ABSTRACT
Monoclonal antibodies directed against an RNA-binding protein from Xenopus oocytes were used to immunoselect messenger ribonucleoprotein (mRNP) particles. RNA was extracted from both the immunoselected and nonselected fractions and was used to direct the synthesis of oligo (dT)-primed 32P-cDNA. These two cDNA preparations were then used to probe Xenopus stage-1 oocyte cDNA libraries to identify sequences that had been specifically coimmunoselected by the antibodies. Three cDNA clones were shown to be derived specifically from the antibody-selected mRNPs. During very early oogenesis (stage 1–2), the RNA-binding protein and the three coselected mRNAs sediment in the nontranslating mRNP region of a sucrose gradient. By oocyte stage 6, the binding protein concentration decreases by as much as 22-fold relative to polyadenylated RNA. At this stage of development, the three mRNAs are found predominantly in the polysome region of a sucrose gradient. These data demonstrate that Xenopus oocytes contain an RNA-binding protein which binds specific message sequences and may regulate their expression.
Introduction
The programme of development in the early embryo of several animal species is directed by maternal messenger RNAs which are present in the egg at the time of fertilization (reviewed in Davidson, 1986). A large portion of this RNA (as much as 90 % in some species) is not assembled into polyribosomes until the completion of meiosis in the oocyte or fertilization in the egg (reviewed in Smith & Richter, 1985). The degree to which maternal mRNA, as opposed to new zygotic transcripts, contributes to early development varies with the organism. For example, maternal mRNA probably plays a minor role in early mouse development since zygotic transcripts are completely responsible for development beyond the 2-cell stage (Johnson, 1981). In contrast, early development in marine invertebrates such as several species of sea urchin (Wilt, 1964; Nemer & Infante, 1965; Davidson, 1986), Urechis (Rosenthal & Wilt, 1986) and Spisula (Rosenthal, Hunt & Ruderman, 1980) is much more dependent upon maternal mRNA. This is also the case with Xenopus and Drosophila, whose zygotic genomes do not become active until the midblastula stage (Bachvarova & Davidson, 1966; Newport & Kirschner, 1982) and cellular blastoderm stage (Zalokar, 1976; Anderson & Lengyel, 1979), respectively.
In many of the organisms mentioned above, there is a precise temporal regulation of specific mRNAs which assemble into polysomes during oogenesis and embryogenesis. Some rather striking examples include the mRNA for cyclin (Evans et al. 1983) and ribonucleotide reductase (Standart et al. 1985) following fertilization in Spisula, and histones (reviewed in Davidson, 1986) in sea urchins. In Xenopus, mRNAs encoding proteins such as fibronectin (Lee, Hynes & Kirschner, 1984), lamin (Stick & Hausen, 1985), the histones (Adamson & Woodland, 1977; Woodland, Flynn & Wyllie, 1979; Flynn & Woodland, 1980) and c-myc (King, Roberts & Eisenman, 1986; Taylor et al. 1986; Godeau, Persson, Gray & Pardee, 1986) are stored as nontranslating mRNPs in oocytes but are translated in embryos (reviewed in Richter, 1987).
Several alternative, though not mutually exclusive, mechanisms could account for the temporally regulated translation of these mRNAs. Incomplete 5’ cap methylation (Caldwell & Emerson, 1985), poly(A) tail length modulation (Rosenthal, Tansey & Ruder-man, 1983) and RNA secondary structure (Pelletier & Sonenberg, 1985) have all been implicated in translational control in other systems and could contribute to translational regulation in Xenopus as well (e.g. Sagata, Shiokawa & Yamana, 1980; Cabada, Darnbrough, Ford & Turner, 1977). Perhaps these or other structural features of mRNA are sites of interaction with proteins which, in turn, regulate translation.
If RNA-binding proteins are indeed responsible for the regulated expression of some mRNAs during early development, they should fulfil certain criteria. First, they should be present only in those cells that have translationally repressed mRNA, i.e. oocytes and early cleavage stage embryos. Second, they should reversibly inhibit translation in vivo. Third, given the fact that not all mRNAs enter polysomes simultaneously during early development (Dworkin & Hershey, 1981; Dworkin & Dworkin-Rastl, 1985; Dworkin, Shrutkowski & Dworkin-Rastl, 1985), these proteins should bind only a subset of RNAs. Using these parameters as testable hypotheses, we have investigated whether RNA-binding proteins might be involved in translational regulation in early Xenopus development. Using both an in vitro protein blotting procedure (Richter & Smith, 1983) and in vivo photocrosslinking (R. E. Swiderski & J. D. Richter, unpublished data), we have demonstrated that oocytes do indeed contain unique sets of polyadenylated RNA (pA+ RNA)-binding proteins. Furthermore, some of these proteins inhibit translation following in vitro reconstitution with mRNA (Richter & Smith, 1984), although the extent to which this reflects the in vivo function of the proteins is not clear. In this report, we present evidence that a particular RNA-binding protein interacts with specific mRNAs, and that these mRNAs enter polysomes at times of development when the binding protein has diminished in amount.
Materials and methods
RNP immunoprecipitation and cDNA synthesis
Dumont (1972) stage-1 to -2 Xenopus oocytes were dispersed from ovarian fragments by gentle swirling in Ca2+-free Barth’s medium (Maniatis, Fritsch & Sambrook, 1982) containing 015 % collagenase (Sigma type I) for about 1 h. They were then collected, washed several times in homogenization buffer (0·02M-Tris-HCl, pH7·4, 0·3M-KCI, 0·002M-MgCl2, 4 μgml−1 polyvinyl sulphate, 0-5% NP-40, 1 mM-dithiothreitol, 500 units ml−1 RNasin), homogenized and centrifuged at 12000g for 15 min at 4°C. The supernatant was collected and incubated with mixed monoclonal antibodies (Richter & Evers, 1984; Lorenz & Richter, 1985) for 2h at room temperature. This 100μl reaction mixture contained approximately 25 μg of pA+ RNA (as determined by the amount of RNA bound by oligo(dT)-cellulose chromatography in a parallel experiment) and sufficient antibody to bind about 5–10 μg of p56 (as determined by the amount of antibody required to immunoselect p56 from a reticulocyte in vitro translation mix primed with stage-1 to -2 oocyte pA+ RNA, assuming the methionine content of p56 is similar to that of total oocyte proteins, i.e. about 2%). 10μl of swollen, gravity-packed protein A Sepharose CL-4B (Sigma) was added to the mix and incubation continued for another 2 h. The beads were then centrifuged briefly in a microfuge and the supernatant was removed and subjected to a second round of the immunoselection. The combined pelleted beads were washed five times with 0·5 ml of homogenization buffer and then both the beads (bound fraction) and supernatant (unbound fraction) were extracted with p-aminosalicylic acid (PAS), SDS, and phenol and chloroform (Kirby, 1965). The RNA from these two fractions was precipitated with ethanol and used for oligo(dT)-primed cDNA synthesis in the presence of 32P-dCTP and -dGTP and avian myeloblastosis virus reverse transcriptase (BRL) (Richter & Smith, 1981). The cDNA greater than about 1500 nucleotides then was used to screen ovary cDNA libraries. Three controls were used for the detection of nonspecific immunoselection. These were the immunoselection of material (1) in the absence of added antibody but in the presence of protein A Sepharose, (2) in the presence of control antibody (mouse IgG) and protein A Sepharose and (3) using deproteinized pA+ RNA plus the antibodies and protein A Sepharose. In each of these controls, essentially no cDNA was synthesized from the selected material as monitored by the incorporation of radioactivity which had a mobility greater than 1500 nucleotides.
cDNA library screening and isolation of clones
Immature Xenopus ovary cDNA libraries inserted in the vectors pUC9 (Lorenz & Richter, 1985) and bacteriophage λgtll were employed for cDNA screening. The phage library included cDNAs with a mean size of 1600–1800 nucleotides which were blunt-end ligated with EcoRI linkers and inserted into the EcoRI site of λgt11 and amplified. The general cloning procedures used have been described (Maniatis et al. 1982; Huynh, Young & Davis. 1985). Approximately 1000 recombinants from each library were plated per 100 mm Petri dish and duplicate nitrocellulose filters (0·45μm pore, Millipore HATF or Schleicher and Schuell BA85) were lifted from each plate. Lysis of colonies or phage was accomplished with NaOH essentially as described by Maniatis et al. (1982). Each filter was hybridized with 32P-cDNA from the bound (B) or unbound (UB) fractions. Colonies and plaques were picked and replated for a second round of cDNA screening. Those colonies and plaques that hybridized with the B or UB cDNAs a second time were picked, the cDNA insert was isolated and used to probe RNA gel blots. 32 total clones were isolated; 20 from the B fraction. 8 from the UB fraction and 4 that were detected in both fractions. Clones from these fractions were chosen at random and used for further analysis.
Polysome gradients, RNA gel blots and protein gel blots
Ovulated Xenopus eggs were fertilized according to Hollinger & Corton (1980) and the total RNA from 4-cell embryos, as well as Dumont (1972) stage-1 to -2 and -6 oocytes was extracted by p-aminosalicylic acid/phenol/ chloroform (Kirby. 1965) or by guanidinium thiocyanate (Taylor et al. 1986). pA+ RNA was isolated by oligo (dT) cellulose chromatography. For the developmental expression experiments, the amount of pA+ RNA from equivalent numbers of oocytes or embryos was loaded per lane (4 yzg of pA+ RNA from stage-6 oocytes and 4-cell embryos and 1 pg from stage-1 to -2 oocytes). This is based on our previous observations that stage-1 to -2 oocytes contain about one-quarter the amount of pA+ RNA of stage-6 oocytes (Lorenz & Richter, 1985).
Polysome sucrose gradients (15% –40%) containing 15–30 stage-6 oocytes or 4-cell embryos, or 200-300 stage-1 to -2 oocytes were centrifuged in 12 ml SW41 tubes at 34000rev. min−1 for 3 h as described (Richter, Smith, Anderson & Davidson, 1984). The RNA in each of seven fractions was extracted with PAS/SDS/phenol/chloroform, precipitated and resolved by formaldehyde-agarose gel electrophoresis. RNA gel blot procedures and nick translation of cDNA probes have been described (Maniatis et al. 1982).
Results
Two independent procedures, a protein-blotting method (Richter & Smith, 1983) and RNA-protein photocrosslinking (Swiderski & Richter, 1987) have demonstrated that Xenopus oocytes and somatic cells contain different sets of RNA-binding proteins, and that different somatic cells contain very similar sets of binding proteins. The presence of unique oocytebinding proteins suggests that they might regulate the expression of maternal mRNA. Monoclonal antibodies have been generated against one of the oocyte-type pA+ RNA-binding proteins, which has a molecular size of about 56×103 (p56) (Richter & Evers, 1984). With these monoclonal antibodies, we have immunoselected mRNPs to determine whether p56 interacts with all, or a subset of, mRNA sequences. The general scheme for obtaining cDNA clones for such mRNAs is outlined in Fig. 1. Approximately 15 000 recombinants were screened with cDNAs from the bound (B, i.e. immunoselected) and unbound (UB, i.e. nonselected) fractions. In the initial screen, 10–15% of the recombinants hybridized specifically with the B cDNA. About three quarters of these hybridized specifically with the B cDNA on a second round of filter hybridization. 32 of the recombinants that hybridized to the B or UB cDNAs, or both, were then picked, the DNA isolated and used to probe RNA gel blots. As detailed in the Materials and methods, no appreciable cDNA could be generated from material immunoselected by (1) deproteinized pA+ RNA, (2) protein A Sepharose only, or (3) nonimmune mouse IgG.
Scheme for the detection of cDNA clones for pA+ RNAs which coimmunoselect with p56.
Fig. 2 shows several gel blots using B and UB RNA probed with four cDNA clones which were isolated as described above. One of the cDNA clones hybridizes to two RNAs that are detected only in the UB fraction (pXRNP.6). Three other cDNA clones, pXRNP.3, pXRNP.10 and pXRNP.l hybridize to RNAs that are specific for the B fraction. Finally, one cDNA clone hybridizes to RNAs in both B and UB fractions (pXRNP.2). Interestingly, this cDNA clone also hybridizes to a second RNA that is detected only in the B fraction.
RNA gel blot of coimmunoselected RNAs. Following the scheme outlined in Fig. 2, approximately 3pg of RNA from the immunoselected fraction (bound or B) and nonselected fraction (unbound or U) was resolved by formaldehyde-agarose gel electrophoresis and blotted to nitrocellulose. The blots then were reacted with nick-translated cDNA clones which were isolated as described in Fig. 2. The approximate sizes of the RNAs are noted (in kilobases). XRNP.l, XRNP.3 and XRNP.10 are bound RNAs and XRNP.6 is an unbound RNA. Exposure of the autoradiograms ranged from 4 to 10 days.
RNA gel blot of coimmunoselected RNAs. Following the scheme outlined in Fig. 2, approximately 3pg of RNA from the immunoselected fraction (bound or B) and nonselected fraction (unbound or U) was resolved by formaldehyde-agarose gel electrophoresis and blotted to nitrocellulose. The blots then were reacted with nick-translated cDNA clones which were isolated as described in Fig. 2. The approximate sizes of the RNAs are noted (in kilobases). XRNP.l, XRNP.3 and XRNP.10 are bound RNAs and XRNP.6 is an unbound RNA. Exposure of the autoradiograms ranged from 4 to 10 days.
The steady-state levels of the transcripts that hybridize to the cDNA clones described above were determined for several stages of development. The RNA gel blots shown in Fig. 3 demonstrate that the amounts of the transcripts are relatively low during early oogenesis and increase by 5-to 10-fold by the 4cell stage. Our analysis of the levels of these transcripts was confined to those molecules that are retained by oligo(dT)-cellulose. Although it is possible that some fraction of these transcripts may be deadenylated during oogenesis, we have never observed a difference in size when total and pA+ RNA are compared (data not shown). Therefore, if deadenylation of these molecules does occur, relatively few adenylate residues would be removed. The developmental profile of these RNAs is very typical of most pA+ RNAs during early Xenopus development (Golden, Schafer & Rosbash, 1980; Baum & Wormington, 1985).
Developmental regulation of RNAs which coimmunoselect with an RNA-binding protein. pA+ RNA was isolated from oocytes (stage 1–2 and stage 6) and 4-cell embryos, resolved by electrophoresis, blotted to nitrocellulose and reacted with some of the nick-translated cDNA clones described in Fig. 2. RNA from equivalent numbers of oocytes and embryos was applied to the gel. XRNP.l, XRNP.3 and XRNP.10 are bound RNAs and XRNP.6 is an unbound RNA. Exposures of autoradiograms ranged from 2 to 6 days.
Developmental regulation of RNAs which coimmunoselect with an RNA-binding protein. pA+ RNA was isolated from oocytes (stage 1–2 and stage 6) and 4-cell embryos, resolved by electrophoresis, blotted to nitrocellulose and reacted with some of the nick-translated cDNA clones described in Fig. 2. RNA from equivalent numbers of oocytes and embryos was applied to the gel. XRNP.l, XRNP.3 and XRNP.10 are bound RNAs and XRNP.6 is an unbound RNA. Exposures of autoradiograms ranged from 2 to 6 days.
The quantitative relationship between. p56 and pA+ RNA during Xenopus development was derived from several experiments and is presented in Table 1. While the absolute mass of p56 decreases during oogenesis, there is a concomitant increase in the mass of pA+ RNA from 12 to 50 ng per oocyte. Therefore, from these data, it can be calculated that in stage-1 to -2 oocytes, there are 1–2 moles of p56 per mole of pA+ RNA but, by stage 6, this ratio decreases by as much as 22-fold.
Translational control of RNAs bound by p56
We have previously demonstrated that p56 is most prevalent in stage-1 to -2 oocytes and has a sedimentation profile characteristic of nontranslating RNPs (i.e. =⩽80S) (Richter & Evers, 1984). p56 also sediments as a nontranslating RNP in stage-6 oocytes (data not shown). We have compared the sedimentation profile of this protein with several of the RNAs shown in Fig. 2 during early development (Fig. 4A,B). XRNP.6, which is not associated with p56, sediments mostly in the polysome region of the sucrose gradient (i.e. >80 S) during all stages of early development (i.e. stage-1 to -2 oocytes to 4-cell embryos). In contrast, three RNAs which have been shown to interact with protein, XRNP.3, XRNP.l and XRNP. 10), sediment mainly in the nontranslating RNP region of the sucrose gradient (i.e. ⩽80 S) during oocyte stage 1–2, and sediment mainly in the polysome region of the gradient during later development (stage-6 and 4-cell embryo).
Sedimentation of coimmunoselected RNAs during development. Postmitochondrial supernatants from stage-1 to -2 and -6 oocytes and 4-cell embryos were centrifuged through a sucrose gradient and the RNA was extracted from each of seven fractions. The RNA was resolved by formaldehyde-agarose gel electrophoresis and hybridized with nick-translated pXRNP.6 and pXRNP.3 (part A) and pXRNP.1 and pXRNP.10 (part B). The direction of sedimentation and 80S monosome peak are denoted. XRNP. 1, XRNP.3 and XRNP.10 are bound RNAs and XRNP.6 is an unbound RNA. Exposure of autoradiograms ranged from 3 to 8 days.
Sedimentation of coimmunoselected RNAs during development. Postmitochondrial supernatants from stage-1 to -2 and -6 oocytes and 4-cell embryos were centrifuged through a sucrose gradient and the RNA was extracted from each of seven fractions. The RNA was resolved by formaldehyde-agarose gel electrophoresis and hybridized with nick-translated pXRNP.6 and pXRNP.3 (part A) and pXRNP.1 and pXRNP.10 (part B). The direction of sedimentation and 80S monosome peak are denoted. XRNP. 1, XRNP.3 and XRNP.10 are bound RNAs and XRNP.6 is an unbound RNA. Exposure of autoradiograms ranged from 3 to 8 days.
Table 2 summarizes the sedimentation data of the RNAs shown in Fig. 4. About two thirds of XRNP.6, which is not associated with p56 (based on the criterion of antibody selection), sediments with polysomes during all stages of development (we assume in these studies that RNAs that sediment in the 80S fraction are nonpolysomal). In contrast, at least three quarters of the RNAs that are associated with p56, XRNP.l, XRNP.3 and XRNP. 10, sediment as nontranslating RNPs in stage-1 to -2 oocytes. By stage 6, however, only 15–20 % of each of these RNAs sediment as RNPs. Thus, when p56 is prevalent, most of the RNAs with which it associates are nontranslating; when the molar ratio of this protein to those RNAs declines (Table 1), the associated RNAs enter polyribosomes.
It should be noted that oogenesis in Xenopus can last as long as three to six months and, therefore, the mRNAs described above could become polysomal at any time during this period. In addition, most pA+ RNA in Xenopus is synthesized throughout oogenesis and is maintained at a relatively constant level by offsetting turnover (Golden et al. 1980). Other mRNAs, however, are synthesized early and are refractive to turnover (Ford, Mathieson & Rosbash, 1977; reviewed in Richter, 1987). As such, we do not know whether the RNAs described in this study are newly synthesized in stage 6 or are derived from an earlier stage.
In stage-1 to -2 oocytes, the binding of specific mRNAs by p56 correlates with their lack of expression at this stage of development. The further observation that p56 is not found in polysomes with these B mRNAs in stage-6 oocytes suggests that it might negatively regulate translation. This possible interpretation would be strengthened if it could be shown that p56 does not associate with B mRNAs in stage-6 oocytes. Accordingly, we have immunoselected mRNPs from stage-6 oocytes (cf. Fig. 1), and have probed the B and UB fractions with cDNA clones from RNAs that are p56-associated in stage 1–2 oocytes. Fig. 5 shows that the stage-1 to -2 bound RNAs XRNP.l, XRNP.3 and XRNP.10 are mostly unassociated with p56 in stage 6.
RNA gel blot of p56-associated RNAs from stage-6 oocytes. Following the scheme outlined in Fig. 1, RNA was extracted from anti-p56 coselected (B) and nonselected (UB) material from stage-6 oocytes. The RNA was resolved by electrophoresis, blotted and probed with nick-translated pXRNP. 1, pXRNP.3. pXRNP.10 (representing bound RNAs).
RNA gel blot of p56-associated RNAs from stage-6 oocytes. Following the scheme outlined in Fig. 1, RNA was extracted from anti-p56 coselected (B) and nonselected (UB) material from stage-6 oocytes. The RNA was resolved by electrophoresis, blotted and probed with nick-translated pXRNP. 1, pXRNP.3. pXRNP.10 (representing bound RNAs).
Discussion
This communication presents evidence that an RNA-binding protein interacts with a subset of the diverse message sequences present in Xenopus oocytes. This protein, p56, was identified initially on protein blots which were reacted with radiolabelled mRNA, and was characterized as decreasing in amount during oogenesis, being absent from somatic tissues, sedimenting in the mRNP fraction of a velocity sedimentation gradient and as binding pA+ RNA (Richter & Smith, 1983). Monoclonal antibodies generated against this protein (Richter & Evers, 1984) were used to immunoselect mRNPs, which were shown to contain specific message sequences. Furthermore, the diminishing levels of the protein during oogenesis correlate with the assembly of specific mRNAs into polysomes.
It is difficult to determine with certainty the number of different message sequences that are associated with p56. Through two rounds of filter hybridization, about 25 % of the colonies or plaques that hybridized initially only with the B cDNA hybridized with both the B and UB cDNAs on a subsequent screen. In addition, many recombinants do not give detectable hybridization signals with 32P-cDNA from either the B or UB fraction. This low frequency of hybridization was also observed by Weeks, Rebagliati. Harvey & Melton (1985), who showed that as many as 40% of oocyte message sequences are present at too low a frequency to be detected by 32P-cDNA filter hybridization. Of the cDNA clones which we do detect, however, about 10% are unique to the B fraction.
The data presented in Fig. 5 and Table 2 demonstrate that three RNAs which immunoselect with monoclonal antibodies directed against p56 are mostly nontranslating during early oogenesis. When the molar concentration of the protein decreases by as much as 22-fold relative to the molar concentration of pA+ RNA by stage 6 (Table 1), these three mRNAs are translated. Thus the presence of p56 correlates with the translatable nature of the mRNAs with which it interacts. While we have no direct evidence that p56 acts as a repressor of translation, it is noteworthy that it is one of seven or eight proteins that have been shown to inhibit translation of a reconstituted mRNP reversibly (Richter & Smith, 1984). Whether p56, or other proteins, inhibits translation in vivo remains to be demonstrated. In addition, we do not know the nucleotides that are bound by p56, although it probably is not poly(A) since we have been able to discriminate between polyadenylated molecules by immunoselection.
p56 is not the only RNA-binding protein that is unique to oocytes (Richter & Smith, 1983). Recently, we have used u.v. irradiation to crosslink covalently pA+ RNA and their associated proteins in living oocytes and somatic cells (R. E. Swiderski & J. D. Richter, unpublished data). Under these stringent conditions, we find that oocytes have unique sets of binding proteins. While it is difficult at the present time to ascribe a function to these proteins, it is provocative that only oocytes (and newly fertilized eggs) have a huge mass of pA+ RNA which is not associated with polysomes (see below).
Two-dimensional gel analysis of polypeptide patterns has shown that new proteins are synthesized during Xenopus oogenesis (Harsa-King. Bender & Lodish. 1979; Younglai, Godeau & Baulieu. 1981). This correlates with our finding of the assembly of specific mRNAs onto polysomes during this stage of development. In an attempt to identify one of these specific mRNAs, we sequenced its corresponding cDNA, which encodes 247 amino acids. Comparison of this predicted amino acid sequence with those of other known proteins has failed to give clues as to its identity. Perhaps sequence analysis of the other RNAs that are bound by p56 will help in their characterization.
One of the most unique characteristics of oocytes is their accumulation of maternal macromolecules which are destined to be used in the early embryo (Smith & Richter, 1985; Davidson. 1986; Richter. 1987). The Xenopus oocyte accumulates up to 50-90ng of pA+ RNA, about 90% of which is not associated with polysomes. Some of this ‘masked’ mRNA, however, assembles into polysomes during early embryogenesis. We have recently obtained evidence that one factor that influences the overall rate of translation in oocytes is the availability of eukaryotic initiation factor 4A (Audet, Goodchild & Richter, 1987). However, it is difficult to reconcile the availability of this factor with the temporal specificity by which many mRNAs enter polysomes. We therefore envision another tier of regulation, possibly proteins which act as repressors of translation. For protein binding to be responsible for the translation of each mRNA, one might expect, in the extreme case, a different protein for each mRNA. However, our data show that several mRNAs are bound by the same protein, and that these mRNAs are found on polysomes when the protein diminishes in amount. Therefore, it seems more likely that there are broad classes of proteins which bind several different mRNAs. To determine if this indeed is the case requires antibodies directed against different developmentally regulated mRNP proteins. We presently are attempting to raise such additional antibodies for use in further mRNP immunoselection experiments similar to those described here.
ACKNOWLEDGEMENTS
We thank Dr T. Thomas (Texas A & M University) for gifts of phage cloning materials and instructions on their use and Drs W. Crain, T. Pederson and R. Swiderski for comments on the manuscript. This work was supported by NIH grants GM34554 and CA40189 to J.D.R. J.D.R. is the recipient of a Cancer Research Scholar Award from the American Cancer Society. Massachusetts Division.