In situ hybridization (ISH) specifically detects mRNA targets and obtains temporal and spatial information of gene expression by means of hybridization with a RNA or, alternatively a DNA probe (Uehara et al., 1993; Nuovo, 2010). The lack of sensitivity to detect low expression of mRNA, the complexity and high expenses of probe design, and the need to set up protocols individually for each probe makes ISH a long, not easily reproducible, discouraging method. In contrast, in situ RT-PCR is performed directly on the tissue and comprises a combination of reverse transcription of mRNA into cDNA, polymerase chain reaction (PCR) of cDNA templates with specific primers and labeled dNTPs, and, finally, immunodetection of the PCR product. It has the added advantage of providing higher sensitivities than ISH, secondary to amplification. Both ISH and RT-PCR have applications to electron microscopy, but combination of mRNA detection with fine morphological analyses is a difficult task (Le Guellec and Frappart, 1993; Le Guellec, 1998; Morel et al., 1998; Cmarko et al., 2014). Pre-embedding methods achieve high signal-to-noise ratios but require longer proteolytic digestion times (tissue-damaging), and result in increased mispriming rates that are detected as an unspecific signal in the nucleus (Cmarko et al., 2014). Post-embedding detection is preferred and often performed on samples embedded in acrylic resins (Lowycril K4M, LR-White) that polymerize by UV light or low temperature, so nucleic acids are preserved. The main drawbacks of acrylic resins are the need for specific equipment and trained personal for the embedding process, with low image quality results and structurally poorly defined cellular structures. Acrylic resins also exhibit less stability under the electron beam compared to conventional epoxy resins (Le Guellec, 1998).

Here, we report an easy, reproducible, and low-damaging in situ RT-PCR immunogold staining protocol for ultrastructural mRNA expression studies that overcomes the lack of specificity and sensitivity of protein and mRNA detection, respectively. We used it to detect vimentin mRNA expression in ependymal cells of the third ventricle (3V) in the adult mouse brain but it is a suitable method for any mRNA and any brain region. Our procedure could be splitted into several stages (diagram in Figure 1): proteinase K digestion, mRNA reverse transcription into cDNA, mRNA-specific PCR with labeled nucleotides, immunogold labeling, and tissue embedding in araldite to, finally obtain ultrathin sections for observation under TEM.

Proteolytic digestion with proteinase K was a critical step because the intensity of the signal varied with concentration, duration of digestion, temperature and at the same time, it had to be carefully controlled to avoid disruption of cellular structures (Nuovo, 1996). After testing different concentrations and temperatures, it was found that 5 μg/ml of proteinase K for 10 min at room temperature preserved the morphological structures and provided a good sensitivity. Reverse transcription to obtain cDNA was performed using random primers, and guaranteed the preservation of the RNA in the form of cDNA. The presence of genomic DNA (gDNA) in the tissue required primer design from exon/exon junctions (junctional primers) to avoid unspecific amplification of gDNA. We chose pairs of primers with similar melting temperatures and a predicted amplicon length of at least 450 bp to minimize diffusion of products through the tissue (Cmarko et al., 2014). Using specific software we confirmed that primers did not form hairpins or loops at working temperatures and we checked that primers did not amplify any region of gDNA. To rule out any homology of the amplicon with other mouse genomic regions a blast search was performed. Gradient PCR in polypropylene tubes of striatal cDNA served us to check the optimal Tm for the tissue reaction that resulted, in our case (vimentin mRNA) to be 60°C.

Under the confocal microscope, biotin staining (representing vimentin mRNA expression, Figure 2A) co-localized specifically with vimentin protein in the wall of the 3V (Figures 2B,C). Biotin signal was restricted to the cell body while protein was in the cell body but also in the basal projections (white arrows in Figure 2B), due to differential subcellular distribution. Gold particles can be seen at electron microscopy because of their high density to the electron beam. We therefore performed gold labeling with mouse anti-biotin primary antibody and colloidal gold-secondary antibodies (Figures 2D,E, Sirerol-Piquer et al., 2012; Gil-Perotin et al., 2016). Images captured at TEM (Figures 2F–K) showed gold particles within the tanycyte implying vimentin-mRNA expression. There was a clear heterogeneity regarding vimentin expression in the 3V with regions of cells devoid of gold particles, and scattered cells with a strong homogeneous signal in the lateral wall (Figure 2F), but also in the floor of the 3V (Figure 2G). Label in positive cells consisted of homogenously distributed gold particles within the cytosol, and occasional clumps of particles over cytosolic dense bodies. Tissue was highly preserved, membranes were structurally well defined, and integrity of cell-to-cell junctions and subcellular structures allowed accurate ultrastructural description (Figures 2H,I). Arcuate neurons were not labeled, because these cells did not express vimentin (Figure 2J). Blood vessel membranes, in contact with tanycytes, had their contours specifically labeled (arrows in Figure 2K). Two negative controls were paramount to test the specificity of the method. We performed PCR without primers and complete immunolabeling (Figure 2D) and PCR with primers but without primary antibody. In both assays, labeling was not detected in the brain parenchyma, except for isolated, and scarce gold particles randomly distributed considered background. We did not observe unspecific signal in the cell nuclei.

Potential pitfalls of the technique and troubleshooting are included in Table 2.

To date, in situ RT-PCR is considered a complex, non-reproducible and aggressive method against cellular structures. In this work, we have adjusted the protocol to simplify its application, guaranteeing its reproducibility and preserving the ultrastructural integrity of the tissue. Hence, this optimized method combines the sensitivity of the PCR, the specificity of the in situ hybridization and the detail of electron microscopy.

In situ PCR quality is very dependent on tissue fixation. While ethanol- or acetone-fixed tissues are not suitable due to problems of DNA preservation and amplicon diffusion, formalin, and other aldehyde agents are recommended as fixative agents (Nuovo, 1996). As Nuovo explains, cross-linking agents constitute an “amplicon migration barrier,” a three-dimensional matrix of proteins and nucleic acids. The target template is fixed in the cytoplasm (in this case, RNA) with no detection of the amplicon in the aqueous amplification solution. Biotynylated nucleotides are shown to prevent amplicon diffusion because the amplicon increases its volume (Morel et al., 1998; Lassner et al., 2013). Amplicon length has also been related to diffusion and lengths between 150 and 500 bp are recommended (Morel et al., 1998; Bagasra and Harris, 2006). In our protocol we have optimized amplicon length to be around 450 bp.

We have acknowledged that permeabilization of the tissue with proteinase K is a key step in this protocol. The pore size of the amplicon migration barrier can be modified by protease digestion, so reagents are able to penetrating into the tissue, but at the same time a careful set up of exposure conditions is required to avoid deterioration of cell membranes. After testing distinct thickness of the sections, time and temperature of incubation, we determined the ideal conditions to guarantee a good ratio between intensity of label and tissue preservation to achieve ultrastructural quality standards. We did not use DNAse, and therefore avoided related tissue damage, because we used junctional primers. We extract from our results that section thickness is critical to resist enzymatic digestion, extreme temperatures during PCR, freezing with methylbutane, and finally, embedding into epoxy resin.

Embedding into resin is a necessary procedure in electron microscopy as it hardens the tissue before sectioning. We followed a conventional epoxy resin embedding protocol that polymerizes samples at a 65–70°C oven for 3 days. It is a relatively standard protocol. On the contrary, previous works have used acrylic resins that polymerize at cold temperatures or under ultraviolet radiation to preserve the integrity of nucleic acids, with the need of more complex embedding protocols (Le Guellec and Frappart, 1993). We demonstrate that the quality of the tissue is maintained using conventional protocols of inclusion in epoxy resin at high temperatures and that nucleic acids are preserved, allowing appropriate intensity of the label.

The main advantage of our method is the detection of specific target mRNA and, most importantly, the localization, identification and characterization of subcellular structures (junctions, organelles, cytoskeleton, synaptic structures) of the cell type expressing the mRNA. If we compare this method with others that are used to detect nucleic acids, such as in situ hybridization, the main advantages are the higher sensitivity, the simpler primer design and lower cost against the high cost and complexity of design, synthesis and/or commercially acquisition of RNA probes. In addition, in situ hybridization requires adjustment of stringency and hybridization conditions for each probe. All these reasons make in situ hybridization a complex technique, difficult to set up or reproduce. If hybridization precedes embedding, it could alter the ultrastructural morphology of the tissue and, if is carried out after embedding, the probe may not bind to nucleic acids that might have been altered. The proposed procedure is also useful if primary antibody of the protein of interest is not available (with fewer costs than generation of the antibody) or if we doubt the specificity of an antibody, since we can analyze the co-localization of mRNA and the protein on the same sample by double immunohistochemistry.

As disadvantages of this method are: (a) the need of a thermocycler adapted for slides to perform controlled PCR (although there are several models and brands available); (b) the necessity of certain expertise in conventional electron microscopy procedures. Nevertheless, for a laboratory that has basic knowledge in electron microscopy, it is a feasible method to carry out; (c) this method is not designed for quantification of mRNA, it is essentially qualitative. The amount of gold particles within the cell might correlate to mRNA expression but it will also depend on RT-PCR conditions or tissue preservation from RNAses; (d) our protocol was performed in a relatively abundant mRNA, and we have not determined the threshold of detection. This requires an individual testing for each target gene, each pair of primers, and distinct RT-PCR conditions, especially for low expressed genes.

Considering the advantages and disadvantages mentioned, the adjustments made attempt to simplify previous protocols and have increased its utility and quality of captured images making it a powerful, relatively easy and fast tool for correlative studies.

SG, LC, and BC wrote the manuscript and together with JG designed the method. LC and JC performed RT-PCR and immunodetection. MD and SG were responsible of ultrastructural analyses. JF contributed with technical assistance, and finally SG supervised the project.