2.1 Vascular or Capillary BBB

The vascular BBB is comprised of the capillary bed and the immediately adjacent arterioles and venules (2). This complex possesses the barrier and other functions described below in Section 4 and most of those in Section 5. Because no brain cell is more than 30–50 nm from a capillary, the vascular barrier is often considered as the exclusive route for drug delivery to the brain. This neglect of the other barriers, however, ignores their unique properties and complex ways in which the various barriers cooperate and interact. Although the cell wall formed by brain endothelial cells (BECs) seems to be the major physical and diffusional barrier between the blood and the CNS, the BECs are in close proximity to other cells of the CNS, forming a complex referred to as the neurovascular unit (3). The two cell types most studied in relation to supporting the BBB are the pericyte and the astrocyte. The pericytes are in intimate contact with the BECs, are connected by peg and socket contacts, and share cytoplasm via gap junctions (4). It is the pericyte that in utero first instructs the BECs to form tight junctions (5, 6). The ratio of pericytes to BECs are 1 to 5–6 in the vascular BBB, a large number of pericytes in comparison to other tissues. The astrocytes also instruct the BECs in the formation of a barrier and form a nearly complete cuff around the capillary (7). Despite this encapsulation, it is usually thought that the astrocytes do not create a second diffusional barrier for small molecules between the capillaries and the other cells in the brain.

2.2 Choroid Plexus: The Blood–Cerebrospinal Fluid Barrier (BCSFB)

Ependymal cells of epithelial origin form a sac that surrounds a vascular plexus that resides in the ventricles of the brain (8). The BECs forming this plexus do not form a barrier but produce an ultrafiltrate that fills the sac formed by the ependymal cells. Tight junctions between the ependymal cells form the barrier function of the BCSFB. The ependymal cells produce about 70% of the CSF with the formation of a transudate. The formation of this transudate is highly regulated and not simply a result of “leakage.” The BCSFB is often thought to be more leaky than the BBB, but this seems to have arisen because of three misinterpretations of experimental results. First, the electrical resistance of the BCSFB was originally reported to be much lower than that of the vascular BBB, but subsequent recalculations that took into account the complete surface area of the ependymal cells with all of its villi found this value to be much closer to that of the vascular BBB. Second, radiographic and dye studies show the choroid plexus to fill with vascular products, but this is to be expected since the barrier function resides at the level of the ependymal cell, not at the level of the BEC. Third, the intense vesicular activity of the choroid plexus has been misinterpreted as residual micropinocytosis when in reality this is part of the process forming the transudate.

Substances entering the ventricles by way of the BCSFB quickly distribute throughout the cranial cerebrospinal fluid (CSF) compartments but do not diffuse into spinal cord CSF very well (9). CSF to brain tissue diffusion is also limited because of the generally poor diffusion of substances within the brain tissue (10). Indeed, diffusion from the ventricular surface into brain tissue is logarithmic, with a 50% decrease in concentration for every 100–300 nm of distance from the CSF (11). The exact diffusion “half-distance” for a substance depends on several characteristics, including its molecular weight and how rapidly it crosses the capillary bed in the brain-to-blood direction (11, 12). As such, substances entering or introduced into the CSF tend to have a periventricular distribution. However, in many species, cell bodies and projections near the ependymal layer of cells lining the ventricles may relay information to deeper areas of the brain (13, 14). Thus, substances entering the CSF may have influences deep within the brain.

2.3 Tanycytic Barrier and Circumventricular Organs (CVOs)

The circumventricular organs (CVOs) are bona fide regions of brain in which the capillary beds do not always possess barrier function (15, 16). The seven regions typically found in mammals are the median eminence, subfornical organ, area postrema, neurohypophysis, pineal, subcommisural organ, and the organum vasculosum of the lamina terminalis. These areas are small, in total forming about 5% of brain weight in the mouse, yet are complex with even the smallest, the area postrema, having defined anatomical regions with and without barrier function. In areas without vascular barrier function, permeabilities can be as high as those found in peripheral tissues and blood-borne products have direct access to their brain cells. They also receive projections from distant brain areas and project to distant brain areas as well. Hence, they form a unique way in which blood-borne products can affect brain functions both in the CVOs and distally. The blood-borne products entering CVOs are not, however, able to diffuse to adjacent brain regions or to distant brain regions because of two factors. First, diffusion within brain tissue is very limited, as discussed above. Secondly, in the adult mammal, the CVOs are delimited by a tanycytic barrier (17-19). The tanycytic barrier is dynamic and is altered with metabolic and reproductive events (19-21).

2.4 Special Barriers

2.5 The Neurovascular Unit

The BECs interact with the various cells of the brain, forming the neurovascular unit (28). Besides pericytes and astrocytes, BECs also interact with neurons, microglia, and mast cells. One area in particular that has emphasized this interaction is that of neuroimmunology (29). Not only is barrier integrity affected, but BEC transporter and secretory functions.

3.1 The Cell Wall

The cell wall of the vascular BBB is formed by the capillaries and consists of the luminal cell membrane, the cytoplasm, and the abluminal cell membrane. The entire width of this wall as measured by electron microscopy is in the range of 100–150 nm; that is, not much more than the diameter of a pinocytotic vesicle.

3.2 Enzymatic Functions

The physical barrier is not the only means by which barrier cells prevent the exchange of substances between the blood and the brain. The BECs also possess enzymatic activity that can degrade molecules attempting to cross the BEC. A classic example is that of the monoamines which are digested by monoamine oxidases in the BECs, thus preventing, for example, dopamine from crossing the BBB (34). Thus, Parkinson's disease can be treated by administering l-dopa, which crosses the BBB using the large neutral amino acid (LNAA) transporter but not by administering dopamine, which is degraded by the BBB (35).

3.3 Brain-to-Blood Efflux

Transport systems can be directed in the brain-to-blood direction and in such cases will hinder or even prevent molecules from accumulating in the brain that would otherwise cross the BBB. Efflux systems exist for a wide range of substances, including peptides (36). The best studied of the efflux systems is P-glycoprotein (P-gp; ABCB1; MDR1; CD243), which transports a large number of structurally diverse compounds (37). Being a P-gp substrate often determines whether a drug will have CNS effects and hence determines its therapeutic profile. A classic example of this is loperamide, which is an opiate and unable to accumulate in brain because of P-gp. As a result, loperamide acts peripherally to induce opiate-related constipation and so is used to treat diarrhea, yet does not induce the CNS effects of morphine, morphine being only a weak substrate for P-gp. In animals without P-gp, loperamide has CNS activities and is as potent as morphine in the induction of analgesia (38).

3.4 Basement Membrane

The basement membrane or extracellular matrix is composed of laminins, collagen, fibronectin, and chondroitin sulfate-rich proteoglycans with hyaluronic acids being the major components. The width of the basement membrane is about 50–80 nm and does not act as a diffusional barrier to most small molecules or biologics. It does, however, hinder the passage of immune cells into the brain. Matrix metalloproteinases secreted by immune cells can digest the basement membrane and so allow these cells to more fully penetrate into the CNS (39, 40). The basement membrane is increasingly viewed as more than a structural feature. Hyaluronic acids, for example, possess immunomodulatory properties with larger molecules having immune-inhibitory and smaller molecules having immune-stimulatory properties (41). Basement membrane properties have been shown to be altered in aging and Alzheimer's disease (42).

4.1 Transcellular Diffusion

Many of the factors that dictate transcellular diffusion are those that determine transmembrane diffusion (43). However, there are some important caveats in considering how substances pass through an entire cell rather than just through a cell membrane.

Small, lipid-soluble molecules are able to readily penetrate cell membranes, including those that form the barrier tissues (44). The octanol/water partition coefficient is often as a measure of lipid solubility. The degree of penetration across the vascular BBB from blood into brain increases as a function of this partition coefficient up to about a ratio of 100, above which penetration decreases. Thoughts for this inverted U-shaped curve are that more lipid-soluble compounds are more tightly bound to plasma proteins and that to penetrate from blood into brain, the solute must not only partition into the lipid environment of the barrier cell's membrane but also from the cell membrane to the aqueous environment of the brain's fluids (45). Thus, substances that are extremely lipid soluble become sequestered in the cell membrane, unable to effectively repartition into brain interstitial fluid.

Other lipid/aqueous coefficients have been used to measure lipid solubility. For example, hexadecane/water is classic for modeling red blood cell permeability. However, these various combinations do not measure exactly the same molecular characteristics and can give variable answers (46); this also means that various characteristics can be deduced from these variables. For example, the heptane/water coefficient when subtracted from the octanol/water coefficient yields an estimate of the H-bond donor acidity (47).

A number of factors modify transcellular diffusion rates (48). One is size so that larger molecules partition less well into cell membranes. It is often stated that there is an absolute cut off of 400–600 Da for crossing the vascular BBB by the mechanism of transcellular diffusion, but this is not true. Instead, there tends to be a molecular weight penalty traditionally represented as an inverse of the square root of the molecular weight. The use of the square root likely arose because it was much easier to compute than the cube root but still a fair estimate of the volumetric radius.

Charged molecules penetrate poorly into lipid membranes and so permeation across the BBB favors molecules that do not ionize at blood pH (7.4). However, the pH of CSF is slightly more acidic, at about 7.3, than plasma (49). As a result, weak bases tend to accumulate in the CSF in comparison to weak acids as they become ionized once in the CSF and so unable to repartition back into the blood.

As a rule, only the portion of a substance not bound to plasma proteins is available to cross the barrier cells by transcellular diffusion (48). Binding can range from low affinity or nonselective as is common with albumin to high affinity, highly selective as exemplified by thyroxine binding to transthyretin. Even for substances crossing the BBB by the transcellular mechanism, the nature of the plasma binding can make a difference for brain sequestration. For example, a substance with a high affinity binding site in the CNS with adequate half-life in blood and reasonable BBB penetration will at equilibrium have levels that are higher in the brain regions with the binding site (50).

Lipinski's rule of five and the work of Wager embodies and expands many of these concepts (51, 52). Unfortunately, these useful rules of thumb have despite the balanced discussions of the authors been overinterpreted and assumed to apply to classes of compounds not in the original data sets or that were originally noted as exceptions to the rules (53). For example, these rules do not apply to small-molecule anti-parasitics nor is it clear how these rules apply to biologics, including small peptides.

The extent to which the above modifying factors apply to biologics is unclear. Peptides, for example, are able to cross the vascular BBB by transcellular diffusion although they are with few exceptions much larger than 400–600 Da (54). Octanol/aqueous measures are usually much less than 1, yet the octanol/aqueous partition coefficient is still a good predictor of the degree to which peptides cross by transcellular diffusion. Plasma protein binding can be substantial for some peptides and this can also influence uptake into the CNS (55, 56). Molecular weight likely also has an influence, although unique possibilities of folding that peptides can have may alter the inverse of the square root rule and maximize hydrogen bonding (57).

4.2 Competitive Transport Mechanisms: Carrier-Mediated, Facilitated Diffusion, Receptor-Mediated Transport, Receptor-Mediated Transcytosis

Terms related to the types of transport across the BBB are largely borrowed from cell biology (43, 58). They are rather loosely applied in part because whereas in cell biology they are used to describe movements across a cell membrane between the cytoplasm and the extracellular fluid, for BBB they relate to transport across the entire cellular barrier. The terms are often misapplied or loosely used; for example, “active” transport is colloquially applied to refer to any type of saturable transport rather than restricted to saturable transport that requires energy. Saturable or mediated transport refers to transport that is mediated by a membrane transport protein. Such transport can require energy (active transport) or not require energy (facilitated diffusion). Another common confusion is the use of “receptor” in transporter terminology. In cellular biology, the term receptor is classically used to refer to a protein with a membrane-spanning domain with a signal sequence. However, in the BBB literature, that term is often used carelessly to refer to any mediated transport. In part, this is due to the widely held assumption that the protein that acts as a substance's receptor will also act as its transporter. For example, insulin binds to and activates cell signaling on BECs by binding to its classic receptor (59-61). Insulin is also transported across the BBB in saturable fashion, but that transporter protein is not the same protein as the receptor (62). The term “transcytosis” refers to transport that involves vesicles but is often applied to the mediated transport of any biologic. Justification for such usage arises from the assumption that any larger molecule will require a vesicle for transport. However, interleukin-2, interleukin-4, and interferon-gamma, ranging in size from 13.5 to 17 kDa, are transported across the T-lymphocyte membrane by a nonvesicular process (63). Because of this loose use of terms in the BBB field, it is easy to assume that the nature of the mediated transport of a substance has been defined, when often it has not.

4.3 Adsorptive Transcytosis

The cellular pathways for adsorptive transcytosis were described by Broadwell and colleagues in the mid-1980s to early 1990s (64), using electron microscopy to map the course of the glycoprotein wheatgerm agglutinin coupled to horseradish peroxidase (WGA-HRP). At that time, adsorptive endocytosis was largely viewed as a pathological event in response to a foreign glycoprotein or lectin binding to BEC glycoproteins. Classically, this glycoprotein–glycoprotein interaction induced vesicles that were routed to the endosome and subsequently back to the luminal membrane and so was viewed as a way for the BEC to cleanse itself. With higher doses of WGA-HRP and by mechanisms still poorly understood, routing to the Golgi apparatus and to the abluminal membrane can occur (65, 66).

Highly charged proteins such as poly-l-lysine, protamine, and cationized albumin induce a similar response (67) and could be used to deliver drugs (68). A poly-l-lysine protein has been designed to transport a peptidase across the BBB (69). Although successful in delivering drug to the brain, the carrier protein was also toxic (70). It is likely that many of the Trojan horse approaches, especially those involving antibodies directed to off-target binding sites on BBB receptors and transporters, are inducing adsorptive transcytosis-like mechanisms rather than receptor-mediated transcytosis. This is an important mechanistic distinction given that adsorptive transcytosis cargos are typically routed through the lysosomal compartment.

Some viruses such as rabies and human immunodeficiency virus-1 (HIV-1) use a process similar to adsorptive transcytosis to enter brain (71, 72). The HIV-1 envelope glycoprotein gp120 binds to the mannose-6 phosphate receptor and rabies to a potassium channel and to receptors for acetylcholine and nerve growth factor (73, 74). This binding induces vesicularization and transport to the endosomes, but as these viruses can survive the lysosomal compartment, viable virus is eventually routed to brain. The various drug delivery approaches that use viral proteins and cell-penetrating peptides, the latter originally derived from the HIV-1 protein TAT, thus likely also are invoking adsorptive endocytosis and adsorptive transcytosis.

4.4 Diapedesis

This is the mechanism by which immune cells cross the barrier cells (75, 76). It involves an elaborate cross-talk between the immune and barrier cells. In some cases, it is thought the immune cells cross between barrier cells, whereas in other cases the immune cells tunnel through the barrier cell. Immune cell trafficking is not dependent on leaking across a disrupted BBB as often stated; such leakage would result in hemorrhage with the entry of millions of erythrocytes for each immune cell. BECs and ependymal cells are both known to be sites of transport, but currently, there are no investigations of immune cell passage at tanycytes (77). Immune cell passage is now appreciated to occur under normal circumstances and not just under immune insult. Diapedesis has elements similar to adsorptive endocytosis in that glycoproteins on the immune cell interact with glycoproteins on the barrier cell. Selectins (Selective Lectins) are involved in the early step of “capture”. Blockade of α₄ integrin and VCAM-1 binding by the antibody natalizumab is highly effective in the treatment of multiple sclerosis (78). Macrophages, which normally cross the BBB and whose trafficking rate into brain can be enhanced by inducing inflammatory events, have been used to deliver a variety of biologics to the brain (79). More recently macrophage-derived exosomes have been used as therapeutic vehicles as well (80).

4.5 Extracellular Pathways

These represent a kind of “functional leak” that allows small amounts of substances to “circumvent” the barriers (81). They are most predominant at the level of the subarachnoid/pial surface and provide access to the Virchow–Robin spaces subpial gray matter. Indeed, substances entering the brain by this route form an easily recognizable pattern on autoradiography. There is also evidence that entry exists at some circumventricular organs, at least in areas devoid of tanycytes (82). The rate of entry is low, essentially identical to that of albumin. Indeed, this seems to be along with CSF production a major route by which albumin enters the brain. As such, this may be a viable route for drug entry. Characteristics of substances shown to enter by this route are large proteins with a long plasma half-life and include IgG and IgM antibodies, erythropoietin, horseradish peroxidase, and complement (81, 83, 84).

6.1 Transcellular Diffusion

This topic and approach have already been discussed extensively above. Most small-molecule drugs cross by this mechanism. It is the easiest to model in silico and in vitro but the effects of the other factors discussed above such as effects of plasma protein binding and transporters in the blood-to-brain or brain-to-blood direction complicates modeling (44).

6.2 Mediated Transport

Few drugs in the pharmacopeia are known to use saturable transport systems to enter the brain. As with receptor and enzyme ligands, minor changes to the candidate drug can alter the binding affinity and so result in altered transport rates. This is likely one reason why “Trojan horse” approaches have been so difficult, in which another compound is bound to a transporter ligand, such as glucose, insulin, or transferrin. The compounds known to date to cross by saturable systems are modestly modified small molecules, such as l-dopa for treatment of Parkinson's, which uses the LNAA carrier, valproic acid which uses monocarboxylate transporter, and verapamil which uses an organic cation-carnitine transporter (113, 114). Biologics may be good targets for such modifications, especially when the region that binds to the receptor differs from the region that binds to the transporter. In such cases, transporter and receptor sites can be independently manipulated so that altered transport rate does not automatically also involve altered receptor function (115-117). Biologics also offer the opportunity of making manipulations that can decrease degradation and increase half-life but preserve transport and receptor activity. This is likely the case with anakinra, an analog of interleukin-1 receptor antagonist (IL-1ra), that is effective against epileptic states otherwise untreatable (118). It likely retains the ability of IL-1ra to cross the BBB and so, unlike the antibody canakinumab which crosses the BBB more poorly, is efficacious (119, 120). Mediated transport occurs for about 30–40% of drugs in the form of being ligands to varying degrees for efflux systems, most notably P-glycoprotein. As noted above, being a P-gp substrate can effectively block CNS activity as exemplified by loperamide. However, many P-gp substrates still exert effects on the CNS, such as morphine and many of the anti-epileptic medications. Efflux systems, including P-gp, are regulated and their altered activity can have consequences for drug efficacy. The enhanced activity of P-gp (and so reduced accumulation of its ligands in the CNS) with inflammation may explain why epileptic drugs are less effective during inflammatory states (121).

6.3 Opening the Blood Brain–Tumor Barrier

Although blood brain–tumor barriers tend to be more leaky than the vascular BBB, they are still much less permeable than the typical peripheral capillary bed. For over 30 years, hyperosmotic opening has been used with significant extension of life expectancy for glioblastoma as well as other primary and metastatic brain cancers (122).

7.1 Disrupting Barrier Integrity

In contrast to the good results that can occur with opening the blood brain–tumor barrier, most approaches to opening the vascular BBB have resulted in significant neurotoxicity. Various approaches and molecules designed to open the barrier, often in some limited way, have been tried. Neurotoxicity has inevitably resulted. Opening the BBB to even small molecules is problematic as so many circulating substances toxic to the CNS are themselves small, such as the adrenergics. Focused ultrasound has recently received a great deal of attention, but induces a proinflammatory reaction similar to that seen with stroke or traumatic brain injury (123).

7.2 Trojan Horse and Related Approaches

The basic idea for a Trojan horse approach is to attach the cargo that one wishes to deliver (that does not cross the BBB) to a delivering molecule that does cross. Two broad approaches have been to use as the delivering molecule an endogenous ligand that normally crosses ( glucose, insulin, transferrin, etc.) or to target a transporter protein with an antibody. Despite the simple elegance of this idea and a great deal of work by a large number of independent scientists and active programs by a number of pharmaceutical companies, product development has mysteriously proved very difficult. Broadly speaking, there are a few identifiable reasons why the Trojan horse approach has proven so difficult. One reason is the choice of delivery agent. Early studies attached glucose to much larger molecules and so likely negated glucose's ability to bind its own receptor. Transferrin, another commonly used delivery vehicle, itself does not cross very rapidly and about 90% of the transferrin taken up by the BEC is recycled back to the luminal, not the abluminal, side (124-126). Another reason that especially applies to antibodies is that they are often targeted to the receptor rather than the transporter protein. As discussed above, it is often assumed that the same protein that acts as a ligand's receptor is also its transporter protein. But when that is not the case, the antibody is targeting the wrong protein. Another problem, especially with off-target antibodies or greatly modified ligands, is that adsorptive endocytosis/transcytosis rather than receptor-mediated transport is induced (127). Although penetration of the BBB is achieved by adsorptive transcytosis, the cargo is often subjected to the lysosomal compartment or other subcellular membrane compartments and so needs to be able to survive these events.

8.1 Competitive or Mediated

8.2 Exosomes

Several studies have shown exosomes to readily cross the BBB. They can effectively deliver therapeutic cargo as well as act as biomarkers (80). Although no comparative studies have been yet reported, currently, it seems that exosomes derived from immune cells cross more rapidly than those derived from other tissues or cancer cells. The immune cell exosomes studied to date use the same selectins to cross the BBB and have transport increased by the same agents as do the immune cells from which they are derived (80). Exosomes from a brain-metastasizing melanoma cell line bound to several proteins of a human immortalized brain endothelial cell line (hCMEC/D3) with CD46 predominating (138). Thus, it seems that not only the proteins used as transporters will vary among exosomes but also the mechanisms for passage with diapedesis likely the mechanism for the reported immune cell exosome and adsorptive transcytosis for the reported melanoma cell line exosome.

8.3 The Barriers as Therapeutic Targets

The barriers themselves can cause or promote disease and so become direct therapeutic targets. Examples already discussed include Allan–Herndon–Dudley syndrome in which thyroid hormone is not transported across the vascular BBB and obesity where altered transport of leptin and ghrelin occur. Many, if not most, of the transporters for regulatory substances and hormones, are modified and so represent therapeutic targets. Examples discussed above include use of adrenergics for affecting transporters for leptin and lysosomal enzymes and use of gemfibrozil to regulate leptin transport indirectly through modulation of triglycerides.

Because of the polarized nature of barrier cells, a substance can act on one side of the barrier to stimulate release of another substance from the other side of the barrier cell. For example, LPS and some cytokines act at the luminal surface of BECs to induce the stimulation of prostaglandins from the abluminal surface, inducing fever (139, 140). The BEC can thus be targeted to modify the CNS action. For example, indomethacin acts on the BEC, blocking BEC release of prostaglandin and so reducing fever (140).

9.1 Intranasal

Many substances and even cells will enter the brain when placed at the level of the cribriform plate where the olfactory bulb protrudes from the cranium (141, 142). Thus, this differs from delivering drugs lower in the nasal cavity for either systemic delivery or treatment of nasal conditions. The major advantage often given for this route is that it “bypasses the BBB.” But the percent of the administered drug taken up by brain is often not much different between the IV and intranasal routes. The main advantage of the intranasal route is that very little of the drug is absorbed into the bloodstream and so the brain is more directly targeted (143). For drugs that have significant peripheral effects, such as insulin, intranasal delivery can be a viable option (144, 145).

Three routes have been discovered by which substances can enter the brain after placement of substances at the level of the cribriform plate: (i) movement through the brain interstitial fluid; (ii) movement through the CSF compartments; (iii) retrograde transmission through the olfactory and trigeminal nerves (141). Each of these routes has a distinct pattern of distribution within the CNS and one route or the other can dominate for a given substance. In addition, based on CNS distribution patterns for more recently studied substances, there are likely other mechanisms of routing. It is currently unclear what dictates which route a substance will use.

Early studies found distribution within the CNS after intranasal delivery was mostly confined to the olfactory bulb with little uptake elsewhere into the CNS. Critics have argued that this makes intranasal delivery useful only when the olfactory bulb is targeted and also that it may not be useful in humans because of their much smaller olfactory bulbs. However, there are now examples of peptides whose uptake into other brain regions such as the hippocampus and hypothalamus are equal to or greater than that of the olfactory bulb. Intranasal delivery has also been shown to be effective in humans (144, 146). Thus, the intranasal route is promising, especially for drugs that are rapidly degraded with systemic administration or have significant peripheral side effects.

Many preclinical studies and several clinical studies have shown the effectiveness of intranasal delivery. One of the most studied and one which illustrates the potential of the intranasal route is that of insulin (147), which has been studied in healthy young males and is in a clinical trial for Alzheimer's disease. Insulin within the CNS supports cognition and evidence suggests a deficiency of CNS insulin action in Alzheimer's disease (148-150). Although insulin crosses the BBB, it cannot be given systemically in amounts therapeutic to the CNS because those doses induce hypoglycemia. The combination of the saturable nature of BBB transport and CNS peripheral resistance further complicates delivery: the CNS resistance means that larger doses are needed in the brain, but the saturable nature of delivery means that there is a limit to blood-to-brain transport capacity. Targets of CNS insulin are likely other than or in addition to the olfactory bulb; preclinical studies show that intranasal insulin reaches the hippocampus and most other brain regions (143).

A few studies indicate that the distribution pattern of a substance given by the intranasal route can be altered by co-administering it with a binding substance, such as a cyclodextrin or albumin. For example, the peptide PACAP when combined with (2-hydroxypropyl)-β-cyclodextrin is routed away from accumulation in the striatum and towards accumulation in the thalamus (151). Insulin when co-administered with albumin had increased uptake by the cortex, but not by the hippocampus (152). Thus, it may be possible to develop combinations so that a therapeutic can be targeted to or away from specific brain targets.

There are many unresolved questions about intranasal delivery. Substances so administered distribute throughout the brain rapidly; how they can do so when diffusion within brain tissue is slow is unclear. What dictates which substances are taken up by brain after intranasal delivery, what determines distribution patterns within brain, and even why there should be a capacity for transfer of compounds at the cribriform plate are largely open questions. Why co-administration with cyclodextrin or albumin alters distribution and how that alteration is achieved are also unresolved questions.

9.2 Intraventricular and Intrathecal

Both of these routes deliver the therapeutic into the CSF space of the CNS. Interest in intrathecal delivery has had a resurgence of interest, especially regarding the delivery of biologics (153). Early experience with the intrathecal route was largely with anesthetics given into the lumbar region (154, 155). As small lipid-soluble drugs, they quickly crossed the capillary beds of the spinal cord in the CNS-to-blood direction, limiting residence time in the CNS and the opportunity to move up the spinal cord. As such, the intrathecal route was discounted for brain delivery. However, large water soluble molecules such as leptin, neurotrophins, and enzymes that are not rapidly transported out of the CNS were shown to slowly move into the cranial compartment and so reach the brain (156-158).

Another problem with delivery of substances into the CSF compartment is that they tend to diffuse only into the periventricular area and not to penetrate deeply into brain. Thus, on first consideration, only a small part of the brain would be influenced by this route. A clearer understanding of the role of the penetrating vessels and glymphatic pathways is leading to some revision of understanding of how the CSF compartment and brain interstitial fluid interact (159). In addition, there has long been evidence that the periventricular region receives nerve terminals and also contains cell bodies that project to deep regions of the brain; thus, substances confined to the periventricular region may still influence distant brain regions (13, 14). Lysosomal enzymes also seem to be able to make their way over time to deep brain regions (156); one hypothesis is that they are recycled among cells which would both hinder their clearance from the CSF and also aid in distribution. However, there are problems with repeatedly performing lumbar punctures in patients and the placement of catheters can be challenging in some clinical conditions, such as obese patients and growing children. Delivery in humans of antisense molecules is currently being conducted for the treatment of amyotrophic lateral sclerosis and for spinal muscular atrophies and cyclodextrin for the treatment of Niemann-Pick C disease (160-162).

9.3 Afferent Nerves

Perhaps the most striking examples of the potential role that afferent nerves can play are the preclinical studies showing that many of the CNS effects of gastrointestinal hormones, cytokines, and lipopolysaccharide are mediated through the afferent vagus (163, 164). This is likely a characteristic of at least the other cranial nerves, as illustrated by work involving the glossopharyngeal nerve (165). The role of the trigeminal nerve in intranasal delivery is discussed above. Whether such retrograde messaging is limited to cranial nerves is unclear.

9.4 Circumventricular Organs

These are small regions of the brain whose capillary beds are deficient in barrier function (166, 167). Classically known to play roles in blood pressure and nausea, they are served by afferent and efferent projections to distant areas of the brain and so likely influence many other behaviors, including feeding. Several drugs likely work in part or whole at the CVOs, particularly those related to blood pressure, anorexia, or nausea. Most are limited by the tanycytic barrier described above and so diffusion out of the CVOs into adjacent brain tissue is limited (17, 82). The tanycytes do possess secretory and transporter functions and so can connect the CVOs and adjacent brain tissues through these more regulated mechanisms (19, 21).

10.1 Pharmacokinetics

In targeting therapeutics to peripheral tissues, the role of pharmacokinetics is appreciated. Often with CNS drug delivery, the role of pharmacokinetics is forgotten in the emphasis on BBB permeability. All other things being equal, improved pharmacokinetics or related issues proportionately increase drug uptake by brain, just as for any other tissue. Circulating half-life, volume of distribution, protein binding kinetics, and related issues all play similar roles for drug delivery to brain as for other tissues. In thinking about pharmacokinetics, it may, however, be more useful to think about the barrier as the target tissue rather than the CNS, especially when transport kinetics are saturable or otherwise nonlinear. For many peptide and regulatory proteins, short half-lives in blood which are often minutes can be a greater obstacle to their therapeutic use than the barriers.

10.2 The Neonatal BBB

Adapted, Not Immature: It is widely assumed including by many who are interested in developing drugs for the neonate that the BBB is leaky to serum proteins until well after birth. The literature shows, however, that fetal brain capillaries attain barrier function to proteins similar to that of the adult soon after neovascularization (168, 169). Pericytes migrate to newly formed blood vessels soon after their formation to induce BBB function (5). In neonates, including rodents, the impermeability of the BBB is essentially the same as in adults. The transporter properties are not identical, however, as illustrated by a difference in the preference for various amino acids or the activity of the mannose-6 phosphate receptor (73, 170, 171). These differences are consistent with the BBB being adaptable to the needs of the CNS. As the needs of the developing CNS are different from those of the adult CNS, so are the transporter properties of the BBB. These differences can have effects on drug efficacy; for example, a low expression of P-glycoprotein may put neonates at higher risk for opiate toxicity (172).

10.3 Efflux vs Nonpermeability

Substrates that are avidly transported in the brain-to-blood direction by an efflux system may accumulate so little in brain as to appear impermeable to the BBB. However, why a substance does not accumulate in brain, whether limited by the physical barrier vs by efflux transport, is critical not only to developing a penetrating therapeutic but also to its behavior in disease states. Obviously, the approach to modification of the parent compound is very different between these two causes, with overcoming efflux requiring a knowledge of the mediator protein's binding properties. In addition, rates of efflux by transporters can vary with changes in physiologic conditions or with disease states. As discussed above, the efflux systems LRP-1 and P-gp are both underactive in Alzheimer's disease. Thus, many endogenous and exogenous substances normally effluxed by these systems are likely accumulating in brain. The converse of the role of efflux systems is that they may prevent many substances from accumulating to therapeutic levels. A classic example of this is for some of the protease inhibitors, which as substrates for P-gp are unable to reach concentrations in the CNS sufficient for the treatment of HIV-1 (173). Inhibition of efflux activity as with oligophosphorothioate antisense molecules can allow substrates to reach therapeutic levels within the CNS (134).