most appropriate probe for your research situation(Anderson 1999: Nath and Johnson, 1998; Temsamani and Agrawal, 1996: Trayhurn, 1996: Mansfield et aL., 1995; Tijssen, 2000). The questions raised in the ensuing discussions demonstrate why only the actual experiment, validated by positive and negative controls, deter mines the best choice The purpose of the following example is to discuss some of the complexities involved in selecting a labeling strategy. Suppose that you have the option of screening a target with a probe generated from the following templates: a 30 base oligo(30mer), a double stranded 800bp DNA fragment, or a double-stranded 2kb fragment 30-mer The 30-mer could be radioactively labeled at the 5 end via T4 polynucleotide kinase(PNK) or at the 3 end via Terminal deoxynucleotidyl transferase(TdT). pNK attaches a single mole cule of radioactive phosphate whereas TdT reactions are usually designed to add 10 or less nucleotides pnK does not produce the hottest probe, since only one radioactive label is attached, but the replacement of unlabeled phosphorous by P will not alter probe structure or specificity. TdT can produce a probe containing more radioactive label, but this gain in signal strength could be offset by altered specificity caused by the addition of multiple nucleotides. A 30mer containing multiple nonradioactive labels ould also be manufactured on a DNA synthesizer, but the pres ence of too many modified bases may alter the probe's hybridiza tion characteristics(Kolocheva et al., 1996) 800bp fragment The double-stranded 800 bp fragment could also be end labeled but labeling efficiency will vary depending on the presence of blunt, recessed, or overhanging termini. Since the complementary strands of the 800 bp fragment can reanneal after labeling,a reduced amount of probe might be available to bind to the target. Unlabeled template will also compete with labeled probes for target binding reducing signal output further. However, probe syn thesis from templates covalently attached to solid supports might overcome this drawback(Andreadis and Chrisey, 2000) Random hexamer- or nanomer- primed and nick translation labeling of the 800bp fragment will generate hotter probes than end labeling. However, they will be heterogeneous in size and specificity, since they originate from random location in the tem Herzer and Englert
most appropriate probe for your research situation (Anderson, 1999; Nath and Johnson, 1998; Temsamani and Agrawal, 1996; Trayhurn, 1996; Mansfield et al., 1995;Tijssen, 2000).The questions raised in the ensuing discussions demonstrate why only the actual experiment, validated by positive and negative controls, determines the best choice. The purpose of the following example is to discuss some of the complexities involved in selecting a labeling strategy. Suppose that you have the option of screening a target with a probe generated from the following templates: a 30 base oligo (30mer), a doublestranded 800bp DNA fragment, or a double-stranded 2kb fragment. 30-mer The 30-mer could be radioactively labeled at the 5¢ end via T4 polynucleotide kinase (PNK) or at the 3¢ end via Terminal deoxynucleotidyl transferase (TdT). PNK attaches a single molecule of radioactive phosphate whereas TdT reactions are usually designed to add 10 or less nucleotides. PNK does not produce the hottest probe, since only one radioactive label is attached, but the replacement of unlabeled phosphorous by 32P will not alter probe structure or specificity. TdT can produce a probe containing more radioactive label, but this gain in signal strength could be offset by altered specificity caused by the addition of multiple nucleotides. A 30mer containing multiple nonradioactive labels could also be manufactured on a DNA synthesizer, but the presence of too many modified bases may alter the probe’s hybridization characteristics (Kolocheva et al., 1996). 800bp fragment The double-stranded 800bp fragment could also be end labeled, but labeling efficiency will vary depending on the presence of blunt, recessed, or overhanging termini. Since the complementary strands of the 800bp fragment can reanneal after labeling, a reduced amount of probe might be available to bind to the target. Unlabeled template will also compete with labeled probes for target binding reducing signal output further. However, probe synthesis from templates covalently attached to solid supports might overcome this drawback (Andreadis and Chrisey, 2000). Random hexamer- or nanomer-primed and nick translation labeling of the 800bp fragment will generate hotter probes than end labeling. However, they will be heterogeneous in size and specificity, since they originate from random location in the tem- 404 Herzer and Englert
plate Probe size can range from about 20 nucleotides to the full- length template and longer(Moran et aL., 1996; Islas, Fairley, and Morgan, 1998). However, the bulk of the probe in most random prime labeling reactions is between 200 and 500nt If the entire 800 bp fragment is complementary to the intended target, a diverse probe population may not be detrimental. If only half the template contains sequence complementary to the target then sensitivity could be reduced. Any attempt to compensate by increasing probe concentration could result in higher back grounds. However, the major concern would be for the stringency of hybridization. Different wash conditions could be required to restore the stringency obtained with a probe sequence completely complementary to the target. 2kb DNA fragment The incorporation of radioactive label into probes generated by andom-primer labeling does not vary significantly between tem plates ranging from 300bp to 2kb, although the average size of probes generated from larger templates is greater(Ambion, Inc unpublished data). Generating a probe from a larger template could be advantageous if it contains target sequence absent from a smaller template The availability of different radioactive and nonradioactive labels could further complicate the situation, but the message re mains the same Visualize the hybridization event before you go to the lab. Consider the possible structures of your labeled probes and compare them to your target(s). Be prepared to change your labeling and hybridization strategies based on your experiments. duration, and resolution of the signal. One could aloc a Lab What Criteria Could You Consider When Selecting a Label One perspective for selecting a label is to compare the strength label's effect on incorporation into the probe, and the impact of the incorporated label on the hybridization of probe to target. The quantity of label incorporated into a probe can also affect the per formance of some labels and the probe's ability to bind its target Many experienced researchers will choose at least two techniques to empirically determine the best strategy to generate a new probe (if possible) Signal Strength and Resolution Signal strength of radioactive and nonradioactive labels is iversely proportional to signal resolution. Nucleic Acid Hybridization 405
plate. Probe size can range from about 20 nucleotides to the fulllength template and longer (Moran et al., 1996; Islas, Fairley, and Morgan, 1998). However, the bulk of the probe in most random prime labeling reactions is between 200 and 500nt. If the entire 800bp fragment is complementary to the intended target, a diverse probe population may not be detrimental. If only half the template contains sequence complementary to the target, then sensitivity could be reduced. Any attempt to compensate by increasing probe concentration could result in higher backgrounds. However, the major concern would be for the stringency of hybridization. Different wash conditions could be required to restore the stringency obtained with a probe sequence completely complementary to the target. 2kb DNA Fragment The incorporation of radioactive label into probes generated by random-primer labeling does not vary significantly between templates ranging from 300bp to 2kb, although the average size of probes generated from larger templates is greater (Ambion, Inc., unpublished data). Generating a probe from a larger template could be advantageous if it contains target sequence absent from a smaller template. The availability of different radioactive and nonradioactive labels could further complicate the situation, but the message remains the same. Visualize the hybridization event before you go to the lab. Consider the possible structures of your labeled probes and compare them to your target(s). Be prepared to change your labeling and hybridization strategies based on your experiments. What Criteria Could You Consider When Selecting a Label? One perspective for selecting a label is to compare the strength, duration, and resolution of the signal. One could also consider the label’s effect on incorporation into the probe, and the impact of the incorporated label on the hybridization of probe to target.The quantity of label incorporated into a probe can also affect the performance of some labels and the probe’s ability to bind its target. Many experienced researchers will choose at least two techniques to empirically determine the best strategy to generate a new probe (if possible). Signal Strength and Resolution Signal strength of radioactive and nonradioactive labels is inversely proportional to signal resolution. Nucleic Acid Hybridization 405
Radioactive Isotope signal strength diminishes in the order: P>>P>S> 3H. When sensitivity is the primary concern, as when searching for a low-copy gene, P is the preferred isotope Tritium is too weak for most blotting applications, but a nucleic acid probe labeled with multiple tritiated nucleotides can produce a useful, highly resolved signal without fear of radiolytic degradation of the probe H and"S are used for applications such as in situ hybridiza tion(ish) where resolution is more essential than sensitivity. The resolution of P is similar to 3S, but Ausubel et al. ( 1993)cites an improved signal-to-noise ratio when"P is applied in ISH Signal strengths of nonradioactive labels are difficult to compare. It is more practical to assess sensitivity instead of signal strength. The resolution of nonradioactive signals is also more omplicated to quantify because resolution is a function of signal strength at the time of detection and most nonradioactive signals weaken significantly over time. Therefore the length of exposure to film must be considered within any resolution discussion Background fluorescence or luminescence from the hybridi zation membrane has to be considered as well. Near-infrared dyes are superior due to low natural near-infrared occurrence (Middendorf, 1992). Some dyes emit in the far red 2700nm (Cy7, Alexa Fluor 549, allophycocyanin) Older nonradioactive, colorimetric labeling methods suffered from resolution problems because the label diffused within the membrane. Newer substrates, especially some of the precipitating chemifluorescent substrates, alleviate this problem. Viscous components such as glycerol are often added to substrates to limit diffusion effects. Colorimetric substrates and some chemilumi nescent substrates will impair resolution if the reaction proceeds beyond the recommended time or when the signal is too strong Hence background can increase dramatically due to substrate diffusion Mohandas ghandi said that there is more to life than increas- its speed(John-Roger and McWilliams, 1994), and the sam olds true for detection systems. Most nonradioactive systems deliver a signal within minutes or hours, but this speed is useless if the system cant detect a low-copy target Searching for a single copy gene with a P labeled probe might require an exposure of several weeks, but at least the target is ultimately identified 406 Herzer and Englert
Radioactive Isotope signal strength diminishes in the order: 32P > 33P > 35S > 3 H. When sensitivity is the primary concern, as when searching for a low-copy gene, 32P is the preferred isotope. Tritium is too weak for most blotting applications, but a nucleic acid probe labeled with multiple tritiated nucleotides can produce a useful, highly resolved signal without fear of radiolytic degradation of the probe. 3 H and 35S are used for applications such as in situ hybridization (ISH) where resolution is more essential than sensitivity. The resolution of 33P is similar to 35S, but Ausubel et al. (1993) cites an improved signal-to-noise ratio when 33P is applied in ISH. Nonradioactive Signal strengths of nonradioactive labels are difficult to compare. It is more practical to assess sensitivity instead of signal strength. The resolution of nonradioactive signals is also more complicated to quantify because resolution is a function of signal strength at the time of detection, and most nonradioactive signals weaken significantly over time. Therefore the length of exposure to film must be considered within any resolution discussion. Background fluorescence or luminescence from the hybridization membrane has to be considered as well. Near-infrared dyes are superior due to low natural near-infrared occurrence (Middendorf, 1992). Some dyes emit in the far red ≥700nm (Cy7, Alexa Fluor 549, allophycocyanin). Older nonradioactive, colorimetric labeling methods suffered from resolution problems because the label diffused within the membrane. Newer substrates, especially some of the precipitating chemifluorescent substrates, alleviate this problem. Viscous components such as glycerol are often added to substrates to limit diffusion effects. Colorimetric substrates and some chemiluminescent substrates will impair resolution if the reaction proceeds beyond the recommended time or when the signal is too strong. Hence background can increase dramatically due to substrate diffusion. Detection Speed Mohandas Ghandi said that there is more to life than increasing its speed (John-Roger and McWilliams, 1994), and the same holds true for detection systems. Most nonradioactive systems deliver a signal within minutes or hours, but this speed is useless if the system can’t detect a low-copy target. Searching for a singlecopy gene with a 32P labeled probe might require an exposure of several weeks, but at least the target is ultimately identified. 406 Herzer and Englert
Will you need to detect a signal from your blot several times over a period of hours or days? Are you pursuing a low-copy target that requires an exposure time of days or weeks? Would ou prefer a short-lived signal t id stripping a blot pri re-probing? Some nonradioactive detection systems allow for the quick inactivation of the enzyme that generates the signal, eliminating the stripping step prior to re-probing(Peterhaensel, Obermaier, and Rueger, 1998). The effects and implications of stripping are discussed in greater detail later in the chapter The practical lifetime of common radiolabels is several days to weeks, and is dependent on the label, the ligand, and its environ ment, as discussed in Chapter 6, Working Safely with Radioactive Materials. Some nonradioactive systems based on alkaline phos phatase can generate signals lasting 10 days without marked reduction(personal observation). Some chemifluorescent systems generate a fluorescent precipitate capable of producing a cumu lative signal, much like isotopes. The functional lifetime of fuorescent labels will vith the chemical nature of the fluorescent tag and the methodology of the application. For example, signal duration of a fluorescent tag could be defined by the number of times the chromophore can be excited to produce a fluorescent emission. Some tags can only be excited/scanned once or a few times, while others are much more stable. Consult the manufacturer of the labeled product for this type of stability information. In systems where an enyzme cat alyzes the production of a reagent required for fluorescence, the enzyme's half-life and sufficient presence of fresh substrate can limit the duration of the sign Will the Label Be efficiently Incorporated into the Probe? The effects of label size, location, and linkage method on the corporation of nucleotides into DNA or RNA are enzyme dependent and can be difficult to predict. Small side chains can inhibit nucleic acid synthesis(Racine, Zhu, and Mamet-Bratley, 993), while larger groups such as biotin might have little or no effect(Duplaa et al., 1993; Richard et al., 1994).In general, nucleotides labeled with isotopes of atoms normally present in nucleotides(P, P,H, C)will be incorporated by DNA and RNa polymerases more efficiently than nucleotides labeled with isotopes of nonnative atoms. Commercial polyermases are frequently engineered to overcome such incorporation bias Nucleic Acid Hybridization 407
Signal Duration Will you need to detect a signal from your blot several times over a period of hours or days? Are you pursuing a low-copy target that requires an exposure time of days or weeks? Would you prefer a short-lived signal to avoid stripping a blot prior to re-probing? Some nonradioactive detection systems allow for the quick inactivation of the enzyme that generates the signal, eliminating the stripping step prior to re-probing (Peterhaensel, Obermaier, and Rueger, 1998). The effects and implications of stripping are discussed in greater detail later in the chapter. The practical lifetime of common radiolabels is several days to weeks, and is dependent on the label, the ligand, and its environment, as discussed in Chapter 6,“Working Safely with Radioactive Materials.” Some nonradioactive systems based on alkaline phosphatase can generate signals lasting 10 days without marked reduction (personal observation). Some chemifluorescent systems generate a fluorescent precipitate capable of producing a cumulative signal, much like isotopes. The functional lifetime of fluorescent labels will vary with the chemical nature of the fluorescent tag and the methodology of the application. For example, signal duration of a fluorescent tag could be defined by the number of times the chromophore can be excited to produce a fluorescent emission. Some tags can only be excited/scanned once or a few times, while others are much more stable. Consult the manufacturer of the labeled product for this type of stability information. In systems where an enyzme catalyzes the production of a reagent required for fluorescence, the enzyme’s half-life and sufficient presence of fresh substrate can limit the duration of the signal. Will the Label Be Efficiently Incorporated into the Probe? The effects of label size, location, and linkage method on the incorporation of nucleotides into DNA or RNA are enzymedependent and can be difficult to predict. Small side chains can inhibit nucleic acid synthesis (Racine, Zhu, and Mamet-Bratley, 1993), while larger groups such as biotin might have little or no effect (Duplaa et al., 1993; Richard et al., 1994). In general, nucleotides labeled with isotopes of atoms normally present in nucleotides (32P, 33P, 3 H, 14C) will be incorporated by DNA and RNA polymerases more efficiently than nucleotides labeled with isotopes of nonnative atoms. Commercial polyermases are frequently engineered to overcome such incorporation bias. Nucleic Acid Hybridization 407
Some applications will exploit impaired incorporation of labeled nucleotides(Alexandrova et aL., 1991). Fluorescein attached to position 5 of cytosine in dCTP inhibits terminal transferase and causes addition of only one to two labeled dNTPs at the 3 end of DNA. Fluorescein-or biotin-riboUTP have been similarly applied Igloi and Schiefermayr, 1993) If no specific data exist regarding incorporation efficiency of your labeled nucleotide-labeling enzyme combination, contact the manufacturer of both products. They will likely be able to provide you with a starting point from which you can optimize your labeling reactions. Will the Label Interfere with the Probe's ability to bind to the Target? Hybridization efficiency can be altered by a labels chemical structure,its location within the probe, the linker that connects he label to the ligand, and the quantity of label within the probe. Isotopes of elements present in nucleic acids in vivo might not directly alter the probe's structure, but as described below(and Chapter 6), a label's radioactive emissions can fragment a probe The importance of label location is demonstrated by comparing the hybridization efficiencies of probes labeled with Cy5Morigi nating at either C5 or the primary amine attached to C4 of dCTP. Probes labeled throughout their sequence with the C5-linked label hybridize efficiently, and are commonly applied in micro- array applications(Lee et aL., 2000). Probes containing the label attached to the amine group at C4 do not hybridize efficiently to their targets. The purpose of the C4 amine-label is the addition of a single molecule of Cys dctP to the 3 end of a sequencing primer(Ansorge et al., 1992). A molecular model to accurately predict the effects of labels on analog conformation, hydrogen bonding, stacking interactions, and hybrid helical geometry has been proposed(Yuriev, Scott, and Hanna, 1999) The C5 position is also preferred for dUTP(Petrie, Oshevskil, Kumarev, and grachev, 1989). C5 is such an attractive labeling site because it does not contribute to base-pairing by hydrogen bonding, and at least some linkers seem to allow posi tioning of the label attached in position 5 so that helix formation is not impaired. But bulky tags linked to pyrimidines in the C5 position still interfere to some degree with hybridization because of steric hindrance. Other sites on purines and pyrimidines have been used as tag or label attachment points. However, they have only been shown to work as primers, not for internal labeling strategies(Srivastava, Raza, and Misra, 1994) Herzer and Englert
Some applications will exploit impaired incorporation of labeled nucleotides (Alexandrova et al., 1991). Fluorescein attached to position 5 of cytosine in dCTP inhibits terminal transferase and causes addition of only one to two labeled dNTPs at the 3¢ end of DNA. Fluorescein- or biotin-riboUTP have been similarly applied (Igloi and Schiefermayr, 1993). If no specific data exist regarding incorporation efficiency of your labeled nucleotide–labeling enzyme combination, contact the manufacturer of both products. They will likely be able to provide you with a starting point from which you can optimize your labeling reactions. Will the Label Interfere with the Probe’s Ability to Bind to the Target? Hybridization efficiency can be altered by a label’s chemical structure, its location within the probe, the linker that connects the label to the ligand, and the quantity of label within the probe. Isotopes of elements present in nucleic acids in vivo might not directly alter the probe’s structure, but as described below (and in Chapter 6), a label’s radioactive emissions can fragment a probe. The importance of label location is demonstrated by comparing the hybridization efficiencies of probes labeled with Cy5TM originating at either C5 or the primary amine attached to C4 of dCTP. Probes labeled throughout their sequence with the C5-linked label hybridize efficiently, and are commonly applied in microarray applications (Lee et al., 2000). Probes containing the label attached to the amine group at C4 do not hybridize efficiently to their targets. The purpose of the C4 amine-label is the addition of a single molecule of Cy5 dCTP to the 3¢ end of a sequencing primer (Ansorge et al., 1992). A molecular model to accurately predict the effects of labels on analog conformation, hydrogen bonding, stacking interactions, and hybrid helical geometry has been proposed (Yuriev, Scott, and Hanna, 1999). The C5 position is also preferred for dUTP (Petrie, 1991; Oshevskii, Kumarev, and Grachev, 1989). C5 is such an attractive labeling site because it does not contribute to base-pairing by hydrogen bonding, and at least some linkers seem to allow positioning of the label attached in position 5 so that helix formation is not impaired. But bulky tags linked to pyrimidines in the C5 position still interfere to some degree with hybridization because of steric hindrance. Other sites on purines and pyrimidines have been used as tag or label attachment points. However, they have only been shown to work as primers, not for internal labeling strategies (Srivastava, Raza, and Misra, 1994). 408 Herzer and Englert