Biomimetic amplification of nanoparticle homing to tumors

CREKA Peptide.

We used a tumor-homing peptide to construct targeted nanoparticles. We identified this peptide by in vivo screening of phage-displayed peptide libraries (5, 11) for tumor homing in tumor-bearing MMTV-PyMT transgenic breast cancer mice (12). The most frequently represented peptide sequence in the selected phage preparation was CREKA (Cys-Arg-Glu-Lys-Ala). We synthesized the CREKA peptide with a fluorescent dye attached to the N terminus and studied the in vivo distribution of the peptide in tumor-bearing mice. Intravenously injected CREKA peptide was readily detectable in the PyMT tumors, and in MDA-MB-435 human breast cancer xenografts, minutes to hours after the injection. The peptide formed a distinct meshwork in the tumor stroma [supporting information (SI) Fig. 5A], and it also highlighted the blood vessels in the tumors. The CREKA peptide was essentially undetectable in normal tissues. In agreement with the microscopy results, whole-body imaging using CREKA peptide labeled with the fluorescent dye Alexa Fluor 647 revealed peptide accumulation in the breast cancer xenografts and in the bladder, reflecting elimination of excess peptide into the urine (SI Fig. 5B).

Tumors contain a meshwork of clotted plasma proteins in the tumor stroma and the walls of vessels, but no such meshwork is detectable in normal tissues (13–15). The mesh-like pattern produced by the CREKA peptide in tumors prompted us to study whether clotted plasma proteins might be the target of this peptide. We tested the peptide in fibrinogen knockout mice, which lack the fibrin meshwork in their tumors. Like previously identified clot-binding peptides (15), intravenously injected CREKA peptide failed to accumulate in B16F1 melanomas grown in the fibrinogen null mice but formed a brightly fluorescent meshwork in B16F1 tumors grown in normal littermates of the null mice (Fig. 1 A and B). In agreement with this result, the CREKA phage, but not the control insertless phage, bound to clotted plasma proteins in vitro (Fig. 1C). These results establish CREKA as a clot-binding peptide. Its structure makes it an attractive peptide to use in nanoparticle targeting because, unlike our other clot-binding peptides, which are cyclic 10 amino acid peptides (15), CREKA is linear and contains only 5 aa. Moreover, the sulfhydryl group of the single cysteine residue is not required to provide binding activity and can be used to couple the peptide to other moieties.

Peptide-Coated Nanoparticles.

We coupled fluorescein-labeled CREKA or fluorescein onto the surface of 50-nm superparamagnetic, amino dextran-coated iron oxide (SPIO) nanoparticles. Such particles have been extensively characterized with regard to their chemistry, pharmacokinetics, and toxicology and are used as MRI contrast agents (16–18). Coupling of the fluorescein-labeled peptides to SPIO produced strongly fluorescent particles. Releasing the peptide from the particles by hydrolysis increased the fluorescence further by a factor of ≈3 (data not shown). These results indicate that the proximity of the fluorescein molecules at the particle surface causes some quenching of the fluorescence. Despite this quenching, fluorescence from the coupled fluorescein peptide was almost linearly related to the number of peptide molecules on the particle (SI Fig. 6), allowing us to track the number of peptide moieties on the particle by measuring particle fluorescence and to use fluorescence intensity as a measure of the concentration of particles in samples. We used CREKA-SPIO with at least 8,000 peptide molecules per particle in our in vivo experiments. The CREKA-SPIO nanoparticles bound to mouse and human plasma clots in vitro, and the binding was inhibited by the free peptide (Fig. 1D). The nanoparticles distributed along a fibrillar meshwork in the clots (Fig. 1D Inset). These results show that the particle-bound peptide retains its binding activity toward clotted plasma proteins.

Tumor Homing vs. Liver Clearance of CREKA-SPIO.

Initial experiments showed that intravenously injected CREKA-SPIO nanoparticles did not accumulate effectively in MDA-MB-435 breast cancer xenografts. In contrast, a high concentration of particles was seen in reticuloendothelial system (RES) tissues (Fig. 2A Upper). Because the free CREKA peptide effectively homes to these tumors (SI Fig. 5), we hypothesized that the RES uptake was a major obstacle to the homing of the nanoparticles. We confirmed the role of the RES in the clearance of CREKA-SPIO by depleting RES macrophages in the liver with liposomal clodronate (19). This treatment caused an ≈5-fold prolongation in particle half-life (Fig. 2B). Particulate material is eliminated from the circulation, because certain plasma proteins bind to the particles and mediate their uptake by the RES (opsonization) (20, 21). Injecting decoy particles that eliminate plasma opsonins is another commonly used way of blocking RES uptake (22, 23). We tested liposomes coated with chelated Ni²⁺ as a potential decoy particle because we surmised that iron oxide and Ni²⁺ would attract similar plasma opsonins, and Ni-liposomes could therefore deplete them from the systemic circulation. Indeed, SDS/PAGE analysis showed that significantly less plasma protein bound to SPIO in the blood of mice that had been pretreated with Ni-liposomes (results not shown).

Intravenously injected Ni-liposomes prolonged the half-life of the SPIO and CREKA-SPIO in the blood by a factor of ≈5 (Fig. 2B). The Ni-liposome pretreatment, whether done 5 min or 48 h before the injection of CREKA-SPIO, greatly increased the tumor homing of the nanoparticles, which primarily localized in tumor blood vessels (Fig. 2 A Lower and D). The local concentration of particles could be so high that the brownish color of iron oxide was visible in the optical microscope (Fig. 2C Right), indicating that the fluorescent signal observed in tumor vessels was from intact CREKA-SPIO. Fewer particles were seen in the liver after the Ni-liposome pretreatment, but accumulation in the spleen was unchanged or even enhanced (Fig. 2A). Other organs contained minor amounts of CREKA-SPIO particles or no particles at all, whether Ni-liposomes were used or not (Fig. 2D). Although Ni-liposomes were quite effective, they caused some deaths in the tumor mice, prompting us to also test plain liposomes as decoy particles. These liposomes prolonged the blood half-life and tumor homing of subsequently injected CKEKA-SPIO (Fig. 2B and data not shown), suggesting the existence of a common clearance mechanism for liposomes and SPIO. However, although the plain liposomes showed no apparent toxicity, they were less effective as decoy particles than the Ni-liposomes. Reducing the Ni content of the liposomes may provide a suitable compromise between toxicity and efficacy.

Nanoparticle-Induced Blood Clotting in Tumor Vessels.

CREKA-SPIO particles administered after liposome pretreatment primarily colocalized with tumor blood vessels, with some particles appearing to have extravasated into the surrounding tissue (Fig. 3A Top). Significantly, up to 20% of tumor vessel lumens were filled with fluorescent masses. These structures stained for fibrin (Fig. 3A Middle), suggesting that they are blood clots impregnated with nanoparticles. In some of the blood vessels, the CREKA-SPIO nanoparticles were distributed along a meshwork (Fig. 3A Middle Inset), presumably formed of fibrin and associated proteins, and similar to the pattern shown in Fig. 1D Inset.

Among the non-RES tissues, the kidneys and lungs contained minor amounts of specific CREKA-SPIO fluorescence (Fig. 2D). However, low-magnification images, which reveal only blood vessels with clots in them, showed no clotting in these tissues, with the exception of very rare clots in the kidneys (SI Fig. 7). Despite massive accumulation of nanoparticles in the liver, we saw no colocalization between fibrin(ogen) staining and CREKA-SPIO fluorescence in liver vessels (SI Fig. 8). Moreover, liver tissue from a noninjected mouse also stained for fibrin(ogen) (SI Fig. 8B), presumably reflecting fibrinogen production by hepatocytes. Thus, the clotting induced by CREKA-SPIO nanoparticles is essentially confined to tumor vessels.

Nanoparticles can cause platelet activation and enhance thrombogenesis (24, 25). Some CREKA-SPIO nanoparticles (<1%) recovered from blood were associated with platelets (SI Fig. 9A), but staining for a platelet marker showed no colocalization between the platelets and CREKA-SPIO nanoparticles in tumor vessels (Fig. 3A Bottom). We also induced thrombocytopenia by injecting mice with an anti-CD41 monoclonal antibody (26) and found no noticeable effect on CREKA-SPIO homing to the MDA-MB-435 tumors (SI Fig. 9B). These results indicate that platelets are not involved in the homing pattern of CREKA-SPIO.

The deep infiltration of clots by nanoparticles suggested that these clots must have formed at the time particles were circulating in blood rather than before the injection. We tested this hypothesis with intravital confocal microscopy, using DiI-labeled erythrocytes as a flow marker. There was time-dependent clot formation and obstruction of blood flow in tumor blood vessels with parallel entrapment of CREKA-SPIO in the forming clots (Fig. 3B; and see SI Movie 1).

We next tested whether the clotting-inducing effect was specific for SPIO particles, or could be induced with a different CREKA-coated particle. We used liposomes into which we incorporated fluorescein-CREKA peptide that was coupled to lipid-tailed polyethylene glycol. Like CREKA-SPIO, the CREKA-liposomes selectively homed to tumors and colocalized with fibrin within tumor vessels (Fig. 3C), suggesting that CREKA liposomes are also capable of causing clotting in tumor vessels. No clotting was seen when control SPIO or control liposomes were injected in the tumor mice.

Clotting-Amplified Tumor Targeting.

We next studied the contribution of clotting to the accumulation of CREKA-SPIO in tumor vessels. Quantitative analysis of tumor magnetization with a superconducting quantum interference device (SQUID) (Fig. 4A), and measurement of the fluorescence signal (data not shown) revealed ≈6-fold greater accumulation of CREKA-SPIO in Ni-liposome-pretreated mice compared with PBS-pretreated mice. Aminated SPIO control particles did not significantly accumulate in the tumors (Fig. 4A).

The SQUID measurements revealed that injecting heparin, which is a strong clotting inhibitor, before injection of CREKA-SPIO reduced tumor accumulation of nanoparticles by >50% (Fig. 4A). Microscopy showed that heparin reduced the fibrin-positive/CREKA-SPIO-positive structures within tumor vessels but that the particles still bound along the walls of the vessels, presumably to preexisting fibrin deposits (a representative image is shown in Fig. 4B). Separate quantification of the homing pattern showed that heparin did not significantly reduce the number of vessels with nanoparticles bound to the vessel walls, but essentially eliminated the intravascular clotting (Fig. 4C). Thus, the binding of CREKA-SPIO to tumor vessels does not require the clotting activity that is associated with these particles, but clotting improves the efficiency of the tumor homing.

The clotting induced by CREKA-SPIO caused a particularly strong enhancement of tumor signal in whole-body scans. CREKA-SPIO nanoparticles labeled with Cy7, a near-infrared fluorescent compound, effectively accumulated in tumors (Fig. 4D, image on the left, arrow), with a significant signal from the liver as well (arrowhead). The reduction in the tumor signal obtained with heparin (Fig. 4D, image on the right) appeared greater in the fluorescence measurements than the 50% value determined by SQUID, possibly because the concentrated signal from the clots enhanced optical detection of the fluorescence. These results suggest that the clotting induced by CREKA-SPIO provides a particular advantage in tumor imaging.

Phage Screening, Tumors, and Peptides.

In vivo screening of a peptide library with the general structure of CX₇C, where C is cysteine and X is any amino acid, was carried out as described (5) by using 65- to 75-day-old transgenic MMTV PyMT mice (31). These mice express the polyoma virus middle T antigen (MT) under the transcriptional control of the mouse mammary tumor virus (MMTV), leading to the induction of multifocal mammary tumors in 100% of carriers. MDA-MB-435 tumors in nude mice and peptide synthesis have been described (32, 33). B16F1 murine melanoma tumors were grown in fibrinogen null mice (34) and their normal littermates and were used when they reached 0.5–1 cm in size (15).

Nanoparticles and Liposomes.

Amino group-functionalized dextran-coated superparamagnetic iron oxide nanoparticles (50-nm nanomag-d-SPIO; Micromod Partikeltechnologie, Rostock, Germany) were coupled with CREKA peptide by using a cross-linker. The final coupling ratio was 30 nmol of fluorescein-labeled peptide molecules per milligram of iron oxide, or 8,000 peptides per particle. For near-infrared labeling with Cy7, ≈20% of the amines were derivatized with Cy7-NHS ester (GE Healthcare Bio-Sciences, Piscataway, NJ), and the remaining amines were used for conjugating the peptide. Details on the SPIO and the preparation of liposomes are described in SI Materials and Methods. Clodronate was purchased from Sigma (St. Louis, MO) and incorporated into liposomes as described (19).

Nanoparticle Injections.

For i.v. injections, the animals were anesthetized with i.p. Avertin, and liposomes (2 μmol of DSPC) and/or nanoparticles (1–4 mg of Fe per kg of body weight) were injected into the tail vein. The animals were killed 5–24 h after injection by cardiac perfusion with PBS under anesthesia, and organs were dissected and analyzed for particle homing. To suppress liver macrophages, mice were intravenously injected with liposomal clodronate suspension (100 μl per mouse), and the mice were used for experiments 24 h later. All animal work was reviewed and approved by the Burnham Institute's Animal Research Committee.

Phage and Nanoparticle Binding to Clots.

Phage binding to clotted plasma proteins was determined as described (15). CREKA-SPIO and control SPIO were added to freshly formed plasma clots in the presence or absence of free CREKA peptide. After 10 min of incubation, the clots were washed four times in PBS, transferred to a new tube and digested in 100 μl of concentrated nitric acid. The digested material was diluted in 2 ml of distilled water, and the iron concentration was determined by using inductively coupled plasma–optical emission spectroscopy (ICP-OES; PerkinElmer, Norwalk, CT).

Plasma Protein Binding to Nanoparticles, Particle Clearance from the Blood, Intravital Microscopy and Magnetic Measurements.

For information on plasma protein binding and the other methods, see SI Materials and Methods.