Localization of GFP in Frozen Sections from Unfixed Mouse Tissues: Immobilization of a Highly Soluble Marker Protein by Formaldehyde Vapor

Abstract

Green fluorescent protein (GFP) and its variants, such as enhanced GFP (EGFP), have been introduced into mammalian cells by transgenes, e.g., to distinguish donor from host cells after transplantation. Free GFP is extremely soluble and leaks out from liquid-covered cryostat sections so that fixation of whole organs before sectioning has been mandatory. This precludes the analysis of serial sections with respect to fixation-sensitive enzyme activities and antigens. We describe here a vapor fixation for sections from unfixed cryostat blocks of tissue that allows unrestricted enzyme and immunohistochemistry on adjacent sections, as demonstrated for cross-striated muscle and other tissues from EGFP transgenic “green mice” and for a transplantation experiment.

In experimental embryology and transplantation research, it is important to distinguish cell and tissue contributions from different sources, e.g., the two partners used to produce embryo aggregation chimeras or donor from host in the case of cell or organ grafts. Stable genetic differences are preferable to vital dye or radioactive labeling because they are not diluted by cell division, and histological markers (like nuclear, cytoplasmic, or membrane differences) are more informative than biochemical markers (such as alloenzymes) that require homogenization of the tissue to be analyzed. In recent years, transgenic labels have been used extensively, in particular bacterial β-galactosidase localized in the cytoplasm (LacZ transgene) or in the nucleus (nLacZ) and green fluorescent protein (GFP) from the jellyfish Aequorea victoria and its artificial variants such as enhanced GFP (EGFP). GFP has the advantage of being a “vital dye” whose presence can be monitored in living organisms and whole unfixed organs, by appropriate UV illumination (Okabe et al. 1997). However, because of GFP's high solubility, conventional cryostat sections require the prior fixation of organs or whole embryos, and any variation of that method has been discouraged, e.g., Okabe: “don't even think about this” (http://133.1.15.131/tg/observation.cfm). We did think about an alternative. Based on our experience with the histochemistry of the highly soluble calcium-binding protein parvalbumin (Füchtbauer et al. 1991), we have succeeded in providing a more flexible method to monitor GFP-labeled tissues adjacent to unlabeled control tissues. Preliminary reports on experiments using this method have been published (Jockusch et al. 2000).

In this study we used “green mice” provided by Professor Masaru Okabe (Osaka University) through Professor Melitta Schachner (ZMNH Hamburg). These mice are transgenic for an enhanced green fluorescent protein (EGFP, hitherto designated “GFP”) construct in which the GFP cDNA is downstream to the chicken β-actin promoter. Non-GFP controls were NMRI mice (Harlan-Winkelmann; Borchen, Germany). Whole organs were either shock-frozen in liquefied propane (-190C) or fixed in an excess of 3.7% formaldehyde and calcium- and magnesium-free PBS (CMF-PBS) for 24 hr at 4C with gentle rocking (“bath fixation,” the method recommended by Okabe; http:// 133.1.15.131/tg/greenmouse.cfm), followed by 24 hr in CMF-PBS. Organs from GFP and non-GFP (“0”) mice were tied together with sewing thread before fixation to yield “artificial chimeric blocks.” In one case, a specimen from a transplantation experiment was used (Jockusch et al. 2000; the surgical procedures had been performed under anesthesia according to the German law for the protection of animals, with the permit of the local authorities).

Cryostat sections (8-μm thick) from unfixed frozen tissues were fixed for 20 min under a drop of 3.7% formaldehyde in CMF-PBS and then processed conventionally (“drop fixation”). For vapor fixation, cryostat sections on SuperFrostPlus adhesive microscope slides (Menzel-Gläser; Braunschweig, Germany) were dried for about 5 min and then exposed, in a tightly closed plastic dish, for 2–12 hr at −20C to the vapor of filter paper soaked with 37% formaldehyde (commercial stock solution; Sigma, F-1268, Dreieich, Germany). Sections were usually counterstained with Hoechst 33342 (1 μg/ml; Sigma, B 2261) and embedded in Elvanol (10 g Mowiol 4–88 in 40 ml CMF-PBS and 20 ml glycerol). Laminin and myosin heavy chain (MyHC) immunostaining was performed with a polyclonal anti-laminin (Sigma; L-9393) using an appropriate horseradish peroxidase-labeled second antibody which was developed with diaminobenzidine (DAB) and a monoclonal anti-MyHC MF20 (Bader et al. 1982) with a Cy3-labeled second antibody, respectively. Succinate dehydrogenase (SDH) histochemistry was performed according to Nachlas et al. (1957). For visualization of overall morphology, sections were stained with hematoxylin/eosin (HE). A Zeiss Axiophot epifluorescence microscope equipped with a digital Nikon Coolpix 990 camera was used for observation and documentation. Images were processed using Adobe Photoshop 5.0 (Adobe Systems; Tucson, AZ).

Figure 1

Different formaldehyde fixation and staining methods applied to serial sections from blocks containing both GFP-positive and -negative tissues. (A-E,I,J,M,N) Composite blocks with GFP-positive to the left and corresponding control tissue to the right mounted side by side. (A-E) Skeletal muscle. (F-H) C₂C₁₂ myogenic cells transplanted under the kidney capsule of GFP nude mouse. (I,J) Cardiac muscle (ventricular wall). (K,L) Stomach wall from GFP mouse only. (M,N) Pancreas GFP/control. (A,D) Bath fixation of tissues. (B,I) Drop fixation. (C,G,J,K,M) Vapor fixation of cryostat sections and (E,F,L,N) unfixed sections. (A-C,G,I-K,M) GFP fluorescence, in A'-C' combined with Hoechst fluorescence (blue) to show nuclei. (D,E) SDH (E, blue) and laminin staining (DAB; E, brownish); (F) MyHC immunochemistry (MF20, Cy3). (H) Overlay of F and G. (L,N) Hematoxylin/eosin. (D,E) Nomarski optics. Arrows in C,E, oxidative muscle fibers; in F,G, m, body wall muscle (GFP-positive), c, C₂C₁₂-derived myotubes (GFP-negative), some co-fused with host cells (GFP-positive); k, kidney (GFP-positive); in L, sm, smooth muscle; ge, gastric epithelium. In C the fluorescence signal was reduced to show the difference between fibers. In J it was intensified to exaggerate the effect of GFP overspill. Bar: A-E,I-N = 100 μm; F-H = 200 μm.

Methods of fixation of GFP-expressing tissues were compared with respect to the following parameters: overall morphology, retention of GFP in GFP-positive tissue regions, “overspill of label” from GFP-positive to neighboring unlabeled tissue, and the possibility of parallel analysis of the same tissue block. To this end, we have produced, from several organs that express high levels of GFP “chimeric blocks” that consisted of pairwise arrangements GFP-labeled and unlabeled tissues in close contact: skeletal muscle, cardiac muscle, stomach, and pancreas (Figure 1).

The bath fixation served as the positive control. Retention of GFP in all tissues was optimal and there was no overspill. Cryostat sections from prefixed skeletal muscle, however, were more difficult to handle than unfixed cryostat sections, and muscle fibers showed a tendency to shrink (Figure 1A vs 1C). The possibility to perform enzyme and immunohistochemistry on adjacent sections was drastically reduced because of the previous fixation, as exemplified by SDH and anti-laminin staining (Figure 1D vs 1E).

Embedding in paraffin greatly improves the quality of sections and yields high resolution (Walter et al. 2000), but the disadvantage of prefixation remains. Fixation of cryostat sections in a drop of liquid 3.7% formaldehyde was insufficient to retain the GFP in the large exposed cytoplasmic areas in skeletal muscle. Most of the GFP was lost before fixation became efficient, and the residual GFP was spilled over the section and not properly confined to the GFP-positive fibers (Figure 1B). We therefore adopted the method we developed earlier for parvalbumin (Füchtbauer et al. 1991) and exposed the sections to concentrated formaldehyde vapor at −20C. With vapor fixation we obtained excellent retention of GFP in positive muscle fibers, with almost no background in negative tissue except for low-level autofluorescence (Figure 1C). Owing to the brief thawing period during adhesion of the sections to the glass slide, there was a very narrow overspill zone of about 1/10 diameter of a muscle fiber into the neighboring GFP-negative muscle. The great advantage of this method is that any enzyme and immunocytochemical method can be applied to unfixed adjacent serial sections, as exemplified by combined SDH and laminin staining (Figure 1E), as well as general MyHC histochemistry (Figure 1F). Other important methods to characterize muscle fiber types would be myosin ATPase and isoform-specific MyHC immunohistochemistry (cf. Füchtbauer et al. 1991), both impossible in formaldehyde-prefixed tissue, which does, however, allow staining of β-galactosidase (cf. Eberhard et al. 2002). Any fixation can of course be combined with a Hoechst staining of nuclei (Figures 1A'-1C').

In tissues with smaller cells, in which the cytoplasmic area exposed to liquid fixative is not as large as in skeletal muscle fibers, such as smooth muscle of stomach and secretory cells in pancreas, liquid fixation of cryostat sections gave surprisingly poor results, whereas vapor fixation showed high retention, low background, and little if any marginal overspill (Figures 1K-1N).

As a possible application, the distinction of host from donor muscle cells after transplantation of the myogenic cell line C₂C₁₂ under the kidney capsule of a GFP nude mouse is shown (Figures 1F-1H). Because mature and regenerating host myofibers were found close to the transplantation site, their distinction from donor-derived myotubes was crucial. Whereas the former were unambiguously GFP-positive, the latter were GFP-negative.

The following observations have been made with regard to the differential expression of GFP in our stock of Okabe's green mice. All types of muscle- skeletal, cardiac, and smooth-showed very high levels of GFP expression. In mixed muscle, the oxidative small-diameter IIA fibers had somewhat higher GFP levels than large-diameter glycolytic IIB fibers (Figures 1C and 1E, arrows). Embryonic and neonatal skeletal muscle had lower GFP levels than adult muscle (Eberhard et al. 2002). There was background autofluorescence in cardiac and oxidative skeletal muscle cells but this was much lower than the GFP fluorescence, even in neonates (Eberhard et al. 2002). In general, these cases of autofluorescence did not interfere with the evaluation of GFP signals. Other tissues with high GFP fluorescence included the gastric epithelia in the stomach wall and the acini of the pancreas, whereas that in kidney was somewhat lower (Figures 1K-1N).

Skeletal muscle is an exceptionally favorable material to be studied in serial sections because it is easy to identify and follow individual fibers over several hundred micrometers. It is now possible to combine this traditional technique with the modern tool of GFP labeling. In contrast to whole-organ fixation, vapor fixation of single cryostat sections allows an unlimited application of standard immunocytochemical methods to adjacent sections in which cells of interest have been recognized by the GFP label. In comparison to bath fixation with subsequent paraffin embedding, the resolution at the subcellular level may be lower, but this is limited anyway by the high solubility of GFP in the cytoplasm. In the most important biomedical model organism, the mouse, our method can be applied to cell or organ transplantation (Jockusch et al. 2000), embryo chimeras (Eberhard et al. 2002), and the study of cell type-specific promoter activities in conjunction with GFP as a reporter of expression.