Nature and consequences of mammalian brain and CSF efflux transporters: four decades of progress

When I began my studies of ligand transport through the blood-brain (BBB) and blood-CSF barrier (BCSFB) in 1971, there was substantial knowledge of the anatomy and histology of the BBB and BCSFB as outlined in Davson’s (1970) magisterial book. Then, as now, the BBB and BCSFB were understood to depend on tight junctions between the cerebral capillaries (BBB), and the choroid plexus (CP) epithelial and arachnoid membrane cells (BCSFB), respectively. Moreover, Davson (1970) artfully explained the theory that ligand passage between blood and brain or CSF depended on lipid solubility, molecular size and degree of ionization. However, Davson pointed out there was newer data that iodide, penicillin, and other organic acids were transported out of CSF faster than expected (Becker 1961; Pappenheimer et al. 1961;Welch 1962; Fishman 1964). Moreover, there was also data that the excised CP incubated in artificial CSF concentrated iodide over 20 times by an energy dependent process (Welch 1962). Was the CP responsible for clearing iodide from CSF in vivo? We now know it is (Davson 1970).

To my eye, at that time, one of the most striking comparisons of steady-state ligand distributions in the CNS was between penicillin G and ascorbic acid (AA) (Spector and Lorenzo 1973, 1974), both water-soluble charged molecules of similar size and shape. In animals and humans, CSF concentrations of penicillin G and AA were ∼ one and four-hundred percent those of the plasma concentrations, respectively. This was obviously not consistent with the then current distribution theories. Moreover, the ratio between brain and plasma for these two molecules was even greater. Further doubts arose (about the then current distribution theories) from inspection of Hammarstrom’s (1966) whole body mouse autoradiographic pictures of (¹⁴C) ascorbic acid after intravenous injection. I was surprised to see the CP ‘light up’ first, then the CSF and over time the brain closest to CSF. Finally the whole brain lit up uniformly like a white ball. It certainly looked like the CP was transferring AA from blood into CSF; the AA was then entering brain from the CSF. We now know this occurs in mammals, why and how (Spector and Lorenzo 1973; Spector and Johanson 2006; Spector 2009).

We also know there are multiple (hundreds) of separate transporters (Table 1), many of these in the cerebral capillaries or CP (Hediger et al. 2004). Over the last 40 years, there has been substantial progress in understanding these transporters. In this review we will focus on three efflux transporters in the CNS at the BBB and/or BCSFB (Table 1b). In the past, much of the work in this area has been done using traditional techniques, e.g., in vitro radiochemical studies with isolated CP and brain endothelial cells to study ligand uptake, two chambered studies using tight-junctioned monolayers of forebrain capillary endothelial cells or CP epithelial cells to show directionality of transport between apical (lumen in capillaries and CSF-side in CP) and basolateral sides, ventriculo-cisternal perfusion studies to measure influx or efflux from the CSF compartment, brain efflux studies, intraventricular injection studies and whole animal distribution studies. Moreover, in vitro in freshly removed CP, Bresler et al. (1979) and we (Spector 1980b) were able to observe visually (with fluorescein and riboflavin, respectively, using fluorescence microscopy) ligands going from the medium, then being concentrated in the cytoplasm of CP epithelial cells and finally being concentrated on the basolateral side of the CP. Subsequent studies with confocal fluorescence microscopy confirmed and extended these results (Breen et al. 2002). These studies clearly showed there were two steps: a concentrative transport step at the apical membrane and a second (concentrative) efflux step at the basolateral membrane of the CP (Fig. 1). In other words, there were two systems involved – one at the CP epithelial apical membrane ‘in’ and a second one at the basolateral membrane ‘out’. In later studies it became clear that both riboflavin and fluorescein were transported on the same CP apical system – the organic acid transporter 3 (see Table 1 and below). Unlike AA which is transported into CSF from blood via CP (Spector and Johanson 2006), fluorescein and riboflavin are transported out of CSF into blood through CP by a complex bi-step system. However, the AA system in CP is also a bi-step system with sodium-dependent vitamin C transporter 2 transporting AA into CP from the blood side and by an unknown mechanism releasing AA from the CSF side.

From these and other data, specific efflux transporters were postulated at the BBB or BCSFB. However, a substantial amount of confusion arose about transporter specificity, species differences, location and implications for human biology and pharmacology.

During this time, in parallel, clinical neurochemists and others solved certain problems phenomenologically. For example, clinicians established cephalothin was not useful (in fact it was dangerous) for the treatment of bacterial meningitis whereas the larger cephalosporin, ceftriaxone and the related imepenem were useful even though the causative bacteria were equally sensitive to these three antibiotics in vitro (Spector 1990). It was not until many years later that basic scientists explained the neurochemical basis for these results (Spector 1986; Suzuki et al. 1987, 1989: See below). Another example is the use of cancer chemotherapy drugs – many of which are extremely neurotoxic (e.g., vinblastine, taxol) yet could be safely given to patients. It was only with the discovery of the existence and function of P-glycoprotein (P-GP) (Table 1) at the BBB that this mystery was at least partially solved. Finally, it was known for years that collie dogs but not other dogs were exquisitely sensitive to the neurotoxicity of ivermectin, an anti-parasitic drug. Why? Once again it was ultimately established that collie dogs were P-gp knock-outs (See below).

Over the last 15 years, there has been tremendous progress with the application of modern neurochemical, immunological and molecular biological techniques. Although the early workers were sagacious in postulating a large number of specific systems at the BBB or BCSFB – systems that transport therapeutically important or toxic ligands out of brain and CSF, until recently the evidence was mainly indirect (e.g., kinetic.) In this paper we will review some of the newer direct results – focusing exclusively on efflux transporters in the CNS.

But before beginning we should now briefly place the BBB and BCSFB in proper perspective. Then we must explain the criteria necessary to draw correct conclusions in transport biology in the CNS – a complex area – to avoid the mistakes of the past.

We now know the BBB and BCSFB have four general functions; nutrition of the brain, homeostasis, protection of the brain especially from toxins, and bidirectional signaling. This is done by mechanisms shown in Table 2. As noted above, our focus will be on number 3, Table 2.

Efflux transporters: criteria for functionality in vivo

What are the criteria for identifying and characterizing a protein as an efflux transporter at the BBB and/or the BCSFB? We and others have suggested the methodological criteria outlined in Table 3 (Spector 2000; Pritchard and Miller 2005). These criteria are stringent and burdensome but there are reasons for this. For example, unlike the nutrient (influx) carriers, which have fairly strict ligand preferences (e.g., the CP AA transporter basically transports only AA; its stereoisomer has only 10% the affinity for the transporter – sodium-dependent vitamin C transporter 2) (Spector and Johanson 2006), some of the efflux transporters are almost unbelievably promiscuous (See Table 4) and there is overlapping specificity in some cases as discussed below. Moreover, in a few cases, even inserting the same transporter gene into Xenopus eggs or alternatively mammalian cells, one obtains not just quantitative kinetic differences, but qualitative differences (Pritchard and Miller 2005) making interpretation problematic. Hence, to be certain of conclusions, one needs to have all the criteria derived from the methods and techniques in Table 3– consistent and coherent.

The reasons for focusing on organic acid transporter 3 (oat-3; non-human), P-gp and peptide transporter-2 (pept-2) (Table 1) are four-fold. First, beside different specificities, they have different mechanisms. Second, they are located in different places: oat-3 is located on both the CP apical and abluminal brain capillary membrane; pept-2 is located on the apical choroid plexus membrane but not in or on cerebral capillaries; P-gp is located at the luminal brain capillaries wall. Third, all three transporters can meet most of the criteria in Table 3 [e.g., there are healthy knockout (KO) mice for all three]. Moreover, there is more convincing data for these three, in my view, than for most of the other CNS efflux transporters, and individual compounds can be found specific for these transporters (e.g., penicillin for oat-3; cefodroxil for pept-2 and ivermectin for P-pg) (Table 4). Finally, the unraveling of the transport mysteries of these three brought extremely important practical medical and pharmacological data.

In what follows, I will not follow the historical patterns of discovery but will review data that are relevant to these transporters today. At the end in the general discussion, I will briefly review the ontogeny (development) of these transporters, the speculation for the existence of these transporters (e.g., from an evolutionary point of view) as well as fit those transporters into modern life – especially with regard to medicine and toxicology. I will also make comments on our ability to predict the fate of exogenous compounds vis a vis the systems in Table 4, and our ability to use the techniques involved in the characterization of those transporters as pharmaceutical screening methods.

P-glycoprotein

The characteristics of P-gp in cerebral capillaries are shown in Table 10 (Schinkel et al. 1994; Jonker et al. 1999; Sadeque et al. 2000; Sasongko et al. 2005; Hsiao et al. 2006; Daood et al. 2008; Geyer et al. 2008; Aller et al. 2009;Liow et al. 2009). In humans, P-GP is detectable in cerebral capillaries at ∼ 22 weeks of gestation (Daood et al. 2008). One of the most striking characteristics is the extreme promiscuity of this transporter (Table 4) (Tang et al. 2009). It transports molecules from molecular weight of 400 to ∼ 4000 out of brain. It is interesting that P-gp is ∼ twice as large as oat-3 and pept-2 consistent with it transporting much larger molecules (Table 4). P-gp also lowers ligand concentration to a lesser extent in CSF (Tang et al. 2009). This degree of promiscuity is unusual in nature since most enzymes, receptors and transporters are more specific – some only processing one ligand. Both the human and mouse transporter tend to favor more lipid soluble molecules. There is some evidence that the mouse transporter is more ‘powerful’in vivo; there is also ∼ twice as much p-gp transporter RNA in mouse brain capillaries (Syvänen et al. 2009; Warren et al. 2009).

Table 10. Characteristics of rodent P-gp in CNS

(1) Rodent genes – Expressed as protein 140–170 kDa

(2) Messenger RNA – localized in brain capillaries – not CP

(3) Transport protein – hydropathy analysis suggests 12 transmembrane domains in two 6 domain ‘halfs’– putative structure and mechanism – see text and Fig. 2

(4) Transporter location: luminal side of brain capillary

(5) In vitro studies

Monolayers of capillary endothelial cell show clear abluminal to luminal transport

(6) Rodent transporter expressed in mammalian cells and Xenopus oocytes

(7) In vivo techniques show P-gp ligands excluded from brain; inhibitors of P-gp increase brain ligand levels

(8) KO studies in mice (and collie dogs) show loss of capillary P-gp; in vivo ivermectin mouse brain levels in KO mice are 30 times (relative to plasma levels) higher than in wild type mice. CSF levels of P-gp ligands in KO mice also increase although not as much as brain

(9) Coherent and consistent studies with P-GP in:

Direct studies in vivo and in vitro

Inhibition studies

Specificity and mechanism partially defined – see text

(10) Human studies consistent with animal data: 87% sequence homology mouse to human; P-GP located in human cerebral capillaries; similar kinetics to mouse; but some evidence exists that P-GP in humans not as active as in mice (see text)

In vivo, in both animals and humans, P-GP serves a protective role – keeping potentially very toxic substances and drugs out of brain (e.g., ivermectin, loperamide and domperidone, a D-2 blocker; Table 4). A related D-2 blocker, metaclopromide more freely enters brain and has long term severe chronic toxicity with a black box USA Food and Drug Administration (FDA) warning. Also in cancer chemotherapy, many chemotherapeutic drugs are kept out of the brain by P-GP – this may either be helpful or harmful depending on the clinical situation.

Although educated guesses could be made to predict P-gp ligands, until Aller et al. 2009, the mechanism for P-gp activity was only known in rough outline. If an investigator really wished to know, that person would have to study the compound. (See discussion.)

A recent surprising finding in vivo is the synergistic interactions between P-gp and breast cancer resistance protein (bcrp or BCRP; ABCG2) at the BBB (Zhou et al. 2009). Both proteins are present in the luminal capillary membranes of mice and humans, and the ligand specificities of P-gp and bcrp overlap. An example of synergy was documented in vivo in mice (Zhou et al. 2009). In bcrp KO mice, there is little effect on imatinib distribution in brain compared to wild type mice whereas in P-pg KO mice, they have four times (relative to plasma) more drug in brain than wild type mice. However, in double KO mice, i.e., those without P-gp and bcrp, there is 28 times more drug in brain relative to wild type mice. BCRP is a ‘half’ transporter with only six trans-membrane domains and one ATP binding site, but as described above can synergistically interact with P-gp by an unknown mechanism (Zhou et al. 2009). This complicates the interpretation of single gene P-gp KO mice.

P-glycoprotein also plays a role in non-sedating anti-histamine H1 blocker pharmacology in the human CNS (Chen et al. 2003). The original histamine H1 blockers, e.g., diphenhydramine (DPH), were sedating (Goldberg et al. 1987; Tsuji 2005). In fact, although a nuisance to allergy sufferers, this property is employed to this day in over-the-counter sleep aids. DPH, although a charged tertiary amine at pH 7.4, is rapidly cleared from blood into brain by a saturable mechanism at the BBB (Goldberg et al. 1987; Tsuji 2005). DPH is not a ligand for P-gp. However, the non-sedating anti-histamine H1 blockers are ligands for P-gp (Chen et al. 2003), and thus somewhat excluded from brain. This property partially explains their non-sedating properties.

Finally, the molecular mechanisms of action of P-gp have been recently clarified by Aller et al. (2009) with their unraveling of the x-ray structure of apo-P-gp at 3.8 angstroms. P-gp shows an internal cavity of 6000 angstroms cubed with one ATP binding site on each half of the molecule (See Fig. 2). Inside the main cavity there are smaller distinct binding sites with definite stereo-selectivity. Ligands enter from the inner leaflet of the lipid bilayer and are then ejected into the blood (Fig. 2). The process is driven by ATP hydrolysis. The reader is encouraged to read and view the elegant figures in the text and supplementary material of this triumph of structural biology (Aller et al. 2009). With this work Aller et al. (2009) have helped explain the broad specificity and location of this carrier at the capillary lumen. Similar studies also need to be done with oat-3 and pept-2 to understand their broad specificity.

Discussion

In discussing brain and CSF concentrations of substances (i.e., the distribution of exogenous ligands in brain and CSF relative to blood, especially at steady state), the rate of ligand entry, exit, tissue binding, and intra-brain and CP metabolism are the four key factors. In this paper we have mainly discussed ligands where CNS metabolism and tissue binding are not important factors (with a few exceptions like ceftriaxone which is heavily bound to plasma proteins). Our focus has been on BBB and BCSFB efflux transporters which, as described above, can radically alter brain and/or CSF concentrations of myriad ligands, in some cases by a factor of 10–30 times.

As noted in the text and Tables 3, 5, 9, and 10, the scientific transport data supporting our interpretations of oat-3, pept-2 and P-gp in animals and humans are strong, plentiful and coherent. There is no doubt these three separate transporters are present in mammalian species at the BBB and/or BCSFB and have important functions.

It is also clear that lipid solubility, molecular size and charge frequently do not determine ligand distribution in brain. Hundreds if not thousands of compounds are exceptions. For example, in normal mice and dogs, because of P-gp, the brain concentration of ivermectin is ∼ 30 times lower (relative to plasma) than in KO animals (Schinkel et al. 1994). For certain ligands in the CNS like ivermectin for P-gp, cefadroxil for pept-2 and penicillin G for oat-3, the effects are large and important.

With regard to protective transport functions of oat-3, pept-2 and P-gp in the CNS, it is interesting to compare the kidney tubules, and the brain capillaries and CP epithelial cells; the former protect the body as a whole and the later two, the brain. In the tubules, there are substantial concentrations of oat-3, pept-2 and P-gp; oat-3 on blood side secretes ligands (e.g., riboflavin, penicillin) into tubular cells and then the ligands are transferred into urine by an unknown mechanism (Spector 1982); pept-2 reabsorbs di- and tripeptides from urine into tubular cells for intracellular hydrolysis to amino acids (conservation), and P-gp ejects many ligands back into urine. Thus, the polarity is different in the kidneys but the functions are analogous to those in cerebral capillaries and CP epithelial cells. In a metaphorical sense, the choroid plexus could be conceived of as a kidney tubule – like structure in the brain.

What is somewhat surprising is that KO mice without oat-3, pept-2, and P-gp and collie dogs without P-pg are normal at birth, remain normal and reproduce normally unless challenged with certain ligands. This suggests that these three transporters are not essential to life, or possibly other mechanisms take over when these are absent. As noted above, all three of these transporters are normally present at birth. There is also one piece of data (as noted above) that at least one living human has a non-functional PEPT-2 transporter (Terada et al. 2004).

If KO animals are generally normal, why do these systems with their very broad specificity exist at the BBB and BCSFB? There are two schools of thought: one is that these systems were originally involved in physiological functions (e.g., in both kidney and CP). For example, when the concentration of riboflavin in blood or CSF rose above a certain limit, it would be secreted by OAT-3 into the urine or blood by OAT-3 in the tubules or CP epithelial cells, respectively. Gradually, the specificity broadened as animals with more protective transporter functions survived the toxins in the environment – particularly in food. This theory was originally proposed by Levy (Spector 1982). The second theory is that from the start these systems provided both ‘house cleaning’ of endogenous waste products like metabolites of serotonin and catechol amines, and also performed protective functions from exogenous agents/toxins (Deguchi et al. 2006).

However the extremely broad specificity of these transporters arose, it is remarkable. One would never anticipate that penicillin, riboflavin and cimetidine would travel on the same transporter (Table 4). But now knowing this, what are the consequences? For therapy? For toxicology? For drug interactions?

For therapy, these efflux systems are very useful if one is using agents that are brain-toxic and the therapist does not want the agent in the brain (e.g., ivermectin; loperamide; cimetidine). On the other hand, if one is attempting to treat brain disorders (e.g., brain cancer; meningitis), since many effective chemotherapeutic agents are transported out of the CNS by these transporters, they interfere with therapy.

In toxicology, these transporters (Table 4) without question protect the brain from many potential toxins. Even penicillin G is neurotoxic if the plasma levels are too high. For animals and presumably primitive man who ingested many toxins, one can imagine the great utility of these systems at the BBB and BCSFB.

Two related questions that were frequently raised in the past can now be answered. First, although there are clearly ligand interactions in vivo with oat-3, pept-2 and P-gp in animals, are these present and important in humans, especially drug interactions? A related question is: are the levels and functions of these three transporters comparable in people and animals.

In terms of drug interactions, in both rodents and humans cyclosporin (at a clinically tolerable concentration) increases the concentration of verapamil (a P-GP, ligand) in brain comparably (∼ 2 times) (Sasongko et al. 2005; Hsiao et al. 2006). Similarly, quinidine depressed breathing in humans given loperamide, a narcotic agent normally excluded from the brain by P-GP (Table 4) (Sadeque et al. 2000). Thus there is the potential for drug interactions in humans – especially with P-gp. Whether such drug interactions can be useful therapeutically remains to be seen.

In terms of quantitative similarity of these efflux transporters in animals and humans, with OAT-3, there is substantial human evidence that penicillin G distributes itself in CSF as it does in animals, strongly supporting a comparable OAT-3 mechanism. Also, when measured directly the concentration of mRNA for P-GP in human brain capillaries is about one-half that of rodents (Warren et al. 2009). However, there are some data using positron emission tomography scanning that shows some P-GP ligands are better kept out of rodent than human brain – thus implying more P-GP ‘activity’in vivo in rodent CNS (Syvänen et al. 2009).

For an individual compound, can one predict whether it might be transported out of the CNS by OAT-3, PEPT-2 or P-GP? As discussed above, there are broad rules about the specificity of these transporters, but to obtain quantitative, reliable data, simple in vitro techniques can be employed. For example, for oat-3 and pept-2 one can see whether the labeled ligand is accumulated by the isolated CP in vitro. If so, one can perform ligand competition experiments or can study monolayers of CP epithelial cells expressing either pept-2 or oat-3 on the apical side. Similarly, in vitro studies with monolayers of cerebral endothelial cells expressing P-gp can be used as described above. In vivo, there are multiple, sophisticated techniques (Table 3).

Also, there are some data that the concentration, location (on the membrane or intracellularly) and activity of these transporters might be altered in disease states and/or by immunological mediators. The reader is referred to Miller et al. (2008) for a discussion of this literature.

In summary, the reader can see the fantastic progress made over the last 40 years (e.g., from simple notions that penicillin and iodide are cleared from the CSF too fast by an unknown mechanism). Our current knowledge base has expanded remarkably. In some cases, (e.g., riboflavin and fluorescein) one can actually watch the ligand crossing the CP epithelial cells from one side to the other, and then out. This is especially striking with riboflavin (Spector 1980b). There are, however, still several glaring holes in the fabric of our knowledge. One persistent hole is how unmetabolized ligands of oat-3 and pept-2 efflux the other side of the CP or, in the case of oat-3, brain capillary endothelial cells. There are many candidates but none have been unequivocally established. Another is a lack of understanding on a molecular basis how oat-3 and pept-2 work. These problems are solvable with more work. The future of basic and clinical neurochemistry in this area is bright indeed. Moreover, the practical consequences and applications are and will be even more important.