• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of pnasPNASInfo for AuthorsSubscriptionsAboutThis Article
Proc Natl Acad Sci U S A. May 1, 2007; 104(18): 7594–7599.
Published online Apr 26, 2007. doi:  10.1073/pnas.0702170104
PMCID: PMC1857226
From the Cover
Medical Sciences

Targeted delivery of proteins across the blood–brain barrier

Abstract

Treatment of many neuronal degenerative disorders will require delivery of a therapeutic protein to neurons or glial cells across the whole CNS. The presence of the blood–brain barrier hampers the delivery of these proteins from the blood, thus necessitating a new method for delivery. Receptors on the blood–brain barrier bind ligands to facilitate their transport to the CNS; therefore, we hypothesized that by targeting these receptors, we may be able to deliver proteins to the CNS for therapy. Here, we report the use of the lentivirus vector system to deliver the lysosomal enzyme glucocerebrosidase and a secreted form of GFP to the neurons and astrocytes in the CNS. We fused the low-density lipoprotein receptor-binding domain of the apolipoprotein B to the targeted protein. This approach proved to be feasible for delivery of the protein and could possibly be used as a general method for delivery of therapeutic proteins to the CNS.

Keywords: fusion protein, lentiviral vectors, low-density lipoprotein receptor

The blood–brain barrier (BBB) controls the passage of substances from the blood into the CNS. Thus, a major challenge for treatment of brain disorders is to overcome the impediment of delivery of therapeutic macromolecules to the brain. Although direct injection of therapeutic proteins may be a reasonable approach for treatment of localized neural degenerative disorders that involve discrete anatomical structures within the brain (1), the treatment of many neurological disorders requires the delivery of a therapeutic protein or peptide to the whole CNS. Delivery of these proteins is hampered by the tight regulation of the BBB. Current protocols for delivery of viral vector-mediated gene delivery involve the stereotaxic injection of the vector to the CNS, resulting only in localized gene expression (2). The small size of the mouse may allow more widespread expression with as few as five injections across the whole brain; however, the larger size of the human brain would require far too many injections to be clinically feasible. Thus a new approach for targeting these proteins to the CNS across the BBB is required for future treatments of widespread neural degenerative conditions.

Vascular distribution of a therapeutic protein would be a preferred method. The human brain contains on the order of 100 million capillaries containing a surface area of ≈12 m2 (3). Nearly every neuron in the brain has its own capillary, with an average distance from capillary to neuron of 8–20 μm (4). Thus, delivery of a therapeutic protein to neurons across the capillary membrane would be a method of choice. However, delivery of proteins by vascular distribution to the CNS is not possible due to the presence of the BBB, which is composed of a tightly sealed layer of endothelial cells and numerous astrocytic processes that regulate the passage and diffusion of proteins such as growth factors from the blood stream to the CNS. Small molecules on the order of 400–500 Da as well as some small lipid-soluble proteins can pass across the BBB unassisted; however, transport of almost all larger proteins to the CNS occurs via receptor-mediated transcytosis to the CNS (5). Well characterized BBB receptors include the following: low-density lipoprotein (LDL) receptor (LDLR), transferrin receptor, and insulin-like growth factor receptor (6).

The LDLR family is a group of cell-surface receptors that bind lipoprotein complexes for internalization to the lysosomes. The family is composed of ≈10 different receptors that are expressed in a tissue-specific manner and primarily bind apolipoprotein complexes (79). The apolipoproteins, of which the two most prominent members are apolipoprotein B (ApoB) and apolipoprotein E (ApoE), function to bind lipids in the blood stream and target them for lysosomal degradation. Apolipoproteins bind to the LDLR on the cell surface of the targeted cell, and then the complex is endocytosed. Conversion to an early endosome and subsequent lowering of the compartmental pH result in release of the apolipoprotein and recycling of the receptor to the cell surface. In contrast, at the BBB, LDLR binds apolipoproteins, resulting in transcytosis to the abluminal side of the BBB, where, presumably, the apolipoprotein is released to be taken up by neurons and/or astrocytes (reviewed in refs. 3 and 10).

We therefore hypothesized that if a secreted form of a protein that can bind to the LDLR can be produced in one central location or depot organ, such as muscle or liver, upon release in the blood stream, then the recombinant protein will bind to the LDLR and transcytose to the CNS. In this instance, the lentivirus (LV) vector system can prove useful because i.v. and i.p. delivery of a nonreplicating LV vector efficiently delivers genes to the liver and spleen, allowing these organs to function as the sites of expression and secretion of a therapeutic protein (11).

Results

The LDLR-binding domain of ApoB was fused to the C terminus of a mouse cDNA capable of generating the secreted form of GFP (sGFPmApoB) or the secreted form of glucocerebrosidase (sGCmApoB; the enzyme deficient in the lysosomal disorder Gaucher's disease) (Fig. 1). This recombinant protein can be expressed from the LV vector delivered by a single i.p. injection. To determine the extent of delivery of sGCmApoB and sGFPmApoB to the CNS, adult mice (>8 weeks) were used to mimic the BBB environment observed in the patient (12).

Fig. 1.
Schematic drawing of lentiviral vectors. The glucocerebrosidase [or enhanced GFP (eGFP)] gene was cloned with the preprotrypsin secretory signal (ss) at the N terminus, and the myc epitope tag was cloned at the C terminus. The ApoB LDLR-binding domain ...

Mice were examined 14 days after vector delivery to allow for expression and uptake of the recombinant protein, without allowing sufficient time for a strong antibody response to the newly introduced protein to commence. Recombinant protein was easily visible in the liver and spleen of mice that had been injected with either the LV–sGCm or LV–sGCmApoB vectors (Fig. 2). In the liver, recombinant protein was observed primarily in cells lining the hepatic sinuses, such as perisinusoidal cells, Kupffer cells, and endotheliocytes, and, to a lesser extent, in hepatocytes, as noted previously (11). In the spleen, germinal centers appeared to contain the majority of recombinant protein. Because of the small cytoplasm-to-nucleus ratio of these cells, we have tentatively identified them as lymphocytes. There did not appear to be any recombinant protein taken up in the cells of the lung. Sections from animals injected with saline did not show staining in any tissue (Fig. 2 GI).

Fig. 2.
Immunofluorescence staining of recombinant protein in peripheral tissues of injected mice. Sections (20 μm) of liver (A, D, and G), spleen (B, E, and H), and lung (C, F, and I) were stained for recombinant protein with the myc epitope tag (green) ...

The recombinant sGCmApoB and sGFPmApoB, stained with myc antibody (green), showed extensive uptake of the proteins in the CNS only when the ApoB LDLR-binding domain was fused to the protein. To confirm that staining was specific for neurons, we used two neuron-specific markers, calbindin (Fig. 3) and NeuN (Fig. 4). Calbindin is a 28-kDa membrane-bound calcium channel protein found primarily in the Purkinje cells of the cerebellum (13) and the interneurons of the cortex (14) and, to a lesser degree, in other regions of the CNS. NeuN is a neuron-specific nuclear protein present in nearly all neurons of the CNS, with the exception of the Purkinje cells of the cerebellum (15, 16).

Fig. 3.
Immunofluorescence staining of recombinant protein in the cerebellum of injected mice. Brain sections (40 μm) were stained for calbindin (red), myc epitope of the recombinant proteins (green), LDLR (purple), and nuclei (blue). (AD) A ...
Fig. 4.
Immunofluorescence staining of recombinant protein in the brains of injected mice. Brain sections (40 μm) were stained for NeuN (red), myc epitope of the recombinant proteins (green), LDLR (purple), and nuclei (blue). (AD) A representative ...

The recombinant proteins sGCmApoB and sGFPmApoB (green in Fig. 3 C and G) were found throughout the Purkinje cell layer (red in Fig. 3 B and F) of the cerebellum but not in the adjacent granular layer (blue in Fig. 3 A and E). This expression pattern corresponded with the expression of the LDLR (purple in Fig. 3 D and H), which suggests that the LDLR is necessary for uptake of the recombinant protein in the Purkinje cells. Protein uptake was restricted to recombinant protein containing the ApoB LDLR, because the control sGCm was not taken up by cells in the cerebellum (Fig. 3 IK). Recombinant protein (yellow arrows) was observed sporadically throughout the cerebellum but was not associated with calbindin (Fig. 3J). This is probably staining of the protein associated with capillaries of the BBB.

NeuN is a nuclear marker associated with the vast majority of neurons of the CNS. The recombinant proteins sGCmApoB and sGFPmApoB (green in Fig. 4 C and G) costained with the NeuN neuronal marker (red in Fig. 4 B and F) as well as the LDLR (purple in Fig. 4 D and H). In contrast, no staining of the sGCm control protein was observed with the NeuN neuronal protein (Fig. 4 IK). Recombinant protein was observed throughout the CNS, with greater concentrations found in the cortex, striatum, and olfactory bulb and lower concentrations in the granular layer of the hippocampus and the cerebellum. Thus, addition of the ApoB LDLR-binding domain facilitated uptake of the recombinant protein in both NeuN and calbindin subsets of neurons, and this uptake was associated with the expression of LDLR.

To visualize the passage of the recombinant protein across the BBB, we stained CNS sections with glial fibrillary acidic protein (GFAP), a marker of astrocytes and lectin, a protein that binds to the endothelial cells of the BBB (see pictorial representation in Fig. 5M). Because astrocytes are known to express the LDLR (17), we suspected that we would observe uptake of the recombinant protein by GFAP-positive cells. The GFAP (red) cells take up the recombinant proteins sGCmApoB and sGFPmApoB (green in Fig. 5 C and G). The endothelial cells are visualized in purple (Fig. 5 D, H, and L), and the adjacent GFAP-positive astrocytes on the CNS side of the BBB are visualized as red (Fig. 5 B, F, and J). The addition of the ApoB LDLR-binding sequence facilitates passage of the recombinant protein across the endothelial cells (purple) of the BBB because no protein can be observed within the vessels or associated with the endothelial cells. This protein is then taken up by astrocytes, as can be determined in the merged picture (yellow in Fig. 5 A and E). In contrast, the sGCm protein lacking the ApoB LDLR-binding domain can be observed within the vessel lumen but is not directly associated with the endothelial cells of the BBB nor with the astrocytes located adjacent to the endothelial cells (Fig. 5I Inset).

Fig. 5.
Immunofluorescence staining of recombinant protein in the endothelial cells in the brains of injected mice. Brain sections (40 μm) were stained for GFAP (red), myc epitope of the recombinant proteins (green), lectin (purple), and nuclei (blue). ...

Binding of ApoB to the LDLR results in targeting of the protein to the lysosome, where the LDLR is released and recycled to the cell surface. To determine whether the recombinant protein containing the ApoB LDLR-binding domain would behave in a similar fashion, brain sections were stained for the sGCmApoB (green in Fig. 6C) as well as the lysosomal membrane marker lysosomal-associated membrane protein (LAMP)1 (red in Fig. 6 B and E). Recombinant sGCmApoB protein appears in the characteristic punctate location across the whole cytoplasm and costained with LAMP1 at a perinuclear location (Fig. 6A). This confirmed the targeting of the recombinant protein to the lysosomes. No costaining was observed with sGCm protein lacking the ApoB LDLR-binding domain (Fig. 6 DF).

Fig. 6.
Colocalization of sGCmApoB with LAMP1. Brain sections (40 μm) were stained for LAMP1 (red), myc-tagged recombinant protein (green), and nuclei (blue). (AC) Cells from a mouse injected with LV–sGCmApoB. (DF) Cells from ...

Sections of the CNS that were stained for calbindin, NeuN, or GFAP were randomly examined blindly to determine the percentage of cells that appeared to take up the recombinant protein. Sections from mice that had been injected with the LV–sGCmApoB virus were observed over at least 10 fields (Table 1). Calbindin cells were counted in two separate regions of the brain. In the cerebellum, the calbindin-positive cells (Purkinje cells) all appeared to take up the recombinant protein. In contrast, only 55% of the calbindin-positive cells of the cortex (interneurons) appeared to take up the recombinant protein. Approximately 30% of NeuN-positive neurons took up the recombinant protein, with the greatest percentage occurring in the striatum and the lowest percentage in the hippocampus, where little or no uptake was observed. Astrocytes (GFAP-positive cells) almost universally appeared to take up the protein, with 84% of the cells staining for the recombinant protein.

Table 1.
Percentage of cells examined in the brain that were positive for the recombinant protein

Discussion

In this report, we have described the delivery of proteins to the CNS by addition of the LDLR-binding domain of ApoB. Delivery of the lentivector expressing the recombinant protein by i.p. injection was sufficient to deliver protein to the CNS across the BBB. This transport was specific to the protein with the ApoB LDLR domain, because the control protein sGCm did not cross the BBB when delivered in the same manner. Because delivery of the lentivector by i.v. or i.p. injection was performed with no appreciable difference in delivery efficiency, the technically simpler i.p. delivery route was performed for all experiments. Although the ApoB LDLR sequence is 38 aa, the length did not appear to greatly affect delivery or function of the recombinant protein. The ApoE LDLR-binding site containing amino acids 152–168 (18, 19) functioned similarly to the ApoB sequence used in this report (data not shown).

The BBB is an effective way to protect the brain from the many chemicals flowing around the body and common infections, yet the brain needs certain nutrients and essential molecules, such as cholesterol, for normal functioning. We have taken advantage of the mechanisms that allows the brain to import lipids from the blood stream. One of the mechanisms involves the use of the LDL family of receptors, which are expressed on most cells in the body, including endothelial cells, which make up the BBB. Specificity of uptake of LDL complexes is determined by the apolipoprotein bound to the complex, such that ApoB binds specifically to the LDLR and megalin (7, 9). LDLR is found primarily in the liver and the adrenal gland and, to a lesser extent, in the brain, muscle, and lung (reviewed in ref. 10). Examination of nonneuronal tissues showed elevated levels of the glucocerebrosidase enzyme in the liver, spleen, and muscle (data not shown) and little or no increased activity in the lung. This specificity of tissue distribution closely mimics the expression levels of the LDLR. In an extension of these studies, we delivered an ApoB fusion lysosomal enzyme to the CNS of knockout mice with the protocol described here (manuscript in preparation). We observed a 50% increase in enzyme activity in the CNS of these animals over control animals that received the lysosomal enzyme without the ApoB fusion. Thus, the ApoB LDLR-binding peptide can be used to deliver a protein to many tissues of the body in addition to delivery to the CNS.

The use of the LDLR as a target for BBB transport will prove most useful in the treatment of lysosomal storage diseases that involve neural degeneration, such as Gaucher's disease. Currently, enzyme-replacement therapy is an effective treatment for the lysosomal storage diseases; however, there is no effective treatment for the neuronal degenerative component (20). The infused enzyme is unable to cross the BBB and, therefore, has no access to the affected neurons and astrocytes. Delivery of a lentiviral vector expressing the fusion protein described in this report may prove effective in treating the neuronal degenerative component of many metabolic disorders, while, at the same time, still effectively treating the peripheral component of the disease. Preliminary results indicate that delivery of an ApoB-fused lysosomal enzyme can prevent some of the neuronal loss associated with mucopolysaccharidoses diseases (B.J.S., unpublished results).

Staining of the liver and spleen indicated significant protein in these two organs. This was further verified by glucocerebrosidase enzyme assay of tissue lysates. Although the immunohistochemistry indicated recombinant protein in both of these tissues, we could not differentiate between cells that were transduced with the LV and expressing the recombinant protein and those cells that had taken up the protein because of LDLR-mediated endocytosis. In contrast, the recombinant protein observed in the CNS could have come from only transcytosis of the protein across the BBB because the LV is not known to cross the BBB. In addition, only the proteins fused to the ApoB LDLR-binding domain were observed in neurons or astrocytes. The localization of the recombinant protein to specific regions or cell types within the brain appears to correlate with the expression of the LDLR, as is most evident in the cerebellum. In this region, Purkinje cells stained positive for LDLR as well as the recombinant protein in contrast to the adjacent granular layer that did not stain for the LDLR.

Although we chose to target the LDLR for transport of the recombinant proteins to the CNS, it may be useful to examine other receptors on the BBB for targeting. The LDLR, similar to the transferrin receptor that is also expressed on the BBB, targets bound proteins to the lysosome. Other receptors may be more useful in delivering soluble growth factors that would be effective at the extracellular level. Alternatively, addition of a peptide-cleavage sequence may be possible to release the therapeutic protein from the targeting sequence once it has entered the CNS. In this instance, γ- or β-secretase (21) may be attractive targets for releasing the protein from its targeting peptide.

Previous attempts to target proteins for transport across the BBB have all relied on the same technology. An antibody against a receptor expressed on the BBB is bound to the targeted protein in vitro before i.v. delivery (2224). This method allows delivery of the targeted protein to neurons; however, it does not allow for continuous delivery of the targeted protein without repeated i.v. injections. The system we have described here utilizes the liver as a depot organ to express and distribute the recombinant protein for sustained delivery to neurons and astrocytes of the CNS. It remains to be seen whether the level of delivery from the lentivector can achieve physiological levels; however, the first step in targeting these proteins without the need of intracranial injection or BBB-disrupting drugs now appears possible.

Methods

Cloning.

The human glucocerebrosidase gene (American Type Culture Collection, Manassas, VA) was cloned in frame into the pcDNA3.1–myc–His vector (Invitrogen, Carlsbad, CA), generating a C-terminal myc–His-tagged protein. Amino acids 3371–3409 of human ApoB were cloned into the BamHI and HindIII sites of this plasmid, replacing the His tag. Finally, the secretory leader sequence of preprotrypsin (pFLAG–CMV-1; Sigma, St. Louis, MO) was cloned in frame at the ApaI site at the N terminus to aid in secretion of the protein. This construct was designated pcDNA–sGCmApoB. A similar construct was generated lacking the ApoB sequence and was designated pcDNA–sGCm.

These constructs were cloned into the third generation self-inactivating LV vector (25), with the chicken β-actin promoter driving expression producing the vectors LV–sGCm and LV–sGCmApoB (Fig. 1). A similar vector was generated with the enhanced GFP reporter gene in place of the glucocerebrosidase gene generating the vector LV–sGFPmApoB. LV vectors were generated from these constructs with the TAT-less system, as described previously (25).

Immunohistochemistry.

Approximately 1 × 109 infectious units of each virus, as determined by p24 ELISA (PerkinElmer Life Sciences, Boston, MA), were injected i.p. into mice. Fourteen days after vector delivery, mice were perfused transcardially via the left ventricle with ice-cold PBS, followed by 4% paraformaldehyde. Frozen brains were sectioned on a microtome at 40 μm along the sagittal plane, and sections were stored in tissue-collection buffer (2:2:1 glycerol/ethylene glycol/0.1 M phosphate buffer) at 4°C until staining. Sections from noninjected mice as well as the three injected groups of mice were stained with a mouse monoclonal anti-myc (Santa Cruz Biotechnology, Santa Cruz, CA) or rabbit polyclonal anti-myc antibody (Santa Cruz Biotechnology) to visualize the sGCm, sGCmApoB, or sGFPmApoB proteins and goat anti-LDLR (Santa Cruz Biotechnology). Cell-specific staining was performed with rabbit anti-calbindin (Chemicon, Temecula, CA), mouse anti-NeuN, guinea pig anti-GFAP (Advanced ImmunoChemical, White City, CA), or biotinylated tomato lectin (Vector Labs, Burlingame, CA). In addition, some sections were stained for the lysosomal membrane marker protein LAMP1 (Santa Cruz Biotechnology). These were visualized with Alexa Fluor secondary antibodies or streptavidin (Molecular Probes, Eugene, OR). All sections were counterstained with DAPI. CNS sections were viewed under a Leica confocal microscope (Leica, Deerfield, IL) and photographed.

The liver, lung, and spleen were collected from the perfused mice and embedded in OCT (Tissue Tek, Hatfield, PA). The organs were then sectioned on a cryostat at 20-μm thickness and stained with a rabbit anti-myc antibody (Santa Cruz Biotechnology) to visualize recombinant protein. Sections were counterstained with ToPro3 (Molecular Probes) to visualize nuclear DNA and then photographed on a Zeiss confocal microscope (Zeiss, Oberkochen, Germany).

Acknowledgments

We thank Dr. F. H. Gage (The Salk Institute for Biological Studies) for encouragement, help, discussions, and the generous gifts of reagents. I.M.V. is an American Cancer Society Professor of Molecular Biology and is supported, in part, by grants from the National Institutes of Health and the H.N. and Frances C. Berger Foundation. B.J.S. was a George E. Hewitt Foundation for Medical Research fellow and was also supported by a National Institutes of Health Training Grant.

Abbreviations

ApoB
apolipoprotein B
ApoE
apolipoprotein E
BBB
blood–brain barrier
GFAP
glial fibrillary acidic protein
LAMP
lysosomal-associated membrane protein
LDL
low-density lipoprotein
LDLR
LDL receptor
LV
lentivirus
sGC
secreted glucocerebrosidase
sGFP
secreted GFP
sGFPmApoB
sGFP myc epitope-tagged ApoB
sGCmApoB
sGC myc epitope-tagged ApoB.

Footnotes

The authors declare no conflict of interest.

See Commentary on page 7315.

References

1. Blomer U, Ganser A, Scherr M. Adv Exp Med Biol. 2002;513:431–451. [PubMed]
2. Suhr ST, Gage FH. Arch Neurol. 1999;56:287–292. [PubMed]
3. Bickel U, Yoshikawa T, Pardridge WM. Adv Drug Deliv Rev. 2001;46:247–279. [PubMed]
4. Schlageter KE, Molnar P, Lapin GD, Groothuis DR. Microvasc Res. 1999;58:312–328. [PubMed]
5. Rubin LL, Staddon JM. Annu Rev Neurosci. 1999;22:11–28. [PubMed]
6. Pardridge WM. Mol Biotechnol. 2005;30:57–70. [PubMed]
7. Stefansson S, Chappell DA, Argraves KM, Strickland DK, Argraves WS. J Biol Chem. 1995;270:19417–19421. [PubMed]
8. Brown MS, Goldstein JL. Science. 1986;232:34–47. [PubMed]
9. Boren J, Lee I, Zhu W, Arnold K, Taylor S, Innerarity TL. J Clin Invest. 1998;101:1084–1093. [PMC free article] [PubMed]
10. Hussain MM, Strickland DK, Bakillah A. Annu Rev Nutr. 1999;19:141–172. [PubMed]
11. Pfeifer A, Kessler T, Yang M, Baranov E, Kootstra N, Cheresh DA, Hoffman RM, Verma IM. Mol Ther. 2001;3:319–322. [PubMed]
12. Urayama A, Grubb JH, Sly WS, Banks WA. Proc Natl Acad Sci USA. 2004;101:12658–12663. [PMC free article] [PubMed]
13. Batini C. Arch Ital Biol. 1990;128:127–149. [PubMed]
14. Hof PR, Glezer II, Conde F, Flagg RA, Rubin MB, Nimchinsky EA, Vogt Weisenhorn DM. J Chem Neuroanat. 1999;16:77–116. [PubMed]
15. Wolf HK, Buslei R, Schmidt-Kastner R, Schmidt-Kastner PK, Pietsch T, Wiestler OD, Blumcke I. J Histochem Cytochem. 1996;44:1167–1171. [PubMed]
16. Mullen RJ, Buck CR, Smith AM. Development (Cambridge, UK) 1992;116:201–211. [PubMed]
17. Pitas RE, Boyles JK, Lee SH, Hui D, Weisgraber KH. J Biol Chem. 1987;262:14352–14360. [PubMed]
18. Weisgraber KH, Innerarity TL, Harder KJ, Mahley RW, Milne RW, Marcel YL, Sparrow JT. J Biol Chem. 1983;258:12348–12354. [PubMed]
19. Raffai RL, Dong LM, Farese RV, Jr, Weisgraber KH. Proc Natl Acad Sci USA. 2001;98:11587–11591. [PMC free article] [PubMed]
20. Nixon RA, Cataldo AM. Ann NY Acad Sci. 1993;679:87–109. [PubMed]
21. Dingwall C. J Clin Invest. 2001;108:1243–1246. [PMC free article] [PubMed]
22. Friden PM, Walus LR, Watson P, Doctrow SR, Kozarich JW, Backman C, Bergman H, Hoffer B, Bloom F, Granholm AC. Science. 1993;259:373–377. [PubMed]
23. Shi N, Pardridge WM. Proc Natl Acad Sci USA. 2000;97:7567–7572. [PMC free article] [PubMed]
24. Zhang Y, Pardridge WM. J Pharmacol Exp Ther. 2005;313:1075–1081. [PubMed]
25. Naldini L, Blomer U, Gage FH, Trono D, Verma IM. Proc Natl Acad Sci USA. 1996;93:11382–11388. [PMC free article] [PubMed]

Articles from Proceedings of the National Academy of Sciences of the United States of America are provided here courtesy of National Academy of Sciences
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Links

  • MedGen
    MedGen
    Related information in MedGen
  • PubMed
    PubMed
    PubMed citations for these articles
  • Substance
    Substance
    PubChem Substance links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...