drug–dna interaction protocols

DNase I requires the presence of dlvalent metal ions, particularly magnesium, and so Its action can be stopped by adding EDTA The enzyme has more than one bmdmg site for dlvalent catlons

Footprmting is essentially a protection assay, m which cleavage of DNA is inhibited at discrete locations by the sequence specific binding of a hgand or protein In this technique, a DNA fragment of known sequence and length (typi- cally a restriction fragment of 100-200 bp), which has been selectively radiola- beled at one end of one strand, IS lightly dtgested by a suitable endonucleolytic probe m the presence and absence of the drug under investigation The cleav- age agent is prevented from cutting around the drug-binding sites so that, when the products of reaction are separated on a denaturing polyacrylamide gel and exposed to autoradiography, the position of the ligand can be seen as a gap m the otherwise continuous ladder of bands (see Fig 1) In this figure, cleavage

at position “a” will produce, after denaturing the DNA, one long fragment (9 bases) corresponding to the left hand strand, and two short fragments (7 bases and 2 bases) from cleavage of the right hand strand Since the bands are located

by autoradiography, only the shortest of these species bearing the radioactive label will be visualized The condittons of the cleavage reaction are adjusted so that, on average, each DNA fragment is cut no more than once As a result, each of the bands on the autoradiograph is produced by a single cleavage event, i.e., single-hit kmetics If an excessive amount of cleavage agent is used, then

From Methods m Molecular Biology, Vol 90 Drug-DNA Interactron Protocols

Edited by K R Fox Humana Press Inc , Totowa NJ

1

DNase I footprmtmg has been successfully employed for mdentrfymg or conlirmmg the preferred DNA binding sites for several hgands mcludmg acti- nomycm (2-4), mtthramycin (5), quinoxalme antrbrotrcs (6,7), daunomycm (8,9), nogalamycin (1/J), vartous minor groove binding agents (2,3,12), and triplex binding ohgonucleottdes (12,13) Various other cleavage agents, both enzymrc and chemical, have also been used as footprinting probes for drug- DNA interactions including micrococcal nuclease (24), DNase II (6,15), cop- per phenanthrolme (16,17), methtdiumpropyl-EDTA.Fe(II) (MPE) (18-21), uranyl photocleavage (22,23), and hydroxyl radicals (24-26) Each of these has a different cleavage mechanism, revealmg different aspects of drug-DNA interactions

An ideal footprmtmg agent should be sequence neutral and generate an even ladder of DNA cleavage products in the absence of the hgand This property is almost achieved by certain chemical probes, such as MPE and hydroxyl radi- cals However, the most commonly used cleavage agent (because of its cost and ease of use) 1s the enzyme DNase I, which produces an uneven cleavage pattern that varies according DNA sequence and local structure (see Subhead- ing 1.2.) Cleavage at mdrvtdual phosphodiester bonds can vary by over an order

DNase I Foo tprinting 3

of magnitude m a manner determined by both local and global DNA structure (27,28) In addltlon, drugs that modrfy DNA structure can induce enhanced DNase I activity m regions surroundmg their binding sites if they alter the DNA structure so as to render it more suscepttble to cleavage (3,6,15,29,30) This IS most frequently seen m regions that are particularly refractory to cleav- age m the drug-free controls

1.2 DNase I

DNase I 1s a monomeric glycoprotem of mol wt 30,400 It IS a double strand- specific endonuclease, which introduces single strand nicks m the phosphodiester backbone, cleaving the 03’-P bond Single stranded DNA is degraded at least four orders of magmtude more slowly (32,32) The enzyme requires divalent cations and shows opttmal actlvlty m the presence of calcmm and magnesium (33) Although it cuts all phosphodiester bonds, and it does not possess any simple sequence dependency, its cleavage pattern 1s very uneven and 1s thought

to reflect variations m DNA structure (27,34) In particular, A, * T, tracts and GC-rich regions are poor substrates for the enzyme The most important fac- tors affecting Its cleavage are thought to be mmor groove width (27,28) and DNA flexibility (35,36)

Several crystal structures have been determined for both the enzyme and its complex with oligonucleotides (37-42) These show that DNase I bmds by inserting an exposed loop mto the DNA minor groove, Interacting with the phosphate backbone, as well as the walls of the groove This explains why cleavage is poor in regions, such as A,, * T, tracts on account of their narrow minor groove, to which the enzyme cannot bind An additional feature of these crystal structures 1s that the DNA 1s always bent by about 2 lo toward the major groove, away from the enzyme If this bendmg 1s a necessary feature of the catalytic reaction, then rigid regions, such as GC-rich sequences, may be refrac- tory to cleavage However, these factors do not explain the very different cutting rates that are often observed at adjacent dinucleotide steps It 1s possible that this

is determined by precise orientation of the sclssile phosphodlester bond, How- ever, the crystal structures show that there may be other specific interactions between the exposed loop and DNA bases removed from the cutting site In particular, tyrosme-76 mteracts with the base 2 posItIons to the 5’ side of the cutting site and arginme-4 1 binds to the base at position -3 This latter mterac- tion 1s sterically hindered by a GC base pair in thts position By examining the characteristics of several good DNase I cleavage sites, Herrera and Chaires (43) suggested that the best cleavage site was WYWIWVN (where W = A or T,

Y = C or T, and V = any base except T)

The DNA-binding surface of DNase I covers about 10 bp, i.e., one complete turn the DNA helix This has tmportant consequences for interpreting

footprmtmg results and explams the observatton that the enzyme overesttmates drug-binding site sizes Although DNA bases he perpendtcular to the hellcal axis, they are mclmed relative to the phosphodtester backbone As a result, clos- est phosphates, postttoned across the minor groove, are not attached to a single base pan, but are staggered by about 2-3 bases m the 3’ direction This is illus- trated m Fig 2A, m which the DNA has been drawn lookmg along the minor groove, showmg the inclmatton of the DNA base pans Since DNase I (hatched box) binds across this groove, its bmdmg sate on the top strand 1s located 2 bases

to the 3’ side of that on the lower strand When a DNA-binding hgand is added (filled box in Fig 2B), it can be seen that the closest approach of the enzyme is not the same on each strand DNase I can approach closer to the enzyme on the lower strand; the region of the upper strand protected extends by about 2 bases beyond the actual ligand-bmdmg sate As a result, DNase I footprmts are stag- gered by about 2-3 bases m the 3’ direction across the two strands

2 Materials

2.1 DNase I

For most footprintmg experiments the DNase I does not need to be espe- cially pure There 1s ltttle advantage m purchasmg HPLC-pure, RNase-free enzymes Currently purchased 1s the type IV enzyme, from bovme pancreas, from Sigma (St Louis, MO) This should be dtssolved m 0.15 MNaCl contam- ing 1 mMMgC1, at a concentratton of 7200 Kumtz U/mL Thts can be stored at -20°C, and is stable to frequent freezing and thawing The enzyme 1s diluted to workmg concentrattons immedtately before use; the remainder of the diluted enzyme should be discarded

DNase I Footprinting 5 Table 1

Sequence of the tyrT DNA Fragment

2.2 Choice of DNA Fragment

2.2.1 Natural DNA Fragments

For footprinting experiments, the length of fragment used depends on both convenience (how easily a specific fragment can be generated) and the resolu- tion limit of the polyacrylamide gels The chosen fragment length is typically between 50 and 200 bp Although different laboratories have adopted different natural fragments as standard substrates for footprmtmg experiments, a few have been used more widely Among these are the 160 bp tyrT fragment (sequence shown m Table 1) t&8)), the EcoRI-PvuII fragments from PBS (Stratagene) (4&M), and several fragments from pBR322 (HindIII-HueIII, HindIII-AM,

or EcoRI-RsaI) The plasmids from which these can be prepared are available from commercial sources or from the author’s laboratory In many ways it would be convenient if a few fragments did become recognized standards, since this would facilitate direct comparison of the relattve specrfictttes of hgands prepared in different laboratories Since many sequence selective small mol- ecules have recognition sites of between 2 and 4 bp, there is a reasonable prob- ability that their preferred sites will be present in a lOO- to 200-bp restriction fragment However, it should be noted that there are 2 different bp, 10 different dmucleotides, 32 trmucleotides, 136 tetranucleotides, 5 12 pentanucleotides, and 2080 hexanucleotides It can therefore be seen that the chance of finding a particular binding site within a given DNA fragment becomes more remote the greater the selectivity of the ligand A further complicatmg factor is that, although many ltgands spectfically recognize only a dmucleotlde step, their binding affinity is often influenced by the nature of the surrounding bases,

6 Fox which alter the local DNA structure (47-49) It IS therefore possible that using

a natural fragment may fail to detect the optimum bmdmg sites for the most selective hgands This becomes especially relevant since many novel synthetic ligands possess enhanced sequence recogmtton properties, with binding sites

of eight or more base pairs

2.2.2 Synthetic Oligonucleotides

As explamed, although footprmtmg experiments with natural DNA frag- ments provide a reasonable estimate of a ligand’s preferred bmdmg sites, these are complicated by the limited number of sequences studied, together with ambiguities over the exact bmdmg site within a larger footprmt The next step

m confirmmg the sequence preference may be to prepare a synthetic DNA fragment containing the putative binding site and to use this as a substrate for footprmting experiments (50,51) In addition, for compounds that have been produced as the result of rational design, one may be able to predict their pre- ferred bmdmg site Synthesis of suitable length ohgonucleotides (50 bases or longer) IS now routine However, the results obtained with short oligonucle- otides need to be interpreted with caution and rigorously controlled for several reasons First, binding sites located close to the ends of short ohgonucleotides may not adopt the same configuration as when located within longer sequences because of “end effects.” Second, smce the synthetic fragments will contam only one or two binding sites, it is necessary to ensure that other sequences with equal or greater affinity have not been excluded This can be investigated

by comparing the mteraction with other closely related sequences, m which one or two bases m or around the cognate sequence are altered m turn Analy- sis is simphfied further if the variant sites are contamed withm the same DNA fragment

2.2.3 Synthetic Fragments

A frequent variant on the above is to clone the synthetic oligonucleottdes mto longer DNA fragments This removes the problems associated with end effects and provides other common flanking sequences to which ligand bind- ing can be compared An added advantage is that, once it has been cloned, the sequence can be readily isolated from bacteria The authors usually clone syn- thetic ohgonucleotides mto the BamHI site of pUC plasmids They have pre- pared a wide range of such cloned inserts, containing central GC, CG, or (A/T),, sites (11,15,29,30), which are available from the authors’ laboratory on request DNA fragments contammg the synthetic inserts can be prepared and radiola- beled at either end (see Subheading 3.2.) by isolatmg the modified polylmker Once again a proper analysis will requtre fragments contammg both cognate and closely related noncognate sequences

DNase I Footprinting 7 2.3 Buffers

2.3.1 Solutrons for Plasmid Preparation

1 Resuspenston solution 50 mM Trts-HCl pH 7 5, contammg 10 mM EDTA

2 Lysis solution 0.1% SDS, 0.1 MNaOH

3 Neutralization solutton 3 M potassium acetate, 2 A4 acettc acid

2.3.2 Genera/ Buffers

1 10 mA4Tris-HCl, pH 7 5, contannng 0 1 mA4EDTA This is used for dtssolvmg DNA

2 10 mM Trts-HCl, pH 7.5, containing 10 mA4 NaCl This is used for preparing drug solutions

3 DNase I buffer 20 mMNaC1,2 mM MgCl*, 2 mM MnC&

2.3.3 Reagents for Electrophoresis

1 TBE electrophorests buffer This should be made up as a 5X stock solutton con- taining 108 g Tns, 55 g Boric acid, and 9.4 g EDTA made up to 2 L with water

2 Acrylamide solutions Polyacrylamide sequencing gels are made from a mixture containing acrylamtde*btsacrylamtde in the ratio 19.1 Because of the toxic nature

of these compounds acrylamide solution are best purchased from a commerctal supplier (National Diagnostics [Atlanta, GA], Anachem [Luton, Beds, UK]) and should be used according to the manufacturers mstructions

3 DNase I stop solution Formamide containing 10 mM EDTA and 0 1% (w/v) bromophenol blue

3 Methods

3.1 Plasmid Preparation

Several methods are available for preparing plasmid DNA, which IS suitable for restriction digestion and radiolabeling, including several commerctal kits (including Qiagen or Wizard) and caesium chloride density gradient centrifu- gation It 1s beyond the scope of this article to review the relative merits of each procedure, except to note that in many instances it is not necessary to generate high purity plasmid preparations Since the radtolabeled restrtction fragments are eventually isolated and purified by gel electrophoresis, prior purification of the plasmids may not be necessary, so long as the preparations do not contain nucleases or any agents that inhibit restriction enzymes or polymerases As a result, plasmtds are usually prepared by standard alkaline lysts procedures, fol- lowed by extraction with phenol/chloroform A very brief protocol for extract-

mg pUC plasmids 1s described as follows:

1 Grow 50 mL bacteria overnight

2 Spin culture at 3000g (I e., 5000 rpm m a Beckman JA20 rotor) for 5 mm m Oakridge tube

Remove the supernatant and add 0 6 vol of lsopropanol

Spin at 17,OOOg (12,000 rpm) for 15 mm

Remove the supernatant and wash the crude DNA pellet with 5-10 mL 70% etha- nol Transfer the pellet to an Eppendorf tube and dry

Redissolve pellet m 0 5 mL 10 mA4 Tns-HCl, pH 7 5, containing 0.1 mM EDTA and 100 pg/mL RNase Leave at 37°C to dissolve for at least 30 mm

Extract twice with 0 5 mL phenol/chloroform (phenol forms the bottom layer and should be discarded) The interface will probably be very messy, leave the Junk behind

Remove any dissolved phenol by extracting twice with 0 5 mL ether (which forms the top layer and should be discarded) Allow excess ether to evaporate by stand- ing at 37°C for a few minutes

Precipitate with ethanol, dry and dissolve m 100-l 50 JJL Tns-HCI, pH 7 5, con- taining 0.1 mM EDTA

3.2 Radiolabeling the DNA

DNA fragments can be efficiently labeled at either the 5’ end (using poly- nucleotlde kmase) or 3’ end using a DNA polymerase However, the results of DNase I digestion are easiest to interpret for 3’-end-labeled fragments Smce DNase I cuts the 03’-P bond, the products of dlgestlon possess a 3’-hydroxyl and 5’-phosphate group In contrast, Maxam-Gilbert sequencing reactions, which are used as markers in footprmtmg gels (see Subheading 3.3.), leave phosphate groups on both sides of the cleavage pomt (52) As a result, the radlolabeled products of DNase I cleavage and Maxam-Gilbert sequencmg reactions will be identical if the DNA 1s labeled at the 3’ end (i.e., both possess a phosphate at the 5’ end) However, if the DNA 1s labeled at the 5’ end then the labeled DNase I products will possess an extra phosphate group and so run slightly faster than the correspondmg Maxam-Gllbert products Although this difference 1s often over- looked in footprmtmg gels, it becomes significant for short fragments for which the difference m mobility may be as great as 2-3 bands For enzymes that cut the O-5’ bond, such as DNase II and mtcrococcal nuclease, 5’-end-labeled fragments comlgrate with the Maxam-Gilbert marker lanes

3.2.1 3’-End Labeling with Reverse Transcriptase

The production of 3’-end-labeled DNA fragments can be achieved by cut- ting with a restrlction enzyme that generates sticky ends with 3’-overhanging

DNase I Footprmting 9 ends, followed by filling m with a polymerase using a suitable [a-32P]-dNTP The fragment of interest IS then released from the remamder of the plasmid by cleaving with a second enzyme that cuts the other side of the region of interest The two restriction enzymes usually cut at single locatlons in the plasmid, though this 1s not necessary so long as the various radiolabeled fragments can

be separated from each other The most commonly used polymerase is the Klenow fragment However, it is found that the most efficient labeling is achieved using AMV reverse transcriptase, even though this 1s actually an RNA-dependent DNA polymerase However, not all commercially sources of this enzyme are equally reltable; consistent results are obtained with reverse transcrlptase from Promega or Pharmacia

3 2 1 .l RESTRICTION DIGESTION AND a’-END LABELING

Using the aforementioned procedure for DNA isolation, the followmg 1s used for generating radlolabeled Hindlll-EcoRl polylmker fragments from pUC plasmids

1 Mix 30 pL plasmld (about 50 pg DNA) with 10 pL of 10X restrlctlon enzyme buffer (as supplied by the manufacturer), 45 PL water

2 Add 3 pL HzndIII (A/AGCTT) and incubate at 37°C for 2 h

3 Add 1 PL [a-32P]-dATP (3000 Wmmol, Amersham) together with 1 PL reverse transcriptase and Incubate for a further 1 h

4 The reverse transcriptase IS then Inactivated (to prevent further mcorporatlon of radiolabel at the 3’ end of the second restrlctlon site) by heatmg at 65°C for 5 mm

5 After cooling to 37”C, 3 pL EcoRI (G/AATTC) is added and the mixture mcu- bated for a further 1-2 h In this case, the DNA can be labeled on the opposite strand by reversing the order of addition of EcoRI and HzndIII

If the second enzyme produces blunt ends or sticky ends with 5’ overhangs,

or if the 3’ overhangs sites can not be filled m with dATP, then all the enzymes can be added simultaneously Examples of such combinations for pUC polylinker fragments are HzndlII-SacI, and EcoRI-&I The @rT fragment can

be prepared by simultaneous digestion with EcoRl and Aval In this instance the EcoRl end is labeled with [a-32P]-dATP, whereas the Aval end can be labeled with [a-32P]dCTP Although various enzymes are supplied with dlffer- ent reaction buffers, it 1s found that there IS usually no need to change buffers between the first and second enzymes

6 The mixture of radlolabeled fragments is preclpltated by addmg 10 PL of 3 M sodium acetate and 300 pL ethanol, followed by centrlfugatlon m a suitable microfuge, at top speed The pellet 1s washed with 70% ethanol, dried and dls- solved m 15-20 FL Tris-HCl containing 0 1 mA4 EDTA Then 4 PL of loading dye (20% F~oll, 10 mA4EDTA, 4 1% [w/v] bromophenol blue) is added before

10 Fox loading onto a polyacrylamide gel (typically 6-8%) The gel should be run cold,

so as not to denature the DNA, it is usually run 0 3-mm-thick, 40-cm-long gels in

1 X TBE at 800 V Samples are loaded into slots 10 mm wide by 15 mm deep After the bromophenol blue has reached the bottom of the gel (about 2 h), the plates are separated and the gel covered with Saran wrap Scanning the gel with a hand-held Geiger counter should give a reading off scale (1 e , at least 3000 cps) over the radiolabeled bands The precise location of the radiolabeled bands is

determined by short (2-10 min) autoradiography This autoradlograph IS placed under the glass plates and used to locate the band of Interest, which IS cut out using a sharp razor blade

3.2.1.2 EXTRACTION OF RADIOLABELED DNA FRAGMENTS

The simplest, cheapest, and most efficient method for extracting radio-

labeled DNA fragments from polyacrylamlde gel slices IS by diffusion Place a

small glass wool plug m the bottom of a 1 mL (PlOOO) pipet tip and seal the bottom end with parafilm Add the gel slice containing the radiolabeled DNA and cover this with 10 mA4 Tris-HCl, pH 7 5, containing 10 mM EDTA (about

300 pL is sufficient) Cover the top of the pipet tip with parafilm and incubate

at 37°C with gentle agitation This is usually incubated overmght, though most

of the DNA elutes after 2 h Remove the parafilm from the top and bottom of the tip and expel the buffer mto an Eppendorf tube using a pipet and/or low- speed centrifugation (15OOg m an Eppendorf centrifuge) The gel slice should

be retamed in the pipet tip by the plug of glass wool, though a small amount of polyacrylamide does occasionally come through This can be removed by cen- trifugation For fragments shorter than 200 bp, this procedure recovers about 95% of the radiolabel m the gel slice, though the efficiency decreases for longer fragments The DNA should then be precipitated with ethanol and redissolved m Tris-HCI containing 0.1 mA4 EDTA so as to generate at least 10 cps per pL on

a hand-held counter For most footprintmg experiments it is not necessary to know the absolute DNA concentration, since this is vamshmgly small The important factor is concentration of the radiolabel, which should be sufficient

to produce an autoradiograph within l-2 d exposure

3.3 Maxam-Gilbert Marker Lanes

Bands in the DNase I digestion patterns are identified by comparison with suitable marker lanes Since each DNA fragment produces a characteristic sequence dependent digestion pattern, it is sometimes possible to identify the bonds by comparison with a previous (published) pattern

3.3.1 G-Tracks

The simplest and most commonly used marker lane is the dimethylsulfate- piperidme marker specific for guanine (52) Since the procedure is more time-

DNase I Footprintmg 11 consuming than DNase I digestion itself, it is usual to prepare sufficient quantity of

“G-track” for several footprmting experiments with the batch of radiolabeled DNA Add 10 uL radiolabeled DNA to 200 pL of 10 mA4 Tris-HCl, pH 7.5, con- tammg 10 mM NaCl To this add 1 pL dtmethylsulfate and mcubate at room temperature for 1 mm before stopping the reaction by addmg 50 uL of a solution containing 1.5 Msodmm acetate and 1 Mmercaptoethanol followed by 750 pL ethanol Some laboratortes include tRNA in this G-stop, as a coprectpttant, but

it is found that this is not generally necessary Leave the mixture on dry ice for

10 min, then spin at full speed in an Eppendorf centrifuge (12,000g) for 10 min Remove the supernatant and wash the pellet twice with 70% ethanol After drying the pellet, add 50-l 00 yL of 10% (v/v) plperidme and heat at 100°C for between 20 and 30 min Remove the ptpertdme by either lyophilizatton or m a speed-vat Redissolve the sample m loading dye (formamlde containing 10 mJ4 EDTA and 0.1% [w/v] bromophenol blue) so that each electrophorests sample contains about 10 cps

3.3.2 G+A Tracks

Although the preparation of a G-track is reliable, it is time-consummg and mvolves some highly toxic compounds (dimethylsulfate) G+A marker lanes are also widely used and are usually prepared by limited acid depurmation using formtc acid-ptperidme reactions During the DNase I footprintmg work

it was noted that occastonal careless handling of the samples resulted m put-me tracks appearing m the DNase I cleavage lanes This observatton has been used

to establish an empirical method for rapidly preparing G+A marker lanes

To 2 pL of radiolabeled DNA, add 15-20 pL of Trts-HCl, pH 7.5, contam- ing 10 MNaCl and 5 pL of loading dye (formamide containing 10 mM EDTA and 0.1% [w/v] bromophenol blue) Heat at 100°C for about 20 mm in an Ep- pendorf tube, with the cap open This reduces the volume to about 5-6 pL, sufficient for loading onto the gel and generates a clean G+A track Since this method 1s rapid, each marker lane can be freshly prepared while performing the DNase I digestions

3.4 DNase I Footprinting

3.4 I Basic Footprinting Protocol

The basic procedure for DNase I footprinting is quick and snnple (hence its popu- larity as a footpnnting agent) and can readily be adapted to suit a range of conditions

1 Mix 2 uL radiolabeled DNA (dissolved m 10 mMTrrs-HCl, pH 8.0, contannng 0.1 rniI4 EDTA) with 2 uL ligand (dissolved in a surtable buffer, such as 10 n&I Trrs-HCl, pH 7.5, containing 10 m&I NaCl) See Note 5 for suitable hgand concentrations

12 Fox

2 Leave this to equihbrate for an approprtate length of time For most small hgands, such as minor groove binding ligands or simple intercalators, the interaction with DNA is very fast, though some hgands require in excess of 30 mm for equiltb- rium distribution

3 Start the digestion by adding 2 PL DNase I (dissolved in 2 mM MgCI,, 2 mM MnCl,, 20 mM NaCi)

4 After 1 minute stop the reaction by adding 3 pL of formamide containing 10 mh4 EDTA and 0 1% (w/v) bromophenol blue

The concentratton of DNase I requtred will depend on the reaction condrttons, 1-e , temperature, pH, DNA concentration, tonic strength This should be adjusted emptrtcally so as to give suitable extent of dtgestton (see Notes l-4) It 1s typically found that, at 20°C with 10 mM NaCl, a suitable enzyme concentra- tion is about 0.03 Kunitz U/rnL (i.e., dilute 2 PL of stock DNase I [7200 U/mL] m

1 mL DNase I buffer, followed by adding 2 l.rL of this dtlutton to a further 1 mL buffer Each of these dilutions should be mixed gently, avotdmg vtgorous agi- tation) The enzyme should be freshly diluted immediately before use

3.5 Electrophoresis and Autoradiography

1 After DNase I digestion the samples should be denatured by boiling for about

3 mm, before loading onto a denaturing polyacrylamide gel Samples can be loaded directly from the boiling conditions, though excessive heating can pro- duce some depurmation However, it is probably best rapidly to cool the samples

on ice before loading For most footprmtmg reactions there ts no need to use sharks teeth combs, and simple slots are sufficient

Denaturing polyacrylamide gels (6-l 2% depending on fragment length) should contam 8 M urea and are run m 1X TBE buffer, For some CC-rich DNAs these denaturing conditions are not harsh enough and some bands are compressed Thts can be alleviated by including formamtde (up to 30%) m the gel mixture and can be further improved by prerunning the gel for 30 mm before use Formamtde contammg gels run slightly slower than conventtonal gels and should be of a slightly higher percentage For footprmtmg expert- ments 0.3-mm-thick gels are normally used that are 40 cm long; these are run

at 1500 V until the bromophenol blue reaches the bottom (about 2 h) The gels should be run hot, maintaining the DNA m a denatured form Although many modern electrophoresls tanks are thermostatically controlled, “smtling” of the lanes can also be avoided by clamping a metal plate over the glass surface, ensuring an even dtstributton of heat

2 After electrophoresis the plates are separated and the gel is soaked in 10% (v/v) acetic acid This serves to fix the DNA and remove much of the urea, prior to drying Each 2 L of 10% acetic acid can be used to fix up to three gels

Although rigorous quantitative analysis is required for assessing the relative binding affinity at different sites, and for measuring bmdmg constants, the locatton of drug-induced footprmts can usually be directly assessed by visual mspectlon Quantitative analysis requires additional equipment (densitometer

or phosphorimager) and 1s beyond the scope of this chapter (see Chapter 2) However, since DNase I footprmts are necessarily larger than the actual hgand binding site, on account of the size of the enzyme, both visual and quantitative analyses leave some uncertainties The footprint will be larger than the binding site, and this too may be larger than the recognition site For example, although actinomycm D specttically recognizes the dmucleotide GpC, tt covers about 4 bp and protects about 6 bases from DNase I cleavage For small hgands that rec- ognize only 2 or 3 bp, and which may generate several discrete footprmts on any given DNA fragment, the ambiguity concermng the exact bindmg can often

be resolved by determmmg the sequences that are common to each of the foot- prints Additional mformatton is gleaned by comparmg the location of the foot- prints on each of the DNA strands, visualized by performing separate experiments with DNA labeled on each strand Since DNase I footprmts are staggered in the 3’ direction by 2-3 bases, the exact binding site will be located toward the 5’ end of each footprint and will be contained m the region of over- lap protected on both strands If there are still uncertamtres about the sequence recognitton properties, then it may be necessary to synthesize (a series of) syn- thetic fragments that contam putative binding sites based on the preliminary footprinting data An example of this is the AT-selective bifuncttonal intercalator TANDEM Footprmting experiments with natural DNA fragments confirmed the AT-selectivity, but could not determine whether the recognition site was ApT or TpA (7) This was resolved by producmg fragments contain- ing a series of different AT-rich binding sites, i.e., ATAT, TATA, TTAA, and AATT (53) These demonstrated that the recognition sate is TpA not ApT An alternative strategy is to use another footprmting agent such as MPE, hydroxyl radicals, mrcrococcal nuclease, DNase II, or uranyl radicals, though these suf- fer to different degrees from the same problems of locating the exact ligand binding site

3.7 A Worked Example

Figure 3 shows DNase I digestion of the tyr?” DNA fragment m the pres- ence of varying concentratrons of the AT-selective anttbiotrc distamycm The

14 Fox

20-

Fig 3 DNase I footprinting of distamycin on the 160 bp Qv-T DNA fragment, whose sequence is presented in Table 2 The EcoRI-AvaI fragment is labeled at the 3’ end of the EcoRI site The distamycin concentration (pA4) is shown at the top of the lanes Each pair of lanes corresponds to cleavage by the enzyme for 1 and 5 min

sequence of this DNA fragment is presented in Table 1 The DNA fragment in Fig 3 has been obtained by digesting with EcoRI and AvaI and has been labeled

at the 3’ end of the EcoRI site with a-32P dATP, using reverse transcriptase, revealing the bottom strand in Table 1 Since this fragment has been widely used as a footprinting substrate, the bands have been assigned by comparison with other published data Samples have been removed from the digestion mix- ture at times of 1 and 5 min This figure will be used to illustrate several aspects

D Nase I Foo tpnn tmg 75 Cleavage is also poor around position 100, m a GC-rich block In addition some positions are cut much better than the surroundmg bonds (e.g., 41, 69, and 81), whereas others are cut less well (e.g., 39,58, 83) The poor cutting m the AT-rich regions of the control presents an obvious problem for this hgand that 1s AT-selective since the binding sites correspond to regions where there is little or no cleavage m the control

Visual inspection reveals that distamycm has altered the DNase I cleavage pattern Clear protections from DNase I cleavage are evident at the lowest hgand concentration (0.2 PM) at positions 26-32 and 43-50 These sites corre- spond to regtons that are poor sites of DNase I cleavage in the control Other regions of protection can be seen at 1 and 5 )L&! at 56-68, 78-89, and around

110 Each of these positions corresponds to an AT-rich sequence The first contains two distamycin bmdmg sites (TTA and TAAA) that produce a single overlappmg footprint, as does the second (AAT and ATAT), whereas the third contams a single site TTAT At concentrations of 25 and 100 uM most of the cleavage in the lower portion of the fragment is protected It can be seen that each of these protections is staggered by 2-3 bases in the 3’ (lower) direction relative to the actual binding site For example, the protection around posi- non 60 extends down at least as far as posttion 56, whereas the AT-bindmg site ends at position 59 In contrast, the 5’ (upper) end of the footprmt is coincident with the edge of the binding sites (position 69) As a result of the overlapping footprints, and the poor cleavage of the enzyme around some bmdmg sites, it is not possible to determine the ligand bmdmg site size from these footprmts

The intensity of certain bands is increased at distamycm concentrations of

5 wand above, especrally at positions 72/73,94/95, and 99/l 00, each of which

is located m a GC-rich region Indeed at the highest lrgand concentration the bands at 72/73 and 94/95 are the only cleavage products remainmg These regions

of enhanced cleavage have previously been interpreted as arising from ligand induced changes in DNA structure (4) However, in view of small amount of free DNA available for enzyme cleavage these enhancements could simply reflect changes in the ratio of free DNA to enzyme (54,55) Since most of the enzyme binding sites are occupied by the ligand, the relative concentration of enzyme at these sites will be much greater, hence the greater cleavage effi- ciency (see Note 8)

It should be noted that, in this example, the 5-min lanes are overdigested; only a small proportion of the DNA is uncut As a result, bands toward the top

of the gel are much lighter, whereas those toward the bottom are overrepre- sented, since they arise from multrple cleavage events Although it is still pos- sible to discern the footprmting sites m the lower portion, this is less clear m the upper part, and could certamly not be used in any quantitative analysis

16 Fox Table 2

The Effect of Various Conditions on the Relative Concentration

of DNase I Required in Footprinting Experiments

Temperature concentration strength concentration pH concentration

of the enzyme The latter 1s generally varied A rough guide for the effect of various condltlons on the relative concentration of DNase I required IS presented

m Table 2 For mitral experiments it 1s often worth performing a time course for the enzyme digestion, increasing the volume of the reactants and removing all- quots e g., say, 1, 5, and 30 mm

2 DNase I requires the presence of dlvalent metal ions, particularly magnesium, and so Its action can be stopped by adding EDTA The enzyme has more than one bmdmg site for dlvalent catlons, though only one of these 1s at the catalytic site The literature on the preferred metal ions IS confusing with various claims for different sites for calcium and/or magnesium suggestmg that both ions are required However, good cleavage is observed with either calcium or magnesium, although slrghtly higher enzyme concentrations are reqmred when using calcmm alone Since manganese has been shown to increase the rate of digestton, equlmolar concentrations of manganese and magnesium are generally used It IS found that the cleavage pattern 1s largely unaffected by the nature of the divalent metal Ion, even though crystallographic data has suggested an alternative bmdmg site for manganese that might produce a different cleavage pattern In contrast, mllllmo- lar concentrations of ions such as Co*+ and Zn*+ inhlblt the activity of DNase I

3 DNase I 1s reasonably tolerant to a variety of organic solvents mcludmg metha- nol, ethanol, and dlmethylsulfoxlde (DMSO) This 1s useful since many DNA- bindmg ligands are only sparmgly soluble m water and must be prepared as stock solutions in various other solvents DMSO concentrations as high as 40% require

a threefold higher enzyme concentration, though this does modify the cleavage pattern, increasing the cuttmg m regions that are poor substrates for DNase I, such as polydA tracts

DNase I Footprinting

4 A glance at the literature reveals that many laboratories include known concen- tration of unlabeled carrier DNA m the footprmtmg reaction This is only neces- sary for experiments m which the absolute DNA concentration 1s needed (I e , some forms of quantitative footprmtmg analysis) and can be omitted for most experiments However, one advantage of mcludmg a fixed concentration of car- rler DNA IS that the concentration of DNase I required to produce a given level of cleavage does not vary between experiments m which the absolute amount of radlolabeled DNA may not be constant

5 In most footprintmg reactions the concentration of the target DNA IS vamshmgly small (nanomolar) whereas the DNA bmdmg ligand IS present m mlcromolar amounts The extent of bmdmg is, therefore, not determined by the stolchlometric ratio of drug to DNA, but by the equlhbrmm bmdmg constant In this regard footprinting reactions resemble typlcal pharmacological experiments, m which the concentration of the target site IS small and unknown and m which the prob- ability of each site being occupied is 50% at a ligand concentration equivalent to the equlhbrium dlssoclatlon constant Since many hgands bmd to DNA with affm- ties of between 1 and 100 PM’, drug concentrations between 1 and 100 @4 are usually examined For drugs that bmd more tightly, lower ligand Concentrations should be explored It IS generally best to test a range of hgand concentrations, extending down to a concentration at which the digestion IS not noticeably affected High hgand concentrations (100 CLM) often mhlblt DNase I digestion throughout the DNA fragment, this could be the result of nonspecific interaction with DNA

or direct inhibition of the enzyme itself

6 A major problem with using DNase I as a footprmtmg tool IS that the enzyme cuts different sequences with efficiencies that can vary over two orders of magm- tude These variations can be both local, m which isolated bonds are cut better or worse than average, or global, where long DNA regions are cut poorly In gen- eral, polydA polydT tracts are poor substrates for DNase I, on account of their narrow minor grooves GC-rich regions are also cut poorly, probably because they are more rigid and resist the bending that may be an important part of the DNase I catalytic reaction In addition, RpY steps are generally cut better than YpR Llgands that bind to those regions that are cut poorly by DNase I, produce footprmts that are difficult to detect The only way round this problem 1s to use a different footprmting probe

7 A similar problem 1s encountered when assessing the exact size of a footprint if bands at the edges of the footprint are cut poorly m the control Although this may be clarified by examining the cleavage of the other strand, the ambiguity often remains so that the footprmting site size can usually only be quoted to within

an accuracy of +l base

8 As well as producing footprmts, many hgands also generate enhanced DNase I cleavage m regions surrounding their binding sites These have been explained m two different ways, each of which is correct in different circumstances First, these may arise from drug-induced changes m DNA structure, which are propa- gated mto neighboring regions, and which render the DNA more susceptible to

18 Fox DNase I cleavage Second, they may simply reflect a change in the ratio of free DNA to enzyme m the presence of the ligand (5455) These two posslblhtles can only be properly dlstingulshed by quantitative footprmtmg experiments How- ever, a few other factors may indicate which is occurrmg Enhancements artsmg from changes m the ratio of free DNA to enzyme should be constant at all points

to which the hgand 1s not bound, whereas those that are directly attributable to hgand bmdmg will be located closest to the hgand bmdmg sites A further posse- blhty, which 1s rarely considered, 1s that of llgand-induced protections from enzyme cleavage, m surrounding regions

9 An apparently mmor detail, which 1s rarely addressed, concerns the hgand con- centration Does this refer to the actual concentration before or after adding the DNase 17 For a hgand m fast exchange with the DNA, a new equlhbrmm will rapidly be established after the small dilution because of the addltlon of the enzyme

In contrast, if the dlssoclatlon IS slow compared with the time course of the dlges- tion, then the dlstrlbutlon of the hgand will resemble the startmg condltlons throughout the reaction In the former case the hgand concentration should be that after adding the DNase I, whereas m the latter case this should refer to the concentration before In theory, the answer to the question requires some prior knowledge of the kinetics of hgand bmdmg, though m practice one or other 1s consistently adopted

10 Unwanted bands sometimes appear m the lanes, which clearly do not arlse from enzyme digestion These may be contaminants m the DNA preparation and can

be checked by running a sample of DNA that has not been digested with the enzyme Artlfactual bands, particularly depurmatlon products, can be produced

by the bollmg procedure These can be obviated by mcludmg a small amount of sodium hydroxide (l-2 m44) in the stop solution

11 Since DNase I cuts from the minor groove, protections are easiest to Interpret for llgands that also bind m this groove, sterlcally inhibiting enzyme activity How- ever, major groove bmdmg agents, such as triplex-formmg ohgonucleotldes, also generate clear DNase I footprints (12,13) In this case cleavage mhlbltlon cannot result from sterlc hmderance, but must arise from changes in the DNA structure and/or rigidity and are, therefore, less easily interpreted It should be noted that the footprmtmg pattern should still be staggered across the two strands by about 2-3 bases m the 3’ direction since this is a function of the cleavage agent, rather than the ligand under mvestlgatlon Agents that cut from the major groove would

be expected to generate a 5’ stagger

12 Another ambiguity m DNase I footprinting gels, which 1s rarely addressed, con- cerns the numbering/assignment of the cleavage products Although this would seem to be a trivial problem the uncertainty arises because, whereas most DNA sequences number the bases, DNase I cleavage products correspond to the phosphodlester bonds When Maxam-Gilbert markers are used alongslde DNase

I cleavage of 3’-end-labeled fragments, each band m the marker lane (X) comlgrates with the band corresponding to cleavage of the phosphodlester bond

on the 3’ side, 1 e , the XpY step

DNase I Foo tprin tmg 79

13 By adapting the simple footprmtmg protocol it can also be used for measurmg slow kinetic parameters, by removing samples from a reaction mixture and sub- jecting to short DNase I footprintmg (48,49)

14 It IS possible that some sequence selective compounds will not produce DNase I footprints if they are in rapid exchange with the DNA In such cases footprints can be induced by lowermg the temperature, thereby increasing then persistence time on the preferred binding sites (56)

Scamrov, A V and Beabealashvilh, R Sh (1983) Bmdmg of actmomycm D to DNA revealed by DNAase I footprintmg FEBS Lett 164, 97-101

Fox, K R and Warmg, M J (1984) DNA structural variatrons produced by actr- nomycm and distamycm as revealed by DNAase I footprmtmg Nuclezc Aczds Res 12,9271-9285

Fox, K R and Howarth, N R (1985) Investigations into the sequence-selective bmd- ing of muhramycm and related ligands to DNA Nuclezc Aczds Res 13,8695-87 14 Low, C M L , Drew, H R , and Waring, M J (1984) Sequence-specific binding

of echmomycm to DNA evidence for conformational changes affecting flanking sequences Nucleic Acids Res 12, 48654879

Low, C M L , Olsen, R K., and Warmg, M J (1984) Sequence preferences m the binding to DNA of triostm A and TANDEM as reported by DNase I footprmtmg FEBS Lett 176,4 14-4 19

Chaires, J B., Fox, K R , Herrera, J E., Britt, M , and Warmg, M J (1987) Site and sequence specificity of the daunomycin-DNA interaction Blochemzstry 26,8227-8236 Chanes, J B , Herrera, J E , and Waring, M J (1990) Preferential bindmg of daunomycm to S’(A/T)CG and S’(A/T)GC sequences revealed by footprmtmg titration experiments Brochemzstry 29, 614556153

Fox, K R and Warmg, M J (1986) Nucleotide sequence bmdmg preferences of nogalamycin investigated by DNase I footprintmg Bzochemzstry 25,4349-4356 Abu-Daya, A , Brown, P M., and Fox, K R (1995) DNA sequence preferences

of several AT-selecttve minor groove binding hgands Nucleic Acids Res 23, 3385-3392

Cooney, M., Czernuszewicz, G., Pastel, E H , Flmt, S J., and Hogan, M E (1988) Site-specific oligonucleotide binding represses transcription of the human c-myc gene in vitro Science 241,456459

20 Fox

13 Cheng, A -J and van Dyke, M W (1994) Oltgodeoxyribonucleotide length and sequence effects on mtermolecular purine-purme-pyrimidme triple-helix forma- tlon Nucleic Acids Res 22,4742-4747

14 Fox, K R and Waring, M J (1987) The use of micrococcal nuclease as a probe for drug-binding sites on DNA Blochrm Bzophys Acta 909, 145-l 55

15 Cons, B M G and Fox, K R (1990) The GC-selective hgand mtthramycm alters the structure of (AT), sequences flankmg its bmding sites FEBS Lett

264, lo&104

16 Stgman, D S (1990) Chemical nucleases Brochemlstry 29,9097-9105

17 Spassky, A and Slgamn, D S (1985) Nuclease acttvny of 1,lO phenanthrolme- copper ion conformational analysis and footprmting of the lac operon Blochem- wtry 24,8050-8056

18 Van Dyke, M W , Hertzberg, R P , and Dervan, P B (1982) Map of distamycin, netropsm and actmomycm binding sites on heterogeneous DNA DNA cleavage inhibition patterns with methidmmpropyl-EDTA-Fe(I1) Proc Nat1 Acad Scz USA 79,5470-5474

19 Van Dyke, M W and Dervan, P B (1983) Chromomycin, mithramycm and ohvomycin binding sites on heterogeneous deoxyribonucleic acid Footprintmg with (methidmmpropyl-EDTA)Iron(II) Biochemutry 22,2373-2377

20 Hertzberg, J P and Dervan, P B (1984) Cleavage of DNA with methidmmpropyl- EDTA-Iron(I1) reaction conditions and product analyses Blochemlstry 23, 3934-3945

2 1 Van Dyke, M W and Dervan, P B (1983) Methidmmpropyl-EDTA.Fe(II) and DNase I footprmtmg report different small molecule bmdmg site sizes on DNA Nuclerc Acids Res 10,5555-5567

22 Nielsen, P E., Jeppesen, C., and Buchardt, 0 (1988) Uranyl salts as photochemi- cal agents for cleavage of DNA and probing of protein DNA contacts FEBS Lett

235, 122-124

23 Nielsen, P E., Hiort, C , Sonmchsen, S H., Buchardt, O., Dahl, O., and Norden,

B (1993) DNA bmdmg and photocleavage by uranyl(VI)(UOZ2’) salts J Am Chem Sot 114,4967-4975

24 Cons, B M G and Fox, K R (1989) High Resolution hydroxyl radtcal footprmting of the bmdmg of mtthramycin and related antibiotics to DNA Nucleic Acids Res 17,5447-5459

25 Churchill, M E A , Hayes, J J , and Tullms, T D (1990) Detection of drug binding to DNA by hydroxyl radical footprintmg Relationship of distamycm binding sites to DNA structure and positioned nucleosomes on 5s RNA genes of Xenopus Biochemistry 29,6043-6050

26 Portugal, J and Warmg, M J (1987) Hydroxyl radical footprmtmg of the sequence- selective bmdmg of netropsm and distamycin to DNA FEBS Lett 225, 195-200

27 Drew, H R and Travers, A A (1984) DNA structural variations m the E colz tyrT promoter Cell 37,491-502

28 Drew, H R (1984) Structural specificrues of five commonly used DNA nucleases

J Mel Bzol 176,535-557

DNase I Footprintmg 21

29 Waterloh, K and Fox, K R (199 I) The effects of actmomycm on the structure of

dA, * dT, and (dA-dT), regions surroundmg its GC bmding site: a footprintmg study J Biol Chem 266,6381-6388

30 Waterloh, K and Fox, K R (1991) Interaction of echmomycm with A,, T, and (AT), regions flanking its CG bmding site Nucleic Acids Res 19,67 19-6724

3 1 Laskowskr, M (197 1) Deoxyrlbonuclease I, in The Enzymes, vol 4 (Boyer, P D ,

ed ), Academtce, London, pp 289-3 11,

32 Kumtz, M (1950) Crystallme deoxyribonuclease I isolation and general proper- ties spectrophotometric method for the measurement of deoxyribonuclease activ- ity J Gen Physzol 33, 349-369

33 Price, P A (1975) The essential role of Cazf m the activity of bovine pancreatic deoxyribonuclease J Blol Chem 250, 1981-1986

34 Lomonossoff, G P , Butler, P J G , and Klug, A (198 1) Sequence-dependent variation m the conformation of DNA J Mol BIO~ 149,745-760

35 Hogan, M E., Roberson, M W., and Austin, R H (1989) DNA flexibility varia- tion may dominate DNase I cleavage Proc Nat1 Acad Scz USA 86,9273-9277

36 Brukner, I., Jurukovski, V , and Savic, A (1990) Sequence-dependent structural variations of DNA revealed by DNase I Nuclezc Aczds Res 18, 89 l-894

37 Suck, D., Oefner, C , and Kabasch, W (1984) Three-dimensional structure of bovine pancreatic DNAase I at 2.5A resolution EMBO J 3, 2423-2430

38 Suck, D and Oefner, C (1986) Structure of DNaseI at 2A resolution suggests a mechanism for bmdmg to and cuttmg DNA Nature 321,62(X-625

39 Oefner, C and Suck, D (1986) Crystallographic refinement and structure of DNAase I at 2A resolution J Mol Blol 192, 605432

40 Suck, D , Lahm, A, and Oefner, C (1988) Structure refined to 2A of anicked octanulceotide complex with DNAase I Nature 332,464-468

4 1 Weston, S A , Lahm, A , and Suck, D (1992) X-ray structure of the DNase I- d(GGTATACC)2 complex at 2 3k resolution J Mol Bzol 226, 1237-1256

42 Lahm, A and Suck, D (1991) DNase I-induced DNA conformation 2A structure

of a DNase I-octamer complex J Mel Bzol 221, 645-667

43 Herrera, J E and Chaires, J B (1994) Characterization of preferred Deoxyribo- nuclease I cleavage sites J Mol Bzol 236,405-411

44 Bailly, C., Donker, I O., Gentle, D., Thornalley, M., and Warmg, M J (1994) Sequence selective binding to DNA of cis- and trans- butamidme analogues of the anti-Pneumocystis carmn pneumonia drug pentamidme MoZ Pharm 46,

47 Waterloh, K and Fox, K R (1992) Secondary (non-GpC) bmdmg sites for acti- nomycin on DNA Blochzm Biophys Acta 1131,300-306

22 Fox

48 Fletcher, M C and Fox, K R (1993) Vtsuahsmg the kmettcs of dtssoctatron ofactinomycin from mdtvrdual bmdmg sites m mixed sequence DNA by DNase I footprmting Nucleic Acids Res 21, 1339-l 344

49 Fletcher, M C and Fox, K R (1996) Dtssoctatton kmettcs of echmomycm from CpG sites m different sequence envtronment Bzochemzstry 35, 1064-l 075

50 Huang, Y -Q , Rehfuss, R P , LaPlante, S R., Boudreau, E Borer, P N , and Lane, M J (1988) Actmomycm D Induced DNAase I cleavage enhancement caused by sequence specttic propagation of an altered DNA structure Nuclezc Aczds Res 16, 11,125-l 1,139

5 1 Bishop, K D , Borer, P N , Huang, Y -Q., and Lane, M J (1991) Actmomycm D induced DNase I hypersensitivtty and asymmetrtc structure transmission m a DNA hexadecamer Nucleic Aclds Res 19, 87 l-875

52 Maxam, A M and Gilbert, W (1980) Sequencmg end labelled DNA wtth base- specific chemical cleavages Methods Enzymol 65,499-560

53 Lavesa, M., Olsen, R K , and Fox, K R (1993) Sequence spectfic bmdmg of [N- MeCys3,N-MeCys’] TANDEM to TpA Blochem I 289,605-607

54 Ward, B Rehfuss, R , Goodisman, J , and Dabrowtak, J C (1988) Rate enhance- ments m the DNase I footprmting experiment Nucfezc Aczds Res 16, 1359-l 369

55 Ward, B Rehfuss, R Goodisman, J., and Dabrowtak, J D (1988) Determination

of netropsm-DNA bmdmg constants from footprmtmg data Bzochemzstry 27,

1198-1205

56 Fox, K R and Warmg, M J (1987) Footprmtmg at low temperatures* evidence that ethidmm and other sample mtercalators can drscrimmate between different nucleottde sequences Nucleic Aczds Res 15,49 l-507

2

Quantitative DNA Footprinting

James C Dabrowiak, Jerry Goodisman, and Brian Ward

1 Introduction

Footprmting analysis has been used to identify the bmdmg sites of drugs and other hgands bound to DNA molecules (see Chapter 1) (1-3) It is particu- larly useful for equilibrium bmdmg drugs or hgands that leave no record of their residence position on DNA In the footprmtmg procedure, the hgand- DNA complex is exposed to an agent or probe that can cleave DNA, and the ohgonucleotide products from the cleavage reaction are separated using, for example, electrophoresis m a polyacrylamide gel If the hgand, when bound, inhibits cleavage by the probe, the ohgonucleotides that terminate at the hgand binding site will be underrepresented among the products analyzed using the sequencing gel This appears as omissions or “footprmts” m the spots on the sequencing autoradiogram

In quantitative footprmtmg, digests are carried out using different concen- trations of drug Then the drug binding can be seen as a decrease in the intensity

of a spot (corresponding to a particular cleavage site) with drug concentration Since the autoradiographic spot mtensities are directly proportional to oligo- nucleotide concentrations, they give the proportion of sites occupied by drug

so that from the dependence of spot mtenstty on drug concentration one may obtain the drug (or protein) bmdmg constant for a particular site, i.e., as a func- tion of sequence

In this chapter, we outline the approach used to obtain binding constants for drugs bound to DNA In Subheading 3.1., the experiment is reviewed and, m Subheadings 3.2.-3.3., the theory behind quantitative footprmtmg analysis is outlined The method is illustrated with published results (46) for the DNA sequence shown m Fig, 1 (Subheading 3.4.), with new results for ohgonucle- otide duplexes having only a single site (Subheading 3.5.) The drug used m

From Methods m Molecular Bology, Vol 90 Drug-DNA Interact/on Protocols

Edlted by K R Fox Humana Press Inc , Totowa, NJ

23

Fig 1, The sequence of a 139-bp fragment from pBR 322 DNA Strong and weak binding sites for ActD are indicated by filled and hatched rectangles, respectively (6)

both cases IS actinomycin D (ActD) Quantitative footprinting analysis is also ap- phed to determination of the dissociation constant of a triple helix formed from an ohgonucleotlde and a lineanzed double-stranded plasmld (Subheading 3.6.)

2 Materials

The materials and equipment necessary for quantitative footprinting analy- SIS are readily avallable m most biochemical laboratories The DNA substrate can be obtained from restriction cleavage of natural DNA& synthesized or generated using PCR It 1s advisable to purify the end-labeled DNA, using a gel to remove labeling reagents that may interfere with the equilibria being measured (7) If calf thymus DNA 1s to be added to the mixture, it should be deproteinized and sonicated prior to use No special treatment of the enzyme DNase I 1s necessary However, all commercial preparations of the enzyme slowly degrade m solution with time For this reason, calibrated stocks of DNase I should be stored at -20°C until needed (8) The sequencing gel, after electrophore- SIS, can be analyzed with a phosphorlmagmg device or by autoradlography/ mlcrodensitometry to obtain quantities proportional to DNA concentrations The concentrations can be used to measure ligand binding constants according

to the method outlined in Subheaading 3.2

3 Methods

3.7 The Footprinting Experiment: General Considerations

The interpretation of the quantitative footprmtmg experiment 1s slmphfied when one terminates the cleavage reaction with -80% of the full-length DNA uncleaved This ensures that the products are the result of a single cleavage m the full-length fragment of DNA In this “single-hit” regime, the amount of each ohgomer 1s proportional to the probability of cleavage at the correspondmg

Quantitative DNA Footprintmg

a titration curve, one wants more points for drug concentrations for which the occupation probabthty of a site varies, and fewer for drug concentrations cor- responding to zero occupation or complete occupation Afterward, experiments are performed using drug concentrattons in the range identified From quanti- tation of the resulting gel, one obtains spot intenstties as a function of sequence and drug concentration In principle, one has carried out a series of digests of identical DNA fragments in the presence of varying amounts of drug, but otherwise under identical conditions The “total cut” plot, the sum of the spot intensities as a function of drug concentration, is shown for actmomycin D interacting with a 139-bp fragment from pBR 322 DNA m Fig 2 (4)

To account for lane-to-lane differences, a “total cut” plot, the sum of all cleavage products vs the drug concentration, 1s constructed Since this plot is a smooth function of drug concentration, deviattons from the curve are due to experimental error A least-square-fit stratght line is shown m Fig 2; in many

cases, a horizontal line, i.e., total cut = constant, fits the data as well as a func-

tion contammg more parameters To correct for experimental error, all spot

26 Dabrowiak et al mtenstties m a lane are multiplied by a common factor, the ratlo of the value

of the smooth curve to the actual spot intensity sum After making this correc- tion, one constructs a plot of spot intensity vs drug concentration for each oh- gomer These plots are referred to as “footprinting plots.”

Some footprmting plots for ActD bmding to a 139-mer are shown in Fig 3 Their shapes can be explamed by noting that the spot Intensity is proportional

to the rate of cleavage at a nucleotide position because the digest time IS con- stant The rate of cleavage at site 1, m turn, may be written as:

(rate), = k, [probe], where k, is the rate constant for cleavage at site i, and [probe],, the effective concentration of cleavage agent at that site, may depend on drug concentration For a nucleotide position within a drug-bmdmg site, [probe], decreases as drug IS added to the system because this increases the probability that drug ~111

be bound at that site and cleavage agent cannot bmd where drug is already present This is the classic footprmtmg phenomenon; it predicts a monotomc decrease of spot mtensity with drug concentration If the drug-bmding constant

is much larger than that for probe, so that a drug molecule always displaces a probe, [probe], will be proportional to I-vk, where vk IS the fraction of sites 1 having drug bound For a nucleotide site not within a drug bmdmg site, the spot mtensity should not depend on drug concentration

For relatively long DNA molecules, spot mtensities correspondmg to sites between drug-bmdmg sites are observed to mcrease as drug is added to the system If DNA is not saturated with probe, increased cleavage with added drug may occur because bound drug decreases the amount of cleavage agent at drug binding sites and, hence, must increase tt elsewhere, i.e., m bulk solution and at sites not blocked by drug The increase m [probe], is important when the ratio of probe concentration to DNA concentration 1s small It ~111 not occur when the DNA is saturated wtth probe It is also possible that drug binding induces a structural change in the DNA, changing the cleavage rate constant k, This could lead to either an increase or a decrease in (rate),, superposed on the mass-action effect just mentioned Intercalating drugs like ActD are more likely

to cause a distortion m DNA upon bmdmg than are groove-bmdmg drugs like netropsin (9) Alterations in cleavage rate constants k, may explain apparent oscll- lations seen in some footprintmg plots for low drug concentrations, O-2 l.tA4, Fig 3 Some of the footprinting plots shown seem to be composites: cleavage is first enhanced and then mhlbtted by increased drug concentration Nonmonotomc footprmting plots arise for cleavage sites within secondary drug-binding sites, with lower bmdmg constants than primary drug-binding sites The explanation

is that, for lower drug concentrations, drugs bind at the primary sites, displacmg probe and leading to enhanced cleavage at other sites At higher drug concen-

Quantitative DNA Foo tprin tmg

tratlons, drug binds at the secondary sites, thus blocking probe binding and decreasing cleavage

Footprinting plots are like titration curves except that fractronal occupation

is being plotted agamst total drug concentratron rather than against free-drug concentration Although footprinting plots are not linear, it IS sometimes useful

to fit their behavior at low drug concentration to a straight line Then one can

28 Dabrowiak et al calculate “mitial relative slopes,” i.e., the slope of the lme divided by the intercept Plotted as a function of sequence of the fragment, the mitral relative slopes clearly show the positions of drug binding They may also point to possible drug- induced structural changes and enhancements caused by the mass-action effect 3.2 Constructing the Model

The first step m constructmg a model for drug bmdmg and tts effect on cleavage is to inspect footprmting plots or initial relative slopes One can deduce the size of the inhibition region, i.e., how many sites are blocked by binding of a drug molecule One wants footprmting plots for as many sites as possible, but, for some sites, if DNase I is the cutting agent, mtensmes will be low, and reliable mformation on the effect of drug concentratton on cleavage will be difficult to obtam Another comphcation is that single-site resolution cannot be obtained over the entire sequencing autoradiogram Resolution along the DNA helix decreases as ohgomer length increases, so one obtains less mformatton about bmdmg sites that are far from the radiolabel

The size of the mhibmon region depends on the drug as well as the probe For ActD, the preferred intercalation site for the phenoxazone rmg 1s 5’-GC-3’, with the two cychc pentapeptides displaced 1 bp to either side of the mtercala- tion site (9) A single ActD molecule would thus cover about 4 bp of DNA However, DNase I has a small loop, important to the binding process, that fits mto the minor groove of DNA, and spans 3-4 bp of DNA (20) Smce the cata- lytic site is located at the end of the loop, there IS an mhibmon region of 3-4 bp

to the 3’ side of a bound ActD molecule The enzyme-DNA contacts on the other side of the catalytic site are relatively weak, so the enzyme can probably cleave near the 5’ edge of a drug site Thus, the inhibition region would be 7-8

bp m length For an isolated drug-binding site j, the concentratton of bound drug is related to the bmdmg constant KJ by:

K, = cJ (c - c,>Do Here, c, 1s the concentration of sites J at which drug is bound, c is the total concentration of sites J (equal to the concentration of DNA m molecules), and

D, is the free-drug concentration The probability that a probe site i wtthm the inhibition region of drug site j is blocked by drug, v,, is equal to c,/c Then [probe], is proportional to:

1 - v, = (1 + K,D,)-’

If two drug sites j and k, with bmdmg constants K, and &, are near enough to each other so that a probe site 1 may be blocked by drug bmdmg at either site J

or site k:

Quantitative DNA Footprintmg 29

l-vl=il-~h-y-~=(~ +K,p):l +KkD 0 0 >

If, however, two drug sites j and k are so close that drug binding at one pre- vents drug binding at the other, the probability of drug bmdmg at site j IS K,D,l (1 + K,D, + &D,,), with a similar expression for the probability of binding at site k Then the probability that there is no drug blocking probe bmdmg at probe site i 1s:

(1 + KJDo +K,D,) Because the drug sites are not independent, one cannot srmply multiply together the probabilmes that each one is vacant

The total concentration of probe at all cleavage sites depends on the drug- probe competition If the total cut IS constant as a function of drug concentra- tion, the system behaves as if the amount of probe available for binding to DNA is unchanged by the addition of drug Added drug decreases probe con- centration at drug-bindmg sites of the DNA and increases it at other sites If the number of avallable probe molecules is small compared to U, the number of unblocked sites, the concentration of probe at such sites should be inversely proportional to U This effect IS represented by dividing [probe], by (1 -K.&, where cb IS the total concentration of drug bound to DNA and the “enhance- ment constant” K, is a parameter whose value is determined along with drug- bindmg constants

If the probe and the hgand are specific for certain sites, or the DNA mol- ecule IS so small that drug and probe compete for a common site, one can consider explicitly the competmve equilibrium between drug and probe The probabtlity that probe is bound at a site is determmed by the stmultaneous equilibrium expressions*

(c - cl - q,,)Do (c - Cl - CpJPo

Here, c, is the concentration of sites I with drug bound and c,,, the concentration

of sites i with probe bound, so that c-c, -cP, is the concentration of empty sites

I D, and PO are the concentrations of unbound drug and unbound probe, so tgb§, for isolated sites:

with D, and Pt the (known) total drug and probe concentrattons The above equations are solved simultaneously to get c, and c,,,

30 Dabrowiak et al Often, unlabeled carrier DNA is present in addition to the DNA fragment whose cleavage products are measured The carrier, if present m large excess over the fragment, determines the concentration of free drug present m the system One has to model drug bmdmg by the carrier because the free-drug concentration enters the bindmg constant expressions for the radiolabeled frag- ment If the carrier DNA has approximately the same base-pair composition as the fragment, the number of strong drug-bmdmJpsites on the carrier can be estimated from the number on the fragment Otherwise, one may represent drug bmdmg to the carrier in terms of an effective concentration of strong sttes on the carrier, c,, and an average binding constant for these sites, KC Then the amount of drug bound to the carrier, cb, 1s gtven fi :

of c, and Kc may be determined reliably A fragment may have weak-bmding sites in addition to strong sites If one goes to high enough drug concentration so that binding to the weaker sites 1s important, one must consider similar weak sites on the carrier

To summarize, there are a number of causes for an alteration m the cleavage rate of a probe in drug-DNA footprmtmg experiments Although enhancements are often attributed to drug- or protein-induced structural changes in DNA, there are other factors that can affect the cleavage rate without changing the cleavage rate constant

To calculate i,,, [probe], 1s evaluated for thejth total-drug concentration using given values of the parameters K,, c, Kc, and K,, The drug-bmndmg constants deter- mine the probability that site 1 is blocked by drug, and [probe], is proportional to 1

Quantitative DNA Footprin ting

minus thts probability The free-drug concentration, required to calculate the probability of drug binding, must be calculated from the known total drug con- centration Dt by combmmg 0, = D, - ,S,c, with the equiltbrmm-constant expres- stons If a carrier is present, binding to the carrier must also be considered usmg a single site of concentration c, and equilibrium constant Kc, unless the drug concentrations used are high enough to require modeling weak sites on the carrier as well The cleavage rate constants k,, which multiply [probe], to give the cleavage rate, are additional parameters; smce these are linear, the best values to use for them can be determined analytically For the nonlinear parameters, a systematic search algorithm in multiparameter space is required The search procedure used 1s the Simplex Method

It is conceivable that there are several relative minima for D, and there IS no way to guarantee that the minimum found is the true absolute mmimum To gain confidence in the result, one may carry out the simplex search several times with different starting points

If the deviations of the calculated i,] from the experimental intensities I,J are no larger in size than the fluctuations of I,J from one drug concentratton to the next, the model is said to fit the data, Naturally, use of a model wtth more parameters (for example, describing cooperative or anticooperative drug bmdmg) will always give a smaller value of D, but it makes no sense to increase the number of param- eters when the deviations )I,, -i,, ] are already smaller than the experimental error The more parameters used, moreover, the less likely it will be that they will be mutually independent, which will make the search procedure work less well

To assign a precision to the values of parameters determined m this way, one can examme the effect of a change m the value of one parameter on the value of D It is usually found that changing an equilibrium constant by a few percent changes D by 10% or more Naturally, if the bmdmg constants have different magnitudes, D is least sensitive to the values of the smallest Note also that one can generally change a parameter by more than a few percent without much changing D, if other parameters are allowed to adjust then values 3.4 Application to a Multisite Problem

The approach outlined in Subheadings 3.2.-3.3 is now applied to the multisite restriction fragment of Fig 1 The authors’ first work concentrated (4,s) on the strong drug-binding sites, usmg only footprmtmg data for low drug concentration, but a later study, using data for more sites and for a wider range of drug concentration, obtained binding constants for the weaker sues as well (6) For this system, drug binding to DNA causes the cleavage agent, DNase

I, to redistribute to DNA sites not blocked by bound drug (11)

The DNase I footprmtmg experiments were carried out in the presence of calf thymus DNA as a carrier (193 @4 in bp), usmg as many as 26 different

32 Dabrowiak et al ActD concentrations from 0 to 38.8 pJ4 The mtenstties of spots corresponding

to cut fragments were obtained by microdensitometric scanning of the sequenc-

mg autoradiogram The resulting total-cut plot, shown m Fig 2, was used to correct for lane-to-lane vartattons The linear fit, shown m Fig 2, was 247 1 - 0.825c,, with a mean-square devtatton of 19.9; fittmg to a quadratic function gave a mean-square devlatton of 18.4 The average of the total cleavage was 239.1, with a mean-square devtatton of 22.2 The decrease m mean-square deviation with more complicated functtons is not stattsttcally significant, and

it 1s concluded that total cleavage IS essentially constant as a function of drug concentration

Total-cut corrected mtenstttes were used to construct footprmtmg plots for

69 sites on the 139-mer, a few of which are shown m Fig 3 To determine the bmdmg regions, “imtlal relative slopes” were obtamed and plotted vs site Drug-bmdmg regions were clearly apparent as negative initial relative slopes surrounded by large positive slopes The sequences at which the negative slopes appeared indicated that ActD binds most strongly to S-GC-3’ and that the mhi- bttion region extends from -3 bp to the 5’ side of G to -2 bp to the 3’ side of C, for a total length of 6-7 bp

The first study, focusing on the stronger drug-binding sites, included data from

32 sites, excluding sites for which spot mtensmes were too low for reliable mea- surement These sites were: 54-56, 60, 62-69, 71-72, 85, 87, 98, 99, 102, 103,

106, 112, 114, 120, 124, 128, 133, 136, 138, 143, 145, and 161 (Fig 1) As mentioned, resolution decreases for the longer fragments Sites 54-56, 85, 87,

106, 112, 114, 120, 143, and 145 behaved as enhancement sites, whereas the others showed inhibition of cleavage due to drug bmdmg Cutting at sites from

6 1 to 72 was inhibited by drug binding to the GC’s at 63-64 and 69-70; cuttmg at sites from 98 to 105 was inhibited by drug binding to the GC’s at 10 1 - 102 and 103- 104; sites from 133 to 138 were influenced by the GC drug-binding site at 137- 138; and site 16 1 was influenced by the GC site at 160- 16 1 Sites 124 and

128 exhibited inhibition, but less than those just listed It is believed they point to

a weak drug-binding site that is not 5’-GC-3’, but has the sequence 5’-CGTC-3’ Note that the drug-binding sites at 101-102 and 103- 104 are expected to be mutually exclusive: they cannot bind drug simultaneously However, the sites at 63-64 and 69-70 seem far enough apart to permit simultaneous drug binding at both There is weak binding to the sequence 5’-GGC-3’ This is shown m the footprinting plots for sites having this sequence on the 139-mer

Data were used for 19 drug concentrations from 0 to 12.4 pM, so that there were

608 data points, to be used to determine rune nonlinear parameters, These were the seven drug-binding constants, for the six GC sites and the site at 124- 127, the drug binding constant to the caner, and an enhancement constant K, The values of all the nme parameters were determined by calculating intensities to

Quantitative DNA Footprinting 33

be compared with all 608 measured intensities and minimizing D with respect

to the parameters

Since the concentration of carrier DNA (193 @4 in bp) far exceeds the concentration of fragment DNA (estimated to be 0.2 FM m bp), the free-drug concentratton for any total-drug concentration IS mainly determined by the equihbrium for carrier DNA The free-drug concentrations were used in the drug-fragment DNA equilibrium expressions to calculate fractional occupa- tion of sites and hence inhibition of binding of cleavage agent To estimate the concentration of strong drug-binding sites on the carrier, it is noted that, in a 114-bp segment of the 139-mer, there are five strong ActD sites (excluding the site at 124-127) Then, if the bases in the (calfthymus) carrier DNA are dtstrib- uted like the bases in the fragment DNA, 193 pA4 bp concentration should provide a strong-site concentration of (5/l 14)( 193) = 8.5 pA4, or considering the two mutually exclusive sites as a single s&-(4/1 14)( 193) = 6.8 PM If the carrier is considered as a random arrangement of base pairs with a fraction 0.6 being A or T and a fraction 0.4 G or C, the probability of tindmg a G or a C at

an arbitrary position is 0.2, and the probabihty of finding a GC with no G to the 3’ side (i.e., not GGC) is (0.8)(0.2)(0.2) = 0.032 Then the concentration of strong actinomycin sites on the carrier is estimated as (0.032) (193 PM) = 6.2 @4 One can also determine this concentration, c,, using the footprrntmg data, by making c, an additional parameter to be varied m the mmlmizatton of D It was found that D went through a minimum as a function of c, at c, = 4.7 @4, In this case, c, and K, are not mutually dependent

The footprmting plots for the higher drug concentrations (> 10 PM) suggest there are many additional bmdmg sites on the fragment, with lower binding constants than those considered so far Such sites must exist on the carrier as well, and be considered in a model to explain footpnntmg data for higher total- drug concentrations Their mclusion lowers the free-drug concentration D, for any total-drug concentration, and leads to higher apparent binding constants to fragment sites Therefore, weak-binding carrier sites were added with total effective concentration c, and average binding constant K, to the model, If c,

is the concentration of drug bound to weak sites:

Kw = (cw ~&A7 The total drug concentration then satisfies:

Dt = Do + cb + c = Do + ccKQO + cwKwDo

1 + K,D,, 1 + K,D,

In these calculations, c, = 5 @4 and first estimated c, from a consideratron of the relative numbers of strong and weak sites on the fragment Later, c,, K,,

34 Dabrowiak et al and K, were determmed along with the other nonhnear parameters by mimmiz- ing D, giving c, near 10 pM

The values found for the fragment-binding constants for the strong sites were somewhat lower than those reported by Chen (12) for ActD bmdmg to small oligonucleotide duplexes, measured optically Phase partition studies of Winkle and Krugh (13) on polymeric DNAs such as poly dG-poly dG yielded bmdmg constants consistent with those obtained from the footprinting experiments This suggests that small ohgomers have higher binding constants for this drug than do polymeric DNAs As noted, the footprintmg data show that ActD binds

to the sequence S-CGTC-3’, that does not contam a 5’-GC-3’ site This is con- sistent with the report of Snyder et al (14) that two ActD molecules bmd to the self-complementary duplex d(CGTCGACG)* The binding is cooperative and the complex exhibits aberrant spectroscopic and calorimetric behavior, sug- gesting that binding at this site is different from that at sites having 5’-GC-3’ The apparent binding constant reported by Snyder et al (14) is 1.5 x 107M-‘, about two orders of magnitude higher than this value, perhaps because of the effect of DNA length or to the fact that the spectroscopic/calorimetric experi- ments measure two events that are cooperative

The analysis does not consider drug-induced structural changes in DNA If drug binding at one site causes a structural change, it could affect cleavage within a second drug-binding site, and change the appearance of the correspondmg footprintmg plots, There are, m most of the footprmtmg plots, noticeable at an ActD concentration

of -2 rnM, which may be the result of structural changes on the fragment and/or the carrier DNA For example, it is known that mtercalation of ActD bends the DNA helix (15) This could release hgand to solution or decrease the free-hgand concentra- tion by enhanced binding The anomalous footprinting plots for sites such as 58 and

59 (Fig 3) may also be the result of alterations m DNA structure The mitral relative slopes of the plots for 58 and 59 are, respectively, above and below what is expected from the simple mass-action mechanism A DNA-cleaving metalloporphyrm, like DNase I, shows anomalous cleavage rates m this region of the 139-mer in ActD footprinting experiments (4), Since groove width and DNA flexibility are known to affect DNase I cleavage, mtercalation by ActD at nearby sites could affect cleavage in this region

A later analysis of footprinting data for this system used data for 26 actinomycin concentrations from 0 to 38.8 rnM, m order to identify the weaker binding sites and derive their binding constants Since the actmomycin concentrations were large enough to show binding to the weak sites, it was necessary to include weak as well as strong sites in modeling the carrier This work also allowed for closer consideration of possible structural changes m DNA The HzndIIIINczI 139-bp restriction fragment from pBR-322 DNA was end-labeled at position 33(A) for one set of experiments and at position

Quan tita We DNA Footprin ting 35 172(G) for a second set (6) In the second set, the labeled fragments produced

by cleavage at the higher-numbered sites were shorter, yielding better site reso- lution for these sites However, only 10 drug concentrations were used in these experiments: 2.48, 3.40, 4.86, 6.93, 9.89, 14.1, 20.2,28.8,41.0, and 58.7 PM After rejecting data for sites showing very low or unreliable spot intensities, data were retained for 54 sites for the A-label gel, and 43 sites for the G-label gel For the A-label experiment, the footprinting plots were similar to those of the G-label experiment: some showed a decrease in cleavage with increased drug concentration, correspondmg to drug bindmg Interfering with cleavage

by the enzyme Those showing the most rapid decreases were associated with the strongest binding sites Some sites showed an increase in cleavage for low drug concentration, followed by a decrease, explained by nearby weak drug sites not occupied by drug until the drug concentration reached a high value Other sites showed a rapid increase or enhancement m cleavage with drug con- centration, believed to be caused by the mass-action effect, bound drug dis- placing cleavage agent to sites where no drug 1s bound A few sates showed only

a slow Increase in cleavage with drug concentration, interpreted as pointing to very weak drug sites for which drug bindmg canceled some of the enhance- ment effect because of mass action The weak binding sites found from this qualitative analysis of the footprintmg plots had sequences* GGC (at 76-78), CCG (at 80-82), GGC (at 119-121), CCGT (at 123-126), CCC (at 129-131), GGC (at 143-145), GGC (at 149-151), GCCGG (at 160-164), and other sequences near 86 and 112 The strong binding sites were those identified m the previous work: the sequence GC at 63-64 69-70 101-102, 103-104, 137-138, and 160-161

The footprinting data from the G-label gel were analyzed using the model of strong and weak drug-binding sites developed from the analysis of the A-label gel The data from the G-label gel showed more scatter than the data from the A-label gel, as can be seen on comparing the total-cut plots (6) Interestingly, there seems to be a drop-off in the total cut near drug concentration of 20

@4, suggesting that this is a real effect Because there were fewer data points, fewer rehable values for binding constants were obtained from the G-label gel The bmding constants are compared to those from the A-label gel in Table 1 It should be noted that site resolution for the A-label gel is highest for smaller site numbers and the reverse for the G-label gel Therefore, the first few binding constants will be determined more reliably from the A-label gel, and the last few will be determined more reliably from the G-label gel, In general, binding constants from the two analyses agree to within a factor of two (note that the binding constants span two orders of magnitude), except for the TGCT site at 62-65, for which one must take the value from the A- label gel as the valid one

36 Dabrowiak et al Table 1

ActD-Binding Constants on 139-bp Restriction Fragment, in (@W-1 (6) Posrtion Sequence From A-label gel From G-label gel

Another problem 1s that the values of all drug-binding constants depend to some extent on how the carrier is modeled In this work, the carrier was considered to have both strong and weak sites, requrrmg four parameters, two (average) bmd- ing constants and two (effective) concentrations The concentration of strong sites was fixed at 5 @4, based on earlier work, and the other three parameters were varied Their values, determined by mmimtzatton of D, were 10 PM, 1.1 x 1 O7 W*, and 4.7 x 1 O5 M-i, respectively

Although the average deviation between experimental and calculated inten- sities approached the estimated experimental error, the deviations in certain footprintmg plots remained significant Some experimental plots had shapes that could not be explamed by the model For example, mtensities for site 59, Fig 3, modeled as an enhancement site, are roughly constant for drug concen-

Quan tita We DNA Foo tprin ting 37 trations ~20 pA4, and also constant, but at about double the original value, for concentrations ~30 piH Other footprmting plots seem to be responding to drug binding, but are not near any site at which drug could reasonably be expected

to bind Also, many footprintmg plots show a small but abrupt decrease in Intensity near 2 PMdrug concentration, followed by an abrupt increase These effects were considered in a second publication, which attempted to show how one could distmguish between enhancements caused by structural effects and the mass-action effect (II)

3.5 Single-Site Problem: ActD Binding to Dodecamers

The analysis for cleavage of small, single-site, oligonucleotides by DNase I

IS given here (16) Footprmtmg titration studies were performed on several different self-complementary 16-bp sequences containing actmomycm-binding sites The sequences for a single strand were:

GC 1: 5’-CTTTTTTGCAAAAAAG-3’

GC 1 AT S-CATATATGCATATATG-3’

Intensities corresponding to cut fragments of various lengths, as well as the full-length, uncut fragment, were measured for different concentrations of ActD Several sets of intensities were collected for each olrgomer Many included intensities for cleavage at all sites from 5 through 16 (uncut ohgomers); for some, lack of resolution made it necessary to combme intensities for several sites The nucleotide positions on the duplexes are numbered from left to right

on the sequences shown above The concentration of DNase I was -0.1 @4, the concentration of hexadecamers was 0.625 uA4, and the concentration of actinomycin varied from 0 to 3 1 uLM m some data sets and from 0 to 100 ~IV m others To correct for loading errors and differences in digest time, the total cut was calculated and fitted to a linear (decreasing) function of drug concentra- tion Intensities for each drug concentration were then corrected as discussed

m Subheading 3.2 From plots of corrected intensities vs drug concentration,

it was easy to determine which cleavage sites are blocked by drug

For GCl, intensities for sites 10 through 7 decreased strongly with mcreas- ing drug concentration, and intensities for sites 11 and 6 less strongly, mdicat- ing the end of the blockage region This means that the blockage region is less than 8 bp long, with GC (at sites 8 and 9) approximately in the middle, For

GC 1 AT, inhibition of cleavage by drug was evident for sites 5 through 12, and less evident for site 13

The data for the inhibition sites on GCl were analyzed according to the competitive-binding model, in which each site can be empty, occupied by drug,

or occupied by probe (DNase I), and the probability of cleavage (and hence

38 Dabrowiak et al spot intensity) is proporttonal to the occupation by probe For each total-drug concentration D,, one solves the stmultaneous equthbrmm expressions:

(C - CVb - CVp) (Dt - CVb)

and Kp = CVP

(C - Cvl, - C’.‘p) (P, - Cvp)

to obtam the concentration of probe bound at a site Here, c is the concentration

of sites, nb and nP are the fraction of sites wtth drug and probe bound, respec- ttvely, and Pt IS the total probe concentration The amount of fragment pro- duced by cleavage at a site is assumed proportional to n,, Binding constants for both probe (K,, assumed the same for all sites) and drug (K) are determmed by seeking the values of these parameters, whtch give the best fit of calculated to experimental intensities Since only mhibttion sites are considered, t-t,, is the same for all sites, so the theoretical curves of spot mtensity vs drug concentra- tion for different sites differ only by a multtphcative constant

It was found that mtenstties for cut fragments did not approach zero when the actinomycm concentration approached zero, mdicatmg that fragments of length less than 16 were present m the origmal DNA To represent this, it was assumed that the intensity of fragment I for total drug concentration D, is*

Here A, and B, are constants (different for different sites) to be determined by fitting to experimental intenstttes, B, giving the intensity because of fragments

of length i present in the original DNA

For analyzing the GCI data, intensities for drug-bmdmg sites 6 through 11 were used Most data sets mvolve 2 1 drug concentrations, so there are 126 data pomts In addition to the drug-bindmg constant K and the probe-binding con- stant Z$,, there are 12 lmear parameters, A, and B, for each sue 1 Values of parameters are chosen to mnumize the sum of the squared deviations of calculated from expertmental mtensittes For GC 1 AT, data for five or six drug- bmding sttes were used, since mtensmes for mdivtdual sites could not always

be resolved Most of the data sets include mtensities for 21 drug concentra- tions Some representative results are shown m Fig 4

The determined values of K for GCl (four sets of expertments used) and GCl AT (five sets of experiments) are given m Table 2 Values of tP determined from the GC 1 AT experiments are also gtven For GC 1, the average K 1s 0.180 pm’ with the root-mean-square deviation from the average 0.082 PM-‘, For GC 1 AT, the average K is 0 168 pm’ with the root-mean-square deviation from the average 0.021 pm’ It does not seem that there is a significant difference be- tween the drug-binding constants for GC 1 and GC 1 AT In contrast, actinomy- tin binding constants for strong sttes on restrictton fragments vary widely, dependmg on the sttes netghbormg the GC

Quantitative DNA Footpnnting

Table 2 Binding Constant K for ActD Binding

to Oligonucleotide Duplexes in @Pi (16)

3.6 Measurement of Triple-Helix Dissociation Constant

using a Type IIS Restriction Enzyme as Probe

The footprinting method has been used to determme the dtssociatton con- stant of a triple heltx, formed by interaction of the ohgonucleottde dT,, and the 272%bp plasmid pA20 shown below (17) The plasmid was constructed to contain a target sequence for dT2a as well as three cleavage sites for the type IIS restriction enzyme Eco571, at bp 429, 1375, and 2423, the first lying wtthm the dTZO target sequence Cleavage and end-labeling of the plasmid at the Me1 site (position 183) produced the doubly end-labeled lmeartzed plasmtd shown below (The binding region for dT,, is shown by x’s)

xxxxxxx

Limited (single-hit) digests of linear pA20 with Eco571 could produce seven possible labeled fragments; m fact, they produced mostly fragments of lengths

Intensities of bands 1 and 2 were measured for fourteen values of [dT,,]: 0 and 13 concentrations from 0.001 to 4.16 PM After subtracting background (intensities m the absence of enzyme), the intensity for band 2 for each [dT2e] was divided by the corresponding intensity for band 1 to produce the experi- mental points plotted m Figure 5 These were used to find the dissociation equilibrium constant Kd, where:

Kd = ([TWPIY[-U

Here, D refers to duplex, T refers to triplex (i.e., duplex with dTZO bound), and [TFO] is the concentration of free dT,, (TFO = triple-helix-forming oligonucle- otide) If only duplex, and not triplex, can be cleaved by Eco571, the intensity

of band 2 (I), divided by the intensity in the absence of drug (IO), should be equal to the fraction of duplexes with no dTZO bound Therefore:

1’ [D] + [T] & + [TFO]

As indicated above, the concentration of free dTZO was considered to be equal

to the total concentration of dT2,

The assumption that I/I0 IS equal to the fraction of duplexes with no dTZO bound is valid when the binding constant of dT,a to the duplex is much larger than the binding constant of Eco571 to its sites on the duplex Then bound ltgand always displaces probe To show that this 1s the case m these measure- ments, the mtensitres of bands 1 and 2 were measured for 15 concentrations of Eco571, in the absence of dT2e and in the presence of 0.076 pMdT,a For band

I, intensities were not changed by the dT,,; for band 2, dTZO reduced mtensmes

by the same factor for all Eco571 concentrations, as should be the case tf bound ligand always displaces probe