 |
Description  |
|
|
TECHNICAL FIELD
The subject of this invention belongs to the field of molecular biology.
TECHNICAL PROBLEM
Genomes range in size from about 4.times.10.sup.6 base pairs (bp) in E.
coli to 3.times.10.sup.9 bp in mammals. Determination of the primary
structure, i.e., sequence, of the entire human genome, is a challenge of
the 20th century. A further challenge for biology is the determination of
the entire genomic sequence for characteristic species of the living
world. It would allow qualitative progress in explaining the function and
evolution of organisms. It would also be a great step forward in the
explanation and treatment of many diseases, in the food industry and in
the entire field of biotechnology.
STATE OF THE ART
Prior Art
Recombinant DNA technology has allowed the multiplication and isolation of
short fragments of genomic DNA (from 200 to 500 bp) whereby a sufficient
quantity of material for determination of the nucleotide sequence may be
obtained in a cloned fragment. The sequence is determined on
polyacrylamide gels which separate DNA fragments in the range of 1 to 500
bp, differing in length by one nucleotide. Distinguishing among the four
nucleotides is achieved in two ways: (1) by specific chemical degradation
of the DNA fragment at specific nucleotides, in accordance with the Maxam
and Gilbert method (Maxam, A. M. and Gilbert, W., 1977, Proc. Natl. Acad.
Sci., 74, 560); or (2) utilizing the dideoxy sequencing method described
by Sanger (Sanger, F., et al., 1977, Proc. Natl. Acad. Sci., 74, 5463).
Both methods are laborious, with competent laboratories able to sequence
approximately 100 bp per man per day. With the use of electronics
(computers and robots), sequencing can be accelerated by several orders of
magnitude. The idea of sequencing the whole human genome has been
discussed at many scientific meetings in the U.S.A. (Research News, 1986,
Science, 232, 1598-1599). The general conclusion was that sequencing is
possible only in big, organized centers (a sequencing factory) and that it
would take about 3 billion dollars and at least ten years. For the time
being the Japanese are the most advanced in organizing components of one
such center. Their sequencing center has the capacity of about 1 million
bp a day at the price of about 17.cent. per bp (Commentary, 1987, Nature,
325,771-772). Since it is necessary to sequence three lengths of a genome,
because of random formation of cloned fragments of about 500 bp, 10
billion bp could be sequenced in approximately 30 years in such a center,
i.e., it would take 10 such centers to sequence the human genome in
several years.
DESCRIPTION OF THE INVENTION
Our process of sequencing, compared with the existing ones, involves an
entirely different logic, and is applicable specifically for determining a
sequence of complex DNA fragments and/or molecules (more than 1 million
base pairs). It is based upon specific hybridization of oligonucleotide
probes (ONPs), having a length of 11 to 20 nucleotides.
Conditions of hybridization of ONPs have been found under which complete
homology with the target sequence is differentiated from a single base
pair mismatch (Wallace, R. B., et al., 1979, Nucleic Acid Res., 6,
3543-3557). If the method of hybridization with 3M tetramethyl ammonium
chloride is used, the melting point of the hybrid is dependent only on the
ONP length, not its GC content (Wood, W. I., et al., 1985, Proc. Natl.
Acad. Sci. U.S.A., 82, 1585-1588). Thus, hybridization under such
conditions unequivocally determines the sequence. Hybridization of genomic
DNA multiplied as subclones of convenient length with a sufficient number
of ONP's will allow the entire genome to be sequenced with the aid of
Computerized assembly of these detected sequences. We believe this method
is an order of magnitude quicker and cheaper than technique presently
being developed. Therefore, it is more suitable for genomic sequencing of
all characteristic species.
In order to apply this method, it is necessary to optimize the length,
sequence and number of ONPs, the length and number of subclones and the
length of the pooled DNA which may represent a hybridizing sample.
Eleven-mer ONPs are the shortest oligonucleotides that can be successfully
hybridized. This means a priori that 4.sup.11 (4,194,304) ONPs are needed
to detect each sequence. The same number of independent hybridizations
would be necessary for each subclone or a pool of subclones. Positively
hybridizing ONPs would be ordered through overlapping 10-mers. This
results in the DNA sequence of the given subclone.
The process of assembling a subclone sequence is interrupted when the
overlapping 10-mer is repeated in the given subclone. Thus, uninterrupted
sequences are found only between repeated 10-mers or between longer
oligonucleotide sequences (ONSs). These fragments of a subclone sequence
(SF) cannot always be ordered in an unambiguous linear order without
additional information. Therefore, it is important to determine the
probable number of SFs (Nsf) distributed along certain length of DNA; this
can be achieved through the application of probability calculations.
The distribution of ONSs along a randomly formed DNA sequence is binomial.
The average distance between identical neighboring ONSs (A) depends only
on the ONS length (L), and is given as: A=4.sup.L. The probability of
having ONSs repeated N times in a fragment of the length of Lf bp is given
as:
P(N,Lf)=C(N,Lf).times.(1/A).sup.N .times.(1-1/A).sup.Lf (1)
where C(N,Lf) represents the number of the N class combinations consisting
of Lf elements. The expected number of different ONSs having the length L
or average distance A, which are repeated N times in a Lf bp fragment, is
given as the product P(N,Lf).times.A. If a sequence is assembled through
the overlapping length L or through average distance Ao, then Nsf in a
fragment Lf is represented as:
Nsf=1+Ao.times..SIGMA.N.times.P(N,Lf), N.gtoreq.2 (2)
If all 11-mers (4.times.10.sup.6) are used, about 3 SFs are to be expected
per Lf of the length of 1.5 kb. We shall return to the problem of ordering
SFs later.
The required synthesis of 4.times.10.sup.6 11-mers, is impracticable for
sequencing by hybridization (SBH). However, it is unsuitable to omit a
significant number of ONPs (more than 25%), because it leads to gaps in
the sequence. A far better way to decrease the number of independent ONP
syntheses and of independent hybridizations is to use ordered ONP groups.
This method requires sequencing of shorter fragments, but no gaps appear
in the resultant sequence. A forty-fold decrease in the number of
syntheses and hybridizations requires a seven-fold increase in the number
of subclones.
The use of ordered ONP groups, in an informative respect, is the same as
using shorter ONPs. For instance, there are 65,536 different 8-mer ONSs.
Since 8-mer ONSs, according to the current knowledge, cannot form a stable
hybrid, the 11-mer group can be used as an equivalent. In other words, all
11-mers in a group have one 8-mer in common, so that the information
obtained concerns only its presence or absence in the target DNA. Each of
the anticipated groups of 11-mers contains 64 ONPs of the (N2)N8(N1) type
(5'.fwdarw.3' orientation is as written, /Nx/means x unspecified bases,
and y/ means y specified bases. A sequence can be detected with about
65,000 such groups. If equation (2) is applied, then DNA fragments 200 bp
long are expected to have 3 SFs on average. Due to dispersion, some
fragments of this length will have 10 and more SFs.
Because of its unrandom GC and dinucleotide composition, ONPs of the
(N2)N8(N1) type are not very convenient for sequencing mammalian DNA. The
more AT bases contained in the common sequence of an ONP group, the longer
it should be. Taking this into account, there remain three types of
suitable probes: (N1)N10, where N10 stands for all 10-mers not containing
G+C; (N1)N9(N1), where N9 stands for all 9-mers containing 1 or 2 G+C; and
(N2)N8(N1), where N8 represents all 8-mers containing 3 or more G+C. About
81,000 such ONP groups are necessary. The average value of their Ao(Aao)
is about 30,000. Equal Ao value in random DNA requires about 130,000 ONPs
of the (N2)N8(A or T) and (N2)N8(or rG) types. These ONP groups allow
sequencing of fragments 300 bp long, with 3 SFs at average. This increase
of 25% in the number of syntheses allows a multifold decrease in the
number of necessary SCs (see below). Apart from these probes it is
necessary to synthesize an additional 20,000 ONPs in Order to (1) solve
the problem of repetitive sequences, (2) confirming the ends of inserts
and (3) supplementing information lost due to the fact that it is
unfeasible to use ONP's which hybridize with vector DNA.
Repetitive sequences, or, generally speaking, ONSs repeated in tandems and
having the length of one or more bp (AAAAAAA . . . TCTCTCTC . . . TGATGATG
. . . ) represent a problem in sequencing by hybridization. The above
mentioned probes cannot determine length of repetitive sequences that are
longer than the common part of a ONP group. Therefore, the precise
determination of repetitive ONSs up to 18 bp long, that represent the
largest part of these ONPs, requires application of the following ONPs:
160 NP An and Tn, where the value of n stretches between 11 and 18 bp,
ONP (AT), ONP Cn and G.sub.n where n is valued from 9 to 18 bp, ONP (AT)n
where n takes on the values of (12, 14, 16, 18), 25 ONPs (AC)n, (AG)n
(TC)n (TG)n and (CG)n, where n is valued (10, 12, 14, 16, 18), 60 ONPs of
the (N1 N2 N3)n type which encompass all trinucleotides and n is (12, 15,
18), 408 ONPs which include all 5-mers in tandems, having the length of 15
and 18 bp, 672 ONPs consisting of 6-mer tandems up to 18 bp long, and 2340
ONPs consisting of 7-mer tandems 18 bp long. The total number of these
ONPs is 3725.
In order to confirm the ends of DNA inserts in an subclone, it is necessary
to synthesize an additional 2048 ONPs of the N6(N5) and (N5)N6 types,
where N6 represents sequences of the vector ends, and (N5) represents all
possible 5-mers in both cases.
The problem of the vector DNA can be solved in two ways. One is
prehybridization with cold vector DNA shortened for 7 bp at both sides of
the cloning site. The other method is to omit ONPs complementary to the
vector DNA. Since phage M13 has been chosen as the most suitable phage
vector (see below), it would eliminate the use of approximately 7,000
ONPs. This is a significant percentage (11%) if 65,000 ONP's [(N2)N8(N1)]
are utilized. It can be decreased to about 3% if, instead of the given
7,000 ONPs, an additional 21,000 ONPs of the (N1)(N.sup.0 1)N8(N1) type
are used, where N8is 7,000 M13 8-mers, and (N.sup.0 1) represents each of
the trinucleotides not present by the given 8-mer.
The calculations, supra, are related to the sequencing of single stranded
DNA. To sequence double stranded DNA it is not necessary to synthesize
both complementary ONPs. Therefore, the number of necessary ONPs can be
halved. Yet, due to advantages of the M13 system, we will discuss further
the method of sequencing the single stranded DNA. In this case, gaps of
unread sequences will appear in the subclones, because half the ONPs are
used. However, a gap in one subclone will be read in the subclone
containing the complementary chain. In a representative subclone library
each sequence is repeated about 10 times on average. This means that it is
probable to have each sequence cloned in both directions, i.e., that it
will be read on both DNA strands. This allows the use of uncomplementary
ONPs with only an increase in algorithmic computation. Thus, the total
number of necessary ONPs would be approximately 50,000. If an M13 vector,
able to package both strands either simultaneously or successively could
be devised, the use of uncomplementary ONPs would not impose any
additional requirement.
All subclone andr all pools of subclones hybridize with all anticipated
ONPs. In this way a set of positively hybridizing s is attained for each
subclone, i.e., subclone pool. These ONPs are ordered in sequences by
overlapping their common sequences, which are shorter than the ONPs by
only one nucleotide. In order to detect overlapping ONPs faster, it is
necessary to determine in advance which ONP overlap maximally with each
synthesized ONP. Thus, each ONP will obtain its subset of ONPs: (ONPa,
ONPb, ONPc, ONPd) 5' ONP.times.3' (ONPe, ONPf, ONPg, ONPh). Ordering must
follow the route of detecting which one out of the four ONPs from the 5'
side, and which one out of four from the 3' side positively hybridize with
the given subclone, i.e., pool. Assembling continues until two positively
overlapping ONPs for the last assembled ONP are found. Thereby, one SF is
determined. When all SFs are extended to the maximum, the process is
terminated.
The number of SFs is increased for a larger length of DNA by using
described ONP groups. Generally, unequivocal ordering can be achieved with
3 SFs per subclone when the SFs are counted in the same way as Nsf is
calculated in equation (2). These SFs are recognized as the two placed at
the ends of an insert with third placed, logic of SFs cannot be The
ordering of SFs cannot be solved for a convenient length of a subclone
because it would be too short. Our solution is the mutual ordering of SFs
and large numbers of subclones, with the possibility of using subclone
pools as one sample of hybridization and/or competitive hybridization of
labeled and unlabeled ONPs.
To obtain the maximum extended sequence by SBH on subclones that may be
used repeatedly, it is necessary to use three subclone libraries in the
M13 vector, with 0.5-7 kb inserts and with inserts of different size,
which consist of two sequences: their distance in genomic DNA should be
about 100 kb. The first library serves primarily to order SFs. These
subclones cannot be preserved for later experimentation. These subclones
enter hybridization as pools obtained by simultaneous infection during or
after growth of a phage. The second library is the basic one. The
subclones it contains are convenient for further use. The length of 7 kbp
represents the current limit of the size of an insert suitable for
successful cloning in M13. The function of the third library is to
regularly associate parts of sequences separated by highly homologous
sequences longer than 7 kbp as well as uncloned DNA fragments into an
undivided sequence.
Hybridization of subclones of all libraries with ONPs, and computation of
SFs is followed by mutual ordering of the SFs and subclones. The basic
library is the first to be ordered. Overlapping subclones are detected by
the content of the whole or of a part of starting SFs of the starting
subclone. Generally, all mammalian SFs having a length of 20 bp or more
are suitable. The average SF length of these subclones is calculated on
the basis of equation (2) is 12 bp. This means that there are enough SFs
of suitable length. Besides, it is known among which SFs--most often there
are two--is the one which continues from the starting SF. In this case
both sequences are examined; the correct one is among them, and
overlapping subclones are detected by it. On the basis of content of other
SFs, exact displacement of overlapping subclones in relation to the
starting subclone is determined. The linear arrangement of subsets (SSFs)
is achieved by detecting of all subclones overlapping with the starting
subclone. SSFs are defined by neighboring ends of overlapping subclones
(either beginning-beginning, beginning-end or end-end). The process of
overlapping subclones is resumed by the SF taken from the jutting SSF of
the most jutting subclone. The process of assembling is interrupted when
encountered with an uncloned portion of DNA, or, similar to forming of
SFs, with a repeated sequence longer than 7 kbp. The maximum size of
groups of ordered overlapping subclones 7 bp long is obtained by this
method, as well as linear arrangement of the SSFs of their SFs.
The length of DNA contained in a SSF an is fundamental in this procedure.
This length depends on the number of SSFs, which is equivalent to the
number of subclone ends, i.e., two times the number of subclones. A
representative library of DNA fragments of 1,000,000 bp requires 700 SCs
of 7 kbp. Therefore, the average size of a SSF is 715 bp. The real average
number of SFs within an SSF is not 1/10 the number of SFs in a subclone 7
kbp long nor is it dependent on the length of subclone. Instead, the real
number is dependent on the length of an SSF. According to equation (2),
for the length of 715 bp and an Aao of 30,000 bp (the values for described
ONPs), 16 SFs with an average length of 45 bp are expected on average. The
ordering of SFs within obtained SSFs is performed through the 0.5 kb
subclone library. This method does not require all subclones to be
individual; the use of a subclone pool is sufficient. The subclone in a
respective pool are informative if they do not overlap with each other.
Informatively and technologically suitable is a 10 kb pool of cloned DNA,
although it is not the limit. The number of such necessary subclones or
pools is such that the maximum size of an SSF formed by them can be no
longer than 300 bp. The ONPs proposed for this length are expected to give
3 SFs (equation 2), which, as explained, may be unequivocally ordered.
Utilizing binomial distribution has enabled the derivation of the equation
:
Nsa=Nsc(1-2Nsc/Nbp).sup.Lms
where Nsa is the number of SSFs greater than Lms, Nsc is the number of bp
per fragment or molecule of the DNA being sequenced, Lms stands for the
size of on SSF, which an average gives the number of 3 SF, which can be
ordered; in this case it amounts 300 bp. On the basis of this equation it
is determined that 25,000 subclones of 0.5 kb are necessary for a DNA
fragment 1 million bp long. The number of 10 kb pools is 1250. Average
size of an SSF obtained with these subclones is 20 bp.
The ordering of SFs is performed by computer detection of pools containing
subclones which overlap with the starting SSF. This detection is performed
on the basis of the content of a part of the whole SF, randomly chosen
from the starting SSF. Contents of other SFs from the starting SSF
determine the overlapping proportion of subclones of 5 kb. Due to their
high density, the arrangement of SFs in the starting SSF is also
determined. At the end of this process, the sequence of each group of
ordered subclones of 7 kb is obtained, as well as information about which
pool contains the subclone of 0.5 kb bearing the determined sequence. At a
certain small number of loci the sequence will either be incomplete or
ambiguous. Our calculations show that on average it is not less than one
locus per million bp, including 30% of randomly distributed undetected
ONSs. These loci are sequenced by convenient treatments of subclones which
bear them, followed by repeated SBH or competitive hybridization with
suitably chosen pairs of labeled and unlabeled ONPs or by classical method
or by the advanced classical method.
The procedure of competitive hybridization will be explained by the example
of the sequence 7 bp long repeated twice. In this case two SFs end and two
others begin with the repeating sequence TTAAAGG which is underlined.
5' NNNNNNNNNNNNCATTAAAAGG3'
5' NNNNNNNNNNNNCGTTAAAAGG3'
5' TTAAAAGGTACNNNNNNN3'
5' TTAAAAGGCCGNNNNNNN3'
Prehybridization wish surplus of an unmarked ONP, e.g.,
5'(N2)CATTAAAAG(N1)3', which cannot hybridize with 5'NNCGTAAAAGG3' due to
one uncomplementary base prevents one of the labelled ONPs-5'
(N2)AAAAGGTAC(N1)3' or 5' (N2)AAAAGGCCG(N1)3'--from the subsequent
hybridization. A pair of mutually competing probes defines a pair of SFs
which follow one another. This can be confirmed by an alternative choice
of a suitable ONP pair. This procedure may be applied on all repeating
ONSs having the length of up to 18 bp. In order to use it for the ordering
of a multitude of SFs, prehybridization must be separated from
hybridization in both time and space. Therefore, the stability of a hybrid
with unlabelled ONP is important. If such stability cannot be achieved by
appropriate concentrations of ONPs and by choice of hybridizing
temperatures, then a covalent link should be formed between a cold ONP and
complementary DNA by UV radiation in presence of psoralen. Alternatively,
one might use ONPs which carry reactive groups for covalent linking.
The subclones of the third library are used to link the sequenced portions
into a uniform sequence of the entire DNA fragment being sequenced.
Approximately 170 subclones are required for 1 million bp. These and other
numbers calculated for 1 million bp increase linearly with length for
longer DNA fragments. Since these subclones contain sequences which are
distanced at 100 kbp on average, they allow jumping over repeated or
uncloned sequences, the size of which increases up to 100 kb. This is done
by detecting which of the two sequenced portions contains sequences
located in one subclone from the library.
The experimental requirement of this method is to have the total number of
50,000 ONPs and hybridizations, and 2120 separated hybridizing subclone
samples per DNA fragment approximately 1,000,000 bp long.
The libraries described, supra, are made in the phage vector, M13. This
vector allows easy cloning of DNA inserts from 100 to 7,000 bp long, and
gives high titer of excreted recombinant phage without bacterial cell
lysis. If a bacterial culture is centrifuged, a pure phage preparation is
recovered. Additionally, the bacterial sediment can be used for repeated
production of the phage as long as the bacterial cell pellet is
resuspended in an appropriate nutrient media. With the addition of alkali,
DNA separates from the protein envelope and is simultaneously denatured
for efficient "dot blot" formation and covalent linking to nylon filters
on which hybridization is performed. The quantity of DNA obtained from
several milliliters of a bacterial culture is sufficient to hybridize one
subclone with all the ONPs. A suitable format for cultivation and robot
application on filters are plates similar to micro-titer plates, of
convenient dimensions and number of wells. Application of DNA subclones on
filters is performed by a robot arm. Even the largest genomes can be
satisfactorily sequenced by a robot arm supplied with 10,000 uptake
micropipets. After the DNA solution has been removed from the plates, the
micropipets are positioned closer to each other by a mechanism reducing
the distance between them to 1 mm. The quantity of DNA suitable for 10,000
subclones is then applied to the filter simultaneously. This procedure is
repeated with the same 10,000 subclones as many times as necessary. The
same procedure is then repeated with all other subclones, their groups
numbering 10,000 each. The number of "imprints" of one such group with
50,000 ONPs is approximately 1,000, since each filter can be washed and
reused 50 times.
The hybridization is performed in cycles. One cycle requires at maximum one
day. All the subclones are hybridized with a certain number of ONPs in one
cycle. In order to have the hybridization completed within a reasonable
period, an experiment in each cycle should require approximately 1,000
containers, each with one ONP. For the purpose of saving ONPs, a smaller
volume of hybridizing liquid is used, and therefore filters are added in
several turns. Filters from all hybridizing containers are collected in
one container, and are simultaneously processed further i.e., they are
washed and biotin is used for labelling of ONPs instead of radioactive
particles, colored reactions are developed. All subclones required for
sequencing (up to 10 kb in length) can be hybridized in containers with
the dimensions 20.times.20.times.20 cm without having to repeat individual
cycles.
Hybridization sequencing of fragments cloned in plasmid vectors can be
performed one of two ways: by (1) colony hybridization or (2) "dot blot"
hybridization of isolated plasmid DNA. In both cases, 2,000 or 3,000
different ONPs in the vector sequence will not be utilized, i.e., will not
be synthesized.
Colony hybridization is presumably faster and cheaper than "dot blot"
hybridization. Moreover, colony hybridization requires specific conditions
to annul the effect of hybridization with bacterial DNA. The marking of
probes giving high sensitivity to hybridizations should be done in order
to reduce the general background and to allow use of minimal number of
bacterial colonies. Marking of ONPs should, however, be via
biotinylization, for the benefit of easy and lasting marking in the last
step of synthesis. The sensitivity attained (Al-Hakin, A. H. and Hull, R.,
1986, Nucleic Acid Research, 14:9965-9976) permits the use of at least 10
times fewer bacterial colonies than standard protocols.
In order to avoid "false positive" hybridization caused by homology of ONP
with a bacterial sequence, and in order to use short probes such as
11-mers, which are repeated twice on average in the bacterial chromosome,
plasmid vectors that give the maximum number of copies per cell should be
used. High copy plasmid vectors, such as pBR322, are amplified to 300-400
copies per bacterial cell when grown in the presence of chloramphenicol.
(Lin Chao, S. and Bremer, L., 1986, Mol. Gen. Genet., 203, 150-153).
Plasmids pAT and pUC have at least twice the efficiency of multiplication
(Twigg, A. J. et al., 1987, Nature, 283, 216-218). Therefore, we can
assume that under optimal conditions even 500 copies of plasmid per
bacterial cell can be attained. The additional sequences within the
hymeric plasmids will surely cause a decrease in plasmid copies per cell,
especially in the presence of poisonous sequences. Therefore, operations
should be performed with about 200 copies of chimeric plasmid per
bacterial cell. This means that, an average, the signal would be 100 times
stronger with every 11-mer if there is a complementary sequence on the
plasmid as well. That is sufficient difference for hybridization with
bacterial DNA not even to be registered, when a small quantity of DNA
(i.e., small colonies) is used.
Using a binomial distribution we determined how many ONPs will be repeated
more than 10 times in bacterial chromosomes due to random order. Such ONPs
will give unreliabl | | |
