• 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 25, 1999; 96(11): 5973–5977.
Colloquium Paper

Use of plant roots for phytoremediation and molecular farming


Alternative agriculture, which expands the uses of plants well beyond food and fiber, is beginning to change plant biology. Two plant-based biotechnologies were recently developed that take advantage of the ability of plant roots to absorb or secrete various substances. They are (i) phytoextraction, the use of plants to remove pollutants from the environment and (ii) rhizosecretion, a subset of molecular farming, designed to produce and secrete valuable natural products and recombinant proteins from roots. Here we discuss recent advances in these technologies and assess their potential in soil remediation, drug discovery, and molecular farming.

Biotechnology is transforming world agriculture, adding new traits to crop plants at a greatly accelerated rate. Plants are becoming more efficient producers of food, fiber, medicines, and construction materials. In addition to these conventional uses, biotechnology opens doors to unique uses of plants that are gaining greater acceptance from the public and attention from the scientific community. These so-called “value-added” uses include phytoremediation, the use of plants to remove pollutants from the environment or to render them harmless (1), and molecular farming (phytomanufacturing), the use of plants for the production of valuable organic molecules and recombinant proteins (2, 3). Because of the growing number of commercially successful applications and the lack of serious environmental concerns, both technologies are gaining acceptance from the scientific community, the general public, and regulators.

With the exception of root crops, plant roots are less utilized and studied than shoots. However, this situation may be changing because of the emerging biotechnologies described below that exploit the ability of plants to transport valuable molecules into and out of their roots. These root-based technologies include metal phytoextraction, a subset of phytoremediation, which uses plants to remove toxic heavy metals from soil; and rhizosecretion, a subset of molecular farming, which relies on the ability of plant roots to exude valuable compounds. Both technologies exploit plants’ innate biological mechanisms for human benefit.


Giant underground networks formed by the roots of living plants function as solar-driven pumps that extract and concentrate essential elements and compounds from soil and water. Absorbed substances are used to support reproductive function and carbon fixation within shoots. Metal phytoextraction relies on metal-accumulating plants to transport and concentrate polluting metals, such as lead, uranium, and cadmium, from the soil into the harvestable aboveground shoots (1, 4, 5). Hydroponically grown plant roots can also directly absorb, precipitate, and concentrate toxic metals from polluted effluents in a process termed rhizofiltration (6). Chelate-assisted phytoextraction (1) has been successfully used to remove lead from contaminated soils using specially selected varieties of Indian mustard (Brassica juncea L.). These varieties combine high shoot biomass with the enhanced ability of roots to adsorb EDTA-chelated lead from soil solution and transport it into the shoots. The transpiration stream is likely to be the main carrier of soluble chelated metal to the shoots, where water is transpired while metal accumulates (5). Chelate-assisted phytoextraction was also successfully used to phytoextract uranium (7).

One strategy for increasing the efficiency of phytoextraction is to increase metal translocation to the shoot by increasing plant transpiration. Earlier research showed that wind enhances metal flux to the shoots, while compounds that block transpiration (i.e., abscisic acid) block metal accumulation in the shoots (8). Spontaneous or chemically induced mutants with increased stomatal transpiration were isolated from various plant species, including tomato (9), Arabidopsis (10), and barley (11). To determine whether genetically increased transpiration would increase the efficiency of phytoextraction, (M1) seeds of B. juncea were mutagenized with ethyl methanesulfonate (EMS), and mature plants were self-pollinated to obtain M2 seeds.

Ten- to fourteen-day-old M2 seedlings were screened by excising a middle leaf from each plant, laying it flat in a well-aerated room, and visually assessing the degree of tissue dehydration after 1 or 2 hours. Plants whose leaves wilted (lost water) faster than others were saved and rescreened later in hydroponics and in soil for increased transpiration to confirm the results of the initial screen. After screening 20,000 M2 seedlings, 47 plants with significantly increased leaf transpiration rates were identified. Line M-30, in which the transpiration rate exceeded that of the wild-type plants by 130% in soil and by 75% in hydroponics, was tested for its phytoextraction performance in lead-contaminated soil amended with 2.5 mmol of EDTA per kg of soil. This high-transpiration line phytoextracted 104% more lead than the wild-type B. juncea, making it a good candidate for field optimization and use.

Increased resistance to metal is another important trait that can improve the efficiency of phytoextraction. Varieties of B. juncea with greater metal tolerance should grow better in metal-contaminated sites and survive longer after metal uptake is induced by chelate application to the soil. Substantial research has been directed toward isolating genes that are involved in metal biology, e.g., metallothioneins or transporters. Interestingly, some increases in cadmium tolerance were observed in transgenic plants overexpressing the human metallothionein-II gene (12).

Valuable metal-resistance traits can be found in metal hyperaccumulating plants that are endemic to soils naturally enriched with heavy metals. These plants can accumulate exceedingly high amounts of essential and nonessential heavy metals in their foliage, to levels that are highly toxic to most other plants (13). For example, several Thlaspi species can accumulate Ni and Zn, to 1–5% of its dry biomass. This is an order of magnitude greater than concentrations of these metals in the nonaccumulating plants growing nearby. The prevention of herbivory and disease is thought to be the main function of this unique phenomenon (14, 15). It recently has been established that the ability of T. goesingense Halacsy to hyperaccumulate metals is the result of high resistance to the metals rather than the greater rates of metal uptake (16). Unfortunately, most hyperaccumulating species are not suitable for phytoextraction for several reasons: (i) metals that are primarily accumulated (Ni, Zn, and Cu) are not among the most important environmental pollutants; (ii) most have very low biomass and capricious growth habits unsuitable for monoculture; and (iii) agronomic practices and crop protection measures for their cultivation have not been developed. However, many metal-hyperaccumulating species belong to Brassicaceae (mustard) family, and thus are related to B. juncea, the preferred plant for phytoextraction of lead. Unfortunately, B. juncea, while exhibiting a high capacity for metal uptake and translocation, is not very resistant to high levels of lead or other heavy metals in its foliage. Therefore, chelate-assisted phytoextraction is very toxic to B. juncea, requiring harvesting several days after chelate application.

Unfortunately, no genes conferring metal resistance were identified in any of the hyperaccumulating species, precluding the possibility of direct gene transfer. Thus, an attempt was made to introduce metal resistant traits into the high-biomass Pb accumulator B. juncea using somatic hybridization. Thlaspi caerulescens, a known Ni and Zn hyperaccumulator, was selected as one of the parents for both symmetric and asymmetric hybrids in which T. caerulescens protoplasts were irradiated with x rays before fusion. Eighteen hybrids were regenerated, all showing a phenotype intermediate between those of the parents. Two asymmetric hybrids were found to be fertile. One of these hybrids (60/31) had vigorous growth, characteristic of B. juncea, and contained Thlaspi-specific repetitive DNA sequences, as demonstrated by Southern hybridization. (As expected, total DNA from B. juncea parent did not hybridize with Thlaspi-specific probes). Hybrid 60/31 displayed dramatically increased resistance when germinated and grown in Pb-, Ni-, and Zn-contaminated soil (Fig. (Fig.1).1). The amount of Pb that the hybrid was able to phytoextract on a dry weight basis was similar to that of both parents. However, the total amount of Pb phytoextracted by each hybrid plant was much greater because of the greater biomass produced on the contaminated soil. Interestingly, the growth habits and biomass of B. juncea and the 60/31 hybrid did not differ much when the plants were grown in noncontaminated fertile soil (data not shown).

Figure 1
Asymmetric somatic hybrid 60/31 (B) and its parents Brassica juncea (A) and Thlaspi caerulescens (C) growing in soil containing 800 mg/kg lead, 328 mg/kg nickel, and 7,600 mg/kg zinc.


Phytoextraction exploits the ability of plant roots to remove unwanted contaminants from their environment. But could the reverse of this process also be exploited? Could roots make valuable compounds and deliver them into their environment? At present, most of the recombinant proteins or valuable natural products used as fine chemicals, pharmaceuticals, crop protection compounds, cosmetic ingredients, etc. are extracted from plants by using solvents. This method requires expensive purification of the active ingredients from complex mixtures of organic molecules and proteins, making downstream processing and purification of individual components difficult and costly. Extracting plants is also a “batch” process whereby the plant is harvested, and its continual ability to synthesize chemicals is not utilized. Natural rubber and maple syrup are rare examples of continuous manufacturing processes, which produce much larger amounts of valuable plant product over the lifetime of the plant.

Rhizosecretion of Natural Products.

In addition to accumulating biologically active chemicals, plant roots continuously produce and secrete compounds into their immediate environment (rhizosphere). While up to 10% of photosynthetically fixed carbon is secreted from the roots (17, 18), the systematic study of chemical composition of root exudates from diverse plant species has not been undertaken. Not surprisingly, few compounds that were identified in root exudates were shown to play an important role in several biological processes. For example, isoflavonoids and flavonoids present in the root exudates of a variety of legume plants activate the Rhizobium genes responsible for the nodulation process (19, 20) and, possibly, for vesicular–arbuscular mycorrhiza (VAM) colonization (21). Strigol, a germination stimulant for the parasitic plant Striga asiatica, has been found in the root exudates of many cereals (22). A variety of plants produce herbicidal allelochemicals that may inhibit growth and germination of neighboring plants (2325). In addition, root-secreted compounds called phytosiderophores may be involved in the acquisition of essential plant nutrients from soils (2628) and in defense against toxic metals such as aluminum (29).

Intuition and limited published data (30) suggest that root-secreted compounds should have a wide spectrum of biological activities including protection against biotic and abiotic stresses. Survival of delicate and physically unprotected root cells may depend on their continuous “underground chemical warfare” against a hostile and constantly changing environment teeming with bacteria and fungi preying on any organic material in soil. The unexplored chemical diversity of root exudates is an obvious place to search for novel biologically active compounds including antimicrobials. Our biochemical analysis of root exudates from 120 plant species can be summarized as follows: (i) each plant species studied exuded a distinct set of compounds, which is a unique biochemical fingerprint for a given species (Fig. (Fig.2 2 AC); (ii) root exudates are relatively simple mixtures, in comparison to solvent extracts of plant tissue, making the isolation of the active molecules an easier task; (iii) root exudates are devoid of pigments and tannins, known to interfere in activity screens, and do not contain large quantities of biologically inert structural compounds; and (iv) the chemical composition of root exudates is very different from that of conventional methanolic extracts of root tissue.

Figure 2
HPLC profiles of nonelicited root exudates of three plant species collected in distilled water (AC) and root exudates of Brassica juncea collected in distilled water (D) or in distilled water supplemented with 1 mM AgNO3 (E) or 500 mM H ...

We have also observed that exudate chemical diversity can be greatly increased by the elicitation process, which is known to alter secondary metabolism in plants exposed to various physical and chemical treatments. Phytoalexins, antimicrobial compounds produced in plants and tissue cultures in response to disease causing agents or their chemical components, are probably the best studied elicited defense compounds in plants (31). Unfortunately, little is known about elicited compounds in root exudates, with the exception of a recent report on isoflavonoid exudation from the roots of white lupine (30). We observed that chemical or physical elicitors stimulate roots of various plants to exude an array of compounds not detected in the “nonelicited” exudates (Fig. (Fig.22 DF). On the other hand, the same elicitor will trigger the production of different compounds in different plant species. In addition, elicitation may dramatically increase the quantities of certain compounds in the exudates. It can be hypothesized that elicitors mimic the effects of stresses on the hydroponically grown roots, activating biochemical defense systems and resulting in quantitative and qualitative changes in the composition of the exudates.

To demonstrate the presence of antimicrobial compounds in root exudates, a screening protocol was designed in which 10 μl of concentrated exudate solution was transferred into a small cavity in agar poured into 24-well microtiter plates. The tested microorganisms were plated in each well before the cavity was made. Exudates from 480 species, each treated with 2–4 elicitors, were tested in this system for the inhibition of growth of selected bacteria and fungi (Fig. (Fig.3).3). The following percentage of exudates showed moderate to strong activity against tested microorganisms: Escherichia coli (3.4%), Staphylococcus aureus (4.3%), Pseudomonas aeruginosa (0.4%), Penicillium notatum (0.8%), and Saccharomyces cerevisiae (0.6%).

Figure 3
Antimicrobial activity of root exudates. The exudates showing activity (indicated with red arrows) against Staphylococcus aureus ssp. aureus (A) were from Tagetes minuta (column 1, Asteraceae) and Eriastrum densiflorum var. austromontana (column ...

In addition to exudates, hydroponically cultivated plant roots also provide a unique source of biologically active compounds. We have also observed that elicitation, both quantitatively and qualitatively, alters the HPLC profiles of secondary metabolites in roots of many plant species (data not shown). Most likely, these changes are subsequently reflected by the dramatic alterations in the rhizosecreted compounds.

Why Root Exudates?

The above observations suggest that root exudates represent a new and functionally enriched source of biologically active compounds. Elicitation of hydroponically grown roots adds another unexplored dimension to the chemical diversity normally hidden in silent parts of the plant genomes. In addition to shedding light on dark corners of plant biology, the systematic study of root exudates may be valuable to the global pharmaceutical industry, which still heavily relies on novel sources of chemical diversity to discover new drugs in an ever-accelerating race against time. Twenty five percent of all prescriptions dispensed from pharmacies in the United States contain active ingredients extracted from higher plants (32). However, methods of harvesting chemical diversity of plant-derived compounds often follows hunter–gatherer strategies. Extracts of plant material haphazardly collected in various places around the world are eventually acquired by pharmaceutical companies, which put them through sophisticated high-throughput screens that use an increasing array of molecular targets. This primitive prospecting process does not provide a reliable and reproducible source of natural products that can be easily resupplied after a novel activity is found. The mismatch between the beginning of the drug development pipeline and what follows creates an opportunity for developing new pharmaceutical agents from plants using more standardized, scientific approaches that favor biologically active molecules over structural components and major metabolites. Tissue culture-based production of natural products, often combined with elicitation, is one of the recently developed strategies for “increasing the size of the needle in the haystack.” However, plant tissue cultures are expensive, slow growing, and relatively deficient of secondary metabolites, presumably because of their nondifferentiated nature. Rhizosecretion, on the other hand, may produce a more cost-effective and diverse source of chemical compound mixtures for the identification of novel biologically active compounds. In addition, rhizosecretion, a nondestructive and continuous process, may provide a constant supply of these compounds over the lifetime of a plant.

Rhizosecretion of Recombinant Proteins.

The ease of transformation and cultivation make plants suitable for manufacturing many recombinant proteins. Indeed, numerous heterologous (recombinant) proteins have been produced in plant leaves, fruits, roots, tubers, and seeds (3335), and are targeted to different subcellular compartments, such as the cytoplasm, endoplasmic reticulum (ER), or apoplastic space (36). Plants are capable of carrying out acetylation, phosphorylation, and glycosylation as well as other posttranslational protein modifications required for the biological activity of many eukaryotic proteins. However, the extraction and purification of proteins from biochemically complex plant tissues is a laborious and expensive process that presents a major obstacle to large-scale protein manufacturing in plants. In attempts to overcome this problem, secretion-based systems utilizing transgenic plant cells or plant organs aseptically cultivated in vitro have been investigated (3739). However, these in vitro systems, which include hairy roots, may be expensive, slow-growing, unstable, and relatively low-yielding. Until now, these disadvantages precluded the use of in vitro plant systems for the commercial manufacturing of recombinant proteins.

Can rhizosecretion be used for the continuous manufacturing of recombinant proteins? The nondestructive rhizosecretion process may provide high yields of recombinant proteins over the lifetime of a plant and facilitate their downstream purification, combining the advantages of the whole plant and in vitro protein expression systems. Indeed, roots of living plants are known to secrete proteins. For example, large amounts of acid phosphatase are released from the roots of many plants during phosphate deficiency (40). We attempted to “rhizosecrete” the following three heterologous proteins of different origins from Nicotiana tabacum L.; green fluorescent protein (GFP) of the jellyfish Aequorea victoria, human placental secreted alkaline phosphatase (SEAP), and xylanase from the thermophylic bacterium Clostridium thermocellum.

All three of these proteins were rhizosecreted from transgenic plants when their expression was controlled by a strong root-expressed promoter and targeted by a secretory signal peptide (Fig. (Fig.4).4). Daily rhizosecretion of GFP, released into fresh medium unprotected from proteolysis, reached 2 μg/g root dry weight, while SEAP rhizosecretion, quantified from its activity, reached 20 μg/g root dry weight, a significant amount considering that no attempts to optimize rhizosecretion had been made thus far. It is likely that methods for increasing protein expression and secretion will be developed along with plant varieties optimized for the rhizosecretion of recombinant proteins.

Figure 4
Rhizosecretion of jellyfish green fluorescent protein (GFP)(A), human placental alkaline phosphatase (SEAP)(B), and bacterial (Clostridium thermocellum) xylanase (C) from the roots of transgenic Nicotiana tabacum L. (A) To direct GFP into the secretory ...

Data suggest that plant roots can continuously produce and secrete biologically active recombinant proteins of different origins. The rhizosecretion system offers a simplified method for the isolation of recombinant proteins from simple hydroponic medium rather than from complex plant extracts. As with rhizosecretion of natural products, protein rhizosecretion can be operated continuously without destroying the plant, thus producing a higher total yield of the recombinant protein over the life of the transgenic plant. In addition, recombinant biopharmaceutical proteins purified from root exudates are less likely to be contaminated with pathogenic viruses that may be present in the milk or urine of transgenic animals. Rhizosecretion also borrows from many well developed and tested methods of commercial hydroponic plant cultivation, and therefore, will be relatively easy to scale up.


While the evolution of plant shoots followed primarily “introverted” paths by perfecting physical barriers between themselves and the environment, roots had to be more “extroverted” in their relationship with soil. This requirement created a unique set of biological mechanisms, which until recently, were understudied and underutilized. Phytoextraction and rhizosecretion are starting to change this, while allowing scientists to take a radically new look at the darkest corners of plant biology. These technologies also open the doors to the value-added, nonagricultural uses of plants, which will continue to expand in the new century.

Neither phytoextraction nor rhizosecretion will directly contribute to feeding world population in the next century. However, these technologies will improve the quality of life for many people if their development continues. The future challenge for metal phytoextraction is to further reduce the cost and increase the spectrum of metals amenable to this technology. This goal can be achieved by creating superior plant varieties for phytoextraction by using genetic engineering to introduce valuable traits into plants, developing better agronomic protocols for their cultivation, and designing safer and more effective soil amendments. A recent, and probably the only, example of the successful use of genetic engineering applied to metal phytoremediation is the use of bacterial mercuric reductase (merA) gene to achieve mercuric ion reduction in transgenic Arabidopsis (41) and yellow poplar plants (42). Elemental mercury produced in transgenic plants is much less toxic than ionic mercury and can be volatilized from transgenic plants in a process termed phytovolatilization, which is related to phytoextraction.

The future challenge for rhizosecretion lies in the successful development of effective and safe pharmaceuticals from the collection of biologically active lead molecules secreted by the roots, and in large-scale, cost-effective manufacturing of recombinant proteins. The aging population and ever-growing demand for better pharmaceuticals should foster the use green plants as sources of new drug discovery, biotransformation, and in some cases, manufacturing. Thus, more effective utilization of immense biosynthetic capacity of plants based on their inexpensive and renewable nature will present major opportunities for plant researchers in the next century.


1. Salt D E, Smith R D, Raskin I. Annu Rev Plant Physiol Plant Mol Biol. 1998;49:643–668. [PubMed]
2. Nawrath C, Poirier Y, Somerville C. Mol Breeding. 1995;1:105–122.
3. Franken E, Teuschel U, Hain R. Curr Opin Biotechnol. 1997;8:411–416. [PubMed]
4. Kumar P B A N, Dushenkov V, Motto H, Raskin I. Environ Sci Technol. 1995;29:1232–1238. [PubMed]
5. Vassil A D, Kapulnik Y, Raskin I, Salt D E. Plant Physiol. 1998;117:447–453. [PMC free article] [PubMed]
6. Dushenkov V, Kumar P B A N, Motto H, Raskin I. Env Sci Technol. 1995;29:1239–1245. [PubMed]
7. Huang J W, Blaylock M J, Kapulnik Y, Ensley B D. Environ Sci Technol. 1998;32:2004–2008.
8. Salt D E, Prince R C, Pickering I J, Raskin I. Plant Physiol. 1995;109:1427–1433. [PMC free article] [PubMed]
9. Tal M. Plant Physiol. 1966;41:1387–1391. [PMC free article] [PubMed]
10. Koornneef M, Reuling G, Karssen C M. Physiol Plant. 1984;61:377–383.
11. Raskin I, Ladyman J A R. Planta. 1988;173:73–78. [PubMed]
12. Misra S, Gedamu L. Theor Appl Genet. 1989;78:161–168. [PubMed]
13. Baker A J M, Brooks R R. Biorecovery. 1989;1:81–126.
14. Boyd R S, Martens S N. Oikos. 1994;70:21–25.
15. Boyd R S, Shaw J J, Martens S N. Am J Bot. 1994;81:294–300.
16. Kramer U, Smith R D, Wenzel W W, Raskin I, Salt D E. Plant Physiol. 1997;115:1641–1650. [PMC free article] [PubMed]
17. Johansson G. Soil Biol Biochem. 1992;24:427–433.
18. Shepherd T, Davies H V. Ann Bot. 1993;72:155–163.
19. Peters N K, Long S R. Plant Physiol. 1988;88:396–400. [PMC free article] [PubMed]
20. Maxwell C A, Phillips D A. Plant Physiol. 1990;93:1552–1558. [PMC free article] [PubMed]
21. Tsai S M, Phillips D A. Appl Environ Microbiol. 1991;57:1485–1488. [PMC free article] [PubMed]
22. Siame B A, Weerasuriya Y, Wood K, Ejeta G, Butler L G J. Agric Food Chem. 1993;41:1486–1491.
23. Friebe A, Schulz M, Kuck P, Schnabel H. Phytochemistry. 1995;38:1157–1159.
24. Yu J Q, Matsui Y J. Chem Ecol. 1994;20:21–31. [PubMed]
25. Inoue M, Nishimura H, Li H H, Mizutani J. J Chem Ecol. 1992;18:1833–1840. [PubMed]
26. Miyasaka S C, Buta J G, Howell R K, Foy C D. Plant Physiol. 1991;96:737–743. [PMC free article] [PubMed]
27. Lipton D S, Blanchar R W, Blevins D G. Plant Physiol. 1987;85:315–317. [PMC free article] [PubMed]
28. Mori S, Nishizawa N, Kawai S, Sato Y, Takagi S J. Plant Nutr. 1987;10:1003–1011.
29. de la Fuente J B M, Ramirez-Rodriguez V, Cabrera-Ponce J B L, Herrera- Estrella L. Science. 1997;276:1566–1568. [PubMed]
30. Gagnon H, Ibrahim R K. Phytochemistry. 1997;44:1463–1467.
31. Dixon R A. Biol Rev. 1986;61:239–291.
32. Farnsworth N R, Morris R W. Am J Pharm. 1976;148:46–52. [PubMed]
33. McGarvey, P. B., Hammond, J., Dienelt, M. M., Hooper, D. C., Fu, Z. F., Dietzschold, B., Koprowski, H. & Michaels, F. H. (1995) Bio/Technology 1484–1487. [PubMed]
34. Van Engelen F A, Schouten A, Molthoff J W, Roosien J, Salinas J, Dirkse W G, Schots A, Bakker J, Gommers F J, Jongsma M A, et al. Plant Mol Biol. 1994;26:1701–1710. [PubMed]
35. Sonnewald U, Hajirezaei M-R, Kossmann J, Heyer A, Trethewey R N, Willmitzer L. Nat Biotechnol. 1997;15:794–797. [PubMed]
36. Conrad U, Fiedler U. Plant Mol Biol. 1998;38:101–109. [PubMed]
37. Li J, Hegeman E, Hanlon R W, Lacy G H, Cenbow D M, Grabau E A. Plant Physiol. 1997;114:1103–1111. [PMC free article] [PubMed]
38. Firek S, Draper J, Owen M R L, Gandecha A, Cockburn B, Whitelam G C. Plant Mol Biol. 1993;23:861–870. [PubMed]
39. Wongsamuth R, Doran P M. Biotechnol Bioeng. 1997;54:401–415. [PubMed]
40. Li M, Tadano T. Soil Sci Plant Nutr. 1996;42:753–763.
41. Rugh C L, Wilde H D, Stack N M, Thompson D M, Summers A O, Meagher R B. Proc Natl Acad Sci USA. 1996;93:3182–3187. [PMC free article] [PubMed]
42. Rugh C L, Senecoff J F, Meagher R B, Merkle S A. Nat Biotechnol. 1998;16:925–928. [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


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...