ABSTRACT
Based on a high-performance liquid chromatographic pump, we have built a device that
allows recirculation of DNA through a 63-[mu]m orifice with ensuing fractionation to a minimum fragment size of
~300 base pairs. Residence time of the DNA fragments in the converging flow created by a sudden contraction was found to be sufficiently long to allow extension of the DNA molecules into a highly extended conformation and, hence, breakage to occur at midpoint. In most instances, 30 passages sufficed to obtain a narrow size distribution, with >90% of the
fragments lying within a 2-fold size distribution. The shear rate required to achieve breakage was found to be inversely proportional to the 1.0 power of the molecular weight. Compared with a restriction
digest, up to 40% of all fragments could be cloned directly, with only marginal
improvements in cloning efficiency having been observed upon prior end repair
with Klenow, T4 polymerase or T4 polynucleotide kinase. Sequencing revealed a
fairly random distribution of the fragments.
Several methods have been described for random fragmentation of DNA. These
methods, often used for library preparation and subcloning prior to DNA
sequence analysis, include passage through the small orifice of a hypodermic
needle (
1
-
3
) or of a high pressure spray atomizer (
4
), nebulization (
5
), sonic treatment (
6
), stirring in a blender (
7
,
8
), as well as partial digestion by restriction endonucleases (
9
) or treatment with DNase I in the presence of manganese ions (
10
). While all of these methods have been used successfully to prepare random DNA
fragments for further manipulation and analysis, each has difficulties and
limitations. Shearing by passage through a hypodermic needle or by stirring
fails to generate fragments small enough (1-2 kb) for efficient cloning into M13 vector. Nebulization requires large
volumes of DNA solution. Further, leaks in the nebulizer are common and almost
unavoidable (
5
). Sonication, on the other hand, is difficult to reproduce, requires relatively
large amounts of DNA and yields a relatively broad size distribution and,
hence, a low yield of fragments useful for cloning and sequencing (
5
). Low cloning efficiencies have also been attributed to damage inflicted on DNA
by the action of hydroxyl radicals which are known to arise as a result of
thermal dissociation of water induced by pressures due to ultrasonic cavitation
(
11
). Methods not based on the generation of hydrodynamic shear, such as the use of
restriction enzymes, also have a number of disadvantages. First, some regions
of DNA sequence have very few restriction sites and would be underrepresented
in the resulting clone banks. Secondly, several different restriction enzymes
are necessary to obtain fragments which overlap each other properly to complete
the DNA sequence analysis. Thirdly, many restriction fragments are quite small
and, hence, yield only very little information upon sequencing. DNase I
overcomes some of these difficulties, as it cleaves with very little sequence
specificity. But due to the wide size distribution of the resultant fragments,
the yield of fragments having the appropriate length for cloning and subsequent
sequence analysis tends to be small.
In an effort to minimize template DNA preparation tasks and simplify the
preparation of libraries for shotgun sequencing, we developed a method
employing an HPLC pump with a manifold valve, that allows recirculation of the
DNA that is to be sheared. Major advantages of the method are a greater
randomness of fragmentation sites and a >90% yield of fragments over a 2-fold size range that can be easily pre-selected by varying the flow-rate.
[lambda]DNA (Cat. No. 25250) and a 1 kb DNA ladder (Cat. No. 15615) were purchased from GIBCO BRL/Life Technologies (Gaithersburg, MD, USA). Cosmid DNA from human Y chromosome was isolated from the
original
Escherichia coli
kindly provided by Dr Vollrath (Stanford University, Stanford, CA, USA). The
shearing buffer was identical to the buffer used to store [lambda]DNA, and contained 10 mM Tris-HCl, pH 7.4, 5 mM NaCl and 0.1 mM Na
2
EDTA (all from Sigma, St Louis, MO, USA). The same source of reagents was used
to prepare TE-buffer (10 mM Tris-HCl, pH 8.0, 1 mM EDTA) and 5* gel loading solution (25% w/v Ficoll, 0.05% w/v bromophenol
blue, 0.05% w/v xylene cyanole FF).
Shearing was carried out in a Waters Model 510 HPLC pump (Marlborough, MA), that
had been equipped with a three port inlet manifold assembly (Part No. 25712,
Waters). Polyetheretherketone (PEEK) tubing (1/16'' O.D.) and precut stainless steel capillary tubing (1/16'' O.D.) of inner diameters ranging from 0.0025 to
0.04'', as well as various unions and tees with the appropriate 1/16'' O.D. fittings were obtained from either Rainin
Instrument Company (Woburn, MA) or Upchurch Scientific (Oak Harbor, WA). The
shearing buffer was vacuum filtered through a disposable sterile bottle top
filter with a 0.22-[mu]m cellulose acetate membrane (Part No. 25970-33, Corning Glass Works, Corning, NY, USA). In addition, the buffer was
continuously degassed by passing it through two channels in series of a flow-through degasser placed in-line between the reservoir and the pump inlet (DG-1310, Rainin Instrument Company). As the installation of
tubings of various inner diameters results in changes in back pressure and,
consequently, in the piston volume to be compressed before delivery can start,
pressure-dependent deviations from the preset flow-rate do occur. Actual flow-rates for a given combination of tubings were determined by
collecting the solvent in a volumetric flask and noting the time. Subsequently, the volume flow/min was calculated and given in the figure legends. DNA
concentrations were determined by means of a Spectronic 1001 split-beam spectrophotometer (Bausch & Lomb Inc., Rochester, NY, USA).
Buffered solutions of 0.5-80 [mu]g DNA were loaded with a Hamilton microliter syringe (Reno, NE, USA)
into the 50- or 100-[mu]l sample loop of a seven-port sample injection valve (Rheodyne, Model 7725i,
Cotati, CA, USA). Following the switching of the valve from LOAD to INJECT,
with the manifold valve in the RECYCLE position, mobile phase was flushed
through the loop for 30 s at a flow-rate of 0.5 ml/min. Then it was returned to the LOAD position, in order to
keep the internal volume of the device as small as possible. Finally, the flow-rate was adjusted to obtain fragments of the desired size range. After a
given period of time, the manifold valve was switched to the COLLECT position
and 0.7-1.0 ml of sheared DNA was collected. The eluate was precipitated with 0.1
vol 3 M sodium acetate, pH 7.0 and 1 vol isopropanol. The sample was mixed,
incubated at -30oC for at least 90 min and spun at 16 000
g
for 30 min. The pellet was washed twice with 75% ethanol, dried under vacuum
and, finally, resuspended in 16 [mu]l of TE buffer and 4 [mu]l of 5* sample loading solution. Between shearing runs of DNA derived
from the same source, the device was flushed with several milliliters of
shearing buffer, the manifold valve being in WASTE position. Between the
shearing of different DNA samples, the device was rinsed first with 0.2 M HCl
followed by 0.2 M NaOH to hydrolyze any remaining DNA in the device and, hence,
to avoid carry-over. Care has to be taken to completely flush out the sodium hydroxide
solution with shearing buffer prior to the next injection. Otherwise,
denaturation of the DNA will result in a second band of fragments exhibiting
roughly half the expected size. This band contains single-stranded DNA molecules that break twice as easily as double-stranded DNA (
12
,
13
).
Sheared DNA samples were separated on a 1% agarose slab gel (SeaKem GTG Agarose,
FMC BioProducts, Rockland, ME) in 45 mM Tris-borate buffer, pH 8.3, containing 1 mM EDTA. Separations were carried out
in a horizontal agarose submarine unit (Model MGU-200T, C.B.S. Scientific Co., Del Mar, CA, USA) at a constant voltage of ~115 V. Size of the sheared DNA fragments was computed by digitizing
agarose slab gels with the IS-1000 Digital Imaging System from Alpha Innotech Corp., San Leandro, CA, USA. The image was calibrated with DNA fragments of known size, and then a cubic spline interpolation function was used to
map digitized coordinates to fragment length in base pairs.
Cloning efficiency was determined by two methods: as the ratio of tetracycline-sensitive (tet
s
) to ampicillin-resistant (amp
r
) transformants arising when DNA fragments were blunt-end ligated into the unique
Eco
RV site within the tet
r
gene of pBR322; or alternatively, as the efficiency with which amp
r
transformants were obtained when increasing quantities of the DNA fragments
were ligated to alkaline phosphatase-treated linearized pBR322.
Comparisons were made between the cloning efficiency of untreated HPLC-sheared [lambda]DNA fragments, those treated with Klenow [DNA fragments at 0.2 [mu]M, in 50 mM Tris-HCl, pH 7.5, 10 mM MgCl
2
, 1 mM DTT, 50 [mu]g/ml BSA and 25 [mu]M each dNTP with 5 U Klenow (United States Biochemicals, Cleveland, OH)
in a 20 [mu]l volume for 15 min at 30oC], T4 DNA polymerase [DNA fragments at 0.2 [mu]M, in 33 mM Tris-acetate, pH 7.9, 66 mM potassium acetate, 10 mM magnesium acetate, 0.5 mM DTT, 0.1 mg/ml BSA and 25 [mu]M each dNTP with 6 U T4 DNA polymerase (United States
Biochemicals) for 5 min at 37oC], or T4 polynucleotide kinase (see below), and [lambda]DNA fragments obtained from
Hin
cII digestion (New England Biolabs, Beverly, MA). [lambda]DNA fragments obtained from
Pst
I and/or
Eco
47I digestion (New England Biolabs) with/without Klenow or T4 DNA polymerase
treatment were concurrently cloned as controls for efficient end-repair. T4 polynucleotide kinase (5-10 U/20 [mu]l reaction volume; Life Technologies, New England Biolabs or
United States Biochemicals) was utilized to treat HPLC-sheared [lambda]DNA fragments (for 30 min at 37oC) under forward reaction conditions (0.2 [mu]M DNA fragments, 50 mM Tris-HCl, pH 7.5, 10 mM MgCl
2
, 5 mM DTT, 1.5 mM spermidine, 0.1 mM EDTA and 0.1 mM ATP), exchange reaction
conditions (0.2 [mu]M DNA fragments, 50 mM imidazole-HCl, pH 6.4, 12 mM MgCl
2
, 0.5 [mu]M ATP, 0.3 mM ADP and 15 mM [beta]-mercaptoethanol), and optimal 3'-phosphatase conditions (0.2 [mu]M DNA fragments, 0.1 M MES, pH 6.5, 10 mM MgCl
2
, 10 mM [beta]-mercaptoethanol and 50 [mu]g/ml BSA); efficiency of incorporation under the first two
conditions was monitored with [[gamma]
32
P]ATP (NEN Dupont).
After
Eco
RV digestion (New England Biolabs), alkaline phosphatase-treated vector (10 [mu]g linearized pBR322 in 400 [mu]l 10 mM Tris-HCl, pH 8.0, incubated for 30 min at 37oC with 0.04 U alkaline phosphatase from Boehringer Mannheim)
was phenol-chloroform extracted three times, chloroform extracted once, and ethanol
precipitated. The concentrations of treated/untreated fragments and
phosphatased/phosphorylated vector were determined immediately before use in
ligation by spectrophotometry (estimating 50 [mu]g/ml double-stranded DNA for an OD
260
of 1). Ligations were peformed overnight at 16oC in a 20 [mu]l volume with 10.4 nM vector and an appropriate amount of fragment (62.1 nM
when cloning into phosphorylated vector; 0-270 nM in incremental amounts when cloning into phosphatased vector), in
50 mM Tris-HCl, pH 7.8, 10 mM MgCl
2
, 10 mM DTT, 1 mM ATP and 50 [mu]g/ml BSA, using 400 U T4 DNA ligase (New England Biolabs). After heat inactivation (10 min at 70oC), 3 [mu]l of each ligation mix was transformed into Subcloning Efficiency
DH5[alpha] competent cells (Life Technologies). Transformants were selected on LB plates containing 50 [mu]g/ml ampicillin; tetracyline-sensitive clones were screened by patching on LB plates
containing 20 [mu]g/ml tetracycline. Transformation of vector alone ligations determined
background in these assays.
For determining the sequence of the fragment ends, 0.4-5 [mu]g of cosmid or P1 cloned human DNA was sheared in 10 mM Tris-HCl, pH 8.0, 5 mM NaCl and 1 mM EDTA. After 8 min of shearing, 1 ml of sheared fragments was collected into a
Centricon 30 (Amicon, Beverly, MA) concentrator. The samples were concentrated to 50 [mu]l by centrifugal dialysis according to the manufacturer's instructions. One
ml of TE (10 mM Tris, 1 mM EDTA) was added to each centricon and the
centrifugal concentration repeated. Vector DNA was prepared by digestion of
M13mp19 (New England Biolabs) with
Sma
I (New England Biolabs) and treatment with Calf Intestinal Alkaline Phosphatase according to the manufacturer's instructions (Boehringer Mannheim). In some cases the sample was cloned directly; for later libraries the sheared
DNA was first cleaned up by gel purification and isolated using a Geneclean II
kit (Bio101, La Jolla, CA). Ligations were carried out in 20 [mu]l with 20 ng of vector and 10-100 ng of sheared DNA. The ligations were incubated for 10-16 h at 16oC with T4 DNA ligase (New England Biolabs) in 50 mM Tris-HCl, pH 7.8, 10 mM MgCl
2
, 10 mM DTT, 500 [mu]M ATP, 25 [mu]g/ml BSA, at 65oC for 5 min, electroporated into DH10B cells and then plated on DH5[alpha] F' plating cells (both from Gibco BRL Life Technologies) prepared according to
Maniatis
et al
. (
9
). A total of 1000-40 000 recombinant clones were produced in each library. Without gel
purification, the libraries contained 5-10% of short inserts (<400 bp, in some cases as short as 80 bp). From each cosmid or P1 clone,
between 1000 and 3000 clones were sequenced using the dye-primer sequencing chemistry from the Applied Biosystems Division of Perkin-Elmer (Foster City, CA, USA).
For DNA subfragment cloning, adapters with a CAAAC overhang were either ligated directly to the sheared DNA fragments overnight at 16oC with T4 DNA ligase, or after end-repair of the DNA fragments with T4 DNA polymerase at 11oC for 60 min. Following ligation, the DNA subfragments were
separated from adapter dimers by means of slab gel electrophoresis on a low-melting 0.8% agarose gel. Subsequently, they were ligated into M13mp18
vector at 16oC overnight with T4 DNA ligase and transformed into NM 522
E.coli
cells (Promega, Madison, WI, USA). Subclone DNAs were prepared by lysis of the
phages in ethanol-butanol (5:2, v/v), purification of the released DNA with Cleanascite
(Affinity Technologies, New Brunswick, NJ, USA), followed by ethanol
precipitation. Subclones were sequenced on a Catalyst 800 Cycle Sequencer (Perkin-Elmer, Applied Biosystems Division) using a linear amplification method with fluorescent dye-primers, followed by electrophoresis on a 4.75% acrylamide gel run
on an Applied Biosystems Model 373 Stretch Sequencer.
Figure
1
shows a scheme of the apparatus used for shearing DNA in a recirculating point-sink flow system. The device had been developed originally for the preparative chromatographic separation and isolation of an intermediate in the total synthesis of vitamin B12 (
14
). The corner stone of the apparatus is the manifold valve, which allows to
direct the solvent to the WASTE, COLLECT or RECYCLE position respectively. In
the RECYCLE mode, the pump does not draw any solvent from the solvent
reservoir. Instead, solvent is recirculated through the system. The dead volume
of the system in the RECYCLE mode is ~1 ml as determined by the photometric measurement of DNA dilution occurring over time. These measurements also revealed no significant loss of DNA during shearing, independently from the set flow-rate.
In contrast with studies by Bowman and Davidson (
15
) as well as Pyeritz
et al.
(
12
), we were able to observe that an increase in ionic strength had a significant
effect on the breakage rate. A higher shear rate was required to obtain
identical breakage rates when shearing was carried out in 200 mM instead of 5
mM sodium chloride. This accords well with the observation of Yew and Davidson
that the rate of breakage at a given shear stress increases with decreasing
ionic strength, approximately as [Na
+
]
-1.6
(
16
). It is also in agreement with the well established effect of ionic strength on
intrinsic viscosity of a DNA solution, which in the case of T7 DNA has been
reported to increase by a factor of ~1.5 when the salt concentration is decreased from 0.2 to 0.005 M (
17
). Hence, it can be concluded that DNA attains a more coiled conformation the
higher the salt concentration. Consequently, a higher shear stress is required to obtain fragments of identical size distribution with an increase of the concentration of sodium
chloride in the shearing buffer from 0.005 to 0.2 M NaCl (Fig.
5
).
The physics of polymer stretching and breakage have been the subject of numerous
reviews (
20
-
22
). While details of the kinetics have not been resolved, the dynamics of
breakage are driven by the hydrodynamic drag force on the molecule and the
spring force internal to the molecule. All current models of long chain
molecules in flow assume the drag on the molecule is proportional to the local
relative velocity of the surrounding fluid. While DNA molecules are in a coiled
state in the absence of flow or in sufficiently weak flows, their central
portions disentangle as soon as a critical velocity gradient is reached, and
become more and more straightened out along the direction of flow at the
expense of the two end portions which remain curled (
21
). This non-uniform stretching of polymers, with a taut central portion and two
randomly coiled end portions, occurs for the simple reason that the viscous
pull due to extensional flow is not uniform along the length of the chain, but
roughly parabolic, being zero at the ends and maximum in the middle (
23
). This also explains the observed 2-fold size distribution of our data (for multiple breakage passes) and that
of Bowman and Davidson (
15
). The latter reported a mean single molecule break at half length with a
standard deviation of 12.5%. Thus if a very long molecule were subjected to
repeated breaks, each in the middle +-12.5%, the limiting result would be a population of molecules having a 2-fold size distribution.
It was also noticed that by moving the 0.0025'' ID tubing farther downstream of the tee outlet, the resulting
mean terminal fragment length was significantly larger than was previously the
case at the same flow-rate (Fig.
7
). The only apparent change to the flow kinetics is the distance between the 90o flow bend in the tee and the region of high shear. In order to gain more
insight into the observation that the proximity of the shearing contraction to
the tee affected the mean size of the sheared DNA, we constructed scale fluid
models of the tee and passageway, matching the critical fluid dynamic scaling
parameter, namely the Reynolds number. By dye injection and qualitative
observation, we noted that the fully developed velocity profile of the fluid
entering the tee was distorted into a pair of counter-rotating vorticies, with axes oriented in the bulk flow direction, as the
flow turned the corner and entered the contraction. We speculate, based on this
observation, that the vorticity associated with fully developed tube flow
(causing tumbling of the DNA molecules) has been restructured so that DNA
molecules are more likely to be aligned with the direction of maximum
extension.
Comparison of the expected nucleotide composition with that observed for 576
fragment ends generated from shearing clones of human DNA and cloning the
fragments directly without any further enzymatic treatment into a
Sma
I cut, phosphatased vector, revealed that Gs and Cs are overrepresented ([chi]
[1]
= 22.7,
P
< 0.001). Looking at the base pairs flanking the cleavage sites, rather than
base composition, AT/GC and CG/GC were overrepresented [[chi]
[1]
of 11.7 and 13.22 (
P
< 0.001) respectively]. On the other hand, AA/TT ([chi]
[1]
= 12.5,
P
< 0.001) and TA/AT ([chi]
[1]
= 8.17,
P
< 0.005) were found to be underrepresented. The reason for this bias is not
completely clear. It appears that GC-rich DNA cleaves more readily. However, the data may also reflect reduced
efficiency with which fragments having terminal thymines and adenines can be
cloned. Effects of ultrasound on nucleic acid bases has shown that sonochemical
changes induced mainly by hydroxy radicals decrease in the order thymine >
cytosine > guanine > adenine (
11
). It can not be excluded that in the point-sink flow system used in the present study, thermal and pressure induced dissociation of water does occur. The occurrence of radicals would also explain why the percentage of clonable fragments
can not be increased significantly by enzymatic end repair.
However, from the schematic representation of M13mp18 subclones obtained from a
sheared cosmid containing a human Y chromosome insert, it is obvious that the
distribution of clones is sufficiently random to ensure complete coverage (Fig.
8
a). Only two ~300-bp regions of the ~41-kb cosmid were covered only by one M13 clone on one
strand. The rest of the cosmid sequence was covered by multiple clones on both
strands.
Since the method employs chromatographic equipment, it lends itself easily to
automation, using an electric valve for automating the flushing of lines
between shearing runs, an autosampler for injecting the samples, and a fraction
collector for collecting the fragments. Work is also underway to develop a
simpler dedicated system for DNA shearing based on a syringe pump, that should
considerably reduce the cost of equipment.
We are grateful to Dr Richard Hyman for the preparation of DNA libraries, and to
Drs Richard Stephens and Sue Kalman for provision of the restriction fragment from
Chlamydia trachomatis
. We also thank Dr George Homsy for helpful fluid mechanics discussions. This
work was supported by a grant from NIH (1P01-HG00205-05).
*To whom correspondence should be addressed at: Department of Genetics, Room M-311, Stanford University, Stanford, CA 94305-5120, USA. Email: oefner@genome.stanford.edu
REFERENCES
Return