The murine model is widely used to study factors involved in the etiology of mammary carcinoma. It became evident as early as 1936 ( 1) that a filterable agent, i.e. , a virus, is involved in at least some mouse strains in the causation of mammary cancer. Over the past four decades, experimental systems have been developed in numerous mouse strains, and in almost every strain studied, a mouse mammary tumor virus (MMTV) has been revealed ( 2, 3). Table 1 surveys the variations in incidence of spontaneously occurring mammary tumors, latent period to tumor, types of tumors produced, and whether mouse mammary tumor virions are observed in mammary tumors by electron microscopy. It should be noted that mammary tumors appear "early" (before one year) in the high incidence mouse strains C3H, RIII, and GR, and "late" (after one year) in low and moderate incidence strains C3HfC57BL and BALB/c.
A question that has remained unanswered concerning the origin of mammary oncogenesis in the mouse is: How many mouse mammary tumor viruses are there? There are at least two considerations involved in this question: (a) each mouse strain contains its own mouse mammary tumor virus or viruses, which are partially related or distinct from other viruses of other mouse strains, and (b) there is only one MMTV and each mouse strain exerts its own control over various properties of this virus, such as its expression or its virulence. The answer to this question is essential to our understanding the etiology of the disease in each system studied.
A recent achievement has been the ability to productively infect heterologous cells with various MMTV isolates ( 4, 5). The MMTVs grown in feline cells share nucleic acid and immunologic properties with their mouse grown counterparts. These reagents have now made possible the delineation between viral-coded vs. host-derived entities. In the nucleic acid hybridization and immunologic studies presented here, both murine and feline grown MMTVs were employed with identical results, thus demonstrating the viral-coded nature of the nucleic acids and proteins being studied.
The major surface component of the MMTV virion is a 52,000d glycoprotein (gp52). The other major protein components of the MMTV virion are a 36,000d glycoprotein (gp36), and the 28,000 (p28), 14,000 (p14), and 10,000 (p10) dalton polypeptides. With the development of sensitive radioimmunoassays for the whole MMTV virion ( 6), or for purified MMTV polypeptides ( 7, 8, 9, 10), it has become possible to precisely analyze similarities or differences among MMTVs from different mouse strains. Several investigators have previously shown that the gp52 of MMTVs share group-specific antigenic determinants ( 7, 8, 9, 10). Experiments have recently been described ( 11, 12) that also demonstrate type-specific antigenic determinants on the gp52 molecule.
MMTVs from RIII and C3H mice, i.e., MMTV(RIII) and MMTV(C3H), were used in a competitive radioimmunoassay (RIA) to compete for the binding of anti-MMTV(C3H) to [125I]-labeled MMTV(C3H) virions. In this system, increasing amounts of MMTV(RIII) competitor gave a shallower slope than that given by the isologous MMTV(C3H) ( Figure 1A). At the highest input of competing MMTV(RIII) protein employed, i.e., 100 micrograms, only 75% inhibition was obtained while less than 1 microgram of MMTV(C3H) resulted in the same competition. In a "group-specific" assay using anti-MMTV(C3H) vs. [125I]MMTV(RIII), both viruses competed identically, i.e., with comparable inputs and with the same slope.
The addition of increasing amounts of MMTV(GR) into the anti-MMTV(C3H) vs. [125I]MMTV(C3H) system did not cause complete inhibition of the precipitation of [125I]MMTV(C3H) ( Figure 1B). Even at high inputs of protein, MMTV (GR) was incapable of competing for all the antibodies binding to MMTV(C3H). The anti-MMTV(C3H) serum, therefore, appears to contain an antibody population that is directed toward antigenic determinants that are present in MMTV(C3H) but not in MMTV(GR). To further amplify the type-specific reactions observed, anti-MMTV(C3H) serum was absorbed with MMTV(GR) and the immune precipitate was removed by centrifugation. The resulting absorbed serum retained its ability to bind [125I]MMTV(C3H). This binding could be completely inhibited by the addition of MMTV(C3H) or MMTV(C3H)gp52, but was not inhibited by the addition of up to 10,000 ng MMTV(GR) as competitor ( Figure 1C). Additional type-specific reactivities among the various MMTVs also exist. These include differences between MMTV(C3H) and the endogenous MMTV of C3H obtained from C3HfC57BL mice ( 11, 12).
The type and group specificities of MMTVs grown in feline cells ( 5) were indistinguishable from the reactivities observed with murine grown MMTVs, thus providing strong evidence that the MMTV gp52 antigens are viral coded. The analysis of feline grown MMTV further excludes the possibilities that the observed antigenic differences were due to either differences in murine antigenic determinants of the different mouse strains producing the virus, or to host-coded differences in glycosylation of virions.
The identification of the type-specific differences for different MMTVs is now being used in several laboratories to monitor the host's immune response to mammary tumorigenesis, as well as in studies seeking trans-species reactivities with MMTVs. These studies further delineate the molecular diversity of viruses that can be involved in the etiology of mammary carcinoma within a given species.
We have recently developed a competition RIA for the 28,000d major internal protein (p28) of MMTV ( 13). When this assay was conducted with high antibody dilutions for maximum sensitivity, no differences were observed among MMTVs from RIII, GR, C3H, or C3HfC57BL mice or with mammary tumor extracts from those mice. To demonstrate type-specific reactivities associated with the MMTV p28 polypeptide, assay conditions of low antibody dilution were used.
The binding of antisera prepared against the p28 of MMTV from RIII mice to [125I]MMTV(RIII) p28 could be completely inhibited by the addition of 1 microgram of purified MMTV(RIII) p28 ( Figure 2A). The addition of increasing amounts of MMTV(C3H) p28 also competed for this binding, but with a shallower slope characteristic of a cross-reacting ( 14) but not identical antigen ( Figure 2A). Changing the radioactive antigen from [125I]MMTV(RIII) p28 to [125I]MMTV(C3H) p28 and maintaining the same antibody dilution revealed that both MMTV(RIII) p28 and MMTV(C3H) p28 competed identically in the radioimmunoassay ( Figure 2B). Thus, the MMTV p28 appears to contain both indistinguishable, i.e., group-specific, antigenic determinants as well as distinguishable, i.e., type-specific, antigenic determinants ( 13).
Both type-specific and group-specific reactivities were retained when MMTVs were used that were grown in the same feline CrFK cell line ( Figure 2C). All three of the MMTVs grown in feline cells, however, competed identically in the anti-MMTV(RIII) p28 vs. [125I]MMTV(C3h) 'group specific" p28 radioimmunoassay ( Figure 2D).
Using the combination of gp52 and p28 RIAs, levels of MMTV gene expression can now be analyzed in terms of both viral glycoproteins and nonglycoproteins. This type of comparison is important in light of recent evidence of noncoordinate polypeptide chain initiation of glycosylated vs. nonglycosylated MMTV proteins in infected cells ( 15). These combined radioimmunoassays should further elucidate the mechanisms of MMTV replication and gene expression in murine mammary tumors.
Tryptic peptide analyses ( 16) have been performed on the 52,000 and 36,000 dalton glycoproteins and the nonglycosylated 28,000, 14,000 and 10,000 dalton proteins of the highly oncogenic MMTVs of C3H, RIII, and GR mice ( 17). Each virus was grown in both murine and feline cells to ensure the viral-coded nature of each peptide analyzed. The gp36-38 and p14 peptide maps of all three MMTVs were indistinguishable. Both the gp52 ( Figure 3) and the p28 ( Figure 4) of MMTV(C3H), however, could be clearly distinguished from the corresponding proteins of MMTV(RIII) and MMTV(GR), regardless of whether the viruses were grown in feline or murine cells. The p10 of MMTV(RIII), on the other hand, was clearly different from that of MMTV(C3H) and MMTV(GR) ( 17). Therefore, tryptic peptide analysis of three MMTV proteins, gp52, p28, and p10, can serve to distinguish these three viruses from one another. These studies represent further strain-specific markers for several MMTV gene products. Thus, as for type-C retroviruses of the mouse ( 16), MMTVs form a multigene family, the final extent of which is not yet known.
Studies employing both MMTV radioactive 60-70S RNA and MMTV cDNA probes have demonstrated that MMTV proviral sequences are present in the DNA of all strains of laboratory mice. Copy numbers are in the low repetitive range for normal tissues such as liver and are higher in mammary tumors ( 18, 19, 20, 21, 22, 23).
Biological studies have shown that MMTVs can be transmitted in different mouse strains either by the milk of via the germline ( 1, 2, 3). Occasionally, MMTV may also be transmitted by male seminal fluids to females, which in turn can transfer the virus to their progeny via the milk ( 2, 3). Other modes of transmission are, of course, possible. The question that arises is: Can one distinguish if an MMTV has been introduced into a given mouse via the germline (i.e., as a germinal provirus or virogene) or via some non-germline mechanism, such as via the placenta, milk, seminal fluid, or as a plasmid: The term "horizontal transmission" is not used here to due to the confusion that would arise from such modes of viral transmission as via the placenta or as a plasmid in a germ cell. If an MMTV was introduced into a mouse via the germline, one would expect to find MMTV proviral sequences equally distributed in the DNA of all tissues of that mouse. On the other hand, if an MMTV was introduced into a mouse by some other mechanism, one would expect to see an uneven distribution of MMTV sequences in the DNA of different tissues of that mouse. To address these points, we used the technique of molecular hybridization.
MMTV(C3H) was isolated from supernatant fluids of a C3H mammary tumor cell line. The 60-70S RNA from these virions was purified and iodinated to a specific activity of approximately 2 x 107 cpm per microgram as described previously ( 20). This RNA was then hybridized at various C0t values to DNA from C3H mammary tumor cells and DNA from an apparently normal C3H liver; hybridization to sheep DNA was used as a control. C0t is defined as the product of the DNA concentration in moles of deoxyribonucleotide per liter and time in seconds. As assayed by resistance to ribonuclease (RNase A and T1) digestion, hybridization to sheep and other nonrodent DNAs remained consistently at less than 6% up to a C0t of 35,000 and was thus scored as nonspecific background. Hybridization between the iodinated MMTV(C3H) 60-70S RNA and DNA from C3H mammary tumors was approximately 60% ( Figure 5A). This value was consistently higher than the maximum extent of hybridization between this MMTV(C3H) RNA and DNA from livers or other normal organs of C3H mice ( Figure 5A).
The C0t1/2 value was approximately 380 for the hybridization between MMTV(C3H) [125I]RNA and C3H mammary tumor DNA, and approximately 440 for C3H liver DNA. For comparison, poly A enriched C3H cellular RNA (selected by poly U-Sepharose chromatography), representing messenger RNA, was also iodinated and hybridized to C3H liver DNA. The C0t1/2 value obtained using this RNA was approximately 3,100, thus representing the value obtained with "unique" DNA. The results depicted in Figure 5A demonstrate that both the C3H mammary tumor cell line and C3H liver contain MMTV proviral sequences in the low repetitive range ( 20). The lower C0t1/2 value obtained with the C3H mammary tumor cell line DNA indicates that there are more MMTV proviral sequences in this DNA than in the DNA of the C3H liver. The differences in final percent hybridization, i.e., approximately 60% for the C3H tumor cell line DNA, and approximately 50% for the C3H liver DNA, however, may be indicative of one or a combination of two phenomena: (a) there are quantitatively more MMTV proviral sequences in the mammary tumor DNA than the liver DNA; and (b) the DNA of the C3H mammary tumor cells contain sequences of the MMTV genome that are not found in the DNA of C3H liver cells. To answer this question, recycling experiments were performed.
To determine if there are any MMTV sequences in mammary tumors that are not present in the DNA of an apparently normal organ, i.e., the liver of a C3H mouse, iodinated MMTV(C3H) 60-70S RNA was first hybridized to a vest excess of C3H liver DNA. Liver DNA was chosen because murine livers have been shown to be negative for most MMTV markers ( 2, 3). MMTV(C3H) 60-70S RNA (300,000 cpm) was first annealed to 30 mg of normal C3H liver DNA at 68°C to a C0t of 20,000 as described previously ( 20). The unhybridized single stranded [125I]RNA eluting from the hydroxyapatite column at 0.14 M sodium phosphate was termed "recycled RNA." This RNA was then concentrated and reannealed to C3H mammary tumor DNA and to C3H liver DNA to a C0t of 20,000. As can be seen in Figure 5B, the recycled MMTV(C3H) RNA failed to hybridize above background levels to the DNA of normal C3H livers. This demonstrates that the recycling procedure effectively removed all portions of the MMTV(C3H) [125I]RNA that are complementary to the DNA of normal C3H liver. The recycled [125I]RNA, however, hybridizes extensively with DNA from C3H mammary tumor cells ( Figure 5B), thus demonstrating the existence of MMTV sequences in C3H mammary tumor DNA that are absent in liver DNA of that same mouse ( 20).
Studies were conducted to determine the natural distribution of the recycled MMTV sequences in various murine DNAs. The recycled MMTV(C3H) [125I]RNA was first hybridized to DNA from RIII livers and RIII mammary tumors that arise early in life. Complementary sequences were found in the DNA of all "early" mammary tumors tested but not in the DNA of livers ( Table 2).
The GR strain of mice is of great interest as a result of genetic studies that have demonstrated the transmission of MMTV in this strain as a one-gene dominant characteristic ( 2). The recycled MMTV(C3H) [125i]RNA was hybridized to GR mammary tumor and liver DNAs. Complementary sequences are present in the DNAs of both GR mammary tumors and GR livers ( Table 2) as well as other GR organs tested. DNA of "late" occurring mammary tumors of the low and moderate incidence strains BALB/c and C3HfC57BL ( Table 1) were also analyzed for the presence of the recycled sequences in their DNA and were consistently found negative. Since the C3HfC57BL strain was originated by C3H mice foster nursed on C57BL mothers, a strain devoid of overt MMTV in its milk ( 2, 3), this is further evidence that the recycled sequences are part of the milk transmitted MMTV(C3H) and are not germline transmitted.
Livers of BALB/c, C57BL/6N, and C57BL/10SCN mice were shown to contain some MMTV proviral information ( 18, 19, 21, 22, 23); they do contain, however, sequences homologous to the MMTV(C3H) recycled RNA ( Table 2).
It thus appears that the virus or viruses responsible for causing early mammary tumors in C3H, RIII, and GR mice are easily distinguishable from the virus or viruses causing late mammary tumors in C3HfC57BL mice and BALB/c mice. Furthermore, a virus similar to the highly oncogenic non-germline transmitted C3H virus appears to be a germinal provirus in the DNA of the GR strain, the only strain in which genetic evidence has been presented for a one-gene dominant characteristic for MMTV. One possible explanation, therefore, is that a virus similar to the non-germline transmitted viruses of C3H and RIII has become integrated as an endogenous virus of GR.
In view of the varied distribution of MMTV proviral sequences in the mouse, we set out to determine if MMTV proviral sequences are present in other rodents. Furthermore, to determine if sequences related to, but not identical to, MMTV could be detected in the DNAs of other rodents, conditions of hybridization were relaxed by lowering the temperature at which hybridizations were carried out from 68°C to 54°C, and the assay of resulting RNA-DNA duplexes was accomplished by raising the salt concentration from the standard conditions of 2 x SSC (1 x SSC is 0.15 M NaCl and 0.015 M Na citrate) to 8 x SSC. Specificity is still maintained using relaxed conditions since no hybridization above 6% background is observed to DNA obtained from bovine and canine tissue or E. coli.
Molecular hybridization experiments were carried out to determine the presence of "MMTV-specific" sequences (using standard conditions) and "MMTV-related" sequences (using relaxed conditions of hybridization and an RNAse assay) in other Mus species. Results are given in Table 3 where values have been normalized by subtracting the 6% background hybridization level and using as 100% the hybridization observed between MMTV [125I]60-70S RNA and the homologous C3H mammary tumor cell line DNA. As can be seen in Table 3, when standard conditions of hybridization and RNase assay are used, the subspecies of Mus musculus, i.e., Mus musculus molossinus and Mus musculus Peru Atteck, appear to contain less MMTV sequences as endogenous provirus. When standard conditions of hybridization are used, however, no MMTV-specific sequences are detected in the DNAs of other Mus species, i.e., Mus cervicolor or Mus caroli ( 24).
When relaxed conditions of hybridization and RNAse assay are used, the degree of hybridization detected using DNAs of C3H or RIII Mus musculus tissues remains approximately the same as does the degree of hybridization to Mus musculus molossinus and Mus musculus Peru Atteck DNAs. Using these relaxed conditions, however, MMTV-related sequences are now detected in the DNAs of Mus cervicolor and Mus caroli mice ( Table 3). Kinetic analysis of this hybridization reveals C0t1/2 values of 150 and 600 for Mus cervicolor and Mus caroli DNAs, respectively, as compared to 500 for Mus musculus and 3,100 for murine "unique" sequences ( 24). Thermal analysis studies revealed Δ-Tm values 3.2 and 3.4°C lower for hybrids formed between MMTV RNA and DNA from Mus cervicolor and Mus caroli, respectively, than observed with Mus musculus DNAs. These results are in general agreement with a recent report ( 21) using MMTV cDNA probes.
MMTV-related DNA in Rats. As seen in Table 3, five different laboratory strains of rats, as well as feral rats, all contained approximately the same degree of MMTV-related information in their DNA when using relaxed conditions of hybridization and RNase assay. DNAs from various organs of a feral rat were also tested for the presence of MMTV-related information and they all contained the same degree of MMTV-related information ( 24). The sequences detected thus appear to be endogenous, i.e., germline transmitted in rats.
The MMTV 60-70S [3H]RNA, obtained from virions from the supernatant fluids of MMTV-infected feline (Fel) cells ( 5), was also used in these experiments and gave the same results as the MMTV RNA obtained from virus grown in murine cells. This RNA was used to rule out the possibility that normal murine or feline cellular RNA or DNA was contaminating the MMTV 60-70S RNA preparation. Thus, the hybridization observed to rat DNA is interpreted as the result of the presence of nucleic acid sequences related to the mouse mammary tumor virus genome.
Fisher, or F344, rats from several colonies were also examined and found to contain the same degree of MMTV-related information in their DNA as other rat strains ( Table 3). However, F344 rats obtained from certain colonies appeared to contain additional MMTV-related information in their DNA, indicating a possible infection and integration by MMTV or an MMTV-related virus. The nature and distribution of these additional sequences are currently being investigated.
To determine at what frequency the MMTV-related sequences are present in rat DNA, MMTV radioactive 60-70S RNA was hybridized to the DNAs of five different laboratory strains of rats as well as C3H mouse liver and mammary tumor DNA. As seen in Figure 6, the kinetics of hybridization to the DNAs of several strains were extremely similar. The C0t1/2 values obtained for all five strains were approximately 800 ( 24). The hybridization of poly A selected radioactive rat RNA to "unique" rat DNA gave a C0t1/2 value of approximately 2,500, indicating that the endogenous MMTV-related sequences in rat DNAs are present in the "low repetitive" range.
To determine the fidelity of hybrids formed between MMTV radioactive RNA and rat DNA, the thermal stability of the RNA-DNA duplexes formed was analyzed by hydroxylapatite chromatography. The Tm values observed were 5° lower than those obtained with homologous Mus musculus DNAs.
A virus, morphologically distinct from the type-B MMTVs of the laboratory mouse Mus musculus, has been identified in the milk of Mus cervicolor popaeus mice ( 25, 26). Group-specific radioimmunoassays for the gp52 ( Figure 7A) and the p28 ( Figure 7B) of MMTV demonstrate that this new virus shares some antigenic determinants with both of these MMTV proteins ( 25). This reactivity is clearly different, however, from that observed with all MMTVs tested from M. musculus. The Mus cervicolor B-type virus has a density of 1.16 g/ml in sucrose and a virion-associated DNA polymerase with a divalent cation preference for Mg2+ over Mn2+. Competitive molecular hybridization experiments showed little if any nucleic acid homology with MMTVs. Radioimmunoassays also clearly differentiate this virus from the other viruses previously identified from M. cervicolor: M432, CERV-CI, and CERV-II ( 26, 27). These studies thus identify the first virus from another species that is immunologically related to the MMTVs of Mus musculus. Similar particles were also observed in a spontaneous M. cervicolor mammary tumor ( 25).
Milk of some feral and various inbred strains of M. musculus has previously been shown to be a source of MMTV, regardless of the mode of transmission of the virus ( 1, 2, 3, 28). Similarly, M. cervicolor milk appears to be a good source of B-type virus, particularly those mice from the Tak province of Thailand. Experiments are now in progress, involving use of this resource, to develop "interspecies" RIAs for the p28 and gp52 viral proteins to use as probes for the detection of antigen expression related to mammary tumor viruses in other rodent and more distantly related species.
1. Bittner, J.J. (1936). Science 84: 162.
2. Bentvelzen, P. (1972). Int. Rev. Exp. Path. 11: 259.
See also
PubMed.
3. Nandi, S., and McGrath, C.M. (1973). Adv. Cancer Res. 17: 353.
4. Lasfargues, E.Y., Lasfargues, J.C., Dion, A.S., Greene, A.E., and Moore, D.H. (1976). Cancer Res. 36: 67.
See also
PubMed.
5. Howard, D.K., Colcher, D., Teramoto, Y.A., Young, J.M., and Schlom, J. (1977). Cancer Res. 37: 2969.
See also
PubMed.
6. Cardiff, R.D., Puentes, M.J., Teramoto, Y.A., and Lund, J.K. (1974). J. Virol. 14: 1293.
See also
PubMed.
7. Parks, W.P., Howk, R.S., Scolnick, E.M., Oroszlan, S., and Gilden, R.V. (1974). J. Virol. 13: 1200.
See also
PubMed.
8. Verstraeten, A.A., vanNie, R., Kwa, H.G., and Hageman, Ph.C. (1975). Int. J. Cancer 15: 270.
See also
PubMed.
9. Ritzi, E., Baldi, A., and Spiegelman, S. (1976). Virology 75: 188.
See also
PubMed.
10. Sheffield, J.B., Daly, T., Dion, A.S., and Taraschi, N. (1977). Cancer Res. 37: 1480.
See also
PubMed.
11. Teramoto, Y.A., Kufe, D., and Schlom, J. (1977). Proc. Natl. Acad. Sci. USA 74: 3564.
See also
PubMed.
12. Teramoto, Y.A., Kufe, D., and Schlom, J. (1977). J. Virol. 24: 525.
See also
PubMed.
13. Teramoto, Y.A., and Schlom, J. (1978). Cancer Res. 38: 1990.
See also
PubMed.
14. Hunter, W.M. (1973). In Handbook of Experimental Immunology (D.M. Weir, ed.), p. 171. Blackwell Scientific Publications, Oxford.
15. Schochetman, G., and Schlom, J. (1976). Virology 73: 431.
See also
PubMed.
16. Elder, J.H., Jensen, F.C., Bryant, M.L., and Lerner, R.A. (1977). Nature 267: 23.
See also
PubMed.
17. Gautsch, J.W., Lerner, R., Howard, D., Teramoto, Y.A., and Schlom, J. (1978). J. Virol. 27: 688.
See also
PubMed.
18. Varmus, H.E., Bishop, J.M., Nowinski, R.C., and Sarkar, N. (1972). Nature New Biol. 238: 189.
See also
MGI.
19. Varmus, H.E., Stavnezer, J., Medeiros, E., and Bishop, J.M. (1975). In Comparative Leukemia Research 1973, Leukemogenesis (Y. Ito and R.M. Dutcher, eds.), p. 451. University of Tokyo Press, Tokyo.
20. Drohan, W., Kettman, R., Colcher, D., and Schlom, J. (1977). J. Virol. 21: 986.
See also
PubMed.
21. Morris, V.L., Medeiros, E., Ringold, G.M., Bishop, J.M., and Varmus, H.E. (1977). J. Mol. Biol. 114: 73.
See also
PubMed.
22. Schlom, J., Colcher, D., Drohan, W., and Kettmann, R. (1978). In Progress in Experimental Tumor Research (F. Homberger, ed.), 21: 140. S. Karger, Basel.
See also
PubMed.
23. Cardiff, R. (1978). In Origins of Inbred Mice (H.C. Morse III, ed.). Academic Press, New York.
24. Drohan, W., and Schlom, J. (1979). J. Natl. Cancer Inst. 62: 1279.
See also
PubMed.
25. Schlom, J., Hand, P., Teramoto, Y.A., Callahan, R., Todaro, G., and Schidlovsky, G. J. Natl. Cancer Inst. 61: 1509.
See also
PubMed.
26. Callahan, R., and Todaro, G. (1978). In Origins of Inbred Mice (H.C. Morse III, ed.). Academic Press, New York.
27. Callahan, R., Sherr, C.J., and Todaro, G.J. (1977). Virology 80: 401.
See also
PubMed.
28. Rongey, R.W., Hiavacova, A., Lara, S., Estes, J., and Gardner, M.B. (1973). J. Natl. Cancer Inst. 50: 1581.
See also
PubMed.