Generally, polymers that are used for preparation of nanoparticles fall into two major categories: natural polymers and synthetic polymers. A number of natural polymers such as heparin, dextran, albumin, gelatine, alginate, collagen, and chitosan have been intensively investigated. Synthetic polymers including polyethylene glycol (PEG), polyglutamic acid (PGA), polylactic acid (PLA), polycarprolactone (PCL) and N-(2-hydroxypropyl)-methacrylamide copolymer (HPMA) have been exploited as well. General requirements for those polymers are biocompatibility, biodegradability, and their capacity to be functionalized (2).

The formation of polymeric nanoparticles has been summarized by several review articles (3,4). In most cases, the polymeric nanoparticle consists of two parts, a hydrophobic core which serves as the container for anticancer agents and a hydrophilic shell which stabilizes the nanoparticle in aqueous environments.

The drug can be loaded into polymeric nanoparticles through two methods: by physical entrapment or by chemical conjugation. A hydrophobic interaction between the core of the polymeric nanoparticle and the drug molecule allow the drug to be entrapped in the nanoparticle core. For instance, deoxychilic acid-modified heparin can self-assemble into 100~200 nm nanoparticles (5) and its hydrophobic core can be used to entrap 4~12% of the total weight of doxorubicin (6) When the drug molecule is covalently conjugated onto the polymer, the chemical properties of the linker between the drug and polymer are critical. If the linker is too stable, drug release may be delayed, while if the linker is too unstable, drug may be released before the nanoparticle reaches the tumor. Therefore, a proper linker is very important to the drug-polymer conjugate. A variety of pH-sensitive linkers have been developed such as hydrozone and cis-aconityl (7,8). These chemical bonds are stable in the blood circulation system (pH=7), but quickly decompose and release drug molecules inside the tumor where pH values typically drop below 5.5. Disulfide bonds are also very attractive because they can be cleaved by glutathione. The intracellular level of glutathione is much higher than its extracellular level, therefore, the disulfide linker is relatively stable while in blood circulation and becomes unstable and releases the drug molecules once it is internalized by cells (9,10).

It is important to note that dendrimers, synthetic superma cromolecules with highly branched repeated three dimensional structures, have emerged as important materials for biological application due to their unique features such as the precise control of size and shape, uncommon physical properties, controlled degradation, and the ability to place numerous functional groups on their periphery and/or core (11,12). There are more than 50 different types of dendrimers (13). Among them, polyamidoamine and poly (propylenemine) have been commercialized and used extensively as biomaterials in gene and drug delivery (14,15) and for nanoparticle encapsulation (16,17) in imaging (18,19).

Liposomes are self-assembling spherical particles with a membrane composed of phospholipid bilayers. The size of liposomes can range from 25 nm to 10 µm depending on the preparation method. They have been studied as candidates for drug delivery for the last 50 years since being first discovered by Bangham (20). The synthesis of liposomal nanoparticles has been reviewed by Bellare, et al (21). Drug delivery systems based on unmodified liposomes are limited by their short blood circulation time. This is mainly due to the fast clearance of liposomes by macrophages of the reticuloendothelial system (RES) (22). The second generation of polymer-coated liposomes can dramatically increase blood circulation times from several minutes up to 3 days.

Recently, several novel nanotechnology concepts have been applied to the development of a new generation of anti-cancer drug delivery systems. Gold nanoparticles can be synthesized through the reduction of HAlCl₄ with a very narrow polydispersity. Several gold nanoparticle anticancer drug delivery systems have been reported and showed good in vitro results (23).Although it seems that gold is inert under physiological environments, the long term toxicity of gold nanoparticles remains an unanswered question. One attractive property of gold nanoparticles is that gold concentrations are naturally low in animal bodies, allowing the convenient use of this nanoparticle model for in vivo pharmacokinetic and biodistribution studies (24).

It is worth mentioning that a gold nanorod formulation is showing very promising potential as a photothermal therapy agent. Gold nanorod can generate heat when it is radiated by a near infra-red (IR) laser (wavelengt > 650 nm). At this range, the laser is relatively safe to the tissue and organs. Once the gold nanorod has accumulated inside the tumor through passive/active targeting, it can be heated locally up to 43℃ (25) by radiation with a near IR laser to destroy the tumor without causing damage to surrounding healthy tissues (27).

Iron oxide nanoparticles have been clinically used as imaging agents for MRI. Recently a number of research groups have investigated them as drug carriers while retaining their imaging functions (26-28). One unique advantage of iron oxide nanoparticle delivery systems is that they can be delivered in a targeted manner to a desired region by applying an external magnetic field.

Table 1 lists nanoparticles that have been used in the clinic and utilize passive targeting to achieve their selective delivery to tumors. Passive targeting takes advantage of the inherent size of nanoparticles and the unique properties of tumor vasculature (30-33). As tumors grow and begin to outstrip the available supply of oxygen and nutrients, they release cytokines and other signaling molecules that recruit new blood vessels to the tumor in a process called angiogenesis. Unlike the tight blood vessels in normal tissues, angiogenic blood vessels in tumor tissues have gaps as large as 600 to 800 nm between adjacent endothelial cells (34,35). This defective vascular architecture coupled with poor lymphatic drainage induces an enhanced permeability and retention effect (EPR) (36,37), Through these gaps, nanoparticles can selectively accumulate into the tumor interstitium (38) (Fig. 2).

In general, the accumulation of nanoparticles in tumor tissues is dependent on several factors including the size, surface characteristics, and circulation half-life of the nanoparticles and the degree of angiogenesis of the tumor (39). It is speculated that nanoparticles with a size between 10 and 100 nm will be optimal for tumor accumulation. For example, smaller polymeric micelles (20 nm) have been shown to accumulate more readily in tumors than larger liposomes (100 nm) (40,41). Proper surface characteristics and longer circulation times of nanoparticles can also improve tumor uptake, as described earlier. The unmodified phospholipid surface of liposomes can attract plasma proteins and thus recognition by the mononuclear phagocytic system (MPS), resulting in their rapid clearance from the circulation. This property impedes the distribution of liposome-associated drugs to solid tumors. Surface-modified (stealth) liposomes have solved the problem of fast clearance from the circulation, yielding liposomes with a significantly increased half-life in the blood (42,43). Dramatically reduced clearance rates have also been obtained with other nanoparticles such as Abraxane (44), Xyotax (45) and IT-101 (46). Tumor vascularization also affects nanoparticle accumulation; usually nanoparticles accumulate poorly in poorly vascularized tumors, small pre-angiogenic tumors, or large necrotic tumors.

As drug delivery systems, nanoparticles have shown an ability to improve pharmacokinetics, pharmacodynamics, efficacy, and to reduce the toxicity of associated drugs (40). For example, Abraxane (ABI-007), an albumin-bound nanoparticle of paclitaxel (Taxol) which has been approved for the treatment of metastatic breast cancer, showed significant greater efficacy than free paclitaxel in a phase III clinic trial (47). Despite the increased dose of paclitaxel in the Abraxane group, the incidence of grade 4 neutropaenia was significantly lower than in patients treated with free paclitaxel. Pharmacokinetic studies also showed that paclitaxel clearance and the volume of distribution were higher for Abraxane than for paclitaxel: Clearance was 13 litres per hour per m² for Abraxane versus 14.76 litres per hour per m² for paclitaxel (p=0.048), and distribution was 663.8 litres per m² for Abraxane versus 433.4 litres per m² for paclitaxel (p=0.04) (44). Similar to Abraxane, NK105, a micellar nanoparticle formulation of paclitaxel also showed improved pharmacokinetics, pharmacodynamics, efficacy, and reduced toxicity as compared with free paclitaxel in a preclinical study and a phase I trial (48,49). The plasma area under the curve (AUC) value was approximately 90-fold greater for NK105 than for free paclitaxel and the tumor AUC value was 25-fold higher for NK105 than for free paclitaxel in an animal model (48). In patients, the plasma AUC of NK105 at 180 mg/m² was approximately 30-fold greater than that of the conventional formulation of paclitaxel (49). NK105 showed significantly more potent antitumor activity in a human colorectal cancer cell line HT-29 xenograft than free paclitaxel, due to enhanced accumulation of the drug in the tumor (48). The phase I trial showed that NK105 was well tolerated and effective in patients with pancreatic cancer (50). These differences in pharmacokinetic properties may contribute to the increased drug accumulation inside the tumor observed with nanoparticles compared with the corresponding free drugs. Other nanoparticles currently used in the clinic or undergoing clinic trials also showed an improved pharmacokinetic profile compared with the respective free drugs, such as Doxil, a PEG-liposome loaded with doxorubicin (DOX) (51), SP1049C, a pluronic micelle loaded with DOX (40), NK911, a PEG-Asp micelle loaded with DOX (52), and Xyotax, a polyglutamic acid nanoparticle carrying paclitaxel (45).

The nanoparticles listed in Table 1 that have been used in the clinic so far mostly utilize the EPR effect of tumors and the tumor microenvironment to promote their selective delivery to tumors. However, certain limitations of non-targeted nanoparticles as a drug delivery system still remain. For example, in the case of the EPR effect, although poor lymphatic drainage helps the extravasated drugs to be enriched in the tumor interstitium, it also induces drug outflow from the cells as a result of higher osmotic pressure in the interstitium, which eventually leads to drug redistribution in some portions of the cancer tissue (53). Most importantly, accumulation merely within the tumor microenvironment by the EPR effect may not always correlate with therapeutic efficacy since internalization into the tumor cells is required for most anticancer drugs to exert their biological functions. To overcome these limitations, a rational approach is to incorporate a targeting moiety on the nanoparticle surface. The targeting moiety is expected to bind a tumor-associated antigen or receptor and facilitate the delivery of nanoparticles to the intracellular site of drug action, enabling a greater therapeutic effect (Fig. 3). Recent preclinical studies have shown that targeted nanoparticles have better antitumor activity compared with non-targeted nanoparticles (54-57). Although targeted nanoparticles may not always mediate an increase in tumor drug accumulation when compared with non-targeted nanoparticles, targeted nanoparticles show greater intracellular drug delivery to cancer cells than non-targeted nanoparticles, resulting in dramatically increased antitumor efficacy (54-56). These findings suggest that the primary role of the targeting ligands is to enhance cellular uptake into cancer cells and to minimize cellular uptake in normal cells.

Although current studies have shown that the use of targeted nanoparticles as a drug delivery system is a promising strategy to treat human cancers, it is still in its early stage of development. Clinical data using targeted nanoparticles are limited since most targeted nanoparticles have not yet reached the clinic. Only a few targeted nanoparticles are currently under clinical investigation. One is MCC-465, which is an immunoliposome-encapsulated doxorubicin (Dox). The liposome is tagged with PEG and the F(ab')₂ fragment of the human monoclonal antibody GAH, which recognizes a cell surface molecule on various types of cancer cells (58).Phase I studies have indicated that the PK parameters of MCC-465 differ from those of free Dox, but were very similar to those of Doxil (non-targeted liposome-encapsulated Dox) in humans. In terms of skin toxicity, the patients who received MCC-465 did not experience any severe skin toxicity such as palmar-plantar erythrodysesthesia (PPE) or mucositis, unlike the patients who received Doxil (58). Besides MCC-465, other examples of targeted therapeutic nanoparticles include MBP-426 which contains the cytotoxic platinum-based drug oxaliplatin in a liposome (59), SGT-53, a liposome containing a plasmid coding for the tumor suppressor p53 (60), and CALAA-01, a polymer-siRNA conjugate (61,62) (Table 2) . These nanoparticles all target the transferrin receptor which is upregulated in many types of cancer (63).

Selection of the target receptor or antigen on cancer cells is crucial for the optimal design of targeted nanoparticles. In general, cell-surface antigens and receptors should have several properties that render them particularly suitable as tumor-specific targets. First, they should be abundantly and uniquely expressed on tumor cells, but negligibly or less expressed on normal cells. Second, they should have a high density on tumor cells.

A targeting ligand should selectively and successfully transport nanoparticles into targeted cancer cells. It is believed that internalization of nanoparticles after binding to targeted tumor cells is necessary for good therapeutic responses, so whether the targeted nanoparticles can be internalized is an important issue in the selection of proper targeting ligand. Use of a ligand that can not trigger the internalization process may result in drug release outside the cell and its redistribution to the surrounding normal tissues.

A variety of targeting ligands, including antibodies, antibody fragments, peptides, growth factors, and aptamers (64), have been used to facilitate the uptake of carriers into target cells (65-76).