Gene, 18 (1982) 289-296 Elsevier Biomedical Press

289

Construction of a broad host range cosmid cloning vector and its use in the genetic analysis of Rhizobium mutants

(Recombinant DNA; Rhizobium meliloti; clone bank; X COS)

Alan M. Friedman *, Sharon R. Long *, Susan E. Brown, William J. Buikema and

Frederick M. Ausubel * *

Department of Cellular and Developmental Biology, Harvard University, Cambridge, MA 02138 (U.S.A.)

(Received January 2Oth, 1982) (Accepted March 5th, 1982)

SUMMARY

We have constructed a cosmid derivative of the low copy-number broad host-range cloning vector pRK290 (Ditta et al., 1980) by inserting a 1.6kb Bg/II fragment containing X cos into the unique BglII site in pRK290. The new vector, pLAFR1, is 21.6 kb long, confers tetracycline resistance, contains a unique EcoRI site, and can be mobilized into and stably replicates within many Gram-negative hosts. We constructed a clone bank of Rhizobium meliloti DNA in pLAFR1 using a partial EcoRI digest. The mean insert size was 23.1 kb. When the clone bank was mated (en masse) from Escherichia coli to various R.

meliloti auxotrophic mutants, tetracycline-resistant (Tc’) transconjugants were obtained at frequencies ranging from 0.1 to 0.8, and among these, prototrophic colonies were obtained at frequencies ranging from 0.001 to 0.007. pLAFR1 cosmids were mobilized from R. meliloti prototrophic colonies into E. coli and then reintroduced into R. meliloti auxotrophs. In most cases, 100% of these latter Tc’ transconjugants were prototrophic.

INTRODUCTION

In this paper we describe the construction and use of a broad host-range “cosmid” vector which

* Current addresses: (A.M.F.) ERCO, Cambridge, MA 02138 (U.S.A.); (S.R.L.) Department of Biological Sciences, Stanford University, Stanford, CA 94305 (U.S.A.). ** To whom correspondence and reprint requests should be sent. Abbreviations: Ap, ampicillin; kb, kilobase pairs; Km, kanamycin; Nm, neomycin; Tc, tetracycline.

can be useful in the genetic analysis of a large number of Gram-negative species. The new vector is derived from the broad host-range cloning sys- tem developed by Ditta et al. (1980). This system consists of two plasmids derived from the P-group plasmid RK2, a 56-kb low copy number plasmid which confers resistance to Tc, Nm, Km and Ap and which is self-transmissible to and replicates within most Gram-negative bacteria. The first de- rivative of RK2 is the 20-kb plasmid, pRK290, which confers Tc’, is mobilizable but not self- transmissible (mob+ tru-), and contains single

0478-l 119/82/OOOt-0000/$02.75 6 1982 Elsevier Biomedical Press

290

recognition sites for EcoRI and BgIII. The second derivative of RK2, pRK2013 (Figurski and Helin- ski, 1979), contains the RK2 tra functions and Nm’ gene ligated to a ColEl replicon. and will mobilize pRK290 in tram to a variety of Gram- negative recipients.

We have modified pRK290 by ligating a BglII fragment, purified from the cosmid pHC79 and containing the X cos site, into the unique BglII recognition site of pRK290. The resulting plasmid, pLAFR1, is an EcoRI cloning vector which when ligated to DNA fragments in the size range of 15-31 kb can be packaged in vitro into X phage heads and infected into E. coli at high efficiency. In the presence of the helper plasmid pRK2013, a pLAFR1 derivative can be conjugated into a suit- able Gram-negative recipient where the hybrid cosmid is stably maintained as an R-prime. In addition, a clone bank in pLAFR1 can be con- jugated en masse into a mutant recipient strain and, if a suitable genetic selection is available, clones carrying the wild-type allele of the mutated gene can be obtained by selecting for complemen- tation of the mutant phenotype. We describe here the use of pLAFR1 in the genetic analysis of R. meliloti, the nitrogen-fixing endosymbiont of al- falfa.

TABLE I

Bacterial strains

MATERIALS AND METHODS

(a) Bacterial strains and media

The genotypes of the bacterial strains used are listed in Table I. The following media have been described previously: LB, a rich medium for E. coli and R. meliloti (Miller, 1972); M9 glucose and M9 sucrose, minimal media for E. co/i (Miller. 1972) and R. meliloti (Meade and Signer, 1977) respectively; X broth, a rich medium for growing phage X (Miller, 1972); and LAM, a plate medium for phage X infection (Hohn. 1979).

(b) DNA isolation

Preparative amounts of plasmid DNAs were isolated by a modification of the cleared lysate procedure of Clewell and Helinski (1969) using 1% Triton X-100 instead of Brij-58 deoxycholate. Small amounts of plasmid DNAs were isolated by a Triton-lysis/phenol-extraction procedure (Klein et al., 1980) or a Triton-lysis “boiling” method (Holmes and Quigley, 198 1). Total R. meliloti DNA was isolated as described by Meade et al. (1982).

Strain Relevant genotype Source or ref.

E. coli HBlOl

NS428

NS433

MM294[pRK20 131

R. melilori 1021 1023 1031 3359

3390

WA, hsdR, hsdbf, str’ pm, leu XAamllb2red3 cIS57Sam7 XEam4b2red3 cI857Sam7 contains pRK20 13

str’ 1021 met::Tn5 1021 thi::TnS a-3, &p-33, his-39 pan-41, spc- 1, ri/- 1 mm-90 m-3, Pp-33, mm-57 spc- 1, his-39, pyr-80 rif- 1

Boyer and Roulland-Dussoix, 1969

Sternberg et al., 1977

Sternberg et al.. 1977 Sternberg et al., 1977 Ruvkun and Ausubel, 198 1

Meade et al., 1982 Meade et al., 1982 S. Brown, unpublished Meade et al., 1977

Meade et al., 1977

291

(c) DNA biochemistry

Restriction endonucleases were purchased from Bethesda Research Laboratories and used accord- ing to the manufacturer’s instructions. Agarose gel electrophoresis was carried out in a horizontal apparatus as described (Riedel et al., 1979). DNA was transferred from agarose gels to sheets of nitrocellulose (Schleicher and Schuell, BA85) as described by Southern (1975), incorporating the limited depurination step of Wahl et al. (1979). DNA was labeled in vitro using the nick-transla- tion procedure described by Maniatis et al. (1975) and labeled DNA probes were hybridized to nitrocellulose sheets using the method of Botchan et al. (1976). DNA ligations were performed with T4 DNA ligase (prepared by the method of Tait et al., 1980) as previously described (Cannon et al., 1977). E. coli was transformed with plasmid DNA using the calcium shock method of Cohen et al. (1972).

(d) Size fractionation of R. meliloti DNA

A partial EcoRI digest of total R. meliloti DNA was layered onto a linear gradient of 5-25% sucrose (Schwartz-Mann Ultrapure in 10 mM Tris . HCl, pH 8.0, 1 mM EDTA, 100 mM NaCl). After spin- ning in an SW40 rotor (Beckman Instruments) at 20000 rev./mm for 17 h, the tube was punctured at the bottom and fractions of 0.2 ml were col- lected. The size of DNA present in each fraction was determined by .agarose gel electrophoresis.

(e) Lambda in vitro packaging

Packaging extracts were prepared by the method of Hohn (1979) using the X lysogenic strains NS428 and NS433 described by Stemberg et al. (1977). Aliquots of packaging extract (50 ~1) were stored in 1.5 ml eppendorf centrifuge tubes at -70°C. Extracts were thawed on ice and 1 ~1 of 0.1 M ATP was added, followed by l-5 ~1 of the DNA to be packaged. The mixture was stirred gently with a 10 ~1 micropipette, spun briefly in a clinical centrifuge at top speed, and then incubated at 37°C for 1 h. A second 50 ~1 aliquot of packaging extract was mixed with 5 ~1 of DNase I (Worthing- ton; 100 pg/ml in 50 % TMN + gelatin; 50%

glycerol) and 2.5 ~1 of 0.5 M MgCl,, and 20 ~1 of this second mixture was added to each tube of packaging extract containing DNA. Incubation was continued for an additional 30 min at 37°C. The reaction was then stopped by the addition of 0.5 ml SM (100 mM NaCl, 5 mM MgSO,, 50 mM Tris . HCl, pH 7.5, 0.01% gelatin) and 2 drops of CHCl,. The tube was spun at 400 X g for 2 min and the supematant stored at 4°C until use.

(f) Bacterial genetics

Recombinant DNA molecules containing X cos and packaged in vitro in X heads were used to infect E. coli strain HBlOl as described by Hohn (1979). Tri-parental matings using E. coli strain MM294 [pRK2013] as a source of the mobilizing plasmid pRK2013, were performed as described by Ruvkun and Ausubel(198 1).

RESULTS

(a) Construction of pLAFR1

The plasmid pRK290 is a 20-kb derivative of the resistance transfer factor RK2, which confers tetracycline resistance, and contains single sites for EcoRI and BglII. Because of its broad host range and its ability to be mobilized in tram, pRK290 is a useful vector for many Gram-negative species. We reasoned that the construction of a “cosmid” derivative of pRK290 that could be packaged into phage X heads in vitro and introduced into E. coli

by infection would be advantageous because packaging in phage heads imposes a size selection on the recombinant inserts with the result that all clones in a gene library would be approximately the same size. Moreover, because of the large average insert size (approx. 23 kb), a cosmid clone bank in pLAFR1 of a bacterial species with a genome size equivalent to that of E. coli (4100 kb) would only need to contain about 820 members to guarantee 99% completeness. [This estimate was based on the equation P = 1 - (1 - f )N, where P is the probability of a given sequence being in the clone bank, f is the fraction of the genome of the average insert, and N is the number of clones in

292

p LAFR 1

21.6 kb

Fig. 1. Restriction map of the cosmid pLAFR1. Sal, SalI; Bs, BsrEII; Bg, &/II; R, EcoRI; rlx, relaxation complex site.

the bank (Clark and Carbon, 1979).] A convenient source of X cos DNA is the cosmid

pHC79 (Hohn and Collins, 1980), a 6.4-kb deriva- tive of pBR322 which confers tetracycline and ampicillin resistance (both contained within a 4.8 kb BglII fragment) and which carries the A cos site on a 1.6-kb Bg/II fragment. We wished to insert only the 1.6-kb cos containing BglII fragment into the BglII site of pRK290. To diminish interference from the large 4.8-kb BglII fragment of pHC79, we digested pHC79 with PstI, which cuts the 4.8-kb fragment but not the desired 1.6-kb frag- ment, and BglII. This mixture was ligated to BglII-digested pRK290, and the ligation products were used to transform HBlOl. Tc’ transformants were seiected and then screened for ApS. Plasmid DNA was isolated and characterized from 48 Tc’Ap” colonies. Out of these, one strain yielded a plasmid containing a 1.6-kb Bg/II fragment in addition to pRK290, and this plasmid was desig- nated pLAFR1.

To confirm its structure, pLAFR1 DNA was digested with BglII and EcoRI, the fragments were transferred to nitrocellulose by the method of Southern (1975) and the filter was hybridized with nick-translated pHC79 DNA. The resulting auto- radiogram confirmed that the small fragment in pLAFR1 was derived from pHC79 (data not shown).

The orientation of the 1.6-kb fragment within pLAFR1 was determined using the enzyme Bst Ell, which cuts asymmetrically within the 1.6-kb BglII fragment and has no site within pRK290. By cleaving the plasmid with EcoRI and Bst EII or with Sal1 and BstEII. the map shown in Fig. 1 was determined.

(b) Construction of a clone bank of R. meliloti DNA

R. meliloti DNA, partially digested with EcoRI, was fractionated by size (see MATERIALS AND

METHODS) and those fractions containing frag- ments 22 to 35 kb were pooled and concentrated. This DNA was ligated to EcoRI cleaved pLAFR1 at a concentration of approx. 400 pg/ml R. meli- loti DNA and about 50 pg/ml vector DNA; the high ratio of insert to vector was chosen to reduce the probability of vector-vector ligation. The ligated mix was packaged in phage lambda heads as described in MATERIALS AND METHODS and used to infect HBlOl. From the number of Tc’ transductants selected, we calculated that the clone bank contained approx. 15 000 independent cos- mids. 24 Tc’ colonies chosen at random were used

1 2 3 4 5 6 7 8 9 10 1112 Fig. 2. Restriction digests of representative cosmid clones from

an R. melilori clone bank. Agarose gel electrophoresis of par-

tially purified plasmid DNAs cleaved with EcoRI. Lane I : intact X DNA plus X DNA cleaved with EcoRI. Lane 2:

pLAFR1 DNA. Lanes 3- 12: DNA isolated from 10 randomly

chosen Tc’ colonies arising as a result of infection with a

pLAFRl-R. meliloti DNA ligation mix packaged in vitro in X

heads.

as sources to isolate plasmid DNA using the small-scale isolation procedures described in MATERIALS AND METHODS. Digestion of these plasmids with EcoRI (Fig. 2) produced in all cases a fragment that co-migrated with pLAFR1 and in 23 out of 24 cases produced other fragments, presumably inserts. One out of the 24 strains yielded only a pLAFRl-sized fragment, and was probably a vector dimer. Using phage A fragments as markers, the total length of insert DNA was measured. This was found to vary from 12.8 kb to 30.0 kb; the mean was 23.1 kb. Thus the total length of DNA being packaged appears to be 34.4 to 5 1.6 kb, which corresponds well with the re- ported values for size selectivity in lambda packag- ing in vitro (Hohn, 1979).

(c) Use of the clone bank to complement aux-

otrophs

The cosmid clone bank was mated (en masse) into various auxotrophic strains of R. meliloti, selecting transconjugants by Tc’ and by comple- mentation of the appropriate auxotrophy. Results of these matings are presented in Table II. Assum- ing that the R. meliloti genome is the same as the E. coli genome (approx. 4100 kb), the maximum fraction of clones expected to complement a par- ticular mutation is 0.0056 for 23-kb inserts. Fre-

TABLE II

293

quencies of transfer of pRK290, pLAFR1, and the hybrids formed from pLAFR1 were similar, and ranged from 0.1 to 0.9. The fraction of the clone bank found to complement particular mutations varied between 0.001 and 0.007. These results indi- cate that the bank is relatively complete and con- tains a more or less random collection of cloned fragments, at least with respect to basic metabolic functions. Whether the restoration of prototrophic growth in these transconjugants is due to true complementation or to marker rescue was determined by both physical and genetic analysis of the plasmids they contained.

Plasmids from several complemented aux- otrophs were isolated using the small-scale isola- tion procedures described in MATERIALS AND METHODS, digested with EcoRI, and examined by agarose gel electrophoresis. In the case of plasmids isolated from several independent Met+ and Thi+ transconjugants of strains 1023 and 1031, common EcoRI restriction fragments were observed within each group of plasmids (data not shown). Several of the cosmids were transferred to E. coli by conjugation selecting for Tc’ (see MATERIALS AND METHODS). Each of these E. coli transconjugants was used as a donor to conjugate the cosmids back into either R. meliloti strain 1023 or 1031. The R.

meliloti transconjugants were subsequently tested for complementation of the appropriate aux-

Frequencies of conjugation of pRK290, pLAFR1 and pLAFRl-based clone bank into various R. meliloti auxotrophs

Donor Recipient Selection Number of transconjugants/ number of recipients

Fraction in bank which complement auxotrophy a

HBlOl[pRIU’O] 1023(met) Tc’ 0.80 HBlOl[pLAFRl] 1023(met) Tc’ 0.90 HBlOl[Bank] 1023(met) Tc’ 0.63 HBlOl[Bank] 1023(met) Tc’+Met+ 0.0032 0.005 1 HBlOl[Bank] 103l(thi) Tc’ 0.47 HBlOl[Bank] 103l(thi) Tc’+Thi+ 0.0012 0.0025 HBlOl[Bank] 3359(pyr, his, trp) Tc’ 0.13 HBlOl[Bank] 3359(pyr, his, trp) Tc’+Pyr+ 0.00013 0.001 HBlOl[Bank] 3359(pyr, his, trp) Tc’+Trp+ 0.001 0.0076 HBlOl[Bank] 339o(Pan, his, trp) Tc’ 0.82 HBlOl[Bank] 339tXPan his, trp) Tc’+Pan+ 0.0036 0.0044 HBlOl[Bank] 339o(Pan, his, tip) Tc’+Trp+ 0.0054 0.0053

B Fraction of clones that complement was calculated by dividing the frequency of Tc-complemented transconjugants by the frequency of Tc’ transconjugants.

294

TABLE III

Properties of recombinant plasmids isolated from prototrophic transconjugants a

Parental strain Marker complemented Clone No. Ratio of prototrophic colonies to Tc’ colonies in second mating to parental auxotrophic strain

1023 met 1 2 3

1031 rhi 1 2

3

3390

3359

Pan

trP

@P

PY’

59/60 Met+ 52/52 Met+ 52/52 Met+

23/23 Thi+ 3/ 3 Thi+

IO/10 Thi+

52/52 Pan+ 52/52 Pan+

O/52 Tip+

36/36 Trp + O/32 Tip+ O/36 Pyr’

a Plasmids were conjugated from “complemented” parental auxotrophs into E. coli HBlOl and re-conjugated into the same R. melilcm

auxotrophs.

atrophic mutation. As shown in Table III, three independent prototrophic colonies of strain 1023 and three such colonies of strain 103 1 each yielded a plasmid which conferred prototrophy again upon transfer into 1023 or 1031, respectively.

Similar experiments were also carried out with the trp, pan, and pyr mutations in strains 3359 and 3390. As seen with strains 1023 and 1031, most plasmids isolated from 3359 or 3390 shared re- striction fragments in common (data not shown), and when these plasmids were conjugated into E. coli and subsequently re-conjugated into the ap- propriate R. meliloti auxotroph, they comple- mented the expected auxotrophic mutation (Table III). However, in contrast to strains 1023 and 1031, approximately half of the primary 3359 and 3390 prototrophic transconjugants contained a plasmid which did not share any restriction frag- ments in common with other complementing plasmids. When these plasmids were conjugated into E. coli and then re-introduced into 3359 or 3390, they failed to complement. Examples of three such cases are given in Table III.

DISCUSSION

Cosmid cloning was used to generate a clone bank of plasmids which are functionally equiva- lent to R-primes. In the experiments reported here, about 8000 clones were generated per microgram of insert DNA; this represents about a ten-fold increase over the frequency obtained by Ditta et al. (1980) for a pRK290 clone bank derived by transformation. Efficiency in cloning may be im- portant in situations where the availability of in- sert DNA is limiting. For example, R. meliloti contains an extremely large extrachromosomal ele- ment, the “megaplasmid,” which contains the nitrogenase and other symbiotic genes (Banfalvi et al., 1982; Rosenberg et al., 1982; S.R. Long and W.J. Buikema, unpublished). This plasmid is dif- ficult to purify in large quantities so preparing a complete clone bank of it may depend upon effi- cient cloning techniques.

The advantage of cloning Rhizobium DNA in a vector such as pRK290 or pLAFR1, instead of in a ColEl-based plasmid or cosmid, is that the clone bank can be used in genetic studies because the RK2 replicon will function in Rhizobium cells. We found that a clone bank in the cosmid vector pLAFR1 could be used to directly complement a

295

variety of auxotrophic mutations in R. meliloti at frequencies ranging from 0.001 to 0.007. Recently, we have shown that this procedure can be used to directly clone genes involved in nodulation by conjugation of the clone bank into R. meliloti

nodulation mutants and by selecting for transcon- jugants which regain the ability to nodulate plants (Long and Ausubel, 1982).

The failure to isolate cosmids from a small percentage of Tc’ prototrophic transconjugants which would subsequently complement the ex- pected R. meliloti auxotrophy can be explained in at least two ways. It is possible that these Tc’ prototrophic transconjugants contained reversions of the auxotrophic mutation, even though no re- vertants were observed in the controls. Another possibility is that recombination had occurred be- tween a cosmid and the chromosome, transferring a wild-type allele to the chromosome, and that the recombinant cosmid was subsequently replaced by a second unrelated cosmid. The data in Table III suggest that there might have been some allele specificity in terms of the number of false positive complementing cosmid clones (for example, 2 out of 3 presumptive trp containing cosmids were negative upon remating). An insufficient number of individual clones have been examined, however, to substantiate this hypothesis.

Because the modification of pRK290 carried out to make pLAFR1 has not affected any known genes for replication or mobilization and does not appear to affect the stability of pLAFR1 in R. meliloti, the host range of pLAFR1 should be as wide as that of its parent plasmid. This wide host range, together with the large insert size and ef- ficiency of transfer possible by using lambda packaging, should make this a useful vector for various applications in cloning genes of Gram-negative bacteria.

ACKNOWLEDGEMENTS

We thank R. Blackman for Charon 4A DNA used for assaying packaging mixtures, G.B. Ruvkun for helpful discussions,. and R. Hyde for assistance in preparing the manuscript. S.R.L. was the recipient of an NIH postdoctoral fellowship

and this work was supported by NSF grants No. PCM7806834 and No. PCM8104492 awarded to F.M.A.

REFERENCES

Banfalvi, Z., Sakanyan, V., Koncz, C., Kiss, A., Dusha, I. and Kondorosi, A.: Location of nodulation and nitrogen fixa- tion genes on a high molecular weight plasmid in Rhizobium meliloti. Mol. Gen. Genet. 184 (1982) 318-325.

Botchan, M., Topp, W. and Sambrook, J.: The arrangement of simian virus 40 sequences in the DNA of transformed cells. Cell 9 (1976) 287-296.

Boyer, H.B. and Roulland-Dussoix, D.: A complementation analysis of the restriction and modification of DNA in Escherichia coli. J. Mol. Biol. 41 (1969) 459-472.

Cannon, F.C., Riedel, G.E. and Ausubel, F.M.: Recombinant plasmid that carries part of the nitrogen fixation (ni/) gene cluster of Klebsiellapneumoniae. Proc. Natl. Acad. Sci. USA 74 (1977) 2963-2967.

Clark, L. and Carbon, J.: Selection of specific clones from colony banks by suppression or complementation tests, in Wu, R. (Ed.), Methods in Enzymology, Vol. 68. Academic Press, New York, 1979, pp. 396-408.

Clewell, D.B. and Helinski, D.R.: Supercoiled circular DNA- protein complex in Escherichia coli: Purification and in- duced conversion to an open circular DNA form. Proc. Natl. Acad. Sci. USA 62 (1969) 1159-l 166.

Cohen, S.N., Chang, A.C.Y. and Hsu, L.: Non-chromosomal antibiotic resistance in bacteria: Genetic transformation of E. cob by R-factor DNA. Proc. Natl. Acad. Sci. USA 69 (1972) 2110-2114.

Ditta, G., Stanfield, S., Corbin, D. and Helinski, D.R.: Broad host range DNA cloning system for Gram-negative bacteria: Construction of a gene bank of Rhizobium meliloti. Proc. Natl. Acad. Sci. USA 77 (1980) 7347-7351.

Figurski, D.H. and Helinski, D.R.: Replication of an origin- containing derivative of plasmid RK2 dependent on a plasmid function provided in trans. Proc. Natl. Sci. USA 76 (1979) 1648-1652.

Hohn, B.: rn vitro packing of X and cosmid DNA, in Wu, R. (Ed.), Methods in Enzymology, Vol. 68. Academic Press, New York, 1979, pp. 299-309.

Hobn, B. and Collins, J.: A small cosmid for efficient cloning of large DNA fragments. Gene 11 (1980) 291-298.

Holmes, D.S. and Quigley, M.: A rapid boiling method for the preparation of bacterial plasmids. Analytical B&hem. 114 (1981) 193-197.

Klein, R.D., Selsing, E. and Well, R.D.: A rapid microscale technique for isolation of recombinant plasmid DNA sui- table for restriction enzyme analysis. Plasmid 3 (1980) 88-91.

Long, S.R., Buikema, W.J. and Ausubel, F.M.: Cloning of Rhirobium meliloti modulation genes by direct complementa- tion of Nod- mutants. Nature (1982) in press.

296

Maniatis, T., Jeffrey, A. and Kleid, D.: Nucleotide sequence of the rightward operator of phage lambda. Proc. Natl. Acad. Sci. USA 72 (1975) 1184-1188.

Meade, H. and Signer, E.: Genetic mapping of Rhkobium meliloti. Proc. Natl. Acad. Sci. USA 74 (1974) 2076-2078.

Meade, H.M., Long, S.R., Ruvkun, G.B., Brown, S.E. and Ausubel, F.M.: Physical and genetic characterization of symbiotic and auxotrophic mutants of Rhizobium meliloti induced by transposon Tn5 mutagenesis. J. Bacterial. 149 (1982) 114-122.

Miller, J.: Experiments in Molecular Genetics. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, 1972.

Riedel, G.E., Ausubel, F.M. and Cannon, F.C.: Physical map of chromosomal nitrogen fixation (nij) genes of Klebsiellu pneumoniae. Proc. Natl. Acad. Sci. USA 76 (1979) 2866- 2870.

Rosenberg, C., Boistard, P., Denarie, J. and Casse-Delbart, F.L.: Genes controlling early and late functions in symbio- sis are located on a megaplasmid in Rhizobium meliloti. Mol. Gen. Genet. 184 (1982) 326-333.

Ruvkun, G.B. and Ausubel, F.M.: A general method for site- directed mutagenesis in prokaryotes. Nature 289 (1981) 85-88.

Southern, E.M.: Detection of specific sequences among DNA fragments separated by gel electrophoresis. J. Mol. Biol. 98 (1975) 503-517.

Sternberg, N., Tiemeier, D. and Enquist, L.: In vitro packaging of a XDam vector containing EcoRI DNA fragments of Escherrchia coli and phage Pl. Gene 1 (1977) 2555280.

Tait, R.C., Rodriguez, R.L. and West, R.W.: The rapid purifi- cation of T4 DNA ligase from a hT4 lig lysogen. J. Biol. Chem. 255 (1980) 813-815.

Wahl, G.M., Stern, M. and Stark, G.R. Efficient transfer of large DNA fragments from agarose gels to diazobenzy- loxymethyl-paper and rapid hybridization by using dextran sulfate. Proc. Natl. Acad. Sci. USA 76 (1979) 3683-3687.

Communicated by S.R. Kushner.