The development and refinement of sensitive single-cell technology, such as in situ hybridization,4-6 has provided one way to overcome the dilution problem, and, perhaps more importantly, has created a window through which to view events of the virus life cycle in the context of the normal cell populations in latent and persistent infections. Not only is it possible to detect the presence of the infectious agent in very few cells within a background of normal cells and tissues, but it is also possible to identify the infected cell type and the nature of the host response to infection. This knowledge is essential for understanding the mechanism of pathogenesis, which will ultimately lead to the design of more effective treatments and preventive measures. However, in the case of true latency, where the cell contains only one or a few copies of transcriptionally silent viral genomes, or where transcriptional activity is minimal, the signal to noise ratio may be too low for confident assessment of infected cells even by in situ hybridization, and certainly not in the time frame essential to diagnostic methodology.
The recent development of the polymerase chain reaction (PCR)7,8 has allowed the amplification of viral sequences represented at extremely low frequencies within extracted nucleic acid pools (one copy of viral DNA per 70,000 cells)9 to levels detectable by gel electrophoresis and hybridization. Because this technology is very powerful in its ability to specifically amplify a minor population of sequences, it holds great promise for the future of viral diagnostics. By the same token, however, the technology is inherently plagued by potential problems of sample contamination, such as carry-over PCR product from previous analyses. In addition, while this technique increases the sensitivity of detection relative to other populational analysis tools, one loses both qualitative and quantitative information regarding the specific cell type infected as well as the spatial organization and numbers of those cells. The logical extension of cellular precision and target expansion involves the coupling of in situ technology with the PCR, which provides a new level of sensitivity essential to both diagnosis and the study of latent viral infections. We10-13 and others14-20 have developed conditions whereby nucleic acid sequences can now be amplified within the environment of the cell. Each cell, fixed either in suspension or on a solid support, and either as a single cell or in the context of surrounding tissue, functions individually as a reaction chamber for the PCR. With proper fixation and permeabilization conditions, the oligonucleotide primers and other reaction components are able to diffuse into the cells, and, upon thermal cycling, are able to amplify available specific target sequences. Product DNA is retained within the source cell and is readily detectable by standard in situ hybridization. Coupled with histological analysis and immunohistochemistry, it is now possible to confidently identify and characterize latent and persistent infections and to determine their pathogenetic context.
In this experimental system, we infected monolayers of fibroblasts, derived from explants of the choroid plexus of spring lambs, that support permissive replication of visna-maedi virus. The time course and extent of viral DNA and RNA synthesis in these sheep choroid plexus (SCP) cells is well documented,25 and as a consequence, the cells could be harvested and fixed at time points post-infection where we could control the approximate copy number of viral DNA and RNA per cell. For the majority of this work, the cells were fixed and collected one to three hours after infection when they contained 1-2 copies of viral DNA (the reverse transcription product of incoming viral genomes), or late in the life cycle when they contained several hundred copies of viral DNA (a consequence of superinfection). With these two extremes as boundaries of the range of detectability, we were able to optimize the conditions of amplification and compare the sensitivity of detection of the product to that achieved with conventional in situ hybridization for unamplified DNA. Because the quality of the PCR product may be greatly influenced by slight variations in reaction conditions, we worked to minimize the changes necessary for amplification of nucleic acids within fixed cells relative to amplification of purified nucleic acids. In order to take advantage of the precise thermal control and heat transfer of available thermal cyclers which accommodate microcentrifuge tubes, the technique was first developed10 using cells fixed in a suspension.
Using visna-infected SCP cells, we were able to establish conditions of fixation, pretreatment and amplification which allowed us to detect one or two copies of viral DNA per cell by in situ hybridization with a radiolabeled probe (Figure 1A).10 As evidence of specificity for the amplification and detection of visna DNA, we observed no signal over background on slides under the following conditions: 1) with uninfected cells amplified with visna-specific primers and hybridized with visna-specific probe; 2) with infected cells amplified with visna-specific primers and hybridized with a heterologous (HIV) probe (Figure 1B) or a visna probe derived from an unamplified region of the genome; 3) with infected cells amplified with heterologous primers and hybridized with visna-specific probe; 4) when an essential component (such as the polymerase) was omitted from the amplification reaction; or 5) when hybridization was carried out for unamplified DNA. The number of copies of amplified product DNA and, consequently, the estimated efficiency of amplification was determined by counting the silver grains over infected cells. The correlation between grain count and nucleic acid copy number was established by quantitative hybridization (solution or blot) for viral nucleic acids extracted from a parallel sample of infected cells.5,6,25 Statistical analysis of grain counts indicated that we were able to achieve approximately 12% efficiency in the amplification reaction or as much as a one hundred-fold amplification of the target sequence (a level more than sufficient for reliable detection by in situ hybridization). We attributed the observed variability in amplification efficiency from cell to cell within the population to asynchronous infection and reverse transcription processes, as well as to differences in the intracellular microenvironments of the DNA targets resulting from the fixation process which affects the accessibility of reaction components.
Figure 1. In situ amplification in single cells in solution: detection of viral DNA in visna-infected SCP cells. At twenty hours post-infection cells were trypsinized, fixed as a suspension in paraformaldehyde, washed with PBS and resuspended in PCR reaction mixture containing a multiple primer set (MPS) representing 1200 bases of the LTR and gag region of the visna genome. Following 50 amplification cycles, the cells were deposited onto denhardt-coated slides and subjected to in situ hybridization with radiolabeled visna-specific (A.) or heterologous (B.) probes.10 C. When amplification was carried out using a single primer pair designed to generate a 600 base pair product, hybridization with the visna-specific probe revealed a high extracellular background of silver grains as well as an accumulation of signal at the periphery of the cell. We interpret this artifact as a tendency of small amplimers to diffuse out of the cells under these conditions of amplification, since larger products generated by the MPS or single primer pairs spaced farther apart were more highly localized over the cell. 3 hour exposures. (A-C, x270)
Following the establishment of conditions to amplify and detect a single copy of viral DNA in cultured cells, we began the more important refinement of this method for the detection of viral DNA in individual cells in tissue sections.11 For these experiments, we exploited an in vivo model of viral infection of the lungs of sheep.26 The pulmonary system is a natural route of infection for visna-maedi where viral gene products accumulate over time to levels that elicit and sustain an inflammatory response. In the lung, the massive infiltration of inflammatory cells into the delicate interstitial spaces of the alveoli ultimately leads to tissue destruction, fibrosis and impaired blood-gas exchange. These events are manifest as shortness of breath, or dyspnea, which is reflected in the Icelandic term maedi. In the experimental model, in order to compress these pathological changes in time and space, we introduced a large viral inoculum (1 x 109 plaque-forming units of visna, produced in culture and concentrated by ultracentrifugation) into a defined segment of the lung of a sedated lamb via bronchoscopy. Progress of the infection was followed initially by in situ hybridization of a radiolabeled visna-specific probe to pulmonary alveolar macrophages (PAMs) recovered by bronchial lavage, since we knew from previous work26 that visna-maedi virus infects this cell type. Two weeks after infection, the animal was sacrificed and the lung tissue was removed, embedded in paraffin, sectioned, and prepared for the in situ procedures. Using this protocol, we found that, with the exception of the development of fibrosis, within this short period of time we could reproduce the type of histopathology in the lung that normally takes months to years to develop following natural infection (formation of lymphoid follicles, perivascular cuffing and interstitial inflammation; Figure 2A).
Figure 2. In situ amplification in tissue sections: detection of viral DNA in visna-infected sheep lung. A. Experimentally-infected tissues exhibited typical visna-induced histopathology characterized by the infiltration of PAMs and inflammatory cells into the alveoli (alv) and the accumulation of lymphocytes and monocytes around the blood vessels (bv) and bronchioles (br) with the formation of lymphoid follicles (LF).11 Eight-micron sections from formalin-fixed paraffin-embedded tissues were subjected to thermal cycling in the presence of a PCR mixture containing the visna MPS which targets the LTR and gag gene. Subsequent in situ hybridization with a radiolabeled visna-specific probe for amplified sequences revealed the presence of viral DNA in a majority of the bronchial epithelial cells (be) in regions of inflammation (B.) while epithelium from a region without inflammatory changes shows a background level of silver grains (C.)(B-C, 24 hr exposure). D. In situ hybridization for the presence of viral RNA demonstrates that an occasional cell (\xd8 )of the bronchial epithelium, in the region of inflammation, is transcriptionally active (2 day exposure). cart, cartilaginous ring; mc, macrophage. (A, x80; B-D, x240)
In order to amplify target sequences in tissue sections attached to a solid support, we had to modify the use of existing thermal cycling equipment to accommodate glass slides. After numerous experiments employing the heating block design found on many thermal cyclers, we focused on developing the use of a programmable circulating-air oven. While adaptations of the standard block-type thermal cycler have been used successfully,14-17 a major disadvantage is its lack of capacity to simultaneously process many slides.
We optimized pretreatments and reaction conditions and were able to amplify target viral DNA sequences to levels easily detectable by in situ hybridization in individual infected cells within a tissue section (Figure 2B).11 With these tissues, we achieved an approximate thirty-fold amplification which, while less efficient than amplification in cells in suspension, was well within levels required for confident detection of single copy genomes in infected cells. Surprisingly, we found in addition to PAMs, an expected host cell for visna-maedi virus, a new host cell in the bronchiolar epithelium. A majority of the bronchiolar epithelial cells in regions of heavy inflammation contained viral DNA. In situ hybridization for viral RNA in these same regions revealed transcripts not only in the PAMs, as again expected, but also in a very small percentage of the bronchiolar epithelial cells (Figure 2D), which corroborated the in situ PCR result demonstrating the presence of viral DNA in a cell type which had not previously been described as a host for visna. Consistent with the in vivo down-regulation of gene expression previously reported for visna,24,26,27 the number of copies of viral RNA in the transcriptionally active cells was ten- to one hundred-fold lower than that of permissively-infected cells in culture.25
Tissues were obtained from the experimentally-infected lamb approximately two weeks after infection. Following sacrifice and necropsy, the lungs were removed, en bloc, inflated with a solution of 10% buffered-formalin and fixed for a period of 72 hours. Tissue segments were dissected from the infected and uninfected lobes, dehydrated in 80% (v/v) ethanol for at least 24 hours, and, using standard techniques, were embedded in paraffin for purposes of thin sectioning, histopathological examination and in situ analyses. Eight-micron serial sections of the embedded tissues were attached to Denhardt-coated glass slides6 with a 3% (v/v) solution of Elmer's white glue in deionized distilled water, dried, and deparaffinized.
Cultured cells may be carried through the amplification process on a solid support in a manner similar to tissue sections, in which case the cells are fixed in formalin or paraformaldehyde, pelleted, embedded in paraffin and further treated as described for paraffin-embedded tissue.15,20 Alternatively, trypsinized and washed cells may be resuspended in PBS-CMF and deposited onto Denhardt-coated slides, either by cytocentrifugation or simply by spotting and drying a small volume of the cell suspension (typically, 30,000-50,000 cells in 10 µl). These cells are subsequently fixed by immersing the slides in fresh 4% (w/v) paraformaldehyde in PBS-CMF for twenty minutes at ambient temperature. They are then rinsed in PBS-CMF and dehydrated by passage through a series of graded alcohols (5 minutes in each of 70%, 80% and absolute ethanol).
PCR reaction solution (10 mM Tris-HCl, pH8.3, 50 mM KCl, 1.5 mM MgCl2, 0.01% (w/v) gelatin, 200 µM deoxynucleoside triphosphates, 1 µM oligonucleotide primers and 5% [v/v] Taq DNA polymerase [5 units/µl]) was pipetted onto the slides to fully hydrate the tissue sections (10-40 µl, depending on the dimensions of the section). The tissues were then covered with siliconized glass coverslips, excess solution was blotted to prevent the coverslips from sliding off of the tissue, and the slides were placed in heat-sealable plastic bags (101.6 x 152.4 x 0.1 mm), 2 slides per bag. Mineral oil (5-6 ml) was added to each bag, the air was removed and the bags were sealed. The oil both facilitates and stabilizes heat transfer to the slide as well as prevents the dehydration of the reaction mixture during thermal cycling. The bags with slides were positioned vertically in a polypropylene 80-pin rack which was placed inside of a thermal cycling circulating-air oven. The thermal sensor which is mounted on a glass slide, was similarly sealed in a plastic bag with oil and placed in the rack along with the samples. The slides were subjected to 25 cycles of amplification (94°C for 2 minutes, 42°C for 2 minutes and 72°C for 15 minutes). They were then removed from the bags, drained and rinsed in two to three changes of chloroform (5 minutes each) to remove residual oil film. After the chloroform had evaporated, the coverslips were lifted with forceps for the addition of fresh PCR reaction mixture and Taq DNA polymerase. Fresh coverslips were placed over the sections and the slides were once again sealed in oil-filled bags and placed in the oven for another set of 25 cycles. More recently, sequences in tissue sections have been successfully amplified with a single addition of Taq DNA polymerase and 30-50 cycles. A newer model of the thermal cycling oven (BioOven II; BioTherm Corporation, Fairfax, VA) is equipped with a sample rack which holds slides in a horizontal position. Sample preparation for this instrument simply requires a spot of fingernail enamel to anchor the coverslip to the slide over the tissue sample and a thin layer of mineral oil along the edge of the coverslip to prevent dehydration of the reaction mixture underneath. After enzymatic amplification was complete, the slides were removed from the bags, processed through chloroform as described above, the coverslips were removed under PBS-CMF. The slides were washed for 10 minutes in PBS-CMF and dehydrated in a series of graded alcohols.
Others have reported the successful use of non-isotopic probes (specifically, biotinylated14,16 or digoxygenin-labeled15,19,20) and the direct incorporation of digoxygenin dUTP15,19,20 or fluorescence-tagged synthetic oligonucleotides18 in the in situ amplification reaction.
The increased sensitivity and cellular specificity provided by in situ amplification will also facilitate the early detection of HIV infections in adults. Previous studies43,44 suggest that there may be a prolonged period of silent infection prior to seroconversion. The use of in situ PCR in prospective studies of seronegative individuals at risk of acquiring HIV-1 will assist not only in bringing early treatment to infected individuals, but also in providing information about the very early events of infection, which will support the process of defining the immunopathogenesis of infection and of developing early preventive and therapeutic interventions. Coupled with fluorescence-activated cell sorting (FACS), amplification of target sequences with fluorescent primers could very quickly and efficiently identify extremely infrequent infected cells within fluid samples such as blood or spinal fluid. This approach has already been successfully used for sorting and analyzing populations of cells in which specific viral nucleic acids45 or cellular messenger RNAs have been amplified in situ.18 Indeed, there is enormous potential in the application of in situ PCR for the efficient screening of body fluids or tissues for any infectious agent (viral, bacterial, fungal, etc.) for which a unique nucleic acid, DNA or RNA, would be present in cells.
Since the identification of HIV as the etiologic agent of AIDS, investigators have attempted to quantitate infection of the peripheral blood and lymphoid organs and to correlate viral burden with the clinical stages of disease. Early estimates of 10-4 to 10-5 infected peripheral blood mononuclear cells (PBMC) were produced by in situ hybridization for viral RNA.46 These values appeared to be too low to be responsible for the dramatic decline in the CD4+ population and ultimate production of immunodeficiency, and in retrospect, this technique alone was insufficiently sensitive to detect latently infected or severely down-regulated members of the population. An increase in sensitivity (at least ten-fold) and accuracy was achieved by using a variety of techniques from limiting dilution assays47 to PCR.48,49 With conventional PCR, however, all information regarding cell type, copy number per cell and state of latency or expression is lost. By using in situ PCR on peripheral blood samples, Bagasra, et al.16 has been able to obviate these problems and, in addition, has demonstrated an approximate ten-fold increase in sensitivity over the conventional techniques.
While there is a rapid and profound drop in the amount of virus present in PBLs and the plasma fraction of individuals shortly after infection, it is not clear whether the viral load present in the PBLs necessarily represents a comparable decrease in the virus present in lymphoid organs. Recent in situ hybridization studies have discovered high concentrations of HIV-1 RNA in extracellular association with follicular dendritic cells (FDCs) in the germinal centers of lymphoid tissues from adult patients at early stages of infection, and consequently it has been suggested that the lymphoid organs serve as major reservoirs for HIV-1.50,51 It has been unclear, however, whether or not these cells are actually infected with HIV. By using in situ PCR and a double-labeling technique52 which combines in situ hybridization and immunohistochemistry, we have been able to confirm the association of viral RNA with the FDCs in the germinal centers of lymph nodes, and to demonstrate a remarkable intracellular reservoir of HIV DNA in the macrophage and CD4+ lymphocyte population (20-30% of cells are positive for viral DNA) residing in and migrating through the lymph node follicles (Figure 4).13 As expected, only a fraction of these infected cells were transcriptionally active and contained viral RNA at levels ten- to one hundred-fold below those of permissively infected cells. This is consistent with information from in vivo studies demonstrating restricted gene expression in lentiviruses.23,24 By in situ PCR, we have also conclusively demonstrated that the FDCs do not contain viral DNA and are therefore not a significant infected cell type in the population. Taking into account electron microscopy data13 and studies on the tissue distribution of viral antigens, it appears likely that significant numbers of virions are present in immune complexes on the surface of the FDCs and it is this "pool" of virus which may ultimately function as an extracellular reservoir for the continued infection of CD4+ lymphocytes which are trafficked through the lymphoid organs.
Figure 4. In situ amplification and detection of HIV DNA in lymph nodes. Lymph node biopsies from HIV-infected individuals were fixed in formalin, embedded in paraffin, sectioned and analyzed by in situ PCR and in situ hybridization for the occurrence and distribution of HIV DNA- and RNA-containing cells.13 A multiple primer set (sequences derived from the HIV HXB2 complete genome) containing six primers (3 sense and 3 antisense) representing approximately 1551 nucleotides of the gag gene of HIV-1 was used in the amplification: 5'-GGAACCCACTGCTTAAGCCT-3' (bases 507-526), 5'-GCGTCAGTATTAAGCGGGGG-3' (bases 801-820), 5'-GTTTTCAGCATTATCAGATG-3' (bases 1301-1320), 5'-CATGGTGTTTAAATCTTGTG-3' (bases 1350-1331), 5'-CCCTGGCCTTAACCGAATTT-3' (bases 860-841), 5'-TTGGTGTCCTTCCTTTCCAC-3' (bases 2054-2035). A. A combination of a long radiographic exposure times (4-11 days) and low magnification emphasizes the presence of HIV DNA (black grains) predominantly in cells in the follicular mantle surrounding the germinal center (GC). B. In situ hybridization without amplification reveals HIV RNA primarily over the follicular dendritic cells (FDCs) of the germinal center. The diffuse distribution of silver grains overlying the processes of the FDCs is more clearly seen by the use of epipolarized illumination makes the silver grains appear white in panel B. (A, x110; B, x160)
We have been working with various tissue specimens from HIV-infected individuals to characterize virus-host interactions in terms of the latently-infected cell population, the cellular host range, and level of viral transcriptional activity. Prior to the lymph node studies described above, our initial experiments with in situ PCR on human samples were performed on tissue sections from a biopsy of an adenocarcinoma taken from an HIV-seropositive individual. In situ amplification revealed that a significant proportion (up to 34%) of the lymphocytes and mononuclear cells (identified by morphological criteria and by immunohistochemistry on adjacent sections with cell-specific antibodies) infiltrating the tumor were positive for viral DNA, and, with few exceptions, were transcriptionally silent.12 Since the CD4+ cells make up about one third of this inflammatory population, it is likely that nearly all of the CD4+ cells in this tumor are latently infected and, thus, represent a potential reservoir for the dissemination of virus over a prolonged period of time. Surprisingly, in situ hybridization for RNA (without amplification) showed that the tumor cells, which were epithelial in origin and which lacked detectable CD4 by immunocytochemistry, contained high levels of viral RNA and appeared to be replicating virus permissively while the infected lymphocyte population exhibited few cells which were transcriptionally active for HIV. These epithelial cells may illustrate the existence of an alternative cell surface receptor and therefore an expanded host range for HIV. Without the in situ PCR these examples of 1) infection of a different cell type and 2) two dramatically different virus life styles (permissive vs. down-regulated) within the same tissue environment, would have otherwise gone undetected.
Clinical events such as death and the development of opportunistic infections have been commonly used as endpoints in the early therapeutic trials of HIV-1 infection. The major limitation of this approach has been the length of time (measured in years) that is required for the studies to show clinical significance. Most recently, clinical trials are incorporating viral measurements, such as changes in viral load and development of resistance, to define study endpoints. By providing a more sensitive and direct estimate of the number of infected blood cells (viral burden) present in HIV-1 infected individuals, the use of in situ PCR would be extremely valuable in monitoring changes in viral load secondary to the use of antiretrovirals in monotherapy as well as in combination therapy. Currently, available techniques being evaluated in multi-center clinical trials include quantitative cultures and quantitative PCR of peripheral blood leukocytes and plasma fractions.
A number of recent studies have described the isolation of HIV-1 strains resistant to zidovudine (ZDV),53,54 didanosine (ddI),54 zalcitabine (ddC),55 as well as other non-nucleoside reverse transcriptase inhibitors. Genotypic characterization of these resistant isolates has shown the presence of specific mutations of the pol gene.53-55 These studies have been performed using PCR to amplify the gene region, followed by cloning and sequencing strategies, or secondary PCR reactions using primers specifically designed to detect the mutations of interest. A major limitation of these studies is that they do not provide a clear estimate of the proportion of mutant viruses present in a given individual. Because multiple quasispecies of HIV-1 coexist in the infected individual,56 the implications of finding a resistant genotype are unclear. With the design of appropriate primers and probes, in situ PCR could be applied to quantitate the infected cells containing viruses with specific mutations as well as to define, temporally and spatially, their emergence in the population. The more sensitive single-cell approach, as opposed to a clinical or cultural definition where the mutation has become the predominant variant with concomitant risks of passage of the resistant phenotype, will aide in the fine-tuning of combined therapies already in clinical trials57 and in the development of new treatment regimens involving alternation of therapeutic agents. Zidovudine-resistant variants are at a growth disadvantage compared to wild type, and appear as the primary variant in therapy only because of their ability to replicate under modified nucleoside pressure. A diagnostic tool such as in situ PCR would represent a significant advance for the understanding and management of infections caused by resistant HIV. This has not only therapeutic, but also epidemiologic implications, in that transmission of ZDV-resistant virus to adult hosts has been documented.58 If the emergence of such variants could be suppressed through early and accurate detection and substitution of therapies, one could possibly eliminate or at least control the pandemic consequences of drug resistant strains.
As in situ PCR combines the extreme sensitivity and specificity of nucleic acid amlification with the exquisite localization offered by in situ techniques to not only verify the presence of a viral agent, but also to provide information about the context in which it exists, it will be an invaluable tool for research as well as diagnostic purposes. Armed with the ability to reveal a single copy of viral nucleic acid in a single cell, and coupled with histology and immunohistochemistry for the detection of cellular antigens, one is well positioned to determine the cellular host range of the virus, the absolute numbers of cells that are infected, the percentage of infected cells that are transcriptionally active, and therefore the size of the latently infected reservoir. One can also begin to correlate viral burden with disease state and follow the effects of therapies at the cellular level. Relationships between these qualitative and quantitative values can lead to the generation and refinement of mathematical models of viral pathogenesis,11,56,59,60 the predictions of which must ultimately lead to a clearer understanding of virus-host interactions (including both primary and secondary effects), more accurate diagnosis of infection and staging of disease, and the development of novel and effective therapies and vaccines.
Finally, in addition to its use as a diagnostics and research tool for infections and diseases of known viral etiology, in situ PCR will prove extremely useful in the search for links between viruses and diseases of unknown origin (eg. multiple sclerosis).