate both peptides and drugs. A prostate cancer cell-targeting HPMA copolymer was designed using N-methacryloyl-glycylglycine-4-nitrophenyl ester (MA-GG-ONp) as a comonomer.⁶¹ The ONp functionality was used to facilitate peptide incorporation, and unreacted ONp groups were reacted with hydrazine hydrate to facilitate doxorubicin incorporation. The use of the hydrazine bond between the polymer and DOX enabled pH-responsive release and resulted in significant anti-proliferative activity in prostate cancer cells relative to unconjugated free doxorubicin.⁶¹ However, a limitation to multivalent polymer conjugates as drug delivery systems is the need to functionalize drugs for incorporation if no functional groups, such as the amine in doxorubicin, exist.

Multivalent Nanoparticles

Some therapeutic compounds lack functionalities for conjugation and are difficult to deliver with polymer conjugates. Nanoparticles (NP) can be used to deliver these compounds: specifically, liposomes for both hydrophobic and hydrophilic molecules (Figure 2D) and micelles for hydrophobic molecules (Figure 2E). NPs are colloidal suspensions of amphiphilic polymers and are inherently multivalent. Careful polymer selection provides functionalities amenable to peptide incorporation, either pre- or post-particle assembly. Peptide conjugation after NP assembly may lead to a more even distribution of peptides across the particle surface but may limit accessibility of functional groups and cause poor functionalization efficiency. Peptide conjugation to polymers before NP assembly provides greater control over degree of functionalization but does not necessarily homogeneously display peptides, as aggregates may form “islands” of peptide on the particle surface. Examples of both strategies are summarized in Table 2.

Multivalent peptide-nanoparticle (NP) conjugates

The array of commercially available reagents, well documented synthesis schemes, and flexibility of drug loading has made peptide-targeted NPs a common drug delivery approach. The approach is particularly suited to cancer therapeutics because the ~100–200 nm sizes of liposomes and micelles enable extravasation and passive accumulation in leaky tumor vasculature via the enhanced permeability and retention (EPR) effect.⁸¹ It is also suited to conditions with an inflammatory component, such as fractures.⁶² However, there is room for improvement, as some systems with active peptide-targeting exhibit only two-fold increased target accumulation⁶² and achieve similar end point outcomes⁷⁶ as untargeted systems.

A method of improving NP targeting is to incorporate two different targeting peptides.^{64, 82} One peptide could be targeted to proteins within the extracellular microenvironment, and the other could be targeted to receptors on the surface of target cells. This strategy may mitigate off-target cell activation, as phage-derived peptides can bind to more than their intended target receptors.⁸³ Targeting by multivalent NPs can be further enhanced by multivalent peptide display on multivalent NPs (Figure 2D2). Monomeric and tetrameric peptides were incorporated into liposomal NPs to investigate the effects of both peptide valence and NP valence on binding.⁸⁴ Liposomes bearing similar numbers of peptides showed that tetrameric peptides enhance cell uptake relative to monomeric and that increased tetrameric peptide density further increased specificity, suggesting multivalence is beneficial.⁸⁴

Perspective

Phage display is a powerful technique to identify targeting peptides, but it comes with limitations. False positives can arise through selection of phages bound to plastic substrates or blocking agents, as well as selection of phages that amplified due to propagation advantages.⁸⁵ Even a small difference (~10%) in growth rates or infectivity can disrupt the diversity of a phage library.⁸⁶ For example, the commercially available New England Biolab Ph.D.−7 phage display library, generated using a degenerate 7-mer library, produces clones enhanced with proline and depleted with cysteine, suggesting a propagation advantage to clones containing proline residues.⁸⁷ The heptapeptide HAIYPRH underscores this issue, as it has been identified in at least 14 independent biopanning experiments and accounts for 0.26% of peptides that pass one round of selection and amplification.⁸⁷ Next generation sequencing (NGS), which has become more accessible with technological advancements, may address these concerns. NGS applied after each selection round may identify sequences with high target affinity that are not selected following amplification due to inferior infectivity or growth.^{87, 88}

An interesting future direction in phage display technology is selective modification by non-standard amino acids⁸⁹ such as selenocysteine (Sec), the 21^st amino acid. Although chemically related to cysteine (Cys), Sec has a sidechain pKa lower than that of Cys (~5 vs ~8) and is more nucleophilic at physiological pH (~7.4). This enables functionalization of the Sec thiol within the foreign phage sequence without affecting other coat proteins. Beech et al. employed this strategy to incorporate small molecule binders of one type of G protein-coupled receptor, the adenosine A₁ receptor, into phages.⁹⁰ The molecule, N⁶-Octylaminoadenosine, retained affinity to A₁ receptors despite conjugation to phage by its primary amine; moreover, avidity effects led to a 14-fold lower EC₅₀ in activating downstream Akt signaling. Similarly, Li et al. replaced the original disulfide bonds of a tumor-targeting peptide, Lyp-1, with diseleno bonds to increase serum stability and saw a reduction in the IC₅₀ of liposomal doxorubicin that led to 75% tumor growth inhibition.⁹¹ An application of this approach would be to perform a biopan of peptide(Sec)-displaying phages against a target cell type to identify cell-specific targeting peptides before modifying the selected phages with small molecule agonists intended to act on the target cells. Then, either a synthetic peptide(Sec)-drug conjugate or the modified coat protein⁹² could be used for cell-specific small molecule delivery.

Conclusions

Phage display has progressed from a tool for cloning genes¹¹ to a tool for developing targeted drug delivery systems. The versatility of panning against any molecule, protein, or cell enables the design of targeting peptides that are highly specific to target tissues. After identifying specific peptides, various chemistries are used to incorporate peptides into multivalent drug conjugates that exhibit high target affinity. Careful experimental design and post-hoc analyses are necessary to mitigate false positives that can arise during the processes of selection and amplification. Phage display technology is highly accessible with the diverse commercial availability of random peptide libraries, leading to successful applications in targeted drug delivery.