Abstract
Oligonucleotides that carry a detectable label can be used to probe for mRNA targets in in situ hybridization experiments. Oligonucleotide probes (OPs) have several advantages over cDNA probes and riboprobes. These include the easy synthesis of large quantities of probe, superior penetration of probe into cells and tissues, and the ability to design gene- or allele-specific probes. One significant disadvantage of OPs is poor sensitivity, in part due to the constraints of adding and subsequently detecting multiple labels per oligonucleotide. In this study, we compared OPs labeled with multiple detectable haptens (such as biotin, digoxigenin, or fluorescein) to those directly conjugated with horseradish peroxidase (HRP). We used branching phosphoramidites to add from two to 64 haptens per OP and show that in cells, 16-32 haptens per OP give the best detection sensitivity for mRNA targets. OPs were also made by directly conjugating the same oligonucleotide sequences to HRP. In general, the HRP-conjugated OPs were more sensitive than the multihapten versions of the same sequence. Both probe designs work well both on cells and on formaldehyde-fixed, paraffin-embedded tissues. We also show that a cocktail of OPs further increases sensitivity and that OPs can be designed to detect specific members of a gene family. This work demonstrates that multihapten-labeled and HRP-conjugated OPs are sensitive and specific and can make superior in situ hybridization probes for both research and diagnostic applications.
Keywords
I
Most ISH studies have been done with long (>100 nt) probes derived from cloned DNA fragments (Panoskaltsis-Mortari and Bucy 1995; Speel et al. 1995; Nath and Johnson 1997). Single-stranded cDNAs or riboprobes can be synthesized and labeled by incorporating a nucleotide containing an isotope or a detectable hapten (i.e., biotin, fluorescein, digoxigenin, DNP). The complexity of long probes increases their sensitivity because many haptens can be incorporated into each probe molecule. However, there are disadvantages to long probes. Often the desired sequence is known but the cloned fragment may not be available. Long probes are not always sequence-specific for the target and so might cross-hybridize to similar sequences within a cell, an outcome that cannot always be predicted before a study is initiated. In addition, probes synthesized by enzymes are not identical preparations, possibly resulting in inconsistent ISH outcomes.
Several studies report that adding multiple detectable haptens to an OP increases its sensitivity in a hybridization assay (Taneja and Singer 1990; Harper et al. 1997; Hougaard et al. 1997; Nath and Johnson, 1997). There are several haptens and functional groups that can be attached to an oligonucleotide via either a polymerase or a chemical condensation. For example, several labels (e.g., dig-11-dUTP) can be added to the 3' end of an oligonucleotide by “tailing” with terminal deoxynucleotide transferase (TdT). Alternatively, there are chemical labeling methods by which haptens can be attached to an oligonucleotide (Forster et al. 1985). Hapten-containing phosphoramidites can be added at the 5' or 3' end of an OP, but labels added to internal positions must be conjugated postsynthetically. Unfortunately, this additional condensation chemistry is tedious and results in low yields of purified final product. Oligonucleotide probes called “branched” DNA (bDNA) have been made by iterative addition of a branching phosphoramidite followed by coupling with a hapten (Horn et al. 1989). These make sensitive probes in liquid-phase hybridization assays but their use in ISH has not been reported. In addition, reporter enzymes such as horseradish peroxidase (HRP or POD) and alkaline phosphatase (AP) can be conjugated to oligonucleotides with a bifunctional linker molecule (Jablonski et al. 1986; Ruth 1994).
We are interested in making highly sensitive ISH probes for mRNA detection in fixed cell and tissue preparations. We also want to manufacture large, reproducible quantities of probe inexpensively. Here we describe OPs that have a high density of biotin, fluorescein, or digoxigenin attached at the 5' end via multiple layers of a branching phosphoramidite. For ISH, we determined that the optimal number of haptens per OP is 16-32 for probes to mRNA targets. OPs directly conjugated with HRP were also synthesized and tested. The expression of several genes was successfully detected in both cells and tissues with multihapten and HRP OPs. For both probe types, we show that a cocktail composed of several different OP sequences can be used to increase sensitivity. Control experiments confirm that the OPs directed against specific mRNA targets hybridize to the intended sequence.
Materials and Methods
Oligonucleotide Synthesis Reagents
Oligonucleotide synthesis reagents were purchased from PE-Applied Biosystems (Foster City, CA) and Glen Research (Sterling, VA). The SYM branching phosphoramidite was obtained from BioGenex (San Ramon, CA). Biotin phosphoramidites were purchased from Glen Research. Fluorescein phosphoramidites were purchased from Glen Research and BioGenex. Amino-modifier phosphoramidites were from Glen Research. Digoxigenin-NHS-ester was from Roche Molecular Biochemicals (Indianapolis, IN).
Oligonucleotide Synthesis
The sequences of all oligonucleotides used in this study are presented in Figure 1. Oligonucleotides were synthesized on a PE-Applied Biosystems 394 DNA Synthesizer by standard DNA phosphoramidite chemistry. For the multihapten OPs, one to siμlevels of branches were added at the 5' end using the SYM branching phosphoramidite (Figures 2A and 2B). Each level of branching doubles the number of 5' sites available for further branching. The cycle parameters were identical for the DNA portion of the chain. Coupling times were extended to 3 min during the addition of branches. Double coupling was used after addition of two cycles of branching. Triple coupling was used for oligonucleotides with siμlevels of branches (64 hapten labels). After branching, the multiple 5' ends were labeled with either biotin or fluorescein phosphoramidite.
Digoxigenin-labeled probes were synthesized by making 5'-branched oligonucleotides as described above and then adding the Glen Research amino-modifier to the multiple 5' ends. The digoxigenin was conjugated onto the probe using an NHS-ester reaction. The branched oligonucleotide was dissolved in 45 μl of water and 5 μl of 10 × TBE, pH 8.4, buffer. Digoxigenin-NHS-ester (1-1.25 mg) was dissolved in 5 μl of dimethylformamide and the two solutions were mixed and incubated overnight at room temperature (RT).
The synthesized OPs were dried in a speed-vac and purified by HPLC (Beckman System Gold with a diode array detector; Beckman, Fullerton, CA). Biotin-labeled oligonucleotides were purified on a Vydac C4 semi-prep HPLC column with a 12-15% gradient of acetonitrile over 15 min. Multifluorescein OPs were purified over the same HPLC column with a 12-17% acetonitrile gradient over 5 min. Multidigoxigenin OPs were purified with a Brownlee Aquapore ODS column and a 0-40% acetonitrile gradient over 30 min.
The γ-1 antisense oligonucleotide (Figure 1) was tailed at the 3' end with digoxigenin-dUTP and dATP using the Dig-Tailing Kit (Roche Molecular Biochemicals) according to the manufacturer's instructions. The tailed OP was purified by ethanol precipitation. The OP was sized by agarose gel electrophoresis; the tail length was ~50-100 nt. The agarose gel was blotted to a positively charged membrane (Roche Molecular Biochemicals) and digoxigenin incorporation was confirmed by colorimetric detection (not shown).
Compilation of the OP sequences used. AS is antisense orientation and S is sense orientation. The GenBank accession number for each gene is γ-globin, X55656, β-globin, V00500, γ-globin, V00508, transferrin receptor, M11507, βHCG, N2711, hPLH, V00573.
Tissue Procurement
Human fetal and adult liver samples and chorionic villous tissue were obtained from nonprofit research tissue banks, either Anatomic Gift Foundation (Woodbine, GA) or Advanced Bioscience Resources (Alameda, CA). Bone marrow was obtained from Poietic Technologies (Gaithersburg, MD). The samples were collected in accordance with the guidelines of the Dept. of Health and Human Services (regulations for the protection of human subjects), the National Organ Transplant Act, and the Uniform Anatomical Gift Acts, as well as other federal and state regulations.
Schematic of the multihapten and HRP-conjugated OP design. (
Cell and Tissue Preparation
A suspension of almost pure fetal nucleated red blood cells (nRBCs) was prepared by mechanically dissociating a 1015-week fetal liver with a scalpel. The liver debris was removed by filtering and the nRBCs were washed and resuspended in PBE (PBS + 0.5% BSA + 5 mM EDTA). A total of 1 × 105 cells were deposited on a positively charged slide by cytocentrifugation (Shandon Lipshaw; Pittsburgh, PA). The cells were dried briefly and fixed for 20 min in 4% formaldehyde/5% acetic acid/1 × PBS (Raap et al. 1994). The slides were washed twice in PBS for 10 min and once in sterile water. After the final wash, the slides were dried and stored at −20C until use.
Fresh chorionic villous tissue (first trimester) was dissected into small pieces and the pieces were placed in cold 4% paraformaldehyde/2.5 mM EGTA. After 12-16-hr fixation, the tissue pieces were embedded in paraffin and 5-μm sections were prepared on Probe-On-Plus (Fisher Scientific; Pittsburgh, PA) slides. The tissue sections were prepared by Paragon Biotech (Baltimore, MD).
mRNA In Situ Hybridization and Detection in Cells
The nRBCs were permeabilized with 0.1-1 μg/ml proteinase K (PK; Roche Molecular Biochemicals) in TBST (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.2% Tween-20) for 10 min at 20C. The PK was inactivated by immersion in PBS + 0.2% glycine for 5 min. In some experiments, the cells were treated for 30 min with the Universal Blocker (Kirkegaard & Perry; Gaithersburg, MD) to inhibit endogenous peroxidase activity. The slides were washed in PBS for 10 min and water for 10 min and then air-dried.
Multihapten OPs were diluted to 0.01 to 0.1 ng/μl (0.001-0.01 pmol/μl) in hybridization buffer (25% deionized formamide, 3 × SSPE, 10% dextran sulfate, 1 × Denhardt's reagent, 0.2% CHAPs detergent, 50 μg/ml ssDNA, and 50 μg/ml tRNA). Probe mix (15 μl) was applied to each cell spot and the spot was covered with a 12-mm circular glass coverslip. In some experiments, the slides were placed on an 80C hotplate for 5 min to denature the nucleic acids. The slides were incubated for 3 hr to overnight at 42C in a humid chamber. The slides were washed at 37C either twice for 10 min each in 0.2 × SSC or twice for 10 min each in 30% formamide/2 × SSC. Nonspecific protein binding sites were blocked by incubating the slides in 1 × Fluorescent Dig Enhancer Blocking Buffer (Roche Molecular Biochemicals) for 30 min. Depending on the hapten being detected, the slides were treated for 3 hr to overnight with a 1:500 dilution in TBST of SA-POD, anti-digoxigenin POD, or anti-fluorescein POD (all detection conjugates were from Roche Molecular Biochemicals). Finally, the slides were washed three times in TBST for 5 min each to wash away the unbound conjugate. The peroxidase enzyme activity was detected using either the TSA-DIRECT (GreenFISH) or the TSA-DIRECT (Cyanine-3 FISH) kits (NEN; Boston, MA) according to the manufacturer's protocol. All TSA reactions were done for 30 min at RT. Vectashield + DAPI (Vector Labs; Burlingame, CA) was applied to the cells and the signals were visualized with either the Quips Imaging System and Software (Vysis; Downers Grove, IL) or with the Open-labs software (Improvision; Coventry, UK).
HRP OPs were prepared by Synthetic Genetics (San Diego, CA) and diluted to 1 μM concentration of oligonucleotide. The concentrated stock was diluted to 0.01 μM in an aqueous hybridization buffer (2 × SSC, 10 mM Tris-HCl, pH 7.5, 10% dextran sulfate, 2 × Denhardt's reagent, 0.5% CHAPs detergent, 50 μg/ml ssDNA, and 50 μg/ml tRNA). When a cocktail of probes was used, a maximum of 1 μl of probe mixture was added per 20 μl of hybridization buffer. Probe mix (15 μl) was applied to the cell spot and covered with a 12-mm circular glass coverslip. The slides were hybridized for 3 hr to overnight. Excess probe was washed off with two 10-min incubations in 0.2 × SSC at 37C. After two 5-min washes in TBST, the HRP OP was detected with the TSA-DIRECT (GreenFISH) kit. Alternatively, in some experiments TSA-INDIRECT (biotin) (NEN) deposition was followed by incubation with 10 μg/ml SA-Alexa 488 (Molecular Probes; Eugene, OR). The signals were captured by digital imaging as described above.
For the double detection of γ- and ∊-globin mRNAs, 10-week fetal liver nRBCs were hybridized with Ep-104 labeled with HRP and a cocktail of five antisense OPs to γ-globin labeled with biotin (Figure 1). The control cells were hybridized with sense OPs labeled with HRP and biotin. The hybridization was done in the buffer described above for the HRP OPs. After an overnight incubation at 42C, the slides were washed twice in 0.2 × SSC for 10 min each at 37C. The HRP OP was detected first with TSA-DIRECT (Green-FISH). After detection, the slides were washed with PBS and the residual HRP activity was eliminated by treating the slides with Universal Blocker for 10 min. The slides were incubated in 1 × Fluorescent Dig Enhancer Blocking Buffer for 30 min. Biotin was detected by incubating the slides in SA-POD as described above. The slides were incubated in TSA-DIRECT (cyanine-3 FISH) for 30 min. After washing, the slides were mounted with Vectashield + DAPI.
Tissue Preparation and mRNA In Situ Hybridization
Five-μm sections of chorionic villous tissue were treated with 10 μg/ml proteinase K for 12 min at 20C. The slides were washed for 5 min each in PBS + 0.2% glycine, PBS, and water. The tissues were air-dried. For the multihapten OPs, 20 μl of hybridization buffer containing 0.01 ng/μl OP was applied to the tissue. The slides were hybridized overnight at 42C in a humid chamber. The slides were washed in 0.2 × SSC twice for 10 min at 37C and then placed in 1 × Fluorescent Dig Enhancer Blocking Buffer for 30 min. The slides were incubated in a 1:500 dilution in TBST of either SA-AP (Roche Molecular Biochemicals) or anti-digoxigenin AP (Roche Molecular Biochemicals) for 2 hr and then washed in TBST for 5 min. After two additional washes in TBST, pH 9.5 (50 mM Tris-HCl, pH 9.5, 150 mM NaCl, 10 mM MgCl2, 0.2% Tween-20), the slides were incubated for 2 hr in a NBT/BCIP substrate solution (Roche Molecular Biochemicals). The AP reaction was stopped by washing the slides in water for 5 min. The tissue was counterstained in Nuclear Fast Red (Vector Labs) for five min, dehydrated in ethanol, and mounted in Vectamount (Vector Labs).
A cocktail of five hPLH HRP OPs (Figure 1) was made by adding 0.2 μl of each probe to 20 μl of the HRP hybridization buffer described above for cells. Twenty μl of the probe was added to the tissues and overlaid with a coverslip. The hybridization, wash, and detection steps were done as described above for the HRP OPs in cells.
Northern Blots
Total RNA was isolated from 16-18-week fetal liver nRBCs using a modified guanidinium isothiocyanate procedure (Qiagen; Hilden, Germany). Total RNA from human adult peripheral blood was purchased from Biochain Institute (San Leandro, CA). Twenty μg of each RNA sample was separated in a 1.1% denaturing agarose gel containing 2% formaldehyde; the running buffer was 1 × MOPS buffer (20 mM MOPS, 5 mM sodium acetate, and 1 mM EDTA). The gels were blotted in a downward-transfer blotting system onto Nytran nylon membranes (Schleicher & Schuell; Keene, NH) in 20 × SSC. After transfer, the membrane was rinsed in 2 × SSC and the RNA was crosslinked by baking at 80C for 1.5 hr.
The RNA blots were hybridized with multibiotin OPs using the same hybridization and wash conditions as described above for the ISH slides. After washing, the blots were incubated in 1 × Fluorescent Dig Enhancer Blocking Buffer for 30 min at RT. The biotin was detected with a 1:10,000 dilution of SA-AP diluted in TBST for 30 min at RT. The blots were immersed in Attophos substrate (Amersham; Chicago, IL) and the resulting chemifluorescence was detected with a Storm PhosphorImager (Molecular Dynamics; Sunnyvale, CA).
Results
Design Strategy for Multihapten OPs
Our goal was to make large quantities of pure hybridization probe in a cost-efficient manner. We also wanted sensitive probes that could detect low copy number mRNA sequences in cells and tissues. OPs were designed in which multiple labels were added to the 5' end of an oligonucleotide by the sequential addition of a branching phosphoramidite. The theoretical number of haptens that can be added is a log function of the number of branches introduced (i.e., 2n = x, where n equals the number of branches and × equals the number of haptens) (see Figures 2A and 2B). Multihapten OPs were made containing biotin, fluorescein, or digoxigenin as the label. We prepared a digoxigenin-tailed OP for comparison to the multidigoxigenin probes.
We reasoned that steric hindrance among multiple detection conjugates might preclude the recognition of closely spaced haptens on each OP. We therefore made OPs in which linker arms of various lengths (up to 18 atoms) and hydrophobicities were added between each branch to increase the molecular distance between each hapten. We hoped that the increased distance among haptens might reduce the adverse consequences of steric hindrance among several closely spaced detection conjugates. None of these additions increased the signal strength of the γ-1-globin OP tested on fetal nRBCs (not shown).
Detection Sensitivity of Multihapten Probes
For mRNA detection, we chose first to target the γ-, β-, and ∊-hemoglobin sequences expressed in nRBCs. γ-globin is found mainly in definitive nRBCs that develop in the fetal liver. β-globin is found both in definitive nRBCs of the developing liver and in nRBCs in adult bone marrow. ∊-globin is expressed predominantly in fetal primitive nRBCs derived from the blood islands in the yolk sac. The OPs were hybridized to formaldehyde-fixed, proteinase K-permeabilized cells as described in Materials and Methods. The optimal probe concentration was determined to be 0.1 ng/μl for the digoxigenin-tailed, multibiotin, and multidigoxigenin OPs. The optimal concentration for the multifluorescein OPs was 0.01 ng/μl. There was unacceptable background signal on both cells and the glass slide at higher concentrations of probe (not shown). A pre-denaturation of 5 min at 80C was done in some experiments, but because this treatment caused an increase in background and no increase in hybridization signal it was omitted in later experiments. The optimal hybridization time was about 3 hr, although overnight hybridizations gave excellent results. In all cases, the label was detected with the appropriate horseradish peroxidase (HRP)-labeled detection conjugate, and TSA-DIRECT (GreenFISH;fluorescein) was added to visualize the bound HRP. TSA-DIRECT (cyanine-3 FISH) also worked well for detection. The globin mRNAs are very abundant in nRBCs and a strong hybridization signal was usually obtained with a single 30-mer OP that contained multiple labels.
Hybridizations to fetal nRBCs were done with multibiotin γ-globin OPs containing from 2-64 biotins attached at the 5' end (Figure 2B). In general, there was weak to no signal for antisense OPs containing two to four biotins, probably because not enough detectable hapten was available to visualize. The multi-8 biotin probe was always detected, although the signal was higher for the multi-16 or multi-32 biotin probes of the same sequence. A strong signal was typically seen for probes with 16-32 biotin labels, as shown in Figures 3A and 3B. There was no consistent difference in signal strength between the multi-16 and multi-32 probes. In different trials with several cell preparations, one or the other gave a slightly more intense signal. The multi-64 biotin probe consistently gave a weak signal (Figure 3C), generally similar to the signal intensity obtained with the multi-4 or multi-8 versions. We did not try probes that contain 128 or more biotins. Sense strand OPs containing from two to 64 biotins did not show a detectable signal, indicating that the increasing biotin density was not responsible for the signals observed with the antisense strand probes (Figure 3D and not shown).
Four additional γ-globin OPs targeted to other regions of the mRNA were synthesized with the same labeling strategy (Figures 1 and 2B). Each individual OP gave approximately the same signal intensity (<two-fold different), indicating that the results described above were not idiosyncratic to the model sequence used (not shown). In addition, when several probes were combined into a cocktail and hybridized, the signal strength increased and was roughly proportional to the number of probes used. Compare the signal intensities for one or a cocktail of five OPs as shown in Figures 3A and 3E.
ISH results with the multifluorescein, multidigoxigenin, and multibiotin OPs. TSA-DIRECT (GreenFISH) was used as the substrate for HRP; the hybridization signal appears as green speckles in the cytoplasm. DAPI was used to counterstain the nuclei blue. (
Using the BioGenex SYM, the γ-1-globin OP was synthesized with multiple fluorescein and digoxigenin haptens. We added either 16 or 32 fluoresceins to the 30-mer sequence and hybridized the resulting OPs to fetal nRBCs. When either the γ-globin multi-16 or multi-32 fluorescein probe was hybridized at the standard concentration of 0.1 ng/μl, a strong nonspecific signal was observed for both the sense and antisense probes (not shown). In addition, the fluorescein-labeled probes stuck to the surface of the glass slide, resulting in severe background “speckling.” When the concentration of the probe was reduced 10-fold to 0.01 ng/μl, most of the nonspecific signal on the cells and glass slide disappeared, as shown in Figure 3F. At the lower concentration for the multifluorescein OPs, the hybridization signal was roughly equivalent to the multibiotin probes.
We were initially concerned that adding many haptens to an oligonucleotide might affect its solubility in water or in a formamide-based hybridization buffer. Therefore, our initial OPs were synthesized with biotin and fluorescein instead of the more hydrophobic digoxigenin. However, we also made multidigoxigenin OPs and found no adverse effects on solubility with 30-mer OPs that contained up to 16 digoxigenin residues. The multidigoxigenin γ-1 OP was compared to the same oligonucleotide sequence which had a ~100 nt tail sequence containing ~4-5 digoxigenin residues. As shown in Figure 3G, we hybridized the multi-8 digoxigenin γ-1-globin antisense OP to fetal nRBCs and detected a strong hybridization signal. The signal intensity was slightly less than that observed with the digoxigenin-tailed probe shown in Figure 3H.
We made oligonucleotide probes to different mRNA targets to further assess both probe specificity and sensitivity. We made a cocktail of three different multi-16 biotin antisense oligonucleotides that targeted β-globin mRNA, and hybridized these to whole adult bone marrow, fetal nRBCs and peripheral blood leukocytes. Whole bone marrow contains 10-30% nRBCs, each of which expresses β-globin mRNA. The remaining ~80% of cells that are leukocytes or their precursors that do not contain β-globin mRNA. The β-globin sense probe is negative on all cell types (Figure 3I; and results not shown). As shown in Figure 3J, the β-globin OP cocktail detected the nRBC subset of bone marrow cells, suggesting that the probes were specific for β-globin mRNA. During definitive erythropoiesis, there is a switch from γ-globin to β-globin expression that occurs during weeks 10-20 of gestation. As shown in Figure 3K, fetal nRBCs contain some β-globin in addition to the much more abundant γ-globin mRNA. As expected, the leukocytes in peripheral blood did not contain detectable levels of β-globin mRNA (Figure 3L).
HRP-labeled OPs
Oligonucleotides directly conjugated to a reporter enzyme also make sensitive hybridization probes (Farquharson et al. 1992; Schmidt et al. 1997; van de Corput et al. 1998a,b). Generally, a bifunctional linker is used to tether the 5' end of an oligonucleotide to reactive amines on either HRP or AP. We made a series of OPs with HRP conjugated to the 5' end of the oligonucleotide (Figure 2C). Because HRP enzyme activity is decreased after prolonged incubation in organic solvents (van Gijlswijk et al. 1996), we used hybridization and wash buffers that did not contain formamide (Farquharson et al. 1992). The HRP probes are easier to use than the multihapten OPs because a detection conjugate is not necessary and hence an intermediate step is eliminated.
OPs to γ-globin, β-globin, ∊-globin, and transferrin receptor mRNAs (Figure 1) were conjugated to HRP and hybridized to fetal liver blood. As shown in Figure 4A, a single antisense OP to γ-globin shows a strong signal in fetal nRBCs; the sense probe is negative (not shown). When an additional antisense OP to γ-globin was included in the hybridization, the signal intensity was about doubled (Figure 4B). We also created a cocktail of three antisense OPs to β-globin and hybridized this to bone marrow. We found a few cells strongly positive; the signal intensity of three OPs was greater than that for a single OP (not shown). In general, the HRP-conjugated OPs were about three- to fivefold more sensitive than the multihapten versions of the same sequence. In addition, the nonspecific background signal was consistently lower than that typically seen for the multihapten OPs.
Not all mRNAs are as abundant as the globin sequences expressed in blood precursor cells. For some studies, the detection of low to moderately abundant mRNAs is necessary. We found that single OPs were not sensitive enough to detect the transferrin receptor mRNA in fetal nRBCs. Whereas the transferrin receptor protein is abundant on the surface of blood cells, the mRNA is moderately abundant, similar to some housekeeping genes such as actin and GAPDH (not shown). We made a cocktail of 15 antisense OPs across the length of the transferrin receptor mRNA and hybridized the mixture to fetal blood cells and to bone marrow cells. A cocktail of eight sense probes gave no hybridization signal, as shown in Figure 4C. In Figure 4D, the cocktail of 15 antisense OPs to transferrin receptor produced a strong signal in the cytoplasm of fetal nRBCs. When the same probe cocktail was hybridized to adult whole bone marrow, a subset of cells was positive and these were similar in number and in morphology to nRBCs (Figure 4E). No signal was observed in peripheral blood leukocytes (not shown). This result indicates that the signal strength of individual OPs is additive and that the cocktails make sufficiently sensitive probes to detect moderately abundant mRNAs.
ISH results for several of the OPs on cells and tissues. (
In addition to the cytoplasmic signal, one or two discrete nuclear signals were detected with the transferrin receptor probe cocktail in most fetal nRBCs and in some bone marrow cells (see Figure 4D). When the probe cocktail was split into seven or eight different sequences, the nuclear signals were still observed (not shown), albeit at a lower intensity. This suggests that the signals were not a consequence of an individual probe hybridizing to a repetitive DNA sequence. When the cells were pretreated with RNase A, both the cytoplasmic and nuclear signals disappeared, as shown in Figure 4F. Because the nuclear signals are RNase-sensitive, they are presumably the sites of transcription and transcript accumulation in the nucleus. A similar observation for a different gene has been described (van de Corput et al. 1998a).
An advantage of OPs over long ssDNA or riboprobes is the ability to make allele- or gene-specific probes (Taneja and Singer 1990; Chumbley et al. 1993). ∊-globin is an embryonic hemoglobin that is synthesized predominantly in the primitive nucleated erythroblasts found in blood islands of the yolk sac. Fetal or γ-globin is predominantly expressed in definitive erythroblasts that develop in the fetal liver. We synthesized an OP to the 3' UTR of ∊-globin mRNA that has 25 and 27 nucleotide changes between β- and γ-globin mRNAs, respectively. Under stringent hybridization and wash conditions the probe is specific for ∊-globin mRNA. We compared the hybridization pattern of the ∊-globin and γ-globin OPs in 10-week fetal blood cells, as shown in Figure 4G. As expected, most of the cells are definitive nRBCs that express γ-globin. However, a small percentage of cells express the mRNA for ∊-globin. The ∊-expressing nRBCs cells are a mixture of primitive nRBCs circulating in the fetal liver and a small percentage of definitive nRBCs. A similar finding using antibodies specific for ∊- and γ-globin proteins was recently reported (Mesker et al. 1998; Luo et al. 1999). The results indicate that OPs can be designed to detect the expression of specific members from a gene family.
mRNA Detection in Tissues
The strategy outlined above was used to synthesize OPs for the detection of mRNAs in formaldehyde-fixed, paraffin-embedded tissues. We synthesized OPs specific for human chorionic gonadotrophin β-chain (βHCG) and human placental lactogen hormone (hPLH) mRNAs, genes that are highly expressed in some cell types of the developing trophoblast (Wide et al. 1988; Hoshina et al. 1984; Sakbun et al. 1990). The first-trimester chorionic villi are composed of an outer layer of multinucleate syncytiotrophoblasts, an inner layer of mononuclear villous cytotrophoblasts, and a stroma that contains fibroblasts, macrophages, and blood cells. As shown in Figure 4H, a single antisense multidigoxigenin OP for βHCG hybridizes specifically to the syncytiotrophoblast layer, in agreement with previously reported data for expression of βHCG (Wide et al. 1988,1989). The same pattern of hybridization is observed for a multi-16 biotin probe (Figure 4I). The sense OPs for each hapten are negative, indicating that the antisense signal is not due to the probe sticking to the tissue (Figure 4J; and results not shown). The results show that the multihapten probes are sensitive hybridization probes in tissues as well as in cells.
We also synthesized HRP-conjugated OPs to the hPLH mRNA and hybridized a cocktail of five different OPs (Figure 1) to chorionic villous tissue. No signal was observed for the sense HRP OP used as a control (Figure 4K). As shown in Figure 4L, a strong signal for hPLH mRNA is observed in the syncytiotrophoblast layer with the antisense probe cocktail, consistent with the results described in the literature (Sakbun et al. 1990). These results show that HRP-conjugated OPs are sensitive hybridization probes on both cells and tissues.
Target Specificity of Oligonucleotide Probes
The interpretation of hybridization signals for mRNA targets is difficult because there is no true sequence control for the antisense probe. The corresponding sense strand is often used, but its sequence is unrelated to the antisense strand; each strand could independently cross-hybridize to other cellular RNAs. An ISH signal only shows whether or not the probe is present in a cell. This can occur either by nonspecific “sticking” of the probe to cellular macromolecules or by hybridization to some sequence (its cognate or not). To show sequence specificity, we ran Northern blots of RNA derived from the same cell populations as were used for the ISH experiments. The Northern blots were done with the same OP sequences, hybridization buffers, and wash conditions used for the ISH experiments.
Northern blots of target mRNAs probed with the multibiotin probes. Total RNA from fetal nRBCs (FBL) and adult peripheral blood (PB) was separated and blotted to a nylon membrane. The sense and antisense multi-2 biotin γ-S and γ-1-globin OPs were hybridized and detected by chemifluorescence. The antisense OP shows a band at ~0.5 kb in size, the correct length for γ-globin mRNA. No other mRNA or rRNA is detected under the hybridization and wash conditions used. As expected, no band for γ-globin mRNA is found in RNA from adult peripheral blood. The γ-S OP shows no signal in either RNA preparation.
We initially attempted Northern blots with the multi-16 or multi-32 γ-1-globin probe and detected weak to no signal, even though labeled riboprobes indicated that there was abundant detectable mRNA present (not shown). As shown in Figure 5, when the multi-2 biotin γ-globin antisense OP was hybridized to a Northern blot of blood cell total RNAs, a strong signal for globin mRNA was detected at the correct length. No other mRNAs or rRNAs were detected, indicating that the probe is specific for its target sequence with the given hybridization and wash conditions. The sense probe did not detect any RNA, as expected. The same hybridization pattern was observed for the γ-1 antisense HRP OP (not shown). In addition, the cocktail of three multi-16 biotin OPs for β-globin also detected an mRNA of the expected size (not shown).
Discussion
In situ hybridization is a powerful method to visualize the presence of target DNA or RNA sequences in cells and tissues. The technique has been widely used in both research and diagnostic applications. Some of the key concerns in all ISH protocols are sensitivity, specificity, signal-to-noise level, preservation of morphology, and ease of probe preparation. We have developed protocols in which multihapten and HRP-conjugated OPs are used to detect moderate and abundant sequences in cells and tissues. The protocols are streamlined for speed and ease of execution. Cell morphology is largely preserved in the protocols described.
Probe Design and Synthesis
All of the OP sequences used in this study were 26- to 30-mers having a 45-60% GC content. Our basic strategy for multihapten OP synthesis was to add as many haptens per sequence as was chemically feasible. We accomplished this through the iterative addition of branching phosphoramidites and the subsequent condensation of detectable haptens such as biotin, fluorescein, or digoxigenin. The multibiotin and multifluorescein probes were synthesized completely “online” and required only minimal purification and handling. Each synthesis of 0.2 μM produced about a milligram of product, enough probe for several million assays. Although we did not do rigorous stability assays, the multibiotin probes gave equivalent ISH results and HPLC tracings even after several months and multiple freeze-thaws.
The HRP OP conjugates were obtained commercially. The HRP enzyme was conjugated to the 5' end of the oligonucleotide through a bifunctional linker and purified by size-exclusion HPLC. All of the HRP OP conjugates were stable for at least several months at 4C.
Sensitivity of Multihapten and HRP-conjugated OPs
The sensitivity of a hybridization probe is generally proportional to its complexity because more labels (isotopic or nonisotopic) can be added per nucleotide subunit. For this reason, the most sensitive probes have been RNA or ssDNAs transcribed or replicated from long template sequences. For hapten labels, theory suggests that probe sensitivity is directly proportional to the number of haptens per probe molecule. However, experience shows that the optimal separation between labels is about 15-20 nt; the space between the haptens reduces the likelihood of steric hindrance among detection complexes.
There are additional parameters that could affect the detection of OPs. The probe must penetrate into the cell or tissue before reaching a target sequence. The efficiency at which this occurs is a function of probe size and shape, the type of fixative used, the extent of fixation (for crosslinking fixatives), the extent of permeabilization (e.g., by proteinase K), and the cell or tissue type. For example, long probes are often reduced in length by hydrolysis to improve penetration into cells or tissues. In addition, even if a probe hybridizes to its target, some hapten labels might be occluded by proximal proteins or other cellular macromolecules and thus be inaccessible to the detection conjugate.
We made a series of oligonucleotide probes to γ-hemoglobin mRNA that had from 2-64 haptens attached to the 5' end via one or more branching phosphoramidites. The probes were hybridized to fetal nRBCs, cells in which γ-globin mRNA is very abundant. If all of the haptens on each probe were equally accessible to the detection complexes, the signal intensity would be proportional to the number of haptens per OP. The signal intensity for the multibiotin OPs did increase with increasing biotin number up to 16 or 32 biotins per OP. A multidigoxigenin OP that contained eight haptens was about as sensitive as the digoxigenin-tailed version of the same sequence, suggesting that multiple digoxigenin residues were being detected on the multihapten OPs. Taken together, these data show that multiple haptens are detected for each OP.
When the number of biotins was increased to 64 per OP, the hybridization signal decreased significantly. The molecular weight of the multi-64 biotin probe is 58 kD, much less than the 85 kD for the SA-HRP detection complex that we used to detect the OP. However, its structure probably is not a random coil and probably has a larger effective radius because of the rigid branch. We believe that the shape of the multi-64 biotin probe precluded either its entry into cells or its accessibility to target mRNAs. Alternatively, the 64 biotins and their linkers might create a “cloud” around the oligonucleotide and prevent its hybridization to the target mRNA.
Interestingly, the signal from a cocktail of two multi-16 biotin OPs is greater than that from a single multi-32 biotin probe. This suggests that signal strength is dependent on factors other than simply the number of biotins added per probe per experiment. Some of the biotins on each OP are probably occluded from detection by one or more mechanisms that include steric hindrance of adjacent detection complexes or proximal cellular macromolecules.
We tested probes synthesized with either biotin, fluorescein, or digoxigenin haptens. No consistent difference in the hybridization signal strength could be observed among the three haptens. As described in the Results, we did observe a significant increase in nonspecific background signal for the multifluorescein probes in both cells and on the glass slide. We remedied this problem by decreasing the probe concentration. The increased background was likely due to “stickiness” of the fluorescein moiety onto the glass and onto the more hydrophobic cellular components.
The HRP OPs to γ-globin gave a strong signal, generally a few-fold more than the multihapten OPs of the identical sequence. In addition, the nonspecific background signal was generally less than that observed for the multi-hapten probes. The molecular weight of the HRP OP conjugate is about 55 kD (assuming a single 30-mer oligonucleotide per enzyme). This complex has more mass than a multi-16 biotin OP (21 kD) but has less mass than detection complexes that range from SA-HRP (~85 kD) to anti-digoxigenin AP (~150 kD). Therefore, if probe or detection complex accessibility to target is a limiting factor in detection sensitivity, the smaller mass of the HRP OPs will enhance their passage through the cell membrane and across internal cellular macromolecules. If these hypotheses are correct, the HRP OPs are more sensitive because each probe that hybridizes to its target is detected and the OP does not need to be recognized by a larger secondary detection complex. Taking all of our results into consideration, we prefer the HRP OPs to the multihapten versions because of the increased signal-to-noise ratio and the decreased number of detection steps. However, for ISH procedures that incorporate high temperatures or organic solvents, the multihapten OPs are the preferred probe type. In addition, OPs that contain haptens are useful because they can be combined with HRP-labeled OPs in a single experiment to allow the simultaneous detection of multiple mRNA targets (for example, see Figure 4G).
Signal intensity is also a function of the number of different probe sequences added to the hybridization mix (Harper et al. 1997; Larsson and Hougaard 1993; Taneja and Singer 1990; van de Corput et al. 1998a,b). We showed that a cocktail of five multibiotin 30-mer OPs gave a significantly stronger signal than any individual probe. In addition, a cocktail of either two γ-globin HRP OPs or 15 transferrin receptor HRP OPs result in a stronger signal than the individual OPs. Although quantification of signal strength is difficult, the signal appeared to increase linearly with the number of probe sequences added to the cocktail. For each of the targets tested, the individual probe sequences were separated by at least 50 nt along the length of the mRNA sequence, limiting or precluding steric hindrance between adjacent probes.
Probe Specificity
The signals from ISH experiments can be difficult to interpret. First, the signal should be in the correct cellular compartment, either the cytoplasm for mRNA or the nucleus for DNA. Second, a signal can be from hybridization to target mRNA, but can also be from cross-hybridization to a related or unrelated (e.g., rRNA) RNA. Sometimes an artifactual signal can occur from sequence-specific probe “stickiness” to cellular macro-molecules. These alternatives are not mutually exclusive, and therefore a signal can be a mixture of specific hybridization and artifact. For mRNA detection, the sense-strand probe does not provide a true control for specific hybridization because its sequence is unrelated to the antisense sequence. At best, it provides a general control for probe type and the cleanliness of the detection procedure. For controls to show probe specificity, we tested several sense-strand sequences, cells known to be negative for the target sequence and multiple probes for the same target sequence. Some of the OPs were applied to Northern blots.
For the mRNA detection experiments, we tested 10 different sense probes that contained from 2-64 haptens or HRP (Figures 1, 3, and 4; and results not shown). When used at the optimal probe concentration, none of the sense-strand probes gave a consistent detectable hybridization signal. In addition, none of the γ-globin or ∊-globin antisense probes gave a hybridization signal on adult leukocytes or adult whole bone marrow, cell types that are negative for the target mRNA. As shown, the cocktail of three β-globin probes detected nRBCs in adult bone marrow but did not hybridize to any leukocyte cells. We tested five γ-globin and three β-globin probes separately, and all gave essentially the same signal intensity, strongly suggesting that the OPs were specifically hybridizing to their intended target. The results described were consistent for both the multihapten and the HRP version of the same OPs. In addition, we showed evidence that γ- and ∊-globin HRP OPs were specific for their cognate mRNA in fetal nRBCs.
We hybridized some of the oligonucleotide probes to Northern blots of total RNA extracted from the same cell populations as were used for the ISH experiments. The purpose of this test was to ascertain whether or not another cellular RNA was present that could hybridize to the probe and generate an artifactual signal. The same hybridization and wash conditions were used for the Northern blot and ISH procedures. The ISH experiments suggested that the most sensitive probes had either 16 or 32 haptens. However, we found that these densely labeled probes gave weak to no signal on total RNAs in which the target sequence was abundant. We redid the Northern blots with the same OPs labeled with two haptens. These probes gave strong target-specific signals, indicating that the ISH signals are due to target mRNA hybridization and not to cross-hybridization. We do not know the reason for the reduced sensitivity of the multi-16 or multi-32 probes on membrane blots, but perhaps the high density of biotins causes a shape or charge distribution that prevents the probe from entering the membrane and hybridizing to the target sequence. We also tested the γ-globin HRP OP and found that only a single mRNA species of the correct length hybridized, again indicating that the probe was specific for its target (not shown). Taken together, the control experiments show that the antisense probes are specific for their target mRNA in the ISH experiments shown.
Conclusions
ISH is a technique that has broad applications in both research and diagnostic procedures in which an mRNA or DNA sequence needs to be localized to a particular cell type. Because there are many different cell and tissue types that are candidates for ISH, as well as many different target sequences, there is not a single best procedure for all systems. In this study, we examined and optimized novel strategies for synthesizing OPs. By using a symmetrical branching phosphoramidite, we attached multiple haptens to the 5' end of oligonucleotides in several configurations. The optimal OP had 16 or 32 biotins or fluoresceins attached via several layers of the SYM branching phosphoramidite obtained from BioGenex. We also tested HRP OP conjugates in which a reporter enzyme is added directly to the 5' end of an oligonucleotide. The HRP OPs were usually more sensitive than the multihapten versions, probably because of better accessibility to target mRNAs and lower background noise. For both the multihapten OPs and HRP OPs, a cocktail of several probe sequences was shown to further enhance sensitivity. The OPs were easily synthesized and purified, were cost-effective, and were sensitive and specific when tested on mRNA targets in either cells or tissues. For research and diagnostic applications that require in situ hybridization, the OP synthesis and detection strategies described will be broadly applicable.
Footnotes
Acknowledgements
We thank Drs Yathi Naidu and Kris Kalra, who provided useful comments on OP design and synthesis. We also thank Dr Bill Harriman for useful comments about manuscript content.