[1 II RESTRICTIONENDONUCLEASES 113

[11] P r o p e r t i e s a n d U s e s o f R e s t r i c t i o n E n d o n u c l e a s e s
By JOAN E. BROOKS

This chapter focuses on the properties and uses of restriction endonu-

cleases. The large battery of endonucleases now commercially available
will first be described in terms of nomenclature and properties. The ne~xt
section will cover basic methods for their use in mapping and genetic
engineering experiments, and a final section will focus on the more de-
tailed aspects of selecting endonucleases for generating ends compatible
with subsequent steps of construction.

Definition of Restriction Endonucleases and Nomenclature

Many bacteria contain systems to guard against invasion of foreign
DNA. The cells contain specific endonucleases that make double-strand
scissions in invading DNA unless the DNA has been previously been
modified, usually by the appropriate DNA methylase (but see this volume
[12]). The endonuclease with its accompanying methylase is called a re-
striction modification system, i-3
Three distinct types of restriction modification systems have been
characterized on the basis of the subunit composition, cofactor require-
ments, and type of DNA cleavage. 4 Type I systems are the most complex.
The endonucleases contain three different types of subunits and require
Mg 2÷, ATP, and S-adenosylmethionine for DNA cleavage. Their recogni-
tion sites are complex, and DNA cleavage occurs at nonspecific sites 400-
7000 base pairs from the recognition site. Type III systems are somewhat
less complex. The endonucleases contain only two types of subunits.
Although Mg 2÷ and ATP are required for DNA cleavage as in type I
enzymes, S-adenosylmethionine stimulates enzymatic activity without
being an absolute requirement. DNA cleavage again occurs distal to the
recognition site, but only by 25-27 base pairs. Type II systems are much
simpler. The endonucleases contain only one type of subunit, and Mg 2÷
alone is required for DNA cleavage. DNA cleavage occurs at specific
sites within or adjacent to the enzyme's recognition site. It is therefore
this class of restriction endonucleases that has proved useful to molecular
biologists.
1 W. A r b e r , Prog. Nucleic Acid Res. Mol. Biol. 14, 1 (1974).
2 H. O . S m i t h , Science 2 0 5 , 455 (1979).
3 p. M o d r i c h , Q. Reo. Biophys. 12, 315 (1979).
4 R. Y u a n , Annu. Rev. Biochern. 50, 285 (1981).

Copyright © 1987 by Academic Press, Inc.

METHODS IN ENZYMOLOGY, VOL. 152 All rights of reproduction in any form reserved.
 114 RESTRICTION ENZYMES [11]

A large number of site-specific endonucleases with cofactor require-

ments and cleavage sites similar to known type II restriction endonu-
cleases have been found in many bacteria. For the vast majority of these,
there is no evidence that restriction of foreign DNA occurs in vivo, but by
analogy with bona fide type II restriction endonucleases these have been
termed restriction endonucleases. 5
In 1973, when it became clear that a large number of these endonu-
cleases existed and would be extensively utilized, Smith and Nathans 6
proposed the nomenclature system that is still followed. A three-letter
abbreviation for the parent organism (Hin for Haemophilus influenzae or
Bam for Bacillus amyloliquefaciens), an additional letter if necessary to
identify strain or serotype (Hind or BamH), and then a Roman numeral to
reflect the order of identification or characterization (HindlII or BamHI).
A complete list of named and characterized restriction endonucleases has
been compiled and is updated yearly by Roberts. 7
Because of their practical uses for molecular dissection of DNA, the
restriction endonucleases have been characterized primarily with respect
to their recognition sequences and cleavage specificity rather than their
protein properties. 8 The majority recognize sequences 4-6 nucleotides in
length, but some have now been found with 7- and 8-base recognition
sites. 7 Most, but not all, recognition sites contain a dyad axis of symmetry
and in most cases all the bases within the site are uniquely specified. (The
symmetrical recognition sequence of these endonucleases have been
termed palindromes.) Those with degenerate or "relaxed" specificities
can recognize multiple bases at some positions. Endonucleases with sym-
metrical recognition sites generally cleave symmetrically within or adja-
cent to the recognition site, but those that recognize asymmetric sites
tend to cut at a distance from the recognition site.
In addition to recognition site, the nature of the DNA cleavage is of
paramount importance to uses of restriction endonucleases because the
nature of the ends determines the suitability of the fragments for subse-
quent procedures. All restriction endonucleases cleave their DNA sub-
strates to form 5'-phosphate and 3'-hydroxyl termini on each strand. 9 The

5 R. J. Roberts, CRC Crit. Rev. Biochem. 4, 123 (1976).

6 H. O. Smith and D. Nathans, J. Mol. Biol. 81, 419 (1973).
7 R. J. Roberts, Nucleic Acids Res. 13, r165 (1985).
8 p. Modrich and R. J. Roberts, in "Nucleases" (S. Linn and R. J. Roberts, eds.), p. 109.
Cold Spring Harbor Lab., Cold Spring Harbor, New York, 1982.
9 The one possible exception found to date is NciI which has been reported to liberate 3'-
phosphate and 5'-hydroxyl termini after digestion o f D N A (A. W. Hu and A. H. Marschel,
unpublished observations). However, recent experimentation using highly concentrated
T4 DNA ligase shows that the NciI ends will religate, suggesting they do have intact 5'-
phosphate groups (E. Rosenvold, unpublished observations). Further experimentation
needs to be done.
[11] RESTRICTION ENDONUCLEASES 1 15

breaks can be staggered, generating either 5'-phosphate extensions on

each strand or Y-hydroxyl extensions on each strand, or they can be
opposed, generating "blunt" ends. (See section below on Generating the
Proper Ends for Subsequent Steps.)
Among the more than 400 different endonucleases isolated from bacte-
rial strains many share common specificities. Restriction endonucleases
which recognize identical sequences have been called "isoschizomers"
by Roberts.5 Although the recognition sequences of isoschizomers are the
same, they may vary with respect to the site of cleavage (XrnaI vs
Small°), sensitivity to methylation (MspI vs HpalIll), and in cleavage
rates at various sites (XhoI vs PaeR7 112).

Restriction Endonuclease Selection and Use

Choice of Enodnuclease
The size of DNA fragments to be generated is often a primary consid-
eration in choosing a restriction endonuclease. Fragments of a few hun-
dred bases in length are useful for fine-structure restriction mapping, for
"shotgunning" small DNA fragments into vectors such as MI3 for DNA
sequence analysis, or for generation of more or less random breaks by
partial digestion for site-specific mutagenesis or cloning experiments.13
Fragments of 1-I0 kilobases (kb) are useful for mapping larger DNA
regions and for cloning whole genes complete with introns (from euka-
ryotes) and control sequences into plasmid or h phage vectors. 14 Still
larger fragments 5-50 kb in length are suitable for cloning into cosmid
vectors, cloning whole operons, or genome "walking. ''~5 Finally, tech-
niques currently available to resolve DNA fragments of 1 megabase (1000
kb) or larger can be used to separate and purify whole or partial chromo-
somes for mapping and cloning. 16,17
It is difficult to predict the frequency of occurrence of any restriction
endonuclease cleavage site within a new DNA substrate. Within a per-
fectly random DNA sequence that contained 50% GC, a 4-base recogni-
tion sequence would occur on average every 256 bases, a 6-base sequence
would appear on average at 4-kb intervals, and an 8-base sequence would

l0 S. A. Endow and R. J. Roberts, J. Mol. Biol. 112, 521 (1977).

11 C. Waalwijk and R. A. Flavell, Nucleic Acids Res. 5, 3231 (1978).
12 T. R. Gingeras and J. E. Brooks, Proc. Natl. Acad. Sci. U.S.A. 80, 402 (1983).
t3 j. Messing, this series, Vol. 101, p. 20.
t4 A.-M. Frischauf, this volume [17].
15 A. G. DiLella and S. L. C. Woo, this volume [18].
16 D. Schwartz and C. R. Cantor, Cell (Cambridge, Mass.) 37, 67 (1984).
i7 G. F. Carle, M. Frank, and M. V. Olsen, Science 232, 65 (1986).
116 RESTRICTIONENZYMES [1 1]
occur at 65-kb intervals. These probabilities are, of course, substantially
different for DNA with a skewed GC content (actual DNA can vary from
22 to 73% GCIS). The level of nonrandomness increases at higher levels of
complexity: some di- and trinucleotides occur at distinctly nonrandom
ratios. A well-known example of underrepresentation of a dinucleotide is
the CpG dinucleotide in mammalian DNA. 19In addition, the high level of
repetitive DNA in higher eukaryotes further biases restriction site repre-
sentations. 20,21
Theoretical predictions of site frequency, based on known base com-
position and repetitive sequences, are being compiled for a variety of
organisms. 22 This knowledge should be useful in endonuclease selection,
especially for experiments requiring megabase-sized fragments.

Choice of Restriction Endonucleases

Restriction endonucleases are relatively stable proteins. Their purifi-
cation to homogeneity is not often necessary and optimal reaction condi-
tions are rarely ascertained. A typical restriction reaction will contain, in
addition to the DNA substrate and restriction endonuclease(s), Tris
buffer, Mg 2÷, NaC1, 2-mercaptoethanol, and bovine serum albumin. All
restriction endonucleases require Mg 2÷ as a cofactor, and most are active
in the pH range of 7.2-7.6. The predominant difference among the en-
donucleases is their dependence on ionic strength. Most manufacturers
now include tables in their catalogs that give activities of their endonu-
cleases in standardized buffers with varying ionic strengths. 23,24These are
particularly valuable for molecular biologists who use dozens of endonu-
cleases and may want to do simultaneous double and tripe digests on the
same DNA substrate.
Aside from ionic strength and cation preferences, restriction endonu-
cleases may vary in temperature optima. Most restriction digests are
routinely done at 37°, but a few endonucleases (notably SmaI) prefer a
lower incubation temperature and several, mainly those isolated from
thermophiles (such as TaqI), require much higher temperatures.

is W. M. Normore, H. S. Shapiro, and P. Setlow, in "CRC Handbook of Biochemistry and

Molecular Biology" (G. D. Fasman, ed.), p. 65. CRC Press, Cleveland, Ohio, 1976.
19 j. Josse, A. D. Kaiser, and A. Kornberg, J. Biol. Chem. 237, 864 (1981).
2o K. Tartoff, Annu. Rev. Genet. 9, 355 (1975).
21 M. Botchan, G. McKenna, and P. A. Sharp, Cold Spring Harbor Symp. Quant. Biol. 38,
383 (1974).
22 M. Nelson and M. McClelland, in "Gene Amplification and Analysis" (J. G. Chirikjian,
ed.), Vol. 5 (in press).
23 R. D. Wells, R. D. Klein, and C. K. Singleton, in "The Enzymes" (P. D. Boyer, ed.), 3rd
ed., Vol. 14, p. 157. Academic Press, New York, 1981.
24 R. Fuchs and R. Blakesley, this series, Vol. 100, p. 3.
[11] RESTRICTIONENDONUCLEASES 117
Units of endonuclease activity are usually measured with h DNA as a
substrate using the (nonstandardized) buffer recommended by the manu-
facturer for the particular endonuclease. Endonuclease activity also var-
ies greatly with the DNA substrate, and activity for each site can be
modified by the neighboring sequences. The classic example of differen-
tial cleavage rate is the nearly 10-fold difference in reaction rate observed
between two EcoRI sites in h DNA. 25 A greater than 50-fold difference in
cleavage rates exists among sites for both NarI and NaeI on pBR322
DNA. 26 Furthermore, PaeR7 I does not cut one of its canonical sites on
adenovirus 2 DNA. j2 Rate variability can be different for different iso-
schizomers; XhoI does not show the same variability as PaeR7 1 on Ad 2
DNA. 12And, in most cases, such rate differences have not been systemat-
ically investigated.
Therefore, titration of previously untried endonucleases on one's par-
ticular DNA substrate under reaction conditions to be used is recom-
mended, especially if the reaction buffer is different from the manufactur-
er's prescribed buffer. If it is desirable to dilute the restriction
endonuclease before use, this should be done in the manufacturer's sug-
gested dilution buffer rather than assay buffer to maintain activity.
The activity of most restriction endonucleases is also adversely af-
fected by the proximity of the recognition site to the end of a DNA
molecule. 24 This parameter is especially important when one is trying to
cleave at adjacent sites within a linker region of a cloning vector.
A word should be mentioned about star activity)7 This phenomenon
was first observed and has been best studied with EcoRI endonuclease. 28
Under "altered" reaction conditions (including low ionic strength, high
pH, Mn 2÷ substitution of Mg2+, presence of organic solvents such as
glycerol or dimethyl sulfoxide, and high endonuclease concentration),
EcoRI can be shown to cleave at noncanonical sites (within sequences
other than GAATTC28-32). These sites are not random, however. They all

25 M. Thomas and R. W. Davis, J. Mol. Biol. 91, 315 (1975).

26 G. Wilson, D. Comb, L. Greenough, and 1. Schildkraut, unpublished observations.
27 p. Modrich, CRC Crit. Rev. Biochem. 13, 287 (1982).
28 B. Polisky, P. Greene, D. E. Garfin, B. J. McCarthy, H. M. Goodman, and H. W. Boyer,
Proc. Natl. Acad. Sci. U.S.A. 72, 3310 (1975).
29 Recognition sequences are written 5' ~ 3'; one strand only is given. Other nucleic acid
nomenclature includes the following conventions. Only bases are indicated (deoxyribo-
nucleotide residues are often not differentiated from their ribonucleotide counterparts).
The presence of phosphodiester bonds is understood (internal phosphates are usually not
written). Oligonucleotide sequences which terminate in either a 5'- or 3'-phosphate mono-
ester are explicitly stated, e.g., p(T)6 means a hexamer of thymidylic acid residues with a
5'-terminal phosphate. A vertical arrow ( ~ ) indicates a cleavage site usually generating a
terminal 5'-phosphate (see ref. 9 for a possible exception).
30 M. Hsu and P. Berg, Biochemistry 17, 131 (1978).
118 RESTRICTIONENZYMES [1 1]

have the sequence AATT. Even under star conditions, canonical EcoRI
sites (GAATTC) are cleaved first and a hierarchy exists as to the order in
which secondary sites are cleaved. ",34 It is extremely difficult to generate
complete star digests, so star activity is often more problematic than
useful.
Under star conditions many other restriction endonucleases have been
reported to have secondary star activity. These include HindlII, HhaI,
BsuI, XbaI, SalI, PstI, BamHI, and SstI. 24 However, since no other
endonuclease has been purified and characterized as well as EcoRI, it is
often difficult to determine whether the secondary activity is, in fact,
associated with the aforementioned endonucleases. It is possible that
some star activities are actually traces of contaminating second endonu-
cleases. To avoid star activity, manufacturers recommend keeping glyc-
erol and endonuclease concentrations low, ionic strength high, and di-
gestion times as short as possible.
A thorough guide for the preparation of DNA substrates and use of
restriction endonucleases is given in a previous volume of this series. 24 It
also includes a useful troubleshooting appendix which addresses the types
of problems most frequently encountered in the use of restriction endonu-
cleases.

Mapping
Most applications of recombinant DNA technology are facilitated by
the generation of a physical map of restriction sites within the DNA being
studied. A number of strategies can be used to construct these maps, and
a combination of several are often necessary to obtain maps that are
sufficiently accurate and detailed.
To start a map, one usually digests the DNA of interest with a series of
single restriction endonucleases. Products of the endonuclease digestion
are resolved by analytical gel electrophoresis on agarose or polyacryl-
amide g e l s . 35 Visualization of the DNA is achieved by ethidium bromide
staining followed by ultraviolet light monitoring, or by autoradiography of

31 T. I. Tikchonenko, E. E. Karamov, B. A. Zavizion, and B. S. Naroditsky, Gene 4, 195

(1978).
3z M. Nasri and D. Thomas, Nucleic Acids Res. 14, 811 (1986).
33 C. P. Woodbury, O. Hagenbuchle, and P. H. yon Hippel, J. Biol. Chem. 225, 11534
(1980).
34 R. C. Gardner, A. J. Howarth, J. Messing, and R. Shepard, DNA 1, 109 (1982).
~ R. C. Ogden and D. A. Adams, this volume [8].
[11] RESTRICTIONENDONUCLEASES 119

radioactively labeled DNA substrates. To estimate size, a standardization

curve is first generated by plotting the relative mobility of DNA fragments
of known molecular weight against the log of their molecular weights.
Then the mobility of each DNA fragment in question is fitted to the
c u r v e . 36 After resolving the restriction products on a gel, it is easy to
determine how many cleavage sites occur on the DNA in question. On a
linear DNA molecule, the first cleavage will yield two fragments and then
each additional site adds one more fragment. When starting with a circu-
lar DNA substrate, however, the first cleavage will linearize and it will
remain one fragment; then each additional cut will yield one more
fragment.
It is difficult to size circular DNA substrates accurately on a gel.
Supercoiled molecules migrate aberrantly fast and differ in mobility from
nicked circles. It is usually more accurate to linearize the DNA before
attempting to estimate its size.
The first step in ordering the DNA fragments for a map is generating a
limit digest with the endonuclease of choice. A characteristic feature of
limit digests is that all cleavage products are present in equimolar
amounts. Therefore, when uniformly labeled DNA is cleaved the amount
of radioactivity in each limit product is directly proportional to its size.
Similarly, when cleaved DNA is stained with ethidium bromide and visu-
alized under ultraviolet light, the intensity of the ultraviolet absorption by
each fragment is directly proportional to its size.
The most frequent problem encountered in restriction site mapping is
identifying partial digestion products and double bands (or "doublets")
on a gel. In both cases a band of variant intensity is produced. "Partials"
usually appear as bands that are too light (or underlabeled); if one adds up
the total DNA length of all the fragments one gets too large a value.
Sometimes partials can be eliminated by digesting the DNA substrate for
a longer time with more endonuclease.
Doublets occur when two products are similarly sized and therefore
unresolved on the gel. The doublet band appears more intense (or radio-
active) than its neighbors; fragments add up to too small a value. Some-
times double bands can be separated on a higher resolution gel. Often
their presence can be verified by digesting one member away with a
second endonuclease.
Another problem that might occur when attempting to generate a re-
striction map is that of DNA fragments migrating aberrantly on a gel. For
example, on polyacrylamide gels, high-molecular-weight fragments with
high GC contents migrate considerably faster than fragments with low GC

36 K. J. Danna, this series, Vol. 65, p. 449.

120 RESTRICTIONENZYMES [11]
contents. 37 Anomalous fragment mobilities can occur on both polyacryl-
amide and agarose gel matrices, and the problem is particularly severe
with short DNA fragments. The sizing of small fragments is more accu-
rately done on denaturing gels. 38
After assessing the fragment profile for a number of restriction en-
donucleases on the DNA substrate of interest, a whole variety of mapping
techniques may be employed. Most commonly the DNA is then digested
with combinations of endonucleases whose single profiles have been char-
acterized. Also individual fragments may be isolated and then digested
with various other endonucleases.
Another relatively easy and popular mapping method involves labeling
a DNA fragment specifically at one terminus and then partially digesting
the fragment in a series of reactions with various endonucleases that cut
the substrate several times. 39 In a variation of this method, DNA frag-
ments generated by partial endonucleolytic digestions can be isolated and
then individually digested to completion to establish fragment order. 34
Finally, a ~ N A fragment to be mapped can be digested to different
extents with a processive double-stranded exonuclease (most commonly
Bal 31). DNA isolated at each time point can then be digested by one or
more restriction endonucleases. Fragments disappear in the order in
which the restriction site appeared on the DNA. 4°
Using a combination of mapping methods and a good number of en-
donucleases it is relatively easy to generate a restriction map that is
detailed and accurate down to the 100-200 base pair level.

Generating the Proper Ends for Subsequent Steps

Table I contains a list of commercially available restriction endonu-
cleases. They have been grouped on the basis of recognition and cleavage
site, and they will be discussed mainly in these groups. Group 1 endonu-
cleases are those that cut defined, palindromic sequences. Group 2 en-
donucleases are those that cut interrupted palindromes. Group 3 contains
endonucleases recognizing "relaxed" specificities. Group 4 endonuclease
are those that recognize nonpalindromic sequences.
Within each group, class A endonucleases cleave to produce 5' exten-
sions, class B to produce 3' extensions, and class C, blunt ends.

37 R. S. Zeiger, R. Salomon, C. W. Dingman, and A. C. Peacock, Nature (London), New

Biol. 238, 65 (1972).
3a T. Maniatis, A. Jeffrey, and H. Van de Sande, Biochemistry 14, 3787 (1975).
39 H. O. Smith and M. L. Birnstiel, Nucleic Acids Res. 3, 2387 (1976).
4o R. J. Legerski, J. L. Hodnett, and H. B. Gray, Jr., Nucleic Acids Res. 5, 1445 (1978).
[11] RESTRICTION ENDONUCLEASES 121

TABLE I
COMMERCIALLY AVAILABLE RESTRICTION ENDONUCLEASES

Endonuclease Recognition sequence

Group 1. Endonucleases cutting within palindromic sequences

A. 5' Extensions
Tetrameric recognition site
1. ItinPI G J, C G C
2. HpaII, MspI C J, C G G
3. MboI, Sau3AI,
NdeI, NdeII + G A T C
4. TaqI T ~ C G A
5. MaeI C $ T A G
6. MaeII A J, C G T
Hexameric recognition site
7. NarI G G ~ C G C C
8. XmaI, XcyI C J, C C G G G
9. BspMII T J, C C G G A
10. XmaIII, EagI C $ G G C C G
I I. BssHII G $ C G C G C
12. MluI A ~ C G C G T
13. EcoRI G ~ A A T T C
14. NdeI C A $ T A T G
15. BamHI, BstI G ~ G A T C C
16. BglII A $ G A T C T
17. BclI T $ G A T C A
18. ClaI, BanIII A T J, C G A T
19. XhoI, PaeR7I C J, T C G A G
20. SalI G J, T C G A C
21. AsuII T T $ C G A A
22. ItindIII A $ A C G T T
23. NheI G $ C T A G C
24. SpeI A ~ C T A G T
25. XbaI T $ C T A G A
26. AvrII C J, C T A G G
27. NcoI C $ C A T G G
28. Asp718 G $ G T A C C
Octameric recognition site
29. NotI G C ~ G G C C G C
B. 3' Extensions
Tetrameric recognition site
1. HhaI, CfoI G C G ~, C
2. NlaIII C A T G ],
Hexameric recognition site
3. SacII, SstII C C G C ~ G G
4. ApaI G G G C C ~, C
5. SphI GC AT G ,1, C

(continued)
122 I~STP.ICTION ENZYMES [11]

TABLE I (continued)

Endonuclease Recognition sequence

6. KpnI GG TAC ~C
7. PvuI, XorII, RspI CG AT C ~,G
8. SacI, SstI GA GC T ~ C
9. PstI CT GC A ~ G
10. NsH AT GC A ~ T
11. AatII GA CG T ~ C
C. Blunts
Tetrameric recognite site
1. FnuDII, Thai CG ~C G
2. HaeIII, PalI GG ~,C C
3. AluI AG J,C T
4. RsaI GT ~,A C
5. DpnI G m*A ~ T C
Hexameric recognition site
6. NaeI GC C G G C
7. Sinai C C C G G G
8. StuI, AatI A G G C C T
9. Bali T G G C C A
10. NruI T C G C G A
11. FspI, MstI T G C G C A
12. PvulI CA G C T G
13. SnaBI T A C ,~ G T A
14. ScaI A G T A C T
15. EcoRV G A T A T C
16. HpaI GT T A A C
17. SspI A A T A T T
18. DraI T T T A A A
Group 2. Endonucleases cutting interrupted palindromes
A. 5' Extensions
1. DdeI C T N A G
2. Fnu4HI G C N G C
3. Hinfl G A N T C
4. MaeIII G T N AC
5. Sau96I G G N C C
6. ScrFI C C ,~ N G G
7. BstEH G G T N ACC
8. MstII, CvnI, Saul C C T N A GG
9. T t h l l l I G A C N N N G T C
B. 3' Extensions
1. DraIII C A C N N N J, G T G
2. BglI G C C N NN N N G G C
3. BstXI C CANN NN N N T G G
4. SfiI G GCCN NN N N G G CC
[11] RESTRICTION ENOONUCLEASES 123

TABLE I (continued)

Endonuclease Recognition sequence

C. Blunts
1. N l a l V G G N ~ N C C
2. Xmnl, Asp700 G A AN N J, N N T T C
Group 3. Endonucleases recognizing sequences with degeneracies
A. 5' Extensions
Pentameric recognition
site
1. EcoRII ~, C C A/T G G
2. BstNI, ApyI C C J, A/T G G
3. Avail G ~ G A/TC C
4. NciI C C ~ C/GG G
Hexameric recognition site
5. XholI Pu ~ G A T C Py
6. AhalI G Pu ~, C G Py C
7. AvaI, NsplII C ~ Py C G Pu G
8. BanlI G ~ Pu G C PyC
9. BanI G J, G Py Pu C C
10. AccI G T J, A / C G / T A C
11. StyI C ~ C A/T A/T G G
Heptameric recognition site
12. EcoOl09, DralI Pu G ~ G N C C Py
13. PpuMI Pu G ~ G A/T C C Py
14. RsrlI C G ~ G A/TC C G
B. 3' Extensions
Hexameric recognition site
1. HaelI PuG C G C ~ Py
2. HgiAI GT/A G C T/A ~ C
3. Bsp1286, N s p l I G G / A / T G C C/A/T ~ C
C. Blunts
1. HinclI, HindlI G T Py ~ Pu A C
Group 4. Endonucleases recognizing nonpalindromes
A. Cutting within sequence,
3' extension
1. BsmI GAA TG C N~,
CTT AC 1' G N
B. Cutting away
5' Extensions
1. HphI GGT G A N8 $
CCA C T N7 1'
2. MbolI GAA G A N8
CTT C T N7 1'

(continued)
124 RESTRICTIONENZYMES [ 11 ]
TABLE I (continued)

Endonuclease Recognition sequence

3' Extensions
1. BbvI GCA GC N8
CGT CG N12 1"
2. FokI GGA T G N9
CCT AC N13 1'
3. HgaI GAC GC N5 ,~
CTG CG N10 ~'
4. SfaNI G C A T C N5 $
CGT AG N9 1'
5. BspMI AC C T GC N4
TGGA CG N8 1'
C. Blunts
1. Mnll C C T C N7
GG A G N7 ~'

Ligation
In genetic engineering one of the most useful features of DNA cleaved
by restriction endonucleases is the ability to ligate different fragments
together at their common restriction sites. Only the sequence of the DNA
within the single-stranded region (if any) of the restriction cleavage site
has to be complementary for ligation to occur. The surrounding DNA
sequence does not affect ligation. Many factors affect the ease of ligation
of DNA molecules but an overriding one is whether protruding or "blunt"
ends are generated. In addition, the length of the protruding end and
stability of the hydrogen-bonded structure formed are important. Any
contaminating phosphatase, endonuclease, or exonuclease activity in a
restriction endonuclease will also markedly reduce the ability of the ends
generated to be ligated.
In Table I, group 1 endonucleases are those that cut defined (nonde-
generate) sequences. Each such endonuclease generates ends that are
compatible with each other. Within this group, class A and class B en-
donucleases producing 5' and 3' extensions, respectively, generate ends
that ligate, in general, with high efficiency. In both groups, as a rule, 4-
base extensions ligate better than 2-base extensions, and extensions with
GC pairs ligate more readily than those with AT. 41 Group C endonu-
cleases produce blunt ends, with no extensions. Although they are advan-

4~ N. P. Higgins and N. R. Cozzarelli, this series, Vol. 68, p. 50.

[11] RESTRICTIONENDONUCLEASES 125

tageous in the sense that any endonuclease within this class produces
ends that are compatible with all other members of the class, blunt ends
are much more difficult to ligate than either type of extension. The ligation
reaction usually requires 20 to 100 times more T4 DNA ligase and higher
DNA concentrations for the ends to rejoin efficiently. 42
Within class B there is the interesting case of PvuI. This endonuclease
cuts at the sequence CGAT ~ CG, leaving a 2-base 3' extension. The sites
religate at high efficiency unless the DNA has come from a dam +bacterial
host. N6-Methyladenine within the sequence CGmeATCG does not inter-
fere with PvuI cleavage, but it inhibits ligation of the site. 43 This is the
only known case where methylation affects ligation but not cleavage. The
effect appears to be mediated by an altered melting temperature of the AT
base pair. 44
Group 2 endonucleases are those that cut interrupted palindromes.
Ligation of fragments with 3' or 5' extensions generated by those endonu-
cleases is usually not possible due to the heterogeneity of sequence within
the interrupted part of the site. Blunt cutters, like NlaIV and XmnI which
cut symmetrically within the interruption, produce DNA fragments that
can be rejoined with regeneration of the recognition site.
Group 3 endonucleases are those recognizing "relaxed" specificities.
Only a subset of such fragments will be compatible and able to ligate. The
exceptions are XhoII, AhaII, and HaeII, where the degeneracies lie out-
side the protruding ends, and HincII, which generates symmetrical, blunt,
and therefore compatible ends.
Within groups 2 and 3, a word should be said about ScrFI, Fnu4HI,
BstNI, and Tthl 1lI. These endonucleases all cleave to generate a single
base extension. These are extremely difficult to ligate--much more so
even than blunt ends. Such DNA is best treated to remove or fill in the
base before subsequent steps. 45
Finally, group 4 endonucleases are those recognizing nonpalindromic
sequences. All contain at least one degeneracy within the cleavage site,
and most cleave away from their recognition site. Therefore it is not
usually possible to get the generated fragments to religate. If, however,
the ends of the fragments are treated to generate flush ends, the recogni-
tion sites are left intact and always suitable for recutting. The exception
within this group is BsmI, where cleavage occurs within the recognition
site on one strand.

42 V. Sgaramella, Proc. Natl. Acad. Sci. U.S.A. 69, 3389 (1972).

43 Specific methylation sites are designated by a superscript " m e " 5' to the modified base.
44 j. j. Sninsky and M. Myers, unpublished observations.
45 I. Schildkraut, unpublished observations.
126 RESTRICTION ENZYMES l l 1]

Compatible Ends
Often it is desirable to join DNA cut by two different restriction en-
donucleases together. This is possible with a number of different restric-
tion endonucleases that generate the same type of ends. However, in
general, when this is done, a new chimeric site is formed that is no longer
cleavable by the original endonucleases used.
The largest group of compatible ends are those generated by any en-
donuclease whose cleavage leaves a blunt end. Endonucleases generating
other families of compatible ends are listed in Table II.

Other Subsequent Steps

Many manipulations other than directly ligating restriction fragments
can be carried out on cleaved DNA fragments. The types of manipula-
tions possible depend on the ends that have been generated.
5' Extensions. The 3'-OH recessed ends can readily be filled in to form
blunt ends or to carry label by the addition of Klenow enzyme and the
appropriate deoxynucleotides.46 Sometimes it is desirable to cleave a re-
striction site that produces a 5' extension, fill in with Klenow, and then
religate the site. This procedure will often generate a new restriction site.
A list of sites that can be generated in this manner is given in Table III.
The 5' ends are also readily labeled by first removing the phosphate
group with alkaline phosphatase and then replacing the phosphate moiety
with [y-32p]ATP and T4 polynucleotide kinase. 46 Alternatively, 5' exten-
sions can be readily removed by treatment of the DNA with $1 or mung
bean nucleases. 47,48
Nucleotides from the recessed 3'-hydroxyl end of a 5'-phosphate ex-
tension may be sequentially removed by treatment with exonuclease 111. 46
3' Extensions. There is no counterpart of Klenow treatment that can
be used to fill in 5' recessed ends. The 3' extension can be digested away
by treatment with T4 DNA polymerase in the absence of deoxynu-
cleotides, or Klenow enzyme in the presence of deoxynucleotides. Alter-
natively, like 5' ends, the 3' protrusions can be digested away with mung
bean or $1 nuclease. 3' Extensions are also excellent substrates for tailing
with terminal transferase. 49This can often result in the regeneration of the
restriction recognition site; when using the PstI or SphI sites on pBR322,
G and C tailing respectively, will restore the site on the recombinant

46 F. Cobianchi and S. H. Wilson, this volume [10].

47 V. M. Vogt, Eur. J. Biochern. 33, 192 (1973).
'~ D. Komalski, W. D. Kroeker, and M. Laskowski, Sr., Biochemistry 15, 4457 (1976).
49 W. H. Eschenfeldt and S. L. Berger, this volume [37].
[11] RESTRICTION ENDONUCLEASES 127

TABLE II
ENDONUCLEASESGENERATINGENDS COMPATIBLE
FOR LIGATION

Endonucleases Cleavage site

5' CG family
1. MspI, HpalI C J, C G G
2. TaqI T ~, C G A
3. HinPI G ~ C G C
4. MaelI A J, C G T
5. AhalI GPu ~, C GPyC
6. ClaI, BanIII AT J, C G A T
7. NarI GG ~, CGC C
8. AsuII T T ~, C G A A
5' GATC family
1. MboI, Sau3AI, NdeII ~, G A T C
2. XholI Pu ~, G A T C Py
3. BamHI, BstI G ~, G A T C C
4. BglII A ~, G A T C T
5. BclI T ~ G A T CA
5' CTAG family
1. NheI G J, C T A G C
2. XbaI T J, C T A GA
3. SpeI A ~, CTA GT
4. AvrlI C ~, C T A GG
5' CCGG family
1. BspMII T ~ C C G G A
2. XmaI C $ CCG GG
3. NotI G C ], C C G G G C
5' TCGA family
1. XhoI, PaeR7I C ], T C G A G
2. Sail G $ T C G A C
3' CATG family
1. NIalV C AT G ~,
2. SphI G C AT G ~, C
3' TGCA family
1. NsiI A T GCA ~,T
2. PstI C T G C A J, G

plasmid. To r e m o v e nucleotides from the 5' end o f a recessed 5' strand,

o n e treats with h e x o n u c l e a s e . 46
Blunt Ends. All flush ends are suitable for ligation to one another. In
addition, o n e can easily create a n e w restriction site at these ends by
ligating phosphorylated linker or adaptor molecules to their Y - O H ends
before ligating the fragments together, s°

50 R. Wu et al., this volume [38].

 128 RESTRICTION ENZYMES [11]

TABLE III
5' EXTENSION FILL-IN SITES GENERATED

New site
Endonucleases Cleavage site Fill in generated

1. MboI, Sau3AI, N d e l I J, G A T C G A T C G A T C ClaI

2. BamHI, BstI G ~, G A T C C G G A T C G A T C C ClaI
3. BclI T J, GAT CA T GA TCGAT C A ClaI
4. BgllI A J, GAT CT A GA T CGAT C T ClaI
5. XholI Pu ~, G A T C Py PuGA T C GAT C Py ClaI
6. HinPI G ~ C G C G C G C G C BssHI
7. NarI GG J, C G C C G G C G C G C C BssHI
8. BssHI G ~ C G C G C G C G C G C G C G C BssHI
9. MluI A ~ C G C G T A C G C G C G C G T BssHI
10. AvrII C ~ C T A G G C C T A G C T A G G AluI
11. NheI G ,~ CTA GC G CT AGCTA G C AluI
12. SpeI A ~, CTA GT ACT AGCTA G T AluI
13. XbaI T J, CTA GA T CT AGCTA G A Alul
14. TaqI T J, C G A T C G C G A NruI
15. ClaI, BanIII AT ~ C G A T AT C G C G A T NruI
16. AsuII TT ~ C G A A T T C G C GA A NruI
17. BspMII T $ GGC CA T GG CCGGC C A NaeI
18. XmaIII, EagI C J, G G C C G C G G C C G G C C G NaeI
19. N o d GC $ GGC CG C GC GG CCGGC C GC NaeI
20. XhoI, PaeRTI C ~ T C G A G C T C G A T C G A G PvuI
21. SalI G ~ T C G AC G T C GAT C G A C PvuI
22. HpaII, MspI C ~ C G G C C G C G G SaclI
23. AhaII G P u ,[ C G P y C CPuCGCGPyC FnuDII
24. X m a I C J, C C G G G C C C G G C C G G G XmaIII
25. MaeII A ~ C G T A C G C G T MlulI
26. HindIII A ~, A C G T T A A G C T A G C T T NheI
27. Asp718 G ~ GTA C C G GT ACGTA C C SnaBI
28. NcoI C ~ C AT G G C CA T GC A T G G NsiI
29. EcoRI G J, A A T T C G A A T T A A T T C XmnI
30. MaeI C ~ TAG CTATAG (none)
31. NdeI CA ~, T A T G C AT AT A T G (none)
[11] RESTRICTIONENDONUCLEASES 129

Some new uses are also being made of the more exotic "degenerate"
and "nonpalindromic" endonucleases. Neither group is heavily used in
typical recombinant DNA work because of the heterogeneity at their
cleavage sites. But in the case of the degenerate endonucleases, new
cleavage specificities are being elicited by first methylating the substrate
with a methylase of overlapping specificity before restriction digestion.
For example, pBR322 is cleaved at two positions by H i n c I I endonuclease
(GTPyPuAC). 51 However, if the DNA is first methylated by M. TaqI
(TCGA), one site is blocked, resulting in cleavage at a single unique site in
the ampicillin gene (GTCAAC). 52
Szybalski has recently developed a method to cleave cloned sin-
gle-stranded DNA at any site desired using the nonpalindromic en-
donucleases. An oligonucleotide adaptor is synthesized that has a hair-
pin containing the recognition site for the endonuclease and a long
single-stranded region. The single-stranded region contains the sequence
complementary to the site to be cleaved, at a precise distance from the
recognition site. The oligonucleotide is hybridized to the single-stranded
DNA substrate and then treated with endonuclease, resulting in cleavage.
This method has been successfully used with F o k I endonuclease. 53,54

Summary
It is clear that we have still not exhausted all the restriction endonu-
clease specificities to be found in nature. Recently discovered B s m I is the
first endonuclease recognizing a nonpalindromic sequence that cleaves
within the site. 55 Certainly other endonucleases belonging to this class will
soon be discovered. More endonucleases are now being sought that rec-
ognize longer recognition sequences, because large fragments can now be
readily separated by pulse-field electrophoresis. New sources of endonu-
cleases are also being found; for example, a group of viruses that grow on
Chlorella algae produce type II-like site-specific endonucleases. 56
As the number and variety of known restriction endonucleases in-
crease, the number and variety of applications keep pace. There is still no
end in sight.

5~py indicates pyrimidines;Pu, purines.

52M. Nelson, C. Christ, and I. Schildkraut,Nucleic Acids Res. 12, 5165 (1985).
53W. Szybalski,Gene 40, 169 (1985).
54A. J. Podhajskaand W. Szybalski,Gene 40, 175 (1985).
55C. Christ and D. Ingalls, unpublishedobservations.
56y. Xia, D. E. Burbank, L. Uher, D. Rabussay, and J. L. van Etten, Mol. Cell. Biol. 6,
1430 (1986).