We selected three peptides to show that it is possible to identify drug-targeting candidates for CNS drug delivery from a phage peptide library. Our biopanning strategy for targeting generated over two dozen new sequences that we predicted would be biased for luminal targeting of CP epithelial cells because the screen used CP explants freshly dissected from donor mice. The three peptides were selected because they were found to have either increased representation in the limited number of clones sequenced (e.g. R3.9) or alternatively whose sequences appeared after 3rd and 4th rounds of biopanning. These peptides, along with EGF , are proven capable of targeting CP epithelial cells in vitro, ex vivo and in vivo. Whether they can target cells for transduction after i.c.v. injection like EGF  remains to be established.
In 1985, Smith  conceived of phage display as a method whereby short nucleic acid sequences of DNA could be inserted into the coding sequence of the M13 phage glll gene to generate particles that display on their capsid surface a peptide-plll fusion protein. Reasoning that the displayed peptides would confer phage with new intrinsic activities, he proposed that it should be possible to introduce random and known sequences of DNA into the glll gene to create particles with new activities. Over the last twenty years, we and other investigators have been adapting this original phage display technique to identify novel peptides with different specificities and activities (reviewed in [27, 28]), for example, peptides have that confer physical stability to particles in organic solvents like chloroform, decrease complement activation of macromolecules in blood, modify immunogenicity, alter viral tropism in vitro and in vivo, internalize particles and transduce cells, promote transcytosis in vitro and in vivo and even promote transmigration of particles across cell barriers in vitro and in vivo [29–53].
Our laboratories' focus has been to identify and exploit ligands that internalize into target cells with a long-term objective of improving the specificity of drug, protein and gene-based medicine delivery. To this end, we have re-engineered phage vectors for increased binding to mammalian cells and monitored their entry into cells by immunohistochemistry for internalization drug delivery, transfection for nucleotide delivery and transduction for gene expression [15–19, 25, 54–58]. It seemed reasonable to propose that these methods might deployed towards CNS drug delivery because ligands like EGF can target the phage particle to cells .
Several years ago, investigators  proposed that it might be possible to target therapeutics to the CNS by exploiting one of the unique features of the choroid plexus: it exists at the interface between blood and CSF. If possible, drugs could then be designed to act like ascorbic acid and translocate across the CP epithelium for entry into CSF and access to brain parenchyma [59–61]. Alternatively, drugs could be targeted to the CP epithelium so as to modify its hydrodynamic and homeostatic functions in controlling CSF production and composition respectively. We show here that when phage are re-engineered so that they can target the choroid plexus epithelium, they internalize into epithelial cells when tested in vitro, ex vivo and in vivo. The targeting is ligand specific, concentration and time dependant.
The possibility of exploiting epithelial routes of drug delivery to the CNS is not new. For example, we originally proved the concept of CP targeting with the ligand EGF  but it was clear from the outset that this ligand is not ideally suited for CP targeting. First, EGF has intrinsic activity and is a naturally occurring growth factor. Second, while EGF may be selective for epithelial cells in the brain as described here, but there are several alternative cell types that express EGF receptors including astrocytes, activated glia and endothelial cells. The specificity that we observe with i.c.v. injections is likely due to the compartmentalization of phage in the ventricular space and the limited number of cell types available for targeting. Herenu et  also explored the ependymal route for insulin like growth factor-1 (IGF-1) gene delivery in an attempt to circumvent the need to transport IGF-1 from blood into brain by injecting peptide i.c.v. [63, 64]. In this case, they exploited adenoviral selectivity to infect epithelial cells when injected i.c.v. to increase the levels of IGF1 in CSF and mimic the concentrations achieved by the injection of protein into CSF. Presumably, adenoviral-retargeting strategies that modify its pharmacokinetics, biological activity and even potency could be used to improve its use as a gene delivery agent to the CNS.
In the study described here, we have not studied the mechanism of internalization or the specific trafficking of the particles once they enter the cell. The fact that the peptide ligand can prevent internalization suggests specificity for cell binding. The observation that the particles are internalized in a ligand dependant fashion supports the hypothesis that phage bind to cells via cognate receptors to the peptides displayed. The identity of these receptors remains unknown but clearly phage display should be considered as a viable method to identify novel portals of entry into the CNS via delivery to CPe cells.
Using the CP as a doorway into the CNS has also gained momentum with significant difficulties in overcoming the blood brain barrier [1, 65, 66]. Just as there are reports that propose using brain's endothelium as a target for CNS drug delivery using molecular signatures[44, 46, 48] in brain endothelia, Chen et al  showed that CNS-directed enzyme therapy was feasible by targeting epitope-modified adeno-associated virus. While the effects are certainly local, the wide distribution of endothelium in the CNS would ensure wide distribution of drug throughout the brain, if all endothelia were targeted. For similar reasons, there is considerable interest in targeting the choroid epithelium [5, 59]. First, the design and structure of barriers are not homogenous throughout the brain and areas like the arcuate nucleus are open to CSF while others, like median eminence, are open to the portal blood . Second, the choroid plexus is actively modulating CSF content and composition in response to the environment  which regulate brain state and neuroactive peptide distribution in CSF. Third, the choroid plexus can play a prominent role in neuroprotection  because of special axonal endings that are formed at CSF contacting neurons [70–72] which presumably enable communication between brain parenchymal cells and CSF. Finally, ependymal cells, which divide asymmetrically and transfer progeny into the subventricular zone activated by injury  are in direct contact with CSF and CP produced neuroactive agents in CSF. In as much as this subependymal zone is a neurogenic niche  that is influenced by CP epithelial cells and an ependymal layer controlling of CSF hydrodynamics and content, these cells are nearly ideally localized to modulate CNS functions and treat brain disease if they can be targeted to produce therapeutics. As demonstrated by Regev et al, lentiviral transduction of choroid plexus epithelium with genes encoding neuroactive peptides can control CNS function thereby supporting the feasibility of CP-mediated drug delivery.