|[Frontiers in Bioscience 1, c4-15, November 1, 1996]|
THE FOUNDATION OF SUCCESSFUL RT IN SITU PCR|
Director, MGN Medical Research Laboratories, Setauket, New York 11733, USA
Received: 07/25/96; Accepted: 10/02/96; On-line 11/01/96
Pathologists use a limited number of fixatives. By far the most common solution used to preserve tissue and cell samples is 10% buffered formalin. Formalin cross-links the amino groups of proteins and nucleic acids to each other, and thus renders degradative enzymes inoperative (18). Other members of this family include paraformaldehyde and glutaraldehyde. One can visualize how this cross linking process creates an intricate, complicated 3-dimensional "roadway" from the nuclear matrix to the cytoskeleton and endoplasmic reticulum of the cell (1,19,20). This structure certainly could form a physical barrier to molecules that otherwise might diffuse readily through the nucleus and cytoplasm (1,19,20). It may also create a "charge" or "ionic" barrier by rigidly fixing the positively and negatively charged side chains of amino acids in space inside the cell (1,20).
Rarely, pathologists add other ingredients to formalin based fixatives. One such additive is picric acid. This acid is thought to improve preservation of nuclear detail, and thus is favored by some as a fixative for lymphoid lesions or testicular biopsies. An example of a picric acid containing fixative is Bouin's solution. A heavy metal, such as mercury or zinc is sometimes added to formalin and is thought by some to improve the cytologic detail of the nucleus. Zenker's solution is an example of this type of preparation.
After fixation the tissue is embedded in paraffin at 65°C for 4 hours. For the non-Pathologist, the paraffin embedded tissue block hardens when brought to room temperature and 4 µM sections can be prepared by the use of an instrument called a microtome. This obligatory heating step has important implications for in situ PCR, as it induces single stranded DNA gaps which lead to primer independent direct incorporation of the reporter nucleotide during in situ PCR (1,21). This DNA "repair" pathway, which can be eliminated by DNase digestion, precludes target specific direct incorporation of the labeled nucleotide for DNA targets, but not for RNA targets (1,21).
Other less commonly used fixatives in the pathology laboratory include ethanol and acetone, which render degradative proteins inactive by denaturation. These are used mostly in immunohistochemistry, to preserve the antigenicity of certain epitopes.
Solution phase PCR and standard in situ hybridization can be done using unfixed material or samples fixed in formalin, ethanol, or acetone. Picric acid or heavy metal containing solutions should not be used as the tissue fixative for standard PCR or in situ hybridization as prolonged (>8 hour) fixation of the samples is associated with extensive degradation of the DNA, and a weak or no amplification or hybridization signal (18,22). One would then anticipate that either unfixed tissues or samples fixed in either a cross-linking fixative such as formalin, or a denaturing fixative such as ethanol or acetone could be used for RTin situ PCR. However, optimal RT in situ PCR , in my experience, requires formalin fixed material that has been adequately proteased (1,21). This is also true for in situ PCR for DNA targets (1,21). When peripheral blood leukocytes, each of which contain two copies of the bcl-2 gene, were fixed in ethanol or acetone, only 0 to 34% of the cells were found to be positive by in situ PCR using bcl-2 specific primers. Detection of the target in every cell was possible only if the cells were fixed in 10% buffered formalin and digested with protease prior to in situ PCR (1,20).
One may question why ethanol, acetone, or no fixation are suitable for PCR and in situ hybridization but not in situ PCR. Clearly, PCR and in situ PCR both involve the synthesis of DNA whereas in situ hybridization and in situ PCR both necessitate the complexing of a probe/primer with its target; in both instances similar sized molecules (Taq polymerase has equivalent dimensions with a 100 base pair labeled probe) must transverse a complex, 3-dimensional matrix of proteins and nucleic acids. The major difference between in situ PCR and the other two methodologies relates to the compartmentalization of the final product. With in situ hybridization, the probe-target complex is fixed to the nuclear matrix. The amplicon will readily diffuse during solution phase PCR to equivalent concentrations throughout the amplifying solution. in situ PCR is unique in that the target is fixed to the nucleus (or cytoplasm for RNA) but the amplicon can either remain at its site of origin, or diffuse throughout the entire cell or into the amplifying solution, depending on how the cells were fixed prior to in situ PCR.
It is a simple matter to do in situ PCR, save the amplifying solution, and determine under what conditions the amplicon remains restricted to the cell. We have performed such experiments (1,20). After in situ PCR, the only condition which permitted detection of the target in the cell and no detectable amplicon in the amplifying solution was fixation of the cells with a cross linker, such as 10% buffered formalin, followed by optimal protease digestion. Importantly, the amplicon was readily detected in the overlying solution after in situ PCR if the cells were fixed in acetone or ethanol (1,20).
It follows that the cross linking of proteins and nucleic acids by formalin fixation must be creating an "amplicon migration barrier". What is the nature of this barrier? Is it a physical barrier related to pore size, or is it biochemical in the sense of an ionic charge barricade? Obviously, at this stage all one can do is to speculate based on relatively little information.
If this putative barrier was based on the pore size of the cytoplasmic and nuclear membranes, then it should apply equally well for in situ hybridization and in situ PCR. There is no doubt that DNA sequences can be "stuck" inside a cell during in situ hybridization, perhaps by being trapped in these pores. This can be demonstrated by doing in situ hybridization with a 100mer probe using a tissue that does not have the corresponding target. If the probe concentration is too high or the post hybridization wash is not stringent enough, a nonspecific signal will be evident (Figure 1). The nonspecific signal, referred to as background, may be seen either in the nucleus or, more commonly, in the cytoplasm. Is background related to the size of the probe? In my experience, for probes that range from 20 to 300 base pairs, background is not related to the size of the probe (1). This would suggest that pore size may not be the most important variable for understanding this "amplicon migration barrier". It should be stressed that background is easily removed by performing a high stringency wash (Figure 1) which suggests that the movement of these DNA segments from 20 to 300 base pairs is not being restricted by pore size, but rather by biochemical forces such as hydrogen bonding and ionic charges. Such biochemical forces can be influenced during the post hybridization wash by conditions such as temperature, salt concentration, and formamide concentration (1,18). However, it has been well documented that probe size is an important variable for successful in situ hybridization. This was demonstrated in 1985 by Moench et al. using probes that varied from 70 to 780 base pairs in size (19).
Figure 1. Background versus signal with in situ PCR. In either in situ hybridization or in situ PCR, background will result if the labeled DNA sequence binds to cellular proteins or nucleic acids. Background often localizes to the cytoplasm of cells that are known to not contain the target. Background was evident in this HPV negative vulvar biopsy using HPV specific primers, if the post in situ PCR high stringency wash was omitted (A). This background was eliminated if a wash of 10 minutes at 60°C in 15 mM salt was done (B).
Another observation that suggests that pore size may be part, but not the complete explanation for the "amplicon migration barrier" with formalin fixed material is that, to obtain a signal with in situ hybridization, one does not need to protease digest samples fixed with one of the denaturing fixatives. However, especially after prolonged fixation, protease digestion will augment the signal for in situ hybridization if the sample has been fixed in formalin (1). Indeed, as evident in Figure 2, protease digestion destroys morphology in ethanol fixed tissue long before this would be evident if the same tissue was fixed in formalin. One may speculate that the cross linking fixatives creates pores or channels that could physically inhibit movement of the probe/primer to its target. Of course, there are other possible explanations for the need for a protease step for in situ hybridization with formalin fixed samples. The protein-DNA cross-links may need to be removed for the probe to access the target (1,18).
Figure 2. Formalin versus ethanol fixation with in situ hybridization. Formalin fixed tissue requires protease digestion for an optimal signal with in situ hybridization; in this instance, the repetitive alu probe was employed (A). If the same biopsy was fixed in ethanol, a signal is evident without protease digestion (B). If protease digestion is used, the signal was lost and the morphology destroyed (C).
Pore size, may, therefore, explain part of the putative "amplicon migration barrier" induced by cross linking fixatives with in situ PCR but probably is not the most important variable. It is possible that the biochemical correlates of a 3-dimensional matrix of rigid, cross linked proteins and nucleic acids may be a key factor (1). One can liken formalin fixation of a cell to creating a rigid, uniform scaffold where the positively and negatively charged side chains of amino acids would be spaced regularly and be poorly mobile. This can be contrasted with the "floating islands" of denatured proteins that are not as intimately associated with DNA and RNA in cells fixed in acetone or ethanol. In the latter situation, the charged side chains of the amino acids would not be rigidly and uniformly spaced nor in close, formal approximentation with the nucleic acids (Figure 3 and Figure 4). As evident from these figures, two conditions would be required for minimal migration of the amplicon: the presence of a fixed nuclear or cytoplasmic target and a network of positively charged amino acid side chains to inhibit amplicon migration (1,20). Acetone or ethanol fixed material would have a fixed target but, according to this model, would lack the latter feature. It is obvious that this is a simplistic model that needs further and rigorous testing.
Figure 3. Migration of the amplicon during in situ PCR: ethanol or acetone fixation. This figure attempts to explain the loss of the amplicon into the amplifying solution and decreased detection rate with in situ PCR if a denaturing fixative is used. It is speculated that ethanol or acetone fixatives remove the "ionic barrier" of positive charged protein side chains by denaturing the proteins.
Figure 4. Migration of the amplicon during in situ PCR: formalin fixation. This figure attempts to explain the apparent lack of migration of the amplicon into the amplifying solution and 100% detection rate with in situ PCR if formalin fixation is used. It is speculated that cross-links between proteins and nucleic acids create a regular arrangement of positively charged amino acid side chains that limit migration of the amplicon, assuming optimal protease digestion.
Whatever the correct model, it must be stressed that, under optimal formalin fixation and protease digestion, there is minimal migration of the amplicon from its site of origin. As seen in Figure 5, this can be seen as the sharp cytoplasmic signal of mRNA with RT in situ PCR versus the nuclear based signal in the positive control, in which DNA is being synthesized from a nuclear based genomic template. Consistent with a large body of data on RNA trafficking via the nuclear matrix, we have noted different nuclear pathways of premRNAs using RT in situ PCR (1,23). The sharp demarcation between the nuclear (for DNA) and cytoplasmic signal (for RNA) also illustrates the marked inhibition of migration of the amplicon under proper conditions of protease digestion and post PCR high stringency washes. This is further illustrated when doing in situ PCR for viruses that show marked cellular tropisms. For example, parvovirus infects only red blood cell precursors, which can be easily recognized on cytologic grounds. The signal after RT in situ PCR for parvoviral RNA in infected tissues is only evident in these cells under proper conditions, with no evident "back diffusion" to the neighboring white blood cell precursors (1, 15).
Figure 5. Correlation of mRNA expression and histologic features of tissues. The tissue is from a breast cancer. Note the intense nuclear signal in the different cell types (large arrow, carcinoma cells, small arrow, stromal cells) in the positive control where there is no DNase digestion (A). After DNase digestion, and RT in situ PCR for MAP kinase mRNA, only the cancer cells show a signal (B). Also note the cytoplasmic localization of the signal; the nuclei are negative. The signal was lost if hepatitis C primers were used in place of the MaP kinase primers (C).
Another interesting observation with regards the "amplicon migration barrier" is that, with intentional over digestion with protease, one can make a nuclear based signal move to the cytoplasm with in situ PCR in formalin fixed samples (Figure 6). Whether this reflects enlargement of cellular pores to sizes greater than critical migration thresholds or the loss of the protein-DNA tight cross linked network is unclear. The importance of the information available when interpreting in situ PCR with formalin fixed cells and protease digestion from the cellular distribution of the signal cannot be overstated. A cytoplasmic localization for a nuclear based target is due to over digestion by the protease. A nuclear based signal for RT in situ PCR and the test means under digestion in protease, as will be discussed in more detail. A cytoplasmic signal for RT in situ PCR and a nuclear signal for the positive control means adequate protease digestion and the target specific localization of the mRNA of interest.
Figure 6. Migration of the amplicon during PCR in situ hybridization. After optimal protease digestion (30 minutes), a strong nuclear signal is evident in SiHa cells, which contain 1 copy of HPV 16, using PCR in situ hybridization for HPV DNA (A). The signal migrates to the cytoplasm if the protease digestion time is increased to 60 minutes (B). The signal is lost, as is cell morphology, if the protease digestion time was increased to 90 minutes (not shown).
The amount of time the sample is exposed to protease digestion is arguably the most important variable in RT in situ PCR. The complex interconnecting protein-nucleic acid latticework that is created with formalin fixation may be, as just discussed, an essential element for successful in situ PCR, both in terms of detection of the amplicon and preventing its migration from inside the cell. However, the use of this fixative necessitates a protease digestion step. An advantage of the denaturing fixatives is that they do not require the digestion step. Perhaps due to a relatively high probability of migration of the amplicon out of the cell, at this stage in the development of the procedure these fixatives do not appear to allow for detection of the amplicon in all cells that contain the target (1,20,21).
The function of the protease in preparation for standard in situ hybridization and PCR in situ hybridization is to allow for entry of the probe/primer/Taq polymerase to the target sequence. Although this procedure also applies to RT in situ PCR, protease digestion has a further extraordinary function of rendering the entire genomic DNA of the cell non-amplifyable by exposing the DNA to subsequent digestion by DNase. Insufficient protease digestion may result in many persistent DNA-protein cross-links and the DNase may not be able to adequately degrade all the cellular DNA. The result can be the development of a non-specific DNA-repair based signal during the PCR step.
The three most commonly used proteases in diagnostic pathology are proteinase K, pepsin, and trypsin. These and other proteases will allow for successful in situ PCR. It is best to choose one of these proteases and use it exclusively in order to become familiar with its particular nuances. The actual procedure for preparing these protease solutions follows:
Both of these solutions can be frozen in 1 ml aliquots. When frozen, proteinase K solution will maintain activity for many months, or when stored at 4°C. The pepsin (or trypsin) should be used immediately or frozen and thawed within 1 week. When thawed, the pepsin solution should be stored on ice until ready to be used and then warmed to either room temperature or 37°C and used immediately. DEPC water, which is RNase free, is used for RNA work, although the protease would probably degrade any RNase present in the solution.
The most important point regarding protease digestion is the following:
Optimal protease digestion time typically is very different for RT in situ PCR as compared to PCR in situ hybridization for the same tissue.
Specifically, for PCR in situ hybridization, 30 minutes of digestion with pepsin solution is adequate for most formalin fixed cell and tissue preparations, regardless of length of fixation. However, for most tissues, especially those fixed for >8 hours, 30 minutes of pepsin digestion would be suboptimal in RT in situ PCR. This difference presumably is due to the fact that RT in situ PCR requires complete degradation of the genomic DNA template by DNase which, in turn demands extensive disruption of protein-DNA cross-links. Longer fixation times lead to more extensive protein-DNA cross-links and reflect the strong correlation between protease digestion and formalin fixation times for successful RT in situ PCR.
Optimal protease digestion for RT in situ PCR (Figure 5) is defined as that protease digestion time which yields the following using direct incorporation of the reporter nucleotide:
Sub optimal protease digestion for RT in situ PCR is defined as that protease digestion time which yields the following using direct incorporation of the reporter nucleotide:
The intense nuclear signal for the positive control (no DNase) with optimal protease digestion represents the following pathways:
A final and important point concerns the observation that sub optimal protease digestion allows for a signal with the negative control that is usually stronger than that for the corresponding positive control (1,20,21). A possible explanation for this is given in Figure 7. With sub optimal protease digestion, sufficient protein-DNA cross-links exist that interfere with complete DNase digestion of the genomic DNA. However, DNase can enlarge the putative gaps that are the foundation of the primer independent signal (1,20,21). The presence of a relatively strong nuclear based signal with the negative control and test (DNase and RT) in RT in situ PCR tells us to repeat the experiment with increased protease digestion time.
Figure 7. Hypothetical model for the enhancement of the signal in the negative control with inadequate protease digestion. It is postulated that DNase digestion after suboptimal protease digestion may enhance the signal by creating new gaps and/or increasing the size of the pre-existent single stranded gaps that may be repaired by Taq polymerase during PCR.
The definition of over-digestion with a protease is a test result exhibiting:
Proteinase K is much more likely than pepsin or trypsin to lead to over digestion of tissues. In RT in situ PCR in about 30% of tests, proteinase K (1 mg/ml) results in over-digestion of tissues compared with 5% using pepsin (Nuovo GJ, unpublished observations). For this reason, pepsin or trypsin is preferable.
In RT in situ PCR, protease can easily be inactivated by simply washing it off the glass slide. A one minute wash in DEPC water followed by a one minute wash in 100% ethanol will suffice. Do not heat inactivate the protease, as is done for solution phase PCR. Dry heat inactivation of the protease using the tissue on the glass slide will markedly diminish any hybridization signal, either with standard in situ hybridization or in situ PCR (1).
DNase digestion is best accomplished by overnight incubation at 37°C using the following recipe:
2 µl 10x PCR buffer (Perkin Elmer, Norwalk, CT)
2 µl of DNase (10U/µl, Boehringer Mannheim, Indianapolis, IN)
16 µl of DEPC water per tissue section.
Although adequate DNase digestion can be achieved after 8 hours, a longer digestion time, specifically overnight, is recommended (1).
Inadequate DNase digestion, as defined by the persistence of a signal with the negative control (DNase, no RT or RT with nonspecific primers) is most likely due to inadequate protease digestion. Increasing either the DNase digestion time and/or the concentration of the DNase usually will not rectify this problem (Nuovo GJ, unpublished observations). Under these conditions, increasing the protease digestion time will remove all nonspecific DNA synthesis with the negative control (1,20,21).
This section concerns an essential aspect of RT in situ PCR: direct incorporation of the reporter nucleotide. It should be stressed that direct incorporation of the tagged nucleotide can be nonspecific if mis-priming, primer oligomerization, or DNA repair are operative inside the cell. Optimal protease digestion and the subsequent DNase digestion will eliminate these nonspecific pathways. The result is that target specific direct incorporation of the reporter nucleotide into the PCR amplified cDNA can be achieved. However, direct incorporation of the reporter nucleotide is not possible in paraffin embedded tissues for DNA targets because DNA repair is invariably operative due to the heating of the tissue during tissue processing. In such cases, a formal hybridization step is needed after the PCR (1,20,21).
We have shown (Nuovo, GJ unpublished observations) that about 8,000 digoxigenin reporter nucleotides are needed inside a cell for a signal to be evident. Similarly, our calculations has suggested that, with the positive control of RT in situ PCR, over 100,000 reporter nucleotides may be incorporated into the nucleus, primarily via the DNA repair pathway. This correlates with the intense signal evident with the positive control (Figure 5).
We undertook a series of experiments to determine if a signal could be generated using labeled primers. We used HPV 16 primers that had one biotin or digoxigenin per 20mer. The samples were either SiHa cells (1 HPV 16 copy) or paraffin embedded cervical SILs that contained about 100 copies of HPV 16 per infected superficial cell. No signal was evident with the SiHa cells and direct incorporation of the labeled primer using hot start in situ PCR. A weak signal was evident in the cervical SILs that was stronger with the hot start maneuver. We obtained similar results with paraffin embedded placenta tissues and primers specific for the bcl-2 gene, present as two copies per cell. The signal was never as strong as with direct incorporation of the labeled nucleotide. Under these conditions, it was calculated that about 5,000 reporter nucleotides were present per nucleus (1).
To try to circumvent this problem of sensitivity with labeled primers, we obtained HPV 16 primers that contained THREE biotin moieties per primer; such primers are expensive to obtain and must be over 40 bp long. We re-did the experiments with the SiHa cells and the cervical SILs. A signal was seen in only about 10% of the SiHa cells and the signal with the cervical SILs was still not nearly as strong as with either standard in situ hybridization or PCR in situ hybridization. It is evident that, at this stage, labeled nucleotides are preferable to labeled primers for RT in situ PCR. Further, for DNA targets, PCR in situ hybridization is preferable to in situ PCR with labeled primers. However, it is important to remember that, by using frozen, fixed tissue and hot start, one CAN achieve intense target specific incorporation of the reporter nucleotide for DNA targets (1,20,21). This is of no help with paraffin embedded tissues where DNA repair would preclude using reporter nucleotides.